diff options
Diffstat (limited to 'xlators/nfs/server')
27 files changed, 22653 insertions, 0 deletions
diff --git a/xlators/nfs/server/Makefile.am b/xlators/nfs/server/Makefile.am new file mode 100644 index 000000000..a985f42a8 --- /dev/null +++ b/xlators/nfs/server/Makefile.am @@ -0,0 +1,3 @@ +SUBDIRS = src + +CLEANFILES = diff --git a/xlators/nfs/server/src/Makefile.am b/xlators/nfs/server/src/Makefile.am new file mode 100644 index 000000000..62fbf6535 --- /dev/null +++ b/xlators/nfs/server/src/Makefile.am @@ -0,0 +1,24 @@ +xlator_LTLIBRARIES = server.la +xlatordir = $(libdir)/glusterfs/$(PACKAGE_VERSION)/xlator/nfs +nfsrpclibdir = $(top_srcdir)/rpc/rpc-lib/src +server_la_LDFLAGS = -module -avoid-version +server_la_SOURCES = nfs.c nfs-common.c nfs-fops.c nfs-inodes.c \ + nfs-generics.c mount3.c nfs3-fh.c nfs3.c nfs3-helpers.c nlm4.c \ + nlmcbk_svc.c mount3udp_svc.c acl3.c +server_la_LIBADD = $(top_builddir)/libglusterfs/src/libglusterfs.la + +noinst_HEADERS = nfs.h nfs-common.h nfs-fops.h nfs-inodes.h nfs-generics.h \ + mount3.h nfs3-fh.h nfs3.h nfs3-helpers.h nfs-mem-types.h nlm4.h \ + acl3.h + +AM_CPPFLAGS = $(GF_CPPFLAGS) \ + -DLIBDIR=\"$(libdir)/glusterfs/$(PACKAGE_VERSION)/auth\" \ + -I$(top_srcdir)/libglusterfs/src \ + -I$(nfsrpclibdir) -I$(CONTRIBDIR)/rbtree \ + -I$(top_srcdir)/rpc/xdr/src/ + +AM_CFLAGS = -Wall $(GF_CFLAGS) + +AM_LDFLAGS = -L$(xlatordir) + +CLEANFILES = diff --git a/xlators/nfs/server/src/acl3.c b/xlators/nfs/server/src/acl3.c new file mode 100644 index 000000000..08b099b4e --- /dev/null +++ b/xlators/nfs/server/src/acl3.c @@ -0,0 +1,708 @@ +/* + * Copyright (c) 2012 Red Hat, Inc. <http://www.redhat.com> + * This file is part of GlusterFS. + * + * This file is licensed to you under your choice of the GNU Lesser + * General Public License, version 3 or any later version (LGPLv3 or + * later), or the GNU General Public License, version 2 (GPLv2), in all + * cases as published by the Free Software Foundation. + */ + +#ifndef _CONFIG_H +#define _CONFIG_H +#include "config.h" +#endif + +#include "defaults.h" +#include "rpcsvc.h" +#include "dict.h" +#include "xlator.h" +#include "nfs.h" +#include "mem-pool.h" +#include "logging.h" +#include "nfs-fops.h" +#include "inode.h" +#include "nfs3.h" +#include "nfs-mem-types.h" +#include "nfs3-helpers.h" +#include "nfs3-fh.h" +#include "nfs-generics.h" +#include "acl3.h" + + +typedef ssize_t (*acl3_serializer) (struct iovec outmsg, void *args); + +extern void nfs3_call_state_wipe (nfs3_call_state_t *cs); + +extern nfs3_call_state_t * +nfs3_call_state_init (struct nfs3_state *s, rpcsvc_request_t *req, xlator_t *v); + +extern int +nfs3_fh_validate (struct nfs3_fh *fh); + +extern fattr3 +nfs3_stat_to_fattr3 (struct iatt *buf); + +#define acl3_validate_nfs3_state(request, state, status, label, retval) \ + do { \ + state = rpcsvc_request_program_private (request); \ + if (!state) { \ + gf_log (GF_ACL, GF_LOG_ERROR, "NFSv3 state " \ + "missing from RPC request"); \ + rpcsvc_request_seterr (req, SYSTEM_ERR); \ + status = NFS3ERR_SERVERFAULT; \ + goto label; \ + } \ + } while (0); \ + +#define acl3_validate_gluster_fh(handle, status, errlabel) \ + do { \ + if (!nfs3_fh_validate (handle)) { \ + status = NFS3ERR_SERVERFAULT; \ + goto errlabel; \ + } \ + } while (0) \ + + +extern xlator_t * +nfs3_fh_to_xlator (struct nfs3_state *nfs3, struct nfs3_fh *fh); + +#define acl3_map_fh_to_volume(nfs3state, handle, req, volume, status, label) \ + do { \ + char exportid[256], gfid[256]; \ + rpc_transport_t *trans = NULL; \ + volume = nfs3_fh_to_xlator ((nfs3state), handle); \ + if (!volume) { \ + uuid_unparse (handle->exportid, exportid); \ + uuid_unparse (handle->gfid, gfid); \ + trans = rpcsvc_request_transport (req); \ + gf_log (GF_ACL, GF_LOG_ERROR, "Failed to map " \ + "FH to vol: client=%s, exportid=%s, gfid=%s",\ + trans->peerinfo.identifier, exportid, \ + gfid); \ + gf_log (GF_ACL, GF_LOG_ERROR, \ + "Stale nfs client %s must be trying to "\ + "connect to a deleted volume, please " \ + "unmount it.", trans->peerinfo.identifier);\ + status = NFS3ERR_STALE; \ + goto label; \ + } else { \ + gf_log (GF_ACL, GF_LOG_TRACE, "FH to Volume: %s"\ + ,volume->name); \ + rpcsvc_request_set_private (req, volume); \ + } \ + } while (0); \ + +#define acl3_volume_started_check(nfs3state, vlm, rtval, erlbl) \ + do { \ + if ((!nfs_subvolume_started (nfs_state (nfs3state->nfsx), vlm))){\ + gf_log (GF_ACL, GF_LOG_ERROR, "Volume is disabled: %s",\ + vlm->name); \ + rtval = RPCSVC_ACTOR_IGNORE; \ + goto erlbl; \ + } \ + } while (0) \ + +#define acl3_check_fh_resolve_status(cst, nfstat, erlabl) \ + do { \ + xlator_t *xlatorp = NULL; \ + char buf[256], gfid[256]; \ + rpc_transport_t *trans = NULL; \ + if ((cst)->resolve_ret < 0) { \ + trans = rpcsvc_request_transport (cst->req); \ + xlatorp = nfs3_fh_to_xlator (cst->nfs3state, \ + &cst->resolvefh); \ + uuid_unparse (cst->resolvefh.gfid, gfid); \ + snprintf (buf, sizeof (buf), "(%s) %s : %s", \ + trans->peerinfo.identifier, \ + xlatorp ? xlatorp->name : "ERR", \ + gfid); \ + gf_log (GF_ACL, GF_LOG_ERROR, "Unable to resolve FH"\ + ": %s", buf); \ + nfstat = nfs3_errno_to_nfsstat3 (cst->resolve_errno);\ + goto erlabl; \ + } \ + } while (0) \ + +#define acl3_handle_call_state_init(nfs3state, calls, rq, v, opstat, errlabel)\ + do { \ + calls = nfs3_call_state_init ((nfs3state), (rq), v); \ + if (!calls) { \ + gf_log (GF_ACL, GF_LOG_ERROR, "Failed to " \ + "init call state"); \ + opstat = NFS3ERR_SERVERFAULT; \ + rpcsvc_request_seterr (req, SYSTEM_ERR); \ + goto errlabel; \ + } \ + } while (0) \ + + +int +acl3svc_submit_reply (rpcsvc_request_t *req, void *arg, acl3_serializer sfunc) +{ + struct iovec outmsg = {0, }; + struct iobuf *iob = NULL; + struct nfs3_state *nfs3 = NULL; + int ret = -1; + ssize_t msglen = 0; + struct iobref *iobref = NULL; + + if (!req) + return -1; + + nfs3 = (struct nfs3_state *)rpcsvc_request_program_private (req); + if (!nfs3) { + gf_log (GF_ACL, GF_LOG_ERROR, "mount state not found"); + goto ret; + } + + /* First, get the io buffer into which the reply in arg will + * be serialized. + */ + iob = iobuf_get (nfs3->iobpool); + if (!iob) { + gf_log (GF_ACL, GF_LOG_ERROR, "Failed to get iobuf"); + goto ret; + } + + iobuf_to_iovec (iob, &outmsg); + /* Use the given serializer to translate the give C structure in arg + * to XDR format which will be written into the buffer in outmsg. + */ + msglen = sfunc (outmsg, arg); + if (msglen < 0) { + gf_log (GF_ACL, GF_LOG_ERROR, "Failed to encode message"); + goto ret; + } + outmsg.iov_len = msglen; + + iobref = iobref_new (); + if (iobref == NULL) { + gf_log (GF_ACL, GF_LOG_ERROR, "Failed to get iobref"); + goto ret; + } + + ret = iobref_add (iobref, iob); + if (ret) { + gf_log (GF_ACL, GF_LOG_ERROR, "Failed to add iob to iobref"); + goto ret; + } + + /* Then, submit the message for transmission. */ + ret = rpcsvc_submit_message (req, &outmsg, 1, NULL, 0, iobref); + if (ret == -1) { + gf_log (GF_ACL, GF_LOG_ERROR, "Reply submission failed"); + goto ret; + } + + ret = 0; +ret: + if (iob) + iobuf_unref (iob); + if (iobref) + iobref_unref (iobref); + + return ret; +} + + +int +acl3svc_null (rpcsvc_request_t *req) +{ + struct iovec dummyvec = {0, }; + + if (!req) { + gf_log (GF_ACL, GF_LOG_ERROR, "Got NULL request!"); + return 0; + } + rpcsvc_submit_generic (req, &dummyvec, 1, NULL, 0, NULL); + return 0; +} + +int +acl3_getacl_reply (nfs3_call_state_t *cs, getaclreply *reply) +{ + acl3svc_submit_reply (cs->req, (void *)reply, + (acl3_serializer)xdr_serialize_getaclreply); + return 0; +} + + +int +acl3_getacl_cbk (call_frame_t *frame, void *cookie, xlator_t *this, + int32_t op_ret, int32_t op_errno, dict_t *dict, + dict_t *xdata) +{ + nfsstat3 stat = NFS3ERR_SERVERFAULT; + nfs3_call_state_t *cs = NULL; + data_t *data = NULL; + int *p = NULL; + int i = 0; + getaclreply *getaclreply = NULL; + + if (!frame->local) { + gf_log (GF_ACL, GF_LOG_ERROR, "Invalid argument," + " frame->local NULL"); + return EINVAL; + } + cs = frame->local; + getaclreply = &cs->args.getaclreply; + if (op_ret == -1) { + stat = nfs3_cbk_errno_status (op_ret, op_errno); + goto err; + } + + getaclreply->aclentry.aclentry_val = cs->aclentry; + getaclreply->daclentry.daclentry_val = cs->daclentry; + + /* FIXME: use posix_acl_from_xattr() */ + data = dict_get (dict, POSIX_ACL_ACCESS_XATTR); + if (data && (p = data_to_bin (data))) { + /* POSIX_ACL_VERSION */ + p++; + while ((char *)p < (data->data + data->len)) { + getaclreply->aclentry.aclentry_val[i].type = *(*(short **)&p)++; + getaclreply->aclentry.aclentry_val[i].perm = *(*(short **)&p)++; + getaclreply->aclentry.aclentry_val[i].uid = *(*(int **)&p)++; + i++; + } + getaclreply->aclcount = getaclreply->aclentry.aclentry_len = i; + } + i = 0; + + data = dict_get (dict, POSIX_ACL_DEFAULT_XATTR); + if (data && (p = data_to_bin (data))) { + /* POSIX_ACL_VERSION */ + p++; + while ((char *)p < (data->data + data->len)) { + getaclreply->daclentry.daclentry_val[i].type = *(*(short **)&p)++; + getaclreply->daclentry.daclentry_val[i].perm = *(*(short **)&p)++; + getaclreply->daclentry.daclentry_val[i].uid = *(*(int **)&p)++; + i++; + } + getaclreply->daclcount = getaclreply->daclentry.daclentry_len = i; + } + + acl3_getacl_reply (cs, getaclreply); + nfs3_call_state_wipe (cs); + return 0; + +err: + if (getaclreply) + getaclreply->status = stat; + acl3_getacl_reply (cs, getaclreply); + nfs3_call_state_wipe (cs); + return 0; +} + +int +acl3_stat_cbk (call_frame_t *frame, void *cookie, xlator_t *this, + int32_t op_ret, int32_t op_errno, struct iatt *buf, + dict_t *xdata) +{ + nfsstat3 stat = NFS3ERR_SERVERFAULT; + nfs3_call_state_t *cs = NULL; + getaclreply *getaclreply = NULL; + int ret = -1; + nfs_user_t nfu = {0, }; + + if (!frame->local) { + gf_log (GF_ACL, GF_LOG_ERROR, "Invalid argument," + " frame->local NULL"); + return EINVAL; + } + + cs = frame->local; + getaclreply = &cs->args.getaclreply; + + if (op_ret == -1) { + stat = nfs3_cbk_errno_status (op_ret, op_errno); + goto err; + } + + getaclreply->attr_follows = 1; + getaclreply->attr = nfs3_stat_to_fattr3 (buf); + getaclreply->mask = 0xf; + nfs_request_user_init (&nfu, cs->req); + ret = nfs_getxattr (cs->nfsx, cs->vol, &nfu, &cs->resolvedloc, NULL, NULL, + acl3_getacl_cbk, cs); + if (ret == -1) { + stat = nfs3_cbk_errno_status (op_ret, op_errno); + goto err; + } + return 0; +err: + getaclreply->status = stat; + acl3_getacl_reply (cs, getaclreply); + nfs3_call_state_wipe (cs); + return 0; +} + + +int +acl3_getacl_resume (void *carg) +{ + int ret = -1; + nfs3_call_state_t *cs = NULL; + nfsstat3 stat = NFS3ERR_SERVERFAULT; + nfs_user_t nfu = {0, }; + + if (!carg) + return ret; + + cs = (nfs3_call_state_t *)carg; + acl3_check_fh_resolve_status (cs, stat, acl3err); + nfs_request_user_init (&nfu, cs->req); + + ret = nfs_stat (cs->nfsx, cs->vol, &nfu, &cs->resolvedloc, + acl3_stat_cbk, cs); + stat = -ret; +acl3err: + if (ret < 0) { + gf_log (GF_ACL, GF_LOG_ERROR, "unable to open_and_resume"); + cs->args.getaclreply.status = nfs3_errno_to_nfsstat3 (stat); + acl3_getacl_reply (cs, &cs->args.getaclreply); + nfs3_call_state_wipe (cs); + } + + return ret; +} + + +int +acl3svc_getacl (rpcsvc_request_t *req) +{ + xlator_t *vol = NULL; + struct nfs_state *nfs = NULL; + nfs3_state_t *nfs3 = NULL; + nfs3_call_state_t *cs = NULL; + int ret = RPCSVC_ACTOR_ERROR; + nfsstat3 stat = NFS3ERR_SERVERFAULT; + struct nfs3_fh fh, *fhp = NULL; + getaclargs getaclargs; + + if (!req) + return ret; + + acl3_validate_nfs3_state (req, nfs3, stat, rpcerr, ret); + nfs = nfs_state (nfs3->nfsx); + memset (&getaclargs, 0, sizeof (getaclargs)); + getaclargs.fh.n_bytes = (char *)&fh; + if (xdr_to_getaclargs(req->msg[0], &getaclargs) <= 0) { + gf_log (GF_ACL, GF_LOG_ERROR, "Error decoding args"); + rpcsvc_request_seterr (req, GARBAGE_ARGS); + goto rpcerr; + } + fhp = &fh; + acl3_validate_gluster_fh (&fh, stat, acl3err); + acl3_map_fh_to_volume (nfs->nfs3state, fhp, req, + vol, stat, acl3err); + acl3_handle_call_state_init (nfs->nfs3state, cs, req, + vol, stat, rpcerr); + + cs->vol = vol; + acl3_volume_started_check (nfs3, vol, ret, acl3err); + + ret = nfs3_fh_resolve_and_resume (cs, fhp, + NULL, acl3_getacl_resume); + +acl3err: + if (ret < 0) { + gf_log (GF_ACL, GF_LOG_ERROR, "unable to resolve and resume"); + if (cs) { + cs->args.getaclreply.status = stat; + acl3_getacl_reply (cs, &cs->args.getaclreply); + nfs3_call_state_wipe (cs); + } + return 0; + } + +rpcerr: + return ret; +} + +int +acl3_setacl_cbk (call_frame_t *frame, void *cookie, + xlator_t *this, int32_t op_ret, int32_t op_errno, + dict_t *xdata) +{ + nfs3_call_state_t *cs = NULL; + cs = frame->local; + if (op_ret < 0) { + nfsstat3 status = nfs3_cbk_errno_status (op_ret, op_errno); + cs->args.setaclreply.status = status; + } + + acl3svc_submit_reply (cs->req, (void *)&cs->args.setaclreply, + (acl3_serializer)xdr_serialize_setaclreply); + return 0; +} + +int +acl3_setacl_resume (void *carg) +{ + int ret = -1; + nfs3_call_state_t *cs = NULL; + nfsstat3 stat = NFS3ERR_SERVERFAULT; + nfs_user_t nfu = {0, }; + dict_t *xattr = NULL; + + if (!carg) + return ret; + cs = (nfs3_call_state_t *)carg; + acl3_check_fh_resolve_status (cs, stat, acl3err); + nfs_request_user_init (&nfu, cs->req); + xattr = dict_new(); + if (cs->aclcount) + ret = dict_set_static_bin (xattr, POSIX_ACL_ACCESS_XATTR, cs->aclxattr, + cs->aclcount * 8 + 4); + if (cs->daclcount) + ret = dict_set_static_bin (xattr, POSIX_ACL_DEFAULT_XATTR, + cs->daclxattr, cs->daclcount * 8 + 4); + + ret = nfs_setxattr (cs->nfsx, cs->vol, &nfu, &cs->resolvedloc, xattr, + 0, NULL, acl3_setacl_cbk, cs); + dict_unref (xattr); + +acl3err: + if (ret < 0) { + stat = -ret; + gf_log (GF_ACL, GF_LOG_ERROR, "unable to open_and_resume"); + cs->args.setaclreply.status = nfs3_errno_to_nfsstat3 (stat); + acl3svc_submit_reply (cs->req, (void *)&cs->args.setaclreply, + (acl3_serializer)xdr_serialize_setaclreply); + nfs3_call_state_wipe (cs); + } + + return ret; +} + + +int +acl3svc_setacl (rpcsvc_request_t *req) +{ + xlator_t *vol = NULL; + struct nfs_state *nfs = NULL; + nfs3_state_t *nfs3 = NULL; + nfs3_call_state_t *cs = NULL; + int ret = RPCSVC_ACTOR_ERROR; + nfsstat3 stat = NFS3ERR_SERVERFAULT; + struct nfs3_fh fh; + struct nfs3_fh *fhp = NULL; + setaclargs setaclargs; + aclentry *aclentry = NULL; + struct aclentry *daclentry = NULL; + int i = 0; + struct posix_acl_xattr_header *bufheader = NULL; + struct posix_acl_xattr_entry *bufentry = NULL; + + if (!req) + return ret; + aclentry = GF_CALLOC (NFS_ACL_MAX_ENTRIES, sizeof(*aclentry), + gf_nfs_mt_arr); + if (!aclentry) { + goto rpcerr; + } + daclentry = GF_CALLOC (NFS_ACL_MAX_ENTRIES, sizeof(*daclentry), + gf_nfs_mt_arr); + if (!daclentry) { + goto rpcerr; + } + + acl3_validate_nfs3_state (req, nfs3, stat, rpcerr, ret); + nfs = nfs_state (nfs3->nfsx); + memset (&setaclargs, 0, sizeof (setaclargs)); + memset (&fh, 0, sizeof (fh)); + setaclargs.fh.n_bytes = (char *)&fh; + setaclargs.aclentry.aclentry_val = aclentry; + setaclargs.daclentry.daclentry_val = daclentry; + if (xdr_to_setaclargs(req->msg[0], &setaclargs) <= 0) { + gf_log (GF_ACL, GF_LOG_ERROR, "Error decoding args"); + rpcsvc_request_seterr (req, GARBAGE_ARGS); + goto rpcerr; + } + fhp = &fh; + acl3_validate_gluster_fh (fhp, stat, acl3err); + acl3_map_fh_to_volume (nfs->nfs3state, fhp, req, + vol, stat, acl3err); + acl3_handle_call_state_init (nfs->nfs3state, cs, req, + vol, stat, rpcerr); + + cs->vol = vol; + acl3_volume_started_check (nfs3, vol, ret, rpcerr); + + cs->aclcount = setaclargs.aclcount; + cs->daclcount = setaclargs.daclcount; + + if ((cs->aclcount > NFS_ACL_MAX_ENTRIES) || + (cs->daclcount > NFS_ACL_MAX_ENTRIES)) + goto acl3err; + /* FIXME: use posix_acl_to_xattr() */ + /* Populate xattr buffer for user ACL */ + bufheader = (struct posix_acl_xattr_header *)(cs->aclxattr); + bufheader->version = htole32(POSIX_ACL_VERSION); + bufentry = bufheader->entries; + for (i = 0; i < cs->aclcount; i++) { + int uaceuid; + const struct aclentry *uace = &aclentry[i]; + switch (uace->type) { + case POSIX_ACL_USER: + case POSIX_ACL_GROUP: + uaceuid = uace->uid; + break; + default: + uaceuid = POSIX_ACL_UNDEFINED_ID; + break; + } + bufentry->tag = htole16(uace->type); + bufentry->perm = htole16(uace->perm); + bufentry->id = htole32(uaceuid); + + bufentry++; + } + + /* Populate xattr buffer for Default ACL */ + bufheader = (struct posix_acl_xattr_header *)(cs->aclxattr); + bufheader->version = htole32(POSIX_ACL_VERSION); + bufentry = bufheader->entries; + for (i = 0; i < cs->daclcount; i++) { + int daceuid; + int dacetype; + const struct aclentry *dace = &daclentry[i]; + /* + * For "default ACL", NFSv3 handles the 'type' differently + * i.e. by logical OR'ing 'type' with NFS_ACL_DEFAULT. + * Which the backend File system does not understand and + * that needs to be masked OFF. + */ + dacetype = (dace->type & ~(NFS_ACL_DEFAULT)); + switch (dacetype) { + case POSIX_ACL_USER: + case POSIX_ACL_GROUP: + daceuid = dace->uid; + break; + default: + daceuid = POSIX_ACL_UNDEFINED_ID; + break; + } + bufentry->tag = htole16(dacetype); + bufentry->perm = htole16(dace->perm); + bufentry->id = htole32(daceuid); + + bufentry++; + } + + + ret = nfs3_fh_resolve_and_resume (cs, fhp, + NULL, acl3_setacl_resume); + +acl3err: + if (ret < 0) { + gf_log (GF_ACL, GF_LOG_ERROR, "unable to resolve and resume"); + cs->args.setaclreply.status = stat; + acl3svc_submit_reply (cs->req, (void *)&cs->args.setaclreply, + (acl3_serializer)xdr_serialize_setaclreply); + nfs3_call_state_wipe (cs); + GF_FREE(aclentry); + GF_FREE(daclentry); + return 0; + } + +rpcerr: + if (ret < 0) + nfs3_call_state_wipe (cs); + if (aclentry) + GF_FREE (aclentry); + if (daclentry) + GF_FREE (daclentry); + return ret; +} + + + +rpcsvc_actor_t acl3svc_actors[ACL3_PROC_COUNT] = { + {"NULL", ACL3_NULL, acl3svc_null, NULL, 0}, + {"GETACL", ACL3_GETACL, acl3svc_getacl, NULL, 0}, + {"SETACL", ACL3_SETACL, acl3svc_setacl, NULL, 0}, +}; + +rpcsvc_program_t acl3prog = { + .progname = "ACL3", + .prognum = ACL_PROGRAM, + .progver = ACLV3_VERSION, + .progport = GF_NFS3_PORT, + .actors = acl3svc_actors, + .numactors = ACL3_PROC_COUNT, + .min_auth = AUTH_NULL, +}; + +rpcsvc_program_t * +acl3svc_init(xlator_t *nfsx) +{ + struct nfs3_state *ns = NULL; + struct nfs_state *nfs = NULL; + dict_t *options = NULL; + int ret = -1; + char *portstr = NULL; + static gf_boolean_t acl3_inited = _gf_false; + + /* Already inited */ + if (acl3_inited) + return &acl3prog; + + nfs = (struct nfs_state*)nfsx->private; + + ns = nfs->nfs3state; + if (!ns) { + gf_log (GF_ACL, GF_LOG_ERROR, "ACL3 init failed"); + goto err; + } + acl3prog.private = ns; + + options = dict_new (); + + ret = gf_asprintf (&portstr, "%d", GF_ACL3_PORT); + if (ret == -1) + goto err; + + ret = dict_set_dynstr (options, "transport.socket.listen-port", + portstr); + if (ret == -1) + goto err; + ret = dict_set_str (options, "transport-type", "socket"); + if (ret == -1) { + gf_log (GF_ACL, GF_LOG_ERROR, "dict_set_str error"); + goto err; + } + + if (nfs->allow_insecure) { + ret = dict_set_str (options, "rpc-auth-allow-insecure", "on"); + if (ret == -1) { + gf_log (GF_ACL, GF_LOG_ERROR, "dict_set_str error"); + goto err; + } + ret = dict_set_str (options, "rpc-auth.ports.insecure", "on"); + if (ret == -1) { + gf_log (GF_ACL, GF_LOG_ERROR, "dict_set_str error"); + goto err; + } + } + + ret = dict_set_str (options, "transport.address-family", "inet"); + if (ret == -1) { + gf_log (GF_ACL, GF_LOG_ERROR, "dict_set_str error"); + goto err; + } + + ret = rpcsvc_create_listeners (nfs->rpcsvc, options, "ACL"); + if (ret == -1) { + gf_log (GF_ACL, GF_LOG_ERROR, "Unable to create listeners"); + dict_unref (options); + goto err; + } + + acl3_inited = _gf_true; + return &acl3prog; +err: + return NULL; +} diff --git a/xlators/nfs/server/src/acl3.h b/xlators/nfs/server/src/acl3.h new file mode 100644 index 000000000..e0e61281a --- /dev/null +++ b/xlators/nfs/server/src/acl3.h @@ -0,0 +1,31 @@ +/* + * Copyright (c) 2012 Red Hat, Inc. <http://www.redhat.com> + * This file is part of GlusterFS. + * + * This file is licensed to you under your choice of the GNU Lesser + * General Public License, version 3 or any later version (LGPLv3 or + * later), or the GNU General Public License, version 2 (GPLv2), in all + * cases as published by the Free Software Foundation. + */ + +#ifndef _ACL3_H +#define _ACL3_H + +#include "glusterfs-acl.h" + +#define GF_ACL3_PORT 38469 +#define GF_ACL GF_NFS"-ACL" + +/* + * NFSv3, identifies the default ACL by NFS_ACL_DEFAULT. Gluster + * NFS needs to mask it OFF before sending it upto POSIX layer + * or File system layer. + */ +#define NFS_ACL_DEFAULT 0x1000 + +#define NFS_ACL_MAX_ENTRIES 1024 + +rpcsvc_program_t * +acl3svc_init(xlator_t *nfsx); + +#endif diff --git a/xlators/nfs/server/src/mount3.c b/xlators/nfs/server/src/mount3.c new file mode 100644 index 000000000..b0824bf10 --- /dev/null +++ b/xlators/nfs/server/src/mount3.c @@ -0,0 +1,2759 @@ +/* + Copyright (c) 2010-2011 Gluster, Inc. <http://www.gluster.com> + This file is part of GlusterFS. + + This file is licensed to you under your choice of the GNU Lesser + General Public License, version 3 or any later version (LGPLv3 or + later), or the GNU General Public License, version 2 (GPLv2), in all + cases as published by the Free Software Foundation. +*/ + +#ifndef _CONFIG_H +#define _CONFIG_H +#include "config.h" +#endif + +#include "rpcsvc.h" +#include "dict.h" +#include "xlator.h" +#include "mount3.h" +#include "xdr-nfs3.h" +#include "msg-nfs3.h" +#include "iobuf.h" +#include "nfs-common.h" +#include "nfs3-fh.h" +#include "nfs-fops.h" +#include "nfs-inodes.h" +#include "nfs-generics.h" +#include "locking.h" +#include "iatt.h" +#include "nfs-mem-types.h" +#include "nfs.h" +#include "common-utils.h" +#include "store.h" + +#include <errno.h> +#include <sys/socket.h> +#include <sys/uio.h> + + +#define IPv4_ADDR_SIZE 32 + +/* Macro to typecast the parameter to struct sockaddr_in + */ +#define SA(addr) ((struct sockaddr_in*)(addr)) + +/* Macro will mask the ip address with netmask. + */ +#define MASKED_IP(ipv4addr, netmask) \ + (ntohl(SA(ipv4addr)->sin_addr.s_addr) & (netmask)) + +/* Macro will compare two IP address after applying the mask + */ +#define COMPARE_IPv4_ADDRS(ip1, ip2, netmask) \ + ((MASKED_IP(ip1, netmask)) == (MASKED_IP(ip2, netmask))) + +/* This macro will assist in freeing up entire link list + * of host_auth_spec structure. + */ +#define FREE_HOSTSPEC(exp) do { \ + struct host_auth_spec *host= exp->hostspec; \ + while (NULL != host){ \ + struct host_auth_spec* temp = host; \ + host = host->next; \ + if (NULL != temp->host_addr) { \ + GF_FREE (temp->host_addr); \ + } \ + GF_FREE (temp); \ + } \ + exp->hostspec = NULL; \ + } while (0) + +typedef ssize_t (*mnt3_serializer) (struct iovec outmsg, void *args); + +extern void * +mount3udp_thread (void *argv); + +static inline void +mnt3_export_free (struct mnt3_export *exp) +{ + if (!exp) + return; + + if (exp->exptype == MNT3_EXPTYPE_DIR) + FREE_HOSTSPEC (exp); + GF_FREE (exp->expname); + GF_FREE (exp); +} + +/* Generic reply function for MOUNTv3 specific replies. */ +int +mnt3svc_submit_reply (rpcsvc_request_t *req, void *arg, mnt3_serializer sfunc) +{ + struct iovec outmsg = {0, }; + struct iobuf *iob = NULL; + struct mount3_state *ms = NULL; + int ret = -1; + ssize_t msglen = 0; + struct iobref *iobref = NULL; + + if (!req) + return -1; + + ms = (struct mount3_state *)rpcsvc_request_program_private (req); + if (!ms) { + gf_log (GF_MNT, GF_LOG_ERROR, "mount state not found"); + goto ret; + } + + /* First, get the io buffer into which the reply in arg will + * be serialized. + */ + /* TODO: use 'xdrproc_t' instead of 'sfunc' to get the xdr-size */ + iob = iobuf_get (ms->iobpool); + if (!iob) { + gf_log (GF_MNT, GF_LOG_ERROR, "Failed to get iobuf"); + goto ret; + } + + iobuf_to_iovec (iob, &outmsg); + /* Use the given serializer to translate the give C structure in arg + * to XDR format which will be written into the buffer in outmsg. + */ + msglen = sfunc (outmsg, arg); + if (msglen < 0) { + gf_log (GF_MNT, GF_LOG_ERROR, "Failed to encode message"); + goto ret; + } + outmsg.iov_len = msglen; + + iobref = iobref_new (); + if (iobref == NULL) { + gf_log (GF_MNT, GF_LOG_ERROR, "Failed to get iobref"); + goto ret; + } + + ret = iobref_add (iobref, iob); + if (ret) { + gf_log (GF_MNT, GF_LOG_ERROR, "Failed to add iob to iobref"); + goto ret; + } + + /* Then, submit the message for transmission. */ + ret = rpcsvc_submit_message (req, &outmsg, 1, NULL, 0, iobref); + if (ret == -1) { + gf_log (GF_MNT, GF_LOG_ERROR, "Reply submission failed"); + goto ret; + } + + ret = 0; +ret: + if (NULL != iob) + iobuf_unref (iob); + if (NULL != iobref) + iobref_unref (iobref); + + return ret; +} + + +/* Generic error reply function, just pass the err status + * and it will do the rest, including transmission. + */ +int +mnt3svc_mnt_error_reply (rpcsvc_request_t *req, int mntstat) +{ + mountres3 res; + + if (!req) + return -1; + + res.fhs_status = mntstat; + mnt3svc_submit_reply (req, (void *)&res, + (mnt3_serializer)xdr_serialize_mountres3); + + return 0; +} + + +mountstat3 +mnt3svc_errno_to_mnterr (int32_t errnum) +{ + mountstat3 stat; + + switch (errnum) { + + case 0: + stat = MNT3_OK; + break; + case ENOENT: + stat = MNT3ERR_NOENT; + break; + case EPERM: + stat = MNT3ERR_PERM; + break; + case EIO: + stat = MNT3ERR_IO; + break; + case EACCES: + stat = MNT3ERR_ACCES; + break; + case ENOTDIR: + stat = MNT3ERR_NOTDIR; + break; + case EINVAL: + stat = MNT3ERR_INVAL; + break; + case ENOSYS: + stat = MNT3ERR_NOTSUPP; + break; + case ENOMEM: + stat = MNT3ERR_SERVERFAULT; + break; + default: + stat = MNT3ERR_SERVERFAULT; + break; + } + + return stat; +} + + +mountres3 +mnt3svc_set_mountres3 (mountstat3 stat, struct nfs3_fh *fh, int *authflavor, + u_int aflen) +{ + mountres3 res = {0, }; + uint32_t fhlen = 0; + + res.fhs_status = stat; + fhlen = nfs3_fh_compute_size (fh); + res.mountres3_u.mountinfo.fhandle.fhandle3_len = fhlen; + res.mountres3_u.mountinfo.fhandle.fhandle3_val = (char *)fh; + res.mountres3_u.mountinfo.auth_flavors.auth_flavors_val = authflavor; + res.mountres3_u.mountinfo.auth_flavors.auth_flavors_len = aflen; + + return res; +} + +/* Read the rmtab from the store_handle and append (or not) the entries to the + * mountlist. + * + * Requires the store_handle to be locked. + */ +static int +__mount_read_rmtab (gf_store_handle_t *sh, struct list_head *mountlist, + gf_boolean_t append) +{ + int ret = 0; + unsigned int idx = 0; + struct mountentry *me = NULL, *tmp = NULL; + /* me->hostname is a char[MNTPATHLEN] */ + char key[MNTPATHLEN + 11]; + + GF_ASSERT (sh && mountlist); + + if (!gf_store_locked_local (sh)) { + gf_log (GF_MNT, GF_LOG_WARNING, "Not reading unlocked %s", + sh->path); + return -1; + } + + if (!append) { + list_for_each_entry_safe (me, tmp, mountlist, mlist) { + list_del (&me->mlist); + GF_FREE (me); + } + me = NULL; + } + + for (;;) { + char *value = NULL; + + if (me && append) { + /* do not add duplicates */ + list_for_each_entry (tmp, mountlist, mlist) { + if (!strcmp(tmp->hostname, me->hostname) && + !strcmp(tmp->exname, me->exname)) { + GF_FREE (me); + goto dont_add; + } + } + list_add_tail (&me->mlist, mountlist); + } else if (me) { + list_add_tail (&me->mlist, mountlist); + } + +dont_add: + me = GF_CALLOC (1, sizeof (*me), gf_nfs_mt_mountentry); + if (!me) { + gf_log (GF_MNT, GF_LOG_ERROR, "Out of memory"); + ret = -1; + goto out; + } + + INIT_LIST_HEAD (&me->mlist); + + snprintf (key, 9 + MNTPATHLEN, "hostname-%d", idx); + ret = gf_store_retrieve_value (sh, key, &value); + if (ret) + break; + strncpy (me->hostname, value, MNTPATHLEN); + GF_FREE (value); + + snprintf (key, 11 + MNTPATHLEN, "mountpoint-%d", idx); + ret = gf_store_retrieve_value (sh, key, &value); + if (ret) + break; + strncpy (me->exname, value, MNTPATHLEN); + GF_FREE (value); + + idx++; + gf_log (GF_MNT, GF_LOG_TRACE, "Read entries %s:%s", me->hostname, me->exname); + } + gf_log (GF_MNT, GF_LOG_DEBUG, "Read %d entries from '%s'", idx, sh->path); + GF_FREE (me); +out: + return ret; +} + +/* Overwrite the contents of the rwtab with te in-memory client list. + * Fail gracefully if the stora_handle is not locked. + */ +static void +__mount_rewrite_rmtab(struct mount3_state *ms, gf_store_handle_t *sh) +{ + struct mountentry *me = NULL; + char key[16]; + int fd, ret; + unsigned int idx = 0; + + if (!gf_store_locked_local (sh)) { + gf_log (GF_MNT, GF_LOG_WARNING, "Not modifying unlocked %s", + sh->path); + return; + } + + fd = gf_store_mkstemp (sh); + if (fd == -1) { + gf_log (GF_MNT, GF_LOG_ERROR, "Failed to open %s", sh->path); + return; + } + + list_for_each_entry (me, &ms->mountlist, mlist) { + snprintf (key, 16, "hostname-%d", idx); + ret = gf_store_save_value (fd, key, me->hostname); + if (ret) + goto fail; + + snprintf (key, 16, "mountpoint-%d", idx); + ret = gf_store_save_value (fd, key, me->exname); + if (ret) + goto fail; + + idx++; + } + + gf_log (GF_MNT, GF_LOG_DEBUG, "Updated rmtab with %d entries", idx); + + close (fd); + if (gf_store_rename_tmppath (sh)) + gf_log (GF_MNT, GF_LOG_ERROR, "Failed to overwrite rwtab %s", + sh->path); + + return; + +fail: + gf_log (GF_MNT, GF_LOG_ERROR, "Failed to update %s", sh->path); + close (fd); + gf_store_unlink_tmppath (sh); +} + +/* Read the rmtab into a clean ms->mountlist. + */ +static void +mount_read_rmtab (struct mount3_state *ms) +{ + gf_store_handle_t *sh = NULL; + struct nfs_state *nfs = NULL; + int ret; + + nfs = (struct nfs_state *)ms->nfsx->private; + + ret = gf_store_handle_new (nfs->rmtab, &sh); + if (ret) { + gf_log (GF_MNT, GF_LOG_WARNING, "Failed to open '%s'", + nfs->rmtab); + return; + } + + if (gf_store_lock (sh)) { + gf_log (GF_MNT, GF_LOG_WARNING, "Failed to lock '%s'", + nfs->rmtab); + goto out; + } + + __mount_read_rmtab (sh, &ms->mountlist, _gf_false); + gf_store_unlock (sh); + +out: + gf_store_handle_destroy (sh); +} + +/* Write the ms->mountlist to the rmtab. + * + * The rmtab could be empty, or it can exists and have been updated by a + * different storage server without our knowing. + * + * 1. takes the store_handle lock on the current rmtab + * - blocks if an other storage server rewrites the rmtab at the same time + * 2. [if new_rmtab] takes the store_handle lock on the new rmtab + * 3. reads/merges the entries from the current rmtab + * 4. [if new_rmtab] reads/merges the entries from the new rmtab + * 5. [if new_rmtab] writes the new rmtab + * 6. [if not new_rmtab] writes the current rmtab + * 7 [if new_rmtab] replaces nfs->rmtab to point to the new location + * 8. [if new_rmtab] releases the store_handle lock of the new rmtab + * 9. releases the store_handle lock of the old rmtab + */ +void +mount_rewrite_rmtab (struct mount3_state *ms, char *new_rmtab) +{ + gf_store_handle_t *sh = NULL, *nsh = NULL; + struct nfs_state *nfs = NULL; + int ret; + char *rmtab = NULL; + + nfs = (struct nfs_state *)ms->nfsx->private; + + ret = gf_store_handle_new (nfs->rmtab, &sh); + if (ret) { + gf_log (GF_MNT, GF_LOG_WARNING, "Failed to open '%s'", + nfs->rmtab); + return; + } + + if (gf_store_lock (sh)) { + gf_log (GF_MNT, GF_LOG_WARNING, "Not rewriting '%s'", + nfs->rmtab); + goto free_sh; + } + + if (new_rmtab) { + ret = gf_store_handle_new (new_rmtab, &nsh); + if (ret) { + gf_log (GF_MNT, GF_LOG_WARNING, "Failed to open '%s'", + new_rmtab); + goto unlock_sh; + } + + if (gf_store_lock (nsh)) { + gf_log (GF_MNT, GF_LOG_WARNING, "Not rewriting '%s'", + new_rmtab); + goto free_nsh; + } + } + + /* always read the currently used rmtab */ + __mount_read_rmtab (sh, &ms->mountlist, _gf_true); + + if (new_rmtab) { + /* read the new rmtab and write changes to the new location */ + __mount_read_rmtab (nsh, &ms->mountlist, _gf_true); + __mount_rewrite_rmtab (ms, nsh); + + /* replace the nfs->rmtab reference to the new rmtab */ + rmtab = gf_strdup(new_rmtab); + if (rmtab == NULL) { + gf_log (GF_MNT, GF_LOG_ERROR, "Out of memory, keeping " + "%s as rmtab", nfs->rmtab); + } else { + GF_FREE (nfs->rmtab); + nfs->rmtab = new_rmtab; + } + + gf_store_unlock (nsh); + } else { + /* rewrite the current (unchanged location) rmtab */ + __mount_rewrite_rmtab (ms, sh); + } + +free_nsh: + if (new_rmtab) + gf_store_handle_destroy (nsh); +unlock_sh: + gf_store_unlock (sh); +free_sh: + gf_store_handle_destroy (sh); +} + +/* Add a new NFS-client to the ms->mountlist and update the rmtab if we can. + * + * A NFS-client will only be removed from the ms->mountlist in case the + * NFS-client sends a unmount request. It is possible that a NFS-client + * crashed/rebooted had network loss or something else prevented the NFS-client + * to unmount cleanly. In this case, a duplicate entry would be added to the + * ms->mountlist, which is wrong and we should prevent. + * + * It is fully acceptible that the ms->mountlist is not 100% correct, this is a + * common issue for all(?) NFS-servers. + */ +int +mnt3svc_update_mountlist (struct mount3_state *ms, rpcsvc_request_t *req, + char *expname) +{ + struct mountentry *me = NULL; + struct mountentry *cur = NULL; + int ret = -1; + char *colon = NULL; + struct nfs_state *nfs = NULL; + gf_store_handle_t *sh = NULL; + + if ((!ms) || (!req) || (!expname)) + return -1; + + me = (struct mountentry *)GF_CALLOC (1, sizeof (*me), + gf_nfs_mt_mountentry); + if (!me) + return -1; + + nfs = (struct nfs_state *)ms->nfsx->private; + + ret = gf_store_handle_new (nfs->rmtab, &sh); + if (ret) { + gf_log (GF_MNT, GF_LOG_WARNING, "Failed to open '%s'", + nfs->rmtab); + goto free_err; + } + + strncpy (me->exname, expname, MNTPATHLEN); + + INIT_LIST_HEAD (&me->mlist); + /* Must get the IP or hostname of the client so we + * can map it into the mount entry. + */ + ret = rpcsvc_transport_peername (req->trans, me->hostname, MNTPATHLEN); + if (ret == -1) + goto free_err2; + + colon = strrchr (me->hostname, ':'); + if (colon) { + *colon = '\0'; + } + LOCK (&ms->mountlock); + { + /* in case locking fails, we just don't write the rmtab */ + if (gf_store_lock (sh)) { + gf_log (GF_MNT, GF_LOG_WARNING, "Failed to lock '%s'" + ", changes will not be written", nfs->rmtab); + } else { + __mount_read_rmtab (sh, &ms->mountlist, _gf_false); + } + + /* do not add duplicates */ + list_for_each_entry (cur, &ms->mountlist, mlist) { + if (!strcmp(cur->hostname, me->hostname) && + !strcmp(cur->exname, me->exname)) { + GF_FREE (me); + goto dont_add; + } + } + list_add_tail (&me->mlist, &ms->mountlist); + + /* only write the rmtab in case it was locked */ + if (gf_store_locked_local (sh)) + __mount_rewrite_rmtab (ms, sh); + } +dont_add: + if (gf_store_locked_local (sh)) + gf_store_unlock (sh); + + UNLOCK (&ms->mountlock); + +free_err2: + gf_store_handle_destroy (sh); + +free_err: + if (ret == -1) + GF_FREE (me); + + return ret; +} + + +int +__mnt3_get_volume_id (struct mount3_state *ms, xlator_t *mntxl, + uuid_t volumeid) +{ + int ret = -1; + struct mnt3_export *exp = NULL; + + if ((!ms) || (!mntxl)) + return ret; + + LOCK (&ms->mountlock); + list_for_each_entry (exp, &ms->exportlist, explist) { + if (exp->vol == mntxl) { + uuid_copy (volumeid, exp->volumeid); + ret = 0; + goto out; + } + } + +out: + UNLOCK (&ms->mountlock); + return ret; +} + + +int32_t +mnt3svc_lookup_mount_cbk (call_frame_t *frame, void *cookie, + xlator_t *this, int32_t op_ret, int32_t op_errno, + inode_t *inode, struct iatt *buf, dict_t *xattr, + struct iatt *postparent) +{ + mountres3 res = {0, }; + rpcsvc_request_t *req = NULL; + struct nfs3_fh fh = {{0}, }; + struct mount3_state *ms = NULL; + mountstat3 status = 0; + int autharr[10]; + int autharrlen = 0; + rpcsvc_t *svc = NULL; + xlator_t *mntxl = NULL; + uuid_t volumeid = {0, }; + char fhstr[1024], *path = NULL; + + req = (rpcsvc_request_t *)frame->local; + + if (!req) + return -1; + + mntxl = (xlator_t *)cookie; + ms = (struct mount3_state *)rpcsvc_request_program_private (req); + if (!ms) { + gf_log (GF_MNT, GF_LOG_ERROR, "mount state not found"); + op_ret = -1; + op_errno = EINVAL; + } + + if (op_ret == -1) { + gf_log (GF_NFS, GF_LOG_ERROR, "error=%s", strerror (op_errno)); + status = mnt3svc_errno_to_mnterr (op_errno); + } + if (status != MNT3_OK) + goto xmit_res; + + path = GF_CALLOC (PATH_MAX, sizeof (char), gf_nfs_mt_char); + if (!path) { + gf_log (GF_MNT, GF_LOG_ERROR, "Out of memory"); + goto xmit_res; + } + + snprintf (path, PATH_MAX, "/%s", mntxl->name); + mnt3svc_update_mountlist (ms, req, path); + GF_FREE (path); + if (gf_nfs_dvm_off (nfs_state (ms->nfsx))) { + fh = nfs3_fh_build_indexed_root_fh (ms->nfsx->children, mntxl); + goto xmit_res; + } + + __mnt3_get_volume_id (ms, mntxl, volumeid); + fh = nfs3_fh_build_uuid_root_fh (volumeid); + +xmit_res: + nfs3_fh_to_str (&fh, fhstr, sizeof (fhstr)); + gf_log (GF_MNT, GF_LOG_DEBUG, "MNT reply: fh %s, status: %d", fhstr, + status); + if (op_ret == 0) { + svc = rpcsvc_request_service (req); + autharrlen = rpcsvc_auth_array (svc, mntxl->name, autharr, + 10); + } + + res = mnt3svc_set_mountres3 (status, &fh, autharr, autharrlen); + mnt3svc_submit_reply (req, (void *)&res, + (mnt3_serializer)xdr_serialize_mountres3); + + return 0; +} + + +int +mnt3_match_dirpath_export (char *expname, char *dirpath) +{ + int ret = 0; + size_t dlen; + + if ((!expname) || (!dirpath)) + return 0; + + /* Some clients send a dirpath for mount that includes the slash at the + * end. String compare for searching the export will fail because our + * exports list does not include that slash. Remove the slash to + * compare. + */ + dlen = strlen (dirpath); + if (dlen && dirpath [dlen - 1] == '/') + dirpath [dlen - 1] = '\0'; + + if (dirpath[0] != '/') + expname++; + + if (strcmp (expname, dirpath) == 0) + ret = 1; + + return ret; +} + + +int +mnt3svc_mount_inode (rpcsvc_request_t *req, struct mount3_state *ms, + xlator_t * xl, inode_t *exportinode) +{ + int ret = -EFAULT; + nfs_user_t nfu = {0, }; + loc_t exportloc = {0, }; + + if ((!req) || (!xl) || (!ms) || (!exportinode)) + return ret; + + ret = nfs_inode_loc_fill (exportinode, &exportloc, NFS_RESOLVE_EXIST); + if (ret < 0) { + gf_log (GF_MNT, GF_LOG_ERROR, "Loc fill failed for export inode" + ": gfid %s, volume: %s", + uuid_utoa (exportinode->gfid), xl->name); + goto err; + } + + /* To service the mount request, all we need to do + * is to send a lookup fop that returns the stat + * for the root of the child volume. This is + * used to build the root fh sent to the client. + */ + nfs_request_user_init (&nfu, req); + ret = nfs_lookup (ms->nfsx, xl, &nfu, &exportloc, + mnt3svc_lookup_mount_cbk, (void *)req); + + nfs_loc_wipe (&exportloc); +err: + return ret; +} + + +/* For a volume mount request, we just have to create loc on the root inode, + * and send a lookup. In the lookup callback the mount reply is send along with + * the file handle. + */ +int +mnt3svc_volume_mount (rpcsvc_request_t *req, struct mount3_state *ms, + struct mnt3_export *exp) +{ + inode_t *exportinode = NULL; + int ret = -EFAULT; + uuid_t rootgfid = {0, }; + + if ((!req) || (!exp) || (!ms)) + return ret; + + rootgfid[15] = 1; + exportinode = inode_find (exp->vol->itable, rootgfid); + if (!exportinode) { + gf_log (GF_MNT, GF_LOG_ERROR, "Failed to get root inode"); + ret = -ENOENT; + goto err; + } + + ret = mnt3svc_mount_inode (req, ms, exp->vol, exportinode); + inode_unref (exportinode); + +err: + return ret; +} + + +/* The catch with directory exports is that the first component of the export + * name will be the name of the volume. + * Any lookup that needs to be performed to build the directory's file handle + * needs to start from the directory path from the root of the volume. For that + * we need to strip out the volume name first. + */ +char * +__volume_subdir (char *dirpath, char **volname) +{ + char *subdir = NULL; + int volname_len = 0; + + if (!dirpath) + return NULL; + + if (dirpath[0] == '/') + dirpath++; + + subdir = index (dirpath, (int)'/'); + if (!subdir) + goto out; + + if (!volname) + goto out; + + if (!*volname) + goto out; + + /* subdir points to the first / after the volume name while dirpath + * points to the first char of the volume name. + */ + volname_len = subdir - dirpath; + strncpy (*volname, dirpath, volname_len); + *(*volname + volname_len) = '\0'; +out: + return subdir; +} + + +void +mnt3_resolve_state_wipe (mnt3_resolve_t *mres) +{ + if (!mres) + return; + + nfs_loc_wipe (&mres->resolveloc); + GF_FREE (mres); + +} + + +/* Sets up the component argument to contain the next component in the path and + * sets up path as an absolute path starting from the next component. + */ +static char * +setup_next_component (char *path, size_t plen, char *component, size_t clen) +{ + char *comp = NULL; + char *nextcomp = NULL; + + if ((!path) || (!component)) + return NULL; + + strncpy (component, path, clen); + comp = index (component, (int)'/'); + if (!comp) + goto err; + + comp++; + nextcomp = index (comp, (int)'/'); + if (nextcomp) { + strncpy (path, nextcomp, plen); + *nextcomp = '\0'; + } else + path[0] = '\0'; + +err: + return comp; +} + +int32_t +mnt3_resolve_subdir_cbk (call_frame_t *frame, void *cookie, xlator_t *this, + int32_t op_ret, int32_t op_errno, inode_t *inode, + struct iatt *buf, dict_t *xattr, + struct iatt *postparent); + +/* There are multiple components in the directory export path and each one + * needs to be looked up one after the other. + */ +int +__mnt3_resolve_export_subdir_comp (mnt3_resolve_t *mres) +{ + char dupsubdir[MNTPATHLEN]; + char *nextcomp = NULL; + int ret = -EFAULT; + nfs_user_t nfu = {0, }; + uuid_t gfid = {0, }; + + if (!mres) + return ret; + + nextcomp = setup_next_component (mres->remainingdir, + sizeof (mres->remainingdir), + dupsubdir, sizeof (dupsubdir)); + if (!nextcomp) + goto err; + + /* Wipe the contents of the previous component */ + uuid_copy (gfid, mres->resolveloc.inode->gfid); + nfs_loc_wipe (&mres->resolveloc); + ret = nfs_entry_loc_fill (mres->exp->vol->itable, gfid, nextcomp, + &mres->resolveloc, NFS_RESOLVE_CREATE); + if ((ret < 0) && (ret != -2)) { + gf_log (GF_MNT, GF_LOG_ERROR, "Failed to resolve and create " + "inode: parent gfid %s, entry %s", + uuid_utoa (gfid), nextcomp); + ret = -EFAULT; + goto err; + } + + nfs_request_user_init (&nfu, mres->req); + ret = nfs_lookup (mres->mstate->nfsx, mres->exp->vol, &nfu, + &mres->resolveloc, mnt3_resolve_subdir_cbk, mres); + +err: + return ret; +} + + +int32_t +mnt3_resolve_subdir_cbk (call_frame_t *frame, void *cookie, xlator_t *this, + int32_t op_ret, int32_t op_errno, inode_t *inode, + struct iatt *buf, dict_t *xattr, + struct iatt *postparent) +{ + mnt3_resolve_t *mres = NULL; + mountstat3 mntstat = MNT3ERR_SERVERFAULT; + struct nfs3_fh fh = {{0}, }; + int autharr[10]; + int autharrlen = 0; + rpcsvc_t *svc = NULL; + mountres3 res = {0, }; + xlator_t *mntxl = NULL; + char *path = NULL; + + mres = frame->local; + mntxl = (xlator_t *)cookie; + if (op_ret == -1) { + gf_log (GF_NFS, GF_LOG_ERROR, "path=%s (%s)", + mres->resolveloc.path, strerror (op_errno)); + mntstat = mnt3svc_errno_to_mnterr (op_errno); + goto err; + } + + inode_link (mres->resolveloc.inode, mres->resolveloc.parent, + mres->resolveloc.name, buf); + + nfs3_fh_build_child_fh (&mres->parentfh, buf, &fh); + if (strlen (mres->remainingdir) <= 0) { + op_ret = -1; + mntstat = MNT3_OK; + path = GF_CALLOC (PATH_MAX, sizeof (char), gf_nfs_mt_char); + if (!path) { + gf_log (GF_MNT, GF_LOG_ERROR, "Memory allocation " + "failed"); + goto err; + } + snprintf (path, PATH_MAX, "/%s%s", mres->exp->vol->name, + mres->resolveloc.path); + mnt3svc_update_mountlist (mres->mstate, mres->req, + path); + GF_FREE (path); + } else { + mres->parentfh = fh; + op_ret = __mnt3_resolve_export_subdir_comp (mres); + if (op_ret < 0) + mntstat = mnt3svc_errno_to_mnterr (-op_ret); + } +err: + if (op_ret == -1) { + gf_log (GF_MNT, GF_LOG_DEBUG, "Mount reply status: %d", + mntstat); + svc = rpcsvc_request_service (mres->req); + autharrlen = rpcsvc_auth_array (svc, mntxl->name, autharr, + 10); + + res = mnt3svc_set_mountres3 (mntstat, &fh, autharr, autharrlen); + mnt3svc_submit_reply (mres->req, (void *)&res, + (mnt3_serializer)xdr_serialize_mountres3); + mnt3_resolve_state_wipe (mres); + } + + return 0; +} + + + +/* We will always have to perform a hard lookup on all the components of a + * directory export for a mount request because in the mount reply we need the + * file handle of the directory. Our file handle creation code is designed with + * the assumption that to build a child file/dir fh, we'll always have the + * parent dir's fh available so that we may copy the hash array of the previous + * dir levels. + * + * Since we do not store the file handles anywhere, for every mount request we + * must resolve the file handles of every component so that the parent dir file + * of the exported directory can be built. + */ +int +__mnt3_resolve_subdir (mnt3_resolve_t *mres) +{ + char dupsubdir[MNTPATHLEN]; + char *firstcomp = NULL; + int ret = -EFAULT; + nfs_user_t nfu = {0, }; + uuid_t rootgfid = {0, }; + + if (!mres) + return ret; + + firstcomp = setup_next_component (mres->remainingdir, + sizeof (mres->remainingdir), + dupsubdir, sizeof (dupsubdir)); + if (!firstcomp) + goto err; + + rootgfid[15] = 1; + ret = nfs_entry_loc_fill (mres->exp->vol->itable, rootgfid, firstcomp, + &mres->resolveloc, NFS_RESOLVE_CREATE); + if ((ret < 0) && (ret != -2)) { + gf_log (GF_MNT, GF_LOG_ERROR, "Failed to resolve and create " + "inode for volume root: %s", mres->exp->vol->name); + ret = -EFAULT; + goto err; + } + + nfs_request_user_init (&nfu, mres->req); + ret = nfs_lookup (mres->mstate->nfsx, mres->exp->vol, &nfu, + &mres->resolveloc, mnt3_resolve_subdir_cbk, mres); + +err: + return ret; +} + + +/** + * This function will verify if the client is allowed to mount + * the directory or not. Client's IP address will be compared with + * allowed IP list or range present in mnt3_export structure. + * + * @param req - RPC request. This structure contains client's IP address. + * @param export - mnt3_export structure. Contains allowed IP list/range. + * + * @return 0 - on Success and -EACCES on failure. + */ +int +mnt3_verify_auth (rpcsvc_request_t *req, struct mnt3_export *export) +{ + int retvalue = -EACCES; + int ret = 0; + int shiftbits = 0; + uint32_t ipv4netmask = 0; + uint32_t routingprefix = 0; + struct host_auth_spec *host = NULL; + struct sockaddr_in *client_addr = NULL; + struct sockaddr_in *allowed_addr = NULL; + struct addrinfo *allowed_addrinfo = NULL; + + /* Sanity check */ + if ((NULL == req) || + (NULL == req->trans) || + (NULL == export) || + (NULL == export->hostspec)) { + gf_log (GF_MNT, GF_LOG_ERROR, "Invalid argument"); + return retvalue; + } + + host = export->hostspec; + + + /* Client's IP address. */ + client_addr = (struct sockaddr_in *)(&(req->trans->peerinfo.sockaddr)); + + /* Try to see if the client IP matches the allowed IP list.*/ + while (NULL != host){ + GF_ASSERT (host->host_addr); + + if (NULL != allowed_addrinfo) { + freeaddrinfo (allowed_addrinfo); + allowed_addrinfo = NULL; + } + + /* Get the addrinfo for the allowed host (host_addr). */ + ret = getaddrinfo (host->host_addr, + NULL, + NULL, + &allowed_addrinfo); + if (0 != ret){ + gf_log (GF_MNT, GF_LOG_ERROR, "getaddrinfo: %s\n", + gai_strerror (ret)); + host = host->next; + + /* Failed to get IP addrinfo. Continue to check other + * allowed IPs in the list. + */ + continue; + } + + allowed_addr = (struct sockaddr_in *)(allowed_addrinfo->ai_addr); + + if (NULL == allowed_addr) { + gf_log (GF_MNT, GF_LOG_ERROR, "Invalid structure"); + break; + } + + if (AF_INET == allowed_addr->sin_family){ + if (IPv4_ADDR_SIZE < host->routeprefix) { + gf_log (GF_MNT, GF_LOG_ERROR, "invalid IP " + "configured for export-dir AUTH"); + host = host->next; + continue; + } + + /* -1 means no route prefix is provided. In this case + * the IP should be an exact match. Which is same as + * providing a route prefix of IPv4_ADDR_SIZE. + */ + if (-1 == host->routeprefix) { + routingprefix = IPv4_ADDR_SIZE; + } else { + routingprefix = host->routeprefix; + } + + /* Create a mask from the routing prefix. User provided + * CIDR address is split into IP address (host_addr) and + * routing prefix (routeprefix). This CIDR address may + * denote a single, distinct interface address or the + * beginning address of an entire network. + * + * e.g. the IPv4 block 192.168.100.0/24 represents the + * 256 IPv4 addresses from 192.168.100.0 to + * 192.168.100.255. + * Therefore to check if an IP matches 192.168.100.0/24 + * we should mask the IP with FFFFFF00 and compare it + * with host address part of CIDR. + */ + shiftbits = IPv4_ADDR_SIZE - routingprefix; + ipv4netmask = 0xFFFFFFFFUL << shiftbits; + + /* Mask both the IPs and then check if they match + * or not. */ + if (COMPARE_IPv4_ADDRS (allowed_addr, + client_addr, + ipv4netmask)){ + retvalue = 0; + break; + } + } + + /* Client IP didn't match the allowed IP. + * Check with the next allowed IP.*/ + host = host->next; + } + + if (NULL != allowed_addrinfo) { + freeaddrinfo (allowed_addrinfo); + } + + return retvalue; +} + +int +mnt3_resolve_subdir (rpcsvc_request_t *req, struct mount3_state *ms, + struct mnt3_export *exp, char *subdir) +{ + mnt3_resolve_t *mres = NULL; + int ret = -EFAULT; + struct nfs3_fh pfh = GF_NFS3FH_STATIC_INITIALIZER; + + if ((!req) || (!ms) || (!exp) || (!subdir)) + return ret; + + /* Need to check AUTH */ + if (NULL != exp->hostspec) { + ret = mnt3_verify_auth (req, exp); + if (0 != ret) { + gf_log (GF_MNT,GF_LOG_ERROR, + "AUTH verification failed"); + return ret; + } + } + + mres = GF_CALLOC (1, sizeof (mnt3_resolve_t), gf_nfs_mt_mnt3_resolve); + if (!mres) { + gf_log (GF_MNT, GF_LOG_ERROR, "Memory allocation failed"); + goto err; + } + + mres->exp = exp; + mres->mstate = ms; + mres->req = req; + strncpy (mres->remainingdir, subdir, MNTPATHLEN); + if (gf_nfs_dvm_off (nfs_state (ms->nfsx))) + pfh = nfs3_fh_build_indexed_root_fh (mres->mstate->nfsx->children, mres->exp->vol); + else + pfh = nfs3_fh_build_uuid_root_fh (exp->volumeid); + + mres->parentfh = pfh; + ret = __mnt3_resolve_subdir (mres); + if (ret < 0) { + gf_log (GF_MNT, GF_LOG_ERROR, "Failed to resolve export dir: %s" + , mres->exp->expname); + GF_FREE (mres); + } + +err: + return ret; +} + + +int +mnt3_resolve_export_subdir (rpcsvc_request_t *req, struct mount3_state *ms, + struct mnt3_export *exp) +{ + char *volume_subdir = NULL; + int ret = -EFAULT; + + if ((!req) || (!ms) || (!exp)) + return ret; + + volume_subdir = __volume_subdir (exp->expname, NULL); + if (!volume_subdir) + goto err; + + ret = mnt3_resolve_subdir (req, ms, exp, volume_subdir); + if (ret < 0) { + gf_log (GF_MNT, GF_LOG_ERROR, "Failed to resolve export dir: %s" + , exp->expname); + goto err; + } + +err: + return ret; +} + + +int +mnt3svc_mount (rpcsvc_request_t *req, struct mount3_state *ms, + struct mnt3_export *exp) +{ + int ret = -EFAULT; + + if ((!req) || (!ms) || (!exp)) + return ret; + + if (exp->exptype == MNT3_EXPTYPE_VOLUME) + ret = mnt3svc_volume_mount (req, ms, exp); + else if (exp->exptype == MNT3_EXPTYPE_DIR) + ret = mnt3_resolve_export_subdir (req, ms, exp); + + return ret; +} + + +/* mnt3_mntpath_to_xlator sets this to 1 if the mount is for a full +* volume or 2 for a subdir in the volume. +*/ +struct mnt3_export * +mnt3_mntpath_to_export (struct mount3_state *ms, char *dirpath) +{ + struct mnt3_export *exp = NULL; + struct mnt3_export *found = NULL; + + if ((!ms) || (!dirpath)) + return NULL; + + LOCK (&ms->mountlock); + list_for_each_entry (exp, &ms->exportlist, explist) { + + /* Search for the an exact match with the volume */ + if (mnt3_match_dirpath_export (exp->expname, dirpath)) { + found = exp; + gf_log (GF_MNT, GF_LOG_DEBUG, "Found export volume: " + "%s", exp->vol->name); + goto foundexp; + } + } + + gf_log (GF_MNT, GF_LOG_DEBUG, "Export not found"); +foundexp: + UNLOCK (&ms->mountlock); + return found; +} + + +int +mnt3_check_client_net (struct mount3_state *ms, rpcsvc_request_t *req, + xlator_t *targetxl) +{ + + rpcsvc_t *svc = NULL; + rpc_transport_t *trans = NULL; + struct sockaddr_storage sastorage = {0,}; + char peer[RPCSVC_PEER_STRLEN] = {0,}; + int ret = -1; + + if ((!ms) || (!req) || (!targetxl)) + return -1; + + svc = rpcsvc_request_service (req); + + trans = rpcsvc_request_transport (req); + ret = rpcsvc_transport_peeraddr (trans, peer, RPCSVC_PEER_STRLEN, + &sastorage, sizeof (sastorage)); + if (ret != 0) { + gf_log (GF_MNT, GF_LOG_WARNING, "Failed to get peer addr: %s", + gai_strerror (ret)); + } + + ret = rpcsvc_auth_check (svc, targetxl->name, trans); + if (ret == RPCSVC_AUTH_REJECT) { + gf_log (GF_MNT, GF_LOG_INFO, "Peer %s not allowed", peer); + goto err; + } + + ret = rpcsvc_transport_privport_check (svc, targetxl->name, + rpcsvc_request_transport (req)); + if (ret == RPCSVC_AUTH_REJECT) { + gf_log (GF_MNT, GF_LOG_INFO, "Peer %s rejected. Unprivileged " + "port not allowed", peer); + goto err; + } + + ret = 0; +err: + return ret; +} + + +int +mnt3_parse_dir_exports (rpcsvc_request_t *req, struct mount3_state *ms, + char *subdir) +{ + char volname[1024]; + struct mnt3_export *exp = NULL; + char *volname_ptr = NULL; + int ret = -1; + + if ((!ms) || (!subdir)) + return -1; + + volname_ptr = volname; + subdir = __volume_subdir (subdir, &volname_ptr); + if (!subdir) + goto err; + + exp = mnt3_mntpath_to_export (ms, volname); + if (!exp) + goto err; + + ret = mnt3_resolve_subdir (req, ms, exp, subdir); + if (ret < 0) { + gf_log (GF_MNT, GF_LOG_ERROR, "Failed to resolve export dir: %s" + , subdir); + goto err; + } + +err: + return ret; +} + + +int +mnt3_find_export (rpcsvc_request_t *req, char *path, struct mnt3_export **e) +{ + int ret = -EFAULT; + struct mount3_state *ms = NULL; + struct mnt3_export *exp = NULL; + + if ((!req) || (!path) || (!e)) + return -1; + + ms = (struct mount3_state *) rpcsvc_request_program_private (req); + if (!ms) { + gf_log (GF_MNT, GF_LOG_ERROR, "Mount state not present"); + rpcsvc_request_seterr (req, SYSTEM_ERR); + goto err; + } + + gf_log (GF_MNT, GF_LOG_DEBUG, "dirpath: %s", path); + exp = mnt3_mntpath_to_export (ms, path); + if (exp) { + ret = 0; + *e = exp; + goto err; + } + + if (!gf_mnt3_export_dirs(ms)) { + ret = -1; + goto err; + } + + ret = mnt3_parse_dir_exports (req, ms, path); + if (ret == 0) { + ret = -2; + goto err; + } + +err: + return ret; +} + + +int +mnt3svc_mnt (rpcsvc_request_t *req) +{ + struct iovec pvec = {0, }; + char path[MNTPATHLEN]; + int ret = -1; + struct mount3_state *ms = NULL; + mountstat3 mntstat = MNT3ERR_SERVERFAULT; + struct mnt3_export *exp = NULL; + struct nfs_state *nfs = NULL; + + if (!req) + return -1; + + pvec.iov_base = path; + pvec.iov_len = MNTPATHLEN; + ret = xdr_to_mountpath (pvec, req->msg[0]); + if (ret == -1) { + gf_log (GF_MNT, GF_LOG_ERROR, "Failed to decode args"); + rpcsvc_request_seterr (req, GARBAGE_ARGS); + goto rpcerr; + } + + ms = (struct mount3_state *)rpcsvc_request_program_private (req); + if (!ms) { + gf_log (GF_MNT, GF_LOG_ERROR, "Mount state not present"); + rpcsvc_request_seterr (req, SYSTEM_ERR); + ret = -1; + goto rpcerr; + } + + ret = 0; + nfs = (struct nfs_state *)ms->nfsx->private; + gf_log (GF_MNT, GF_LOG_DEBUG, "dirpath: %s", path); + ret = mnt3_find_export (req, path, &exp); + if (ret == -2) { + ret = 0; + goto rpcerr; + } else if (ret < 0) { + ret = -1; + mntstat = MNT3ERR_NOENT; + goto mnterr; + } + + if (!nfs_subvolume_started (nfs, exp->vol)) { + gf_log (GF_MNT, GF_LOG_DEBUG, "Volume %s not started", + exp->vol->name); + ret = -1; + mntstat = MNT3ERR_NOENT; + goto mnterr; + } + + ret = mnt3_check_client_net (ms, req, exp->vol); + if (ret == RPCSVC_AUTH_REJECT) { + mntstat = MNT3ERR_ACCES; + gf_log (GF_MNT, GF_LOG_DEBUG, "Client mount not allowed"); + ret = -1; + goto mnterr; + } + + ret = mnt3svc_mount (req, ms, exp); + if (ret < 0) + mntstat = mnt3svc_errno_to_mnterr (-ret); +mnterr: + if (ret < 0) { + mnt3svc_mnt_error_reply (req, mntstat); + ret = 0; + } + +rpcerr: + return ret; +} + + +int +mnt3svc_null (rpcsvc_request_t *req) +{ + struct iovec dummyvec = {0, }; + + if (!req) { + gf_log (GF_MNT, GF_LOG_ERROR, "Got NULL request!"); + return 0; + } + rpcsvc_submit_generic (req, &dummyvec, 1, NULL, 0, NULL); + return 0; +} + + +mountlist +__build_mountlist (struct mount3_state *ms, int *count) +{ + struct mountbody *mlist = NULL; + struct mountbody *prev = NULL; + struct mountbody *first = NULL; + size_t namelen = 0; + int ret = -1; + struct mountentry *me = NULL; + + if ((!ms) || (!count)) + return NULL; + + /* read rmtab, other peers might have updated it */ + mount_read_rmtab(ms); + + *count = 0; + gf_log (GF_MNT, GF_LOG_DEBUG, "Building mount list:"); + list_for_each_entry (me, &ms->mountlist, mlist) { + namelen = strlen (me->exname); + mlist = GF_CALLOC (1, sizeof (*mlist), gf_nfs_mt_mountbody); + if (!mlist) { + gf_log (GF_MNT, GF_LOG_ERROR, "Memory allocation" + " failed"); + goto free_list; + } + if (!first) + first = mlist; + + mlist->ml_directory = GF_CALLOC (namelen + 2, sizeof (char), + gf_nfs_mt_char); + if (!mlist->ml_directory) { + gf_log (GF_MNT, GF_LOG_ERROR, "Memory allocation" + " failed"); + goto free_list; + } + + strcpy (mlist->ml_directory, me->exname); + + namelen = strlen (me->hostname); + mlist->ml_hostname = GF_CALLOC (namelen + 2, sizeof (char), + gf_nfs_mt_char); + if (!mlist->ml_hostname) { + gf_log (GF_MNT, GF_LOG_ERROR, "Memory allocation" + " failed"); + goto free_list; + } + + strcat (mlist->ml_hostname, me->hostname); + + gf_log (GF_MNT, GF_LOG_DEBUG, "mount entry: dir: %s, host: %s", + mlist->ml_directory, mlist->ml_hostname); + if (prev) { + prev->ml_next = mlist; + prev = mlist; + } else + prev = mlist; + + (*count)++; + } + + ret = 0; + +free_list: + if (ret == -1) { + xdr_free_mountlist (first); + first = NULL; + } + + return first; +} + + +mountlist +mnt3svc_build_mountlist (struct mount3_state *ms, int *count) +{ + struct mountbody *first = NULL; + + LOCK (&ms->mountlock); + { + first = __build_mountlist (ms, count); + } + UNLOCK (&ms->mountlock); + + return first; +} + + +int +mnt3svc_dump (rpcsvc_request_t *req) +{ + int ret = -1; + struct mount3_state *ms = NULL; + mountlist mlist; + mountstat3 mstat = 0; + mnt3_serializer sfunc = NULL; + void *arg = NULL; + + + if (!req) + return -1; + + ms = (struct mount3_state *)rpcsvc_request_program_private (req); + if (!ms) { + rpcsvc_request_seterr (req, SYSTEM_ERR); + goto rpcerr; + } + + sfunc = (mnt3_serializer)xdr_serialize_mountlist; + mlist = mnt3svc_build_mountlist (ms, &ret); + arg = &mlist; + + if (!mlist) { + if (ret != 0) { + rpcsvc_request_seterr (req, SYSTEM_ERR); + ret = -1; + goto rpcerr; + } else { + arg = &mstat; + sfunc = (mnt3_serializer)xdr_serialize_mountstat3; + } + } + + mnt3svc_submit_reply (req, arg, sfunc); + + xdr_free_mountlist (mlist); + ret = 0; + +rpcerr: + return ret; +} + + +int +mnt3svc_umount (struct mount3_state *ms, char *dirpath, char *hostname) +{ + struct mountentry *me = NULL; + int ret = -1; + gf_store_handle_t *sh = NULL; + struct nfs_state *nfs = NULL; + + if ((!ms) || (!dirpath) || (!hostname)) + return -1; + + nfs = (struct nfs_state *)ms->nfsx->private; + + ret = gf_store_handle_new (nfs->rmtab, &sh); + if (ret) { + gf_log (GF_MNT, GF_LOG_WARNING, "Failed to open '%s'", + nfs->rmtab); + return 0; + } + + ret = gf_store_lock (sh); + if (ret) { + goto out_free; + } + + LOCK (&ms->mountlock); + { + __mount_read_rmtab (sh, &ms->mountlist, _gf_false); + if (list_empty (&ms->mountlist)) { + ret = 0; + goto out_unlock; + } + + ret = -1; + list_for_each_entry (me, &ms->mountlist, mlist) { + if ((strcmp (me->exname, dirpath) == 0) && + (strcmp (me->hostname, hostname) == 0)) { + ret = 0; + break; + } + } + + /* Need this check here because at the end of the search me + * might still be pointing to the last entry, which may not be + * the one we're looking for. + */ + if (ret == -1) {/* Not found in list. */ + gf_log (GF_MNT, GF_LOG_TRACE, "Export not found"); + goto out_unlock; + } + + if (!me) + goto out_unlock; + + gf_log (GF_MNT, GF_LOG_DEBUG, "Unmounting: dir %s, host: %s", + me->exname, me->hostname); + + list_del (&me->mlist); + GF_FREE (me); + __mount_rewrite_rmtab (ms, sh); + } +out_unlock: + UNLOCK (&ms->mountlock); + gf_store_unlock (sh); +out_free: + gf_store_handle_destroy (sh); + return ret; +} + + +int +mnt3svc_umnt (rpcsvc_request_t *req) +{ + char hostname[MNTPATHLEN]; + char dirpath[MNTPATHLEN]; + struct iovec pvec = {0, }; + int ret = -1; + struct mount3_state *ms = NULL; + mountstat3 mstat = MNT3_OK; + char *colon = NULL; + + if (!req) + return -1; + + /* Remove the mount point from the exports list. */ + pvec.iov_base = dirpath; + pvec.iov_len = MNTPATHLEN; + ret = xdr_to_mountpath (pvec, req->msg[0]); + if (ret == -1) { + gf_log (GF_MNT, GF_LOG_ERROR, "Failed decode args"); + rpcsvc_request_seterr (req, GARBAGE_ARGS); + goto rpcerr; + } + + ms = (struct mount3_state *)rpcsvc_request_program_private (req); + if (!ms) { + gf_log (GF_MNT, GF_LOG_ERROR, "Mount state not present"); + rpcsvc_request_seterr (req, SYSTEM_ERR); + ret = -1; + goto rpcerr; + } + + ret = rpcsvc_transport_peername (req->trans, hostname, MNTPATHLEN); + if (ret != 0) { + gf_log (GF_MNT, GF_LOG_ERROR, "Failed to get remote name: %s", + gai_strerror (ret)); + goto rpcerr; + } + + colon = strrchr (hostname, ':'); + if (colon) { + *colon= '\0'; + } + gf_log (GF_MNT, GF_LOG_DEBUG, "dirpath: %s, hostname: %s", dirpath, + hostname); + ret = mnt3svc_umount (ms, dirpath, hostname); + + if (ret == -1) { + ret = 0; + mstat = MNT3ERR_NOENT; + } + /* FIXME: also take care of the corner case where the + * client was resolvable at mount but not at the umount - vice-versa. + */ + mnt3svc_submit_reply (req, &mstat, + (mnt3_serializer)xdr_serialize_mountstat3); + +rpcerr: + return ret; +} + + +int +__mnt3svc_umountall (struct mount3_state *ms) +{ + struct mountentry *me = NULL; + struct mountentry *tmp = NULL; + + if (!ms) + return -1; + + if (list_empty (&ms->mountlist)) + return 0; + + list_for_each_entry_safe (me, tmp, &ms->mountlist, mlist) { + list_del (&me->mlist); + GF_FREE (me); + } + + return 0; +} + + +int +mnt3svc_umountall (struct mount3_state *ms) +{ + int ret = -1; + if (!ms) + return -1; + + LOCK (&ms->mountlock); + { + ret = __mnt3svc_umountall (ms); + } + UNLOCK (&ms->mountlock); + + return ret; +} + + +int +mnt3svc_umntall (rpcsvc_request_t *req) +{ + int ret = RPCSVC_ACTOR_ERROR; + struct mount3_state *ms = NULL; + mountstat3 mstat = MNT3_OK; + + if (!req) + return ret; + + ms = (struct mount3_state *)rpcsvc_request_program_private (req); + if (!ms) { + gf_log (GF_MNT, GF_LOG_ERROR, "Mount state not present"); + rpcsvc_request_seterr (req, SYSTEM_ERR); + goto rpcerr; + } + + mnt3svc_umountall (ms); + mnt3svc_submit_reply (req, &mstat, + (mnt3_serializer)xdr_serialize_mountstat3); + + ret = RPCSVC_ACTOR_SUCCESS; +rpcerr: + return ret; +} + + +exports +mnt3_xlchildren_to_exports (rpcsvc_t *svc, struct mount3_state *ms) +{ + struct exportnode *elist = NULL; + struct exportnode *prev = NULL; + struct exportnode *first = NULL; + size_t namelen = 0; + int ret = -1; + char *addrstr = NULL; + struct mnt3_export *ent = NULL; + struct nfs_state *nfs = NULL; + + if ((!ms) || (!svc)) + return NULL; + + nfs = (struct nfs_state *)ms->nfsx->private; + if (!nfs) + return NULL; + + LOCK (&ms->mountlock); + list_for_each_entry(ent, &ms->exportlist, explist) { + + /* If volume is not started yet, do not list it for tools like + * showmount. + */ + if (!nfs_subvolume_started (nfs, ent->vol)) + continue; + + namelen = strlen (ent->expname) + 1; + elist = GF_CALLOC (1, sizeof (*elist), gf_nfs_mt_exportnode); + if (!elist) { + gf_log (GF_MNT, GF_LOG_ERROR, "Memory allocation" + " failed"); + goto free_list; + } + if (!first) + first = elist; + elist->ex_dir = GF_CALLOC (namelen + 2, sizeof (char), + gf_nfs_mt_char); + if (!elist->ex_dir) { + gf_log (GF_MNT, GF_LOG_ERROR, "Memory allocation" + " failed"); + goto free_list; + } + strcpy (elist->ex_dir, ent->expname); + + addrstr = rpcsvc_volume_allowed (svc->options, + ent->vol->name); + elist->ex_groups = GF_CALLOC (1, sizeof (struct groupnode), + gf_nfs_mt_groupnode); + if (!elist->ex_groups) { + gf_log (GF_MNT, GF_LOG_ERROR, "Memory allocation" + " failed"); + goto free_list; + } + /*This check has to be done after checking + * elist->ex_groups allocation check to avoid resource leak; + */ + if (addrstr) + addrstr = gf_strdup (addrstr); + else + addrstr = gf_strdup ("No Access"); + + if (!addrstr) { + goto free_list; + } + elist->ex_groups->gr_name = addrstr; + if (prev) { + prev->ex_next = elist; + prev = elist; + } else + prev = elist; + } + + ret = 0; + +free_list: + UNLOCK (&ms->mountlock); + if (ret == -1) { + xdr_free_exports_list (first); + first = NULL; + } + + return first; +} + + +int +mnt3svc_export (rpcsvc_request_t *req) +{ + struct mount3_state *ms = NULL; + exports elist = NULL; + int ret = -1; + + if (!req) + return -1; + + ms = (struct mount3_state *)rpcsvc_request_program_private (req); + if (!ms) { + gf_log (GF_MNT, GF_LOG_ERROR, "mount state not found"); + rpcsvc_request_seterr (req, SYSTEM_ERR); + goto err; + } + + /* Using the children translator names, build the export list */ + elist = mnt3_xlchildren_to_exports (rpcsvc_request_service (req), + ms); + /* Do not return error when exports list is empty. An exports list can + * be empty when no subvolumes have come up. No point returning error + * and confusing the user. + if (!elist) { + gf_log (GF_MNT, GF_LOG_ERROR, "Failed to build exports list"); + nfs_rpcsvc_request_seterr (req, SYSTEM_ERR); + goto err; + } + */ + + /* Note how the serializer is passed to the generic reply function. */ + mnt3svc_submit_reply (req, &elist, + (mnt3_serializer)xdr_serialize_exports); + + xdr_free_exports_list (elist); + ret = 0; +err: + return ret; +} + +/* just declaring, definition is way down below */ +rpcsvc_program_t mnt3prog; + +/* nfs3_rootfh used by mount3udp thread needs to access mount3prog.private + * directly as we don't have nfs xlator pointer to dereference it. But thats OK + */ + +struct nfs3_fh * +nfs3_rootfh (char* path) +{ + struct mount3_state *ms = NULL; + struct nfs3_fh *fh = NULL; + struct mnt3_export *exp = NULL; + inode_t *inode = NULL; + char *tmp = NULL; + + ms = mnt3prog.private; + exp = mnt3_mntpath_to_export (ms, path); + if (exp == NULL) + goto err; + + tmp = (char *)path; + tmp = strchr (tmp, '/'); + if (tmp == NULL) + tmp = "/"; + + inode = inode_from_path (exp->vol->itable, tmp); + if (inode == NULL) + goto err; + + fh = GF_CALLOC (1, sizeof(*fh), gf_nfs_mt_nfs3_fh); + if (fh == NULL) + goto err; + nfs3_build_fh (inode, exp->volumeid, fh); + +err: + if (inode) + inode_unref (inode); + return fh; +} + +int +mount3udp_add_mountlist (char *host, dirpath *expname) +{ + struct mountentry *me = NULL; + struct mount3_state *ms = NULL; + char *export = NULL; + + ms = mnt3prog.private; + me = GF_CALLOC (1, sizeof (*me), gf_nfs_mt_mountentry); + if (!me) + return -1; + export = (char *)expname; + while (*export == '/') + export++; + + strncpy (me->exname, export, MNTPATHLEN); + strncpy (me->hostname, host, MNTPATHLEN); + INIT_LIST_HEAD (&me->mlist); + LOCK (&ms->mountlock); + { + list_add_tail (&me->mlist, &ms->mountlist); + mount_rewrite_rmtab(ms, NULL); + } + UNLOCK (&ms->mountlock); + return 0; +} + +int +mount3udp_delete_mountlist (char *hostname, dirpath *expname) +{ + struct mount3_state *ms = NULL; + char *export = NULL; + + ms = mnt3prog.private; + export = (char *)expname; + while (*export == '/') + export++; + mnt3svc_umount (ms, export, hostname); + return 0; +} + +/** + * This function will parse the hostip (IP addres, IP range, or hostname) + * and fill the host_auth_spec structure. + * + * @param hostspec - struct host_auth_spec + * @param hostip - IP address, IP range (CIDR format) or hostname + * + * @return 0 - on success and -1 on failure + */ +int +mnt3_export_fill_hostspec (struct host_auth_spec* hostspec, const char* hostip) +{ + char *ipdupstr = NULL; + char *savptr = NULL; + char *ip = NULL; + char *token = NULL; + int ret = -1; + + /* Create copy of the string so that the source won't change + */ + ipdupstr = gf_strdup (hostip); + if (NULL == ipdupstr) { + gf_log (GF_MNT, GF_LOG_ERROR, "Memory allocation failed"); + goto err; + } + + ip = strtok_r (ipdupstr, "/", &savptr); + hostspec->host_addr = gf_strdup (ip); + if (NULL == hostspec->host_addr) { + gf_log (GF_MNT, GF_LOG_ERROR, "Memory allocation failed"); + goto err; + } + + /* Check if the IP is in <IP address> / <Range> format. + * If yes, then strip the range and store it separately. + */ + token = strtok_r (NULL, "/", &savptr); + + if (NULL == token) { + hostspec->routeprefix = -1; + } else { + hostspec->routeprefix = atoi (token); + } + + // success + ret = 0; +err: + if (NULL != ipdupstr) { + GF_FREE (ipdupstr); + } + return ret; +} + + +/** + * This function will parse the AUTH parameter passed along with + * "export-dir" option. If AUTH parameter is present then it will be + * stripped from exportpath and stored in mnt3_export (exp) structure. + * + * @param exp - mnt3_export structure. Holds information needed for mount. + * @param exportpath - Value of "export-dir" key. Holds both export path + * and AUTH parameter for the path. + * exportpath format: <abspath>[(hostdesc[|hostspec|...])] + * + * @return This function will return 0 on success and -1 on failure. + */ +int +mnt3_export_parse_auth_param (struct mnt3_export* exp, char* exportpath) +{ + char *token = NULL; + char *savPtr = NULL; + char *hostip = NULL; + struct host_auth_spec *host = NULL; + int ret = 0; + + /* Using exportpath directly in strtok_r because we want + * to strip off AUTH parameter from exportpath. */ + token = strtok_r (exportpath, "(", &savPtr); + + /* Get the next token, which will be the AUTH parameter. */ + token = strtok_r (NULL, ")", &savPtr); + + if (NULL == token) { + /* If AUTH is not present then we should return success. */ + return 0; + } + + /* Free any previously allocated hostspec structure. */ + if (NULL != exp->hostspec) { + GF_FREE (exp->hostspec); + exp->hostspec = NULL; + } + + exp->hostspec = GF_CALLOC (1, + sizeof (*(exp->hostspec)), + gf_nfs_mt_auth_spec); + if (NULL == exp->hostspec){ + gf_log (GF_MNT, GF_LOG_ERROR, "Memory allocation failed"); + return -1; + } + + /* AUTH parameter can have multiple entries. For each entry + * a host_auth_spec structure is created. */ + host = exp->hostspec; + + hostip = strtok_r (token, "|", &savPtr); + + /* Parse all AUTH parameters separated by '|' */ + while (NULL != hostip){ + ret = mnt3_export_fill_hostspec (host, hostip); + if (0 != ret) { + gf_log(GF_MNT, GF_LOG_WARNING, + "Failed to parse hostspec: %s", hostip); + goto err; + } + + hostip = strtok_r (NULL, "|", &savPtr); + if (NULL == hostip) { + break; + } + + host->next = GF_CALLOC (1, sizeof (*(host)), + gf_nfs_mt_auth_spec); + if (NULL == host->next){ + gf_log (GF_MNT,GF_LOG_ERROR, + "Memory allocation failed"); + goto err; + } + host = host->next; + } + + /* In case of success return from here */ + return 0; +err: + /* In case of failure free up hostspec structure. */ + FREE_HOSTSPEC (exp); + + return -1; +} + +/** + * exportpath will also have AUTH options (ip address, subnet address or + * hostname) mentioned. + * exportpath format: <abspath>[(hostdesc[|hostspec|...])] + */ +struct mnt3_export * +mnt3_init_export_ent (struct mount3_state *ms, xlator_t *xl, char *exportpath, + uuid_t volumeid) +{ + struct mnt3_export *exp = NULL; + int alloclen = 0; + int ret = -1; + + if ((!ms) || (!xl)) + return NULL; + + exp = GF_CALLOC (1, sizeof (*exp), gf_nfs_mt_mnt3_export); + if (!exp) { + gf_log (GF_MNT, GF_LOG_ERROR, "Memory allocation failed"); + return NULL; + } + + if (NULL != exportpath) { + /* If exportpath is not NULL then we should check if AUTH + * parameter is present or not. If AUTH parameter is present + * then it will be stripped and stored in mnt3_export (exp) + * structure. + */ + if (0 != mnt3_export_parse_auth_param (exp, exportpath)){ + gf_log (GF_MNT, GF_LOG_ERROR, + "Failed to parse auth param"); + goto err; + } + } + + + INIT_LIST_HEAD (&exp->explist); + if (exportpath) + alloclen = strlen (xl->name) + 2 + strlen (exportpath); + else + alloclen = strlen (xl->name) + 2; + + exp->expname = GF_CALLOC (alloclen, sizeof (char), gf_nfs_mt_char); + if (!exp->expname) { + gf_log (GF_MNT, GF_LOG_ERROR, "Memory allocation failed"); + goto err; + } + + if (exportpath) { + gf_log (GF_MNT, GF_LOG_TRACE, "Initing dir export: %s:%s", + xl->name, exportpath); + exp->exptype = MNT3_EXPTYPE_DIR; + ret = snprintf (exp->expname, alloclen, "/%s%s", xl->name, + exportpath); + } else { + gf_log (GF_MNT, GF_LOG_TRACE, "Initing volume export: %s", + xl->name); + exp->exptype = MNT3_EXPTYPE_VOLUME; + ret = snprintf (exp->expname, alloclen, "/%s", xl->name); + } + if (ret < 0) { + gf_log (xl->name, GF_LOG_ERROR, + "Failed to set the export name"); + goto err; + } + /* Just copy without discrimination, we'll determine whether to + * actually use it when a mount request comes in and a file handle + * needs to be built. + */ + uuid_copy (exp->volumeid, volumeid); + exp->vol = xl; + + /* On success we should return from here*/ + return exp; +err: + /* On failure free exp and it's members.*/ + if (NULL != exp) { + mnt3_export_free (exp); + exp = NULL; + } + + return exp; +} + + +int +__mnt3_init_volume_direxports (struct mount3_state *ms, xlator_t *xlator, + char *optstr, uuid_t volumeid) +{ + struct mnt3_export *newexp = NULL; + int ret = -1; + char *savptr = NULL; + char *dupopt = NULL; + char *token = NULL; + + if ((!ms) || (!xlator) || (!optstr)) + return -1; + + dupopt = gf_strdup (optstr); + if (!dupopt) { + gf_log (GF_MNT, GF_LOG_ERROR, "gf_strdup failed"); + goto err; + } + + token = strtok_r (dupopt, ",", &savptr); + while (token) { + newexp = mnt3_init_export_ent (ms, xlator, token, volumeid); + if (!newexp) { + gf_log (GF_MNT, GF_LOG_ERROR, "Failed to init dir " + "export: %s", token); + ret = -1; + goto err; + } + + list_add_tail (&newexp->explist, &ms->exportlist); + token = strtok_r (NULL, ",", &savptr); + } + + ret = 0; +err: + GF_FREE (dupopt); + + return ret; +} + + +int +__mnt3_init_volume (struct mount3_state *ms, dict_t *opts, xlator_t *xlator) +{ + struct mnt3_export *newexp = NULL; + int ret = -1; + char searchstr[1024]; + char *optstr = NULL; + uuid_t volumeid = {0, }; + + if ((!ms) || (!xlator) || (!opts)) + return -1; + + uuid_clear (volumeid); + if (gf_nfs_dvm_off (nfs_state (ms->nfsx))) + goto no_dvm; + + ret = snprintf (searchstr, 1024, "nfs3.%s.volume-id", xlator->name); + if (ret < 0) { + gf_log (GF_MNT, GF_LOG_ERROR, "snprintf failed"); + ret = -1; + goto err; + } + + if (dict_get (opts, searchstr)) { + ret = dict_get_str (opts, searchstr, &optstr); + if (ret < 0) { + gf_log (GF_MNT, GF_LOG_ERROR, "Failed to read option" + ": %s", searchstr); + ret = -1; + goto err; + } + } else { + gf_log (GF_MNT, GF_LOG_ERROR, "DVM is on but volume-id not " + "given for volume: %s", xlator->name); + ret = -1; + goto err; + } + + if (optstr) { + ret = uuid_parse (optstr, volumeid); + if (ret < 0) { + gf_log (GF_MNT, GF_LOG_ERROR, "Failed to parse volume " + "UUID"); + ret = -1; + goto err; + } + } + +no_dvm: + ret = snprintf (searchstr, 1024, "nfs3.%s.export-dir", xlator->name); + if (ret < 0) { + gf_log (GF_MNT, GF_LOG_ERROR, "snprintf failed"); + ret = -1; + goto err; + } + + if (dict_get (opts, searchstr)) { + ret = dict_get_str (opts, searchstr, &optstr); + if (ret < 0) { + gf_log (GF_MNT, GF_LOG_ERROR, "Failed to read option: " + "%s", searchstr); + ret = -1; + goto err; + } + + ret = __mnt3_init_volume_direxports (ms, xlator, optstr, + volumeid); + if (ret == -1) { + gf_log (GF_MNT, GF_LOG_ERROR, "Dir export setup failed" + " for volume: %s", xlator->name); + goto err; + } + } + + if (ms->export_volumes) { + newexp = mnt3_init_export_ent (ms, xlator, NULL, volumeid); + if (!newexp) { + ret = -1; + goto err; + } + + list_add_tail (&newexp->explist, &ms->exportlist); + } + ret = 0; + + +err: + return ret; +} + + +int +__mnt3_init_volume_export (struct mount3_state *ms, dict_t *opts) +{ + int ret = -1; + char *optstr = NULL; + /* On by default. */ + gf_boolean_t boolt = _gf_true; + + if ((!ms) || (!opts)) + return -1; + + if (!dict_get (opts, "nfs3.export-volumes")) { + ret = 0; + goto err; + } + + ret = dict_get_str (opts, "nfs3.export-volumes", &optstr); + if (ret < 0) { + gf_log (GF_MNT, GF_LOG_ERROR, "Failed to read option: " + "nfs3.export-volumes"); + ret = -1; + goto err; + } + + ret = gf_string2boolean (optstr, &boolt); + if (ret < 0) { + gf_log (GF_MNT, GF_LOG_ERROR, "Failed to convert" + " string to boolean"); + } + +err: + if (boolt == _gf_false) { + gf_log (GF_MNT, GF_LOG_TRACE, "Volume exports disabled"); + ms->export_volumes = 0; + } else { + gf_log (GF_MNT, GF_LOG_TRACE, "Volume exports enabled"); + ms->export_volumes = 1; + } + + return ret; +} + + +int +__mnt3_init_dir_export (struct mount3_state *ms, dict_t *opts) +{ + int ret = -1; + char *optstr = NULL; + /* On by default. */ + gf_boolean_t boolt = _gf_true; + + if ((!ms) || (!opts)) + return -1; + + if (!dict_get (opts, "nfs3.export-dirs")) { + ret = 0; + goto err; + } + + ret = dict_get_str (opts, "nfs3.export-dirs", &optstr); + if (ret < 0) { + gf_log (GF_MNT, GF_LOG_ERROR, "Failed to read option: " + "nfs3.export-dirs"); + ret = -1; + goto err; + } + + ret = gf_string2boolean (optstr, &boolt); + if (ret < 0) { + gf_log (GF_MNT, GF_LOG_ERROR, "Failed to convert" + " string to boolean"); + } + +err: + if (boolt == _gf_false) { + gf_log (GF_MNT, GF_LOG_TRACE, "Dir exports disabled"); + ms->export_dirs = 0; + } else { + gf_log (GF_MNT, GF_LOG_TRACE, "Dir exports enabled"); + ms->export_dirs = 1; + } + + return ret; +} + + +int +mnt3_init_options (struct mount3_state *ms, dict_t *options) +{ + xlator_list_t *volentry = NULL; + int ret = -1; + + if ((!ms) || (!options)) + return -1; + + __mnt3_init_volume_export (ms, options); + __mnt3_init_dir_export (ms, options); + volentry = ms->nfsx->children; + while (volentry) { + gf_log (GF_MNT, GF_LOG_TRACE, "Initing options for: %s", + volentry->xlator->name); + ret = __mnt3_init_volume (ms, options, volentry->xlator); + if (ret < 0) { + gf_log (GF_MNT, GF_LOG_ERROR, "Volume init failed"); + goto err; + } + + volentry = volentry->next; + } + + ret = 0; +err: + return ret; +} + + +struct mount3_state * +mnt3_init_state (xlator_t *nfsx) +{ + struct mount3_state *ms = NULL; + int ret = -1; + + if (!nfsx) + return NULL; + + ms = GF_CALLOC (1, sizeof (*ms), gf_nfs_mt_mount3_state); + if (!ms) { + gf_log (GF_MNT, GF_LOG_ERROR, "Memory allocation failed"); + return NULL; + } + + ms->iobpool = nfsx->ctx->iobuf_pool; + ms->nfsx = nfsx; + INIT_LIST_HEAD (&ms->exportlist); + ret = mnt3_init_options (ms, nfsx->options); + if (ret < 0) { + gf_log (GF_MNT, GF_LOG_ERROR, "Options init failed"); + return NULL; + } + + INIT_LIST_HEAD (&ms->mountlist); + LOCK_INIT (&ms->mountlock); + + return ms; +} + +int +mount_init_state (xlator_t *nfsx) +{ + int ret = -1; + struct nfs_state *nfs = NULL; + + if (!nfsx) + goto out; + + nfs = (struct nfs_state *)nfs_state (nfsx); + /*Maintaining global state for MOUNT1 and MOUNT3*/ + nfs->mstate = mnt3_init_state (nfsx); + if (!nfs->mstate) { + gf_log (GF_NFS, GF_LOG_ERROR, "Failed to allocate" + "mount state"); + goto out; + } + ret = 0; +out: + return ret; +} + +rpcsvc_actor_t mnt3svc_actors[MOUNT3_PROC_COUNT] = { + {"NULL", MOUNT3_NULL, mnt3svc_null, NULL, 0, DRC_NA}, + {"MNT", MOUNT3_MNT, mnt3svc_mnt, NULL, 0, DRC_NA}, + {"DUMP", MOUNT3_DUMP, mnt3svc_dump, NULL, 0, DRC_NA}, + {"UMNT", MOUNT3_UMNT, mnt3svc_umnt, NULL, 0, DRC_NA}, + {"UMNTALL", MOUNT3_UMNTALL, mnt3svc_umntall, NULL, 0, DRC_NA}, + {"EXPORT", MOUNT3_EXPORT, mnt3svc_export, NULL, 0, DRC_NA} +}; + + + +/* Static init parts are assigned here, dynamic ones are done in + * mnt3svc_init and mnt3_init_state. + * Making MOUNT3 a synctask so that the blocking DNS calls during rpc auth + * gets offloaded to syncenv, keeping the main/poll thread unblocked + */ +rpcsvc_program_t mnt3prog = { + .progname = "MOUNT3", + .prognum = MOUNT_PROGRAM, + .progver = MOUNT_V3, + .progport = GF_MOUNTV3_PORT, + .actors = mnt3svc_actors, + .numactors = MOUNT3_PROC_COUNT, + .min_auth = AUTH_NULL, + .synctask = _gf_true, +}; + + + +rpcsvc_program_t * +mnt3svc_init (xlator_t *nfsx) +{ + struct mount3_state *mstate = NULL; + struct nfs_state *nfs = NULL; + dict_t *options = NULL; + char *portstr = NULL; + int ret = -1; + pthread_t udp_thread; + + if (!nfsx || !nfsx->private) + return NULL; + + nfs = (struct nfs_state *)nfsx->private; + + gf_log (GF_MNT, GF_LOG_DEBUG, "Initing Mount v3 state"); + mstate = (struct mount3_state *)nfs->mstate; + if (!mstate) { + gf_log (GF_MNT, GF_LOG_ERROR, "Mount v3 state init failed"); + goto err; + } + + mnt3prog.private = mstate; + options = dict_new (); + + ret = gf_asprintf (&portstr, "%d", GF_MOUNTV3_PORT); + if (ret == -1) + goto err; + + ret = dict_set_dynstr (options, "transport.socket.listen-port", portstr); + if (ret == -1) + goto err; + ret = dict_set_str (options, "transport-type", "socket"); + if (ret == -1) { + gf_log (GF_NFS, GF_LOG_ERROR, "dict_set_str error"); + goto err; + } + + if (nfs->allow_insecure) { + ret = dict_set_str (options, "rpc-auth-allow-insecure", "on"); + if (ret == -1) { + gf_log (GF_NFS, GF_LOG_ERROR, "dict_set_str error"); + goto err; + } + ret = dict_set_str (options, "rpc-auth.ports.insecure", "on"); + if (ret == -1) { + gf_log (GF_NFS, GF_LOG_ERROR, "dict_set_str error"); + goto err; + } + } + + ret= rpcsvc_create_listeners (nfs->rpcsvc, options, nfsx->name); + if (ret == -1) { + gf_log (GF_NFS, GF_LOG_ERROR, "Unable to create listeners"); + dict_unref (options); + goto err; + } + + if (nfs->mount_udp) { + pthread_create (&udp_thread, NULL, mount3udp_thread, NULL); + } + return &mnt3prog; +err: + return NULL; +} + + +rpcsvc_actor_t mnt1svc_actors[MOUNT1_PROC_COUNT] = { + {"NULL", MOUNT1_NULL, mnt3svc_null, NULL, 0, DRC_NA}, + {"MNT", MOUNT1_MNT, NULL, NULL, 0, DRC_NA }, + {"DUMP", MOUNT1_DUMP, mnt3svc_dump, NULL, 0, DRC_NA}, + {"UMNT", MOUNT1_UMNT, mnt3svc_umnt, NULL, 0, DRC_NA}, + {"UMNTALL", MOUNT1_UMNTALL, NULL, NULL, 0, DRC_NA}, + {"EXPORT", MOUNT1_EXPORT, mnt3svc_export, NULL, 0, DRC_NA} +}; + +rpcsvc_program_t mnt1prog = { + .progname = "MOUNT1", + .prognum = MOUNT_PROGRAM, + .progver = MOUNT_V1, + .progport = GF_MOUNTV1_PORT, + .actors = mnt1svc_actors, + .numactors = MOUNT1_PROC_COUNT, + .min_auth = AUTH_NULL, + .synctask = _gf_true, +}; + + +rpcsvc_program_t * +mnt1svc_init (xlator_t *nfsx) +{ + struct mount3_state *mstate = NULL; + struct nfs_state *nfs = NULL; + dict_t *options = NULL; + char *portstr = NULL; + int ret = -1; + + if (!nfsx || !nfsx->private) + return NULL; + + nfs = (struct nfs_state *)nfsx->private; + + gf_log (GF_MNT, GF_LOG_DEBUG, "Initing Mount v1 state"); + mstate = (struct mount3_state *)nfs->mstate; + if (!mstate) { + gf_log (GF_MNT, GF_LOG_ERROR, "Mount v3 state init failed"); + goto err; + } + + mnt1prog.private = mstate; + + options = dict_new (); + + ret = gf_asprintf (&portstr, "%d", GF_MOUNTV1_PORT); + if (ret == -1) + goto err; + + ret = dict_set_dynstr (options, "transport.socket.listen-port", portstr); + if (ret == -1) + goto err; + ret = dict_set_str (options, "transport-type", "socket"); + if (ret == -1) { + gf_log (GF_NFS, GF_LOG_ERROR, "dict_set_str error"); + goto err; + } + + if (nfs->allow_insecure) { + ret = dict_set_str (options, "rpc-auth-allow-insecure", "on"); + if (ret == -1) { + gf_log (GF_NFS, GF_LOG_ERROR, "dict_set_str error"); + goto err; + } + ret = dict_set_str (options, "rpc-auth.ports.insecure", "on"); + if (ret == -1) { + gf_log (GF_NFS, GF_LOG_ERROR, "dict_set_str error"); + goto err; + } + } + + ret = rpcsvc_create_listeners (nfs->rpcsvc, options, nfsx->name); + if (ret == -1) { + gf_log (GF_NFS, GF_LOG_ERROR, "Unable to create listeners"); + dict_unref (options); + goto err; + } + + return &mnt1prog; +err: + return NULL; +} + +int +mount_reconfigure_state (xlator_t *nfsx, dict_t *options) +{ + int ret = -1; + struct nfs_state *nfs = NULL; + struct mount3_state *ms = NULL; + struct mnt3_export *exp = NULL; + struct mnt3_export *texp = NULL; + + if ((!nfsx) || (!options)) + return (-1); + + nfs = (struct nfs_state *)nfs_state (nfsx); + if (!nfs) + return (-1); + + ms = nfs->mstate; + if (!ms) + return (-1); + + /* + * Free() up the old export list. mnt3_init_options() will + * rebuild the export list from scratch. Do it with locking + * to avoid unnecessary race conditions. + */ + LOCK (&ms->mountlock); + list_for_each_entry_safe (exp, texp, &ms->exportlist, explist) { + list_del (&exp->explist); + mnt3_export_free (exp); + } + ret = mnt3_init_options (ms, options); + UNLOCK (&ms->mountlock); + + if (ret < 0) { + gf_log (GF_MNT, GF_LOG_ERROR, "Options reconfigure failed"); + return (-1); + } + + return (0); +} diff --git a/xlators/nfs/server/src/mount3.h b/xlators/nfs/server/src/mount3.h new file mode 100644 index 000000000..7fc16ed57 --- /dev/null +++ b/xlators/nfs/server/src/mount3.h @@ -0,0 +1,131 @@ +/* + Copyright (c) 2010-2011 Gluster, Inc. <http://www.gluster.com> + This file is part of GlusterFS. + + This file is licensed to you under your choice of the GNU Lesser + General Public License, version 3 or any later version (LGPLv3 or + later), or the GNU General Public License, version 2 (GPLv2), in all + cases as published by the Free Software Foundation. +*/ + +#ifndef _MOUNT3_H_ +#define _MOUNT3_H_ + +#ifndef _CONFIG_H +#define _CONFIG_H +#include "config.h" +#endif + +#include "rpcsvc.h" +#include "dict.h" +#include "xlator.h" +#include "iobuf.h" +#include "nfs.h" +#include "list.h" +#include "xdr-nfs3.h" +#include "locking.h" +#include "nfs3-fh.h" +#include "uuid.h" + +/* Registered with portmap */ +#define GF_MOUNTV3_PORT 38465 +#define GF_MOUNTV3_IOB (2 * GF_UNIT_KB) +#define GF_MOUNTV3_IOBPOOL (GF_MOUNTV3_IOB * 50) + +#define GF_MOUNTV1_PORT 38466 +#define GF_MNT GF_NFS"-mount" + +extern rpcsvc_program_t * +mnt3svc_init (xlator_t *nfsx); + +extern rpcsvc_program_t * +mnt1svc_init (xlator_t *nfsx); + +extern int +mount_init_state (xlator_t *nfsx); + +extern int +mount_reconfigure_state (xlator_t *nfsx, dict_t *options); + +void +mount_rewrite_rmtab (struct mount3_state *ms, char *new_rmtab); + +/* Data structure used to store the list of mounts points currently + * in use by NFS clients. + */ +struct mountentry { + /* Links to mount3_state->mountlist. */ + struct list_head mlist; + + /* The export name */ + char exname[MNTPATHLEN]; + char hostname[MNTPATHLEN]; +}; + +#define MNT3_EXPTYPE_VOLUME 1 +#define MNT3_EXPTYPE_DIR 2 + +/* Structure to hold export-dir AUTH parameter */ +struct host_auth_spec { + char *host_addr; /* Allowed IP or host name */ + int routeprefix; /* Routing prefix */ + struct host_auth_spec *next; /* Pointer to next AUTH struct */ +}; + +struct mnt3_export { + struct list_head explist; + + /* The string that may contain either the volume name if the full volume + * is exported or the subdirectory in the volume. + */ + char *expname; + /* + * IP address, hostname or subnets who are allowed to connect to expname + * subvolume or subdirectory + */ + struct host_auth_spec* hostspec; + xlator_t *vol; + int exptype; + + /* Extracted from nfs volume options if nfs.dynamicvolumes is on. + */ + uuid_t volumeid; +}; + +struct mount3_state { + xlator_t *nfsx; + + /* The buffers for all network IO are got from this pool. */ + struct iobuf_pool *iobpool; + + /* List of exports, can be volumes or directories in those volumes. */ + struct list_head exportlist; + + /* List of current mount points over all the exports from this + * server. + */ + struct list_head mountlist; + + /* Used to protect the mountlist. */ + gf_lock_t mountlock; + + /* Set to 0 if exporting full volumes is disabled. On by default. */ + gf_boolean_t export_volumes; + gf_boolean_t export_dirs; +}; + +#define gf_mnt3_export_dirs(mst) ((mst)->export_dirs) + +struct mount3_resolve_state { + struct mnt3_export *exp; + struct mount3_state *mstate; + rpcsvc_request_t *req; + + char remainingdir[MNTPATHLEN]; + loc_t resolveloc; + struct nfs3_fh parentfh; +}; + +typedef struct mount3_resolve_state mnt3_resolve_t; + +#endif diff --git a/xlators/nfs/server/src/mount3udp_svc.c b/xlators/nfs/server/src/mount3udp_svc.c new file mode 100644 index 000000000..fb59e282c --- /dev/null +++ b/xlators/nfs/server/src/mount3udp_svc.c @@ -0,0 +1,189 @@ +/* + Copyright (c) 2012 Gluster, Inc. <http://www.gluster.com> + This file is part of GlusterFS. + + This file is licensed to you under your choice of the GNU Lesser + General Public License, version 3 or any later version (LGPLv3 or + later), or the GNU General Public License, version 2 (GPLv2), in all + cases as published by the Free Software Foundation. +*/ + + +#include "xdr-nfs3.h" +#include "logging.h" +#include "mem-pool.h" +#include "nfs-mem-types.h" +#include "mount3.h" +#include <stdio.h> +#include <stdlib.h> +#include <rpc/pmap_clnt.h> +#include <string.h> +#include <memory.h> +#include <sys/socket.h> +#include <netinet/in.h> + + +extern struct nfs3_fh* nfs3_rootfh (char *dp); +extern mountres3 mnt3svc_set_mountres3 (mountstat3 stat, struct nfs3_fh *fh, + int *authflavor, u_int aflen); +extern int +mount3udp_add_mountlist (char *host, dirpath *expname); + +extern int +mount3udp_delete_mountlist (char *host, dirpath *expname); + + +/* only this thread will use this, no locking needed */ +char mnthost[INET_ADDRSTRLEN+1]; + +mountres3 * +mountudpproc3_mnt_3_svc(dirpath **dpp, struct svc_req *req) +{ + struct mountres3 *res = NULL; + int *autharr = NULL; + struct nfs3_fh *fh = NULL; + char *tmp = NULL; + + tmp = (char *)*dpp; + while (*tmp == '/') + tmp++; + fh = nfs3_rootfh (tmp); + if (fh == NULL) { + gf_log (GF_MNT, GF_LOG_DEBUG, "unable to get fh for %s", tmp); + goto err; + } + + res = GF_CALLOC (1, sizeof(*res), gf_nfs_mt_mountres3); + if (res == NULL) { + gf_log (GF_MNT, GF_LOG_ERROR, "unable to allocate memory"); + goto err; + } + autharr = GF_CALLOC (1, sizeof(*autharr), gf_nfs_mt_int); + if (autharr == NULL) { + gf_log (GF_MNT, GF_LOG_ERROR, "unable to allocate memory"); + goto err; + } + autharr[0] = AUTH_UNIX; + *res = mnt3svc_set_mountres3 (MNT3_OK, fh, autharr, 1); + mount3udp_add_mountlist (mnthost, *dpp); + return res; + + err: + GF_FREE (fh); + GF_FREE (res); + GF_FREE (autharr); + return NULL; +} + +mountstat3 * +mountudpproc3_umnt_3_svc(dirpath **dp, struct svc_req *req) +{ + mountstat3 *stat = NULL; + + stat = GF_CALLOC (1, sizeof(mountstat3), gf_nfs_mt_mountstat3); + if (stat == NULL) { + gf_log (GF_MNT, GF_LOG_ERROR, "unable to allocate memory"); + return NULL; + } + *stat = MNT3_OK; + mount3udp_delete_mountlist (mnthost, *dp); + return stat; +} + +static void +mountudp_program_3(struct svc_req *rqstp, register SVCXPRT *transp) +{ + union { + dirpath mountudpproc3_mnt_3_arg; + } argument; + char *result = NULL; + xdrproc_t _xdr_argument = NULL, _xdr_result = NULL; + char *(*local)(char *, struct svc_req *) = NULL; + mountres3 *res = NULL; + struct sockaddr_in *sin = NULL; + + sin = svc_getcaller (transp); + inet_ntop (AF_INET, &sin->sin_addr, mnthost, INET_ADDRSTRLEN+1); + + switch (rqstp->rq_proc) { + case NULLPROC: + (void) svc_sendreply (transp, (xdrproc_t) xdr_void, + (char *)NULL); + return; + + case MOUNT3_MNT: + _xdr_argument = (xdrproc_t) xdr_dirpath; + _xdr_result = (xdrproc_t) xdr_mountres3; + local = (char *(*)(char *, + struct svc_req *)) mountudpproc3_mnt_3_svc; + break; + + case MOUNT3_UMNT: + _xdr_argument = (xdrproc_t) xdr_dirpath; + _xdr_result = (xdrproc_t) xdr_mountstat3; + local = (char *(*)(char *, + struct svc_req *)) mountudpproc3_umnt_3_svc; + break; + + default: + svcerr_noproc (transp); + return; + } + memset ((char *)&argument, 0, sizeof (argument)); + if (!svc_getargs (transp, (xdrproc_t) _xdr_argument, + (caddr_t) &argument)) { + svcerr_decode (transp); + return; + } + result = (*local)((char *)&argument, rqstp); + if (result == NULL) { + gf_log (GF_MNT, GF_LOG_DEBUG, "PROC returned error"); + svcerr_systemerr (transp); + } + if (result != NULL && !svc_sendreply(transp, (xdrproc_t) _xdr_result, + result)) { + gf_log (GF_MNT, GF_LOG_ERROR, "svc_sendreply returned error"); + svcerr_systemerr (transp); + } + if (!svc_freeargs (transp, (xdrproc_t) _xdr_argument, + (caddr_t) &argument)) { + gf_log (GF_MNT, GF_LOG_ERROR, "unable to free arguments"); + } + if (result == NULL) + return; + /* free the result */ + switch (rqstp->rq_proc) { + case MOUNT3_MNT: + res = (mountres3 *) result; + GF_FREE (res->mountres3_u.mountinfo.fhandle.fhandle3_val); + GF_FREE (res->mountres3_u.mountinfo.auth_flavors.auth_flavors_val); + GF_FREE (res); + break; + + case MOUNT3_UMNT: + GF_FREE (result); + break; + } + return; +} + +void * +mount3udp_thread (void *argv) +{ + register SVCXPRT *transp = NULL; + + transp = svcudp_create(RPC_ANYSOCK); + if (transp == NULL) { + gf_log (GF_MNT, GF_LOG_ERROR, "svcudp_create error"); + return NULL; + } + if (!svc_register(transp, MOUNT_PROGRAM, MOUNT_V3, + mountudp_program_3, IPPROTO_UDP)) { + gf_log (GF_MNT, GF_LOG_ERROR, "svc_register error"); + return NULL; + } + + svc_run (); + gf_log (GF_MNT, GF_LOG_ERROR, "svc_run returned"); + return NULL; +} diff --git a/xlators/nfs/server/src/nfs-common.c b/xlators/nfs/server/src/nfs-common.c new file mode 100644 index 000000000..f74396ee8 --- /dev/null +++ b/xlators/nfs/server/src/nfs-common.c @@ -0,0 +1,465 @@ +/* + Copyright (c) 2010-2011 Gluster, Inc. <http://www.gluster.com> + This file is part of GlusterFS. + + This file is licensed to you under your choice of the GNU Lesser + General Public License, version 3 or any later version (LGPLv3 or + later), or the GNU General Public License, version 2 (GPLv2), in all + cases as published by the Free Software Foundation. +*/ + +#ifndef _CONFIG_H +#define _CONFIG_H +#include "config.h" +#endif + +#include "rpcsvc.h" +#include "dict.h" +#include "xlator.h" +#include "xdr-nfs3.h" +#include "msg-nfs3.h" +#include "iobuf.h" +#include "nfs-common.h" +#include "nfs-fops.h" +#include "nfs-mem-types.h" +#include "rpcsvc.h" +#include "iatt.h" + +#include <libgen.h> + +xlator_t * +nfs_xlid_to_xlator (xlator_list_t *cl, uint8_t xlid) +{ + xlator_t *xl = NULL; + uint8_t id = 0; + + while (id <= xlid) { + if (!cl) { + xl = NULL; + break; + } + + xl = cl->xlator; + cl = cl->next; + id++; + } + + return xl; +} + + +xlator_t * +nfs_path_to_xlator (xlator_list_t *cl, char *path) +{ + return NULL; +} + + +uint16_t +nfs_xlator_to_xlid (xlator_list_t *cl, xlator_t *xl) +{ + uint16_t xlid = 0; + + if ((!cl) || (!xl)) + return 0; + + while (cl) { + if (xl == cl->xlator) + break; + cl = cl->next; + ++xlid; + } + + return xlid; +} + + +xlator_t * +nfs_mntpath_to_xlator (xlator_list_t *cl, char *path) +{ + char volname[MNTPATHLEN]; + char *volptr = NULL; + size_t pathlen; + xlator_t *targetxl = NULL; + + if ((!cl) || (!path)) + return NULL; + + strncpy (volname, path, MNTPATHLEN); + pathlen = strlen (volname); + gf_log (GF_NFS, GF_LOG_TRACE, "Subvolume search: %s", path); + if (volname[0] == '/') + volptr = &volname[1]; + else + volptr = &volname[0]; + + if (pathlen && volname[pathlen - 1] == '/') + volname[pathlen - 1] = '\0'; + + while (cl) { + if (strcmp (volptr, cl->xlator->name) == 0) { + targetxl = cl->xlator; + break; + } + + cl = cl->next; + } + + return targetxl; + +} + + +/* Returns 1 if the stat seems to be filled with zeroes. */ +int +nfs_zero_filled_stat (struct iatt *buf) +{ + if (!buf) + return 1; + + /* Do not use st_dev because it is transformed to store the xlator id + * in place of the device number. Do not use st_ino because by this time + * we've already mapped the root ino to 1 so it is not guaranteed to be + * 0. + */ + if ((buf->ia_nlink == 0) && (buf->ia_ctime == 0)) + return 1; + + return 0; +} + + +void +nfs_loc_wipe (loc_t *loc) +{ + loc_wipe (loc); +} + + +int +nfs_loc_copy (loc_t *dst, loc_t *src) +{ + int ret = -1; + + ret = loc_copy (dst, src); + + return ret; +} + + +int +nfs_loc_fill (loc_t *loc, inode_t *inode, inode_t *parent, char *path) +{ + int ret = -EFAULT; + + if (!loc) + return ret; + + if (inode) { + loc->inode = inode_ref (inode); + if (!uuid_is_null (inode->gfid)) + uuid_copy (loc->gfid, inode->gfid); + } + + if (parent) + loc->parent = inode_ref (parent); + + if (path) { + loc->path = gf_strdup (path); + if (!loc->path) { + gf_log (GF_NFS, GF_LOG_ERROR, "strdup failed"); + goto loc_wipe; + } + loc->name = strrchr (loc->path, '/'); + if (loc->name) + loc->name++; + } + + ret = 0; +loc_wipe: + if (ret < 0) + nfs_loc_wipe (loc); + + return ret; +} + + +int +nfs_inode_loc_fill (inode_t *inode, loc_t *loc, int how) +{ + char *resolvedpath = NULL; + inode_t *parent = NULL; + int ret = -EFAULT; + + if ((!inode) || (!loc)) + return ret; + + /* If gfid is not null, then the inode is already linked to + * the inode table, and not a newly created one. For newly + * created inode, inode_path returns null gfid as the path. + */ + if (!uuid_is_null (inode->gfid)) { + ret = inode_path (inode, NULL, &resolvedpath); + if (ret < 0) { + gf_log (GF_NFS, GF_LOG_ERROR, "path resolution failed " + "%s", resolvedpath); + goto err; + } + } + + if (resolvedpath == NULL) { + char tmp_path[GFID_STR_PFX_LEN + 1] = {0,}; + snprintf (tmp_path, sizeof (tmp_path), "<gfid:%s>", + uuid_utoa (loc->gfid)); + resolvedpath = gf_strdup (tmp_path); + } else { + parent = inode_parent (inode, loc->pargfid, NULL); + } + + ret = nfs_loc_fill (loc, inode, parent, resolvedpath); + if (ret < 0) { + gf_log (GF_NFS, GF_LOG_ERROR, "loc fill resolution failed %s", + resolvedpath); + goto err; + } + + ret = 0; +err: + if (parent) + inode_unref (parent); + + GF_FREE (resolvedpath); + + return ret; +} + +int +nfs_gfid_loc_fill (inode_table_t *itable, uuid_t gfid, loc_t *loc, int how) +{ + int ret = -EFAULT; + inode_t *inode = NULL; + + if (!loc) + return ret; + + inode = inode_find (itable, gfid); + if (!inode) { + gf_log (GF_NFS, GF_LOG_TRACE, "Inode not found in itable, will try to create one."); + if (how == NFS_RESOLVE_CREATE) { + gf_log (GF_NFS, GF_LOG_TRACE, "Inode needs to be created."); + inode = inode_new (itable); + if (!inode) { + gf_log (GF_NFS, GF_LOG_ERROR, "Failed to " + "allocate memory"); + ret = -ENOMEM; + goto err; + } + + } else { + gf_log (GF_NFS, GF_LOG_ERROR, "Inode not found in itable and no creation was requested."); + ret = -ENOENT; + goto err; + } + } else { + gf_log (GF_NFS, GF_LOG_TRACE, "Inode was found in the itable."); + } + + uuid_copy (loc->gfid, gfid); + + ret = nfs_inode_loc_fill (inode, loc, how); + if (ret < 0) { + gf_log (GF_NFS, GF_LOG_ERROR, "Inode loc filling failed.: %s", strerror (-ret)); + goto err; + } + +err: + if (inode) + inode_unref (inode); + return ret; +} + + +int +nfs_root_loc_fill (inode_table_t *itable, loc_t *loc) +{ + uuid_t rootgfid = {0, }; + + rootgfid[15] = 1; + return nfs_gfid_loc_fill (itable, rootgfid, loc, NFS_RESOLVE_EXIST); +} + + + +int +nfs_parent_inode_loc_fill (inode_t *parent, inode_t *entryinode, char *entry, + loc_t *loc) +{ + int ret = -EFAULT; + char *path = NULL; + + if ((!parent) || (!entry) || (!loc) || (!entryinode)) + return ret; + + ret = inode_path (parent, entry, &path); + if (ret < 0) { + gf_log (GF_NFS, GF_LOG_ERROR, "path resolution failed %s", + path); + goto err; + } + + ret = nfs_loc_fill (loc, entryinode, parent, path); + GF_FREE (path); +err: + return ret; +} + + +/* Returns -1 if parent is not available, return -2 if the entry is not + * available. In case the return is going to be -2, and how = NFS_RESOLVE_CREATE + * it does however fill in the loc so that it can be used to perform a lookup + * fop for the entry. + * On other errors, return -3. 0 on success. + */ +int +nfs_entry_loc_fill (inode_table_t *itable, uuid_t pargfid, char *entry, + loc_t *loc, int how) +{ + inode_t *parent = NULL; + inode_t *entryinode = NULL; + int ret = -3; + char *resolvedpath = NULL; + int pret = -3; + + if ((!itable) || (!entry) || (!loc)) + return ret; + + parent = inode_find (itable, pargfid); + + ret = -1; + /* Will need hard resolution now */ + if (!parent) + goto err; + + uuid_copy (loc->pargfid, pargfid); + + ret = -2; + entryinode = inode_grep (itable, parent, entry); + if (!entryinode) { + if (how == NFS_RESOLVE_CREATE) { + /* Even though we'll create the inode and the loc for + * a missing inode, we still need to return -2 so + * that the caller can use the filled loc to call + * lookup. + */ + entryinode = inode_new (itable); + /* Cannot change ret because that must + * continue to have -2. + */ + pret = nfs_parent_inode_loc_fill (parent, entryinode, + entry, loc); + /* Only if parent loc fill fails, should we notify error + * through ret, otherwise, we still need to force a + * lookup by returning -2. + */ + if (pret < 0) + ret = -3; + } + goto err; + } + + ret = inode_path (parent, entry, &resolvedpath); + if (ret < 0) { + gf_log (GF_NFS, GF_LOG_ERROR, "path resolution failed %s", + resolvedpath); + ret = -3; + goto err; + } + + ret = nfs_loc_fill (loc, entryinode, parent, resolvedpath); + if (ret < 0) { + gf_log (GF_NFS, GF_LOG_ERROR, "loc_fill failed %s", + resolvedpath); + ret = -3; + } + +err: + if (parent) + inode_unref (parent); + + if (entryinode) + inode_unref (entryinode); + + GF_FREE (resolvedpath); + + return ret; +} + + +uint32_t +nfs_hash_gfid (uuid_t gfid) +{ + uint32_t hash = 0; + uint64_t msb64 = 0; + uint64_t lsb64 = 0; + uint32_t a1 = 0; + uint32_t a2 = 0; + uint32_t a3 = 0; + uint32_t a4 = 0; + uint32_t b1 = 0; + uint32_t b2 = 0; + + if (__is_root_gfid (gfid)) + return 0x1; + + memcpy (&msb64, &gfid[8], 8); + memcpy (&lsb64, &gfid[0], 8); + + a1 = (msb64 << 32); + a2 = (msb64 >> 32); + a3 = (lsb64 << 32); + a4 = (lsb64 >> 32); + + b1 = a1 ^ a4; + b2 = a2 ^ a3; + + hash = b1 ^ b2; + + return hash; +} + + +void +nfs_fix_generation (xlator_t *this, inode_t *inode) +{ + uint64_t raw_ctx = 0; + struct nfs_inode_ctx *ictx = NULL; + struct nfs_state *priv = NULL; + int ret = -1; + + if (!inode) { + return; + } + priv = this->private; + + if (inode_ctx_get(inode,this,&raw_ctx) == 0) { + ictx = (struct nfs_inode_ctx *)raw_ctx; + ictx->generation = priv->generation; + } + else { + ictx = GF_CALLOC (1, sizeof (struct nfs_inode_ctx), + gf_nfs_mt_inode_ctx); + if (!ictx) { + gf_log (this->name, GF_LOG_ERROR, + "could not allocate nfs inode ctx"); + return; + } + INIT_LIST_HEAD(&ictx->shares); + ictx->generation = priv->generation; + ret = inode_ctx_put (inode, this, (uint64_t)ictx); + if (ret) { + gf_log (this->name, GF_LOG_ERROR, + "could not store nfs inode ctx"); + return; + } + } +} diff --git a/xlators/nfs/server/src/nfs-common.h b/xlators/nfs/server/src/nfs-common.h new file mode 100644 index 000000000..2e97f1563 --- /dev/null +++ b/xlators/nfs/server/src/nfs-common.h @@ -0,0 +1,81 @@ +/* + Copyright (c) 2010-2011 Gluster, Inc. <http://www.gluster.com> + This file is part of GlusterFS. + + This file is licensed to you under your choice of the GNU Lesser + General Public License, version 3 or any later version (LGPLv3 or + later), or the GNU General Public License, version 2 (GPLv2), in all + cases as published by the Free Software Foundation. +*/ + +#ifndef _NFS_COMMON_H_ +#define _NFS_COMMON_H_ + +#ifndef _CONFIG_H +#define _CONFIG_H +#include "config.h" +#endif + +#include <unistd.h> + +#include "xlator.h" +#include "rpcsvc.h" +#include "iatt.h" +#include "uuid.h" + +//NFS_PATH_MAX hard-coded to 4096 as a work around for bug 2476. +//nfs server crashes when path received is longer than PATH_MAX +#define NFS_PATH_MAX 4096 +#define NFS_NAME_MAX NAME_MAX + +#define NFS_DEFAULT_CREATE_MODE 0600 + +extern xlator_t * +nfs_xlid_to_xlator (xlator_list_t *cl, uint8_t xlid); + +extern uint16_t +nfs_xlator_to_xlid (xlator_list_t *cl, xlator_t *xl); + +extern xlator_t * +nfs_path_to_xlator (xlator_list_t *cl, char *path); + +extern xlator_t * +nfs_mntpath_to_xlator (xlator_list_t *cl, char *path); + +extern int +nfs_zero_filled_stat (struct iatt *buf); + +extern void +nfs_loc_wipe (loc_t *loc); + +extern int +nfs_loc_copy (loc_t *dst, loc_t *src); + +extern int +nfs_loc_fill (loc_t *loc, inode_t *inode, inode_t *parent, char *path); + +#define NFS_RESOLVE_EXIST 1 +#define NFS_RESOLVE_CREATE 2 + +extern int +nfs_inode_loc_fill (inode_t *inode, loc_t *loc, int how); + +extern int +nfs_ino_loc_fill (inode_table_t *itable, uuid_t gfid, loc_t *l); + +extern int +nfs_entry_loc_fill (inode_table_t *itable, uuid_t pargfid, char *entry, + loc_t *loc, int how); + +extern int +nfs_root_loc_fill (inode_table_t *itable, loc_t *loc); + +extern uint32_t +nfs_hash_gfid (uuid_t gfid); + +extern int +nfs_gfid_loc_fill (inode_table_t *itable, uuid_t gfid, loc_t *loc, int how); + +void +nfs_fix_generation (xlator_t *this, inode_t *inode); +#endif diff --git a/xlators/nfs/server/src/nfs-fops.c b/xlators/nfs/server/src/nfs-fops.c new file mode 100644 index 000000000..236b80c76 --- /dev/null +++ b/xlators/nfs/server/src/nfs-fops.c @@ -0,0 +1,1661 @@ +/* + Copyright (c) 2010-2011 Gluster, Inc. <http://www.gluster.com> + This file is part of GlusterFS. + + This file is licensed to you under your choice of the GNU Lesser + General Public License, version 3 or any later version (LGPLv3 or + later), or the GNU General Public License, version 2 (GPLv2), in all + cases as published by the Free Software Foundation. +*/ + +#ifndef _CONFIG_H +#define _CONFIG_H +#include "config.h" +#endif + +#include <grp.h> +#include <pwd.h> + +#include "dict.h" +#include "xlator.h" +#include "iobuf.h" +#include "call-stub.h" +#include "stack.h" +#include "nfs.h" +#include "nfs-fops.h" +#include "inode.h" +#include "nfs-common.h" +#include "nfs3-helpers.h" +#include "nfs-mem-types.h" +#include <libgen.h> +#include <semaphore.h> + +void +nfs_fix_groups (xlator_t *this, call_stack_t *root) +{ + struct passwd mypw; + char mystrs[1024]; + struct passwd *result; + gid_t mygroups[GF_MAX_AUX_GROUPS]; + int ngroups; + int i; + struct nfs_state *priv = this->private; + const gid_list_t *agl; + gid_list_t gl; + + if (!priv->server_aux_gids) { + return; + } + + agl = gid_cache_lookup(&priv->gid_cache, root->uid); + if (agl) { + for (ngroups = 0; ngroups < agl->gl_count; ngroups++) + root->groups[ngroups] = agl->gl_list[ngroups]; + root->ngrps = ngroups; + gid_cache_release(&priv->gid_cache, agl); + return; + } + + /* No cached list found. */ + if (getpwuid_r(root->uid,&mypw,mystrs,sizeof(mystrs),&result) != 0) { + gf_log (this->name, GF_LOG_ERROR, + "getpwuid_r(%u) failed", root->uid); + return; + } + + if (!result) { + gf_log (this->name, GF_LOG_ERROR, + "getpwuid_r(%u) found nothing", root->uid); + return; + } + + gf_log (this->name, GF_LOG_TRACE, "mapped %u => %s", + root->uid, result->pw_name); + + ngroups = GF_MAX_AUX_GROUPS; + if (getgrouplist(result->pw_name,root->gid,mygroups,&ngroups) == -1) { + gf_log (this->name, GF_LOG_ERROR, + "could not map %s to group list", result->pw_name); + return; + } + + /* Add the group data to the cache. */ + gl.gl_list = GF_CALLOC(ngroups, sizeof(gid_t), gf_nfs_mt_aux_gids); + if (gl.gl_list) { + /* It's not fatal if the alloc failed. */ + gl.gl_id = root->uid; + gl.gl_count = ngroups; + memcpy(gl.gl_list, mygroups, sizeof(gid_t) * ngroups); + if (gid_cache_add(&priv->gid_cache, &gl) != 1) + GF_FREE(gl.gl_list); + } + + /* Copy data to the frame. */ + for (i = 0; i < ngroups; ++i) { + gf_log (this->name, GF_LOG_TRACE, + "%s is in group %u", result->pw_name, mygroups[i]); + root->groups[i] = mygroups[i]; + } + root->ngrps = ngroups; +} + +struct nfs_fop_local * +nfs_fop_local_init (xlator_t *nfsx) +{ + struct nfs_fop_local *l = NULL; + + if (!nfsx) + return NULL; + + l = mem_get (nfs_fop_mempool (nfsx)); + + memset (l, 0, sizeof (*l)); + return l; +} + +void +nfs_fop_local_wipe (xlator_t *nfsx, struct nfs_fop_local *l) +{ + if ((!nfsx) || (!l)) + return; + + if (l->iobref) + iobref_unref (l->iobref); + + if (l->parent) + inode_unref (l->parent); + + if (l->inode) + inode_unref (l->inode); + + if (l->newparent) + inode_unref (l->newparent); + + if (l->dictgfid) + dict_unref (l->dictgfid); + + mem_put (l); + + return; +} + +#define nfs_stack_destroy(nfl, fram) \ + do { \ + nfs_fop_local_wipe ((nfl)->nfsx, nfl); \ + (fram)->local = NULL; \ + STACK_DESTROY ((fram)->root); \ + } while (0) \ + + +pthread_mutex_t ctr = PTHREAD_MUTEX_INITIALIZER; +unsigned int cval = 1; + + +int +nfs_frame_getctr () +{ + uint64_t val = 0; + + pthread_mutex_lock (&ctr); + { + if (cval == 0) + cval = 1; + val = cval; + cval++; + } + pthread_mutex_unlock (&ctr); + + return val; +} + + +call_frame_t * +nfs_create_frame (xlator_t *xl, nfs_user_t *nfu) +{ + call_frame_t *frame = NULL; + int x = 0; + int y = 0; + + if ((!xl) || (!nfu) || (nfu->ngrps > NFS_NGROUPS)) + return NULL; + + frame = create_frame (xl, (call_pool_t *)xl->ctx->pool); + if (!frame) + goto err; + if (call_stack_alloc_groups (frame->root, nfu->ngrps) != 0) { + STACK_DESTROY (frame->root); + frame = NULL; + goto err; + } + + frame->root->pid = NFS_PID; + frame->root->uid = nfu->uid; + frame->root->gid = nfu->gids[NFS_PRIMGID_IDX]; + frame->root->lk_owner = nfu->lk_owner; + + if (nfu->ngrps != 1) { + frame->root->ngrps = nfu->ngrps - 1; + + gf_log (GF_NFS, GF_LOG_TRACE,"uid: %d, gid %d, gids: %d", + frame->root->uid, frame->root->gid, frame->root->ngrps); + for(y = 0, x = 1; y < frame->root->ngrps; x++,y++) { + gf_log (GF_NFS, GF_LOG_TRACE, "gid: %d", nfu->gids[x]); + frame->root->groups[y] = nfu->gids[x]; + } + } + + /* + * It's tempting to do this *instead* of using nfu above, but we need + * to have those values in case nfs_fix_groups doesn't do anything. + */ + nfs_fix_groups(xl,frame->root); + +err: + return frame; +} + +#define nfs_fop_handle_frame_create(fram, xla, nfuser, retval, errlabel) \ + do { \ + fram = nfs_create_frame (xla, (nfuser)); \ + if (!fram) { \ + retval = (-ENOMEM); \ + gf_log (GF_NFS, GF_LOG_ERROR,"Frame creation failed");\ + goto errlabel; \ + } \ + } while (0) \ + +/* Look into the inode and parent inode of a loc and save enough state + * for us to determine in the callback whether to funge the ino in the stat buf + * with 1 for the parent. + */ +#define nfs_fop_save_root_ino(locl, loc) \ + do { \ + if (((loc)->inode) && \ + __is_root_gfid ((loc)->inode->gfid)) \ + (locl)->rootinode = 1; \ + else if (((loc)->parent) && \ + __is_root_gfid ((loc)->parent->gfid)) \ + (locl)->rootparentinode = 1; \ + } while (0) + +/* Do the same for an fd */ +#define nfs_fop_save_root_fd_ino(locl, fdesc) \ + do { \ + if (__is_root_gfid ((fdesc)->inode->gfid)) \ + (locl)->rootinode = 1; \ + } while (0) + + +/* Use the state saved by the previous macro to funge the ino in the appropriate + * structure. + */ +#define nfs_fop_restore_root_ino(locl, fopret, preattr, postattr, prepar, postpar) \ + do { \ + if (fopret == -1) \ + break; \ + if ((locl)->rootinode) { \ + if ((preattr)) { \ + ((struct iatt *)(preattr))->ia_ino = 1; \ + ((struct iatt *)(preattr))->ia_dev = 0; \ + } \ + if ((postattr)) { \ + ((struct iatt *)(postattr))->ia_ino = 1; \ + ((struct iatt *)(postattr))->ia_dev = 0; \ + } \ + } else if ((locl)->rootparentinode) { \ + if ((prepar)) { \ + ((struct iatt *)(prepar))->ia_ino = 1; \ + ((struct iatt *)(prepar))->ia_dev = 0; \ + } \ + if ((postpar)) { \ + ((struct iatt *)(postpar))->ia_ino = 1; \ + ((struct iatt *)(postpar))->ia_dev = 0; \ + } \ + } \ + } while (0) \ + +/* If the newly created, inode's parent is root, we'll need to funge the ino + * in the parent attr when we receive them in the callback. + */ +#define nfs_fop_newloc_save_root_ino(locl, newloc) \ + do { \ + if (((newloc)->inode) && \ + __is_root_gfid ((newloc)->inode->gfid)) \ + (locl)->newrootinode = 1; \ + else if (((newloc)->parent) && \ + __is_root_gfid ((newloc)->parent->gfid)) \ + (locl)->newrootparentinode = 1; \ + } while (0) + +#define nfs_fop_newloc_restore_root_ino(locl, fopret, preattr, postattr, prepar, postpar) \ + do { \ + if (fopret == -1) \ + break; \ + \ + if ((locl)->newrootinode) { \ + if ((preattr)) \ + ((struct iatt *)(preattr))->ia_ino = 1; \ + if ((postattr)) \ + ((struct iatt *)(postattr))->ia_ino = 1; \ + } else if ((locl)->newrootparentinode) { \ + if ((prepar)) \ + ((struct iatt *)(prepar))->ia_ino = 1; \ + if ((postpar)) \ + ((struct iatt *)(postpar))->ia_ino = 1; \ + } \ + } while (0) \ + +dict_t * +nfs_gfid_dict (inode_t *inode) +{ + uuid_t newgfid = {0, }; + char *dyngfid = NULL; + dict_t *dictgfid = NULL; + int ret = -1; + uuid_t rootgfid = {0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1}; + + dyngfid = GF_CALLOC (1, sizeof (uuid_t), gf_common_mt_char); + uuid_generate (newgfid); + + if (uuid_compare (inode->gfid, rootgfid) == 0) + memcpy (dyngfid, rootgfid, sizeof (uuid_t)); + else + memcpy (dyngfid, newgfid, sizeof (uuid_t)); + + dictgfid = dict_new (); + if (!dictgfid) { + gf_log (GF_NFS, GF_LOG_ERROR, "Failed to create gfid dict"); + goto out; + } + + ret = dict_set_bin (dictgfid, "gfid-req", dyngfid, sizeof (uuid_t)); + if (ret < 0) { + dict_unref (dictgfid); + dictgfid = NULL; + } + +out: + return dictgfid; +} + +#define nfs_fop_gfid_setup(nflcl, inode, retval, erlbl) \ + do { \ + if (nflcl) { \ + (nflcl)->dictgfid = nfs_gfid_dict (inode); \ + \ + if (!((nflcl)->dictgfid)) { \ + retval = -EFAULT; \ + goto erlbl; \ + } \ + } \ + } while (0) \ + +/* Fops Layer Explained + * The fops layer has three types of functions. They can all be identified by + * their names. Here are the three patterns: + * + * nfs_fop_<fopname> + * This is the lowest level function that knows nothing about states and + * callbacks. At most this is required to create a frame and call the + * fop. The idea here is to provide a convenient way to call fops than + * directly use STACK_WINDs. If this type of interface is used, the caller's + * callback is responsible for doing the relevant GlusterFS state + * maintenance operations on the data returned in the callbacks. + * + * nfs_<fopname> + * Unlike the nfs_fop_<fopname> variety, this is the stateful type of fop, in + * that it silently performs all the relevant GlusterFS state maintenance + * operations on the data returned to the callbacks, leaving the caller's + * callback to just use the data returned for whatever it needs to do with that + * data, for eg. the nfs_lookup, will take care of looking up the inodes, + * revalidating them if needed and linking new inodes into the table, while + * the caller's callback, for eg, the NFSv3 LOOKUP callback can just use + * the stat bufs returned to create file handle, map the file handle into the + * fh cache and finally encode the fh and the stat bufs into a NFS reply. + * + */ + +int32_t +nfs_fop_lookup_cbk (call_frame_t *frame, void *cookie, xlator_t *this, + int32_t op_ret, int32_t op_errno, inode_t *inode, + struct iatt *buf, dict_t *xattr, struct iatt *postparent) +{ + struct nfs_fop_local *local = NULL; + fop_lookup_cbk_t progcbk; + + if (op_ret == 0) { + nfs_fix_generation(this,inode); + } + + nfl_to_prog_data (local, progcbk, frame); + nfs_fop_restore_root_ino (local, op_ret, buf, NULL, NULL, postparent); + if (progcbk) + progcbk (frame, cookie, this, op_ret, op_errno, inode, buf, + xattr, postparent); + + nfs_stack_destroy (local, frame); + return 0; +} + + +int +nfs_fop_lookup (xlator_t *nfsx, xlator_t *xl, nfs_user_t *nfu, loc_t *loc, + fop_lookup_cbk_t cbk, void *local) +{ + call_frame_t *frame = NULL; + int ret = -EFAULT; + struct nfs_fop_local *nfl = NULL; + + if ((!xl) || (!loc) || (!nfu)) + return ret; + + gf_log (GF_NFS, GF_LOG_TRACE, "Lookup: %s", loc->path); + nfs_fop_handle_frame_create (frame, nfsx, nfu, ret, err); + nfs_fop_handle_local_init (frame, nfsx, nfl, cbk, local, ret, err); + nfs_fop_save_root_ino (nfl, loc); + nfs_fop_gfid_setup (nfl, loc->inode, ret, err); + + STACK_WIND_COOKIE (frame, nfs_fop_lookup_cbk, xl, xl, + xl->fops->lookup, loc, nfl->dictgfid); + + ret = 0; +err: + if (ret < 0) { + if (frame) + nfs_stack_destroy (nfl, frame); + } + + return ret; +} + +int32_t +nfs_fop_access_cbk (call_frame_t *frame, void *cookie, xlator_t *this, + int32_t op_ret, int32_t op_errno, dict_t *xdata) +{ + struct nfs_fop_local *nfl = NULL; + fop_access_cbk_t progcbk = NULL; + + nfl_to_prog_data (nfl, progcbk, frame); + if (progcbk) + progcbk (frame, cookie, this, op_ret, op_errno, xdata); + + nfs_stack_destroy (nfl, frame); + return 0; +} + +int +nfs_fop_access (xlator_t *nfsx, xlator_t *xl, nfs_user_t *nfu, loc_t *loc, + int32_t accesstest, fop_access_cbk_t cbk, void *local) +{ + call_frame_t *frame = NULL; + int ret = -EFAULT; + struct nfs_fop_local *nfl = NULL; + uint32_t accessbits = 0; + + if ((!xl) || (!loc) || (!nfu)) + return ret; + + gf_log (GF_NFS, GF_LOG_TRACE, "Access: %s", loc->path); + nfs_fop_handle_frame_create (frame, nfsx, nfu, ret, err); + nfs_fop_handle_local_init (frame, nfsx, nfl, cbk, local, ret, err); + nfs_fop_save_root_ino (nfl, loc); + + accessbits = nfs3_request_to_accessbits (accesstest); + STACK_WIND_COOKIE (frame, nfs_fop_access_cbk, xl, xl, xl->fops->access, + loc, accessbits, NULL); + ret = 0; +err: + if (ret < 0) { + if (frame) + nfs_stack_destroy (nfl, frame); + } + + return ret; +} + +int32_t +nfs_fop_stat_cbk (call_frame_t *frame, void *cookie, xlator_t *this, + int32_t op_ret, int32_t op_errno, struct iatt *buf, + dict_t *xdata) +{ + struct nfs_fop_local *nfl = NULL; + fop_stat_cbk_t progcbk = NULL; + + nfl_to_prog_data (nfl, progcbk, frame); + nfs_fop_restore_root_ino (nfl, op_ret, buf, NULL, NULL, NULL); + if (progcbk) + progcbk (frame, cookie, this, op_ret, op_errno, buf, xdata); + + nfs_stack_destroy (nfl, frame); + return 0; +} + + +int +nfs_fop_stat (xlator_t *nfsx, xlator_t *xl, nfs_user_t *nfu, loc_t *loc, + fop_stat_cbk_t cbk, void *local) +{ + call_frame_t *frame = NULL; + int ret = -EFAULT; + struct nfs_fop_local *nfl = NULL; + + if ((!xl) || (!loc) || (!nfu)) + return ret; + + gf_log (GF_NFS, GF_LOG_TRACE, "Stat: %s", loc->path); + nfs_fop_handle_frame_create (frame, nfsx, nfu, ret, err); + nfs_fop_handle_local_init (frame, nfsx, nfl, cbk, local, ret, err); + nfs_fop_save_root_ino (nfl, loc); + + STACK_WIND_COOKIE (frame, nfs_fop_stat_cbk, xl, xl, xl->fops->stat, + loc, NULL); + ret = 0; +err: + if (ret < 0) { + if (frame) + nfs_stack_destroy (nfl, frame); + } + + return ret; +} + + +int32_t +nfs_fop_fstat_cbk (call_frame_t *frame, void *cookie, xlator_t *this, + int32_t op_ret, int32_t op_errno, struct iatt *buf, + dict_t *xdata) +{ + struct nfs_fop_local *nfl = NULL; + fop_fstat_cbk_t progcbk = NULL; + + nfl_to_prog_data (nfl, progcbk, frame); + nfs_fop_restore_root_ino (nfl, op_ret, buf, NULL, NULL, NULL); + if (progcbk) + progcbk (frame, cookie, this, op_ret, op_errno, buf, xdata); + + nfs_stack_destroy (nfl, frame); + return 0; +} + + +int +nfs_fop_fstat (xlator_t *nfsx, xlator_t *xl, nfs_user_t *nfu, fd_t *fd, + fop_fstat_cbk_t cbk, void *local) +{ + call_frame_t *frame = NULL; + int ret = -EFAULT; + struct nfs_fop_local *nfl = NULL; + + if ((!nfsx) || (!xl) || (!fd) || (!nfu)) + return ret; + + gf_log (GF_NFS, GF_LOG_TRACE, "FStat"); + nfs_fop_handle_frame_create (frame, nfsx, nfu, ret, err); + nfs_fop_handle_local_init (frame, nfsx, nfl, cbk, local, ret, err); + nfs_fop_save_root_fd_ino (nfl, fd); + + STACK_WIND_COOKIE (frame, nfs_fop_fstat_cbk, xl, xl, xl->fops->fstat, + fd, NULL); + + ret = 0; +err: + if (ret < 0) { + if (frame) + nfs_stack_destroy (nfl, frame); + } + + return ret; +} + + +int32_t +nfs_fop_opendir_cbk (call_frame_t *frame, void *cookie, xlator_t *this, + int32_t op_ret, int32_t op_errno, fd_t *fd, dict_t *xdata) +{ + struct nfs_fop_local *nfl = NULL; + fop_opendir_cbk_t progcbk = NULL; + + nfl_to_prog_data (nfl, progcbk, frame); + if (progcbk) + progcbk (frame, cookie, this, op_ret, op_errno, fd, xdata); + nfs_stack_destroy (nfl, frame); + return 0; +} + + +int +nfs_fop_opendir (xlator_t *nfsx, xlator_t *xl, nfs_user_t *nfu, loc_t *pathloc, + fd_t *dirfd, fop_opendir_cbk_t cbk, void *local) +{ + call_frame_t *frame = NULL; + int ret = -EFAULT; + struct nfs_fop_local *nfl = NULL; + + if ((!nfsx) || (!xl) || (!pathloc) || (!dirfd) || (!nfu)) + return ret; + + gf_log (GF_NFS, GF_LOG_TRACE, "Opendir: %s", pathloc->path); + nfs_fop_handle_frame_create (frame, nfsx, nfu, ret, err); + nfs_fop_handle_local_init (frame, nfsx, nfl, cbk, local, ret, err); + + STACK_WIND_COOKIE (frame, nfs_fop_opendir_cbk, xl, xl, + xl->fops->opendir, pathloc, dirfd, NULL); + ret = 0; + +err: + if (ret < 0) { + if (frame) + nfs_stack_destroy (nfl, frame); + } + + return ret; +} + +int +nfs_fop_flush_cbk (call_frame_t *frame, void *cookie, xlator_t *this, + int32_t op_ret, int32_t op_errno, dict_t *xdata) +{ + struct nfs_fop_local *nfl = NULL; + fop_flush_cbk_t progcbk = NULL; + + nfl_to_prog_data (nfl, progcbk, frame); + if (progcbk) + progcbk (frame, cookie, this, op_ret, op_errno, xdata); + + nfs_stack_destroy (nfl, frame); + return 0; +} + + +int +nfs_fop_flush (xlator_t *nfsx, xlator_t *xl, nfs_user_t *nfu, fd_t *fd, + fop_flush_cbk_t cbk, void *local) +{ + call_frame_t *frame = NULL; + int ret = -EFAULT; + struct nfs_fop_local *nfl = NULL; + + if ((!nfsx) || (!xl) || (!fd) || (!nfu)) + return ret; + + nfs_fop_handle_frame_create (frame, nfsx, nfu, ret, err); + nfs_fop_handle_local_init (frame, nfsx, nfl, cbk, local, ret, err); + + STACK_WIND_COOKIE (frame, nfs_fop_flush_cbk, xl, xl, xl->fops->flush, + fd, NULL); + ret = 0; +err: + if (ret < 0) { + if (frame) + nfs_stack_destroy (nfl, frame); + } + + return ret; +} + + +int32_t +nfs_fop_readdirp_cbk (call_frame_t *frame, void *cookie, xlator_t *this, + int32_t op_ret, int32_t op_errno, gf_dirent_t *entries, + dict_t *xdata) +{ + struct nfs_fop_local *nfl = NULL; + fop_readdirp_cbk_t progcbk = NULL; + + nfl_to_prog_data (nfl, progcbk, frame); + if (progcbk) + progcbk (frame, cookie, this, op_ret, op_errno, entries, xdata); + + nfs_stack_destroy (nfl, frame); + + return 0; +} + + +int +nfs_fop_readdirp (xlator_t *nfsx, xlator_t *xl, nfs_user_t *nfu, fd_t *dirfd, + size_t bufsize, off_t offset, fop_readdirp_cbk_t cbk, + void *local) +{ + call_frame_t *frame = NULL; + int ret = -EFAULT; + struct nfs_fop_local *nfl = NULL; + + if ((!nfsx) || (!xl) || (!dirfd) || (!nfu)) + return ret; + + gf_log (GF_NFS, GF_LOG_TRACE, "readdir"); + nfs_fop_handle_frame_create (frame, nfsx, nfu, ret, err); + nfs_fop_handle_local_init (frame, nfsx, nfl, cbk, local, ret, err); + + STACK_WIND_COOKIE (frame, nfs_fop_readdirp_cbk, xl, xl, + xl->fops->readdirp, dirfd, bufsize, offset, 0); + + ret = 0; +err: + if (ret < 0) { + if (frame) + nfs_stack_destroy (nfl, frame); + } + + return ret; +} + + +int32_t +nfs_fop_statfs_cbk (call_frame_t *frame, void *cookie, xlator_t *this, + int32_t op_ret, int32_t op_errno, struct statvfs *buf, + dict_t *xdata) +{ + + struct nfs_fop_local *nfl = NULL; + fop_statfs_cbk_t progcbk = NULL; + + nfl_to_prog_data (nfl, progcbk, frame); + if (progcbk) + progcbk (frame, cookie, this, op_ret, op_errno, buf, xdata); + + nfs_stack_destroy (nfl, frame); + return 0; +} + + +int +nfs_fop_statfs (xlator_t *nfsx, xlator_t *xl, nfs_user_t *nfu, loc_t *pathloc, + fop_statfs_cbk_t cbk, void *local) +{ + call_frame_t *frame = NULL; + int ret = -EFAULT; + struct nfs_fop_local *nfl = NULL; + + if ((!nfsx) || (!xl) || (!pathloc) || (!nfu)) + return ret; + + gf_log (GF_NFS, GF_LOG_TRACE, "Statfs: %s", pathloc->path); + nfs_fop_handle_frame_create (frame, nfsx, nfu, ret, err); + nfs_fop_handle_local_init (frame, nfsx, nfl, cbk, local, ret, err); + + STACK_WIND_COOKIE (frame, nfs_fop_statfs_cbk, xl, xl, + xl->fops->statfs, pathloc, NULL); + ret = 0; +err: + if (ret < 0) { + if (frame) + nfs_stack_destroy (nfl, frame); + } + + return ret; +} + + +int32_t +nfs_fop_create_cbk (call_frame_t *frame, void *cookie, xlator_t *this, + int32_t op_ret, int32_t op_errno, fd_t *fd, inode_t *inode, + struct iatt *buf, struct iatt *preparent, + struct iatt *postparent, dict_t *xdata) +{ + struct nfs_fop_local *nfl = NULL; + fop_create_cbk_t progcbk = NULL; + + if (op_ret == 0) { + nfs_fix_generation(this,inode); + } + + nfl_to_prog_data (nfl, progcbk, frame); + nfs_fop_restore_root_ino (nfl, op_ret, buf, NULL, preparent, + postparent); + if (progcbk) + progcbk (frame, cookie, this, op_ret, op_errno, fd, inode, buf, + preparent, postparent, NULL); + + nfs_stack_destroy (nfl, frame); + return 0; +} + + +int +nfs_fop_create (xlator_t *nfsx, xlator_t *xl, nfs_user_t *nfu, loc_t *pathloc, + int flags, mode_t mode, fd_t *fd, fop_create_cbk_t cbk, + void *local) +{ + call_frame_t *frame = NULL; + int ret = -EFAULT; + struct nfs_fop_local *nfl = NULL; + + if ((!nfsx) || (!xl) || (!pathloc) || (!nfu)) + return ret; + + gf_log (GF_NFS, GF_LOG_TRACE, "Create: %s", pathloc->path); + nfs_fop_handle_frame_create (frame, nfsx, nfu, ret, err); + nfs_fop_handle_local_init (frame, nfsx, nfl, cbk, local, ret, err); + nfs_fop_save_root_ino (nfl, pathloc); + nfs_fop_gfid_setup (nfl, pathloc->inode, ret, err); + + STACK_WIND_COOKIE (frame, nfs_fop_create_cbk, xl, xl, xl->fops->create, + pathloc, flags, mode, 0, fd, nfl->dictgfid); + + ret = 0; +err: + if (ret < 0) { + if (frame) + nfs_stack_destroy (nfl, frame); + } + + return ret; +} + + +int32_t +nfs_fop_setattr_cbk (call_frame_t *frame, void *cookie, xlator_t *this, + int32_t op_ret, int32_t op_errno, struct iatt *pre, + struct iatt *post, dict_t *xdata) +{ + struct nfs_fop_local *nfl = NULL; + fop_setattr_cbk_t progcbk = NULL; + + nfl_to_prog_data (nfl, progcbk, frame); + nfs_fop_restore_root_ino (nfl, op_ret, pre, post, NULL, NULL); + if (progcbk) + progcbk (frame, cookie, this, op_ret, op_errno, pre, post, + xdata); + nfs_stack_destroy (nfl, frame); + return 0; +} + + + +int +nfs_fop_setattr (xlator_t *nfsx, xlator_t *xl, nfs_user_t *nfu, loc_t *pathloc, + struct iatt *buf, int32_t valid, fop_setattr_cbk_t cbk, + void *local) +{ + call_frame_t *frame = NULL; + int ret = -EFAULT; + struct nfs_fop_local *nfl = NULL; + + if ((!nfsx) || (!xl) || (!pathloc) || (!nfu)) + return ret; + + gf_log (GF_NFS, GF_LOG_TRACE, "Setattr: %s", pathloc->path); + nfs_fop_handle_frame_create (frame, nfsx, nfu, ret, err); + nfs_fop_handle_local_init (frame, nfsx, nfl, cbk, local, ret, err); + nfs_fop_save_root_ino (nfl, pathloc); + + STACK_WIND_COOKIE (frame, nfs_fop_setattr_cbk, xl, xl, + xl->fops->setattr, pathloc, buf, valid, NULL); + ret = 0; +err: + if (ret < 0) { + if (frame) + nfs_stack_destroy (nfl, frame); + } + + return ret; +} + + +int32_t +nfs_fop_mkdir_cbk (call_frame_t *frame, void *cookie, xlator_t *this, + int32_t op_ret, int32_t op_errno, inode_t *inode, + struct iatt *buf, struct iatt *preparent, + struct iatt *postparent, dict_t *xdata) +{ + struct nfs_fop_local *nfl = NULL; + fop_mkdir_cbk_t progcbk = NULL; + + if (op_ret == 0) { + nfs_fix_generation(this,inode); + } + + nfl_to_prog_data (nfl, progcbk, frame); + nfs_fop_restore_root_ino (nfl, op_ret, buf, NULL,preparent, postparent); + if (progcbk) + progcbk (frame, cookie, this, op_ret, op_errno, inode, buf, + preparent, postparent, xdata); + nfs_stack_destroy (nfl, frame); + return 0; +} + + +int +nfs_fop_mkdir (xlator_t *nfsx, xlator_t *xl, nfs_user_t *nfu, loc_t *pathloc, + mode_t mode, fop_mkdir_cbk_t cbk, void *local) +{ + call_frame_t *frame = NULL; + int ret = -EFAULT; + struct nfs_fop_local *nfl = NULL; + + if ((!nfsx) || (!xl) || (!pathloc) || (!nfu)) + return ret; + + gf_log (GF_NFS, GF_LOG_TRACE, "Mkdir: %s", pathloc->path); + nfs_fop_handle_frame_create (frame, nfsx, nfu, ret, err); + nfs_fop_handle_local_init (frame, nfsx, nfl, cbk, local, ret, err); + nfs_fop_save_root_ino (nfl, pathloc); + nfs_fop_gfid_setup (nfl, pathloc->inode, ret, err); + + STACK_WIND_COOKIE (frame, nfs_fop_mkdir_cbk, xl, xl, xl->fops->mkdir, + pathloc, mode, 0, nfl->dictgfid); + ret = 0; +err: + if (ret < 0) { + if (frame) + nfs_stack_destroy (nfl, frame); + } + + return ret; +} + + +int32_t +nfs_fop_symlink_cbk (call_frame_t *frame, void *cookie, xlator_t *this, + int32_t op_ret, int32_t op_errno, inode_t *inode, + struct iatt *buf, struct iatt *preparent, + struct iatt *postparent, dict_t *xdata) +{ + struct nfs_fop_local *nfl = NULL; + fop_symlink_cbk_t progcbk = NULL; + + if (op_ret == 0) { + nfs_fix_generation(this,inode); + } + + nfl_to_prog_data (nfl, progcbk, frame); + nfs_fop_restore_root_ino (nfl, op_ret,buf, NULL, preparent, postparent); + if (progcbk) + progcbk (frame, cookie, this, op_ret, op_errno, inode, buf, + preparent, postparent, xdata); + nfs_stack_destroy (nfl, frame); + return 0; +} + +int +nfs_fop_symlink (xlator_t *nfsx, xlator_t *xl, nfs_user_t *nfu, char *target, + loc_t *pathloc, fop_symlink_cbk_t cbk, void *local) +{ + call_frame_t *frame = NULL; + int ret = -EFAULT; + struct nfs_fop_local *nfl = NULL; + + if ((!nfsx) || (!xl) || (!pathloc) || (!target) || (!nfu)) + return ret; + + gf_log (GF_NFS, GF_LOG_TRACE, "Symlink: %s", pathloc->path); + nfs_fop_handle_frame_create (frame, nfsx, nfu, ret, err); + nfs_fop_handle_local_init (frame, nfsx, nfl, cbk, local, ret, err); + nfs_fop_save_root_ino (nfl, pathloc); + nfs_fop_gfid_setup (nfl, pathloc->inode, ret, err); + + STACK_WIND_COOKIE (frame, nfs_fop_symlink_cbk, xl, xl, + xl->fops->symlink, target, pathloc, + 0, nfl->dictgfid); + ret = 0; +err: + if (ret < 0) { + if (frame) + nfs_stack_destroy (nfl, frame); + } + + return ret; +} + + +int32_t +nfs_fop_readlink_cbk (call_frame_t *frame, void *cookie, xlator_t *this, + int32_t op_ret, int32_t op_errno, const char *path, + struct iatt *buf, dict_t *xdata) +{ + struct nfs_fop_local *nfl = NULL; + fop_readlink_cbk_t progcbk = NULL; + + nfl_to_prog_data (nfl, progcbk, frame); + nfs_fop_restore_root_ino (nfl, op_ret, buf, NULL, NULL, NULL); + if (progcbk) + progcbk (frame, cookie, this, op_ret, op_errno, path, buf, + xdata); + nfs_stack_destroy (nfl, frame); + return 0; +} + + +int +nfs_fop_readlink (xlator_t *nfsx, xlator_t *xl, nfs_user_t *nfu, loc_t *pathloc, + size_t size, fop_readlink_cbk_t cbk, void *local) +{ + call_frame_t *frame = NULL; + int ret = -EFAULT; + struct nfs_fop_local *nfl = NULL; + + if ((!nfsx) || (!xl) || (!pathloc) || (!nfu)) + return ret; + + gf_log (GF_NFS, GF_LOG_TRACE, "Readlink: %s", pathloc->path); + nfs_fop_handle_frame_create (frame, nfsx, nfu, ret, err); + nfs_fop_handle_local_init (frame, nfsx, nfl, cbk, local, ret, err); + nfs_fop_save_root_ino (nfl, pathloc); + + STACK_WIND_COOKIE (frame, nfs_fop_readlink_cbk, xl, xl, + xl->fops->readlink, pathloc, size, NULL); + ret = 0; +err: + if (ret < 0) { + if (frame) + nfs_stack_destroy (nfl, frame); + } + + return ret; +} + + +int32_t +nfs_fop_mknod_cbk (call_frame_t *frame, void *cookie, xlator_t *this, + int32_t op_ret, int32_t op_errno, inode_t *inode, + struct iatt *buf, struct iatt *preparent, + struct iatt *postparent, dict_t *xdata) +{ + struct nfs_fop_local *nfl = NULL; + fop_mknod_cbk_t progcbk = NULL; + + if (op_ret == 0) { + nfs_fix_generation(this,inode); + } + + nfl_to_prog_data (nfl, progcbk, frame); + nfs_fop_restore_root_ino (nfl, op_ret,buf, NULL, preparent, postparent); + if (progcbk) + progcbk (frame, cookie, this, op_ret, op_errno, inode, buf, + preparent, postparent, xdata); + nfs_stack_destroy (nfl, frame); + return 0; +} + + +int +nfs_fop_mknod (xlator_t *nfsx, xlator_t *xl, nfs_user_t *nfu, loc_t *pathloc, + mode_t mode, dev_t dev, fop_mknod_cbk_t cbk, void *local) +{ + call_frame_t *frame = NULL; + int ret = -EFAULT; + struct nfs_fop_local *nfl = NULL; + + if ((!nfsx) || (!xl) || (!pathloc) || (!nfu)) + return ret; + + gf_log (GF_NFS, GF_LOG_TRACE, "Mknod: %s", pathloc->path); + nfs_fop_handle_frame_create (frame, nfsx, nfu, ret, err); + nfs_fop_handle_local_init (frame, nfsx, nfl, cbk, local, ret, err); + nfs_fop_save_root_ino (nfl, pathloc); + nfs_fop_gfid_setup (nfl, pathloc->inode, ret, err); + + STACK_WIND_COOKIE (frame, nfs_fop_mknod_cbk, xl, xl, xl->fops->mknod, + pathloc, mode, dev, 0, nfl->dictgfid); + ret = 0; +err: + if (ret < 0) { + if (frame) + nfs_stack_destroy (nfl, frame); + } + + return ret; +} + +int32_t +nfs_fop_rmdir_cbk (call_frame_t *frame, void *cookie, xlator_t *this, + int32_t op_ret, int32_t op_errno, struct iatt *preparent, + struct iatt *postparent, dict_t *xdata) +{ + struct nfs_fop_local *nfl = frame->local; + fop_rmdir_cbk_t progcbk = NULL; + + nfl_to_prog_data (nfl, progcbk, frame); + nfs_fop_restore_root_ino (nfl, op_ret, NULL, NULL, preparent, + postparent); + if (progcbk) + progcbk (frame, cookie, this, op_ret, op_errno, preparent, + postparent, NULL); + nfs_stack_destroy (nfl, frame); + return 0; +} + + + +int +nfs_fop_rmdir (xlator_t *nfsx, xlator_t *xl, nfs_user_t *nfu, loc_t *pathloc, + fop_rmdir_cbk_t cbk, void *local) +{ + call_frame_t *frame = NULL; + int ret = -EFAULT; + struct nfs_fop_local *nfl = NULL; + + if ((!nfsx) || (!xl) || (!pathloc) || (!nfu)) + return ret; + + gf_log (GF_NFS, GF_LOG_TRACE, "Rmdir: %s", pathloc->path); + nfs_fop_handle_frame_create (frame, nfsx, nfu, ret, err); + nfs_fop_handle_local_init (frame, nfsx, nfl, cbk, local, ret, err); + nfs_fop_save_root_ino (nfl, pathloc); + + STACK_WIND_COOKIE (frame, nfs_fop_rmdir_cbk, xl, xl, xl->fops->rmdir, + pathloc, 0, NULL); + ret = 0; +err: + if (ret < 0) { + if (frame) + nfs_stack_destroy (nfl, frame); + } + + return ret; + +} + + +int32_t +nfs_fop_unlink_cbk (call_frame_t *frame, void *cookie, xlator_t *this, + int32_t op_ret, int32_t op_errno, struct iatt *preparent, + struct iatt *postparent, dict_t *xdata) +{ + struct nfs_fop_local *nfl = frame->local; + fop_unlink_cbk_t progcbk = NULL; + + nfl_to_prog_data (nfl, progcbk, frame); + nfs_fop_restore_root_ino (nfl, op_ret, NULL, NULL, preparent, + postparent); + if (progcbk) + progcbk (frame, cookie, this, op_ret, op_errno, preparent, + postparent, xdata); + nfs_stack_destroy (nfl, frame); + return 0; +} + + +int +nfs_fop_unlink (xlator_t *nfsx, xlator_t *xl, nfs_user_t *nfu, loc_t *pathloc, + fop_unlink_cbk_t cbk, void *local) +{ + call_frame_t *frame = NULL; + int ret = -EFAULT; + struct nfs_fop_local *nfl = NULL; + + if ((!nfsx) || (!xl) || (!pathloc) || (!nfu)) + return ret; + + gf_log (GF_NFS, GF_LOG_TRACE, "Unlink: %s", pathloc->path); + nfs_fop_handle_frame_create (frame, nfsx, nfu, ret, err); + nfs_fop_handle_local_init (frame, nfsx, nfl, cbk, local, ret, err); + nfs_fop_save_root_ino (nfl, pathloc); + + STACK_WIND_COOKIE (frame, nfs_fop_unlink_cbk, xl, xl, + xl->fops->unlink, pathloc, 0, NULL); + ret = 0; +err: + if (ret < 0) { + if (frame) + nfs_stack_destroy (nfl, frame); + } + + return ret; + +} + + +int32_t +nfs_fop_link_cbk (call_frame_t *frame, void *cookie, xlator_t *this, + int32_t op_ret, int32_t op_errno, inode_t *inode, + struct iatt *buf, struct iatt *preparent, + struct iatt *postparent, dict_t *xdata) +{ + struct nfs_fop_local *nfl = NULL; + fop_link_cbk_t progcbk = NULL; + + if (op_ret == 0) { + nfs_fix_generation(this,inode); + } + + nfl_to_prog_data (nfl, progcbk, frame); + nfs_fop_restore_root_ino (nfl, op_ret, buf, NULL, preparent, + postparent); + if (progcbk) + progcbk (frame, cookie, this, op_ret, op_errno, inode, buf, + preparent, postparent, xdata); + + nfs_stack_destroy (nfl, frame); + return 0; +} + + +int +nfs_fop_link (xlator_t *nfsx, xlator_t *xl, nfs_user_t *nfu, loc_t *oldloc, + loc_t *newloc, fop_link_cbk_t cbk, void *local) +{ + call_frame_t *frame = NULL; + int ret = -EFAULT; + struct nfs_fop_local *nfl = NULL; + + if ((!nfsx) || (!xl) || (!oldloc) || (!newloc) || (!nfu)) + return ret; + + gf_log (GF_NFS, GF_LOG_TRACE, "Link: %s -> %s", newloc->path, + oldloc->path); + nfs_fop_handle_frame_create (frame, nfsx, nfu, ret, err); + nfs_fop_handle_local_init (frame, nfsx, nfl, cbk, local, ret, err); + nfs_fop_save_root_ino (nfl, newloc); + + STACK_WIND_COOKIE (frame, nfs_fop_link_cbk, xl, xl, xl->fops->link, + oldloc, newloc, NULL); + ret = 0; +err: + if (ret < 0) { + if (frame) + nfs_stack_destroy (nfl, frame); + } + + return ret; +} + + +int32_t +nfs_fop_rename_cbk (call_frame_t *frame, void *cookie, xlator_t *this, + int32_t op_ret, int32_t op_errno, struct iatt *buf, + struct iatt *preoldparent, struct iatt *postoldparent, + struct iatt *prenewparent, struct iatt *postnewparent, + dict_t *xdata) +{ + + struct nfs_fop_local *nfl = NULL; + fop_rename_cbk_t progcbk = NULL; + + nfl_to_prog_data (nfl, progcbk, frame); + /* The preattr arg needs to be NULL instead of @buf because it is + * possible that the new parent is not root whereas the source dir + * could have been. That is handled in the next macro. + */ + nfs_fop_restore_root_ino (nfl, op_ret, NULL, NULL, preoldparent, + postoldparent); + nfs_fop_newloc_restore_root_ino (nfl, op_ret, buf, NULL, prenewparent, + postnewparent); + if (progcbk) + progcbk (frame, cookie, this, op_ret, op_errno, buf, + preoldparent, postoldparent, prenewparent, + postnewparent, xdata); + nfs_stack_destroy (nfl, frame); + return 0; +} + + +int +nfs_fop_rename (xlator_t *nfsx, xlator_t *xl, nfs_user_t *nfu, loc_t *oldloc, + loc_t *newloc, fop_rename_cbk_t cbk, void *local) +{ + call_frame_t *frame = NULL; + int ret = -EFAULT; + struct nfs_fop_local *nfl = NULL; + + if ((!nfsx) || (!xl) || (!oldloc) || (!newloc) || (!nfu)) + return ret; + + gf_log (GF_NFS, GF_LOG_TRACE, "Rename: %s -> %s", oldloc->path, + newloc->path); + nfs_fop_handle_frame_create (frame, nfsx, nfu, ret, err); + nfs_fop_handle_local_init (frame, nfsx, nfl, cbk, local, ret, err); + nfs_fop_save_root_ino (nfl, oldloc); + nfs_fop_newloc_save_root_ino (nfl, newloc); + + STACK_WIND_COOKIE (frame, nfs_fop_rename_cbk, xl, xl, + xl->fops->rename, oldloc, newloc, NULL); + ret = 0; +err: + if (ret < 0) { + if (frame) + nfs_stack_destroy (nfl, frame); + } + + return ret; +} + + +int32_t +nfs_fop_open_cbk (call_frame_t *frame, void *cookie, xlator_t *this, + int32_t op_ret, int32_t op_errno, fd_t *fd, dict_t *xdata) +{ + struct nfs_fop_local *nfl = NULL; + fop_open_cbk_t progcbk = NULL; + + nfl_to_prog_data (nfl, progcbk, frame); + if (progcbk) + progcbk (frame, cookie, this, op_ret, op_errno, fd, xdata); + nfs_stack_destroy (nfl, frame); + + return 0; +} + +int +nfs_fop_open (xlator_t *nfsx, xlator_t *xl, nfs_user_t *nfu, loc_t *loc, + int32_t flags, fd_t *fd, fop_open_cbk_t cbk, + void *local) +{ + call_frame_t *frame = NULL; + int ret = -EFAULT; + struct nfs_fop_local *nfl = NULL; + + if ((!nfsx) || (!xl) || (!loc) || (!fd) || (!nfu)) + return ret; + + gf_log (GF_NFS, GF_LOG_TRACE, "Open: %s", loc->path); + nfs_fop_handle_frame_create (frame, nfsx, nfu, ret, err); + nfs_fop_handle_local_init (frame, nfsx, nfl, cbk, local, ret, err); + + STACK_WIND_COOKIE (frame, nfs_fop_open_cbk, xl, xl, xl->fops->open, + loc, flags, fd, NULL); + ret = 0; +err: + if (ret < 0) { + if (frame) + nfs_stack_destroy (nfl, frame); + } + + return ret; +} + + +int32_t +nfs_fop_writev_cbk (call_frame_t *frame, void *cookie, xlator_t *this, + int32_t op_ret, int32_t op_errno, struct iatt *prebuf, + struct iatt *postbuf, dict_t *xdata) +{ + struct nfs_fop_local *nfl = NULL; + fop_writev_cbk_t progcbk = NULL; + + nfl_to_prog_data (nfl, progcbk, frame); + nfs_fop_restore_root_ino (nfl, op_ret, prebuf, postbuf, NULL, NULL); + if (progcbk) + progcbk (frame, cookie, this, op_ret, op_errno, prebuf, + postbuf, xdata); + + nfs_stack_destroy (nfl, frame); + + return 0; +} + + +int +nfs_fop_write (xlator_t *nfsx, xlator_t *xl, nfs_user_t *nfu, fd_t *fd, + struct iobref *srciobref, struct iovec *vector, int32_t count, + off_t offset, fop_writev_cbk_t cbk, void *local) +{ + call_frame_t *frame = NULL; + int ret = -EFAULT; + struct nfs_fop_local *nfl = NULL; + + if ((!nfsx) || (!xl) || (!fd) || (!vector) || (!nfu) || (!srciobref)) + return ret; + + nfs_fop_handle_frame_create (frame, nfsx, nfu, ret, err); + nfs_fop_handle_local_init (frame, nfsx, nfl, cbk, local, ret, err); + nfs_fop_save_root_fd_ino (nfl, fd); +/* + nfl->iobref = iobref_new (); + if (!nfl->iobref) { + gf_log (GF_NFS, GF_LOG_ERROR, "iobref creation failed"); + ret = -ENOMEM; + goto err; + } + + iobref_add (nfl->iobref, srciob); +*/ + STACK_WIND_COOKIE (frame, nfs_fop_writev_cbk, xl, xl,xl->fops->writev, + fd, vector, count, offset, fd->flags, srciobref, NULL); + ret = 0; +err: + if (ret < 0) { + if (frame) + nfs_stack_destroy (nfl, frame); + } + + return ret; +} + + +int32_t +nfs_fop_fsync_cbk (call_frame_t *frame, void *cookie, xlator_t *this, + int32_t op_ret, int32_t op_errno, struct iatt *prebuf, + struct iatt *postbuf, dict_t *xdata) +{ + struct nfs_fop_local *nfl = NULL; + fop_fsync_cbk_t progcbk = NULL; + + nfl_to_prog_data (nfl, progcbk, frame); + nfs_fop_restore_root_ino (nfl, op_ret, prebuf, postbuf, NULL, NULL); + if (progcbk) + progcbk (frame, cookie, this, op_ret, op_errno, prebuf, + postbuf, xdata); + + nfs_stack_destroy (nfl, frame); + return 0; +} + + + +int +nfs_fop_fsync (xlator_t *nfsx, xlator_t *xl, nfs_user_t *nfu, fd_t *fd, + int32_t datasync, fop_fsync_cbk_t cbk, void *local) +{ + call_frame_t *frame = NULL; + int ret = -EFAULT; + struct nfs_fop_local *nfl = NULL; + + if ((!nfsx) || (!xl) || (!fd)) + return ret; + + nfs_fop_handle_frame_create (frame, nfsx, nfu, ret, err); + nfs_fop_handle_local_init (frame, nfsx, nfl, cbk, local, ret, err); + nfs_fop_save_root_fd_ino (nfl, fd); + + STACK_WIND_COOKIE (frame, nfs_fop_fsync_cbk, xl, xl, + xl->fops->fsync, fd, datasync, NULL); + ret = 0; +err: + if (ret < 0) { + if (frame) + nfs_stack_destroy (nfl, frame); + } + + return ret; +} + + +int32_t +nfs_fop_readv_cbk (call_frame_t *frame, void *cookie, xlator_t *this, + int32_t op_ret, int32_t op_errno, struct iovec *vector, + int32_t count, struct iatt *stbuf, struct iobref *iobref, + dict_t *xdata) +{ + struct nfs_fop_local *nfl = NULL; + fop_readv_cbk_t progcbk = NULL; + + nfl_to_prog_data (nfl, progcbk, frame); + nfs_fop_restore_root_ino (nfl, op_ret, stbuf, NULL, NULL, NULL); + if (progcbk) + progcbk (frame, cookie, this, op_ret, op_errno, vector, count, + stbuf, iobref, xdata); + + nfs_stack_destroy (nfl, frame); + return 0; +} + + +int +nfs_fop_read (xlator_t *nfsx, xlator_t *xl, nfs_user_t *nfu, fd_t *fd, + size_t size, off_t offset, fop_readv_cbk_t cbk, void *local) +{ + call_frame_t *frame = NULL; + int ret = -EFAULT; + struct nfs_fop_local *nfl = NULL; + + if ((!xl) || (!fd) || (!nfu)) + return ret; + + nfs_fop_handle_frame_create (frame, nfsx, nfu, ret, err); + nfs_fop_handle_local_init (frame, nfsx, nfl, cbk, local, ret, err); + nfs_fop_save_root_fd_ino (nfl, fd); + + STACK_WIND_COOKIE (frame, nfs_fop_readv_cbk, xl, xl, xl->fops->readv, + fd, size, offset, 0, NULL); + ret = 0; +err: + if (ret < 0) { + if (frame) + nfs_stack_destroy (nfl, frame); + } + + return ret; +} + +int32_t +nfs_fop_lk_cbk (call_frame_t *frame, void *cookie, xlator_t *this, + int32_t op_ret, int32_t op_errno, struct gf_flock *flock, + dict_t *xdata) +{ + struct nfs_fop_local *nfl = NULL; + fop_lk_cbk_t progcbk = NULL; + + nfl_to_prog_data (nfl, progcbk, frame); + + if (!op_ret) + fd_lk_insert_and_merge (nfl->fd, nfl->cmd, &nfl->flock); + + fd_unref (nfl->fd); + + if (progcbk) + progcbk (frame, cookie, this, op_ret, op_errno, flock, xdata); + + nfs_stack_destroy (nfl, frame); + return 0; +} + + +int +nfs_fop_lk (xlator_t *nfsx, xlator_t *xl, nfs_user_t *nfu, fd_t *fd, + int cmd, struct gf_flock *flock, fop_lk_cbk_t cbk, void *local) +{ + call_frame_t *frame = NULL; + int ret = -EFAULT; + struct nfs_fop_local *nfl = NULL; + + if ((!xl) || (!fd) || (!nfu)) + return ret; + + nfs_fop_handle_frame_create (frame, nfsx, nfu, ret, err); + nfs_fop_handle_local_init (frame, nfsx, nfl, cbk, local, ret, err); + + nfl->cmd = cmd; + nfl->fd = fd_ref (fd); + nfl->flock = *flock; + + STACK_WIND_COOKIE (frame, nfs_fop_lk_cbk, xl, xl, xl->fops->lk, + fd, cmd, flock, NULL); + ret = 0; +err: + if (ret < 0) { + if (frame) + nfs_stack_destroy (nfl, frame); + } + + return ret; +} + +int32_t +nfs_fop_getxattr_cbk (call_frame_t *frame, void *cookie, xlator_t *this, + int32_t op_ret, int32_t op_errno, dict_t *dict, + dict_t *xdata) +{ + struct nfs_fop_local *nfl = NULL; + fop_getxattr_cbk_t progcbk = NULL; + + nfl_to_prog_data (nfl, progcbk, frame); + + if (progcbk) + progcbk (frame, cookie, this, op_ret, op_errno, dict, xdata); + + nfs_stack_destroy (nfl, frame); + return 0; +} + + +int +nfs_fop_getxattr (xlator_t *nfsx, xlator_t *xl, nfs_user_t *nfu, loc_t *loc, + char *name, dict_t *xdata, fop_getxattr_cbk_t cbk, void *local) +{ + call_frame_t *frame = NULL; + int ret = -EFAULT; + struct nfs_fop_local *nfl = NULL; + + if ((!xl) || (!loc) || (!nfu)) + return ret; + + nfs_fop_handle_frame_create (frame, nfsx, nfu, ret, err); + nfs_fop_handle_local_init (frame, nfsx, nfl, cbk, local, ret, err); + + STACK_WIND_COOKIE (frame, nfs_fop_getxattr_cbk, xl, xl, xl->fops->getxattr, + loc, name, NULL); + ret = 0; +err: + if (ret < 0) { + if (frame) + nfs_stack_destroy (nfl, frame); + } + + return ret; +} + + +int32_t +nfs_fop_setxattr_cbk (call_frame_t *frame, void *cookie, xlator_t *this, + int32_t op_ret, int32_t op_errno, dict_t *xdata) +{ + struct nfs_fop_local *nfl = NULL; + fop_setxattr_cbk_t progcbk = NULL; + + nfl_to_prog_data (nfl, progcbk, frame); + + if (progcbk) + progcbk (frame, cookie, this, op_ret, op_errno, xdata); + + nfs_stack_destroy (nfl, frame); + return 0; +} + + +int +nfs_fop_setxattr (xlator_t *nfsx, xlator_t *xl, nfs_user_t *nfu, + loc_t *loc, dict_t *dict, int32_t flags, dict_t *xdata, + fop_setxattr_cbk_t cbk, void *local) +{ + call_frame_t *frame = NULL; + int ret = -EFAULT; + struct nfs_fop_local *nfl = NULL; + + if ((!xl) || (!loc) || (!nfu)) + return ret; + + nfs_fop_handle_frame_create (frame, nfsx, nfu, ret, err); + nfs_fop_handle_local_init (frame, nfsx, nfl, cbk, local, ret, err); + + STACK_WIND_COOKIE (frame, nfs_fop_setxattr_cbk, xl, xl, xl->fops->setxattr, + loc, dict, flags, xdata); + ret = 0; +err: + if (ret < 0) { + if (frame) + nfs_stack_destroy (nfl, frame); + } + + return ret; +} + + +int32_t +nfs_fop_truncate_cbk (call_frame_t *frame, void *cookie, xlator_t *this, + int32_t op_ret, int32_t op_errno, struct iatt *prebuf, + struct iatt *postbuf, dict_t *xdata) +{ + struct nfs_fop_local *nfl = NULL; + fop_truncate_cbk_t progcbk = NULL; + + nfl_to_prog_data (nfl, progcbk, frame); + nfs_fop_restore_root_ino (nfl, op_ret, prebuf, postbuf, NULL, NULL); + if (progcbk) + progcbk (frame, cookie, this, op_ret, op_errno, prebuf, + postbuf, xdata); + + nfs_stack_destroy (nfl, frame); + return 0; +} + + +int +nfs_fop_truncate (xlator_t *nfsx, xlator_t *xl, nfs_user_t *nfu, loc_t *loc, + off_t offset, fop_truncate_cbk_t cbk, void *local) +{ + call_frame_t *frame = NULL; + int ret = -EFAULT; + struct nfs_fop_local *nfl = NULL; + + if ((!nfsx) || (!xl) || (!loc) || (!nfu)) + return ret; + + nfs_fop_handle_frame_create (frame, nfsx, nfu, ret, err); + nfs_fop_handle_local_init (frame, nfsx, nfl, cbk, local, ret, err); + nfs_fop_save_root_ino (nfl, loc); + + STACK_WIND_COOKIE (frame, nfs_fop_truncate_cbk, xl, xl, + xl->fops->truncate, loc, offset, NULL); + + ret = 0; +err: + if (ret < 0) { + if (frame) + nfs_stack_destroy (nfl, frame); + } + + return ret; +} diff --git a/xlators/nfs/server/src/nfs-fops.h b/xlators/nfs/server/src/nfs-fops.h new file mode 100644 index 000000000..44e99c66b --- /dev/null +++ b/xlators/nfs/server/src/nfs-fops.h @@ -0,0 +1,248 @@ +/* + Copyright (c) 2010-2011 Gluster, Inc. <http://www.gluster.com> + This file is part of GlusterFS. + + This file is licensed to you under your choice of the GNU Lesser + General Public License, version 3 or any later version (LGPLv3 or + later), or the GNU General Public License, version 2 (GPLv2), in all + cases as published by the Free Software Foundation. +*/ + +#ifndef _NFS_FOPS_H_ +#define _NFS_FOPS_H_ + +#ifndef _CONFIG_H +#define _CONFIG_H +#include "config.h" +#endif + +#include "dict.h" +#include "xlator.h" +#include "iobuf.h" +#include "call-stub.h" +#include "stack.h" +#include "nfs.h" +#include "nfs-common.h" +#include <semaphore.h> + +/* This structure used to communicate state between a fop and its callback. + * The problem is, when we're calling a fop in the nfs op handler, the callback + * is the NFS protocol's callback and we have to perform all GlusterFS + * inode, inode table, fd_ts and fd table operations in the NFS callback. That + * approach soon gets extremely complicated and confusing because, then we have + * to try and separate in our heads which source lines in those callbacks are + * required for serving the NFS op and which ones are needed for satisfying + * GlusterFS requirements. This structure allows us avoid performing GlusterFS + * state maintenance operations inside the fops layer itself. Now, before + * we call the callback registered by the NFS operation, a hidden fops-layer + * specific callback is called which performs the state maintenance and then + * calls the NFS callback. + * + * These are allocated from a mem-pool stored in the nfs xlator's state. + * i.e. struct nfs_state. + * That is initiated in nfs_init_subvolumes in nfs.c. + */ +struct nfs_fop_local { + /* The local sent along by the user of the fop. */ + void *proglocal; + + /* The address of the callback supplied by the user. After our + * callback is executed this one is called. + * The exact cast destination of this pointer will depend on the + * fop that is being called. + */ + void *progcbk; + + /* Used only for write requests. */ + struct iobref *iobref; + + inode_t *parent; + inode_t *newparent; + inode_t *inode; + + /* Set to 1 by nfs-inodes layer, which uses this to decide whether to + * link the newly allocated inode into the itable, in case the fop was + * successful. + */ + int newinode; + + /* Used by nfs-fops layer in order to determine whether to funge the + * ino in a dir's stbuf. This funging of root ino is needed to ensure + * that the root ino remains 1 even when the NFS server has been + * restarted. Note that in distribute, a fresh lookup and a revalidate + * on the root inode returns two different inode numbers and this we + * need to handle by ourself. + */ + int rootinode; + + /* This member is used to determine whether the new parent of a file + * being renamed is the root directory. If yes, the ino is funged. + */ + int newrootinode; + int newrootparentinode; + + /* Determines whether to funge the ino in the post and pre parent + * stbufs for a file/dir where the parent directory could be the root + * dir. Needed here because of the same reason as above. + */ + int rootparentinode; + + char path[NFS_NAME_MAX + 1]; + char newpath[NFS_NAME_MAX + 1]; + xlator_t *nfsx; + dict_t *dictgfid; + + fd_t *fd; + int cmd; + struct gf_flock flock; +}; + +extern struct nfs_fop_local * +nfs_fop_local_init (xlator_t *xl); + +extern void +nfs_fop_local_wipe (xlator_t *xl, struct nfs_fop_local *l); + +#define nfs_state(nfsxl) (nfsxl)->private +#define nfs_fop_mempool(nfxl) (((struct nfs_state *)nfs_state(nfxl))->foppool) + +#define prog_data_to_nfl(nf,nflocal, fram, pcbk, plocal) \ + do { \ + nflocal = nfs_fop_local_init (nf); \ + if (nflocal) { \ + nflocal->proglocal = plocal; \ + nflocal->progcbk = *VOID(&pcbk); \ + nflocal->nfsx = nf; \ + if (fram) \ + ((call_frame_t *)fram)->local = nflocal;\ + } \ + } while (0) \ + + + +#define nfl_to_prog_data(nflocal, pcbk, fram) \ + do { \ + nflocal = fram->local; \ + fram->local = nflocal->proglocal; \ + pcbk = nflocal->progcbk; \ + } while (0) \ + +#define nfs_fop_handle_local_init(fram,nfx, nfloc, cbck,prgloc,retval,lab) \ + do { \ + prog_data_to_nfl (nfx, nfloc, fram, cbck, prgloc); \ + if (!nfloc) { \ + gf_log (GF_NFS,GF_LOG_ERROR,"Failed to init local");\ + retval = -ENOMEM; \ + goto lab; \ + } \ + } while (0) \ + +extern int +nfs_fop_fstat (xlator_t *nfsx, xlator_t *xl, nfs_user_t *nfu, fd_t *fd, + fop_stat_cbk_t cbk, void *local); + +extern int +nfs_fop_readdirp (xlator_t *nfsx, xlator_t *xl, nfs_user_t *nfu, fd_t *dirfd, + size_t bufsize, off_t offset, fop_readdir_cbk_t cbk, + void *local); +extern int +nfs_fop_lookup (xlator_t *nfsx, xlator_t *xl, nfs_user_t *nfu, loc_t *loc, + fop_lookup_cbk_t cbk, void *local); + +extern int +nfs_fop_create (xlator_t *nfsx, xlator_t *xl, nfs_user_t *nfu, loc_t *pathloc, + int flags, mode_t mode, fd_t *fd, fop_create_cbk_t cbk, + void *local); +extern int +nfs_fop_flush (xlator_t *nfsx, xlator_t *xl, nfs_user_t *nfu, fd_t *fd, + fop_flush_cbk_t cbk, void *local); + +extern int +nfs_fop_mkdir (xlator_t *nfsx, xlator_t *xl, nfs_user_t *nfu, loc_t *pathloc, + mode_t mode, fop_mkdir_cbk_t cbk, void *local); + +extern int +nfs_fop_truncate (xlator_t *nfsx, xlator_t *xl, nfs_user_t *nfu, loc_t *loc, + off_t offset, fop_truncate_cbk_t cbk, void *local); + +extern int +nfs_fop_read (xlator_t *nfsx, xlator_t *xl, nfs_user_t *nfu, fd_t *fd, + size_t size, off_t offset, fop_readv_cbk_t cbk, void *local); + +extern int +nfs_fop_fsync (xlator_t *nfsx, xlator_t *xl, nfs_user_t *nfu, fd_t *fd, + int32_t datasync, fop_fsync_cbk_t cbk, void *local); + +extern int +nfs_fop_write (xlator_t *nfsx, xlator_t *xl, nfs_user_t *nfu, fd_t *fd, + struct iobref *srciobref, struct iovec *vector, int32_t count, + off_t offset, fop_writev_cbk_t cbk, void *local); + +extern int +nfs_fop_open (xlator_t *nfsx, xlator_t *xl, nfs_user_t *nfu, loc_t *loc, + int32_t flags, fd_t *fd, fop_open_cbk_t cbk, + void *local); + +extern int +nfs_fop_rename (xlator_t *nfsx, xlator_t *xl, nfs_user_t *nfu, loc_t *oldloc, + loc_t *newloc, fop_rename_cbk_t cbk, void *local); + +extern int +nfs_fop_link (xlator_t *nfsx, xlator_t *xl, nfs_user_t *nfu, loc_t *oldloc, + loc_t *newloc, fop_link_cbk_t cbk, void *local); + +extern int +nfs_fop_unlink (xlator_t *nfsx, xlator_t *xl, nfs_user_t *nfu, loc_t *pathloc, + fop_unlink_cbk_t cbk, void *local); + +extern int +nfs_fop_rmdir (xlator_t *nfsx, xlator_t *xl, nfs_user_t *nfu, loc_t *pathloc, + fop_rmdir_cbk_t cbk, void *local); + +extern int +nfs_fop_mknod (xlator_t *nfsx, xlator_t *xl, nfs_user_t *nfu, loc_t *pathloc, + mode_t mode, dev_t dev, fop_mknod_cbk_t cbk, void *local); + +extern int +nfs_fop_readlink (xlator_t *nfsx, xlator_t *xl, nfs_user_t *nfu, loc_t *pathloc, + size_t size, fop_readlink_cbk_t cbk, void *local); + +extern int +nfs_fop_symlink (xlator_t *nfsx, xlator_t *xl, nfs_user_t *nfu, char *target, + loc_t *pathloc, fop_symlink_cbk_t cbk, void *local); + +extern int +nfs_fop_setattr (xlator_t *nfsx, xlator_t *xl, nfs_user_t *nfu, loc_t *pathloc, + struct iatt *buf, int32_t valid, fop_setattr_cbk_t cbk, + void *local); + +extern int +nfs_fop_statfs (xlator_t *nfsx, xlator_t *xl, nfs_user_t *nfu, loc_t *pathloc, + fop_statfs_cbk_t cbk, void *local); + +extern int +nfs_fop_opendir (xlator_t *nfsx, xlator_t *xl, nfs_user_t *nfu, loc_t *pathloc, + fd_t *dirfd, fop_opendir_cbk_t cbk, void *local); + +extern int +nfs_fop_stat (xlator_t *nfsx, xlator_t *xl, nfs_user_t *nfu, loc_t *loc, + fop_stat_cbk_t cbk, void *local); + +extern int +nfs_fop_access (xlator_t *nfsx, xlator_t *xl, nfs_user_t *nfu, loc_t *loc, + int32_t accesstest, fop_access_cbk_t cbk, void *local); + +extern int +nfs_fop_lk (xlator_t *nfsx, xlator_t *xl, nfs_user_t *nfu, fd_t *fd, + int cmd, struct gf_flock *flock, fop_lk_cbk_t cbk, void *local); + +extern int +nfs_fop_getxattr (xlator_t *nfsx, xlator_t *xl, nfs_user_t *nfu, loc_t *loc, + char *name, dict_t *xdata, fop_getxattr_cbk_t cbk, void *local); + +extern int +nfs_fop_setxattr (xlator_t *nfsx, xlator_t *xl, nfs_user_t *nfu, + loc_t *loc, dict_t *dict, int32_t flags, dict_t *xdata, + fop_setxattr_cbk_t cbk, void *local); + +#endif diff --git a/xlators/nfs/server/src/nfs-generics.c b/xlators/nfs/server/src/nfs-generics.c new file mode 100644 index 000000000..cb32b7f1b --- /dev/null +++ b/xlators/nfs/server/src/nfs-generics.c @@ -0,0 +1,345 @@ +/* + Copyright (c) 2010-2011 Gluster, Inc. <http://www.gluster.com> + This file is part of GlusterFS. + + This file is licensed to you under your choice of the GNU Lesser + General Public License, version 3 or any later version (LGPLv3 or + later), or the GNU General Public License, version 2 (GPLv2), in all + cases as published by the Free Software Foundation. +*/ + +#ifndef _CONFIG_H +#define _CONFIG_H +#include "config.h" +#endif + + +#include "string.h" + +#include "inode.h" +#include "nfs.h" +#include "nfs-fops.h" +#include "nfs-inodes.h" +#include "nfs-generics.h" +#include "xlator.h" + + +int +nfs_fstat (xlator_t *nfsx, xlator_t *xl, nfs_user_t *nfu, fd_t *fd, + fop_stat_cbk_t cbk, void *local) +{ + int ret = -EFAULT; + + if ((!nfsx) || (!xl) || (!fd) || (!nfu)) + return ret; + + ret = nfs_fop_fstat (nfsx, xl, nfu, fd, cbk, local); + return ret; +} + +int +nfs_access (xlator_t *nfsx, xlator_t *xl, nfs_user_t *nfu, loc_t *pathloc, + int32_t accesstest, fop_access_cbk_t cbk, void *local) +{ + int ret = -EFAULT; + + if ((!nfsx) || (!xl) || (!pathloc) || (!nfu)) + return ret; + + ret = nfs_fop_access (nfsx, xl, nfu, pathloc, accesstest, cbk, local); + + return ret; +} + +int +nfs_stat (xlator_t *nfsx, xlator_t *xl, nfs_user_t *nfu, loc_t *pathloc, + fop_stat_cbk_t cbk, void *local) +{ + int ret = -EFAULT; + + if ((!nfsx) || (!xl) || (!pathloc) || (!nfu)) + return ret; + + ret = nfs_fop_stat (nfsx, xl, nfu, pathloc, cbk, local); + + return ret; +} + + +int +nfs_readdirp (xlator_t *nfsx, xlator_t *xl, nfs_user_t *nfu, fd_t *dirfd, + size_t bufsize, off_t offset, fop_readdir_cbk_t cbk, void *local) +{ + if ((!nfsx) || (!xl) || (!dirfd) || (!nfu)) + return -EFAULT; + + return nfs_fop_readdirp (nfsx, xl, nfu, dirfd, bufsize, offset, cbk, + local); +} + + +int +nfs_lookup (xlator_t *nfsx, xlator_t *xl, nfs_user_t *nfu, loc_t *pathloc, + fop_lookup_cbk_t cbk, void *local) +{ + int ret = -EFAULT; + + if ((!nfsx) || (!xl) || (!pathloc) || (!nfu)) + return ret; + + ret = nfs_fop_lookup (nfsx, xl, nfu, pathloc, cbk, local); + return ret; +} + +int +nfs_create (xlator_t *nfsx, xlator_t *xl, nfs_user_t *nfu, loc_t *pathloc, + int flags, mode_t mode, fop_create_cbk_t cbk, void *local) +{ + int ret = -EFAULT; + + if ((!nfsx) || (!xl) || (!pathloc) || (!nfu)) + return ret; + + ret = nfs_inode_create (nfsx, xl, nfu, pathloc, flags, mode, cbk,local); + return ret; +} + + +int +nfs_flush (xlator_t *nfsx, xlator_t *xl, nfs_user_t *nfu, fd_t *fd, + fop_flush_cbk_t cbk, void *local) +{ + return nfs_fop_flush (nfsx, xl, nfu, fd, cbk, local); +} + + + +int +nfs_mkdir (xlator_t *nfsx, xlator_t *xl, nfs_user_t *nfu, loc_t *pathloc, + mode_t mode, fop_mkdir_cbk_t cbk, void *local) +{ + int ret = -EFAULT; + + if ((!nfsx) || (!xl) || (!pathloc) || (!nfu)) + return ret; + + ret = nfs_inode_mkdir (nfsx, xl, nfu, pathloc, mode, cbk, local); + return ret; +} + + + +int +nfs_truncate (xlator_t *nfsx, xlator_t *xl, nfs_user_t *nfu, loc_t *pathloc, + off_t offset, fop_truncate_cbk_t cbk, void *local) +{ + int ret = -EFAULT; + + if ((!xl) || (!pathloc) || (!nfu)) + return ret; + + ret = nfs_fop_truncate (nfsx, xl, nfu, pathloc, offset, cbk, local); + return ret; +} + +int +nfs_read (xlator_t *nfsx, xlator_t *xl, nfs_user_t *nfu, fd_t *fd, size_t size, + off_t offset, fop_readv_cbk_t cbk, void *local) +{ + return nfs_fop_read (nfsx, xl, nfu, fd, size, offset, cbk, local); +} + +int +nfs_lk (xlator_t *nfsx, xlator_t *xl, nfs_user_t *nfu, fd_t *fd, + int cmd, struct gf_flock *flock, fop_lk_cbk_t cbk, void *local) +{ + return nfs_fop_lk ( nfsx, xl, nfu, fd, cmd, flock, cbk, local); +} + +int +nfs_getxattr (xlator_t *nfsx, xlator_t *xl, nfs_user_t *nfu, loc_t *loc, + char *name, dict_t *xdata, fop_getxattr_cbk_t cbk, void *local) +{ + return nfs_fop_getxattr (nfsx, xl, nfu, loc, name, xdata, cbk, local); +} + +int +nfs_setxattr (xlator_t *nfsx, xlator_t *xl, nfs_user_t *nfu, + loc_t *loc, dict_t *dict, int32_t flags, dict_t *xdata, + fop_setxattr_cbk_t cbk, void *local) +{ + return nfs_fop_setxattr (nfsx, xl, nfu, loc, dict, flags, xdata, cbk, + local); +} + +int +nfs_fsync (xlator_t *nfsx, xlator_t *xl, nfs_user_t *nfu, fd_t *fd, + int32_t datasync, fop_fsync_cbk_t cbk, void *local) +{ + return nfs_fop_fsync (nfsx, xl, nfu, fd, datasync, cbk, local); +} + + +int +nfs_write (xlator_t *nfsx, xlator_t *xl, nfs_user_t *nfu, fd_t *fd, + struct iobref *srciobref, struct iovec *vector, int32_t count, + off_t offset, fop_writev_cbk_t cbk, void *local) +{ + return nfs_fop_write (nfsx, xl, nfu, fd, srciobref, vector, count, + offset, cbk, local); +} + + +int +nfs_open (xlator_t *nfsx, xlator_t *xl, nfs_user_t *nfu, loc_t *pathloc, + int32_t flags, fop_open_cbk_t cbk, void *local) +{ + int ret = -EFAULT; + + if ((!nfsx) || (!xl) || (!pathloc) || (!nfu)) + return ret; + + ret = nfs_inode_open (nfsx, xl, nfu, pathloc, flags, cbk, + local); + return ret; +} + + +int +nfs_rename (xlator_t *nfsx, xlator_t *xl, nfs_user_t *nfu, loc_t *oldloc, + loc_t *newloc, fop_rename_cbk_t cbk, void *local) +{ + int ret = -EFAULT; + + if ((!nfsx) || (!xl) || (!oldloc) || (!newloc) || (!nfu)) + return ret; + + ret = nfs_inode_rename (nfsx, xl, nfu, oldloc, newloc, cbk, local); + return ret; +} + + +int +nfs_link (xlator_t *nfsx, xlator_t *xl, nfs_user_t *nfu, loc_t *oldloc, + loc_t *newloc, fop_link_cbk_t cbk, void *local) +{ + int ret = -EFAULT; + + if ((!nfsx) || (!xl) || (!oldloc) || (!newloc) || (!nfu)) + return ret; + + ret = nfs_inode_link (nfsx, xl, nfu, oldloc, newloc, cbk, local); + return ret; +} + + +int +nfs_unlink (xlator_t *nfsx, xlator_t *xl, nfs_user_t *nfu, loc_t *pathloc, + fop_unlink_cbk_t cbk, void *local) +{ + int ret = -EFAULT; + + if ((!xl) || (!pathloc) || (!nfu)) + return ret; + + ret = nfs_inode_unlink (nfsx, xl, nfu, pathloc, cbk, local); + return ret; +} + + +int +nfs_rmdir (xlator_t *nfsx, xlator_t *xl, nfs_user_t *nfu, loc_t *path, + fop_rmdir_cbk_t cbk, void *local) +{ + int ret = -EFAULT; + + if ((!nfsx) || (!xl) || (!path) || (!nfu)) + return ret; + + ret = nfs_inode_rmdir (nfsx, xl, nfu, path, cbk, local); + return ret; +} + + +int +nfs_mknod (xlator_t *nfsx, xlator_t *xl, nfs_user_t *nfu, loc_t *pathloc, + mode_t mode, dev_t dev, fop_mknod_cbk_t cbk, void *local) +{ + int ret = -EFAULT; + + if ((!nfsx) || (!xl) || (!pathloc) || (!nfu)) + return ret; + + ret = nfs_inode_mknod (nfsx, xl, nfu, pathloc, mode, dev, cbk, local); + return ret; +} + + +int +nfs_readlink (xlator_t *nfsx, xlator_t *xl, nfs_user_t *nfu, loc_t *linkloc, + fop_readlink_cbk_t cbk, void *local) +{ + int ret = -EFAULT; + + if ((!nfsx) || (!xl) || (!linkloc) || (!nfu)) + return ret; + + ret = nfs_fop_readlink (nfsx, xl, nfu, linkloc, NFS_PATH_MAX, cbk, + local); + return ret; +} + + +int +nfs_symlink (xlator_t *nfsx, xlator_t *xl, nfs_user_t *nfu, char *target, + loc_t *linkloc, fop_symlink_cbk_t cbk, void *local) +{ + int ret = -EFAULT; + + if ((!nfsx) || (!xl) || (!linkloc) || (!target) || (!nfu)) + return ret; + + ret = nfs_inode_symlink (nfsx, xl, nfu, target, linkloc, cbk, local); + return ret; +} + + + +int +nfs_setattr (xlator_t *nfsx, xlator_t *xl, nfs_user_t *nfu, loc_t *pathloc, + struct iatt *buf, int32_t valid, fop_setattr_cbk_t cbk, + void *local) +{ + int ret = -EFAULT; + + if ((!nfsx) || (!xl) || (!pathloc) || (!nfu)) + return ret; + + ret = nfs_fop_setattr (nfsx, xl, nfu, pathloc, buf, valid, cbk, local); + return ret; +} + + +int +nfs_statfs (xlator_t *nfsx, xlator_t *xl, nfs_user_t *nfu, loc_t *pathloc, + fop_statfs_cbk_t cbk, void *local) +{ + int ret = -EFAULT; + + if ((!nfsx) || (!xl) || (!pathloc) || (!nfu)) + return ret; + + ret = nfs_fop_statfs (nfsx, xl, nfu, pathloc, cbk, local); + return ret; +} + +int +nfs_opendir (xlator_t *nfsx, xlator_t *fopxl, nfs_user_t *nfu, loc_t *pathloc, + fop_opendir_cbk_t cbk, void *local) +{ + if ((!nfsx) || (!fopxl) || (!pathloc) || (!nfu)) + return -EFAULT; + + return nfs_inode_opendir (nfsx, fopxl, nfu, pathloc, cbk, local); +} + diff --git a/xlators/nfs/server/src/nfs-generics.h b/xlators/nfs/server/src/nfs-generics.h new file mode 100644 index 000000000..01876d68e --- /dev/null +++ b/xlators/nfs/server/src/nfs-generics.h @@ -0,0 +1,168 @@ +/* + Copyright (c) 2010-2011 Gluster, Inc. <http://www.gluster.com> + This file is part of GlusterFS. + + This file is licensed to you under your choice of the GNU Lesser + General Public License, version 3 or any later version (LGPLv3 or + later), or the GNU General Public License, version 2 (GPLv2), in all + cases as published by the Free Software Foundation. +*/ + +#ifndef _NFS_GENERICS_H_ +#define _NFS_GENERICS_H_ + +#ifndef _CONFIG_H +#define _CONFIG_H +#include "config.h" +#endif + +#include "nfs.h" +#include "xlator.h" +#include "nfs-fops.h" +#include "nfs-inodes.h" + +struct nfs_direntcache { + gf_dirent_t entries; /* Head of list of cached dirents. */ + gf_dirent_t *next; /* Pointer to the next entry that + * should be sent by readdir */ + uint64_t prev_off; /* Offset where the next read will + * happen. + */ +}; + +/* WE're trying to abstract the fops interface from the NFS xlator so that + * different NFS versions can simply call a standard interface and have fop + * interface dependent functions be handled internally. + * This structure is part of such an abstraction. The fops layer stores any + * state is requires in the fd. E.g. the dirent cache for a directory fd_t. + */ +typedef struct nfs_fop_fdcontext { + pthread_mutex_t lock; + size_t dirent_bufsize; + off_t offset; + struct nfs_direntcache *dcache; + xlator_t *dirvol; +} nfs_fdctx_t; + +extern int +nfs_fstat (xlator_t *nfsx, xlator_t *xl, nfs_user_t *nfu, fd_t *fd, + fop_stat_cbk_t cbk, void *local); + +extern int +nfs_readdirp (xlator_t *nfsx, xlator_t *xl, nfs_user_t *nfu, fd_t *dirfd, + size_t bufsize, off_t offset, fop_readdir_cbk_t cbk, void *local); + + +extern int +nfs_lookup (xlator_t *nfsx, xlator_t *xl, nfs_user_t *nfu, loc_t *pathloc, + fop_lookup_cbk_t cbk, void *local); + +extern int +nfs_create (xlator_t *nfsx, xlator_t *xl, nfs_user_t *nfu, loc_t *pathloc, + int flags, mode_t mode, fop_create_cbk_t cbk, void *local); + +extern int +nfs_flush (xlator_t *nfsx, xlator_t *xl, nfs_user_t *nfu, fd_t *fd, + fop_flush_cbk_t cbk, void *local); + +extern int +nfs_mkdir (xlator_t *nfsx, xlator_t *xl, nfs_user_t *nfu, loc_t *pathloc, + mode_t mode, fop_mkdir_cbk_t cbk, void *local); + +extern int +nfs_truncate (xlator_t *nfsx, xlator_t *xl, nfs_user_t *nfu, loc_t *pathloc, + off_t offset, fop_truncate_cbk_t cbk, void *local); + +extern int +nfs_read (xlator_t *nfsx, xlator_t *xl, nfs_user_t *nfu, fd_t *fd, size_t size, + off_t offset, fop_readv_cbk_t cbk, void *local); + +extern int +nfs_fsync (xlator_t *nfsx, xlator_t *xl, nfs_user_t *nfu, fd_t *fd, + int32_t datasync, fop_fsync_cbk_t cbk, void *local); + +extern int +nfs_write (xlator_t *nfsx, xlator_t *xl, nfs_user_t *nfu, fd_t *fd, + struct iobref *srciobref, struct iovec *vector, int32_t count, + off_t offset, fop_writev_cbk_t cbk, void *local); + +extern int +nfs_open (xlator_t *nfsx, xlator_t *xl, nfs_user_t *nfu, loc_t *pathloc, + int32_t flags, fop_open_cbk_t cbk, void *local); + +extern int +nfs_rename (xlator_t *nfsx, xlator_t *xl, nfs_user_t *nfu, loc_t *oldloc, + loc_t *newloc, fop_rename_cbk_t cbk, void *local); + +extern int +nfs_link (xlator_t *nfsx, xlator_t *xl, nfs_user_t *nfu, loc_t *oldloc, + loc_t *newloc, fop_link_cbk_t cbk, void *local); + +extern int +nfs_unlink (xlator_t *nfsx, xlator_t *xl, nfs_user_t *nfu, loc_t *pathloc, + fop_unlink_cbk_t cbk, void *local); + +extern int +nfs_rmdir (xlator_t *nfsx, xlator_t *xl, nfs_user_t *nfu, loc_t *pathloc, + fop_rmdir_cbk_t cbk, void *local); + +extern int +nfs_mknod (xlator_t *nfsx, xlator_t *xl, nfs_user_t *nfu, loc_t *pathloc, + mode_t mode, dev_t dev, fop_mknod_cbk_t cbk, void *local); + +extern int +nfs_readlink (xlator_t *nfsx, xlator_t *xl, nfs_user_t *nfu, loc_t *linkloc, + fop_readlink_cbk_t cbk, void *local); + +extern int +nfs_setattr (xlator_t *nfsx, xlator_t *xl, nfs_user_t *nfu, loc_t *pathloc, + struct iatt *buf, int32_t valid, fop_setattr_cbk_t cbk, + void *local); + +extern int +nfs_statfs (xlator_t *nfsx, xlator_t *xl, nfs_user_t *nfu, loc_t *pathloc, + fop_statfs_cbk_t cbk, void *local); + +extern int +nfs_stat (xlator_t *nfsx, xlator_t *xl, nfs_user_t *nfu, loc_t *pathloc, + fop_stat_cbk_t cbk, void *local); + +extern int +nfs_symlink (xlator_t *nfsx, xlator_t *xl, nfs_user_t *nfu, char *target, + loc_t *linkloc, fop_symlink_cbk_t cbk, void *local); + +/* Synchronous equivalents */ + +extern call_stub_t * +nfs_open_sync (xlator_t *xl, nfs_user_t *nfu, loc_t *pathloc, + int32_t flags); + +extern call_stub_t * +nfs_write_sync (xlator_t *xl, nfs_user_t *nfu, fd_t *fd, struct iobuf *srciob, + struct iovec *vec, int count, off_t offset); + +extern call_stub_t * +nfs_read_sync (xlator_t *xl, nfs_user_t *nfu, fd_t *fd, size_t size, + off_t offset); + +extern int +nfs_opendir (xlator_t *nfsx, xlator_t *fopxl, nfs_user_t *nfu, loc_t *pathloc, + fop_opendir_cbk_t cbk, void *local); + +extern int +nfs_access (xlator_t *nfsx, xlator_t *xl, nfs_user_t *nfu, loc_t *pathloc, + int32_t accesstest, fop_access_cbk_t cbk, void *local); +extern int +nfs_lk (xlator_t *nfsx, xlator_t *xl, nfs_user_t *nfu, fd_t *fd, + int cmd, struct gf_flock *flock, fop_lk_cbk_t cbk, void *local); + +extern int +nfs_getxattr (xlator_t *nfsx, xlator_t *xl, nfs_user_t *nfu, loc_t *loc, + char *name, dict_t *xdata, fop_getxattr_cbk_t cbk, void *local); + +extern int +nfs_setxattr (xlator_t *nfsx, xlator_t *xl, nfs_user_t *nfu, + loc_t *loc, dict_t *dict, int32_t flags, dict_t *xdata, + fop_setxattr_cbk_t cbk, void *local); + +#endif diff --git a/xlators/nfs/server/src/nfs-inodes.c b/xlators/nfs/server/src/nfs-inodes.c new file mode 100644 index 000000000..63d5e8a19 --- /dev/null +++ b/xlators/nfs/server/src/nfs-inodes.c @@ -0,0 +1,608 @@ +/* + Copyright (c) 2010-2011 Gluster, Inc. <http://www.gluster.com> + This file is part of GlusterFS. + + This file is licensed to you under your choice of the GNU Lesser + General Public License, version 3 or any later version (LGPLv3 or + later), or the GNU General Public License, version 2 (GPLv2), in all + cases as published by the Free Software Foundation. +*/ + +#ifndef _CONFIG_H +#define _CONFIG_H +#include "config.h" +#endif + +#include "string.h" + +#include "inode.h" +#include "nfs.h" +#include "nfs-inodes.h" +#include "nfs-fops.h" +#include "xlator.h" + +#include <libgen.h> + +#define inodes_nfl_to_prog_data(nflocal, pcbk, fram) \ + do { \ + nflocal = fram->local; \ + fram->local = nflocal->proglocal; \ + *VOID(&pcbk) = nflocal->progcbk; \ + nfs_fop_local_wipe (nflocal->nfsx, nflocal); \ + } while (0) \ + +void +nfl_inodes_init (struct nfs_fop_local *nfl, inode_t *inode, inode_t *parent, + inode_t *newparent, const char *name, const char *newname) +{ + if (!nfl) + return; + + if (inode) + nfl->inode = inode_ref (inode); + + if (parent) + nfl->parent = inode_ref (parent); + + if (newparent) + nfl->newparent = inode_ref (newparent); + + if (name) + strncpy (nfl->path, name, NFS_NAME_MAX); + + if (newname) + strncpy (nfl->newpath, newname, NFS_NAME_MAX); + + return; +} + + +int32_t +nfs_inode_create_cbk (call_frame_t *frame, void *cookie, xlator_t *this, + int32_t op_ret, int32_t op_errno, fd_t *fd, inode_t *inode + , struct iatt *buf, struct iatt *preparent, + struct iatt *postparent, dict_t *xdata) +{ + struct nfs_fop_local *nfl = frame->local; + fop_create_cbk_t progcbk = NULL; + inode_t *linked_inode = NULL; + + if (op_ret == -1) + goto do_not_link; + + linked_inode = inode_link (inode, nfl->parent, nfl->path, buf); + +do_not_link: + /* NFS does not need it, upper layers should not expect the pointer to + * be a valid fd. + */ + fd_unref (fd); + + inodes_nfl_to_prog_data (nfl, progcbk, frame); + if (progcbk) + progcbk (frame, cookie, this, op_ret, op_errno, fd, inode, buf, + preparent, postparent, xdata); + + if (linked_inode) { + inode_lookup (linked_inode); + inode_unref (linked_inode); + } + + return 0; +} + + +int +nfs_inode_create (xlator_t *nfsx, xlator_t *xl, nfs_user_t *nfu, + loc_t *pathloc, int flags, int mode, fop_create_cbk_t cbk, + void *local) +{ + struct nfs_fop_local *nfl = NULL; + int ret = -EFAULT; + fd_t *newfd = NULL; + + if ((!nfsx) || (!xl) || (!pathloc) || (!nfu)) + return ret; + + nfs_fop_handle_local_init (NULL, nfsx, nfl, cbk, local, ret, err); + + newfd = fd_create (pathloc->inode, 0); + if (!newfd) { + gf_log (GF_NFS, GF_LOG_ERROR, "Failed to create new fd"); + ret = -ENOMEM; + goto wipe_nfl; + } + + /* The parent and base name will be needed to link the new inode + * into the inode table. + */ + nfl_inodes_init (nfl, pathloc->inode, pathloc->parent, NULL, + pathloc->name, NULL); + ret = nfs_fop_create (nfsx, xl, nfu, pathloc, flags, mode, newfd, + nfs_inode_create_cbk, nfl); +wipe_nfl: + if (ret < 0) + nfs_fop_local_wipe (xl, nfl); + +err: + return ret; +} + + +int32_t +nfs_inode_mkdir_cbk (call_frame_t *frame, void *cookie, xlator_t *this, + int32_t op_ret, int32_t op_errno, inode_t *inode, + struct iatt *buf, struct iatt *preparent, + struct iatt *postparent, dict_t *xdata) +{ + struct nfs_fop_local *nfl = frame->local; + fop_mkdir_cbk_t progcbk = NULL; + inode_t *linked_inode = NULL; + + if (op_ret == -1) + goto do_not_link; + + linked_inode = inode_link (inode, nfl->parent, nfl->path, buf); + +do_not_link: + inodes_nfl_to_prog_data (nfl, progcbk, frame); + if (progcbk) + progcbk (frame, cookie, this, op_ret, op_errno, inode, buf, + preparent, postparent, xdata); + + if (linked_inode) { + inode_lookup (linked_inode); + inode_unref (linked_inode); + } + + return 0; +} + + +int +nfs_inode_mkdir (xlator_t *nfsx, xlator_t *xl, nfs_user_t *nfu, loc_t *pathloc, + int mode, fop_mkdir_cbk_t cbk, void *local) +{ + struct nfs_fop_local *nfl = NULL; + int ret = -EFAULT; + + if ((!nfsx) || (!xl) || (!pathloc) || (!nfu)) + return ret; + + nfs_fop_handle_local_init (NULL, nfsx, nfl, cbk, local, ret, err); + nfl_inodes_init (nfl, pathloc->inode, pathloc->parent, NULL, + pathloc->name, NULL); + ret = nfs_fop_mkdir (nfsx, xl, nfu, pathloc, mode, nfs_inode_mkdir_cbk, + nfl); + if (ret < 0) + nfs_fop_local_wipe (nfsx, nfl); + +err: + return ret; +} + + +int32_t +nfs_inode_open_cbk (call_frame_t *frame, void *cookie, xlator_t *this, + int32_t op_ret, int32_t op_errno, fd_t *fd, dict_t *xdata) +{ + + struct nfs_fop_local *nfl = NULL; + fop_open_cbk_t progcbk = NULL; + + if ((op_ret == -1) && (fd)) + fd_unref (fd); + /* Not needed here since the fd is cached in higher layers and the bind + * must happen atomically when the fd gets added to the fd LRU. + */ +/* else + fd_bind (fd); +*/ + inodes_nfl_to_prog_data (nfl, progcbk, frame); + if (progcbk) + progcbk (frame, cookie, this, op_ret, op_errno, fd, xdata); + return 0; +} + + +int +nfs_inode_open (xlator_t *nfsx, xlator_t *xl, nfs_user_t *nfu, loc_t *loc, + int32_t flags, fop_open_cbk_t cbk, void *local) +{ + struct nfs_fop_local *nfl = NULL; + fd_t *newfd = NULL; + int ret = -EFAULT; + + if ((!nfsx) || (!xl) || (!loc) || (!nfu)) + return ret; + + newfd = fd_create (loc->inode, 0); + if (!newfd) { + gf_log (GF_NFS, GF_LOG_ERROR, "Failed to create fd"); + ret = -ENOMEM; + goto err; + } + + nfs_fop_handle_local_init (NULL, nfsx, nfl, cbk, local, ret, fd_err); + ret = nfs_fop_open (nfsx, xl, nfu, loc, flags, newfd, + nfs_inode_open_cbk, nfl); + + if (ret < 0) + nfs_fop_local_wipe (xl, nfl); + +fd_err: + if (ret < 0) + if (newfd) + fd_unref (newfd); + +err: + + return ret; +} + + + +int32_t +nfs_inode_rename_cbk (call_frame_t *frame, void *cookie, xlator_t *this, + int32_t op_ret, int32_t op_errno, struct iatt *buf, + struct iatt *preoldparent, struct iatt *postoldparent, + struct iatt *prenewparent, struct iatt *postnewparent, + dict_t *xdata) +{ + struct nfs_fop_local *nfl = NULL; + fop_rename_cbk_t progcbk = NULL; + + nfl = frame->local; + if (op_ret == -1) + goto do_not_link; + + inode_rename (this->itable, nfl->parent, nfl->path, nfl->newparent, + nfl->newpath, nfl->inode, buf); + +do_not_link: + inodes_nfl_to_prog_data (nfl, progcbk, frame); + if (progcbk) + progcbk (frame, cookie, this, op_ret, op_errno, buf, + preoldparent, postoldparent, prenewparent, + postnewparent, xdata); + return 0; +} + + +int +nfs_inode_rename (xlator_t *nfsx, xlator_t *xl, nfs_user_t *nfu, loc_t *oldloc, + loc_t *newloc, fop_rename_cbk_t cbk, void *local) +{ + struct nfs_fop_local *nfl = NULL; + int ret = -EFAULT; + + if ((!nfsx) || (!xl) || (!oldloc) || (!newloc)) + return ret; + + nfs_fop_handle_local_init (NULL, nfsx, nfl, cbk, local, ret, err); + nfl_inodes_init (nfl, oldloc->inode, oldloc->parent, newloc->parent, + oldloc->name, newloc->name); + ret = nfs_fop_rename (nfsx,xl, nfu, oldloc, newloc, nfs_inode_rename_cbk + , nfl); + +err: + if (ret < 0) + nfs_fop_local_wipe (xl, nfl); + + return ret; +} + + +int32_t +nfs_inode_link_cbk (call_frame_t *frame, void *cookie, xlator_t *this, + int32_t op_ret, int32_t op_errno, inode_t *inode, + struct iatt *buf, struct iatt *preparent, + struct iatt *postparent, dict_t *xdata) +{ + struct nfs_fop_local *nfl = NULL; + fop_link_cbk_t progcbk = NULL; + inode_t *linked_inode = NULL; + + if (op_ret == -1) + goto do_not_link; + + nfl = frame->local; + linked_inode = inode_link (inode, nfl->newparent, nfl->path, buf); + +do_not_link: + inodes_nfl_to_prog_data (nfl, progcbk, frame); + if (progcbk) + progcbk (frame, cookie, this, op_ret, op_errno, inode, buf, + preparent, postparent, xdata); + + if (linked_inode) { + inode_lookup (linked_inode); + inode_unref (linked_inode); + } + + return 0; +} + + +int +nfs_inode_link (xlator_t *nfsx, xlator_t *xl, nfs_user_t *nfu, loc_t *oldloc, + loc_t *newloc, fop_link_cbk_t cbk, void *local) +{ + struct nfs_fop_local *nfl = NULL; + int ret = -EFAULT; + + if ((!nfsx) || (!xl) || (!oldloc) || (!newloc) || (!nfu)) + return -EFAULT; + + nfs_fop_handle_local_init (NULL, nfsx, nfl, cbk, local, ret, err); + nfl_inodes_init (nfl, NULL, NULL, newloc->parent, newloc->name, NULL); + ret = nfs_fop_link (nfsx, xl, nfu, oldloc, newloc, nfs_inode_link_cbk, + nfl); + +err: + if (ret < 0) + nfs_fop_local_wipe (xl, nfl); + + return ret; +} + + +int32_t +nfs_inode_unlink_cbk (call_frame_t *frame, void *cookie, xlator_t *this, + int32_t op_ret, int32_t op_errno, struct iatt *preparent, + struct iatt *postparent, dict_t *xdata) +{ + struct nfs_fop_local *nfl = NULL; + fop_unlink_cbk_t progcbk = NULL; + + nfl = frame->local; + + if (op_ret == -1) + goto do_not_unlink; + + inode_unlink (nfl->inode, nfl->parent, nfl->path); + inode_forget (nfl->inode, 0); + +do_not_unlink: + inodes_nfl_to_prog_data (nfl, progcbk, frame); + if (progcbk) + progcbk (frame, cookie, this, op_ret, op_errno, preparent, + postparent, xdata); + return 0; +} + + +int +nfs_inode_unlink (xlator_t *nfsx, xlator_t *xl, nfs_user_t *nfu, loc_t *pathloc, + fop_unlink_cbk_t cbk, void *local) +{ + struct nfs_fop_local *nfl = NULL; + int ret = -EFAULT; + + if ((!nfsx) || (!xl) || (!pathloc) || (!nfu)) + return -EFAULT; + + nfs_fop_handle_local_init (NULL, nfsx, nfl, cbk, local, ret, err); + nfl_inodes_init (nfl, pathloc->inode, pathloc->parent, NULL, + pathloc->name, NULL); + ret = nfs_fop_unlink (nfsx, xl, nfu, pathloc, nfs_inode_unlink_cbk,nfl); + +err: + if (ret < 0) + nfs_fop_local_wipe (xl, nfl); + + return ret; +} + + +int32_t +nfs_inode_rmdir_cbk (call_frame_t *frame, void *cookie, xlator_t *this, + int32_t op_ret, int32_t op_errno, struct iatt *preparent, + struct iatt *postparent, dict_t *xdata) +{ + struct nfs_fop_local *nfl = NULL; + fop_rmdir_cbk_t progcbk = NULL; + + nfl = frame->local; + + if (op_ret == -1) + goto do_not_unlink; + + inode_unlink (nfl->inode, nfl->parent, nfl->path); + inode_forget (nfl->inode, 0); + +do_not_unlink: + inodes_nfl_to_prog_data (nfl, progcbk, frame); + if (progcbk) + progcbk (frame, cookie, this, op_ret, op_errno, preparent, + postparent, xdata); + + return 0; +} + + +int +nfs_inode_rmdir (xlator_t *nfsx, xlator_t *xl, nfs_user_t *nfu, loc_t *pathloc, + fop_rmdir_cbk_t cbk, void *local) +{ + struct nfs_fop_local *nfl = NULL; + int ret = -EFAULT; + + if ((!nfsx) || (!xl) || (!pathloc) || (!nfu)) + return ret; + + nfs_fop_handle_local_init (NULL, nfsx, nfl, cbk, local, ret, err); + nfl_inodes_init (nfl, pathloc->inode, pathloc->parent, NULL, + pathloc->name, NULL); + + ret = nfs_fop_rmdir (nfsx, xl, nfu, pathloc, nfs_inode_rmdir_cbk, nfl); + +err: + if (ret < 0) + nfs_fop_local_wipe (xl, nfl); + return ret; +} + + +int32_t +nfs_inode_mknod_cbk (call_frame_t *frame, void *cookie, xlator_t *this, + int32_t op_ret, int32_t op_errno, inode_t *inode, + struct iatt *buf, struct iatt *preparent, + struct iatt *postparent, dict_t *xdata) +{ + struct nfs_fop_local *nfl = NULL; + fop_mknod_cbk_t progcbk = NULL; + inode_t *linked_inode = NULL; + + nfl = frame->local; + + if (op_ret == -1) + goto do_not_link; + + linked_inode = inode_link (inode, nfl->parent, nfl->path, buf); + +do_not_link: + inodes_nfl_to_prog_data (nfl, progcbk, frame); + if (progcbk) + progcbk (frame, cookie, this, op_ret, op_errno, inode, buf, + preparent, postparent, xdata); + + if (linked_inode) { + inode_lookup (linked_inode); + inode_unref (linked_inode); + } + + return 0; +} + + +int +nfs_inode_mknod (xlator_t *nfsx, xlator_t *xl, nfs_user_t *nfu, loc_t *pathloc, + mode_t mode, dev_t dev, fop_mknod_cbk_t cbk, void *local) +{ + struct nfs_fop_local *nfl = NULL; + int ret = -EFAULT; + + if ((!nfsx) || (!xl) || (!pathloc) || (!nfu)) + return ret; + + nfs_fop_handle_local_init (NULL, nfsx, nfl, cbk, local, ret, err); + nfl_inodes_init (nfl, pathloc->inode, pathloc->parent, NULL, + pathloc->name, NULL); + + ret = nfs_fop_mknod (nfsx, xl, nfu, pathloc, mode, dev, + nfs_inode_mknod_cbk, nfl); + +err: + if (ret < 0) + nfs_fop_local_wipe (xl, nfl); + + return ret; +} + + +int32_t +nfs_inode_symlink_cbk (call_frame_t *frame, void *cookie, xlator_t *this, + int32_t op_ret, int32_t op_errno, inode_t *inode, + struct iatt *buf, struct iatt *preparent, + struct iatt *postparent, dict_t *xdata) +{ + struct nfs_fop_local *nfl = NULL; + fop_symlink_cbk_t progcbk = NULL; + inode_t *linked_inode = NULL; + + nfl = frame->local; + if (op_ret == -1) + goto do_not_link; + + linked_inode = inode_link (inode, nfl->parent, nfl->path, buf); + +do_not_link: + inodes_nfl_to_prog_data (nfl, progcbk, frame); + if (progcbk) + progcbk (frame, cookie, this, op_ret, op_errno, inode, buf, + preparent, postparent, xdata); + + if (linked_inode) { + inode_lookup (linked_inode); + inode_unref (linked_inode); + } + + return 0; +} + + +int +nfs_inode_symlink (xlator_t *nfsx, xlator_t *xl, nfs_user_t *nfu, char *target, + loc_t *pathloc, fop_symlink_cbk_t cbk, void *local) +{ + struct nfs_fop_local *nfl = NULL; + int ret = -EFAULT; + + if ((!nfsx) || (!xl) || (!target) || (!pathloc) || (!nfu)) + return ret; + + nfs_fop_handle_local_init (NULL, nfsx, nfl, cbk, local, ret, err); + nfl_inodes_init (nfl, pathloc->inode, pathloc->parent, NULL, + pathloc->name, NULL); + ret = nfs_fop_symlink (nfsx, xl, nfu, target, pathloc, + nfs_inode_symlink_cbk, nfl); + +err: + if (ret < 0) + nfs_fop_local_wipe (xl, nfl); + + return ret; +} + +int32_t +nfs_inode_opendir_cbk (call_frame_t *frame, void *cookie, xlator_t *this, + int32_t op_ret, int32_t op_errno, fd_t *fd, dict_t *xdata) +{ + + struct nfs_fop_local *nfl = NULL; + fop_open_cbk_t progcbk = NULL; + + if (op_ret != -1) + fd_bind (fd); + + inodes_nfl_to_prog_data (nfl, progcbk, frame); + + if (progcbk) + progcbk (frame, cookie, this, op_ret, op_errno, fd, xdata); + + return 0; +} + + +int +nfs_inode_opendir (xlator_t *nfsx, xlator_t *xl, nfs_user_t *nfu, loc_t *loc, + fop_opendir_cbk_t cbk, void *local) +{ + struct nfs_fop_local *nfl = NULL; + fd_t *newfd = NULL; + int ret = -EFAULT; + + if ((!nfsx) || (!xl) || (!loc) || (!nfu)) + return ret; + + newfd = fd_create (loc->inode, 0); + if (!newfd) { + gf_log (GF_NFS, GF_LOG_ERROR, "Failed to create fd"); + ret = -ENOMEM; + goto err; + } + + nfs_fop_handle_local_init (NULL, nfsx, nfl, cbk, local, ret, err); + ret = nfs_fop_opendir (nfsx, xl, nfu, loc, newfd, + nfs_inode_opendir_cbk, nfl); + +err: + if (ret < 0) { + if (newfd) + fd_unref (newfd); + nfs_fop_local_wipe (xl, nfl); + } + + return ret; +} diff --git a/xlators/nfs/server/src/nfs-inodes.h b/xlators/nfs/server/src/nfs-inodes.h new file mode 100644 index 000000000..ba7a57124 --- /dev/null +++ b/xlators/nfs/server/src/nfs-inodes.h @@ -0,0 +1,76 @@ +/* + Copyright (c) 2010-2011 Gluster, Inc. <http://www.gluster.com> + This file is part of GlusterFS. + + This file is licensed to you under your choice of the GNU Lesser + General Public License, version 3 or any later version (LGPLv3 or + later), or the GNU General Public License, version 2 (GPLv2), in all + cases as published by the Free Software Foundation. +*/ + +#ifndef _NFS_INODES_H_ +#define _NFS_INODES_H_ + +#ifndef _CONFIG_H +#define _CONFIG_H +#include "config.h" +#endif + +#include "dict.h" +#include "xlator.h" +#include "iobuf.h" +#include "call-stub.h" +#include "stack.h" +#include "nfs-fops.h" + + +extern int +nfs_link_inode (inode_t *newi, inode_t *parent, char *name, + struct iatt *newstat); + +extern int +nfs_inode_create (xlator_t *nfsx, xlator_t *xl, nfs_user_t *nfu, + loc_t *pathloc, int flags, int mode, fop_create_cbk_t cbk, + void *local); + +extern int +nfs_inode_mkdir (xlator_t *nfsx, xlator_t *xl, nfs_user_t *nfu, loc_t *pathloc, + int mode, fop_mkdir_cbk_t cbk, void *local); + +extern int +nfs_inode_open (xlator_t *nfsx, xlator_t *xl, nfs_user_t *nfu, loc_t *loc, + int32_t flags, fop_open_cbk_t cbk, + void *local); + +extern int +nfs_inode_rename (xlator_t *nfsx, xlator_t *xl, nfs_user_t *nfu, loc_t *oldloc, + loc_t *newloc, fop_rename_cbk_t cbk, void *local); + +extern int +nfs_inode_link (xlator_t *nfsx, xlator_t *xl, nfs_user_t *nfu, loc_t *oldloc, + loc_t *newloc, fop_link_cbk_t cbk, void *local); + +extern int +nfs_inode_unlink (xlator_t *nfsx, xlator_t *xl, nfs_user_t *nfu, loc_t *pathloc, + fop_unlink_cbk_t cbk, void *local); + +extern int +nfs_inode_rmdir (xlator_t *nfsx, xlator_t *xl, nfs_user_t *nfu, loc_t *pathloc, + fop_rmdir_cbk_t cbk, void *local); + +extern int +nfs_inode_symlink (xlator_t *nfsx, xlator_t *xl, nfs_user_t *nfu, char *target, + loc_t *pathloc, fop_symlink_cbk_t cbk, void *local); + +extern int +nfs_inode_opendir (xlator_t *nfsx, xlator_t *xl, nfs_user_t *nfu, loc_t *loc, + fop_opendir_cbk_t cbk, void *local); + +extern int +nfs_inode_mknod (xlator_t *nfsx, xlator_t *xl, nfs_user_t *nfu, loc_t *pathloc, + mode_t mode, dev_t dev, fop_mknod_cbk_t cbk, void *local); + +extern int +nfs_inode_lookup (xlator_t *nfsx, xlator_t *xl, nfs_user_t *nfu, loc_t *pathloc, + fop_lookup_cbk_t cbk, void *local); +#endif diff --git a/xlators/nfs/server/src/nfs-mem-types.h b/xlators/nfs/server/src/nfs-mem-types.h new file mode 100644 index 000000000..450b6f2fe --- /dev/null +++ b/xlators/nfs/server/src/nfs-mem-types.h @@ -0,0 +1,53 @@ +/* + Copyright (c) 2008-2011 Gluster, Inc. <http://www.gluster.com> + This file is part of GlusterFS. + + This file is licensed to you under your choice of the GNU Lesser + General Public License, version 3 or any later version (LGPLv3 or + later), or the GNU General Public License, version 2 (GPLv2), in all + cases as published by the Free Software Foundation. +*/ + + +#ifndef __NFS_MEM_TYPES_H__ +#define __NFS_MEM_TYPES_H__ + +#include "mem-types.h" + +enum gf_nfs_mem_types_ { + gf_nfs_mt_mountentry = gf_common_mt_end + 1, + gf_nfs_mt_mountbody, + gf_nfs_mt_nfs_state, + gf_nfs_mt_char, + gf_nfs_mt_exportnode, + gf_nfs_mt_groupnode, + gf_nfs_mt_mount3_state, + gf_nfs_mt_write3args, + gf_nfs_mt_nfs3_export, + gf_nfs_mt_nfs3_state, + gf_nfs_mt_entry3, + gf_nfs_mt_entryp3, + gf_nfs_mt_nfs3_fd_entry, + gf_nfs_mt_nfs3_fh, + gf_nfs_mt_nfs_initer_list, + gf_nfs_mt_xlator_t, + gf_nfs_mt_list_head, + gf_nfs_mt_mnt3_resolve, + gf_nfs_mt_mnt3_export, + gf_nfs_mt_int, + gf_nfs_mt_mountres3, + gf_nfs_mt_mountstat3, + gf_nfs_mt_inode_q, + gf_nfs_mt_nlm4_state, + gf_nfs_mt_nlm4_cm, + gf_nfs_mt_nlm4_fde, + gf_nfs_mt_nlm4_nlmclnt, + gf_nfs_mt_nlm4_share, + gf_nfs_mt_aux_gids, + gf_nfs_mt_inode_ctx, + gf_nfs_mt_auth_spec, + gf_nfs_mt_arr, + gf_nfs_mt_end +}; +#endif + diff --git a/xlators/nfs/server/src/nfs.c b/xlators/nfs/server/src/nfs.c new file mode 100644 index 000000000..75c8fe44e --- /dev/null +++ b/xlators/nfs/server/src/nfs.c @@ -0,0 +1,1825 @@ +/* + Copyright (c) 2010-2011 Gluster, Inc. <http://www.gluster.com> + This file is part of GlusterFS. + + This file is licensed to you under your choice of the GNU Lesser + General Public License, version 3 or any later version (LGPLv3 or + later), or the GNU General Public License, version 2 (GPLv2), in all + cases as published by the Free Software Foundation. +*/ + +/* This is the primary translator source for NFS. + * Every other protocol version gets initialized from here. + */ + + +#ifndef _CONFIG_H +#define _CONFIG_H +#include "config.h" +#endif + +#include "defaults.h" +#include "rpcsvc.h" +#include "dict.h" +#include "xlator.h" +#include "nfs.h" +#include "mem-pool.h" +#include "logging.h" +#include "nfs-fops.h" +#include "inode.h" +#include "mount3.h" +#include "nfs3.h" +#include "nfs-mem-types.h" +#include "nfs3-helpers.h" +#include "nlm4.h" +#include "options.h" +#include "acl3.h" +#include "rpc-drc.h" + +#define STRINGIFY(val) #val +#define TOSTRING(val) STRINGIFY(val) + +#define OPT_SERVER_AUX_GIDS "nfs.server-aux-gids" +#define OPT_SERVER_GID_CACHE_TIMEOUT "nfs.server.aux-gid-timeout" + +/* TODO: DATADIR should be based on configure's $(localstatedir) */ +#define DATADIR "/var/lib/glusterd" +#define NFS_DATADIR DATADIR "/nfs" + +/* Forward declaration */ +int nfs_add_initer (struct list_head *list, nfs_version_initer_t init); + +static int +nfs_init_version (xlator_t *this, nfs_version_initer_t init) +{ + int ret = -1; + struct nfs_initer_list *version = NULL; + struct nfs_initer_list *tmp = NULL; + rpcsvc_program_t *prog = NULL; + struct list_head *versions = NULL; + struct nfs_state *nfs = NULL; + gf_boolean_t found = _gf_false; + + if ((!this) || (!this->private) || (!init)) + return (-1); + + nfs = (struct nfs_state *)this->private; + + ret = nfs_add_initer (&nfs->versions, init); + if (ret == -1) { + gf_log (GF_NFS, GF_LOG_ERROR, + "Failed to add protocol initializer"); + goto err; + } + + versions = &nfs->versions; + list_for_each_entry_safe (version, tmp, versions, list) { + prog = version->program; + if (version->init == init) { + prog = init(this); + if (!prog) { + ret = -1; + goto err; + } + version->program = prog; + found = _gf_true; + break; + } + } + + /* program not added */ + if (!found) { + gf_log (GF_NFS, GF_LOG_ERROR, + "Program: %s NOT found", prog->progname); + goto err; + } + + /* Check if nfs.port is configured */ + if (nfs->override_portnum) + prog->progport = nfs->override_portnum; + + gf_log (GF_NFS, GF_LOG_DEBUG, "Starting program: %s", prog->progname); + + ret = rpcsvc_program_register (nfs->rpcsvc, prog); + if (ret == -1) { + gf_log (GF_NFS, GF_LOG_ERROR, "Program: %s init failed", + prog->progname); + goto err; + } + + /* Registration with portmapper is disabled, Nothing to do */ + if (!nfs->register_portmap) + goto err; + + ret = rpcsvc_program_register_portmap (prog, prog->progport); + if (ret == -1) { + gf_log (GF_NFS, GF_LOG_ERROR, + "Program %s registration failed", + prog->progname); + goto err; + } + ret = 0; /* All well */ +err: + return ret; +} + +static int +nfs_deinit_version (struct nfs_state *nfs, nfs_version_initer_t init) +{ + int ret = -1; + struct nfs_initer_list *version = NULL; + struct nfs_initer_list *tmp = NULL; + rpcsvc_program_t *prog = NULL; + struct list_head *versions = NULL; + + if ((!nfs) || (!init)) + return (-1); + + versions = &nfs->versions; + list_for_each_entry_safe (version, tmp, versions, list) { + prog = version->program; + if (version->init == init) { + prog = version->program; + ret = rpcsvc_program_unregister (nfs->rpcsvc, prog); + if (ret != 0) + return (-1); + list_del (&version->list); + GF_FREE (version); + return (0); + } + } + + return (-1); +} + +static int +nfs_reconfigure_acl3 (xlator_t *this) +{ + struct nfs_state *nfs = NULL; + + if ((!this) || (!this->private)) + return (-1); + + nfs = (struct nfs_state *)this->private; + + /* ACL is enabled */ + if (nfs->enable_acl) + return nfs_init_version (this, acl3svc_init); + + /* ACL is disabled */ + return nfs_deinit_version (nfs, acl3svc_init); +} + +static int +nfs_reconfigure_nlm4 (xlator_t *this) +{ + struct nfs_state *nfs = NULL; + + if ((!this) || (!this->private)) + return (-1); + + nfs = (struct nfs_state *)this->private; + + /* NLM is enabled */ + if (nfs->enable_nlm) + return nfs_init_version (this, nlm4svc_init); + + /* NLM is disabled */ + return nfs_deinit_version (nfs, nlm4svc_init); +} + +static int +nfs_program_register_portmap_all (struct nfs_state *nfs) +{ + struct list_head *versions = NULL; + struct nfs_initer_list *version = NULL; + struct nfs_initer_list *tmp = NULL; + rpcsvc_program_t *prog = NULL; + + if (nfs == NULL) + return (-1); + + versions = &nfs->versions; + list_for_each_entry_safe (version, tmp, versions, list) { + prog = version->program; + if (prog == NULL) + continue; + if (nfs->override_portnum) + prog->progport = nfs->override_portnum; + (void) rpcsvc_program_register_portmap (prog, prog->progport); + } + + return (0); +} + +static int +nfs_program_unregister_portmap_all (struct nfs_state *nfs) +{ + struct list_head *versions = NULL; + struct nfs_initer_list *version = NULL; + struct nfs_initer_list *tmp = NULL; + rpcsvc_program_t *prog = NULL; + + if (nfs == NULL) + return (-1); + + versions = &nfs->versions; + list_for_each_entry_safe (version, tmp, versions, list) { + prog = version->program; + if (prog == NULL) + continue; + (void) rpcsvc_program_unregister_portmap (prog); + } + + return (0); +} + +/* Every NFS version must call this function with the init function + * for its particular version. + */ +int +nfs_add_initer (struct list_head *list, nfs_version_initer_t init) +{ + struct nfs_initer_list *new = NULL; + if ((!list) || (!init)) + return -1; + + new = GF_CALLOC (1, sizeof (*new), gf_nfs_mt_nfs_initer_list); + if (!new) { + gf_log (GF_NFS, GF_LOG_ERROR, "Memory allocation failed"); + return -1; + } + + new->init = init; + list_add_tail (&new->list, list); + return 0; +} + + +int +nfs_deinit_versions (struct list_head *versions, xlator_t *this) +{ + struct nfs_initer_list *version = NULL; + struct nfs_initer_list *tmp = NULL; + struct nfs_state *nfs = NULL; + + if ((!versions) || (!this)) + return -1; + + nfs = (struct nfs_state *)this->private; + list_for_each_entry_safe (version, tmp, versions, list) { + /* TODO: Add version specific destructor. + * if (!version->deinit) + goto err; + + version->deinit (this); + */ + if (version->program) + rpcsvc_program_unregister (nfs->rpcsvc, + (version->program)); + + list_del (&version->list); + GF_FREE (version); + } + + return 0; +} + +int +nfs_init_versions (struct nfs_state *nfs, xlator_t *this) +{ + struct nfs_initer_list *version = NULL; + struct nfs_initer_list *tmp = NULL; + rpcsvc_program_t *prog = NULL; + int ret = -1; + struct list_head *versions = NULL; + + if ((!nfs) || (!this)) + return -1; + + gf_log (GF_NFS, GF_LOG_DEBUG, "Initing protocol versions"); + versions = &nfs->versions; + list_for_each_entry_safe (version, tmp, versions, list) { + if (!version->init) { + ret = -1; + goto err; + } + + prog = version->init (this); + if (!prog) { + ret = -1; + goto err; + } + + version->program = prog; + if (nfs->override_portnum) + prog->progport = nfs->override_portnum; + gf_log (GF_NFS, GF_LOG_DEBUG, "Starting program: %s", + prog->progname); + + ret = rpcsvc_program_register (nfs->rpcsvc, prog); + if (ret == -1) { + gf_log (GF_NFS, GF_LOG_ERROR, "Program: %s init failed", + prog->progname); + goto err; + } + if (nfs->register_portmap) { + ret = rpcsvc_program_register_portmap (prog, + prog->progport); + if (ret == -1) { + gf_log (GF_NFS, GF_LOG_ERROR, + "Program %s registration failed", + prog->progname); + goto err; + } + } + + } + + ret = 0; +err: + return ret; +} + + +int +nfs_add_all_initiators (struct nfs_state *nfs) +{ + int ret = 0; + + /* Add the initializers for all versions. */ + ret = nfs_add_initer (&nfs->versions, mnt3svc_init); + if (ret == -1) { + gf_log (GF_NFS, GF_LOG_ERROR, "Failed to add " + "MOUNT3 protocol initializer"); + goto ret; + } + + ret = nfs_add_initer (&nfs->versions, mnt1svc_init); + if (ret == -1) { + gf_log (GF_NFS, GF_LOG_ERROR, "Failed to add " + "MOUNT1 protocol initializer"); + goto ret; + } + + ret = nfs_add_initer (&nfs->versions, nfs3svc_init); + if (ret == -1) { + gf_log (GF_NFS, GF_LOG_ERROR, "Failed to add " + "NFS3 protocol initializer"); + goto ret; + } + + if (nfs->enable_nlm == _gf_true) { + ret = nfs_add_initer (&nfs->versions, nlm4svc_init); + if (ret == -1) { + gf_log (GF_NFS, GF_LOG_ERROR, "Failed to add protocol" + " initializer"); + goto ret; + } + } + + if (nfs->enable_acl == _gf_true) { + ret = nfs_add_initer (&nfs->versions, acl3svc_init); + if (ret == -1) { + gf_log (GF_NFS, GF_LOG_ERROR, "Failed to add " + "ACL protocol initializer"); + goto ret; + } + } + + ret = 0; +ret: + return ret; +} + + +int +nfs_subvolume_started (struct nfs_state *nfs, xlator_t *xl) +{ + int x = 0; + int started = 0; + + if ((!nfs) || (!xl)) + return 1; + + LOCK (&nfs->svinitlock); + { + for (;x < nfs->allsubvols; ++x) { + if (nfs->initedxl[x] == xl) { + started = 1; + goto unlock; + } + } + } +unlock: + UNLOCK (&nfs->svinitlock); + + return started; +} + + +int +nfs_subvolume_set_started (struct nfs_state *nfs, xlator_t *xl) +{ + int x = 0; + + if ((!nfs) || (!xl)) + return 1; + + LOCK (&nfs->svinitlock); + { + for (;x < nfs->allsubvols; ++x) { + if (nfs->initedxl[x] == xl) { + gf_log (GF_NFS, GF_LOG_DEBUG, + "Volume already started %s", + xl->name); + break; + } + + if (nfs->initedxl[x] == NULL) { + nfs->initedxl[x] = xl; + ++nfs->upsubvols; + gf_log (GF_NFS, GF_LOG_DEBUG, "Starting up: %s " + ", vols started till now: %d", xl->name, + nfs->upsubvols); + goto unlock; + } + } + } +unlock: + UNLOCK (&nfs->svinitlock); + + return 0; +} + + +int32_t +nfs_start_subvol_lookup_cbk (call_frame_t *frame, void *cookie, + xlator_t *this, int32_t op_ret, int32_t op_errno, + inode_t *inode, struct iatt *buf, dict_t *xattr, + struct iatt *postparent) +{ + if (op_ret == -1) { + gf_log (GF_NFS, GF_LOG_CRITICAL, "Failed to lookup root: %s", + strerror (op_errno)); + goto err; + } + + nfs_subvolume_set_started (this->private, ((xlator_t *)cookie)); + gf_log (GF_NFS, GF_LOG_TRACE, "Started %s", ((xlator_t *)cookie)->name); +err: + return 0; +} + + +int +nfs_startup_subvolume (xlator_t *nfsx, xlator_t *xl) +{ + int ret = -1; + loc_t rootloc = {0, }; + nfs_user_t nfu = {0, }; + + if ((!nfsx) || (!xl)) + return -1; + + if (nfs_subvolume_started (nfsx->private, xl)) { + gf_log (GF_NFS,GF_LOG_TRACE, "Subvolume already started: %s", + xl->name); + ret = 0; + goto err; + } + + ret = nfs_root_loc_fill (xl->itable, &rootloc); + if (ret == -1) { + gf_log (GF_NFS, GF_LOG_CRITICAL, "Failed to init root loc"); + goto err; + } + + nfs_user_root_create (&nfu); + ret = nfs_fop_lookup (nfsx, xl, &nfu, &rootloc, + nfs_start_subvol_lookup_cbk, + (void *)nfsx->private); + if (ret < 0) { + gf_log (GF_NFS, GF_LOG_CRITICAL, "Failed to lookup root: %s", + strerror (-ret)); + goto err; + } + + nfs_loc_wipe (&rootloc); + +err: + return ret; +} + +int +nfs_startup_subvolumes (xlator_t *nfsx) +{ + int ret = -1; + xlator_list_t *cl = NULL; + struct nfs_state *nfs = NULL; + + if (!nfsx) + return -1; + + nfs = nfsx->private; + cl = nfs->subvols; + while (cl) { + gf_log (GF_NFS, GF_LOG_DEBUG, "Starting subvolume: %s", + cl->xlator->name); + ret = nfs_startup_subvolume (nfsx, cl->xlator); + if (ret == -1) { + gf_log (GF_NFS, GF_LOG_CRITICAL, "Failed to start-up " + "xlator: %s", cl->xlator->name); + goto err; + } + cl = cl->next; + } + + ret = 0; +err: + return ret; +} + + +int +nfs_init_subvolume (struct nfs_state *nfs, xlator_t *xl) +{ + unsigned int lrusize = 0; + int ret = -1; + + if ((!nfs) || (!xl)) + return -1; + + lrusize = nfs->memfactor * GF_NFS_INODE_LRU_MULT; + xl->itable = inode_table_new (lrusize, xl); + if (!xl->itable) { + gf_log (GF_NFS, GF_LOG_CRITICAL, "Failed to allocate " + "inode table"); + goto err; + } + ret = 0; +err: + return ret; +} + +int +nfs_init_subvolumes (struct nfs_state *nfs, xlator_list_t *cl) +{ + int ret = -1; + unsigned int lrusize = 0; + int svcount = 0; + + if ((!nfs) || (!cl)) + return -1; + + lrusize = nfs->memfactor * GF_NFS_INODE_LRU_MULT; + nfs->subvols = cl; + gf_log (GF_NFS, GF_LOG_TRACE, "inode table lru: %d", lrusize); + + while (cl) { + gf_log (GF_NFS, GF_LOG_DEBUG, "Initing subvolume: %s", + cl->xlator->name); + ret = nfs_init_subvolume (nfs, cl->xlator); + if (ret == -1) { + gf_log (GF_NFS, GF_LOG_CRITICAL, "Failed to init " + "xlator: %s", cl->xlator->name); + goto err; + } + ++svcount; + cl = cl->next; + } + + LOCK_INIT (&nfs->svinitlock); + nfs->initedxl = GF_CALLOC (svcount, sizeof (xlator_t *), + gf_nfs_mt_xlator_t ); + if (!nfs->initedxl) { + gf_log (GF_NFS, GF_LOG_ERROR, "Failed to allocated inited xls"); + ret = -1; + goto err; + } + + gf_log (GF_NFS, GF_LOG_TRACE, "Inited volumes: %d", svcount); + nfs->allsubvols = svcount; + ret = 0; +err: + return ret; +} + + +int +nfs_user_root_create (nfs_user_t *newnfu) +{ + if (!newnfu) + return -1; + + newnfu->uid = 0; + newnfu->gids[0] = 0; + newnfu->ngrps = 1; + + return 0; +} + + +int +nfs_user_create (nfs_user_t *newnfu, uid_t uid, gid_t gid, gid_t *auxgids, + int auxcount) +{ + int x = 1; + int y = 0; + + /* We test for GF_REQUEST_MAXGROUPS instead of NFS_FOP_NGROUPS because + * the latter accounts for the @gid being in @auxgids, which is not the + * case here. + */ + if ((!newnfu) || (auxcount > GF_REQUEST_MAXGROUPS)) + return -1; + + newnfu->uid = uid; + newnfu->gids[0] = gid; + newnfu->ngrps = 1; + + gf_log (GF_NFS, GF_LOG_TRACE, "uid: %d, gid %d, gids: %d", uid, gid, + auxcount); + + if (!auxgids) + return 0; + + for (; y < auxcount; ++x,++y) { + newnfu->gids[x] = auxgids[y]; + ++newnfu->ngrps; + gf_log (GF_NFS, GF_LOG_TRACE, "gid: %d", auxgids[y]); + } + + return 0; +} + + +void +nfs_request_user_init (nfs_user_t *nfu, rpcsvc_request_t *req) +{ + gid_t *gidarr = NULL; + int gids = 0; + + if ((!req) || (!nfu)) + return; + + gidarr = rpcsvc_auth_unix_auxgids (req, &gids); + nfs_user_create (nfu, rpcsvc_request_uid (req), + rpcsvc_request_gid (req), gidarr, gids); + + return; +} + +void +nfs_request_primary_user_init (nfs_user_t *nfu, rpcsvc_request_t *req, + uid_t uid, gid_t gid) +{ + gid_t *gidarr = NULL; + int gids = 0; + + if ((!req) || (!nfu)) + return; + + gidarr = rpcsvc_auth_unix_auxgids (req, &gids); + nfs_user_create (nfu, uid, gid, gidarr, gids); + + return; +} + + +int32_t +mem_acct_init (xlator_t *this) +{ + int ret = -1; + + if (!this) + return ret; + + ret = xlator_mem_acct_init (this, gf_nfs_mt_end + 1); + + if (ret != 0) { + gf_log(this->name, GF_LOG_ERROR, "Memory accounting init" + "failed"); + return ret; + } + + return ret; +} + + +struct nfs_state * +nfs_init_state (xlator_t *this) +{ + struct nfs_state *nfs = NULL; + int ret = -1; + unsigned int fopspoolsize = 0; + char *optstr = NULL; + gf_boolean_t boolt = _gf_false; + struct stat stbuf = {0,}; + + if (!this) + return NULL; + + if (!this->children) { + gf_log (GF_NFS, GF_LOG_INFO, + "NFS is manually disabled: Exiting"); + /* Nothing for nfs process to do, exit cleanly */ + kill (getpid (), SIGTERM); + } + + nfs = GF_CALLOC (1, sizeof (*nfs), gf_nfs_mt_nfs_state); + if (!nfs) { + gf_log (GF_NFS, GF_LOG_ERROR, "memory allocation failed"); + return NULL; + } + + nfs->memfactor = GF_NFS_DEFAULT_MEMFACTOR; + if (dict_get (this->options, "nfs.mem-factor")) { + ret = dict_get_str (this->options, "nfs.mem-factor", + &optstr); + if (ret < 0) { + gf_log (GF_NFS, GF_LOG_ERROR, "Failed to parse dict"); + goto free_rpcsvc; + } + + ret = gf_string2uint (optstr, &nfs->memfactor); + if (ret < 0) { + gf_log (GF_NFS, GF_LOG_ERROR, "Failed to parse uint " + "string"); + goto free_rpcsvc; + } + } + + fopspoolsize = nfs->memfactor * GF_NFS_CONCURRENT_OPS_MULT; + /* FIXME: Really saddens me to see this as xlator wide. */ + nfs->foppool = mem_pool_new (struct nfs_fop_local, fopspoolsize); + if (!nfs->foppool) { + gf_log (GF_NFS, GF_LOG_CRITICAL, "Failed to allocate fops " + "local pool"); + goto free_rpcsvc; + } + + nfs->dynamicvolumes = GF_NFS_DVM_OFF; + if (dict_get (this->options, "nfs.dynamic-volumes")) { + ret = dict_get_str (this->options, "nfs.dynamic-volumes", + &optstr); + if (ret < 0) { + gf_log (GF_NFS, GF_LOG_ERROR, "Failed to parse dict"); + goto free_foppool; + } + + ret = gf_string2boolean (optstr, &boolt); + if (ret < 0) { + gf_log (GF_NFS, GF_LOG_ERROR, "Failed to parse bool " + "string"); + goto free_foppool; + } + + if (boolt == _gf_true) + nfs->dynamicvolumes = GF_NFS_DVM_ON; + } + + nfs->enable_nlm = _gf_true; + ret = dict_get_str_boolean (this->options, "nfs.nlm", _gf_true); + if (ret == _gf_false) { + gf_log (GF_NFS, GF_LOG_INFO, "NLM is manually disabled"); + nfs->enable_nlm = _gf_false; + } + + nfs->enable_acl = _gf_true; + ret = dict_get_str_boolean (this->options, "nfs.acl", _gf_true); + if (ret == _gf_false) { + gf_log (GF_NFS, GF_LOG_INFO, "ACL is manually disabled"); + nfs->enable_acl = _gf_false; + } + + nfs->enable_ino32 = 0; + if (dict_get (this->options, "nfs.enable-ino32")) { + ret = dict_get_str (this->options, "nfs.enable-ino32", + &optstr); + if (ret < 0) { + gf_log (GF_NFS, GF_LOG_ERROR, "Failed to parse dict"); + goto free_foppool; + } + + ret = gf_string2boolean (optstr, &boolt); + if (ret < 0) { + gf_log (GF_NFS, GF_LOG_ERROR, "Failed to parse bool " + "string"); + goto free_foppool; + } + + if (boolt == _gf_true) + nfs->enable_ino32 = 1; + } + + if (dict_get (this->options, "nfs.port")) { + ret = dict_get_str (this->options, "nfs.port", + &optstr); + if (ret < 0) { + gf_log (GF_NFS, GF_LOG_ERROR, "Failed to parse dict"); + goto free_foppool; + } + + ret = gf_string2uint (optstr, &nfs->override_portnum); + if (ret < 0) { + gf_log (GF_NFS, GF_LOG_ERROR, "Failed to parse uint " + "string"); + goto free_foppool; + } + } + + if (dict_get(this->options, "transport.socket.listen-port") == NULL) { + if (nfs->override_portnum) + ret = gf_asprintf (&optstr, "%d", + nfs->override_portnum); + else + ret = gf_asprintf (&optstr, "%d", GF_NFS3_PORT); + if (ret == -1) { + gf_log (GF_NFS, GF_LOG_ERROR, "failed mem-allocation"); + goto free_foppool; + } + ret = dict_set_dynstr (this->options, + "transport.socket.listen-port", optstr); + if (ret == -1) { + gf_log (GF_NFS, GF_LOG_ERROR, "dict_set_dynstr error"); + goto free_foppool; + } + } + + if (dict_get(this->options, "transport-type") == NULL) { + ret = dict_set_str (this->options, "transport-type", "socket"); + if (ret == -1) { + gf_log (GF_NFS, GF_LOG_ERROR, "dict_set_str error"); + goto free_foppool; + } + } + + nfs->mount_udp = 0; + if (dict_get(this->options, "nfs.mount-udp")) { + ret = dict_get_str (this->options, "nfs.mount-udp", &optstr); + if (ret == -1) { + gf_log (GF_NFS, GF_LOG_ERROR, "Failed to parse dict"); + goto free_foppool; + } + + ret = gf_string2boolean (optstr, &boolt); + if (ret < 0) { + gf_log (GF_NFS, GF_LOG_ERROR, "Failed to parse bool " + "string"); + goto free_foppool; + } + + if (boolt == _gf_true) + nfs->mount_udp = 1; + } + + nfs->rmtab = gf_strdup (NFS_DATADIR "/rmtab"); + if (dict_get(this->options, "nfs.mount-rmtab")) { + ret = dict_get_str (this->options, "nfs.mount-rmtab", &nfs->rmtab); + if (ret == -1) { + gf_log (GF_NFS, GF_LOG_ERROR, "Failed to parse dict"); + goto free_foppool; + } + } + + /* support both options rpc-auth.ports.insecure and + * rpc-auth-allow-insecure for backward compatibility + */ + nfs->allow_insecure = 1; + if (dict_get(this->options, "rpc-auth.ports.insecure")) { + ret = dict_get_str (this->options, "rpc-auth.ports.insecure", + &optstr); + if (ret < 0) { + gf_log (GF_NFS, GF_LOG_ERROR, "Failed to parse dict"); + goto free_foppool; + } + + ret = gf_string2boolean (optstr, &boolt); + if (ret < 0) { + gf_log (GF_NFS, GF_LOG_ERROR, "Failed to parse bool " + "string"); + goto free_foppool; + } + + if (boolt == _gf_false) + nfs->allow_insecure = 0; + } + + if (dict_get(this->options, "rpc-auth-allow-insecure")) { + ret = dict_get_str (this->options, "rpc-auth-allow-insecure", + &optstr); + if (ret < 0) { + gf_log (GF_NFS, GF_LOG_ERROR, "Failed to parse dict"); + goto free_foppool; + } + + ret = gf_string2boolean (optstr, &boolt); + if (ret < 0) { + gf_log (GF_NFS, GF_LOG_ERROR, "Failed to parse bool " + "string"); + goto free_foppool; + } + + if (boolt == _gf_false) + nfs->allow_insecure = 0; + } + + if (nfs->allow_insecure) { + /* blindly set both the options */ + dict_del (this->options, "rpc-auth-allow-insecure"); + ret = dict_set_str (this->options, + "rpc-auth-allow-insecure", "on"); + if (ret == -1) { + gf_log (GF_NFS, GF_LOG_ERROR, "dict_set_str error"); + goto free_foppool; + } + dict_del (this->options, "rpc-auth.ports.insecure"); + ret = dict_set_str (this->options, + "rpc-auth.ports.insecure", "on"); + if (ret == -1) { + gf_log (GF_NFS, GF_LOG_ERROR, "dict_set_str error"); + goto free_foppool; + } + } + + GF_OPTION_INIT (OPT_SERVER_AUX_GIDS, nfs->server_aux_gids, + bool, free_foppool); + GF_OPTION_INIT (OPT_SERVER_GID_CACHE_TIMEOUT, nfs->server_aux_gids_max_age, + uint32, free_foppool); + + if (gid_cache_init(&nfs->gid_cache, nfs->server_aux_gids_max_age) < 0) { + gf_log(GF_NFS, GF_LOG_ERROR, "Failed to initialize group cache."); + goto free_foppool; + } + + if (stat("/sbin/rpc.statd", &stbuf) == -1) { + gf_log (GF_NFS, GF_LOG_WARNING, "/sbin/rpc.statd not found. " + "Disabling NLM"); + nfs->enable_nlm = _gf_false; + } + + nfs->rpcsvc = rpcsvc_init (this, this->ctx, this->options, 0); + if (!nfs->rpcsvc) { + ret = -1; + gf_log (GF_NFS, GF_LOG_ERROR, "RPC service init failed"); + goto free_foppool; + } + + nfs->register_portmap = rpcsvc_register_portmap_enabled (nfs->rpcsvc); + + this->private = (void *)nfs; + INIT_LIST_HEAD (&nfs->versions); + nfs->generation = 1965; + + ret = 0; + +free_foppool: + if (ret < 0) + mem_pool_destroy (nfs->foppool); + +free_rpcsvc: + /* + * rpcsvc_deinit */ + if (ret < 0) { + GF_FREE (nfs); + nfs = NULL; + } + + return nfs; +} + +int +nfs_drc_init (xlator_t *this) +{ + int ret = -1; + rpcsvc_t *svc = NULL; + + svc = ((struct nfs_state *)(this->private))->rpcsvc; + if (!svc) + goto out; + + ret = rpcsvc_drc_init (svc, this->options); + + out: + return ret; +} + +int +nfs_reconfigure_state (xlator_t *this, dict_t *options) +{ + int ret = 0; + int keyindx = 0; + char *optstr = NULL; + gf_boolean_t optbool; + uint32_t optuint32; + struct nfs_state *nfs = NULL; + char *blacklist_keys[] = { + "nfs.port", + "nfs.transport-type", + "nfs.mem-factor", + NULL}; + + GF_VALIDATE_OR_GOTO (GF_NFS, this, out); + GF_VALIDATE_OR_GOTO (GF_NFS, this->private, out); + GF_VALIDATE_OR_GOTO (GF_NFS, options, out); + + nfs = (struct nfs_state *)this->private; + + /* Black listed options can't be reconfigured, they need + * NFS to be restarted. There are two cases 1. SET 2. UNSET. + * 1. SET */ + while (blacklist_keys[keyindx]) { + if (dict_get (options, blacklist_keys[keyindx])) { + gf_log (GF_NFS, GF_LOG_ERROR, + "Reconfiguring %s needs NFS restart", + blacklist_keys[keyindx]); + goto out; + } + keyindx ++; + } + + /* UNSET for nfs.mem-factor */ + if ((!dict_get (options, "nfs.mem-factor")) && + (nfs->memfactor != GF_NFS_DEFAULT_MEMFACTOR)) { + gf_log (GF_NFS, GF_LOG_INFO, + "Reconfiguring nfs.mem-factor needs NFS restart"); + goto out; + } + + /* UNSET for nfs.port */ + if ((!dict_get (options, "nfs.port")) && + (nfs->override_portnum)) { + gf_log (GF_NFS, GF_LOG_ERROR, + "Reconfiguring nfs.port needs NFS restart"); + goto out; + } + + /* reconfig nfs.mount-rmtab */ + optstr = NFS_DATADIR "/rmtab"; + if (dict_get (options, "nfs.mount-rmtab")) { + ret = dict_get_str (options, "nfs.mount-rmtab", &optstr); + if (ret < 0) { + gf_log (GF_NFS, GF_LOG_ERROR, "Failed to read " + "reconfigured option: nfs.mount-rmtab"); + goto out; + } + gf_path_strip_trailing_slashes (optstr); + } + if (strcmp (nfs->rmtab, optstr) != 0) { + mount_rewrite_rmtab (nfs->mstate, optstr); + gf_log (GF_NFS, GF_LOG_INFO, + "Reconfigured nfs.mount-rmtab path: %s", + nfs->rmtab); + } + + GF_OPTION_RECONF (OPT_SERVER_AUX_GIDS, optbool, + options, bool, out); + if (nfs->server_aux_gids != optbool) { + nfs->server_aux_gids = optbool; + gf_log(GF_NFS, GF_LOG_INFO, "Reconfigured %s with value %d", + OPT_SERVER_AUX_GIDS, optbool); + } + + GF_OPTION_RECONF (OPT_SERVER_GID_CACHE_TIMEOUT, optuint32, + options, uint32, out); + if (nfs->server_aux_gids_max_age != optuint32) { + nfs->server_aux_gids_max_age = optuint32; + gid_cache_reconf (&nfs->gid_cache, optuint32); + gf_log(GF_NFS, GF_LOG_INFO, "Reconfigured %s with value %d", + OPT_SERVER_GID_CACHE_TIMEOUT, optuint32); + } + + /* reconfig nfs.dynamic-volumes */ + ret = dict_get_str_boolean (options, "nfs.dynamic-volumes", + GF_NFS_DVM_OFF); + switch (ret) { + case GF_NFS_DVM_ON: + case GF_NFS_DVM_OFF: + optbool = ret; + break; + default: + optbool = GF_NFS_DVM_OFF; + break; + } + if (nfs->dynamicvolumes != optbool) { + nfs->dynamicvolumes = optbool; + gf_log(GF_NFS, GF_LOG_INFO, "Reconfigured nfs.dynamic-volumes" + " with value %d", optbool); + } + + optbool = _gf_false; + if (dict_get (options, "nfs.enable-ino32")) { + ret = dict_get_str_boolean (options, "nfs.enable-ino32", + _gf_false); + if (ret < 0) { + gf_log (GF_NFS, GF_LOG_ERROR, + "Failed to read reconfigured option: " + "nfs.enable-ino32"); + goto out; + } + optbool = ret; + } + if (nfs->enable_ino32 != optbool) { + nfs->enable_ino32 = optbool; + gf_log(GF_NFS, GF_LOG_INFO, "Reconfigured nfs.enable-ino32" + " with value %d", optbool); + } + + /* nfs.nlm is enabled by default */ + ret = dict_get_str_boolean (options, "nfs.nlm", _gf_true); + if (ret < 0) { + optbool = _gf_true; + } else { + optbool = ret; + } + if (nfs->enable_nlm != optbool) { + gf_log (GF_NFS, GF_LOG_INFO, "NLM is manually %s", + (optbool ? "enabled":"disabled")); + nfs->enable_nlm = optbool; + nfs_reconfigure_nlm4 (this); + } + + /* nfs.acl is enabled by default */ + ret = dict_get_str_boolean (options, "nfs.acl", _gf_true); + if (ret < 0) { + optbool = _gf_true; + } else { + optbool = ret; + } + if (nfs->enable_acl != optbool) { + gf_log (GF_NFS, GF_LOG_INFO, "ACL is manually %s", + (optbool ? "enabled":"disabled")); + nfs->enable_acl = optbool; + nfs_reconfigure_acl3 (this); + } + + ret = 0; +out: + return ret; +} + + +/* + * reconfigure() for NFS server xlator. + */ +int +reconfigure (xlator_t *this, dict_t *options) +{ + int ret = 0; + struct nfs_state *nfs = NULL; + gf_boolean_t regpmap = _gf_true; + + if ((!this) || (!this->private) || (!options)) + return (-1); + + nfs = (struct nfs_state *)this->private; + + /* Reconfigure nfs options */ + ret = nfs_reconfigure_state(this, options); + if (ret) { + gf_log (GF_NFS, GF_LOG_ERROR, + "nfs reconfigure state failed"); + return (-1); + } + + /* Reconfigure nfs3 options */ + ret = nfs3_reconfigure_state(this, options); + if (ret) { + gf_log (GF_NFS, GF_LOG_ERROR, + "nfs3 reconfigure state failed"); + return (-1); + } + + /* Reconfigure mount options */ + ret = mount_reconfigure_state(this, options); + if (ret) { + gf_log (GF_NFS, GF_LOG_ERROR, + "mount reconfigure state failed"); + return (-1); + } + + /* Reconfigure rpc layer */ + ret = rpcsvc_reconfigure_options (nfs->rpcsvc, options); + if (ret) { + gf_log (GF_NFS, GF_LOG_ERROR, + "rpcsvc reconfigure options failed"); + return (-1); + } + regpmap = rpcsvc_register_portmap_enabled(nfs->rpcsvc); + if (nfs->register_portmap != regpmap) { + nfs->register_portmap = regpmap; + if (regpmap) { + (void) nfs_program_register_portmap_all (nfs); + } else { + (void) nfs_program_unregister_portmap_all (nfs); + } + } + + /* Reconfigure drc */ + ret = rpcsvc_drc_reconfigure (nfs->rpcsvc, options); + if (ret) { + gf_log (GF_NFS, GF_LOG_ERROR, + "rpcsvc DRC reconfigure failed"); + return (-1); + } + + return (0); +} + +int +init (xlator_t *this) { + + struct nfs_state *nfs = NULL; + int ret = -1; + + if (!this) + return -1; + + nfs = nfs_init_state (this); + if (!nfs) { + gf_log (GF_NFS, GF_LOG_ERROR, "Failed to init nfs option"); + return -1; + } + + ret = nfs_add_all_initiators (nfs); + if (ret == -1) { + gf_log (GF_NFS, GF_LOG_ERROR, "Failed to add initiators"); + goto err; + } + + ret = nfs_init_subvolumes (nfs, this->children); + if (ret == -1) { + gf_log (GF_NFS, GF_LOG_CRITICAL, "Failed to init NFS " + "exports"); + goto err; + } + + ret = mount_init_state (this); + if (ret == -1) { + gf_log (GF_NFS, GF_LOG_CRITICAL, "Failed to init Mount" + "state"); + goto err; + } + + ret = nlm4_init_state (this); + if (ret == -1) { + gf_log (GF_NFS, GF_LOG_CRITICAL, "Failed to init NLM" + "state"); + goto err; + } + + ret = nfs_init_versions (nfs, this); + if (ret == -1) { + gf_log (GF_NFS, GF_LOG_ERROR, "Failed to initialize " + "protocols"); + /* Do not return an error on this. If we dont return + * an error, the process keeps running and it helps + * to point out where the log is by doing ps ax|grep gluster. + */ + ret = 0; + goto err; + } + + ret = nfs_drc_init (this); + if (ret == 0) + gf_log (GF_NFS, GF_LOG_INFO, "NFS service started"); +err: + + return ret; +} + + +int +notify (xlator_t *this, int32_t event, void *data, ...) +{ + xlator_t *subvol = NULL; + struct nfs_state *priv = NULL; + + subvol = (xlator_t *)data; + + gf_log (GF_NFS, GF_LOG_TRACE, "Notification received: %d", + event); + + switch (event) { + case GF_EVENT_CHILD_UP: + nfs_startup_subvolume (this, subvol); + break; + + case GF_EVENT_CHILD_MODIFIED: + priv = this->private; + ++(priv->generation); + break; + + case GF_EVENT_PARENT_UP: + default_notify (this, GF_EVENT_PARENT_UP, data); + break; + } + + return 0; +} + + +int +fini (xlator_t *this) +{ + + struct nfs_state *nfs = NULL; + + nfs = (struct nfs_state *)this->private; + gf_log (GF_NFS, GF_LOG_DEBUG, "NFS service going down"); + nfs_deinit_versions (&nfs->versions, this); + return 0; +} + +int32_t +nfs_forget (xlator_t *this, inode_t *inode) +{ + uint64_t ctx = 0; + struct nfs_inode_ctx *ictx = NULL; + + if (inode_ctx_del (inode, this, &ctx)) + return -1; + + ictx = (struct nfs_inode_ctx *)ctx; + GF_FREE (ictx); + + return 0; +} + +gf_boolean_t +_nfs_export_is_for_vol (char *exname, char *volname) +{ + gf_boolean_t ret = _gf_false; + char *tmp = NULL; + + tmp = exname; + if (tmp[0] == '/') + tmp++; + + if (!strcmp (tmp, volname)) + ret = _gf_true; + + return ret; +} + +int +nfs_priv_to_dict (xlator_t *this, dict_t *dict) +{ + int ret = -1; + struct nfs_state *priv = NULL; + struct mountentry *mentry = NULL; + char *volname = NULL; + char key[1024] = {0,}; + int count = 0; + + GF_VALIDATE_OR_GOTO (THIS->name, this, out); + GF_VALIDATE_OR_GOTO (THIS->name, dict, out); + + priv = this->private; + GF_ASSERT (priv); + + ret = dict_get_str (dict, "volname", &volname); + if (ret) { + gf_log (this->name, GF_LOG_ERROR, "Could not get volname"); + goto out; + } + + list_for_each_entry (mentry, &priv->mstate->mountlist, mlist) { + if (!_nfs_export_is_for_vol (mentry->exname, volname)) + continue; + + memset (key, 0, sizeof (key)); + snprintf (key, sizeof (key), "client%d.hostname", count); + ret = dict_set_str (dict, key, mentry->hostname); + if (ret) { + gf_log (this->name, GF_LOG_ERROR, + "Error writing hostname to dict"); + goto out; + } + + /* No connection data available yet in nfs server. + * Hence, setting to 0 to prevent cli failing + */ + memset (key, 0, sizeof (key)); + snprintf (key, sizeof (key), "client%d.bytesread", count); + ret = dict_set_uint64 (dict, key, 0); + if (ret) { + gf_log (this->name, GF_LOG_ERROR, + "Error writing bytes read to dict"); + goto out; + } + + memset (key, 0, sizeof (key)); + snprintf (key, sizeof (key), "client%d.byteswrite", count); + ret = dict_set_uint64 (dict, key, 0); + if (ret) { + gf_log (this->name, GF_LOG_ERROR, + "Error writing bytes write to dict"); + goto out; + } + + count++; + } + + ret = dict_set_int32 (dict, "clientcount", count); + if (ret) + gf_log (this->name, GF_LOG_ERROR, + "Error writing client count to dict"); + +out: + gf_log (THIS->name, GF_LOG_DEBUG, "Returning %d", ret); + return ret; +} + +extern int32_t +nlm_priv (xlator_t *this); + +int32_t +nfs_priv (xlator_t *this) +{ + int32_t ret = -1; + + /* DRC needs the global drc structure, xl is of no use to it. */ + ret = rpcsvc_drc_priv (((struct nfs_state *)(this->private))->rpcsvc->drc); + if (ret) { + gf_log (this->name, GF_LOG_DEBUG, "Statedump of DRC failed"); + goto out; + } + + ret = nlm_priv (this); + if (ret) { + gf_log (this->name, GF_LOG_DEBUG, "Statedump of NLM failed"); + goto out; + } + out: + return ret; +} + + +struct xlator_cbks cbks = { + .forget = nfs_forget, +}; + +struct xlator_fops fops; + +struct xlator_dumpops dumpops = { + .priv = nfs_priv, + .priv_to_dict = nfs_priv_to_dict, +}; + +/* TODO: If needed, per-volume options below can be extended to be export ++ * specific also because after export-dir is introduced, a volume is not ++ * neccessarily an export whereas different subdirectories within that volume ++ * can be and may need these options to be specified separately. ++ */ +struct volume_options options[] = { + { .key = {"nfs3.read-size"}, + .type = GF_OPTION_TYPE_SIZET, + .min = GF_NFS3_RTMIN, + .max = GF_NFS3_RTMAX, + .default_value = TOSTRING(GF_NFS3_RTPREF), + .description = "Size in which the client should issue read requests " + "to the Gluster NFSv3 server. Must be a multiple of " + "4KB (4096). Min and Max supported values are 4KB " + "(4096) and 1MB (1048576) respectively. If the " + "specified value is within the supported range but " + "not a multiple of 4096, it is rounded up to the " + "nearest multiple of 4096." + }, + { .key = {"nfs3.write-size"}, + .type = GF_OPTION_TYPE_SIZET, + .min = GF_NFS3_WTMIN, + .max = GF_NFS3_WTMAX, + .default_value = TOSTRING(GF_NFS3_WTPREF), + .description = "Size in which the client should issue write requests " + "to the Gluster NFSv3 server. Must be a multiple of " + "1KB (1024). Min and Max supported values are " + "4KB (4096) and 1MB(1048576) respectively. If the " + "specified value is within the supported range but " + "not a multiple of 4096, it is rounded up to the " + "nearest multiple of 4096." + }, + { .key = {"nfs3.readdir-size"}, + .type = GF_OPTION_TYPE_SIZET, + .min = GF_NFS3_DTMIN, + .max = GF_NFS3_DTMAX, + .default_value = TOSTRING(GF_NFS3_DTPREF), + .description = "Size in which the client should issue directory " + "reading requests to the Gluster NFSv3 server. Must " + "be a multiple of 1KB (1024). Min and Max supported " + "values are 4KB (4096) and 1MB (1048576) respectively." + "If the specified value is within the supported range " + "but not a multiple of 4096, it is rounded up to the " + "nearest multiple of 4096." + }, + { .key = {"nfs3.*.volume-access"}, + .type = GF_OPTION_TYPE_STR, + .value = {"read-only", "read-write"}, + .default_value = "read-write", + .description = "Type of access desired for this subvolume: " + " read-only, read-write(default)" + }, + { .key = {"nfs3.*.trusted-write"}, + .type = GF_OPTION_TYPE_BOOL, + .default_value = "off", + .description = "On an UNSTABLE write from client, return STABLE flag" + " to force client to not send a COMMIT request. In " + "some environments, combined with a replicated " + "GlusterFS setup, this option can improve write " + "performance. This flag allows user to trust Gluster" + " replication logic to sync data to the disks and " + "recover when required. COMMIT requests if received " + "will be handled in a default manner by fsyncing." + " STABLE writes are still handled in a sync manner. " + "Off by default." + + }, + { .key = {"nfs3.*.trusted-sync"}, + .type = GF_OPTION_TYPE_BOOL, + .default_value = "off", + .description = "All writes and COMMIT requests are treated as async." + " This implies that no write requests are guaranteed" + " to be on server disks when the write reply is " + "received at the NFS client. Trusted sync includes " + " trusted-write behaviour. Off by default." + + }, + { .key = {"nfs3.*.export-dir"}, + .type = GF_OPTION_TYPE_PATH, + .default_value = "", + .description = "By default, all subvolumes of nfs are exported as " + "individual exports. There are cases where a " + "subdirectory or subdirectories in the volume need to " + "be exported separately. This option can also be used " + "in conjunction with nfs3.export-volumes option to " + "restrict exports only to the subdirectories specified" + " through this option. Must be an absolute path. Along" + " with path allowed list of IPs/hostname can be " + "associated with each subdirectory. If provided " + "connection will allowed only from these IPs. By " + "default connections from all IPs are allowed. " + "Format: <dir>[(hostspec[|hostspec|...])][,...]. Where" + " hostspec can be an IP address, hostname or an IP " + "range in CIDR notation. " + "e.g. /foo(192.168.1.0/24|host1|10.1.1.8),/host2." + " NOTE: Care must be taken while configuring this " + "option as invalid entries and/or unreachable DNS " + "servers can introduce unwanted delay in all the mount" + " calls." + }, + { .key = {"nfs3.export-dirs"}, + .type = GF_OPTION_TYPE_BOOL, + .default_value = "on", + .description = "By default, all subvolumes of nfs are exported as " + "individual exports. There are cases where a " + "subdirectory or subdirectories in the volume need to " + "be exported separately. Enabling this option allows " + "any directory on a volumes to be exported separately." + "Directory exports are enabled by default." + }, + { .key = {"nfs3.export-volumes"}, + .type = GF_OPTION_TYPE_BOOL, + .default_value = "on", + .description = "Enable or disable exporting whole volumes, instead " + "if used in conjunction with nfs3.export-dir, can " + "allow setting up only subdirectories as exports. On " + "by default." + }, + { .key = {"rpc-auth.auth-unix"}, + .type = GF_OPTION_TYPE_BOOL, + .default_value = "on", + .description = "Disable or enable the AUTH_UNIX authentication type." + "Must always be enabled for better interoperability." + "However, can be disabled if needed. Enabled by" + "default" + }, + { .key = {"rpc-auth.auth-null"}, + .type = GF_OPTION_TYPE_BOOL, + .default_value = "on", + .description = "Disable or enable the AUTH_NULL authentication type." + "Must always be enabled. This option is here only to" + " avoid unrecognized option warnings" + }, + { .key = {"rpc-auth.auth-unix.*"}, + .type = GF_OPTION_TYPE_BOOL, + .default_value = "on", + .description = "Disable or enable the AUTH_UNIX authentication type " + "for a particular exported volume overriding defaults" + " and general setting for AUTH_UNIX scheme. Must " + "always be enabled for better interoperability." + "However, can be disabled if needed. Enabled by" + "default." + }, + { .key = {"rpc-auth.auth-unix.*.allow"}, + .type = GF_OPTION_TYPE_STR, + .default_value = "on", + .description = "Disable or enable the AUTH_UNIX authentication type " + "for a particular exported volume overriding defaults" + " and general setting for AUTH_UNIX scheme. Must " + "always be enabled for better interoperability." + "However, can be disabled if needed. Enabled by" + "default." + }, + { .key = {"rpc-auth.auth-null.*"}, + .type = GF_OPTION_TYPE_BOOL, + .default_value = "on", + .description = "Disable or enable the AUTH_NULL authentication type " + "for a particular exported volume overriding defaults" + " and general setting for AUTH_NULL. Must always be " + "enabled. This option is here only to avoid " + "unrecognized option warnings." + }, + { .key = {"rpc-auth.addr.allow"}, + .type = GF_OPTION_TYPE_INTERNET_ADDRESS_LIST, + .default_value = "all", + .description = "Allow a comma separated list of addresses and/or" + " hostnames to connect to the server. By default, all" + " connections are allowed. This allows users to " + "define a general rule for all exported volumes." + }, + { .key = {"rpc-auth.addr.reject"}, + .type = GF_OPTION_TYPE_INTERNET_ADDRESS_LIST, + .default_value = "none", + .description = "Reject a comma separated list of addresses and/or" + " hostnames from connecting to the server. By default," + " all connections are allowed. This allows users to" + "define a general rule for all exported volumes." + }, + { .key = {"rpc-auth.addr.*.allow"}, + .type = GF_OPTION_TYPE_INTERNET_ADDRESS_LIST, + .default_value = "all", + .description = "Allow a comma separated list of addresses and/or" + " hostnames to connect to the server. By default, all" + " connections are allowed. This allows users to " + "define a rule for a specific exported volume." + }, + { .key = {"rpc-auth.addr.*.reject"}, + .type = GF_OPTION_TYPE_INTERNET_ADDRESS_LIST, + .default_value = "none", + .description = "Reject a comma separated list of addresses and/or" + " hostnames from connecting to the server. By default," + " all connections are allowed. This allows users to" + "define a rule for a specific exported volume." + }, + { .key = {"rpc-auth.ports.insecure"}, + .type = GF_OPTION_TYPE_BOOL, + .default_value = "off", + .description = "Allow client connections from unprivileged ports. By " + "default only privileged ports are allowed. This is a" + "global setting in case insecure ports are to be " + "enabled for all exports using a single option." + }, + { .key = {"rpc-auth.ports.*.insecure"}, + .type = GF_OPTION_TYPE_BOOL, + .default_value = "off", + .description = "Allow client connections from unprivileged ports. By " + "default only privileged ports are allowed. Use this" + " option to enable or disable insecure ports for " + "a specific subvolume and to override the global " + "setting set by the previous option." + }, + { .key = {"rpc-auth.addr.namelookup"}, + .type = GF_OPTION_TYPE_BOOL, + .default_value = "off", + .description = "Users have the option of turning on name lookup for" + " incoming client connections using this option. Use this " + "option to turn on name lookups during address-based " + "authentication. Turning this on will enable you to" + " use hostnames in nfs.rpc-auth-* filters. In some " + "setups, the name server can take too long to reply to DNS " + "queries resulting in timeouts of mount requests. By " + "default, name lookup is off" + }, + { .key = {"nfs.dynamic-volumes"}, + .type = GF_OPTION_TYPE_BOOL, + .default_value = "off", + .description = "Internal option set to tell gnfs to use a different" + " scheme for encoding file handles when DVM is being" + " used." + }, + { .key = {"nfs3.*.volume-id"}, + .type = GF_OPTION_TYPE_STR, + .default_value = "", + .description = "When nfs.dynamic-volumes is set, gnfs expects every " + "subvolume to have this option set for it, so that " + "gnfs can use this option to identify the volume. " + "If all subvolumes do not have this option set, an " + "error is reported." + }, + { .key = {"nfs.enable-ino32"}, + .type = GF_OPTION_TYPE_BOOL, + .default_value = "no", + .description = "For nfs clients or apps that do not support 64-bit " + "inode numbers, use this option to make NFS return " + "32-bit inode numbers instead. Disabled by default, so" + " NFS returns 64-bit inode numbers." + }, + { .key = {"rpc.register-with-portmap"}, + .type = GF_OPTION_TYPE_BOOL, + .default_value = "on", + .description = "For systems that need to run multiple nfs servers, " + "only one registration is possible with " + "portmap service. Use this option to turn off portmap " + "registration for Gluster NFS. On by default" + }, + { .key = {"rpc.outstanding-rpc-limit"}, + .type = GF_OPTION_TYPE_INT, + .min = RPCSVC_MIN_OUTSTANDING_RPC_LIMIT, + .max = RPCSVC_MAX_OUTSTANDING_RPC_LIMIT, + .default_value = TOSTRING(RPCSVC_DEFAULT_OUTSTANDING_RPC_LIMIT), + .description = "Parameter to throttle the number of incoming RPC " + "requests from a client. 0 means no limit (can " + "potentially run out of memory)" + }, + { .key = {"nfs.port"}, + .type = GF_OPTION_TYPE_INT, + .min = 1, + .max = 0xffff, + .default_value = TOSTRING(GF_NFS3_PORT), + .description = "Use this option on systems that need Gluster NFS to " + "be associated with a non-default port number." + }, + { .key = {"nfs.mem-factor"}, + .type = GF_OPTION_TYPE_INT, + .min = 1, + .max = 1024, + .default_value = TOSTRING(GF_NFS_DEFAULT_MEMFACTOR), + .description = "Use this option to make NFS be faster on systems by " + "using more memory. This option specifies a multiple " + "that determines the total amount of memory used. " + "Default value is 15. Increase to use more memory in " + "order to improve performance for certain use cases." + "Please consult gluster-users list before using this " + "option." + }, + { .key = {"nfs.*.disable"}, + .type = GF_OPTION_TYPE_BOOL, + .default_value = "false", + .description = "This option is used to start or stop NFS server" + "for individual volume." + }, + + { .key = {"nfs.nlm"}, + .type = GF_OPTION_TYPE_BOOL, + .default_value = "on", + .description = "This option, if set to 'off', disables NLM server " + "by not registering the service with the portmapper." + " Set it to 'on' to re-enable it. Default value: 'on'" + }, + + { .key = {"nfs.mount-udp"}, + .type = GF_OPTION_TYPE_BOOL, + .default_value = "off", + .description = "set the option to 'on' to enable mountd on UDP. " + "Required for some Solaris and AIX NFS clients. " + "The need for enabling this option often depends " + "on the usage of NLM." + }, + { .key = {"nfs.mount-rmtab"}, + .type = GF_OPTION_TYPE_PATH, + .default_value = DATADIR "/rmtab", + .description = "Set the location of the cache file that is used to " + "list all the NFS-clients that have connected " + "through the MOUNT protocol. If this is on shared " + "storage, all GlusterFS servers will update and " + "output (with 'showmount') the same list." + }, + { .key = {OPT_SERVER_AUX_GIDS}, + .type = GF_OPTION_TYPE_BOOL, + .default_value = "off", + .description = "Let the server look up which groups a user belongs " + "to, overwriting the list passed from the client. " + "This enables support for group lists longer than " + "can be passed through the NFS protocol, but is not " + "secure unless users and groups are well synchronized " + "between clients and servers." + }, + { .key = {OPT_SERVER_GID_CACHE_TIMEOUT}, + .type = GF_OPTION_TYPE_INT, + .min = 0, + .max = 3600, + .default_value = "5", + .description = "Number of seconds to cache auxiliary-GID data, when " + OPT_SERVER_AUX_GIDS " is set." + }, + { .key = {"nfs.acl"}, + .type = GF_OPTION_TYPE_BOOL, + .default_value = "on", + .description = "This option is used to control ACL support for NFS." + }, + { .key = {"nfs.drc"}, + .type = GF_OPTION_TYPE_STR, + .default_value = "on", + .description = "Enable Duplicate Request Cache in gNFS server to " + "improve correctness of non-idempotent operations like " + "write, delete, link, et al" + }, + { .key = {"nfs.drc-size"}, + .type = GF_OPTION_TYPE_INT, + .default_value = "0x20000", + .description = "Sets the number of non-idempotent " + "requests to cache in drc" + }, + { .key = {NULL} }, +}; diff --git a/xlators/nfs/server/src/nfs.h b/xlators/nfs/server/src/nfs.h new file mode 100644 index 000000000..00c7f8046 --- /dev/null +++ b/xlators/nfs/server/src/nfs.h @@ -0,0 +1,135 @@ +/* + Copyright (c) 2010-2011 Gluster, Inc. <http://www.gluster.com> + This file is part of GlusterFS. + + This file is licensed to you under your choice of the GNU Lesser + General Public License, version 3 or any later version (LGPLv3 or + later), or the GNU General Public License, version 2 (GPLv2), in all + cases as published by the Free Software Foundation. +*/ + +#ifndef __NFS_H__ +#define __NFS_H__ + +#ifndef _CONFIG_H +#define _CONFIG_H +#include "config.h" +#endif + +#include "rpcsvc.h" +#include "dict.h" +#include "xlator.h" +#include "lkowner.h" +#include "gidcache.h" + +#define GF_NFS "nfs" + +#define GF_NFS_CONCURRENT_OPS_MULT 15 + +#define GF_NFS_INODE_LRU_MULT 6000 + +#define GF_RPC_MIN_THREADS 1 +#define GF_RPC_MAX_THREADS 16 + +#define GF_NFS_DEFAULT_MEMFACTOR 15 +#define GF_NFS_MIN_MEMFACTOR 1 +#define GF_NFS_MAX_MEMFACTOR 30 + +#define GF_NFS_DVM_ON 1 +#define GF_NFS_DVM_OFF 0 + +/* This corresponds to the max 16 number of group IDs that are sent through an + * RPC request. Since NFS is the only one going to set this, we can be safe + * in keeping this size hardcoded. + */ +#define GF_REQUEST_MAXGROUPS 16 + +/* Callback into a version-specific NFS protocol. + * The return type is used by the nfs.c code to register the protocol. + * with the RPC service. + */ +typedef rpcsvc_program_t *(*nfs_version_initer_t) (xlator_t *nfsx); + +/* List of version-specific protocol initiators */ +struct nfs_initer_list { + struct list_head list; + nfs_version_initer_t init; + rpcsvc_program_t *program; +}; + +struct nfs_state { + rpcsvc_t *rpcsvc; + struct list_head versions; + struct mount3_state *mstate; + struct nfs3_state *nfs3state; + struct nlm4_state *nlm4state; + struct mem_pool *foppool; + unsigned int memfactor; + xlator_list_t *subvols; + + gf_lock_t svinitlock; + int allsubvols; + int upsubvols; + xlator_t **initedxl; + int subvols_started; + int dynamicvolumes; + int enable_ino32; + unsigned int override_portnum; + int allow_insecure; + int enable_nlm; + int enable_acl; + int mount_udp; + char *rmtab; + struct rpc_clnt *rpc_clnt; + gf_boolean_t server_aux_gids; + uint32_t server_aux_gids_max_age; + gid_cache_t gid_cache; + uint32_t generation; + gf_boolean_t register_portmap; +}; + +struct nfs_inode_ctx { + struct list_head shares; + uint32_t generation; +}; + +#define gf_nfs_dvm_on(nfsstt) (((struct nfs_state *)nfsstt)->dynamicvolumes == GF_NFS_DVM_ON) +#define gf_nfs_dvm_off(nfsstt) (((struct nfs_state *)nfsstt)->dynamicvolumes == GF_NFS_DVM_OFF) +#define __gf_nfs_enable_ino32(nfsstt) (((struct nfs_state *)nfsstt)->enable_ino32) +#define gf_nfs_this_private ((struct nfs_state *)((xlator_t *)THIS)->private) +#define gf_nfs_enable_ino32() (__gf_nfs_enable_ino32(gf_nfs_this_private)) + +/* We have one gid more than the glusterfs maximum since we pass the primary + * gid as the first element of the array. + */ +#define NFS_NGROUPS (GF_REQUEST_MAXGROUPS + 1) + +/* Index of the primary gid */ +#define NFS_PRIMGID_IDX 0 + +typedef struct nfs_user_info { + uid_t uid; + gid_t gids[NFS_NGROUPS]; + int ngrps; + gf_lkowner_t lk_owner; +} nfs_user_t; + +extern int +nfs_user_root_create (nfs_user_t *newnfu); + +extern int +nfs_user_create (nfs_user_t *newnfu, uid_t uid, gid_t gid, gid_t *auxgids, + int auxcount); + +extern void +nfs_request_user_init (nfs_user_t *nfu, rpcsvc_request_t *req); + +extern void +nfs_request_primary_user_init (nfs_user_t *nfu, rpcsvc_request_t *req, + uid_t uid, gid_t gid); +extern int +nfs_subvolume_started (struct nfs_state *nfs, xlator_t *xl); + +extern void +nfs_fix_groups (xlator_t *this, call_stack_t *root); +#endif diff --git a/xlators/nfs/server/src/nfs3-fh.c b/xlators/nfs/server/src/nfs3-fh.c new file mode 100644 index 000000000..e199c56dc --- /dev/null +++ b/xlators/nfs/server/src/nfs3-fh.c @@ -0,0 +1,188 @@ +/* + Copyright (c) 2010-2011 Gluster, Inc. <http://www.gluster.com> + This file is part of GlusterFS. + + This file is licensed to you under your choice of the GNU Lesser + General Public License, version 3 or any later version (LGPLv3 or + later), or the GNU General Public License, version 2 (GPLv2), in all + cases as published by the Free Software Foundation. +*/ + +#ifndef _CONFIG_H +#define _CONFIG_H +#include "config.h" +#endif + +#include "rpcsvc.h" +#include "dict.h" +#include "xlator.h" +#include "xdr-nfs3.h" +#include "msg-nfs3.h" +#include "iobuf.h" +#include "nfs3-fh.h" +#include "nfs-common.h" +#include "iatt.h" +#include "common-utils.h" + + +int +nfs3_fh_validate (struct nfs3_fh *fh) +{ + if (!fh) + return 0; + + if (fh->ident[0] != GF_NFSFH_IDENT0) + return 0; + + if (fh->ident[1] != GF_NFSFH_IDENT1) + return 0; + + if (fh->ident[2] != GF_NFSFH_IDENT2) + return 0; + + if (fh->ident[3] != GF_NFSFH_IDENT3) + return 0; + + return 1; +} + + +void +nfs3_fh_init (struct nfs3_fh *fh, struct iatt *buf) +{ + if ((!fh) || (!buf)) + return; + + fh->ident[0] = GF_NFSFH_IDENT0; + fh->ident[1] = GF_NFSFH_IDENT1; + fh->ident[2] = GF_NFSFH_IDENT2; + fh->ident[3] = GF_NFSFH_IDENT3; + + uuid_copy (fh->gfid, buf->ia_gfid); +} + + +struct nfs3_fh +nfs3_fh_build_indexed_root_fh (xlator_list_t *cl, xlator_t *xl) +{ + struct nfs3_fh fh = {{0}, }; + struct iatt buf = {0, }; + uuid_t root = {0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1}; + + if ((!cl) || (!xl)) + return fh; + + uuid_copy (buf.ia_gfid, root); + nfs3_fh_init (&fh, &buf); + fh.exportid [15] = nfs_xlator_to_xlid (cl, xl); + + return fh; +} + + +struct nfs3_fh +nfs3_fh_build_uuid_root_fh (uuid_t volumeid) +{ + struct nfs3_fh fh = {{0}, }; + struct iatt buf = {0, }; + uuid_t root = {0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1}; + + uuid_copy (buf.ia_gfid, root); + nfs3_fh_init (&fh, &buf); + uuid_copy (fh.exportid, volumeid); + + return fh; +} + + +int +nfs3_fh_is_root_fh (struct nfs3_fh *fh) +{ + uuid_t rootgfid = {0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1}; + + if (!fh) + return 0; + + if (uuid_compare (fh->gfid, rootgfid) == 0) + return 1; + + return 0; +} + + +void +nfs3_fh_to_str (struct nfs3_fh *fh, char *str, size_t len) +{ + char gfid[GF_UUID_BUF_SIZE]; + char exportid[GF_UUID_BUF_SIZE]; + + if ((!fh) || (!str)) + return; + + snprintf (str, len, "FH: exportid %s, gfid %s", + uuid_utoa_r (fh->exportid, exportid), + uuid_utoa_r (fh->gfid, gfid)); +} + +void +nfs3_log_fh (struct nfs3_fh *fh) +{ + char gfidstr[512]; + char exportidstr[512]; + + if (!fh) + return; + + gf_log ("nfs3-fh", GF_LOG_TRACE, "filehandle: exportid " + "0x%s, gfid 0x%s", + uuid_utoa_r (fh->exportid, exportidstr), + uuid_utoa_r (fh->gfid, gfidstr)); +} + +int +nfs3_fh_build_parent_fh (struct nfs3_fh *child, struct iatt *newstat, + struct nfs3_fh *newfh) +{ + if ((!child) || (!newstat) || (!newfh)) + return -1; + + nfs3_fh_init (newfh, newstat); + uuid_copy (newfh->exportid, child->exportid); + + return 0; +} + +int +nfs3_build_fh (inode_t *inode, uuid_t exportid, struct nfs3_fh *newfh) +{ + if (!newfh || !inode) + return -1; + + newfh->ident[0] = GF_NFSFH_IDENT0; + newfh->ident[1] = GF_NFSFH_IDENT1; + newfh->ident[2] = GF_NFSFH_IDENT2; + newfh->ident[3] = GF_NFSFH_IDENT3; + uuid_copy (newfh->gfid, inode->gfid); + uuid_copy (newfh->exportid, exportid); + return 0; +} + +int +nfs3_fh_build_child_fh (struct nfs3_fh *parent, struct iatt *newstat, + struct nfs3_fh *newfh) +{ + if ((!parent) || (!newstat) || (!newfh)) + return -1; + + nfs3_fh_init (newfh, newstat); + uuid_copy (newfh->exportid, parent->exportid); + + return 0; +} + + +uint32_t +nfs3_fh_compute_size (struct nfs3_fh *fh) +{ + return GF_NFSFH_STATIC_SIZE; +} diff --git a/xlators/nfs/server/src/nfs3-fh.h b/xlators/nfs/server/src/nfs3-fh.h new file mode 100644 index 000000000..1049cdc96 --- /dev/null +++ b/xlators/nfs/server/src/nfs3-fh.h @@ -0,0 +1,103 @@ +/* + Copyright (c) 2010-2011 Gluster, Inc. <http://www.gluster.com> + This file is part of GlusterFS. + + This file is licensed to you under your choice of the GNU Lesser + General Public License, version 3 or any later version (LGPLv3 or + later), or the GNU General Public License, version 2 (GPLv2), in all + cases as published by the Free Software Foundation. +*/ + +#ifndef _NFS_FH_H_ +#define _NFS_FH_H_ + +#ifndef _CONFIG_H +#define _CONFIG_H +#include "config.h" +#endif + +#include "xlator.h" +#include "xdr-nfs3.h" +#include "iatt.h" +#include <sys/types.h> +#include "uuid.h" + +/* BIG FAT WARNING: The file handle code is tightly coupled to NFSv3 file + * handles for now. This will change if and when we need v4. */ +#define GF_NFSFH_IDENT0 ':' +#define GF_NFSFH_IDENT1 'O' +#define GF_NFSFH_IDENT2 'G' +#define GF_NFSFH_IDENT3 'L' +#define GF_NFSFH_IDENT_SIZE (sizeof(char) * 4) +#define GF_NFSFH_STATIC_SIZE (GF_NFSFH_IDENT_SIZE + (2*sizeof (uuid_t))) + +#define nfs3_fh_exportid_to_index(exprtid) ((uint16_t)exprtid[15]) +/* ATTENTION: Change in size of the structure below should be reflected in the + * GF_NFSFH_STATIC_SIZE. + */ +struct nfs3_fh { + + /* Used to ensure that a bunch of bytes are actually a GlusterFS NFS + * file handle. Should contain ":OGL" + */ + char ident[4]; + + /* UUID that identifies an export. The value stored in exportid + * depends on the usage of gluster nfs. If the DVM is enabled using + * the nfs.dynamic-volumes option then exportid will contain the UUID + * of the volume so that gnfs is able to identify volumes uniquely + * through volume additions,deletions,migrations, etc. + * + * When not using dvm, exportid contains the index of the volume + * based on the position of the volume in the list of subvolumes + * for gnfs. + */ + uuid_t exportid; + + /* File/dir gfid. */ + uuid_t gfid; + /* This structure must be exactly NFS3_FHSIZE (64) bytes long. + Having the structure shorter results in buffer overflows + during XDR decoding. + */ + unsigned char padding[NFS3_FHSIZE - GF_NFSFH_STATIC_SIZE]; +} __attribute__((__packed__)); + +#define GF_NFS3FH_STATIC_INITIALIZER {{0},} + +extern uint32_t +nfs3_fh_compute_size (struct nfs3_fh *fh); + +extern uint16_t +nfs3_fh_hash_entry (uuid_t gfid); + +extern int +nfs3_fh_validate (struct nfs3_fh *fh); + +extern struct nfs3_fh +nfs3_fh_build_indexed_root_fh (xlator_list_t *cl, xlator_t *xl); + +extern int +nfs3_fh_is_root_fh (struct nfs3_fh *fh); + +extern int +nfs3_fh_build_child_fh (struct nfs3_fh *parent, struct iatt *newstat, + struct nfs3_fh *newfh); + +extern void +nfs3_log_fh (struct nfs3_fh *fh); + +extern void +nfs3_fh_to_str (struct nfs3_fh *fh, char *str, size_t len); + +extern int +nfs3_fh_build_parent_fh (struct nfs3_fh *child, struct iatt *newstat, + struct nfs3_fh *newfh); + +extern struct nfs3_fh +nfs3_fh_build_uuid_root_fh (uuid_t volumeid); + +extern int +nfs3_build_fh (inode_t *inode, uuid_t exportid, struct nfs3_fh *newfh); + +#endif diff --git a/xlators/nfs/server/src/nfs3-helpers.c b/xlators/nfs/server/src/nfs3-helpers.c new file mode 100644 index 000000000..9059fc341 --- /dev/null +++ b/xlators/nfs/server/src/nfs3-helpers.c @@ -0,0 +1,3881 @@ +/* + Copyright (c) 2010-2011 Gluster, Inc. <http://www.gluster.com> + This file is part of GlusterFS. + + This file is licensed to you under your choice of the GNU Lesser + General Public License, version 3 or any later version (LGPLv3 or + later), or the GNU General Public License, version 2 (GPLv2), in all + cases as published by the Free Software Foundation. +*/ + +#ifndef _CONFIG_H +#define _CONFIG_H +#include "config.h" +#endif + +#include <inttypes.h> + +#include "xlator.h" +#include "nfs3.h" +#include "nfs3-fh.h" +#include "msg-nfs3.h" +#include "rbthash.h" +#include "nfs-fops.h" +#include "nfs-inodes.h" +#include "nfs-generics.h" +#include "nfs3-helpers.h" +#include "nfs-mem-types.h" +#include "iatt.h" +#include "common-utils.h" +#include <string.h> + +extern int +nfs3_set_root_looked_up (struct nfs3_state *nfs3, struct nfs3_fh *rootfh); + +extern int +nfs3_is_root_looked_up (struct nfs3_state *nfs3, struct nfs3_fh *rootfh); + + +#define nfs3_call_resume(cst) \ + do { \ + if (((cst)) && (cst)->resume_fn) \ + (cst)->resume_fn (cst); \ + } while (0) \ + +#define nfs3_call_resume_estale(csta) \ + do { \ + (csta)->resolve_ret = -1; \ + (csta)->resolve_errno = ESTALE; \ + nfs3_call_resume (csta); \ + } while (0) \ + +struct nfs3stat_strerror { + nfsstat3 stat; + char strerror[100]; +}; + + +struct nfs3stat_strerror nfs3stat_strerror_table[] = { + { NFS3_OK, "Call completed successfully." }, + { NFS3ERR_PERM, "Not owner" }, + { NFS3ERR_NOENT, "No such file or directory" }, + { NFS3ERR_IO, "I/O error" }, + { NFS3ERR_NXIO, "I/O error" }, + { NFS3ERR_ACCES, "Permission denied" }, + { NFS3ERR_EXIST, "File exists" }, + { NFS3ERR_XDEV, "Attempt to do a cross-device hard link"}, + { NFS3ERR_NODEV, "No such device" }, + { NFS3ERR_NOTDIR, "Not a directory" }, + { NFS3ERR_ISDIR, "Is a directory" }, + { NFS3ERR_INVAL, "Invalid argument for operation" }, + { NFS3ERR_FBIG, "File too large" }, + { NFS3ERR_NOSPC, "No space left on device" }, + { NFS3ERR_ROFS, "Read-only file system" }, + { NFS3ERR_MLINK, "Too many hard links" }, + { NFS3ERR_NAMETOOLONG, "Filename in operation was too long" }, + { NFS3ERR_NOTEMPTY, "Directory not empty" }, + { NFS3ERR_DQUOT, "Resource (quota) hard limit exceeded" }, + { NFS3ERR_STALE, "Invalid file handle" }, + { NFS3ERR_REMOTE, "Too many levels of remote in path" }, + { NFS3ERR_BADHANDLE, "Illegal NFS file handle" }, + { NFS3ERR_NOT_SYNC, "Update synchronization mismatch detected" }, + { NFS3ERR_BAD_COOKIE, "READDIR or READDIRPLUS cookie is stale"}, + { NFS3ERR_NOTSUPP, "Operation is not supported" }, + { NFS3ERR_TOOSMALL, "Buffer or request is too small" }, + { NFS3ERR_SERVERFAULT, "Error occurred on the server or IO Error" }, + { NFS3ERR_BADTYPE, "Type not supported by the server" }, + { NFS3ERR_JUKEBOX, "Cannot complete server initiated request" }, + { -1, "IO Error" }, + +}; + + +uint64_t +nfs3_iatt_gfid_to_ino (struct iatt *buf) +{ + uint64_t ino = 0; + + if (!buf) + return 0; + + if (gf_nfs_enable_ino32()) { + ino = (uint32_t )nfs_hash_gfid (buf->ia_gfid); + goto hashout; + } + + /* from posix its guaranteed to send unique ino */ + ino = buf->ia_ino; + +hashout: + return ino; +} + + +void +nfs3_map_deviceid_to_statdev (struct iatt *ia, uint64_t deviceid) +{ + if (!ia) + return; + + ia->ia_dev = deviceid; +} + + +struct nfs3_fh +nfs3_extract_nfs3_fh (nfs_fh3 fh) +{ + struct nfs3_fh gfh; + + memcpy (&gfh, fh.data.data_val, fh.data.data_len); + return gfh; +} + + +struct nfs3_fh +nfs3_extract_lookup_fh (lookup3args *args) +{ + return nfs3_extract_nfs3_fh(args->what.dir); +} + + +char * +nfs3_extract_lookup_name (lookup3args *args) +{ + return args->what.name; +} + + +nfsstat3 +nfs3_errno_to_nfsstat3 (int errnum) +{ + nfsstat3 stat = NFS3_OK; + + switch (errnum) { + + case 0: + stat = NFS3_OK; + break; + + case EPERM: + stat = NFS3ERR_PERM; + break; + + case ENOENT: + stat = NFS3ERR_NOENT; + break; + + case EACCES: + stat = NFS3ERR_ACCES; + break; + + case EEXIST: + stat = NFS3ERR_EXIST; + break; + + case EXDEV: + stat = NFS3ERR_XDEV; + break; + + case ENODEV: + stat = NFS3ERR_NODEV; + break; + + case EIO: + stat = NFS3ERR_IO; + break; + + case ENXIO: + stat = NFS3ERR_NXIO; + break; + + case ENOTDIR: + stat = NFS3ERR_NOTDIR; + break; + + case EISDIR: + stat = NFS3ERR_ISDIR; + break; + + case EINVAL: + stat = NFS3ERR_INVAL; + break; + + case ENOSPC: + stat = NFS3ERR_NOSPC; + break; + + case EROFS: + stat = NFS3ERR_ROFS; + break; + + case EFBIG: + stat = NFS3ERR_FBIG; + break; + + case EMLINK: + stat = NFS3ERR_MLINK; + break; + + case ENAMETOOLONG: + stat = NFS3ERR_NAMETOOLONG; + break; + + case ENOTEMPTY: + stat = NFS3ERR_NOTEMPTY; + break; + + case EFAULT: + stat = NFS3ERR_SERVERFAULT; + break; + + case ENOSYS: + stat = NFS3ERR_NOTSUPP; + break; + + case EBADF: + stat = NFS3ERR_BADTYPE; + break; + + case ESTALE: + stat = NFS3ERR_STALE; + break; + + case ENOTCONN: + stat = NFS3ERR_IO; + break; + + case EDQUOT: + stat = NFS3ERR_DQUOT; + break; + + default: + stat = NFS3ERR_SERVERFAULT; + break; + } + + return stat; +} + +/* + * Special case: If op_ret is -1, it's very unusual op_errno being + * 0 which means something came wrong from upper layer(s). If it + * happens by any means, then set NFS3 status to NFS3ERR_SERVERFAULT. + */ +inline nfsstat3 +nfs3_cbk_errno_status (int32_t op_ret, int32_t op_errno) +{ + if ((op_ret == -1) && (op_errno == 0)) { + return NFS3ERR_SERVERFAULT; + } + + return nfs3_errno_to_nfsstat3 (op_errno); +} + +void +nfs3_fill_lookup3res_error (lookup3res *res, nfsstat3 stat, + struct iatt *dirstat) +{ + + memset (res, 0, sizeof (*res)); + res->status = stat; + if (!dirstat) { + res->lookup3res_u.resfail.dir_attributes.attributes_follow = FALSE; + } + +} + +fattr3 +nfs3_stat_to_fattr3 (struct iatt *buf) +{ + fattr3 fa = {0, }; + + if (buf == NULL) + goto out; + + if (IA_ISDIR (buf->ia_type)) + fa.type = NF3DIR; + else if (IA_ISREG (buf->ia_type)) + fa.type = NF3REG; + else if (IA_ISCHR (buf->ia_type)) + fa.type = NF3CHR; + else if (IA_ISBLK (buf->ia_type)) + fa.type = NF3BLK; + else if (IA_ISFIFO (buf->ia_type)) + fa.type = NF3FIFO; + else if (IA_ISLNK (buf->ia_type)) + fa.type = NF3LNK; + else if (IA_ISSOCK (buf->ia_type)) + fa.type = NF3SOCK; + + if (IA_PROT_RUSR (buf->ia_prot)) + fa.mode |= NFS3MODE_ROWNER; + if (IA_PROT_WUSR (buf->ia_prot)) + fa.mode |= NFS3MODE_WOWNER; + if (IA_PROT_XUSR (buf->ia_prot)) + fa.mode |= NFS3MODE_XOWNER; + + if (IA_PROT_RGRP (buf->ia_prot)) + fa.mode |= NFS3MODE_RGROUP; + if (IA_PROT_WGRP (buf->ia_prot)) + fa.mode |= NFS3MODE_WGROUP; + if (IA_PROT_XGRP (buf->ia_prot)) + fa.mode |= NFS3MODE_XGROUP; + + if (IA_PROT_ROTH (buf->ia_prot)) + fa.mode |= NFS3MODE_ROTHER; + if (IA_PROT_WOTH (buf->ia_prot)) + fa.mode |= NFS3MODE_WOTHER; + if (IA_PROT_XOTH (buf->ia_prot)) + fa.mode |= NFS3MODE_XOTHER; + + if (IA_PROT_SUID (buf->ia_prot)) + fa.mode |= NFS3MODE_SETXUID; + if (IA_PROT_SGID (buf->ia_prot)) + fa.mode |= NFS3MODE_SETXGID; + if (IA_PROT_STCKY (buf->ia_prot)) + fa.mode |= NFS3MODE_SAVESWAPTXT; + + fa.nlink = buf->ia_nlink; + fa.uid = buf->ia_uid; + fa.gid = buf->ia_gid; + fa.size = buf->ia_size; + fa.used = (buf->ia_blocks * 512); + + if ((IA_ISCHR (buf->ia_type) || IA_ISBLK (buf->ia_type))) { + fa.rdev.specdata1 = ia_major (buf->ia_rdev); + fa.rdev.specdata2 = ia_minor (buf->ia_rdev); + } else { + fa.rdev.specdata1 = 0; + fa.rdev.specdata2 = 0; + } + + fa.fsid = buf->ia_dev; + fa.fileid = nfs3_iatt_gfid_to_ino (buf); + + fa.atime.seconds = buf->ia_atime; + fa.atime.nseconds = buf->ia_atime_nsec; + + fa.ctime.seconds = buf->ia_ctime; + fa.ctime.nseconds = buf->ia_ctime_nsec; + + fa.mtime.seconds = buf->ia_mtime; + fa.mtime.nseconds = buf->ia_mtime_nsec; + +out: + return fa; +} + + +post_op_attr +nfs3_stat_to_post_op_attr (struct iatt *buf) +{ + post_op_attr attr = {0, }; + if (!buf) + return attr; + + /* Some performance translators return zero-filled stats when they + * do not have up-to-date attributes. Need to handle this by not + * returning these zeroed out attrs. + */ + attr.attributes_follow = FALSE; + if (nfs_zero_filled_stat (buf)) + goto out; + + attr.post_op_attr_u.attributes = nfs3_stat_to_fattr3 (buf); + attr.attributes_follow = TRUE; + +out: + return attr; +} + + +pre_op_attr +nfs3_stat_to_pre_op_attr (struct iatt *pre) +{ + pre_op_attr poa = {0, }; + + /* Some performance translators return zero-filled stats when they + * do not have up-to-date attributes. Need to handle this by not + * returning these zeroed out attrs. + */ + poa.attributes_follow = FALSE; + if (nfs_zero_filled_stat (pre)) + goto out; + + poa.attributes_follow = TRUE; + poa.pre_op_attr_u.attributes.size = pre->ia_size; + poa.pre_op_attr_u.attributes.mtime.seconds = pre->ia_mtime; + poa.pre_op_attr_u.attributes.mtime.nseconds = pre->ia_mtime_nsec; + poa.pre_op_attr_u.attributes.ctime.seconds = pre->ia_ctime; + poa.pre_op_attr_u.attributes.ctime.nseconds = pre->ia_ctime_nsec; + +out: + return poa; +} + +void +nfs3_fill_lookup3res_success (lookup3res *res, nfsstat3 stat, + struct nfs3_fh *fh, struct iatt *buf, + struct iatt *postparent) +{ + post_op_attr obj, dir; + uint32_t fhlen = 0; + + res->status = stat; + if (fh) { + res->lookup3res_u.resok.object.data.data_val = (void *)fh; + fhlen = nfs3_fh_compute_size (fh); + res->lookup3res_u.resok.object.data.data_len = fhlen; + } + + obj.attributes_follow = FALSE; + dir.attributes_follow = FALSE; + obj = nfs3_stat_to_post_op_attr (buf); + dir = nfs3_stat_to_post_op_attr (postparent); + + res->lookup3res_u.resok.obj_attributes = obj; + res->lookup3res_u.resok.dir_attributes = dir; +} + + +void +nfs3_fill_lookup3res (lookup3res *res, nfsstat3 stat, struct nfs3_fh *newfh, + struct iatt *buf, struct iatt *postparent, + uint64_t deviceid) +{ + + memset (res, 0, sizeof (*res)); + nfs3_map_deviceid_to_statdev (buf, deviceid); + nfs3_map_deviceid_to_statdev (postparent, deviceid); + if (stat != NFS3_OK) + nfs3_fill_lookup3res_error (res, stat, postparent); + else + nfs3_fill_lookup3res_success (res, stat, newfh, buf, + postparent); +} + +struct nfs3_fh +nfs3_extract_getattr_fh (getattr3args *args) +{ + return nfs3_extract_nfs3_fh(args->object); +} + + +void +nfs3_fill_getattr3res (getattr3res *res, nfsstat3 stat, struct iatt *buf, + uint64_t deviceid) +{ + + memset (res, 0, sizeof (*res)); + res->status = stat; + if (stat != NFS3_OK) + return; + + nfs3_map_deviceid_to_statdev (buf, deviceid); + res->getattr3res_u.resok.obj_attributes = nfs3_stat_to_fattr3 (buf); + +} + + +struct nfs3_fh +nfs3_extract_fsinfo_fh (fsinfo3args *args) +{ + return nfs3_extract_nfs3_fh (args->fsroot); +} + + +void +nfs3_fill_fsinfo3res (struct nfs3_state *nfs3, fsinfo3res *res, + nfsstat3 status, struct iatt *fsroot, uint64_t deviceid) +{ + fsinfo3resok resok = {{0}, }; + nfstime3 tdelta = GF_NFS3_TIMEDELTA_SECS; + + memset (res, 0, sizeof (*res)); + res->status = status; + if (status != NFS3_OK) + return; + + nfs3_map_deviceid_to_statdev (fsroot, deviceid); + resok.obj_attributes = nfs3_stat_to_post_op_attr (fsroot); + resok.rtmax = nfs3->readsize; + resok.rtpref = nfs3->readsize; + resok.rtmult = GF_NFS3_RTMULT; + resok.wtmax = nfs3->writesize; + resok.wtpref = nfs3->writesize; + resok.wtmult = GF_NFS3_WTMULT; + resok.dtpref = nfs3->readdirsize; + resok.maxfilesize = GF_NFS3_MAXFILESIZE; + resok.time_delta = tdelta; + resok.properties = GF_NFS3_FS_PROP; + + res->fsinfo3res_u.resok = resok; + +} + + +void +nfs3_prep_lookup3args (lookup3args *args, struct nfs3_fh *fh, char *name) +{ + memset (args, 0, sizeof (*args)); + args->what.dir.data.data_val = (void *)fh; + args->what.name = name; +} + + +void +nfs3_prep_getattr3args (getattr3args *args, struct nfs3_fh *fh) +{ + memset (args, 0, sizeof (*args)); + args->object.data.data_val = (void *)fh; +} + + +void +nfs3_prep_fsinfo3args (fsinfo3args *args, struct nfs3_fh *root) +{ + memset (args, 0, sizeof (*args)); + args->fsroot.data.data_val = (void *)root; +} + + +char * +nfsstat3_strerror(int stat) +{ + int i; + for(i = 0; nfs3stat_strerror_table[i].stat != -1; i++) { + if (nfs3stat_strerror_table[i].stat == stat) + return nfs3stat_strerror_table[i].strerror; + } + + return nfs3stat_strerror_table[i].strerror; +} + + + +void +nfs3_prep_access3args (access3args *args, struct nfs3_fh *fh) +{ + memset (args, 0, sizeof (*args)); + args->object.data.data_val = (void *)fh; +} + +#define POSIX_READ 4 +#define POSIX_WRITE 2 +#define POSIX_EXEC 1 + +uint32_t +nfs3_accessbits (int32_t accbits) +{ + uint32_t accresult = 0; + + if (accbits & POSIX_READ) + accresult |= ACCESS3_READ; + + if (accbits & POSIX_WRITE) + accresult |= (ACCESS3_MODIFY | ACCESS3_EXTEND | ACCESS3_DELETE); + + /* lookup on directory allowed only in case of execute permission */ + if (accbits & POSIX_EXEC) + accresult |= (ACCESS3_EXECUTE | ACCESS3_LOOKUP); + + return accresult; +} + +uint32_t +nfs3_request_to_accessbits (int32_t accbits) +{ + uint32_t acc_request = 0; + + if (accbits & ACCESS3_READ) + acc_request |= POSIX_READ; + + if (accbits & (ACCESS3_MODIFY | ACCESS3_EXTEND | ACCESS3_DELETE)) + acc_request |= POSIX_WRITE; + + /* For lookup on directory check for execute permission */ + if (accbits & (ACCESS3_EXECUTE | ACCESS3_LOOKUP)) + acc_request |= POSIX_EXEC; + + return acc_request; +} +void +nfs3_fill_access3res (access3res *res, nfsstat3 status, int32_t accbits, + int32_t reqaccbits) +{ + uint32_t accres = 0; + + memset (res, 0, sizeof (*res)); + res->status = status; + if (status != NFS3_OK) + return; + + accres = nfs3_accessbits (accbits); + + /* do not answer what was not asked */ + res->access3res_u.resok.access = accres & reqaccbits; +} + +void +nfs3_prep_readdir3args (readdir3args *ra, struct nfs3_fh *fh) +{ + memset (ra, 0, sizeof (*ra)); + ra->dir.data.data_val = (void *)fh; +} + + +int +nfs3_is_dot_entry (char *entry) +{ + int ret = 0; + + if (!entry) + return 0; + + if (strcmp (entry, ".") == 0) + ret = 1; + + return ret; +} + + +int +nfs3_is_parentdir_entry (char *entry) +{ + int ret = 0; + + if (!entry) + return 0; + + if (strcmp (entry, "..") == 0) + ret = 1; + + return ret; +} + + +void +nfs3_funge_root_dotdot_dirent (gf_dirent_t *ent, struct nfs3_fh *dfh) +{ + if ((!ent) || (!dfh)) + return; + + if (nfs3_fh_is_root_fh (dfh) && + nfs3_is_parentdir_entry (ent->d_name)) { + ent->d_ino = 1; + ent->d_stat.ia_ino = 1; + } + + if (nfs3_fh_is_root_fh (dfh) && + nfs3_is_dot_entry (ent->d_name)) { + ent->d_ino = 1; + ent->d_stat.ia_ino = 1; + } + +} + + +entry3 * +nfs3_fill_entry3 (gf_dirent_t *entry, struct nfs3_fh *dfh) +{ + entry3 *ent = NULL; + if ((!entry) || (!dfh)) + return NULL; + + ent = GF_CALLOC (1, sizeof (*ent), gf_nfs_mt_entry3); + if (!ent) + return NULL; + + gf_log (GF_NFS3, GF_LOG_TRACE, "Entry: %s", entry->d_name); + + /* If the entry is . or .., we need to replace the physical ino and gen + * with 1 and 0 respectively if the directory is root. This funging is + * needed because there is no parent directory of the root. In that + * sense the behavior we provide is similar to the output of the + * command: "stat /.." + */ + entry->d_ino = nfs3_iatt_gfid_to_ino (&entry->d_stat); + nfs3_funge_root_dotdot_dirent (entry, dfh); + ent->fileid = entry->d_ino; + ent->cookie = entry->d_off; + ent->name = GF_CALLOC ((strlen (entry->d_name) + 1), sizeof (char), + gf_nfs_mt_char); + if (!ent->name) { + GF_FREE (ent); + ent = NULL; + goto err; + } + strcpy (ent->name, entry->d_name); + +err: + return ent; +} + + +void +nfs3_fill_post_op_fh3 (struct nfs3_fh *fh, post_op_fh3 *pfh) +{ + uint32_t fhlen = 0; + + if ((!fh) || (!pfh)) + return; + + pfh->handle_follows = 1; + fhlen = nfs3_fh_compute_size (fh); + pfh->post_op_fh3_u.handle.data.data_val = (void *)fh; + pfh->post_op_fh3_u.handle.data.data_len = fhlen; +} + + +post_op_fh3 +nfs3_fh_to_post_op_fh3 (struct nfs3_fh *fh) +{ + post_op_fh3 pfh = {0, }; + char *fhp = NULL; + + if (!fh) + return pfh; + + pfh.handle_follows = 1; + + fhp = GF_CALLOC (1, sizeof (*fh), gf_nfs_mt_char); + if (!fhp) + return pfh; + + memcpy (fhp, fh, sizeof (*fh)); + nfs3_fill_post_op_fh3 ((struct nfs3_fh *)fhp, &pfh); + return pfh; +} + + +entryp3 * +nfs3_fill_entryp3 (gf_dirent_t *entry, struct nfs3_fh *dirfh, uint64_t devid) +{ + entryp3 *ent = NULL; + struct nfs3_fh newfh = {{0}, }; + + if ((!entry) || (!dirfh)) + return NULL; + + /* If the entry is . or .., we need to replace the physical ino and gen + * with 1 and 0 respectively if the directory is root. This funging is + * needed because there is no parent directory of the root. In that + * sense the behavior we provide is similar to the output of the + * command: "stat /.." + */ + entry->d_ino = nfs3_iatt_gfid_to_ino (&entry->d_stat); + nfs3_funge_root_dotdot_dirent (entry, dirfh); + gf_log (GF_NFS3, GF_LOG_TRACE, "Entry: %s, ino: %"PRIu64, + entry->d_name, entry->d_ino); + ent = GF_CALLOC (1, sizeof (*ent), gf_nfs_mt_entryp3); + if (!ent) + return NULL; + + ent->fileid = entry->d_ino; + ent->cookie = entry->d_off; + ent->name = GF_CALLOC ((strlen (entry->d_name) + 1), sizeof (char), + gf_nfs_mt_char); + if (!ent->name) { + GF_FREE (ent); + ent = NULL; + goto err; + } + strcpy (ent->name, entry->d_name); + + nfs3_fh_build_child_fh (dirfh, &entry->d_stat, &newfh); + nfs3_map_deviceid_to_statdev (&entry->d_stat, devid); + ent->name_attributes = nfs3_stat_to_post_op_attr (&entry->d_stat); + ent->name_handle = nfs3_fh_to_post_op_fh3 (&newfh); +err: + return ent; +} + + +void +nfs3_fill_readdir3res (readdir3res *res, nfsstat3 stat, struct nfs3_fh *dirfh, + uint64_t cverf, struct iatt *dirstat, + gf_dirent_t *entries, count3 count, int is_eof, + uint64_t deviceid) +{ + post_op_attr dirattr; + entry3 *ent = NULL; + entry3 *headentry = NULL; + entry3 *preventry = NULL; + count3 filled = 0; + gf_dirent_t *listhead = NULL; + + memset (res, 0, sizeof (*res)); + res->status = stat; + if (stat != NFS3_OK) + return; + + nfs3_map_deviceid_to_statdev (dirstat, deviceid); + dirattr = nfs3_stat_to_post_op_attr (dirstat); + res->readdir3res_u.resok.dir_attributes = dirattr; + res->readdir3res_u.resok.reply.eof = (bool_t)is_eof; + memcpy (res->readdir3res_u.resok.cookieverf, &cverf, sizeof (cverf)); + + filled = NFS3_READDIR_RESOK_SIZE; + /* First entry is just the list head */ + listhead = entries; + entries = entries->next; + while (((entries) && (entries != listhead)) && (filled < count)) { + /* + if ((strcmp (entries->d_name, ".") == 0) || + (strcmp (entries->d_name, "..") == 0)) + goto nextentry; + */ + ent = nfs3_fill_entry3 (entries, dirfh); + if (!ent) + break; + + if (!headentry) + headentry = ent; + + if (preventry) { + preventry->nextentry = ent; + preventry = ent; + } else + preventry = ent; + + filled += NFS3_ENTRY3_FIXED_SIZE + strlen (ent->name); +//nextentry: + entries = entries->next; + } + + res->readdir3res_u.resok.reply.entries = headentry; + + return; +} + + +void +nfs3_fill_readdirp3res (readdirp3res *res, nfsstat3 stat, + struct nfs3_fh *dirfh, uint64_t cverf, + struct iatt *dirstat, gf_dirent_t *entries, + count3 dircount, count3 maxcount, int is_eof, + uint64_t deviceid) +{ + post_op_attr dirattr; + entryp3 *ent = NULL; + entryp3 *headentry = NULL; + entryp3 *preventry = NULL; + count3 filled = 0; + gf_dirent_t *listhead = NULL; + int fhlen = 0; + + memset (res, 0, sizeof (*res)); + res->status = stat; + if (stat != NFS3_OK) + return; + + nfs3_map_deviceid_to_statdev (dirstat, deviceid); + dirattr = nfs3_stat_to_post_op_attr (dirstat); + res->readdirp3res_u.resok.dir_attributes = dirattr; + res->readdirp3res_u.resok.reply.eof = (bool_t)is_eof; + memcpy (res->readdirp3res_u.resok.cookieverf, &cverf, sizeof (cverf)); + + filled = NFS3_READDIR_RESOK_SIZE; + /* First entry is just the list head */ + listhead = entries; + entries = entries->next; + while (((entries) && (entries != listhead)) && (filled < maxcount)) { + /* Linux does not display . and .. entries unless we provide + * these entries here. + */ +/* if ((strcmp (entries->d_name, ".") == 0) || + (strcmp (entries->d_name, "..") == 0)) + goto nextentry; + */ + ent = nfs3_fill_entryp3 (entries, dirfh, deviceid); + if (!ent) + break; + + if (!headentry) + headentry = ent; + + if (preventry) { + preventry->nextentry = ent; + preventry = ent; + } else + preventry = ent; + + fhlen = ent->name_handle.post_op_fh3_u.handle.data.data_len; + filled += NFS3_ENTRYP3_FIXED_SIZE + fhlen + strlen (ent->name); +//nextentry: + entries = entries->next; + } + + res->readdirp3res_u.resok.reply.entries = headentry; + + return; +} + + +void +nfs3_prep_readdirp3args (readdirp3args *ra, struct nfs3_fh *fh) +{ + memset (ra, 0, sizeof (*ra)); + ra->dir.data.data_val = (void *)fh; +} + +void +nfs3_free_readdirp3res (readdirp3res *res) +{ + entryp3 *ent = NULL; + entryp3 *next = NULL; + + if (!res) + return; + + ent = res->readdirp3res_u.resok.reply.entries; + while (ent) { + + next = ent->nextentry; + GF_FREE (ent->name); + GF_FREE (ent->name_handle.post_op_fh3_u.handle.data.data_val); + GF_FREE (ent); + ent = next; + } + + return; +} + + +void +nfs3_free_readdir3res (readdir3res *res) +{ + entry3 *ent = NULL; + entry3 *next = NULL; + + if (!res) + return; + + ent = res->readdir3res_u.resok.reply.entries; + while (ent) { + + next = ent->nextentry; + GF_FREE (ent->name); + GF_FREE (ent); + ent = next; + } + + return; +} + +void +nfs3_prep_fsstat3args (fsstat3args *args, struct nfs3_fh *fh) +{ + memset (args, 0, sizeof (*args)); + args->fsroot.data.data_val = (char *)fh; +} + + +void +nfs3_fill_fsstat3res (fsstat3res *res, nfsstat3 stat, struct statvfs *fsbuf, + struct iatt *postbuf, uint64_t deviceid) +{ + post_op_attr poa; + fsstat3resok resok; + + memset (res, 0, sizeof (*res)); + res->status = stat; + if (stat != NFS3_OK) + return; + + nfs3_map_deviceid_to_statdev (postbuf, deviceid); + poa = nfs3_stat_to_post_op_attr (postbuf); + resok.tbytes = (size3)(fsbuf->f_frsize * fsbuf->f_blocks); + resok.fbytes = (size3)(fsbuf->f_frsize * fsbuf->f_bfree); + resok.abytes = (size3)(fsbuf->f_frsize * fsbuf->f_bavail); + resok.tfiles = (size3)(fsbuf->f_files); + resok.ffiles = (size3)(fsbuf->f_ffree); + resok.afiles = (size3)(fsbuf->f_favail); + resok.invarsec = 0; + + resok.obj_attributes = poa; + res->fsstat3res_u.resok = resok; +} + + +int32_t +nfs3_sattr3_to_setattr_valid (sattr3 *sattr, struct iatt *buf, mode_t *omode) +{ + int32_t valid = 0; + ia_prot_t prot = {0, }; + mode_t mode = 0; + + if (!sattr) + return 0; + + if (sattr->mode.set_it) { + valid |= GF_SET_ATTR_MODE; + + if (sattr->mode.set_mode3_u.mode & NFS3MODE_ROWNER) { + mode |= S_IRUSR; + prot.owner.read = 1; + } + if (sattr->mode.set_mode3_u.mode & NFS3MODE_WOWNER) { + mode |= S_IWUSR; + prot.owner.write = 1; + } + if (sattr->mode.set_mode3_u.mode & NFS3MODE_XOWNER) { + mode |= S_IXUSR; + prot.owner.exec = 1; + } + + if (sattr->mode.set_mode3_u.mode & NFS3MODE_RGROUP) { + mode |= S_IRGRP; + prot.group.read = 1; + } + if (sattr->mode.set_mode3_u.mode & NFS3MODE_WGROUP) { + mode |= S_IWGRP; + prot.group.write = 1; + } + if (sattr->mode.set_mode3_u.mode & NFS3MODE_XGROUP) { + mode |= S_IXGRP; + prot.group.exec = 1; + } + + if (sattr->mode.set_mode3_u.mode & NFS3MODE_ROTHER) { + mode |= S_IROTH; + prot.other.read = 1; + } + if (sattr->mode.set_mode3_u.mode & NFS3MODE_WOTHER) { + mode |= S_IWOTH; + prot.other.write = 1; + } + if (sattr->mode.set_mode3_u.mode & NFS3MODE_XOTHER) { + mode |= S_IXOTH; + prot.other.exec = 1; + } + + if (sattr->mode.set_mode3_u.mode & NFS3MODE_SETXUID) { + mode |= S_ISUID; + prot.suid = 1; + } + if (sattr->mode.set_mode3_u.mode & NFS3MODE_SETXGID) { + mode |= S_ISGID; + prot.sgid = 1; + } + if (sattr->mode.set_mode3_u.mode & NFS3MODE_SAVESWAPTXT) { + mode |= S_ISVTX; + prot.sticky = 1; + } + + if (buf) + buf->ia_prot = prot; + /* Create fop still requires the old mode_t style argument. */ + if (omode) + *omode = mode; + } + + if (sattr->uid.set_it) { + valid |= GF_SET_ATTR_UID; + if (buf) + buf->ia_uid = sattr->uid.set_uid3_u.uid; + } + + if (sattr->gid.set_it) { + valid |= GF_SET_ATTR_GID; + if (buf) + buf->ia_gid = sattr->gid.set_gid3_u.gid; + } + + if (sattr->size.set_it) { + valid |= GF_SET_ATTR_SIZE; + if (buf) + buf->ia_size = sattr->size.set_size3_u.size; + } + + if (sattr->atime.set_it == SET_TO_CLIENT_TIME) { + valid |= GF_SET_ATTR_ATIME; + if (buf) + buf->ia_atime = sattr->atime.set_atime_u.atime.seconds; + } + + if (sattr->atime.set_it == SET_TO_SERVER_TIME) { + valid |= GF_SET_ATTR_ATIME; + if (buf) + buf->ia_atime = time (NULL); + } + + if (sattr->mtime.set_it == SET_TO_CLIENT_TIME) { + valid |= GF_SET_ATTR_MTIME; + if (buf) + buf->ia_mtime = sattr->mtime.set_mtime_u.mtime.seconds; + } + + if (sattr->mtime.set_it == SET_TO_SERVER_TIME) { + valid |= GF_SET_ATTR_MTIME; + if (buf) + buf->ia_mtime = time (NULL); + } + + return valid; +} + + +wcc_data +nfs3_stat_to_wcc_data (struct iatt *pre, struct iatt *post) +{ + wcc_data wd = {{0}, }; + + if (post) + wd.after = nfs3_stat_to_post_op_attr (post); + if (pre) + wd.before = nfs3_stat_to_pre_op_attr (pre); + + return wd; +} + +void +nfs3_fill_create3res (create3res *res, nfsstat3 stat, struct nfs3_fh *newfh, + struct iatt *newbuf, struct iatt *preparent, + struct iatt *postparent, uint64_t deviceid) +{ + post_op_attr poa = {0, }; + wcc_data dirwcc = {{0}, }; + + memset (res, 0, sizeof (*res)); + res->status = stat; + if (stat != NFS3_OK) + return; + + nfs3_fill_post_op_fh3 (newfh, &res->create3res_u.resok.obj); + nfs3_map_deviceid_to_statdev (newbuf, deviceid); + poa = nfs3_stat_to_post_op_attr (newbuf); + res->create3res_u.resok.obj_attributes = poa; + nfs3_map_deviceid_to_statdev (preparent, deviceid); + nfs3_map_deviceid_to_statdev (postparent, deviceid); + dirwcc = nfs3_stat_to_wcc_data (preparent, postparent); + + res->create3res_u.resok.dir_wcc = dirwcc; +} + +void +nfs3_prep_create3args (create3args *args, struct nfs3_fh *fh, char *name) +{ + + memset (args, 0, sizeof (*args)); + args->where.dir.data.data_val = (void *)fh; + args->where.name = name; +} + +void +nfs3_prep_setattr3args (setattr3args *args, struct nfs3_fh *fh) +{ + memset (args, 0, sizeof (*args)); + args->object.data.data_val = (void *)fh; +} + + +void +nfs3_fill_setattr3res (setattr3res *res, nfsstat3 stat, struct iatt *preop, + struct iatt *postop, uint64_t deviceid) +{ + wcc_data wcc; + memset (res, 0, sizeof (*res)); + res->status = stat; + if (stat != NFS3_OK) + return; + + nfs3_map_deviceid_to_statdev (preop, deviceid); + nfs3_map_deviceid_to_statdev (postop, deviceid); + wcc = nfs3_stat_to_wcc_data (preop, postop); + res->setattr3res_u.resok.obj_wcc = wcc; +} + + +void +nfs3_prep_mkdir3args (mkdir3args *args, struct nfs3_fh *dirfh, char *name) +{ + + memset (args, 0, sizeof (*args)); + args->where.dir.data.data_val = (void *)dirfh; + args->where.name = name; +} + + +void +nfs3_fill_mkdir3res (mkdir3res *res, nfsstat3 stat, struct nfs3_fh *fh, + struct iatt *buf, struct iatt *preparent, + struct iatt *postparent, uint64_t deviceid) +{ + wcc_data dirwcc; + post_op_attr poa; + + memset (res, 0, sizeof (*res)); + res->status = stat; + if (stat != NFS3_OK) + return; + + nfs3_fill_post_op_fh3 (fh, &res->mkdir3res_u.resok.obj); + nfs3_map_deviceid_to_statdev (buf, deviceid); + poa = nfs3_stat_to_post_op_attr (buf); + nfs3_map_deviceid_to_statdev (preparent, deviceid); + nfs3_map_deviceid_to_statdev (postparent, deviceid); + dirwcc = nfs3_stat_to_wcc_data (preparent, postparent); + res->mkdir3res_u.resok.obj_attributes = poa; + res->mkdir3res_u.resok.dir_wcc = dirwcc; + +} + + +void +nfs3_prep_symlink3args (symlink3args *args, struct nfs3_fh *dirfh, char *name, + char *target) +{ + memset (args, 0, sizeof (*args)); + args->where.dir.data.data_val = (void *)dirfh; + args->where.name = name; + args->symlink.symlink_data = target; +} + + +void +nfs3_fill_symlink3res (symlink3res *res, nfsstat3 stat, struct nfs3_fh *fh, + struct iatt *buf, struct iatt *preparent, + struct iatt *postparent, uint64_t deviceid) +{ + wcc_data dirwcc; + post_op_attr poa; + + memset (res, 0, sizeof (*res)); + res->status = stat; + if (stat != NFS3_OK) + return; + + nfs3_fill_post_op_fh3 (fh, &res->symlink3res_u.resok.obj); + nfs3_map_deviceid_to_statdev (buf, deviceid); + poa = nfs3_stat_to_post_op_attr (buf); + nfs3_map_deviceid_to_statdev (postparent, deviceid); + nfs3_map_deviceid_to_statdev (preparent, deviceid); + dirwcc = nfs3_stat_to_wcc_data (preparent, postparent); + res->symlink3res_u.resok.obj_attributes = poa; + res->symlink3res_u.resok.dir_wcc = dirwcc; + +} + + +void +nfs3_prep_readlink3args (readlink3args *args, struct nfs3_fh *fh) +{ + + memset (args, 0, sizeof (*args)); + args->symlink.data.data_val = (void *)fh; +} + + +void +nfs3_fill_readlink3res (readlink3res *res, nfsstat3 stat, char *path, + struct iatt *buf, uint64_t deviceid) +{ + post_op_attr poa; + + memset (res, 0, sizeof (*res)); + res->status = stat; + + if (stat != NFS3_OK) + return; + + nfs3_map_deviceid_to_statdev (buf, deviceid); + poa = nfs3_stat_to_post_op_attr (buf); + res->readlink3res_u.resok.data = (void *)path; + res->readlink3res_u.resok.symlink_attributes = poa; +} + + +void +nfs3_prep_mknod3args (mknod3args *args, struct nfs3_fh *fh, char *name) +{ + memset (args, 0, sizeof (*args)); + args->where.dir.data.data_val = (void *)fh; + args->where.name = name; + +} + +void +nfs3_fill_mknod3res (mknod3res *res, nfsstat3 stat, struct nfs3_fh *fh, + struct iatt *buf, struct iatt *preparent, + struct iatt *postparent, uint64_t deviceid) +{ + post_op_attr poa; + wcc_data wccdir; + + memset (res, 0, sizeof (*res)); + res->status = stat; + if (stat != NFS3_OK) + return; + + nfs3_fill_post_op_fh3 (fh, &res->mknod3res_u.resok.obj); + nfs3_map_deviceid_to_statdev (buf, deviceid); + poa = nfs3_stat_to_post_op_attr (buf); + nfs3_map_deviceid_to_statdev (preparent, deviceid); + nfs3_map_deviceid_to_statdev (postparent, deviceid); + wccdir = nfs3_stat_to_wcc_data (preparent, postparent); + res->mknod3res_u.resok.obj_attributes = poa; + res->mknod3res_u.resok.dir_wcc = wccdir; + +} + + +void +nfs3_fill_remove3res (remove3res *res, nfsstat3 stat, struct iatt *preparent, + struct iatt *postparent, uint64_t deviceid) +{ + wcc_data dirwcc; + + memset (res, 0, sizeof (*res)); + res->status = stat; + if (stat != NFS3_OK) + return; + + nfs3_map_deviceid_to_statdev (preparent, deviceid); + nfs3_map_deviceid_to_statdev (postparent, deviceid); + dirwcc = nfs3_stat_to_wcc_data (preparent, postparent); + res->remove3res_u.resok.dir_wcc = dirwcc; +} + + +void +nfs3_prep_remove3args (remove3args *args, struct nfs3_fh *fh, char *name) +{ + memset (args, 0, sizeof (*args)); + args->object.dir.data.data_val = (void *)fh; + args->object.name = name; +} + + +void +nfs3_prep_rmdir3args (rmdir3args *args, struct nfs3_fh *fh, char *name) +{ + memset (args, 0, sizeof (*args)); + args->object.dir.data.data_val = (void *)fh; + args->object.name = name; +} + + +void +nfs3_fill_rmdir3res (rmdir3res *res, nfsstat3 stat, struct iatt *preparent, + struct iatt *postparent, uint64_t deviceid) +{ + wcc_data dirwcc; + memset (res, 0, sizeof (*res)); + res->status = stat; + + if (stat != NFS3_OK) + return; + + nfs3_map_deviceid_to_statdev (postparent, deviceid); + nfs3_map_deviceid_to_statdev (preparent, deviceid); + dirwcc = nfs3_stat_to_wcc_data (preparent, postparent); + res->rmdir3res_u.resok.dir_wcc = dirwcc; +} + + +void +nfs3_prep_link3args (link3args *args, struct nfs3_fh *target, + struct nfs3_fh * dirfh, char *name) +{ + memset (args, 0, sizeof (*args)); + args->file.data.data_val = (void *)target; + args->link.dir.data.data_val = (void *)dirfh; + args->link.name = name; +} + + +void +nfs3_fill_link3res (link3res *res, nfsstat3 stat, struct iatt *buf, + struct iatt *preparent, struct iatt *postparent, + uint64_t deviceid) +{ + post_op_attr poa; + wcc_data dirwcc; + + memset (res, 0, sizeof (*res)); + res->status = stat; + if (stat != NFS3_OK) + return; + + nfs3_map_deviceid_to_statdev (preparent, deviceid); + nfs3_map_deviceid_to_statdev (postparent, deviceid); + nfs3_map_deviceid_to_statdev (buf,deviceid); + poa = nfs3_stat_to_post_op_attr (buf); + dirwcc = nfs3_stat_to_wcc_data (preparent, postparent); + res->link3res_u.resok.file_attributes = poa; + res->link3res_u.resok.linkdir_wcc = dirwcc; +} + + + +void +nfs3_prep_rename3args (rename3args *args, struct nfs3_fh *olddirfh, + char *oldname, struct nfs3_fh *newdirfh, char *newname) +{ + memset (args, 0, sizeof (*args)); + + args->from.name = oldname; + args->from.dir.data.data_val = (void *)olddirfh; + args->to.name = newname; + args->to.dir.data.data_val = (void *)newdirfh; + +} + + +void +nfs3_fill_rename3res (rename3res *res, nfsstat3 stat, struct iatt *buf, + struct iatt *preoldparent, struct iatt *postoldparent, + struct iatt *prenewparent, struct iatt *postnewparent, + uint64_t deviceid) + +{ + wcc_data dirwcc; + + memset (res, 0, sizeof (*res)); + res->status = stat; + if (stat != NFS3_OK) + return; + + nfs3_map_deviceid_to_statdev (preoldparent, deviceid); + nfs3_map_deviceid_to_statdev (postoldparent, deviceid); + nfs3_map_deviceid_to_statdev (prenewparent, deviceid); + nfs3_map_deviceid_to_statdev (postnewparent, deviceid); + nfs3_map_deviceid_to_statdev (buf, deviceid); + dirwcc = nfs3_stat_to_wcc_data (preoldparent, postoldparent); + res->rename3res_u.resok.fromdir_wcc = dirwcc; + dirwcc = nfs3_stat_to_wcc_data (prenewparent, postnewparent); + res->rename3res_u.resok.todir_wcc = dirwcc; +} + + +void +nfs3_prep_write3args (write3args *args, struct nfs3_fh *fh) +{ + memset (args, 0, sizeof (*args)); + args->file.data.data_val = (void *)fh; +} + + +void +nfs3_fill_write3res (write3res *res, nfsstat3 stat, count3 count, + stable_how stable, uint64_t wverf, struct iatt *prestat, + struct iatt *poststat, uint64_t deviceid) +{ + write3resok resok; + memset (res, 0, sizeof (*res)); + res->status = stat; + if (stat != NFS3_OK) + return; + + nfs3_map_deviceid_to_statdev (prestat, deviceid); + nfs3_map_deviceid_to_statdev (poststat, deviceid); + resok.file_wcc = nfs3_stat_to_wcc_data (prestat, poststat); + resok.count = count; + resok.committed = stable; + memcpy (resok.verf, &wverf, sizeof (wverf)); + + res->write3res_u.resok = resok; +} + + +void +nfs3_prep_commit3args (commit3args *args, struct nfs3_fh *fh) +{ + memset (args, 0, sizeof (*args)); + args->file.data.data_val = (void *)fh; +} + + +void +nfs3_fill_commit3res (commit3res *res, nfsstat3 stat, uint64_t wverf, + struct iatt *prestat, struct iatt *poststat, + uint64_t deviceid) +{ + memset (res, 0, sizeof (*res)); + res->status = stat; + if (stat != NFS3_OK) + return; + + nfs3_map_deviceid_to_statdev (poststat, deviceid); + nfs3_map_deviceid_to_statdev (prestat, deviceid); + res->commit3res_u.resok.file_wcc = nfs3_stat_to_wcc_data (prestat, + poststat); + memcpy (res->commit3res_u.resok.verf, &wverf, sizeof (wverf)); +} + +void +nfs3_fill_read3res (read3res *res, nfsstat3 stat, count3 count, + struct iatt *poststat, int is_eof, uint64_t deviceid) +{ + post_op_attr poa; + + memset (res, 0, sizeof (*res)); + res->status = stat; + if (stat != NFS3_OK) + return; + + nfs3_map_deviceid_to_statdev (poststat, deviceid); + poa = nfs3_stat_to_post_op_attr (poststat); + res->read3res_u.resok.file_attributes = poa; + res->read3res_u.resok.count = count; + res->read3res_u.resok.eof = is_eof; + res->read3res_u.resok.data.data_len = count; +} + + +void +nfs3_prep_read3args (read3args *args, struct nfs3_fh *fh) +{ + memset (args, 0, sizeof (*args)); + args->file.data.data_val = (void *)fh; +} + + +void +nfs3_fill_pathconf3res (pathconf3res *res, nfsstat3 stat, struct iatt *buf, + uint64_t deviceid) +{ + pathconf3resok resok; + + memset (res, 0, sizeof (*res)); + res->status = stat; + if (stat != NFS3_OK) + return; + + nfs3_map_deviceid_to_statdev (buf, deviceid); + resok.obj_attributes = nfs3_stat_to_post_op_attr (buf); + resok.linkmax = 256; + resok.name_max = NFS_NAME_MAX; + resok.no_trunc = TRUE; + resok.chown_restricted = FALSE; + resok.case_insensitive = FALSE; + resok.case_preserving = TRUE; + + res->pathconf3res_u.resok = resok; +} + + +void +nfs3_prep_pathconf3args (pathconf3args *args, struct nfs3_fh *fh) +{ + memset (args, 0, sizeof (*args)); + args->object.data.data_val = (void *)fh; +} + + +int +nfs3_verify_dircookie (struct nfs3_state *nfs3, fd_t *dirfd, cookie3 cookie, + uint64_t cverf, nfsstat3 *stat) +{ + int ret = -1; + + if ((!nfs3) || (!dirfd)) + return -1; + + /* Can assume that this is first read on the dir, so cookie check + * is successful by default. + */ + if (cookie == 0) + return 0; + + gf_log (GF_NFS3, GF_LOG_TRACE, "Verifying cookie: cverf: %"PRIu64 + ", cookie: %"PRIu64, cverf, cookie); + /* The cookie bad, no way cverf will be zero with a non-zero cookie. */ + if ((cverf == 0) && (cookie != 0)) { + gf_log (GF_NFS3, GF_LOG_TRACE, "Bad cookie requested"); + if (stat) + *stat = NFS3ERR_BAD_COOKIE; + goto err; + } + + /* Yes, its true, our cookie is simply the fd_t address. + * NOTE: We used have the check for cookieverf but VMWare client sends + * a readdirp requests even after we've told it that EOF has been + * reached on the directory. This causes a problem because we close a + * dir fd_t after reaching EOF. The next readdirp sent by VMWare + * contains the address of the closed fd_t as cookieverf. Since we + * closed that fd_t, this readdirp results in a new opendir which will + * give an fd_t that will fail this check below. + */ +/* if ((cverf != (uint64_t)dirfd)) { + gf_log (GF_NFS3, GF_LOG_TRACE, "Cookieverf does not match"); + if (stat) + *stat = NFS3ERR_BAD_COOKIE; + goto err; + } +*/ + gf_log (GF_NFS3, GF_LOG_TRACE, "Cookie verified"); + if (stat) + *stat = NFS3_OK; + ret = 0; +err: + return ret; +} + + +void +nfs3_stat_to_errstr (uint32_t xid, char *op, nfsstat3 stat, int pstat, + char *errstr, size_t len) +{ + if ((!op) || (!errstr)) + return; + + snprintf (errstr, len, "XID: %x, %s: NFS: %d(%s), POSIX: %d(%s)", + xid, op,stat, nfsstat3_strerror (stat), pstat, + strerror (pstat)); +} + +void +nfs3_log_common_call (uint32_t xid, char *op, struct nfs3_fh *fh) +{ + char fhstr[1024]; + + if (THIS->ctx->log.loglevel < GF_LOG_DEBUG) + return; + + nfs3_fh_to_str (fh, fhstr, sizeof (fhstr)); + gf_log (GF_NFS3, GF_LOG_DEBUG, "XID: %x, %s: args: %s", xid, op, + fhstr); +} + + +void +nfs3_log_fh_entry_call (uint32_t xid, char *op, struct nfs3_fh *fh, + char *name) +{ + char fhstr[1024]; + + if (THIS->ctx->log.loglevel < GF_LOG_DEBUG) + return; + nfs3_fh_to_str (fh, fhstr, sizeof (fhstr)); + gf_log (GF_NFS3, GF_LOG_DEBUG, "XID: %x, %s: args: %s, name: %s", xid, + op, fhstr, name); +} + + +void +nfs3_log_rename_call (uint32_t xid, struct nfs3_fh *src, char *sname, + struct nfs3_fh *dst, char *dname) +{ + char sfhstr[1024]; + char dfhstr[1024]; + + if (THIS->ctx->log.loglevel < GF_LOG_DEBUG) + return; + nfs3_fh_to_str (src, sfhstr, sizeof (sfhstr)); + nfs3_fh_to_str (dst, dfhstr, sizeof (dfhstr)); + gf_log (GF_NFS3, GF_LOG_DEBUG, "XID: %x, RENAME: args: Src: %s, " + "name: %s, Dst: %s, name: %s", xid, sfhstr, sname, dfhstr, + dname); +} + + + +void +nfs3_log_create_call (uint32_t xid, struct nfs3_fh *fh, char *name, + createmode3 mode) +{ + char fhstr[1024]; + char *modestr = NULL; + char exclmode[] = "EXCLUSIVE"; + char unchkd[] = "UNCHECKED"; + char guarded[] = "GUARDED"; + + if (THIS->ctx->log.loglevel < GF_LOG_DEBUG) + return; + nfs3_fh_to_str (fh, fhstr, sizeof (fhstr)); + if (mode == EXCLUSIVE) + modestr = exclmode; + else if (mode == GUARDED) + modestr = guarded; + else + modestr = unchkd; + + gf_log (GF_NFS3, GF_LOG_DEBUG, "XID: %x, CREATE: args: %s, name: %s," + " mode: %s", xid, fhstr, name, modestr); +} + + +void +nfs3_log_mknod_call (uint32_t xid, struct nfs3_fh *fh, char *name, int type) +{ + char fhstr[1024]; + char *modestr = NULL; + char chr[] = "CHAR"; + char blk[] = "BLK"; + char sock[] = "SOCK"; + char fifo[] = "FIFO"; + + if (THIS->ctx->log.loglevel < GF_LOG_DEBUG) + return; + nfs3_fh_to_str (fh, fhstr, sizeof (fhstr)); + if (type == NF3CHR) + modestr = chr; + else if (type == NF3BLK) + modestr = blk; + else if (type == NF3SOCK) + modestr = sock; + else + modestr = fifo; + + gf_log (GF_NFS3, GF_LOG_DEBUG, "XID: %x, MKNOD: args: %s, name: %s," + " type: %s", xid, fhstr, name, modestr); +} + + + +void +nfs3_log_symlink_call (uint32_t xid, struct nfs3_fh *fh, char *name, char *tgt) +{ + char fhstr[1024]; + + if (THIS->ctx->log.loglevel < GF_LOG_DEBUG) + return; + nfs3_fh_to_str (fh, fhstr, sizeof (fhstr)); + gf_log (GF_NFS3, GF_LOG_DEBUG, "XID: %x, SYMLINK: args: %s, name: %s," + " target: %s", xid, fhstr, name, tgt); +} + + +void +nfs3_log_link_call (uint32_t xid, struct nfs3_fh *fh, char *name, + struct nfs3_fh *tgt) +{ + char dfhstr[1024]; + char tfhstr[1024]; + + if (THIS->ctx->log.loglevel < GF_LOG_DEBUG) + return; + nfs3_fh_to_str (fh, dfhstr, sizeof (dfhstr)); + nfs3_fh_to_str (tgt, tfhstr, sizeof (tfhstr)); + gf_log (GF_NFS3, GF_LOG_DEBUG, "XID: %x, LINK: args: %s, name: %s," + " target: %s", xid, dfhstr, name, tfhstr); +} + + +void +nfs3_log_rw_call (uint32_t xid, char *op, struct nfs3_fh *fh, offset3 offt, + count3 count, int stablewrite) +{ + char fhstr[1024]; + + if (THIS->ctx->log.loglevel < GF_LOG_DEBUG) + return; + nfs3_fh_to_str (fh, fhstr, sizeof (fhstr)); + if (stablewrite == -1) + gf_log (GF_NFS3, GF_LOG_DEBUG, "XID: %x, %s: args: %s, offset:" + " %"PRIu64", count: %"PRIu32, xid, op, fhstr, offt, + count); + else + gf_log (GF_NFS3, GF_LOG_DEBUG, "XID: %x, %s: args: %s, offset:" + " %"PRIu64", count: %"PRIu32", %s", xid, op, fhstr, + offt, count, + (stablewrite == UNSTABLE)?"UNSTABLE":"STABLE"); + +} + + +int +nfs3_getattr_loglevel (nfsstat3 stat) { + + int ll = GF_LOG_DEBUG; + + switch (stat) { + + case NFS3ERR_PERM: + ll = GF_LOG_WARNING; + break; + + case NFS3ERR_NOENT: + ll = GF_LOG_WARNING; + break; + + case NFS3ERR_ACCES: + ll = GF_LOG_WARNING; + break; + + case NFS3ERR_EXIST: + ll = GF_LOG_WARNING; + break; + + case NFS3ERR_XDEV: + ll = GF_LOG_WARNING; + break; + + case NFS3ERR_NODEV: + ll = GF_LOG_WARNING; + break; + + case NFS3ERR_IO: + ll = GF_LOG_WARNING; + break; + + case NFS3ERR_NXIO: + ll = GF_LOG_WARNING; + break; + + case NFS3ERR_NOTDIR: + ll = GF_LOG_WARNING; + break; + + case NFS3ERR_ISDIR: + ll = GF_LOG_WARNING; + break; + + case NFS3ERR_INVAL: + ll = GF_LOG_WARNING; + break; + + case NFS3ERR_NOSPC: + ll = GF_LOG_WARNING; + break; + + case NFS3ERR_ROFS: + ll = GF_LOG_WARNING; + break; + + case NFS3ERR_FBIG: + ll = GF_LOG_WARNING; + break; + + case NFS3ERR_MLINK: + ll = GF_LOG_WARNING; + break; + + case NFS3ERR_NAMETOOLONG: + ll = GF_LOG_WARNING; + break; + + case NFS3ERR_NOTEMPTY: + ll = GF_LOG_WARNING; + break; + + case NFS3ERR_SERVERFAULT: + ll = GF_LOG_WARNING; + break; + + case NFS3ERR_NOTSUPP: + ll = GF_LOG_WARNING; + break; + + case NFS3ERR_BADHANDLE: + ll = GF_LOG_WARNING; + break; + + case NFS3ERR_STALE: + ll = GF_LOG_WARNING; + break; + + case NFS3ERR_DQUOT: + ll = GF_LOG_WARNING; + break; + + default: + ll = GF_LOG_DEBUG; + break; + } + + return ll; +} + + +int +nfs3_setattr_loglevel (nfsstat3 stat) { + + int ll = GF_LOG_DEBUG; + + switch (stat) { + + case NFS3ERR_NOENT: + ll = GF_LOG_WARNING; + break; + + case NFS3ERR_EXIST: + ll = GF_LOG_WARNING; + break; + + case NFS3ERR_XDEV: + ll = GF_LOG_WARNING; + break; + + case NFS3ERR_NODEV: + ll = GF_LOG_WARNING; + break; + + case NFS3ERR_IO: + ll = GF_LOG_WARNING; + break; + + case NFS3ERR_NXIO: + ll = GF_LOG_WARNING; + break; + + case NFS3ERR_NOTDIR: + ll = GF_LOG_WARNING; + break; + + case NFS3ERR_ISDIR: + ll = GF_LOG_WARNING; + break; + + case NFS3ERR_INVAL: + ll = GF_LOG_WARNING; + break; + + case NFS3ERR_NOSPC: + ll = GF_LOG_WARNING; + break; + + case NFS3ERR_ROFS: + ll = GF_LOG_WARNING; + break; + + case NFS3ERR_FBIG: + ll = GF_LOG_WARNING; + break; + + case NFS3ERR_MLINK: + ll = GF_LOG_WARNING; + break; + + case NFS3ERR_NAMETOOLONG: + ll = GF_LOG_WARNING; + break; + + case NFS3ERR_NOTEMPTY: + ll = GF_LOG_WARNING; + break; + + case NFS3ERR_SERVERFAULT: + ll = GF_LOG_WARNING; + break; + + case NFS3ERR_NOTSUPP: + ll = GF_LOG_WARNING; + break; + + case NFS3ERR_BADHANDLE: + ll = GF_LOG_WARNING; + break; + + case NFS3ERR_STALE: + ll = GF_LOG_WARNING; + break; + + case NFS3ERR_DQUOT: + ll = GF_LOG_WARNING; + break; + + default: + ll = GF_LOG_DEBUG; + break; + } + + return ll; +} + + +int +nfs3_lookup_loglevel (nfsstat3 stat) { + + int ll = GF_LOG_DEBUG; + + switch (stat) { + + case NFS3ERR_PERM: + ll = GF_LOG_WARNING; + break; + + case NFS3ERR_ACCES: + ll = GF_LOG_WARNING; + break; + + case NFS3ERR_EXIST: + ll = GF_LOG_WARNING; + break; + + case NFS3ERR_XDEV: + ll = GF_LOG_WARNING; + break; + + case NFS3ERR_NODEV: + ll = GF_LOG_WARNING; + break; + + case NFS3ERR_IO: + ll = GF_LOG_WARNING; + break; + + case NFS3ERR_NXIO: + ll = GF_LOG_WARNING; + break; + + case NFS3ERR_NOTDIR: + ll = GF_LOG_WARNING; + break; + + case NFS3ERR_ISDIR: + ll = GF_LOG_WARNING; + break; + + case NFS3ERR_INVAL: + ll = GF_LOG_WARNING; + break; + + case NFS3ERR_NOSPC: + ll = GF_LOG_WARNING; + break; + + case NFS3ERR_ROFS: + ll = GF_LOG_WARNING; + break; + + case NFS3ERR_FBIG: + ll = GF_LOG_WARNING; + break; + + case NFS3ERR_MLINK: + ll = GF_LOG_WARNING; + break; + + case NFS3ERR_NAMETOOLONG: + ll = GF_LOG_WARNING; + break; + + case NFS3ERR_NOTEMPTY: + ll = GF_LOG_WARNING; + break; + + case NFS3ERR_SERVERFAULT: + ll = GF_LOG_WARNING; + break; + + case NFS3ERR_NOTSUPP: + ll = GF_LOG_WARNING; + break; + + case NFS3ERR_BADHANDLE: + ll = GF_LOG_WARNING; + break; + + case NFS3ERR_STALE: + ll = GF_LOG_WARNING; + break; + + case NFS3ERR_DQUOT: + ll = GF_LOG_WARNING; + break; + + default: + ll = GF_LOG_DEBUG; + break; + } + + return ll; +} + + +int +nfs3_access_loglevel (nfsstat3 stat) { + + int ll = GF_LOG_DEBUG; + + switch (stat) { + + case NFS3ERR_NOENT: + ll = GF_LOG_WARNING; + break; + + case NFS3ERR_EXIST: + ll = GF_LOG_WARNING; + break; + + case NFS3ERR_XDEV: + ll = GF_LOG_WARNING; + break; + + case NFS3ERR_NODEV: + ll = GF_LOG_WARNING; + break; + + case NFS3ERR_IO: + ll = GF_LOG_WARNING; + break; + + case NFS3ERR_NXIO: + ll = GF_LOG_WARNING; + break; + + case NFS3ERR_NOTDIR: + ll = GF_LOG_WARNING; + break; + + case NFS3ERR_ISDIR: + ll = GF_LOG_WARNING; + break; + + case NFS3ERR_INVAL: + ll = GF_LOG_WARNING; + break; + + case NFS3ERR_NOSPC: + ll = GF_LOG_WARNING; + break; + + case NFS3ERR_ROFS: + ll = GF_LOG_WARNING; + break; + + case NFS3ERR_FBIG: + ll = GF_LOG_WARNING; + break; + + case NFS3ERR_MLINK: + ll = GF_LOG_WARNING; + break; + + case NFS3ERR_NAMETOOLONG: + ll = GF_LOG_WARNING; + break; + + case NFS3ERR_NOTEMPTY: + ll = GF_LOG_WARNING; + break; + + case NFS3ERR_SERVERFAULT: + ll = GF_LOG_WARNING; + break; + + case NFS3ERR_NOTSUPP: + ll = GF_LOG_WARNING; + break; + + case NFS3ERR_BADHANDLE: + ll = GF_LOG_WARNING; + break; + + case NFS3ERR_STALE: + ll = GF_LOG_WARNING; + break; + + case NFS3ERR_DQUOT: + ll = GF_LOG_WARNING; + break; + + default: + ll = GF_LOG_DEBUG; + break; + } + + return ll; +} + + +int +nfs3_readlink_loglevel (nfsstat3 stat) { + + int ll = GF_LOG_DEBUG; + + switch (stat) { + + case NFS3ERR_EXIST: + ll = GF_LOG_WARNING; + break; + + case NFS3ERR_XDEV: + ll = GF_LOG_WARNING; + break; + + case NFS3ERR_NODEV: + ll = GF_LOG_WARNING; + break; + + case NFS3ERR_IO: + ll = GF_LOG_WARNING; + break; + + case NFS3ERR_NXIO: + ll = GF_LOG_WARNING; + break; + + case NFS3ERR_NOTDIR: + ll = GF_LOG_WARNING; + break; + + case NFS3ERR_ISDIR: + ll = GF_LOG_WARNING; + break; + + case NFS3ERR_INVAL: + ll = GF_LOG_WARNING; + break; + + case NFS3ERR_NOSPC: + ll = GF_LOG_WARNING; + break; + + case NFS3ERR_ROFS: + ll = GF_LOG_WARNING; + break; + + case NFS3ERR_FBIG: + ll = GF_LOG_WARNING; + break; + + case NFS3ERR_MLINK: + ll = GF_LOG_WARNING; + break; + + case NFS3ERR_NOTEMPTY: + ll = GF_LOG_WARNING; + break; + + case NFS3ERR_SERVERFAULT: + ll = GF_LOG_WARNING; + break; + + case NFS3ERR_NOTSUPP: + ll = GF_LOG_WARNING; + break; + + case NFS3ERR_BADHANDLE: + ll = GF_LOG_WARNING; + break; + + case NFS3ERR_STALE: + ll = GF_LOG_WARNING; + break; + + case NFS3ERR_DQUOT: + ll = GF_LOG_WARNING; + break; + + default: + ll = GF_LOG_DEBUG; + break; + } + + return ll; +} + +int +nfs3_read_loglevel (nfsstat3 stat) { + + int ll = GF_LOG_DEBUG; + + switch (stat) { + + case NFS3ERR_NOENT: + ll = GF_LOG_WARNING; + break; + + case NFS3ERR_EXIST: + ll = GF_LOG_WARNING; + break; + + case NFS3ERR_XDEV: + ll = GF_LOG_WARNING; + break; + + case NFS3ERR_NODEV: + ll = GF_LOG_WARNING; + break; + + case NFS3ERR_IO: + ll = GF_LOG_WARNING; + break; + + case NFS3ERR_NXIO: + ll = GF_LOG_WARNING; + break; + + case NFS3ERR_NOTDIR: + ll = GF_LOG_WARNING; + break; + + case NFS3ERR_ISDIR: + ll = GF_LOG_WARNING; + break; + + case NFS3ERR_INVAL: + ll = GF_LOG_WARNING; + break; + + case NFS3ERR_NOSPC: + ll = GF_LOG_WARNING; + break; + + case NFS3ERR_ROFS: + ll = GF_LOG_WARNING; + break; + + case NFS3ERR_FBIG: + ll = GF_LOG_WARNING; + break; + + case NFS3ERR_MLINK: + ll = GF_LOG_WARNING; + break; + + case NFS3ERR_NAMETOOLONG: + ll = GF_LOG_WARNING; + break; + + case NFS3ERR_NOTEMPTY: + ll = GF_LOG_WARNING; + break; + + case NFS3ERR_SERVERFAULT: + ll = GF_LOG_WARNING; + break; + + case NFS3ERR_NOTSUPP: + ll = GF_LOG_WARNING; + break; + + case NFS3ERR_BADHANDLE: + ll = GF_LOG_WARNING; + break; + + case NFS3ERR_STALE: + ll = GF_LOG_WARNING; + break; + + case NFS3ERR_DQUOT: + ll = GF_LOG_WARNING; + break; + + default: + ll = GF_LOG_DEBUG; + break; + } + + return ll; +} + + +int +nfs3_write_loglevel (nfsstat3 stat) { + + int ll = GF_LOG_DEBUG; + + switch (stat) { + + case NFS3ERR_NOENT: + ll = GF_LOG_WARNING; + break; + + case NFS3ERR_EXIST: + ll = GF_LOG_WARNING; + break; + + case NFS3ERR_XDEV: + ll = GF_LOG_WARNING; + break; + + case NFS3ERR_NODEV: + ll = GF_LOG_WARNING; + break; + + case NFS3ERR_IO: + ll = GF_LOG_WARNING; + break; + + case NFS3ERR_NXIO: + ll = GF_LOG_WARNING; + break; + + case NFS3ERR_NOTDIR: + ll = GF_LOG_WARNING; + break; + + case NFS3ERR_ISDIR: + ll = GF_LOG_WARNING; + break; + + case NFS3ERR_INVAL: + ll = GF_LOG_WARNING; + break; + + case NFS3ERR_NOSPC: + ll = GF_LOG_WARNING; + break; + + case NFS3ERR_ROFS: + ll = GF_LOG_WARNING; + break; + + case NFS3ERR_FBIG: + ll = GF_LOG_WARNING; + break; + + case NFS3ERR_MLINK: + ll = GF_LOG_WARNING; + break; + + case NFS3ERR_NAMETOOLONG: + ll = GF_LOG_WARNING; + break; + + case NFS3ERR_NOTEMPTY: + ll = GF_LOG_WARNING; + break; + + case NFS3ERR_SERVERFAULT: + ll = GF_LOG_WARNING; + break; + + case NFS3ERR_NOTSUPP: + ll = GF_LOG_WARNING; + break; + + case NFS3ERR_BADHANDLE: + ll = GF_LOG_WARNING; + break; + + case NFS3ERR_STALE: + ll = GF_LOG_WARNING; + break; + + case NFS3ERR_DQUOT: + ll = GF_LOG_WARNING; + break; + + default: + ll = GF_LOG_DEBUG; + break; + } + + return ll; +} + + +int +nfs3_create_loglevel (nfsstat3 stat) { + + int ll = GF_LOG_DEBUG; + + switch (stat) { + + case NFS3ERR_NOENT: + ll = GF_LOG_WARNING; + break; + + case NFS3ERR_EXIST: + ll = GF_LOG_WARNING; + break; + + case NFS3ERR_XDEV: + ll = GF_LOG_WARNING; + break; + + case NFS3ERR_NODEV: + ll = GF_LOG_WARNING; + break; + + case NFS3ERR_IO: + ll = GF_LOG_WARNING; + break; + + case NFS3ERR_NXIO: + ll = GF_LOG_WARNING; + break; + + case NFS3ERR_NOTDIR: + ll = GF_LOG_WARNING; + break; + + case NFS3ERR_ISDIR: + ll = GF_LOG_WARNING; + break; + + case NFS3ERR_INVAL: + ll = GF_LOG_WARNING; + break; + + case NFS3ERR_FBIG: + ll = GF_LOG_WARNING; + break; + + case NFS3ERR_MLINK: + ll = GF_LOG_WARNING; + break; + + case NFS3ERR_NOTEMPTY: + ll = GF_LOG_WARNING; + break; + + case NFS3ERR_SERVERFAULT: + ll = GF_LOG_WARNING; + break; + + case NFS3ERR_NOTSUPP: + ll = GF_LOG_WARNING; + break; + + case NFS3ERR_BADHANDLE: + ll = GF_LOG_WARNING; + break; + + case NFS3ERR_STALE: + ll = GF_LOG_WARNING; + break; + + default: + ll = GF_LOG_DEBUG; + break; + } + + return ll; +} + + +int +nfs3_mkdir_loglevel (nfsstat3 stat) { + + int ll = GF_LOG_DEBUG; + + switch (stat) { + + case NFS3ERR_NOENT: + ll = GF_LOG_WARNING; + break; + + case NFS3ERR_XDEV: + ll = GF_LOG_WARNING; + break; + + case NFS3ERR_NODEV: + ll = GF_LOG_WARNING; + break; + + case NFS3ERR_IO: + ll = GF_LOG_WARNING; + break; + + case NFS3ERR_NXIO: + ll = GF_LOG_WARNING; + break; + + case NFS3ERR_NOTDIR: + ll = GF_LOG_WARNING; + break; + + case NFS3ERR_ISDIR: + ll = GF_LOG_WARNING; + break; + + case NFS3ERR_INVAL: + ll = GF_LOG_WARNING; + break; + + case NFS3ERR_FBIG: + ll = GF_LOG_WARNING; + break; + + case NFS3ERR_MLINK: + ll = GF_LOG_WARNING; + break; + + case NFS3ERR_NOTEMPTY: + ll = GF_LOG_WARNING; + break; + + case NFS3ERR_SERVERFAULT: + ll = GF_LOG_WARNING; + break; + + case NFS3ERR_NOTSUPP: + ll = GF_LOG_WARNING; + break; + + case NFS3ERR_BADHANDLE: + ll = GF_LOG_WARNING; + break; + + case NFS3ERR_STALE: + ll = GF_LOG_WARNING; + break; + + case NFS3ERR_DQUOT: + ll = GF_LOG_WARNING; + break; + + default: + ll = GF_LOG_DEBUG; + break; + } + + return ll; +} + + +int +nfs3_symlink_loglevel (nfsstat3 stat) { + + int ll = GF_LOG_DEBUG; + + switch (stat) { + + case NFS3ERR_XDEV: + ll = GF_LOG_WARNING; + break; + + case NFS3ERR_NODEV: + ll = GF_LOG_WARNING; + break; + + case NFS3ERR_IO: + ll = GF_LOG_WARNING; + break; + + case NFS3ERR_NXIO: + ll = GF_LOG_WARNING; + break; + + case NFS3ERR_NOTDIR: + ll = GF_LOG_WARNING; + break; + + case NFS3ERR_ISDIR: + ll = GF_LOG_WARNING; + break; + + case NFS3ERR_INVAL: + ll = GF_LOG_WARNING; + break; + + case NFS3ERR_FBIG: + ll = GF_LOG_WARNING; + break; + + case NFS3ERR_MLINK: + ll = GF_LOG_WARNING; + break; + + case NFS3ERR_NOTEMPTY: + ll = GF_LOG_WARNING; + break; + + case NFS3ERR_SERVERFAULT: + ll = GF_LOG_WARNING; + break; + + case NFS3ERR_NOTSUPP: + ll = GF_LOG_WARNING; + break; + + case NFS3ERR_BADHANDLE: + ll = GF_LOG_WARNING; + break; + + case NFS3ERR_STALE: + ll = GF_LOG_WARNING; + break; + + case NFS3ERR_DQUOT: + ll = GF_LOG_WARNING; + break; + + default: + ll = GF_LOG_DEBUG; + break; + } + + return ll; +} + + +int +nfs3_mknod_loglevel (nfsstat3 stat) { + + int ll = GF_LOG_DEBUG; + + switch (stat) { + + case NFS3ERR_NOENT: + ll = GF_LOG_WARNING; + break; + + case NFS3ERR_XDEV: + ll = GF_LOG_WARNING; + break; + + case NFS3ERR_NODEV: + ll = GF_LOG_WARNING; + break; + + case NFS3ERR_IO: + ll = GF_LOG_WARNING; + break; + + case NFS3ERR_NXIO: + ll = GF_LOG_WARNING; + break; + + case NFS3ERR_NOTDIR: + ll = GF_LOG_WARNING; + break; + + case NFS3ERR_ISDIR: + ll = GF_LOG_WARNING; + break; + + case NFS3ERR_INVAL: + ll = GF_LOG_WARNING; + break; + + case NFS3ERR_FBIG: + ll = GF_LOG_WARNING; + break; + + case NFS3ERR_MLINK: + ll = GF_LOG_WARNING; + break; + + case NFS3ERR_NOTEMPTY: + ll = GF_LOG_WARNING; + break; + + case NFS3ERR_SERVERFAULT: + ll = GF_LOG_WARNING; + break; + + case NFS3ERR_NOTSUPP: + ll = GF_LOG_WARNING; + break; + + case NFS3ERR_BADHANDLE: + ll = GF_LOG_WARNING; + break; + + case NFS3ERR_STALE: + ll = GF_LOG_WARNING; + break; + + default: + ll = GF_LOG_DEBUG; + break; + } + + return ll; +} + +int +nfs3_remove_loglevel (nfsstat3 stat) { + + int ll = GF_LOG_DEBUG; + + switch (stat) { + + case NFS3ERR_EXIST: + ll = GF_LOG_WARNING; + break; + + case NFS3ERR_XDEV: + ll = GF_LOG_WARNING; + break; + + case NFS3ERR_NODEV: + ll = GF_LOG_WARNING; + break; + + case NFS3ERR_IO: + ll = GF_LOG_WARNING; + break; + + case NFS3ERR_NXIO: + ll = GF_LOG_WARNING; + break; + + case NFS3ERR_NOTDIR: + ll = GF_LOG_WARNING; + break; + + case NFS3ERR_INVAL: + ll = GF_LOG_WARNING; + break; + + case NFS3ERR_NOSPC: + ll = GF_LOG_WARNING; + break; + + case NFS3ERR_FBIG: + ll = GF_LOG_WARNING; + break; + + case NFS3ERR_MLINK: + ll = GF_LOG_WARNING; + break; + + case NFS3ERR_SERVERFAULT: + ll = GF_LOG_WARNING; + break; + + case NFS3ERR_NOTSUPP: + ll = GF_LOG_WARNING; + break; + + case NFS3ERR_BADHANDLE: + ll = GF_LOG_WARNING; + break; + + case NFS3ERR_STALE: + ll = GF_LOG_WARNING; + break; + + case NFS3ERR_DQUOT: + ll = GF_LOG_WARNING; + break; + + default: + ll = GF_LOG_DEBUG; + break; + } + + return ll; +} + + +int +nfs3_rmdir_loglevel (nfsstat3 stat) { + + int ll = GF_LOG_DEBUG; + + switch (stat) { + + case NFS3ERR_EXIST: + ll = GF_LOG_WARNING; + break; + + case NFS3ERR_XDEV: + ll = GF_LOG_WARNING; + break; + + case NFS3ERR_NODEV: + ll = GF_LOG_WARNING; + break; + + case NFS3ERR_IO: + ll = GF_LOG_WARNING; + break; + + case NFS3ERR_NXIO: + ll = GF_LOG_WARNING; + break; + + case NFS3ERR_NOTDIR: + ll = GF_LOG_WARNING; + break; + + case NFS3ERR_INVAL: + ll = GF_LOG_WARNING; + break; + + case NFS3ERR_NOSPC: + ll = GF_LOG_WARNING; + break; + + case NFS3ERR_FBIG: + ll = GF_LOG_WARNING; + break; + + case NFS3ERR_MLINK: + ll = GF_LOG_WARNING; + break; + + case NFS3ERR_SERVERFAULT: + ll = GF_LOG_WARNING; + break; + + case NFS3ERR_NOTSUPP: + ll = GF_LOG_WARNING; + break; + + case NFS3ERR_BADHANDLE: + ll = GF_LOG_WARNING; + break; + + case NFS3ERR_STALE: + ll = GF_LOG_WARNING; + break; + + case NFS3ERR_DQUOT: + ll = GF_LOG_WARNING; + break; + + default: + ll = GF_LOG_DEBUG; + break; + } + + return ll; +} + + +int +nfs3_rename_loglevel (nfsstat3 stat) { + + int ll = GF_LOG_DEBUG; + + switch (stat) { + + case NFS3ERR_XDEV: + ll = GF_LOG_WARNING; + break; + + case NFS3ERR_NODEV: + ll = GF_LOG_WARNING; + break; + + case NFS3ERR_IO: + ll = GF_LOG_WARNING; + break; + + case NFS3ERR_NXIO: + ll = GF_LOG_WARNING; + break; + + case NFS3ERR_NOTDIR: + ll = GF_LOG_WARNING; + break; + + case NFS3ERR_ISDIR: + ll = GF_LOG_WARNING; + break; + + case NFS3ERR_INVAL: + ll = GF_LOG_WARNING; + break; + + case NFS3ERR_NOSPC: + ll = GF_LOG_WARNING; + break; + + case NFS3ERR_FBIG: + ll = GF_LOG_WARNING; + break; + + case NFS3ERR_MLINK: + ll = GF_LOG_WARNING; + break; + + case NFS3ERR_NOTEMPTY: + ll = GF_LOG_WARNING; + break; + + case NFS3ERR_SERVERFAULT: + ll = GF_LOG_WARNING; + break; + + case NFS3ERR_NOTSUPP: + ll = GF_LOG_WARNING; + break; + + case NFS3ERR_BADHANDLE: + ll = GF_LOG_WARNING; + break; + + case NFS3ERR_STALE: + ll = GF_LOG_WARNING; + break; + + case NFS3ERR_DQUOT: + ll = GF_LOG_WARNING; + break; + + default: + ll = GF_LOG_DEBUG; + break; + } + + return ll; +} + + +int +nfs3_link_loglevel (nfsstat3 stat) { + + int ll = GF_LOG_DEBUG; + + switch (stat) { + + case NFS3ERR_XDEV: + ll = GF_LOG_WARNING; + break; + + case NFS3ERR_NODEV: + ll = GF_LOG_WARNING; + break; + + case NFS3ERR_IO: + ll = GF_LOG_WARNING; + break; + + case NFS3ERR_NXIO: + ll = GF_LOG_WARNING; + break; + + case NFS3ERR_INVAL: + ll = GF_LOG_WARNING; + break; + + case NFS3ERR_FBIG: + ll = GF_LOG_WARNING; + break; + + case NFS3ERR_MLINK: + ll = GF_LOG_WARNING; + break; + + case NFS3ERR_NOTEMPTY: + ll = GF_LOG_WARNING; + break; + + case NFS3ERR_SERVERFAULT: + ll = GF_LOG_WARNING; + break; + + case NFS3ERR_NOTSUPP: + ll = GF_LOG_WARNING; + break; + + case NFS3ERR_BADHANDLE: + ll = GF_LOG_WARNING; + break; + + case NFS3ERR_STALE: + ll = GF_LOG_WARNING; + break; + + case NFS3ERR_DQUOT: + ll = GF_LOG_WARNING; + break; + + default: + ll = GF_LOG_DEBUG; + break; + } + + return ll; +} + + +int +nfs3_readdir_loglevel (nfsstat3 stat) { + + int ll = GF_LOG_DEBUG; + + switch (stat) { + + case NFS3ERR_NOENT: + ll = GF_LOG_WARNING; + break; + + case NFS3ERR_EXIST: + ll = GF_LOG_WARNING; + break; + + case NFS3ERR_XDEV: + ll = GF_LOG_WARNING; + break; + + case NFS3ERR_NODEV: + ll = GF_LOG_WARNING; + break; + + case NFS3ERR_IO: + ll = GF_LOG_WARNING; + break; + + case NFS3ERR_NXIO: + ll = GF_LOG_WARNING; + break; + + case NFS3ERR_NOTDIR: + ll = GF_LOG_WARNING; + break; + + case NFS3ERR_ISDIR: + ll = GF_LOG_WARNING; + break; + + case NFS3ERR_INVAL: + ll = GF_LOG_WARNING; + break; + + case NFS3ERR_NOSPC: + ll = GF_LOG_WARNING; + break; + + case NFS3ERR_ROFS: + ll = GF_LOG_WARNING; + break; + + case NFS3ERR_FBIG: + ll = GF_LOG_WARNING; + break; + + case NFS3ERR_MLINK: + ll = GF_LOG_WARNING; + break; + + case NFS3ERR_NAMETOOLONG: + ll = GF_LOG_WARNING; + break; + + case NFS3ERR_NOTEMPTY: + ll = GF_LOG_WARNING; + break; + + case NFS3ERR_SERVERFAULT: + ll = GF_LOG_WARNING; + break; + + case NFS3ERR_NOTSUPP: + ll = GF_LOG_WARNING; + break; + + case NFS3ERR_BADHANDLE: + ll = GF_LOG_WARNING; + break; + + case NFS3ERR_STALE: + ll = GF_LOG_WARNING; + break; + + case NFS3ERR_DQUOT: + ll = GF_LOG_WARNING; + break; + + default: + ll = GF_LOG_DEBUG; + break; + } + + return ll; +} + + +int +nfs3_fsstat_loglevel (nfsstat3 stat) { + + int ll = GF_LOG_DEBUG; + + switch (stat) { + + case NFS3ERR_PERM: + ll = GF_LOG_WARNING; + break; + + case NFS3ERR_NOENT: + ll = GF_LOG_WARNING; + break; + + case NFS3ERR_ACCES: + ll = GF_LOG_WARNING; + break; + + case NFS3ERR_EXIST: + ll = GF_LOG_WARNING; + break; + + case NFS3ERR_XDEV: + ll = GF_LOG_WARNING; + break; + + case NFS3ERR_NODEV: + ll = GF_LOG_WARNING; + break; + + case NFS3ERR_IO: + ll = GF_LOG_WARNING; + break; + + case NFS3ERR_NXIO: + ll = GF_LOG_WARNING; + break; + + case NFS3ERR_NOTDIR: + ll = GF_LOG_WARNING; + break; + + case NFS3ERR_ISDIR: + ll = GF_LOG_WARNING; + break; + + case NFS3ERR_INVAL: + ll = GF_LOG_WARNING; + break; + + case NFS3ERR_NOSPC: + ll = GF_LOG_WARNING; + break; + + case NFS3ERR_ROFS: + ll = GF_LOG_WARNING; + break; + + case NFS3ERR_FBIG: + ll = GF_LOG_WARNING; + break; + + case NFS3ERR_MLINK: + ll = GF_LOG_WARNING; + break; + + case NFS3ERR_NAMETOOLONG: + ll = GF_LOG_WARNING; + break; + + case NFS3ERR_NOTEMPTY: + ll = GF_LOG_WARNING; + break; + + case NFS3ERR_SERVERFAULT: + ll = GF_LOG_WARNING; + break; + + case NFS3ERR_NOTSUPP: + ll = GF_LOG_WARNING; + break; + + case NFS3ERR_BADHANDLE: + ll = GF_LOG_WARNING; + break; + + case NFS3ERR_STALE: + ll = GF_LOG_WARNING; + break; + + case NFS3ERR_DQUOT: + ll = GF_LOG_WARNING; + break; + + default: + ll = GF_LOG_DEBUG; + break; + } + + return ll; +} + +struct nfs3op_str { + int op; + char str[100]; +}; + +struct nfs3op_str nfs3op_strings[] = { + { NFS3_NULL, "NULL"}, + { NFS3_GETATTR, "GETATTR"}, + { NFS3_SETATTR, "SETATTR"}, + { NFS3_LOOKUP, "LOOKUP"}, + { NFS3_ACCESS, "ACCESS"}, + { NFS3_READLINK, "READLINK"}, + { NFS3_READ, "READ"}, + { NFS3_WRITE, "WRITE"}, + { NFS3_CREATE, "CREATE"}, + { NFS3_MKDIR, "MKDIR"}, + { NFS3_SYMLINK, "SYMLINK"}, + { NFS3_MKNOD, "MKNOD"}, + { NFS3_REMOVE, "REMOVE"}, + { NFS3_RMDIR, "RMDIR"}, + { NFS3_RENAME, "RENAME"}, + { NFS3_LINK, "LINK"}, + { NFS3_READDIR, "READDIR"}, + { NFS3_READDIRP, "READDIRP"}, + { NFS3_FSSTAT, "FSSTAT"}, + { NFS3_FSINFO, "FSINFO"}, + { NFS3_PATHCONF, "PATHCONF"}, + { NFS3_COMMIT, "COMMIT"}, +}; + +int +nfs3_loglevel (int nfs_op, nfsstat3 stat) { + + int ll = GF_LOG_DEBUG; + + switch (nfs_op) { + case NFS3_GETATTR: + ll = nfs3_getattr_loglevel (stat); + break; + + case NFS3_SETATTR: + ll = nfs3_setattr_loglevel (stat); + break; + + case NFS3_LOOKUP: + ll = nfs3_lookup_loglevel (stat); + break; + + case NFS3_ACCESS: + ll = nfs3_access_loglevel (stat); + break; + + case NFS3_READLINK: + ll = nfs3_readlink_loglevel (stat); + break; + + case NFS3_READ: + ll = nfs3_read_loglevel (stat); + break; + + case NFS3_WRITE: + ll = nfs3_write_loglevel (stat); + break; + + case NFS3_CREATE: + ll = nfs3_create_loglevel (stat); + break; + + case NFS3_MKDIR: + ll = nfs3_mkdir_loglevel (stat); + break; + + case NFS3_SYMLINK: + ll = nfs3_symlink_loglevel (stat); + break; + + case NFS3_MKNOD: + ll = nfs3_mknod_loglevel (stat); + break; + + case NFS3_REMOVE: + ll = nfs3_remove_loglevel (stat); + break; + + case NFS3_RMDIR: + ll = nfs3_rmdir_loglevel (stat); + break; + + case NFS3_RENAME: + ll = nfs3_rename_loglevel (stat); + break; + + case NFS3_LINK: + ll = nfs3_link_loglevel (stat); + break; + + case NFS3_READDIR: + ll = nfs3_readdir_loglevel (stat); + break; + + case NFS3_READDIRP: + ll = nfs3_readdir_loglevel (stat); + break; + + case NFS3_FSSTAT: + ll = nfs3_fsstat_loglevel (stat); + break; + + case NFS3_FSINFO: + ll = nfs3_fsstat_loglevel (stat); + break; + + case NFS3_PATHCONF: + ll = nfs3_fsstat_loglevel (stat); + break; + + case NFS3_COMMIT: + ll = nfs3_write_loglevel (stat); + break; + + default: + ll = GF_LOG_DEBUG; + break; + } + + return ll; +} + +void +nfs3_log_common_res (uint32_t xid, int op, nfsstat3 stat, int pstat) +{ + char errstr[1024]; + int ll = nfs3_loglevel (op, stat); + + if (THIS->ctx->log.loglevel < ll) + return; + nfs3_stat_to_errstr (xid, nfs3op_strings[op].str, stat, pstat, errstr, sizeof (errstr)); + gf_log (GF_NFS3, ll, "%s", errstr); +} + +void +nfs3_log_readlink_res (uint32_t xid, nfsstat3 stat, int pstat, char *linkpath) +{ + char errstr[1024]; + int ll = nfs3_loglevel (NFS3_READLINK, stat); + + if (THIS->ctx->log.loglevel < ll) + return; + + nfs3_stat_to_errstr (xid, "READLINK", stat, pstat, errstr, sizeof (errstr)); + gf_log (GF_NFS3, ll, "%s, target: %s", + errstr, linkpath); + +} + +void +nfs3_log_read_res (uint32_t xid, nfsstat3 stat, int pstat, count3 count, + int is_eof, struct iovec *vec, int32_t veccount) +{ + char errstr[1024]; + int ll = GF_LOG_DEBUG; + + ll = nfs3_loglevel (NFS3_READ, stat); + if (THIS->ctx->log.loglevel < ll) + return; + nfs3_stat_to_errstr (xid, "READ", stat, pstat, errstr, sizeof (errstr)); + if (vec) + gf_log (GF_NFS3, ll, "%s, count: %"PRIu32", is_eof:" + " %d, vector: count: %d, len: %zd", errstr, count, + is_eof, veccount, vec->iov_len); + else + gf_log (GF_NFS3, ll, "%s, count: %"PRIu32", is_eof:" + " %d", errstr, count, is_eof); +} + + +void +nfs3_log_write_res (uint32_t xid, nfsstat3 stat, int pstat, count3 count, + int stable, uint64_t wverf) +{ + char errstr[1024]; + int ll = nfs3_loglevel (NFS3_WRITE, stat); + + if (THIS->ctx->log.loglevel < ll) + return; + + nfs3_stat_to_errstr (xid, "WRITE", stat, pstat, errstr, sizeof (errstr)); + gf_log (GF_NFS3, ll, "%s, count: %"PRIu32", %s,wverf: %"PRIu64 + , errstr, count, (stable == UNSTABLE)?"UNSTABLE":"STABLE", + wverf); +} + + +void +nfs3_log_newfh_res (uint32_t xid, int op, nfsstat3 stat, int pstat, + struct nfs3_fh *newfh) +{ + char errstr[1024]; + char fhstr[1024]; + int ll = nfs3_loglevel (op, stat); + + if (THIS->ctx->log.loglevel < ll) + return; + nfs3_stat_to_errstr (xid, nfs3op_strings[op].str, stat, pstat, errstr, sizeof (errstr)); + nfs3_fh_to_str (newfh, fhstr, sizeof (fhstr)); + + gf_log (GF_NFS3, nfs3_loglevel (op, stat), "%s, %s", errstr, fhstr); +} + + +void +nfs3_log_readdir_res (uint32_t xid, nfsstat3 stat, int pstat, uint64_t cverf, + count3 count, int is_eof) +{ + char errstr[1024]; + int ll = nfs3_loglevel (NFS3_READDIR, stat); + + if (THIS->ctx->log.loglevel < ll) + return; + nfs3_stat_to_errstr (xid, "READDIR", stat, pstat, errstr, sizeof (errstr)); + gf_log (GF_NFS3, ll, "%s, count: %"PRIu32", cverf: %"PRIu64 + ", is_eof: %d", errstr, count, cverf, is_eof); +} + + +void +nfs3_log_readdirp_res (uint32_t xid, nfsstat3 stat, int pstat, uint64_t cverf, + count3 dircount, count3 maxcount, int is_eof) +{ + char errstr[1024]; + int ll = nfs3_loglevel (NFS3_READDIRP, stat); + + if (THIS->ctx->log.loglevel < ll) + return; + nfs3_stat_to_errstr (xid, "READDIRPLUS", stat, pstat, errstr, sizeof (errstr)); + gf_log (GF_NFS3, ll, "%s, dircount: %"PRIu32", maxcount: %" + PRIu32", cverf: %"PRIu64", is_eof: %d", errstr, dircount, + maxcount, cverf, is_eof); +} + + +void +nfs3_log_commit_res (uint32_t xid, nfsstat3 stat, int pstat, uint64_t wverf) +{ + char errstr[1024]; + int ll = nfs3_loglevel (NFS3_COMMIT, stat); + + if (THIS->ctx->log.loglevel < ll) + return; + nfs3_stat_to_errstr (xid, "COMMIT", stat, pstat, errstr, sizeof (errstr)); + gf_log (GF_NFS3, ll, "%s, wverf: %"PRIu64, errstr, wverf); +} + + +void +nfs3_log_readdir_call (uint32_t xid, struct nfs3_fh *fh, count3 dircount, + count3 maxcount) +{ + char fhstr[1024]; + + if (THIS->ctx->log.loglevel < GF_LOG_DEBUG) + return; + + nfs3_fh_to_str (fh, fhstr, sizeof (fhstr)); + + if (maxcount == 0) + gf_log (GF_NFS3, GF_LOG_DEBUG, "XID: %x, READDIR: args: %s," + " count: %d", xid, fhstr, (uint32_t)dircount); + else + gf_log (GF_NFS3, GF_LOG_DEBUG, "XID: %x, READDIRPLUS: args: %s," + " dircount: %d, maxcount: %d", xid, fhstr, + (uint32_t)dircount, (uint32_t)maxcount); +} + + +int +nfs3_fh_resolve_inode_done (nfs3_call_state_t *cs, inode_t *inode) +{ + int ret = -EFAULT; + + if ((!cs) || (!inode)) + return ret; + + gf_log (GF_NFS3, GF_LOG_TRACE, "FH inode resolved"); + ret = nfs_inode_loc_fill (inode, &cs->resolvedloc, NFS_RESOLVE_EXIST); + if (ret < 0) { + gf_log (GF_NFS3, GF_LOG_ERROR, "inode loc fill failed"); + goto err; + } + + nfs3_call_resume (cs); + +err: + return ret; +} + + +int32_t +nfs3_fh_resolve_entry_lookup_cbk (call_frame_t *frame, void *cookie, + xlator_t *this, int32_t op_ret, + int32_t op_errno, inode_t *inode, + struct iatt *buf, dict_t *xattr, + struct iatt *postparent) +{ + nfs3_call_state_t *cs = NULL; + inode_t *linked_inode = NULL; + + cs = frame->local; + cs->resolve_ret = op_ret; + cs->resolve_errno = op_errno; + + if (op_ret == -1) { + gf_log (GF_NFS3, (op_errno == ENOENT ? GF_LOG_TRACE : GF_LOG_ERROR), + "Lookup failed: %s: %s", + cs->resolvedloc.path, strerror (op_errno)); + goto err; + } else + gf_log (GF_NFS3, GF_LOG_TRACE, "Entry looked up: %s", + cs->resolvedloc.path); + + memcpy (&cs->stbuf, buf, sizeof (*buf)); + memcpy (&cs->postparent, postparent, sizeof (*postparent)); + linked_inode = inode_link (inode, cs->resolvedloc.parent, + cs->resolvedloc.name, buf); + if (linked_inode) { + inode_lookup (linked_inode); + inode_unref (cs->resolvedloc.inode); + cs->resolvedloc.inode = linked_inode; + } +err: + nfs3_call_resume (cs); + return 0; +} + + +int32_t +nfs3_fh_resolve_inode_lookup_cbk (call_frame_t *frame, void *cookie, + xlator_t *this, int32_t op_ret, + int32_t op_errno, inode_t *inode, + struct iatt *buf, dict_t *xattr, + struct iatt *postparent) +{ + nfs3_call_state_t *cs = NULL; + inode_t *linked_inode = NULL; + + cs = frame->local; + cs->resolve_ret = op_ret; + cs->resolve_errno = op_errno; + + if (op_ret == -1) { + gf_log (GF_NFS3, (op_errno == ENOENT ? GF_LOG_TRACE : GF_LOG_ERROR), + "Lookup failed: %s: %s", + cs->resolvedloc.path, strerror (op_errno)); + nfs3_call_resume (cs); + goto err; + } + + memcpy (&cs->stbuf, buf, sizeof(*buf)); + memcpy (&cs->postparent, buf, sizeof(*postparent)); + linked_inode = inode_link (inode, cs->resolvedloc.parent, + cs->resolvedloc.name, buf); + if (linked_inode) { + inode_lookup (linked_inode); + inode_unref (cs->resolvedloc.inode); + cs->resolvedloc.inode = linked_inode; + } + + /* If it is an entry lookup and we landed in the callback for hard + * inode resolution, it means the parent inode was not available and + * had to be resolved first. Now that is done, lets head back into + * entry resolution. + */ + if (cs->resolventry) + nfs3_fh_resolve_entry_hard (cs); + else + nfs3_call_resume (cs); +err: + return 0; +} + + + +/* Needs no extra argument since it knows that the fh to be resolved is in + * resolvefh and that it needs to start looking from the root. + */ +int +nfs3_fh_resolve_inode_hard (nfs3_call_state_t *cs) +{ + int ret = -EFAULT; + nfs_user_t nfu = {0, }; + + if (!cs) + return ret; + + gf_log (GF_NFS3, GF_LOG_TRACE, "FH hard resolution for: gfid 0x%s", + uuid_utoa (cs->resolvefh.gfid)); + cs->hardresolved = 1; + nfs_loc_wipe (&cs->resolvedloc); + ret = nfs_gfid_loc_fill (cs->vol->itable, cs->resolvefh.gfid, + &cs->resolvedloc, NFS_RESOLVE_CREATE); + if (ret < 0) { + gf_log (GF_NFS3, GF_LOG_ERROR, "Failed to fill loc using gfid: " + "%s", strerror (-ret)); + goto out; + } + + nfs_user_root_create (&nfu); + ret = nfs_lookup (cs->nfsx, cs->vol, &nfu, &cs->resolvedloc, + nfs3_fh_resolve_inode_lookup_cbk, cs); + +out: + return ret; +} + + +int +nfs3_fh_resolve_entry_hard (nfs3_call_state_t *cs) +{ + int ret = -EFAULT; + nfs_user_t nfu = {0, }; + + if (!cs) + return ret; + + nfs_loc_wipe (&cs->resolvedloc); + nfs_user_root_create (&nfu); + gf_log (GF_NFS3, GF_LOG_TRACE, "FH hard resolution: gfid: %s " + ", entry: %s", uuid_utoa (cs->resolvefh.gfid), + cs->resolventry); + + ret = nfs_entry_loc_fill (cs->vol->itable, cs->resolvefh.gfid, + cs->resolventry, &cs->resolvedloc, + NFS_RESOLVE_CREATE); + + if (ret == -2) { + gf_log (GF_NFS3, GF_LOG_TRACE, "Entry needs lookup: %s", + cs->resolvedloc.path); + /* If the NFS op is lookup, let the resume callback + * handle the sending of the lookup fop. Similarly, + * if the NFS op is create, let the create call + * go ahead in the resume callback so that an EEXIST gets + * handled at posix without an extra fop at this point. + */ + if (nfs3_lookup_op (cs) || + (nfs3_create_op (cs) && !nfs3_create_exclusive_op (cs))) { + cs->lookuptype = GF_NFS3_FRESH; + cs->resolve_ret = 0; + nfs3_call_resume (cs); + } else { + cs->hardresolved = 1; + nfs_lookup (cs->nfsx, cs->vol, &nfu, &cs->resolvedloc, + nfs3_fh_resolve_entry_lookup_cbk, cs); + } + ret = 0; + } else if (ret == -1) { + gf_log (GF_NFS3, GF_LOG_TRACE, "Entry needs parent lookup: %s", + cs->resolvedloc.path); + ret = nfs3_fh_resolve_inode_hard (cs); + } else if (ret == 0) { + cs->resolve_ret = 0; + nfs3_call_resume (cs); + } + + return ret; +} + +int +nfs3_fh_resolve_inode (nfs3_call_state_t *cs) +{ + inode_t *inode = NULL; + int ret = -EFAULT; + + if (!cs) + return ret; + + gf_log (GF_NFS3, GF_LOG_TRACE, "FH needs inode resolution"); + uuid_copy (cs->resolvedloc.gfid, cs->resolvefh.gfid); + inode = inode_find (cs->vol->itable, cs->resolvefh.gfid); + if (!inode) + ret = nfs3_fh_resolve_inode_hard (cs); + else + ret = nfs3_fh_resolve_inode_done (cs, inode); + + if (inode) + inode_unref (inode); + + return ret; +} + +int +nfs3_fh_resolve_entry (nfs3_call_state_t *cs) +{ + int ret = -EFAULT; + + if (!cs) + return ret; + + return nfs3_fh_resolve_entry_hard (cs); +} + + +int +nfs3_fh_resolve_resume (nfs3_call_state_t *cs) +{ + int ret = -EFAULT; + + if (!cs) + return ret; + + if (cs->resolve_ret < 0) + goto err_resume_call; + + if (!cs->resolventry) + ret = nfs3_fh_resolve_inode (cs); + else + ret = nfs3_fh_resolve_entry (cs); + +err_resume_call: + if (ret < 0) { + cs->resolve_ret = -1; + cs->resolve_errno = EFAULT; + nfs3_call_resume (cs); + ret = 0; + } + + return ret; +} + + +int32_t +nfs3_fh_resolve_root_lookup_cbk (call_frame_t *frame, void *cookie, + xlator_t *this, int32_t op_ret, + int32_t op_errno, inode_t *inode, + struct iatt *buf, dict_t *xattr, + struct iatt *postparent) +{ + nfs3_call_state_t *cs = NULL; + + cs = frame->local; + cs->resolve_ret = op_ret; + cs->resolve_errno = op_errno; + + if (op_ret == -1) { + gf_log (GF_NFS3, GF_LOG_ERROR, "Root lookup failed: %s", + strerror (op_errno)); + goto err; + } else + gf_log (GF_NFS3, GF_LOG_TRACE, "Root looked up: %s", + cs->resolvedloc.path); + + nfs3_set_root_looked_up (cs->nfs3state, &cs->resolvefh); +err: + nfs3_fh_resolve_resume (cs); + return 0; +} + + +int +nfs3_fh_resolve_root (nfs3_call_state_t *cs) +{ + int ret = -EFAULT; + nfs_user_t nfu = {0, }; + + if (!cs) + return ret; + + if (nfs3_is_root_looked_up (cs->nfs3state, &cs->resolvefh)) { + ret = nfs3_fh_resolve_resume (cs); + goto out; + } + + nfs_user_root_create (&nfu); + gf_log (GF_NFS3, GF_LOG_TRACE, "Root needs lookup"); + ret = nfs_root_loc_fill (cs->vol->itable, &cs->resolvedloc); + if (ret < 0) { + gf_log (GF_NFS3, GF_LOG_ERROR, "Failed to lookup root from itable: %s", + strerror (-ret)); + goto out; + } + + ret = nfs_lookup (cs->nfsx, cs->vol, &nfu, &cs->resolvedloc, + nfs3_fh_resolve_root_lookup_cbk, cs); + +out: + return ret; +} + + +int +nfs3_fh_resolve_and_resume (nfs3_call_state_t *cs, struct nfs3_fh *fh, + char *entry, nfs3_resume_fn_t resum_fn) +{ + int ret = -EFAULT; + + if ((!cs) || (!fh)) + return ret; + + cs->resume_fn = resum_fn; + cs->resolvefh = *fh; + cs->hashidx = 0; + + /* Check if the resolution is: + * a. fh resolution + * + * or + * + * b. (fh, basename) resolution + */ + if (entry) { /* b */ + cs->resolventry = gf_strdup (entry); + if (!cs->resolventry) + goto err; + } + + ret = nfs3_fh_resolve_root (cs); +err: + return ret; +} diff --git a/xlators/nfs/server/src/nfs3-helpers.h b/xlators/nfs/server/src/nfs3-helpers.h new file mode 100644 index 000000000..4de1d5623 --- /dev/null +++ b/xlators/nfs/server/src/nfs3-helpers.h @@ -0,0 +1,337 @@ +/* + Copyright (c) 2010-2011 Gluster, Inc. <http://www.gluster.com> + This file is part of GlusterFS. + + This file is licensed to you under your choice of the GNU Lesser + General Public License, version 3 or any later version (LGPLv3 or + later), or the GNU General Public License, version 2 (GPLv2), in all + cases as published by the Free Software Foundation. +*/ + +#ifndef _NFS3_HELPER_H_ +#define _NFS3_HELPER_H_ +#ifndef _CONFIG_H +#define _CONFIG_H +#include "config.h" +#endif + + +#include "xlator.h" +#include "nfs3.h" +#include "nfs3-fh.h" +#include "msg-nfs3.h" +#include "xdr-nfs3.h" + +#include <sys/statvfs.h> + +#define GF_NFS3_FD_CACHED 0xcaced + +extern struct nfs3_fh +nfs3_extract_lookup_fh (lookup3args *args); + +extern char * +nfs3_extract_lookup_name (lookup3args *args); + +extern nfsstat3 +nfs3_errno_to_nfsstat3 (int errnum); + +extern nfsstat3 +nfs3_cbk_errno_status (int32_t, int32_t); + +extern void +nfs3_fill_lookup3res (lookup3res *res, nfsstat3 stat, struct nfs3_fh *newfh, + struct iatt *stbuf, struct iatt *postparent, + uint64_t deviceid); + +extern post_op_attr +nfs3_stat_to_post_op_attr (struct iatt *buf); + +extern struct nfs3_fh +nfs3_extract_getattr_fh (getattr3args *args); + +extern void +nfs3_fill_getattr3res (getattr3res *res, nfsstat3 stat, struct iatt *buf, + uint64_t deviceid); + +extern struct nfs3_fh +nfs3_extract_fsinfo_fh (fsinfo3args *args); + +extern void +nfs3_fill_fsinfo3res (struct nfs3_state *nfs3, fsinfo3res *res, + nfsstat3 status, struct iatt *fsroot,uint64_t deviceid); + +/* Functions containing _prep_ are used specifically to work around + * the memory allocations that happen inside Sun RPC library. + * In that library, there are numerous places where every NFS request + * can result in really tiny malloc calls. I fear the memory fragmentation + * that will follow. After studying the points at and the way in which malloc + * is called in Sun RPC, I've come up with this work-around. It is based on + * the idea that if the user/caller of the xdr_to_XXXXargs functions can provide + * already allocated memory or provide references to memory areas on its stack + * just for the short-term purpose of decoding the message from XDR format, we + * can avoid the memory allocations in Sun RPC. This is based on the fact + * that Sun RPC first checks whether structure members which require memory + * are NULL or not and only then calls malloc. In this case, if the caller + * provided references are non-NULL, then the if-branches containing malloc + * in Sun RPC will be avoided. + * PS: You're not expected to understand this unless you've spent some time + * looking through the glibc/sunrpc sources. + */ +extern void +nfs3_prep_lookup3args (lookup3args *args, struct nfs3_fh *fh, char *name); + +extern void +nfs3_prep_getattr3args (getattr3args *args, struct nfs3_fh *fh); + +extern void +nfs3_prep_fsinfo3args (fsinfo3args *args, struct nfs3_fh *root); + +extern char * +nfsstat3_strerror(int stat); + +extern void +nfs3_prep_access3args (access3args *args, struct nfs3_fh *fh); + +extern void +nfs3_fill_access3res (access3res *res, nfsstat3 status, int32_t accbits, + int32_t reqaccbits); + +extern char * +nfs3_fhcache_getpath (struct nfs3_state *nfs3, struct nfs3_fh *fh); + +extern int +nfs3_fhcache_putpath (struct nfs3_state *nfs3, struct nfs3_fh *fh, char *path); + +extern void +nfs3_prep_readdir3args (readdir3args *ra, struct nfs3_fh *fh); + +extern void +nfs3_fill_readdir3res (readdir3res *res, nfsstat3 stat, struct nfs3_fh *dfh, + uint64_t cverf, struct iatt *dirstat, + gf_dirent_t *entries, count3 count, int is_eof, + uint64_t deviceid); + +extern void +nfs3_prep_readdirp3args (readdirp3args *ra, struct nfs3_fh *fh); + +extern void +nfs3_fill_readdirp3res (readdirp3res *res, nfsstat3 stat, + struct nfs3_fh *dirfh, uint64_t cverf, + struct iatt *dirstat, gf_dirent_t *entries, + count3 dircount, count3 maxcount, int is_eof, + uint64_t deviceid); + +extern void +nfs3_free_readdirp3res (readdirp3res *res); + +extern void +nfs3_free_readdir3res (readdir3res *res); + +extern void +nfs3_prep_fsstat3args (fsstat3args *args, struct nfs3_fh *fh); + +extern void +nfs3_fill_fsstat3res (fsstat3res *res, nfsstat3 stat, struct statvfs *fsbuf, + struct iatt *postbuf, uint64_t deviceid); + +extern int32_t +nfs3_sattr3_to_setattr_valid (sattr3 *sattr, struct iatt *buf, mode_t *omode); +extern void +nfs3_fill_create3res (create3res *res, nfsstat3 stat, struct nfs3_fh *newfh, + struct iatt *newbuf, struct iatt *preparent, + struct iatt *postparent, uint64_t deviceid); + +extern void +nfs3_prep_create3args (create3args *args, struct nfs3_fh *fh, char *name); + +extern void +nfs3_prep_setattr3args (setattr3args *args, struct nfs3_fh *fh); + +extern void +nfs3_fill_setattr3res (setattr3res *res, nfsstat3 stat, struct iatt *preop, + struct iatt *postop, uint64_t deviceid); + +extern void +nfs3_prep_mkdir3args (mkdir3args *args, struct nfs3_fh *dirfh, char *name); + +extern void +nfs3_fill_mkdir3res (mkdir3res *res, nfsstat3 stat, struct nfs3_fh *fh, + struct iatt *buf, struct iatt *preparent, + struct iatt *postparent, uint64_t deviceid); + +extern void +nfs3_prep_symlink3args (symlink3args *args, struct nfs3_fh *dirfh, char *name, + char *target); + +extern void +nfs3_fill_symlink3res (symlink3res *res, nfsstat3 stat, struct nfs3_fh *fh, + struct iatt *buf, struct iatt *preparent, + struct iatt *postparent, uint64_t deviceid); + +extern void +nfs3_prep_readlink3args (readlink3args *args, struct nfs3_fh *fh); + +extern void +nfs3_fill_readlink3res (readlink3res *res, nfsstat3 stat, char *path, + struct iatt *buf, uint64_t deviceid); + +extern void +nfs3_prep_mknod3args (mknod3args *args, struct nfs3_fh *fh, char *name); + +extern void +nfs3_fill_mknod3res (mknod3res *res, nfsstat3 stat, struct nfs3_fh *fh, + struct iatt *buf, struct iatt *preparent, + struct iatt *postparent, uint64_t deviceid); + +extern void +nfs3_fill_remove3res (remove3res *res, nfsstat3 stat, struct iatt *preparent, + struct iatt *postparent, uint64_t deviceid); +extern void +nfs3_prep_remove3args (remove3args *args, struct nfs3_fh *fh, char *name); + +extern void +nfs3_fill_rmdir3res (rmdir3res *res, nfsstat3 stat, struct iatt *preparent, + struct iatt *postparent, uint64_t deviceid); + +extern void +nfs3_prep_rmdir3args (rmdir3args *args, struct nfs3_fh *fh, char *name); + +extern void +nfs3_fill_link3res (link3res *res, nfsstat3 stat, struct iatt *buf, + struct iatt *preparent, struct iatt *postparent, + uint64_t deviceid); + +extern void +nfs3_prep_link3args (link3args *args, struct nfs3_fh *target, + struct nfs3_fh * dirfh, char *name); + +extern void +nfs3_prep_rename3args (rename3args *args, struct nfs3_fh *olddirfh, + char *oldname, struct nfs3_fh *newdirfh, + char *newname); + +extern void +nfs3_fill_rename3res (rename3res *res, nfsstat3 stat, struct iatt *buf, + struct iatt *preoldparent, struct iatt *postoldparent, + struct iatt *prenewparent, struct iatt *postnewparent, + uint64_t deviceid); + +extern void +nfs3_prep_write3args (write3args *args, struct nfs3_fh *fh); + +extern void +nfs3_fill_write3res (write3res *res, nfsstat3 stat, count3 count, + stable_how stable, uint64_t wverf, struct iatt *prestat, + struct iatt *poststat, uint64_t deviceid); + +extern void +nfs3_prep_commit3args (commit3args *args, struct nfs3_fh *fh); + +extern void +nfs3_fill_commit3res (commit3res *res, nfsstat3 stat, uint64_t wverf, + struct iatt *prestat, struct iatt *poststat, + uint64_t deviceid); + +extern void +nfs3_fill_read3res (read3res *res, nfsstat3 stat, count3 count, + struct iatt *poststat, int is_eof, uint64_t deviceid); + +extern void +nfs3_prep_read3args (read3args *args, struct nfs3_fh *fh); + +extern void +nfs3_prep_pathconf3args (pathconf3args *args, struct nfs3_fh *fh); + +extern void +nfs3_fill_pathconf3res (pathconf3res *res, nfsstat3 stat, struct iatt *buf, + uint64_t deviceid); + +extern int +nfs3_cached_inode_opened (xlator_t *nfsxl, inode_t *inode); + +extern void +nfs3_log_common_res (uint32_t xid, int op, nfsstat3 stat, int pstat); + +extern void +nfs3_log_readlink_res (uint32_t xid, nfsstat3 stat, int pstat, char *linkpath); + +extern void +nfs3_log_read_res (uint32_t xid, nfsstat3 stat, int pstat, count3 count, + int is_eof, struct iovec *vec, int32_t vcount); + +extern void +nfs3_log_write_res (uint32_t xid, nfsstat3 stat, int pstat, count3 count, + int stable, uint64_t wverf); + +extern void +nfs3_log_newfh_res (uint32_t xid, int op, nfsstat3 stat, int pstat, + struct nfs3_fh *newfh); + +extern void +nfs3_log_readdir_res (uint32_t xid, nfsstat3 stat, int pstat, uint64_t cverf, + count3 count, int is_eof); + +extern void +nfs3_log_readdirp_res (uint32_t xid, nfsstat3 stat, int pstat, uint64_t cverf, + count3 dircount, count3 maxcount, int is_eof); + +extern void +nfs3_log_commit_res (uint32_t xid, nfsstat3 stat, int pstat, uint64_t wverf); + +extern void +nfs3_log_common_call (uint32_t xid, char *op, struct nfs3_fh *fh); + +extern void +nfs3_log_fh_entry_call (uint32_t xid, char *op, struct nfs3_fh *fh, + char *name); + +extern void +nfs3_log_rw_call (uint32_t xid, char *op, struct nfs3_fh *fh, offset3 offt, + count3 count, int stablewrite); + +extern void +nfs3_log_create_call (uint32_t xid, struct nfs3_fh *fh, char *name, + createmode3 mode); + +extern void +nfs3_log_symlink_call (uint32_t xid, struct nfs3_fh *fh, char *name, char *tgt); + +extern void +nfs3_log_mknod_call (uint32_t xid, struct nfs3_fh *fh, char *name, int type); + +extern void +nfs3_log_rename_call (uint32_t xid, struct nfs3_fh *src, char *sname, + struct nfs3_fh *dst, char *dname); + +extern void +nfs3_log_link_call (uint32_t xid, struct nfs3_fh *fh, char *name, + struct nfs3_fh *tgt); + +extern void +nfs3_log_readdir_call (uint32_t xid, struct nfs3_fh *fh, count3 dircount, + count3 maxcount); + +extern int +nfs3_fh_resolve_entry_hard (nfs3_call_state_t *cs); + +extern int +nfs3_fh_resolve_inode (nfs3_call_state_t *cs); + +extern int +nfs3_fh_resolve_entry (nfs3_call_state_t *cs); + +extern int +nfs3_fh_resolve_and_resume (nfs3_call_state_t *cs, struct nfs3_fh *fh, + char *entry, nfs3_resume_fn_t resum_fn); + +extern int +nfs3_verify_dircookie (struct nfs3_state *nfs3, fd_t *dirfd, cookie3 cookie, + uint64_t cverf, nfsstat3 *stat); + +extern int +nfs3_is_parentdir_entry (char *entry); + +uint32_t +nfs3_request_to_accessbits (int32_t accbits); + +#endif diff --git a/xlators/nfs/server/src/nfs3.c b/xlators/nfs/server/src/nfs3.c new file mode 100644 index 000000000..f914c3193 --- /dev/null +++ b/xlators/nfs/server/src/nfs3.c @@ -0,0 +1,5630 @@ +/* + Copyright (c) 2010-2011 Gluster, Inc. <http://www.gluster.com> + This file is part of GlusterFS. + + This file is licensed to you under your choice of the GNU Lesser + General Public License, version 3 or any later version (LGPLv3 or + later), or the GNU General Public License, version 2 (GPLv2), in all + cases as published by the Free Software Foundation. +*/ + +#ifndef _CONFIG_H +#define _CONFIG_H +#include "config.h" +#endif + +#include "rpcsvc.h" +#include "dict.h" +#include "xlator.h" +#include "mount3.h" +#include "xdr-nfs3.h" +#include "msg-nfs3.h" +#include "iobuf.h" +#include "nfs3.h" +#include "mem-pool.h" +#include "logging.h" +#include "nfs-common.h" +#include "nfs-fops.h" +#include "nfs-inodes.h" +#include "nfs-generics.h" +#include "nfs3-helpers.h" +#include "nfs-mem-types.h" +#include "nfs.h" +#include "xdr-rpc.h" +#include "xdr-generic.h" + +#include <sys/socket.h> +#include <sys/uio.h> +#include <sys/statvfs.h> +#include <time.h> + +#define nfs3_validate_strlen_or_goto(str, len, label, status, retval) \ + do { \ + if ((str)) { \ + if (strlen ((str)) > (len)) { \ + gf_log (GF_NFS3, GF_LOG_ERROR, "strlen "\ + "too long"); \ + status = NFS3ERR_NAMETOOLONG; \ + retval = -ENAMETOOLONG; \ + goto label; \ + } \ + } \ + } while (0); \ + +#define nfs3_validate_nfs3_state(request, state, status, label, retval) \ + do { \ + state = rpcsvc_request_program_private (request); \ + if (!state) { \ + gf_log (GF_NFS3, GF_LOG_ERROR, "NFSv3 state " \ + "missing from RPC request"); \ + status = NFS3ERR_SERVERFAULT; \ + ret = -EFAULT; \ + goto label; \ + } \ + } while (0); \ + + +struct nfs3_export * +__nfs3_get_export_by_index (struct nfs3_state *nfs3, uuid_t exportid) +{ + struct nfs3_export *exp = NULL; + int index = 0; + int searchindex = 0; + + searchindex = nfs3_fh_exportid_to_index (exportid); + list_for_each_entry (exp, &nfs3->exports, explist) { + if (searchindex == index) + goto found; + + ++index; + } + + exp = NULL; + gf_log (GF_NFS, GF_LOG_ERROR, "searchindex=%d not found", searchindex); +found: + return exp; +} + + +struct nfs3_export * +__nfs3_get_export_by_volumeid (struct nfs3_state *nfs3, uuid_t exportid) +{ + struct nfs3_export *exp = NULL; + + list_for_each_entry (exp, &nfs3->exports, explist) { + if (!uuid_compare (exportid, exp->volumeid)) + goto found; + } + + exp = NULL; +found: + return exp; +} + + +struct nfs3_export * +__nfs3_get_export_by_exportid (struct nfs3_state *nfs3, uuid_t exportid) +{ + struct nfs3_export *exp = NULL; + + if (!nfs3) + return exp; + + if (gf_nfs_dvm_off (nfs_state(nfs3->nfsx))) + exp = __nfs3_get_export_by_index (nfs3, exportid); + else + exp = __nfs3_get_export_by_volumeid (nfs3, exportid); + + return exp; +} + + +int +nfs3_export_access (struct nfs3_state *nfs3, uuid_t exportid) +{ + int ret = GF_NFS3_VOLACCESS_RO; + struct nfs3_export *exp = NULL; + + GF_VALIDATE_OR_GOTO (GF_NFS3, nfs3, err); + + exp = __nfs3_get_export_by_exportid (nfs3, exportid); + + if (!exp) { + gf_log (GF_NFS3, GF_LOG_ERROR, "Failed to get export by ID"); + goto err; + } + + ret = exp->access; + +err: + return ret; +} + +#define nfs3_check_rw_volaccess(nfs3state, exid, status, label) \ + do { \ + if (nfs3_export_access (nfs3state,exid)!=GF_NFS3_VOLACCESS_RW){\ + gf_log (GF_NFS3, GF_LOG_ERROR, "No read-write access");\ + status = NFS3ERR_ROFS; \ + goto label; \ + } \ + } while (0) \ + + + +xlator_t * +nfs3_fh_to_xlator (struct nfs3_state *nfs3, struct nfs3_fh *fh) +{ + xlator_t *vol = NULL; + struct nfs3_export *exp = NULL; + + GF_VALIDATE_OR_GOTO (GF_NFS3, nfs3, out); + GF_VALIDATE_OR_GOTO (GF_NFS3, fh, out); + + exp = __nfs3_get_export_by_exportid (nfs3, fh->exportid); + if (!exp) + goto out; + + vol = exp->subvol; +out: + return vol; +} + + +int +nfs3_is_root_looked_up (struct nfs3_state *nfs3, struct nfs3_fh *rootfh) +{ + struct nfs3_export *exp = NULL; + int ret = 0; + + GF_VALIDATE_OR_GOTO (GF_NFS3, nfs3, out); + GF_VALIDATE_OR_GOTO (GF_NFS3, rootfh, out); + + exp = __nfs3_get_export_by_exportid (nfs3, rootfh->exportid); + if (!exp) + goto out; + + ret = exp->rootlookedup; +out: + return ret; +} + + +int +nfs3_set_root_looked_up (struct nfs3_state *nfs3, struct nfs3_fh *rootfh) +{ + struct nfs3_export *exp = NULL; + int ret = 0; + + GF_VALIDATE_OR_GOTO (GF_NFS3, nfs3, out); + GF_VALIDATE_OR_GOTO (GF_NFS3, rootfh, out); + + exp = __nfs3_get_export_by_exportid (nfs3, rootfh->exportid); + if (!exp) + goto out; + + exp->rootlookedup = 1; +out: + return ret; +} + + +#define nfs3_map_fh_to_volume(nfs3state, handle, req, volume, status, label) \ + do { \ + char exportid[256], gfid[256]; \ + rpc_transport_t *trans = NULL; \ + volume = nfs3_fh_to_xlator ((nfs3state), handle); \ + if (!volume) { \ + uuid_unparse (handle->exportid, exportid); \ + uuid_unparse (handle->gfid, gfid); \ + trans = rpcsvc_request_transport (req); \ + GF_LOG_OCCASIONALLY (nfs3state->occ_logger, \ + GF_NFS3, GF_LOG_ERROR, "Failed to map " \ + "FH to vol: client=%s, exportid=%s, " \ + "gfid=%s", trans->peerinfo.identifier, \ + exportid, gfid); \ + GF_LOG_OCCASIONALLY (nfs3state->occ_logger, \ + GF_NFS3, GF_LOG_ERROR, "Stale nfs " \ + "client %s must be trying to connect to"\ + " a deleted volume, please unmount it.",\ + trans->peerinfo.identifier); \ + status = NFS3ERR_STALE; \ + goto label; \ + } else { \ + gf_log (GF_NFS3, GF_LOG_TRACE, "FH to Volume:" \ + "%s", volume->name); \ + rpcsvc_request_set_private (req, volume); \ + } \ + } while (0); \ + + +#define nfs3_validate_gluster_fh(handle, status, errlabel) \ + do { \ + if (!nfs3_fh_validate (handle)) { \ + gf_log (GF_NFS3, GF_LOG_ERROR, "Bad Handle"); \ + status = NFS3ERR_BADHANDLE; \ + goto errlabel; \ + } \ + } while (0) \ + +#define nfs3_check_fh_resolve_status(cst, nfstat, erlabl) \ + do { \ + xlator_t *xlatorp = NULL; \ + char buf[256], gfid[256]; \ + rpc_transport_t *trans = NULL; \ + if ((cst)->resolve_ret < 0) { \ + trans = rpcsvc_request_transport (cst->req); \ + xlatorp = nfs3_fh_to_xlator (cst->nfs3state, \ + &cst->resolvefh); \ + uuid_unparse (cst->resolvefh.gfid, gfid); \ + snprintf (buf, sizeof (buf), "(%s) %s : %s", \ + trans->peerinfo.identifier, \ + xlatorp ? xlatorp->name : "ERR", \ + gfid ); \ + gf_log (GF_NFS3, GF_LOG_ERROR, "%s: %s", \ + strerror(cst->resolve_errno), buf); \ + nfstat = nfs3_errno_to_nfsstat3 (cst->resolve_errno);\ + goto erlabl; \ + } \ + } while (0) \ + +#define nfs3_check_new_fh_resolve_status(cst, nfstat, erlabl) \ + do { \ + xlator_t *xlatorp = NULL; \ + char buf[256], gfid[256]; \ + rpc_transport_t *trans = NULL; \ + if (((cst)->resolve_ret < 0) && \ + ((cst)->resolve_errno != ENOENT)) { \ + trans = rpcsvc_request_transport (cst->req); \ + xlatorp = nfs3_fh_to_xlator (cst->nfs3state, \ + &cst->resolvefh); \ + uuid_unparse (cst->resolvefh.gfid, gfid); \ + snprintf (buf, sizeof (buf), "(%s) %s : %s", \ + trans->peerinfo.identifier, \ + xlatorp ? xlatorp->name : "ERR", \ + gfid); \ + gf_log (GF_NFS3, GF_LOG_ERROR, "%s: %s", \ + strerror(cst->resolve_errno), buf); \ + nfstat = nfs3_errno_to_nfsstat3 (cs->resolve_errno);\ + goto erlabl; \ + } \ + } while (0) \ + + +int +__nfs3_get_volume_id (struct nfs3_state *nfs3, xlator_t *xl, + uuid_t volumeid) +{ + int ret = -1; + struct nfs3_export *exp = NULL; + + GF_VALIDATE_OR_GOTO (GF_NFS3, nfs3, out); + GF_VALIDATE_OR_GOTO (GF_NFS3, xl, out); + + list_for_each_entry (exp, &nfs3->exports, explist) { + if (exp->subvol == xl) { + uuid_copy (volumeid, exp->volumeid); + ret = 0; + goto out; + } + } + +out: + return ret; +} + + +#define nfs3_funge_solaris_zerolen_fh(nfs3st, fhd, enam, nfsst, erl) \ + do { \ + xlator_t *fungexl = NULL; \ + uuid_t zero = {0, }; \ + fungexl =nfs_mntpath_to_xlator ((nfs3st)->exportslist,enam);\ + if (!fungexl) { \ + (nfsst) = NFS3ERR_NOENT; \ + goto erl; \ + } \ + \ + uuid_copy ((fhd)->gfid, zero); \ + (fhd)->gfid[15] = 1; \ + (enam) = NULL; \ + if ((gf_nfs_dvm_off (nfs_state (nfs3st->nfsx)))) \ + (fhd)->exportid[15] = nfs_xlator_to_xlid ((nfs3st)->exportslist, fungexl); \ + else { \ + if(__nfs3_get_volume_id ((nfs3st), fungexl, (fhd)->exportid) < 0) { \ + (nfsst) = NFS3ERR_STALE; \ + goto erl; \ + } \ + } \ + } while (0) \ + + +#define nfs3_volume_started_check(nf3stt, vlm, rtval, erlbl) \ + do { \ + if ((!nfs_subvolume_started (nfs_state (nf3stt->nfsx), vlm))){\ + gf_log (GF_NFS3, GF_LOG_ERROR, "Volume is disabled: %s",\ + vlm->name); \ + rtval = RPCSVC_ACTOR_IGNORE; \ + goto erlbl; \ + } \ + } while (0) \ + + +int +nfs3_export_sync_trusted (struct nfs3_state *nfs3, uuid_t exportid) +{ + struct nfs3_export *exp = NULL; + int ret = 0; + + GF_VALIDATE_OR_GOTO (GF_NFS3, nfs3, err); + + exp = __nfs3_get_export_by_exportid (nfs3, exportid); + if (!exp) + goto err; + + ret = exp->trusted_sync; +err: + return ret; +} + + +int +nfs3_export_write_trusted (struct nfs3_state *nfs3, uuid_t exportid) +{ + struct nfs3_export *exp = NULL; + int ret = 0; + + GF_VALIDATE_OR_GOTO (GF_NFS3, nfs3, err); + + exp = __nfs3_get_export_by_exportid (nfs3, exportid); + if (!exp) + goto err; + + ret = exp->trusted_write; +err: + return ret; +} + +int +nfs3_solaris_zerolen_fh (struct nfs3_fh *fh, int fhlen) +{ + if (!fh) + return 0; + + if (nfs3_fh_validate (fh)) + return 0; + + if (fhlen == 0) + return 1; + + return 0; +} + + +/* Function pointer that represents the generic prototypes of functions used + * to serialize NFS3 message structures into the XDR format. + * For usage, see the nfs3svc_XXX_cbk functions. + */ +typedef ssize_t (*nfs3_serializer) (struct iovec outmsg, void *args); + +nfs3_call_state_t * +nfs3_call_state_init (struct nfs3_state *s, rpcsvc_request_t *req, xlator_t *v) +{ + nfs3_call_state_t *cs = NULL; + + GF_VALIDATE_OR_GOTO (GF_NFS3, s, err); + GF_VALIDATE_OR_GOTO (GF_NFS3, req, err); + GF_VALIDATE_OR_GOTO (GF_NFS3, v, err); + + cs = (nfs3_call_state_t *) mem_get (s->localpool); + if (!cs) { + gf_log (GF_NFS3, GF_LOG_ERROR, "out of memory"); + return NULL; + } + + memset (cs, 0, sizeof (*cs)); + INIT_LIST_HEAD (&cs->entries.list); + INIT_LIST_HEAD (&cs->openwait_q); + cs->operrno = EINVAL; + cs->req = req; + cs->vol = v; + cs->nfsx = s->nfsx; + cs->nfs3state = s; +err: + return cs; +} + +void +nfs3_call_state_wipe (nfs3_call_state_t *cs) +{ + if (!cs) + return; + + if (cs->fd) { + gf_log (GF_NFS3, GF_LOG_TRACE, "fd 0x%lx ref: %d", + (long)cs->fd, cs->fd->refcount); + fd_unref (cs->fd); + } + + GF_FREE (cs->resolventry); + + GF_FREE (cs->pathname); + + if (!list_empty (&cs->entries.list)) + gf_dirent_free (&cs->entries); + + nfs_loc_wipe (&cs->oploc); + nfs_loc_wipe (&cs->resolvedloc); + if (cs->iob) + iobuf_unref (cs->iob); + if (cs->iobref) + iobref_unref (cs->iobref); + if (cs->trans) + rpc_transport_unref (cs->trans); + memset (cs, 0, sizeof (*cs)); + mem_put (cs); + /* Already refd by fd_lookup, so no need to ref again. */ +} + + +#define nfs3_handle_call_state_init(nfs3state, calls, rq, vl ,opstat, errlabel)\ + do { \ + calls = nfs3_call_state_init ((nfs3state), (rq), (vl)); \ + if (!calls) { \ + gf_log (GF_NFS3, GF_LOG_ERROR, "Failed to " \ + "init call state"); \ + opstat = NFS3ERR_SERVERFAULT; \ + goto errlabel; \ + } \ + } while (0) \ + + + +struct iobuf * +nfs3_serialize_reply (rpcsvc_request_t *req, void *arg, nfs3_serializer sfunc, + struct iovec *outmsg) +{ + struct nfs3_state *nfs3 = NULL; + struct iobuf *iob = NULL; + ssize_t retlen = -1; + + nfs3 = (struct nfs3_state *)rpcsvc_request_program_private (req); + if (!nfs3) { + gf_log (GF_NFS3, GF_LOG_ERROR, "NFSv3 state not found in RPC" + " request"); + goto ret; + } + + /* First, get the io buffer into which the reply in arg will + * be serialized. + */ + /* TODO: get rid of 'sfunc' and use 'xdrproc_t' so we + can have 'xdr_sizeof' */ + iob = iobuf_get (nfs3->iobpool); + if (!iob) { + gf_log (GF_NFS3, GF_LOG_ERROR, "Failed to get iobuf"); + goto ret; + } + + iobuf_to_iovec (iob, outmsg); + /* Use the given serializer to translate the give C structure in arg + * to XDR format which will be written into the buffer in outmsg. + */ + /* retlen is used to received the error since size_t is unsigned and we + * need -1 for error notification during encoding. + */ + retlen = sfunc (*outmsg, arg); + if (retlen == -1) { + gf_log (GF_NFS3, GF_LOG_ERROR, "Failed to encode message"); + goto ret; + } + + outmsg->iov_len = retlen; +ret: + if (retlen == -1) { + iobuf_unref (iob); + iob = NULL; + } + + return iob; +} + + + +/* Generic reply function for NFSv3 specific replies. */ +int +nfs3svc_submit_reply (rpcsvc_request_t *req, void *arg, nfs3_serializer sfunc) +{ + struct iovec outmsg = {0, }; + struct iobuf *iob = NULL; + int ret = -1; + struct iobref *iobref = NULL; + + if (!req) + return -1; + + iob = nfs3_serialize_reply (req, arg, sfunc, &outmsg); + if (!iob) { + gf_log (GF_NFS3, GF_LOG_ERROR, "Failed to serialize reply"); + goto ret; + } + + iobref = iobref_new (); + if (!iobref) { + gf_log (GF_NFS3, GF_LOG_ERROR, "failed on iobref_new()"); + goto ret; + } + + ret = iobref_add (iobref, iob); + if (ret) { + gf_log (GF_NFS3, GF_LOG_ERROR, "Failed to add iob to iobref"); + goto ret; + } + + /* Then, submit the message for transmission. */ + ret = rpcsvc_submit_message (req, &outmsg, 1, NULL, 0, iobref); + if (ret == -1) { + gf_log (GF_NFS3, GF_LOG_ERROR, "Reply submission failed"); + goto ret; + } + + ret = 0; +ret: + /* Now that we've done our job of handing the message to the RPC layer + * we can safely unref the iob in the hope that RPC layer must have + * ref'ed the iob on receiving into the txlist. + */ + if (NULL != iob) + iobuf_unref (iob); + if (NULL != iobref) + iobref_unref (iobref); + return ret; +} + + +int +nfs3svc_submit_vector_reply (rpcsvc_request_t *req, void *arg, + nfs3_serializer sfunc, struct iovec *payload, + int vcount, struct iobref *iobref) +{ + struct iovec outmsg = {0, }; + struct iobuf *iob = NULL; + int ret = -1; + int new_iobref = 0; + + if (!req) + return -1; + + iob = nfs3_serialize_reply (req, arg, sfunc, &outmsg); + if (!iob) { + gf_log (GF_NFS3, GF_LOG_ERROR, "Failed to serialize reply"); + goto ret; + } + if (iobref == NULL) { + iobref = iobref_new (); + if (!iobref) { + gf_log (GF_NFS3, GF_LOG_ERROR, "failed on iobref_new"); + goto ret; + } + new_iobref = 1; + } + + ret = iobref_add (iobref, iob); + if (ret) { + gf_log (GF_NFS3, GF_LOG_ERROR, "Failed to add iob to iobref"); + goto ret; + } + + /* Then, submit the message for transmission. */ + ret = rpcsvc_submit_message (req, &outmsg, 1, payload, vcount, iobref); + if (ret == -1) { + gf_log (GF_NFS3, GF_LOG_ERROR, "Reply submission failed"); + goto ret; + } + + ret = 0; +ret: + /* Now that we've done our job of handing the message to the RPC layer + * we can safely unref the iob in the hope that RPC layer must have + * ref'ed the iob on receiving into the txlist. + */ + if (NULL != iob) + iobuf_unref (iob); + if (new_iobref) + iobref_unref (iobref); + return ret; +} + +uint64_t +nfs3_request_xlator_deviceid (rpcsvc_request_t *rq) +{ + struct nfs3_state *nfs3 = NULL; + xlator_t *xl = NULL; + uint64_t devid = 0; + uuid_t volumeid = {0, }; + + if (!rq) + return 0; + + xl = rpcsvc_request_private (rq); + nfs3 = rpcsvc_request_program_private (rq); + if (gf_nfs_dvm_off (nfs_state (nfs3->nfsx))) + devid = (uint64_t)nfs_xlator_to_xlid (nfs3->exportslist, xl); + else { + __nfs3_get_volume_id (nfs3, xl, volumeid); + memcpy (&devid, &volumeid[8], sizeof (devid)); + } + + return devid; +} + + +int +nfs3svc_null (rpcsvc_request_t *req) +{ + struct iovec dummyvec = {0, }; + if (!req) + return RPCSVC_ACTOR_ERROR; + rpcsvc_submit_generic (req, &dummyvec, 1, NULL, 0, NULL); + return RPCSVC_ACTOR_SUCCESS; +} + + +int +nfs3_getattr_reply (rpcsvc_request_t *req, nfsstat3 status, struct iatt *buf) +{ + getattr3res res; + uint64_t deviceid = 0; + + deviceid = nfs3_request_xlator_deviceid (req); + nfs3_fill_getattr3res (&res, status, buf, deviceid); + nfs3svc_submit_reply (req, &res, + (nfs3_serializer)xdr_serialize_getattr3res); + + return 0; +} + + +int32_t +nfs3svc_getattr_lookup_cbk (call_frame_t *frame, void *cookie, xlator_t *this, + int32_t op_ret, int32_t op_errno, inode_t *inode, + struct iatt *buf, dict_t *xattr, + struct iatt *postparent) +{ + nfsstat3 status = NFS3_OK; + nfs3_call_state_t *cs = NULL; + + cs = frame->local; + + /* + * Somewhat counter-intuitively, we don't need to look for sh-failed + * here. Failing this getattr will generate a new lookup from the + * client, and nfs_fop_lookup_cbk will detect any self-heal failures. + */ + + if (op_ret == -1) { + gf_log (GF_NFS, GF_LOG_WARNING, + "%x: %s => -1 (%s)", rpcsvc_request_xid (cs->req), + cs->resolvedloc.path, strerror (op_errno)); + status = nfs3_cbk_errno_status (op_ret, op_errno); + } + else { + nfs_fix_generation(this,inode); + } + + nfs3_log_common_res (rpcsvc_request_xid (cs->req), NFS3_GETATTR, + status, op_errno); + + nfs3_getattr_reply (cs->req, status, buf); + nfs3_call_state_wipe (cs); + + return 0; +} + + +int32_t +nfs3svc_getattr_stat_cbk (call_frame_t *frame, void *cookie, xlator_t *this, + int32_t op_ret, int32_t op_errno, struct iatt *buf, + dict_t *xdata) +{ + nfsstat3 status = NFS3_OK; + nfs3_call_state_t *cs = NULL; + + cs = frame->local; + + if (op_ret == -1) { + gf_log (GF_NFS, GF_LOG_WARNING, + "%x: %s => -1 (%s)", rpcsvc_request_xid (cs->req), + cs->resolvedloc.path, strerror (op_errno)); + status = nfs3_cbk_errno_status (op_ret, op_errno); + } + + nfs3_log_common_res (rpcsvc_request_xid (cs->req), NFS3_GETATTR, + status, op_errno); + + nfs3_getattr_reply (cs->req, status, buf); + nfs3_call_state_wipe (cs); + + return 0; +} + + +int +nfs3_getattr_resume (void *carg) +{ + nfsstat3 stat = NFS3ERR_SERVERFAULT; + int ret = -EFAULT; + nfs_user_t nfu = {0, }; + nfs3_call_state_t *cs = NULL; + uint64_t raw_ctx = 0; + struct nfs_inode_ctx *ictx = NULL; + struct nfs_state *priv = NULL; + + if (!carg) + return ret; + + cs = (nfs3_call_state_t *)carg; + nfs3_check_fh_resolve_status (cs, stat, nfs3err); + nfs_request_user_init (&nfu, cs->req); + /* If inode which is to be getattr'd is the root, we need to do a + * lookup instead because after a server reboot, it is not necessary + * for the root to have been looked up when the getattr on the root is + * sent. AND, this causes a problem for stat-prefetch in that it + * expects even the root inode to have been looked up. + + if (__is_root_gfid (cs->resolvedloc.inode->gfid)) + ret = nfs_lookup (cs->nfsx, cs->vol, &nfu, &cs->resolvedloc, + nfs3svc_getattr_lookup_cbk, cs); + else + ret = nfs_stat (cs->nfsx, cs->vol, &nfu, &cs->resolvedloc, + */ + + if (cs->hardresolved) { + ret = -EFAULT; + stat = NFS3_OK; + goto nfs3err; + } + + /* + * If brick state changed, we need to force a proper lookup cycle (as + * would happen in native protocol) to do self-heal checks. We detect + * this by comparing the generation number for the last successful + * creation/lookup on the inode to the current number, so inodes that + * haven't been validated since the state change are affected. + */ + if (inode_ctx_get(cs->resolvedloc.inode,cs->nfsx,&raw_ctx) == 0) { + ictx = (struct nfs_inode_ctx *)raw_ctx; + priv = cs->nfsx->private; + if (ictx->generation != priv->generation) { + ret = nfs_lookup (cs->nfsx, cs->vol, &nfu, + &cs->resolvedloc, + nfs3svc_getattr_lookup_cbk, cs); + goto check_err; + } + } + + ret = nfs_stat (cs->nfsx, cs->vol, &nfu, &cs->resolvedloc, + nfs3svc_getattr_stat_cbk, cs); + +check_err: + if (ret < 0) { + gf_log (GF_NFS3, GF_LOG_ERROR, "Stat fop failed: %s: %s", + cs->oploc.path, strerror (-ret)); + stat = nfs3_errno_to_nfsstat3 (-ret); + } + +nfs3err: + if (ret < 0) { + nfs3_log_common_res (rpcsvc_request_xid (cs->req), + NFS3_GETATTR, stat, -ret); + nfs3_getattr_reply (cs->req, stat, &cs->stbuf); + nfs3_call_state_wipe (cs); + ret = 0; + } + + return ret; +} + + +int +nfs3_getattr (rpcsvc_request_t *req, struct nfs3_fh *fh) +{ + xlator_t *vol = NULL; + nfsstat3 stat = NFS3ERR_SERVERFAULT; + int ret = -EFAULT; + struct nfs3_state *nfs3 = NULL; + nfs3_call_state_t *cstate = NULL; + + GF_VALIDATE_OR_GOTO (GF_NFS3, req, out); + GF_VALIDATE_OR_GOTO (GF_NFS3, fh, out); + + nfs3_log_common_call (rpcsvc_request_xid (req), "GETATTR", fh); + nfs3_validate_gluster_fh (fh, stat, nfs3err); + nfs3_validate_nfs3_state (req, nfs3, stat, nfs3err, ret); + nfs3_map_fh_to_volume (nfs3, fh, req, vol, stat, nfs3err); + nfs3_volume_started_check (nfs3, vol, ret, out); + nfs3_handle_call_state_init (nfs3, cstate, req, vol, stat, nfs3err); + + ret = nfs3_fh_resolve_and_resume (cstate, fh, NULL,nfs3_getattr_resume); + if (ret < 0) + stat = nfs3_errno_to_nfsstat3 (-ret); + +nfs3err: + if (ret < 0) { + nfs3_log_common_res (rpcsvc_request_xid (req), NFS3_GETATTR, + stat, -ret); + nfs3_getattr_reply (req, stat, NULL); + ret = 0; + nfs3_call_state_wipe (cstate); + } +out: + return ret; +} + + +int +nfs3svc_getattr (rpcsvc_request_t *req) +{ + struct nfs3_fh fh = {{0}, }; + getattr3args args; + int ret = RPCSVC_ACTOR_ERROR; + + if (!req) + return ret; + + nfs3_prep_getattr3args (&args, &fh); + if (xdr_to_getattr3args (req->msg[0], &args) <= 0) { + gf_log (GF_NFS3, GF_LOG_ERROR, "Error decoding args"); + rpcsvc_request_seterr (req, GARBAGE_ARGS); + goto rpcerr; + } + + ret = nfs3_getattr (req, &fh); + if ((ret < 0) && (ret != RPCSVC_ACTOR_IGNORE)) { + gf_log (GF_NFS3, GF_LOG_ERROR, "GETATTR procedure failed"); + rpcsvc_request_seterr (req, SYSTEM_ERR); + ret = RPCSVC_ACTOR_ERROR; + } + +rpcerr: + return ret; +} + + +int +nfs3_setattr_reply (rpcsvc_request_t *req, nfsstat3 stat, struct iatt *preop, + struct iatt *postop) +{ + setattr3res res = {0, }; + uint64_t deviceid = 0; + + deviceid = nfs3_request_xlator_deviceid (req); + nfs3_fill_setattr3res (&res, stat, preop, postop, deviceid); + nfs3svc_submit_reply (req, (void *)&res, + (nfs3_serializer) xdr_serialize_setattr3res); + return 0; +} + + +int32_t +nfs3svc_truncate_cbk (call_frame_t *frame, void *cookie, xlator_t *this, + int32_t op_ret, int32_t op_errno, struct iatt *prebuf, + struct iatt *postbuf, dict_t *xdata) +{ + nfsstat3 stat = NFS3ERR_SERVERFAULT; + struct iatt *prestat = NULL; + nfs3_call_state_t *cs = NULL; + + cs = frame->local; + if (op_ret == -1) { + gf_log (GF_NFS, GF_LOG_WARNING, + "%x: %s => -1 (%s)", rpcsvc_request_xid (cs->req), + cs->resolvedloc.path, strerror (op_errno)); + stat = nfs3_cbk_errno_status (op_ret, op_errno); + goto nfs3err; + } + + /* If the first stat was got from the guarded setattr callback, or + * from an earlier setattr call then we'll need to use that stat + * instead of the preop returned here. + */ + if (cs->preparent.ia_ino != 0) + prestat = &cs->preparent; + else + prestat = prebuf; + + stat = NFS3_OK; +nfs3err: + nfs3_log_common_res (rpcsvc_request_xid (cs->req), NFS3_SETATTR, stat, + op_errno); + nfs3_setattr_reply (cs->req, stat, prestat, postbuf); + nfs3_call_state_wipe (cs); + + return 0; +} + + +int32_t +nfs3svc_setattr_cbk (call_frame_t *frame, void *cookie, xlator_t *this, + int32_t op_ret, int32_t op_errno, struct iatt *preop, + struct iatt *postop, dict_t *xdata) +{ + nfsstat3 stat = NFS3ERR_SERVERFAULT; + int ret = -1; + struct iatt *prebuf = NULL; + nfs_user_t nfu = {0, }; + nfs3_call_state_t *cs = NULL; + + cs = frame->local; + if (op_ret == -1) { + gf_log (GF_NFS, GF_LOG_WARNING, + "%x: %s => -1 (%s)", rpcsvc_request_xid (cs->req), + cs->resolvedloc.path, strerror (op_errno)); + stat = nfs3_cbk_errno_status (op_ret, op_errno); + goto nfs3err; + } + + prebuf = preop; + /* Store the current preop in case we need to send a truncate, + * in which case the preop to be returned will be this one. + */ + cs->preparent = *preop; + + /* Only truncate if the size is not already same as the requested + * truncation and also only if this is not a directory. + */ + if ((gf_attr_size_set (cs->setattr_valid)) && + (!IA_ISDIR (postop->ia_type)) && + (preop->ia_size != cs->stbuf.ia_size)) { + nfs_request_user_init (&nfu, cs->req); + ret = nfs_truncate (cs->nfsx, cs->vol, &nfu, &cs->resolvedloc, + cs->stbuf.ia_size, nfs3svc_truncate_cbk,cs); + + if (ret < 0) + stat = nfs3_errno_to_nfsstat3 (-ret); + } else { + ret = -1; /* Force a reply in the branch below. */ + stat = NFS3_OK; + } + +nfs3err: + if (ret < 0) { + nfs3_log_common_res (rpcsvc_request_xid (cs->req), + NFS3_SETATTR, stat, op_errno); + nfs3_setattr_reply (cs->req, stat, prebuf, postop); + nfs3_call_state_wipe (cs); + } + + return 0; +} + + + +int32_t +nfs3svc_setattr_stat_cbk (call_frame_t *frame, void *cookie, xlator_t *this, + int32_t op_ret, int32_t op_errno, struct iatt *buf, + dict_t *xdata) +{ + + int ret = -EFAULT; + nfsstat3 stat = NFS3ERR_SERVERFAULT; + nfs_user_t nfu = {0, }; + nfs3_call_state_t *cs = NULL; + + cs = frame->local; + if (op_ret == -1) { + gf_log (GF_NFS, GF_LOG_WARNING, + "%x: %s => -1 (%s)", rpcsvc_request_xid (cs->req), + cs->resolvedloc.path, strerror (op_errno)); + stat = nfs3_cbk_errno_status (op_ret, op_errno); + goto nfs3err; + } + + if (buf->ia_ctime != cs->timestamp.seconds) { + gf_log (GF_NFS3, GF_LOG_ERROR, "Timestamps not in sync"); + stat = NFS3ERR_NOT_SYNC; + goto nfs3err; + } + + /* Not a clean way but no motivation to add a new member to local. */ + cs->preparent = *buf; + nfs_request_user_init (&nfu, cs->req); + ret = nfs_setattr (cs->nfsx, cs->vol, &nfu, &cs->resolvedloc,&cs->stbuf, + cs->setattr_valid, nfs3svc_setattr_cbk, cs); + if (ret < 0) + stat = nfs3_errno_to_nfsstat3 (-ret); + +nfs3err: + if (ret < 0) { + nfs3_log_common_res (rpcsvc_request_xid (cs->req), + NFS3_SETATTR, stat, op_errno); + nfs3_setattr_reply (cs->req, stat, NULL, NULL); + nfs3_call_state_wipe (cs); + } + + return 0; +} + + +int +nfs3_setattr_resume (void *carg) +{ + nfsstat3 stat = NFS3ERR_SERVERFAULT; + int ret = -EFAULT; + nfs_user_t nfu = {0, }; + nfs3_call_state_t *cs = NULL; + + if (!carg) + return ret; + + cs = (nfs3_call_state_t *)carg; + nfs3_check_fh_resolve_status (cs, stat, nfs3err); + nfs_request_user_init (&nfu, cs->req); + ret = nfs_setattr (cs->nfsx, cs->vol, &nfu, &cs->resolvedloc, + &cs->stbuf, cs->setattr_valid, + nfs3svc_setattr_cbk, cs); + + if (ret < 0) + stat = nfs3_errno_to_nfsstat3 (-ret); + +nfs3err: + if (ret < 0) { + nfs3_log_common_res (rpcsvc_request_xid (cs->req), + NFS3_SETATTR, stat, -ret); + nfs3_setattr_reply (cs->req, stat, NULL, NULL); + nfs3_call_state_wipe (cs); + } + + return ret; +} + + +int +nfs3_setattr (rpcsvc_request_t *req, struct nfs3_fh *fh, sattr3 *sattr, + sattrguard3 *guard) +{ + xlator_t *vol = NULL; + nfsstat3 stat = NFS3ERR_SERVERFAULT; + int ret = -EFAULT; + struct nfs3_state *nfs3 = NULL; + nfs3_call_state_t *cs = NULL; + + GF_VALIDATE_OR_GOTO (GF_NFS3, req, out); + GF_VALIDATE_OR_GOTO (GF_NFS3, fh, out); + GF_VALIDATE_OR_GOTO (GF_NFS3, sattr, out); + GF_VALIDATE_OR_GOTO (GF_NFS3, guard, out); + + nfs3_log_common_call (rpcsvc_request_xid (req), "SETATTR", fh); + nfs3_validate_gluster_fh (fh, stat, nfs3err); + nfs3_validate_nfs3_state (req, nfs3, stat, nfs3err, ret); + nfs3_map_fh_to_volume (nfs3, fh, req, vol, stat, nfs3err); + nfs3_volume_started_check (nfs3, vol, ret, out); + nfs3_check_rw_volaccess (nfs3, fh->exportid, stat, nfs3err); + nfs3_handle_call_state_init (nfs3, cs, req, vol, stat, nfs3err); + + cs->setattr_valid = nfs3_sattr3_to_setattr_valid (sattr, &cs->stbuf, + NULL); + if (guard->check) { + gf_log (GF_NFS3, GF_LOG_TRACE, "Guard check required"); + cs->timestamp = guard->sattrguard3_u.obj_ctime; + cs->sattrguardcheck = 1; + } else { + gf_log (GF_NFS3, GF_LOG_TRACE, "Guard check not required"); + cs->sattrguardcheck = 0; + } + + if (!cs->setattr_valid) { + ret = -EINVAL; /* Force a reply */ + stat = NFS3_OK; + gf_log (GF_NFS3, GF_LOG_ERROR, "cs->setattr_valid is invalid"); + goto nfs3err; + } + + ret = nfs3_fh_resolve_and_resume (cs, fh, NULL, nfs3_setattr_resume); + if (ret < 0) + stat = nfs3_errno_to_nfsstat3 (-ret); + +nfs3err: + if (ret < 0) { + nfs3_log_common_res (rpcsvc_request_xid (req), NFS3_SETATTR, + stat, -ret); + nfs3_setattr_reply (req, stat, NULL, NULL); + nfs3_call_state_wipe (cs); + /* Ret must be 0 after this so that the caller does not + * also send an RPC reply. + */ + ret = 0; + } +out: + return ret; +} + + + +int +nfs3svc_setattr (rpcsvc_request_t *req) +{ + struct nfs3_fh fh = {{0}, }; + setattr3args args; + int ret = RPCSVC_ACTOR_ERROR; + + GF_VALIDATE_OR_GOTO (GF_NFS3, req, rpcerr); + + nfs3_prep_setattr3args (&args, &fh); + if (xdr_to_setattr3args (req->msg[0], &args) <= 0) { + gf_log (GF_NFS3, GF_LOG_ERROR, "Error decoding args"); + rpcsvc_request_seterr (req, GARBAGE_ARGS); + goto rpcerr; + } + + ret = nfs3_setattr (req, &fh, &args.new_attributes, &args.guard); + if ((ret < 0) && (ret != RPCSVC_ACTOR_IGNORE)) { + gf_log (GF_NFS3, GF_LOG_ERROR, "SETATTR procedure failed"); + rpcsvc_request_seterr (req, SYSTEM_ERR); + ret = RPCSVC_ACTOR_ERROR; + } + +rpcerr: + return ret; + +} + + +int +nfs3_lookup_reply (rpcsvc_request_t *req, nfsstat3 stat, struct nfs3_fh *newfh, + struct iatt *stbuf, struct iatt *postparent) +{ + lookup3res res = {0, }; + uint64_t deviceid = 0; + + deviceid = nfs3_request_xlator_deviceid (req); + nfs3_fill_lookup3res (&res, stat, newfh, stbuf, postparent, deviceid); + return nfs3svc_submit_reply (req, &res, + (nfs3_serializer)xdr_serialize_lookup3res); +} + +int +nfs3_lookup_resume (void *carg); + + +int +nfs3_fresh_lookup (nfs3_call_state_t *cs) +{ + int ret = -EFAULT; + char *oldresolventry = NULL; + + GF_VALIDATE_OR_GOTO (GF_NFS3, cs, err); + gf_log (GF_NFS3, GF_LOG_DEBUG, "inode needs fresh lookup"); + inode_unlink (cs->resolvedloc.inode, cs->resolvedloc.parent, + cs->resolventry); + nfs_loc_wipe (&cs->resolvedloc); + + /* Store pointer to currently allocated resolventry because it gets over + * written in fh_resolve_and_resume. + */ + oldresolventry = cs->resolventry; + cs->lookuptype = GF_NFS3_FRESH; + ret = nfs3_fh_resolve_and_resume (cs, &cs->resolvefh, cs->resolventry, + nfs3_lookup_resume); + /* Allocated in the previous call to fh_resolve_and_resume using the + * same call_state. + */ + GF_FREE (oldresolventry); +err: + return ret; +} + +int +nfs3svc_lookup_cbk (call_frame_t *frame, void *cookie, xlator_t *this, + int32_t op_ret, int32_t op_errno, inode_t *inode, + struct iatt *buf, dict_t *xattr, struct iatt *postparent) +{ + struct nfs3_fh newfh = {{0}, }; + nfsstat3 status = NFS3_OK; + nfs3_call_state_t *cs = NULL; + inode_t *oldinode = NULL; + + cs = frame->local; + if (op_ret == -1) { + gf_log (GF_NFS, + (op_errno == ENOENT ? GF_LOG_TRACE : GF_LOG_WARNING), + "%x: %s => -1 (%s)", rpcsvc_request_xid (cs->req), + cs->resolvedloc.path, strerror (op_errno)); + status = nfs3_cbk_errno_status (op_ret, op_errno); + goto xmit_res; + } + + nfs3_fh_build_child_fh (&cs->parent, buf, &newfh); + oldinode = inode_link (inode, cs->resolvedloc.parent, + cs->resolvedloc.name, buf); +xmit_res: + /* Only send fresh lookup if it was a revalidate that failed. */ + if ((op_ret == -1) && (nfs3_is_revalidate_lookup (cs))) { + op_ret = nfs3_fresh_lookup (cs); + goto out; + } + + nfs3_log_newfh_res (rpcsvc_request_xid (cs->req), NFS3_LOOKUP, status, + op_errno, &newfh); + nfs3_lookup_reply (cs->req, status, &newfh, buf, postparent); + nfs3_call_state_wipe (cs); +out: + if (oldinode) { + inode_lookup (oldinode); + inode_unref (oldinode); + } + return 0; +} + + +int +nfs3svc_lookup_parentdir_cbk (call_frame_t *frame, void *cookie, xlator_t *this, + int32_t op_ret, int32_t op_errno, inode_t *inode, + struct iatt *buf, dict_t *xattr, + struct iatt *postparent) +{ + struct nfs3_fh newfh = {{0}, }; + nfsstat3 status = NFS3_OK; + nfs3_call_state_t *cs = NULL; + uuid_t volumeid = {0, }; + struct nfs3_state *nfs3 = NULL; + + cs = frame->local; + if (op_ret == -1) { + gf_log (GF_NFS, GF_LOG_WARNING, + "%x: %s => -1 (%s)", rpcsvc_request_xid (cs->req), + cs->resolvedloc.path, strerror (op_errno)); + status = nfs3_cbk_errno_status (op_ret, op_errno); + goto xmit_res; + } + + nfs3 = cs->nfs3state; + /* If the buf inode shows that this is a root dir's buf, then the file + * handle needs to be specially crafted, in all other cases, we'll just + * create the handle normally using the buffer of the parent dir. + */ + if (buf->ia_ino != 1) { + nfs3_fh_build_parent_fh (&cs->fh, buf, &newfh); + goto xmit_res; + } + + if (gf_nfs_dvm_off (nfs_state (nfs3->nfsx))) + newfh = nfs3_fh_build_indexed_root_fh (nfs3->exportslist, + cs->vol); + else { + __nfs3_get_volume_id (nfs3, cs->vol, volumeid); + newfh = nfs3_fh_build_uuid_root_fh (volumeid); + } + +xmit_res: + nfs3_log_newfh_res (rpcsvc_request_xid (cs->req), NFS3_LOOKUP, status, + op_errno, &newfh); + nfs3_lookup_reply (cs->req, status, &newfh, buf, postparent); + nfs3_call_state_wipe (cs); + + return 0; +} + + + +int +nfs3_lookup_parentdir_resume (void *carg) +{ + nfsstat3 stat = NFS3ERR_SERVERFAULT; + int ret = -EFAULT; + nfs_user_t nfu = {0, }; + nfs3_call_state_t *cs = NULL; + inode_t *parent = NULL; + + if (!carg) { + gf_log (GF_NFS3, GF_LOG_ERROR, "Invalid argument," + " carg value NULL"); + return EINVAL; + } + + cs = (nfs3_call_state_t *)carg; + nfs3_check_fh_resolve_status (cs, stat, nfs3err); + + /* At this point now, the loc in cs is for the directory file handle + * sent by the client. This loc needs to be transformed into a loc that + * represents the parent dir of cs->resolvedloc.inode. + * + * EXCEPT in the case where the .. is a parent of the root directory. + * In this case we'll be returning the file handle and attributes of the + * root itself. + */ + nfs_request_user_init (&nfu, cs->req); + + /* Save the file handle from the LOOKUP request. We'll use this to + * build the file handle of the parent directory in case the parent is + * not root dir. + */ + cs->fh = cs->resolvefh; + + /* If fh is that of the root, the resolvedloc will already contain + * the loc for root. After that, we'll send lookup for the root dir + * itself since we cannot send the lookup on the parent of root. + * + * For all other cases, we'll send the lookup on the parent of the + * given directory file handle. + */ + if (!nfs3_fh_is_root_fh (&cs->fh)) { + parent = inode_ref (cs->resolvedloc.parent); + nfs_loc_wipe (&cs->resolvedloc); + ret = nfs_inode_loc_fill (parent, &cs->resolvedloc, + NFS_RESOLVE_CREATE); + + if (ret < 0) { + gf_log (GF_NFS3, GF_LOG_ERROR, "nfs_inode_loc_fill" + " error"); + goto errtostat; + } + } + + ret = nfs_lookup (cs->nfsx, cs->vol, &nfu, &cs->resolvedloc, + nfs3svc_lookup_parentdir_cbk, cs); +errtostat: + if (ret < 0) + stat = nfs3_errno_to_nfsstat3 (-ret); + +nfs3err: + if (ret < 0) { + nfs3_log_common_res (rpcsvc_request_xid (cs->req), NFS3_LOOKUP, + stat, -ret); + nfs3_lookup_reply (cs->req, stat, NULL, NULL, NULL); + nfs3_call_state_wipe (cs); + } + + if (parent) + inode_unref (parent); + + return ret; +} + + +int +nfs3_lookup_resume (void *carg) +{ + nfsstat3 stat = NFS3ERR_SERVERFAULT; + int ret = -EFAULT; + nfs_user_t nfu = {0, }; + nfs3_call_state_t *cs = NULL; + struct nfs3_fh newfh = {{0},}; + + if (!carg) { + gf_log (GF_NFS3, GF_LOG_ERROR, "Invalid argument," + " carg value NULL"); + return EINVAL; + } + + cs = (nfs3_call_state_t *)carg; + nfs3_check_fh_resolve_status (cs, stat, nfs3err); + cs->parent = cs->resolvefh; + + if (cs->hardresolved) { + stat = NFS3_OK; + nfs3_fh_build_child_fh (&cs->parent, &cs->stbuf, &newfh); + goto nfs3err; + } + + nfs_request_user_init (&nfu, cs->req); + ret = nfs_lookup (cs->nfsx, cs->vol, &nfu, &cs->resolvedloc, + nfs3svc_lookup_cbk, cs); + if (ret < 0) + stat = nfs3_errno_to_nfsstat3 (-ret); + +nfs3err: + if (ret < 0) { + nfs3_log_common_res (rpcsvc_request_xid (cs->req), NFS3_LOOKUP, + stat, -ret); + nfs3_lookup_reply (cs->req, stat, &newfh, &cs->stbuf, + &cs->postparent); + nfs3_call_state_wipe (cs); + } + + return ret; +} + + +int +nfs3_lookup (rpcsvc_request_t *req, struct nfs3_fh *fh, int fhlen, char *name) +{ + xlator_t *vol = NULL; + nfsstat3 stat = NFS3ERR_SERVERFAULT; + int ret = -EFAULT; + struct nfs3_state *nfs3 = NULL; + nfs3_call_state_t *cs = NULL; + + GF_VALIDATE_OR_GOTO (GF_NFS3, req, out); + GF_VALIDATE_OR_GOTO (GF_NFS3, fh, out); + GF_VALIDATE_OR_GOTO (GF_NFS3, name, out); + + nfs3_log_fh_entry_call (rpcsvc_request_xid (req), "LOOKUP", fh, + name); + nfs3_validate_nfs3_state (req, nfs3, stat, nfs3err, ret); + if (nfs3_solaris_zerolen_fh (fh, fhlen)) + nfs3_funge_solaris_zerolen_fh (nfs3, fh, name, stat, nfs3err); + else + nfs3_validate_gluster_fh (fh, stat, nfs3err); + nfs3_validate_strlen_or_goto (name, NFS_NAME_MAX, nfs3err, stat, ret); + nfs3_map_fh_to_volume (nfs3, fh, req, vol, stat, nfs3err); + nfs3_volume_started_check (nfs3, vol, ret, out); + nfs3_handle_call_state_init (nfs3, cs, req, vol, stat, nfs3err); + + cs->lookuptype = GF_NFS3_REVALIDATE; + ret = nfs3_fh_resolve_and_resume (cs, fh, name, + nfs3_lookup_resume); + + if (ret < 0) { + gf_log (GF_NFS, GF_LOG_ERROR, "failed to start hard reslove"); + stat = nfs3_errno_to_nfsstat3 (-ret); + } + +nfs3err: + if (ret < 0) { + nfs3_log_common_res (rpcsvc_request_xid (req), NFS3_LOOKUP, + stat, + -ret); + nfs3_lookup_reply (req, stat, NULL, NULL, NULL); + nfs3_call_state_wipe (cs); + /* Ret must be 0 after this so that the caller does not + * also send an RPC reply. + */ + ret = 0; + } +out: + return ret; +} + + +int +nfs3svc_lookup (rpcsvc_request_t *req) +{ + char name[NFS_PATH_MAX]; + struct nfs3_fh fh = {{0}, }; + lookup3args args; + int ret = RPCSVC_ACTOR_ERROR; + + GF_VALIDATE_OR_GOTO (GF_NFS, req, rpcerr); + + nfs3_prep_lookup3args (&args, &fh, name); + if (xdr_to_lookup3args (req->msg[0], &args) <= 0) { + gf_log (GF_NFS3, GF_LOG_ERROR, "Error decoding args"); + rpcsvc_request_seterr (req, GARBAGE_ARGS); + goto rpcerr; + } + + ret = nfs3_lookup (req, &fh, args.what.dir.data.data_len, name); + if ((ret < 0) && (ret != RPCSVC_ACTOR_IGNORE)) { + gf_log (GF_NFS3, GF_LOG_ERROR, "LOOKUP procedure failed"); + rpcsvc_request_seterr (req, SYSTEM_ERR); + ret = RPCSVC_ACTOR_ERROR; + } + +rpcerr: + return ret; +} + + +int +nfs3_access_reply (rpcsvc_request_t *req, nfsstat3 status, int32_t accbits, + int32_t reqaccbits) +{ + access3res res; + + nfs3_fill_access3res (&res, status, accbits, reqaccbits); + nfs3svc_submit_reply (req, &res, + (nfs3_serializer)xdr_serialize_access3res); + return 0; +} + + +int32_t +nfs3svc_access_cbk (call_frame_t *frame, void *cookie, xlator_t *this, + int32_t op_ret, int32_t op_errno, dict_t *xdata) +{ + nfsstat3 status = NFS3_OK; + nfs3_call_state_t *cs = NULL; + + cs = frame->local; + + if (op_ret == -1) { + gf_log (GF_NFS, GF_LOG_WARNING, + "%x: %s => -1 (%s)", rpcsvc_request_xid (cs->req), + cs->resolvedloc.path, strerror (op_errno)); + status = nfs3_cbk_errno_status (op_ret, op_errno); + } + nfs3_log_common_res (rpcsvc_request_xid (cs->req), NFS3_ACCESS, status, + op_errno); + nfs3_access_reply (cs->req, status, op_errno, cs->accessbits); + nfs3_call_state_wipe (cs); + + return 0; +} + +int +nfs3_access_resume (void *carg) +{ + nfsstat3 stat = NFS3ERR_SERVERFAULT; + int ret = -EFAULT; + nfs_user_t nfu = {0, }; + nfs3_call_state_t *cs = NULL; + + if (!carg) { + gf_log (GF_NFS3, GF_LOG_ERROR, "Invalid argument," + " carg value NULL"); + return EINVAL; + } + + cs = (nfs3_call_state_t *)carg; + nfs3_check_fh_resolve_status (cs, stat, nfs3err); + cs->fh = cs->resolvefh; + nfs_request_user_init (&nfu, cs->req); + ret = nfs_access (cs->nfsx, cs->vol, &nfu, &cs->resolvedloc, + cs->accessbits, nfs3svc_access_cbk, cs); + if (ret < 0) + stat = nfs3_errno_to_nfsstat3 (-ret); + +nfs3err: + if (ret < 0) { + nfs3_log_common_res (rpcsvc_request_xid (cs->req), NFS3_ACCESS, + stat, -ret); + nfs3_access_reply (cs->req, stat, 0, 0); + nfs3_call_state_wipe (cs); + ret = 0; + } + + return ret; +} + + +int +nfs3_access (rpcsvc_request_t *req, struct nfs3_fh *fh, uint32_t accbits) +{ + xlator_t *vol = NULL; + struct nfs3_state *nfs3 = NULL; + nfsstat3 stat = NFS3ERR_SERVERFAULT; + int ret = -EFAULT; + nfs3_call_state_t *cs = NULL; + + GF_VALIDATE_OR_GOTO (GF_NFS, req, out); + GF_VALIDATE_OR_GOTO (GF_NFS, fh, out); + nfs3_log_common_call (rpcsvc_request_xid (req), "ACCESS", fh); + nfs3_validate_gluster_fh (fh, stat, nfs3err); + nfs3_validate_nfs3_state (req, nfs3, stat, nfs3err, ret); + nfs3_map_fh_to_volume (nfs3, fh, req, vol, stat, nfs3err); + nfs3_volume_started_check (nfs3, vol, ret, out); + nfs3_handle_call_state_init (nfs3, cs, req, vol, stat, nfs3err); + cs->accessbits = accbits; + + ret = nfs3_fh_resolve_and_resume (cs, fh, NULL, nfs3_access_resume); + if (ret < 0) + stat = nfs3_errno_to_nfsstat3 (-ret); + +nfs3err: + if (ret < 0) { + nfs3_log_common_res (rpcsvc_request_xid (req), NFS3_ACCESS, + stat, -ret); + nfs3_access_reply (req, stat, 0, 0); + nfs3_call_state_wipe (cs); + ret = 0; + } +out: + return ret; +} + + +int +nfs3svc_access (rpcsvc_request_t *req) +{ + struct nfs3_fh fh = {{0}, }; + access3args args; + int ret = RPCSVC_ACTOR_ERROR; + + if (!req) + return ret; + + nfs3_prep_access3args (&args, &fh); + if (xdr_to_access3args (req->msg[0], &args) <= 0) { + gf_log (GF_NFS3, GF_LOG_ERROR, "Error decoding args"); + rpcsvc_request_seterr (req, GARBAGE_ARGS); + goto rpcerr; + } + + ret = nfs3_access (req, &fh, args.access); + if ((ret < 0) && (ret != RPCSVC_ACTOR_IGNORE)) { + gf_log (GF_NFS3, GF_LOG_ERROR, "ACCESS procedure failed"); + rpcsvc_request_seterr (req, SYSTEM_ERR); + ret = RPCSVC_ACTOR_ERROR; + } + +rpcerr: + return ret; +} + + +int +nfs3_readlink_reply (rpcsvc_request_t *req, nfsstat3 stat, char *path, + struct iatt *buf) +{ + readlink3res res = {0, }; + uint64_t deviceid = 0; + + deviceid = nfs3_request_xlator_deviceid (req); + nfs3_fill_readlink3res (&res, stat, path, buf, deviceid); + nfs3svc_submit_reply (req, (void *)&res, + (nfs3_serializer)xdr_serialize_readlink3res); + + return 0; +} + + +int32_t +nfs3svc_readlink_cbk (call_frame_t *frame, void *cookie, xlator_t *this, + int32_t op_ret, int32_t op_errno, const char *path, + struct iatt *buf, dict_t *xdata) +{ + nfsstat3 stat = NFS3ERR_SERVERFAULT; + nfs3_call_state_t *cs = NULL; + + cs = frame->local; + if (op_ret == -1) { + gf_log (GF_NFS, GF_LOG_WARNING, + "%x: %s => -1 (%s)", rpcsvc_request_xid (cs->req), + cs->resolvedloc.path, strerror (op_errno)); + stat = nfs3_cbk_errno_status (op_ret, op_errno); + goto nfs3err; + } + + stat = NFS3_OK; + +nfs3err: + nfs3_log_readlink_res (rpcsvc_request_xid (cs->req), stat, op_errno, + (char *)path); + nfs3_readlink_reply (cs->req, stat, (char *)path, buf); + nfs3_call_state_wipe (cs); + + return 0; +} + + +int +nfs3_readlink_resume (void *carg) +{ + nfsstat3 stat = NFS3ERR_SERVERFAULT; + int ret = -EFAULT; + nfs3_call_state_t *cs = NULL; + nfs_user_t nfu = {0, }; + + if (!carg) + return ret; + + cs = (nfs3_call_state_t *)carg; + nfs3_check_fh_resolve_status (cs, stat, nfs3err); + nfs_request_user_init (&nfu, cs->req); + ret = nfs_readlink (cs->nfsx, cs->vol, &nfu, &cs->resolvedloc, + nfs3svc_readlink_cbk, cs); + if (ret < 0) + stat = nfs3_errno_to_nfsstat3 (-ret); + +nfs3err: + if (ret < 0) { + nfs3_log_common_res (rpcsvc_request_xid (cs->req), + NFS3_READLINK, stat, -ret); + nfs3_readlink_reply (cs->req, stat, NULL, NULL); + nfs3_call_state_wipe (cs); + } + + return ret; +} + + +int +nfs3_readlink (rpcsvc_request_t *req, struct nfs3_fh *fh) +{ + xlator_t *vol = NULL; + nfsstat3 stat = NFS3ERR_SERVERFAULT; + int ret = -EFAULT; + struct nfs3_state *nfs3 = NULL; + nfs3_call_state_t *cs = NULL; + + if ((!req) || (!fh)) { + gf_log (GF_NFS3, GF_LOG_ERROR, "Bad arguments"); + return -1; + } + + nfs3_log_common_call (rpcsvc_request_xid (req), "READLINK", fh); + nfs3_validate_gluster_fh (fh, stat, nfs3err); + nfs3_validate_nfs3_state (req, nfs3, stat, nfs3err, ret); + nfs3_map_fh_to_volume (nfs3, fh, req, vol, stat, nfs3err); + nfs3_volume_started_check (nfs3, vol, ret, out); + nfs3_handle_call_state_init (nfs3, cs, req, vol, stat, nfs3err); + + ret = nfs3_fh_resolve_and_resume (cs, fh, NULL, nfs3_readlink_resume); + if (ret < 0) + stat = nfs3_errno_to_nfsstat3 (-ret); + +nfs3err: + if (ret < 0) { + nfs3_log_common_res (rpcsvc_request_xid (req), NFS3_READLINK, + stat, -ret); + nfs3_readlink_reply (req, stat, NULL, NULL); + nfs3_call_state_wipe (cs); + /* Ret must be 0 after this so that the caller does not + * also send an RPC reply. + */ + ret = 0; + } +out: + return ret; +} + + +int +nfs3svc_readlink (rpcsvc_request_t *req) +{ + struct nfs3_fh fh = {{0}, }; + readlink3args args; + int ret = RPCSVC_ACTOR_ERROR; + + if (!req) + return ret; + + nfs3_prep_readlink3args (&args, &fh); + if (xdr_to_readlink3args (req->msg[0], &args) <= 0) { + gf_log (GF_NFS3, GF_LOG_ERROR, "Error decoding args"); + rpcsvc_request_seterr (req, GARBAGE_ARGS); + goto rpcerr; + } + + ret = nfs3_readlink (req, &fh); + if ((ret < 0) && (ret != RPCSVC_ACTOR_IGNORE)) { + gf_log (GF_NFS3, GF_LOG_ERROR, "READLINK procedure failed"); + rpcsvc_request_seterr (req, SYSTEM_ERR); + ret = RPCSVC_ACTOR_ERROR; + } + +rpcerr: + return ret; +} + + +int +nfs3_read_reply (rpcsvc_request_t *req, nfsstat3 stat, count3 count, + struct iovec *vec, int vcount, struct iobref *iobref, + struct iatt *poststat, int is_eof) +{ + read3res res = {0, }; + uint64_t deviceid = 0; + + deviceid = nfs3_request_xlator_deviceid (req); + nfs3_fill_read3res (&res, stat, count, poststat, is_eof, deviceid); + if (stat == NFS3_OK) { + xdr_vector_round_up (vec, vcount, count); + /* iob can be zero if the file size was zero. If so, op_ret + * would be 0 and count = 0. + */ + + if (count != 0) { + nfs3svc_submit_vector_reply (req, (void *)&res, + (nfs3_serializer) + xdr_serialize_read3res_nocopy, + vec, vcount, iobref); + } else + + nfs3svc_submit_reply (req, (void *)&res, + (nfs3_serializer) + xdr_serialize_read3res_nocopy); + } else + nfs3svc_submit_reply (req, (void *)&res, + (nfs3_serializer) + xdr_serialize_read3res_nocopy); + + return 0; +} + + +int32_t +nfs3svc_read_cbk (call_frame_t *frame, void *cookie, xlator_t *this, + int32_t op_ret, int32_t op_errno, struct iovec *vector, + int32_t count, struct iatt *stbuf, struct iobref *iobref, + dict_t *xdata) +{ + nfsstat3 stat = NFS3ERR_SERVERFAULT; + int is_eof = 0; + nfs3_call_state_t *cs = NULL; + + cs = frame->local; + if (op_ret == -1) { + gf_log (GF_NFS, GF_LOG_WARNING, + "%x: %s => -1 (%s)", rpcsvc_request_xid (cs->req), + cs->resolvedloc.path, strerror (op_errno)); + stat = nfs3_cbk_errno_status (op_ret, op_errno); + goto err; + } else + stat = NFS3_OK; + + if (op_errno == ENOENT) + is_eof = 1; + +err: + nfs3_log_read_res (rpcsvc_request_xid (cs->req), stat, op_errno, + op_ret, is_eof, vector, count); + nfs3_read_reply (cs->req, stat, op_ret, vector, count, iobref, stbuf, + is_eof); + nfs3_call_state_wipe (cs); + + return 0; +} + + +int +nfs3_read_fd_resume (void *carg) +{ + nfsstat3 stat = NFS3ERR_SERVERFAULT; + int ret = -EFAULT; + nfs_user_t nfu = {0, }; + nfs3_call_state_t *cs = NULL; + + if (!carg) + return ret; + + cs = (nfs3_call_state_t *)carg; + nfs3_check_fh_resolve_status (cs, stat, nfs3err); + nfs_request_user_init (&nfu, cs->req); + ret = nfs_read (cs->nfsx, cs->vol, &nfu, cs->fd, cs->datacount, + cs->dataoffset, nfs3svc_read_cbk, cs); + if (ret < 0) + stat = nfs3_errno_to_nfsstat3 (-ret); +nfs3err: + if (ret < 0) { + nfs3_log_common_res (rpcsvc_request_xid (cs->req), NFS3_READ, + stat, -ret); + nfs3_read_reply (cs->req, stat, 0, NULL, 0, NULL, NULL, 0); + nfs3_call_state_wipe (cs); + } + + return ret; +} + + +int +nfs3_read_resume (void *carg) +{ + nfsstat3 stat = NFS3ERR_SERVERFAULT; + int ret = -EFAULT; + nfs3_call_state_t *cs = NULL; + fd_t *fd = NULL; + + if (!carg) + return ret; + + cs = (nfs3_call_state_t *)carg; + nfs3_check_fh_resolve_status (cs, stat, nfs3err); + fd = fd_anonymous (cs->resolvedloc.inode); + if (!fd) { + gf_log (GF_NFS3, GF_LOG_ERROR, "Failed to create anonymous fd"); + goto nfs3err; + } + + cs->fd = fd; + nfs3_read_fd_resume (cs); + ret = 0; +nfs3err: + if (ret < 0) { + nfs3_log_common_res (rpcsvc_request_xid (cs->req), NFS3_READ, + stat, -ret); + nfs3_read_reply (cs->req, stat, 0, NULL,0, NULL, NULL, 0); + nfs3_call_state_wipe (cs); + } + + return ret; +} + +int +nfs3_read (rpcsvc_request_t *req, struct nfs3_fh *fh, offset3 offset, + count3 count) +{ + xlator_t *vol = NULL; + nfsstat3 stat = NFS3ERR_SERVERFAULT; + int ret = -EFAULT; + struct nfs3_state *nfs3 = NULL; + nfs3_call_state_t *cs = NULL; + + if ((!req) || (!fh)) { + gf_log (GF_NFS3, GF_LOG_ERROR, "Bad arguments"); + return -1; + } + + nfs3_log_rw_call (rpcsvc_request_xid (req), "READ", fh, offset, + count, -1); + nfs3_validate_gluster_fh (fh, stat, nfs3err); + nfs3_validate_nfs3_state (req, nfs3, stat, nfs3err, ret); + nfs3_map_fh_to_volume (nfs3, fh, req, vol, stat, nfs3err); + nfs3_volume_started_check (nfs3, vol, ret, out); + nfs3_handle_call_state_init (nfs3, cs, req, vol, stat, nfs3err); + + cs->datacount = count; + cs->dataoffset = offset; + ret = nfs3_fh_resolve_and_resume (cs, fh, NULL, nfs3_read_resume); + if (ret < 0) + stat = nfs3_errno_to_nfsstat3 (-ret); + +nfs3err: + if (ret < 0) { + nfs3_log_common_res (rpcsvc_request_xid (req), NFS3_READ, stat, + -ret); + nfs3_read_reply (req, stat, 0, NULL,0, NULL, NULL, 0); + nfs3_call_state_wipe (cs); + ret = 0; + } +out: + return ret; +} + + +int +nfs3svc_read (rpcsvc_request_t *req) +{ + struct nfs3_fh fh = {{0}, }; + read3args args; + int ret = RPCSVC_ACTOR_ERROR; + + if (!req) + return ret; + + nfs3_prep_read3args (&args, &fh); + if (xdr_to_read3args (req->msg[0], &args) <= 0) { + gf_log (GF_NFS3, GF_LOG_ERROR, "Error decoding args"); + rpcsvc_request_seterr (req, GARBAGE_ARGS); + goto rpcerr; + } + + ret = nfs3_read (req, &fh, args.offset, args.count); + if ((ret < 0) && (ret != RPCSVC_ACTOR_IGNORE)) { + gf_log (GF_NFS3, GF_LOG_ERROR, "READ procedure failed"); + rpcsvc_request_seterr (req, SYSTEM_ERR); + ret = RPCSVC_ACTOR_ERROR; + } + +rpcerr: + return ret; +} + + +int +nfs3_write_reply (rpcsvc_request_t *req, nfsstat3 stat, count3 count, + stable_how stable, uint64_t wverf, struct iatt *prestat, + struct iatt *poststat) +{ + write3res res = {0, }; + uint64_t deviceid = 0; + + deviceid = nfs3_request_xlator_deviceid (req); + nfs3_fill_write3res (&res, stat, count, stable, wverf, prestat, + poststat, deviceid); + nfs3svc_submit_reply (req, (void *)&res, + (nfs3_serializer)xdr_serialize_write3res); + + return 0; +} + +int32_t +nfs3svc_write_fsync_cbk (call_frame_t *frame, void *cookie, xlator_t *this, + int32_t op_ret, int32_t op_errno, struct iatt *prebuf, + struct iatt *postbuf, dict_t *xdata) +{ + struct nfs3_state *nfs3 = NULL; + nfsstat3 stat = NFS3ERR_SERVERFAULT; + nfs3_call_state_t *cs = NULL; + + cs = frame->local; + nfs3 = rpcsvc_request_program_private (cs->req); + + if (op_ret == -1) { + gf_log (GF_NFS, GF_LOG_WARNING, + "%x: %s => -1 (%s)", rpcsvc_request_xid (cs->req), + cs->resolvedloc.path, strerror (op_errno)); + stat = nfs3_cbk_errno_status (op_ret, op_errno); + } else + stat = NFS3_OK; + + nfs3_log_write_res (rpcsvc_request_xid (cs->req), stat, op_errno, + cs->maxcount, cs->writetype, nfs3->serverstart); + nfs3_write_reply (cs->req, stat, cs->maxcount, cs->writetype, + nfs3->serverstart, &cs->stbuf, postbuf); + nfs3_call_state_wipe (cs); + return 0; +} + + + +/* + * Before going into the write reply logic, here is a matrix that shows the + * requirements for a write reply as given by RFC1813. + * + * Requested Write Type || Possible Returns + * ============================================== + * FILE_SYNC || FILE_SYNC + * DATA_SYNC || DATA_SYNC or FILE_SYNC + * UNSTABLE || DATA_SYNC or FILE_SYNC or UNSTABLE + * + * Write types other than UNSTABLE are together called STABLE. + * RS - Return Stable + * RU - Return Unstable + * WS - Write Stable + * WU - Write Unstable + * + *+============================================+ + *| Vol Opts -> || trusted-write| trusted-sync | + *| Write Type || | | + *|-------------||--------------|--------------| + *| STABLE || WS | WU | + *| || RS | RS | + *|-------------||--------------|--------------| + *| UNSTABLE || WU | WU | + *| || RS | RS | + *|-------------||--------------|--------------| + *| COMMIT || fsync | getattr | + *+============================================+ + * + * + */ +int32_t +nfs3svc_write_cbk (call_frame_t *frame, void *cookie, xlator_t *this, + int32_t op_ret, int32_t op_errno, struct iatt *prebuf, + struct iatt *postbuf, dict_t *xdata) +{ + nfsstat3 stat = NFS3ERR_SERVERFAULT; + nfs3_call_state_t *cs = NULL; + struct nfs3_state *nfs3 = NULL; + + cs = frame->local; + nfs3 = rpcsvc_request_program_private (cs->req); + if (op_ret == -1) { + gf_log (GF_NFS, GF_LOG_WARNING, + "%x: %s => -1 (%s)", rpcsvc_request_xid (cs->req), + cs->resolvedloc.path, strerror (op_errno)); + stat = nfs3_cbk_errno_status (op_ret, op_errno); + goto err; + } + + stat = NFS3_OK; + cs->maxcount = op_ret; + +err: + nfs3_log_write_res (rpcsvc_request_xid (cs->req), stat, + op_errno, cs->maxcount, cs->writetype, + nfs3->serverstart); + nfs3_write_reply (cs->req, stat, cs->maxcount, + cs->writetype, nfs3->serverstart, prebuf, + postbuf); + nfs3_call_state_wipe (cs); + + return 0; +} + + +int +__nfs3_write_resume (nfs3_call_state_t *cs) +{ + int ret = -EFAULT; + nfs_user_t nfu = {0, }; + + if (!cs) + return ret; + + nfs_request_user_init (&nfu, cs->req); + /* It is possible that the RPC record contains more bytes than + * than the size of write requested in this request. This means, + * that in the RPC message buffer, there could be more bytes + * beyind the @count bytes. Since @payload is referring to the write + * data directly inside the RPC request buffer(..since we performed a + * no-copy deXDRing..), we might end up writing more data than + * requested, because till now payload.iov_len accounts for all the + * bytes not just the write request bytes. These extra bytes are present + * as a requirement of the XDR encoding to round up the all string and + * opaque data buffers to multiples of 4 bytes. + */ + cs->datavec.iov_len = cs->datacount; + ret = nfs_write (cs->nfsx, cs->vol, &nfu, cs->fd, cs->iobref, + &cs->datavec, 1, cs->dataoffset, nfs3svc_write_cbk, + cs); + + return ret; +} + + +int +nfs3_write_resume (void *carg) +{ + nfsstat3 stat = NFS3ERR_SERVERFAULT; + int ret = -EFAULT; + nfs3_call_state_t *cs = NULL; + fd_t *fd = NULL; + + if (!carg) + return ret; + + cs = (nfs3_call_state_t *)carg; + nfs3_check_fh_resolve_status (cs, stat, nfs3err); + fd = fd_anonymous (cs->resolvedloc.inode); + if (!fd) { + gf_log (GF_NFS3, GF_LOG_ERROR, "Failed to create anonymous fd"); + goto nfs3err; + } + + cs->fd = fd; /* Gets unrefd when the call state is wiped. */ + +/* + enum stable_how { + UNSTABLE = 0, + DATA_SYNC = 1, + FILE_SYNC = 2, + }; +*/ + switch (cs->writetype) { + case UNSTABLE: + break; + case DATA_SYNC: + fd->flags |= O_DSYNC; + break; + case FILE_SYNC: + fd->flags |= O_SYNC; + break; + } + + ret = __nfs3_write_resume (cs); + if (ret < 0) + stat = nfs3_errno_to_nfsstat3 (-ret); +nfs3err: + if (ret < 0) { + nfs3_log_common_res (rpcsvc_request_xid (cs->req), NFS3_WRITE, + stat, -ret); + nfs3_write_reply (cs->req, stat, 0, cs->writetype, 0, NULL, + NULL); + nfs3_call_state_wipe (cs); + } + return ret; +} + + +int +nfs3_write (rpcsvc_request_t *req, struct nfs3_fh *fh, offset3 offset, + count3 count, stable_how stable, struct iovec payload, + struct iobref *iobref) +{ + xlator_t *vol = NULL; + nfsstat3 stat = NFS3ERR_SERVERFAULT; + int ret = -EFAULT; + struct nfs3_state *nfs3 = NULL; + nfs3_call_state_t *cs = NULL; + + if ((!req) || (!fh) || (!payload.iov_base)) { + gf_log (GF_NFS3, GF_LOG_ERROR, "Bad arguments"); + return -1; + } + + nfs3_log_rw_call (rpcsvc_request_xid (req), "WRITE", fh, offset, + count, stable); + nfs3_validate_gluster_fh (fh, stat, nfs3err); + nfs3_validate_nfs3_state (req, nfs3, stat, nfs3err, ret); + nfs3_map_fh_to_volume (nfs3, fh, req, vol, stat, nfs3err); + nfs3_volume_started_check (nfs3, vol, ret, out); + nfs3_check_rw_volaccess (nfs3, fh->exportid, stat, nfs3err); + nfs3_handle_call_state_init (nfs3, cs, req, vol, stat, nfs3err); + cs->datacount = count; + cs->dataoffset = offset; + cs->writetype = stable; + cs->iobref = iobref; + cs->datavec = payload; + + + ret = nfs3_fh_resolve_and_resume (cs, fh, NULL, nfs3_write_resume); + if (ret < 0) + stat = nfs3_errno_to_nfsstat3 (-ret); + +nfs3err: + if (ret < 0) { + nfs3_log_common_res (rpcsvc_request_xid (req), NFS3_WRITE, + stat, -ret); + nfs3_write_reply (req, stat, 0, stable, 0, NULL, NULL); + nfs3_call_state_wipe (cs); + ret = 0; + } +out: + return ret; +} + +#define NFS3_VECWRITE_READFHLEN 1 +#define NFS3_VECWRITE_READFH 2 +#define NFS3_VECWRITE_READREST 3 + +#define NFS3_WRITE_POSTFH_SIZE 20 + + +int +nfs3svc_write_vecsizer (int state, ssize_t *readsize, char *base_addr, + char *curr_addr) +{ + int ret = 0; + uint32_t fhlen = 0; + uint32_t fhlen_n = 0; + + if (state == 0) { + ret = NFS3_VECWRITE_READFHLEN; + *readsize = 4; + } else if (state == NFS3_VECWRITE_READFHLEN) { + fhlen_n = *(uint32_t *)(curr_addr - 4); + fhlen = ntohl (fhlen_n); + *readsize = xdr_length_round_up (fhlen, NFS3_FHSIZE); + ret = NFS3_VECWRITE_READFH; + } else if (state == NFS3_VECWRITE_READFH) { + *readsize = NFS3_WRITE_POSTFH_SIZE; + ret = NFS3_VECWRITE_READREST; + } else if (state == NFS3_VECWRITE_READREST) { + ret = 0; + *readsize = 0; + } else + gf_log ("nfs", GF_LOG_ERROR, "state wrong"); + + return ret; +} + + +int +nfs3svc_write (rpcsvc_request_t *req) +{ + struct nfs3_fh fh = {{0}, }; + write3args args; + int ret = RPCSVC_ACTOR_ERROR; + + if (!req) + return ret; + nfs3_prep_write3args (&args, &fh); + if (xdr_to_write3args (req->msg[0], &args) <= 0) { + gf_log (GF_NFS3, GF_LOG_ERROR, "Error decoding args"); + rpcsvc_request_seterr (req, GARBAGE_ARGS); + goto rpcerr; + } + + /* To ensure that the iobuf for the current record does not + * get returned to the iobpool, we need to keep a reference for + * ourselves because the RPC call handler who called us will unref its + * own ref of the record's iobuf when it is done handling the request. + */ + + ret = nfs3_write (req, &fh, args.offset, args.count, args.stable, + req->msg[1], rpcsvc_request_iobref_ref (req)); + if ((ret < 0) && (ret != RPCSVC_ACTOR_IGNORE)) { + gf_log (GF_NFS3, GF_LOG_ERROR, "WRITE procedure failed"); + rpcsvc_request_seterr (req, SYSTEM_ERR); + ret = RPCSVC_ACTOR_ERROR; + } + +rpcerr: + return ret; +} + + +int +nfs3_create_reply (rpcsvc_request_t *req, nfsstat3 stat, struct nfs3_fh *newfh, + struct iatt *newbuf, struct iatt *preparent, + struct iatt *postparent) +{ + create3res res = {0, }; + uint64_t deviceid = 0; + + deviceid = nfs3_request_xlator_deviceid (req); + nfs3_fill_create3res (&res, stat, newfh, newbuf, preparent, postparent, + deviceid); + nfs3svc_submit_reply (req, (void *)&res, + (nfs3_serializer)xdr_serialize_create3res); + return 0; +} + + +int32_t +nfs3svc_create_setattr_cbk (call_frame_t *frame, void *cookie, xlator_t *this, + int32_t op_ret, int32_t op_errno, + struct iatt *preop, struct iatt *postop, dict_t *xdata) +{ + nfsstat3 stat = NFS3ERR_SERVERFAULT; + nfs3_call_state_t *cs = NULL; + + cs = frame->local; + if (op_ret == -1) { + gf_log (GF_NFS, GF_LOG_WARNING, + "%x: %s => -1 (%s)", rpcsvc_request_xid (cs->req), + cs->resolvedloc.path, strerror (op_errno)); + stat = nfs3_cbk_errno_status (op_ret, op_errno); + goto nfs3err; + } + + stat = NFS3_OK; +nfs3err: + nfs3_log_newfh_res (rpcsvc_request_xid (cs->req), NFS3_CREATE, stat, + op_errno, &cs->fh); + nfs3_create_reply (cs->req, stat, &cs->fh, postop, &cs->preparent, + &cs->postparent); + nfs3_call_state_wipe (cs); + + return 0; +} + + +int32_t +nfs3svc_create_cbk (call_frame_t *frame, void *cookie, xlator_t *this, + int32_t op_ret, int32_t op_errno, fd_t *fd, inode_t *inode, + struct iatt *buf, struct iatt *preparent, + struct iatt *postparent, dict_t *xdata) +{ + nfsstat3 stat = NFS3ERR_SERVERFAULT; + int ret = -EFAULT; + nfs_user_t nfu = {0, }; + nfs3_call_state_t *cs = NULL; + inode_t *oldinode = NULL; + + cs = frame->local; + if (op_ret == -1) { + gf_log (GF_NFS, GF_LOG_WARNING, + "%x: %s => -1 (%s)", rpcsvc_request_xid (cs->req), + cs->resolvedloc.path, strerror (op_errno)); + stat = nfs3_cbk_errno_status (op_ret, op_errno); + goto nfs3err; + } + + nfs3_fh_build_child_fh (&cs->parent, buf, &cs->fh); + oldinode = inode_link (inode, cs->resolvedloc.parent, + cs->resolvedloc.name, buf); + + /* Means no attributes were required to be set. */ + if (!cs->setattr_valid) { + stat = NFS3_OK; + ret = -1; + goto nfs3err; + } + + cs->preparent = *preparent; + cs->postparent = *postparent; + nfs_request_user_init (&nfu, cs->req); + uuid_copy (cs->resolvedloc.gfid, inode->gfid); + ret = nfs_setattr (cs->nfsx, cs->vol, &nfu, &cs->resolvedloc,&cs->stbuf, + cs->setattr_valid, nfs3svc_create_setattr_cbk, cs); + if (ret < 0) + stat = nfs3_errno_to_nfsstat3 (-ret); + +nfs3err: + if (oldinode) { + inode_lookup (oldinode); + inode_unref (oldinode); + } + + if (ret < 0) { + nfs3_log_newfh_res (rpcsvc_request_xid (cs->req), NFS3_CREATE, + stat, op_errno, &cs->fh); + nfs3_create_reply (cs->req, stat, &cs->fh, buf, preparent, + postparent); + nfs3_call_state_wipe (cs); + } + + return 0; +} + +int +nfs3_create_common (nfs3_call_state_t *cs) +{ + int ret = -EFAULT; + int flags = 0; + nfs_user_t nfu = {0, }; + uid_t uid = 0; + gid_t gid = 0; + + if (!cs) + return ret; + + if (cs->createmode == GUARDED) + flags = (O_RDWR | O_EXCL); + else + flags = O_RDWR; + + if (gf_attr_uid_set (cs->setattr_valid)) { + uid = cs->stbuf.ia_uid; + cs->setattr_valid &= ~GF_SET_ATTR_UID; + } else + uid = rpcsvc_request_uid (cs->req); + + if (gf_attr_gid_set (cs->setattr_valid)) { + gid = cs->stbuf.ia_gid; + cs->setattr_valid &= ~GF_SET_ATTR_GID; + } else + gid = rpcsvc_request_gid (cs->req); + + nfs_request_primary_user_init (&nfu, cs->req, uid, gid); + /* We can avoid sending the setattr call later if only the mode is + * required to be set. This is possible because the create fop allows + * us to specify a mode arg. + */ + if (cs->setattr_valid & GF_SET_ATTR_MODE) { + cs->setattr_valid &= ~GF_SET_ATTR_MODE; + ret = nfs_create (cs->nfsx, cs->vol, &nfu, &cs->resolvedloc, + flags, cs->mode, nfs3svc_create_cbk, cs); + } else + ret = nfs_create (cs->nfsx, cs->vol, &nfu, &cs->resolvedloc, + flags, NFS_DEFAULT_CREATE_MODE, + nfs3svc_create_cbk, cs); + + return ret; +} + + +int32_t +nfs3svc_create_stat_cbk (call_frame_t *frame, void *cookie, xlator_t *this, + int32_t op_ret, int32_t op_errno, struct iatt *buf, + dict_t *xdata) +{ + int ret = -EFAULT; + nfsstat3 stat = NFS3ERR_SERVERFAULT; + nfs_user_t nfu = {0, }; + nfs3_call_state_t *cs = NULL; + + cs = frame->local; + nfs_request_user_init (&nfu, cs->req); + if (op_ret == -1) { + gf_log (GF_NFS, GF_LOG_WARNING, + "%x: %s => -1 (%s)", rpcsvc_request_xid (cs->req), + cs->resolvedloc.path, strerror (op_errno)); + ret = -op_errno; + stat = nfs3_cbk_errno_status (op_ret, op_errno); + goto nfs3err; + } + + if ((cs->stbuf.ia_mtime == buf->ia_mtime) && + (cs->stbuf.ia_atime == buf->ia_atime)) { + gf_log (GF_NFS3, GF_LOG_DEBUG, + "Create req retransmitted verf %x %x", + cs->stbuf.ia_mtime, cs->stbuf.ia_atime); + stat = NFS3_OK; + nfs3_fh_build_child_fh (&cs->parent, buf, &cs->fh); + } else { + gf_log (GF_NFS3, GF_LOG_DEBUG, + "File already exist new_verf %x %x" + "old_verf %x %x", cs->stbuf.ia_mtime, + cs->stbuf.ia_atime, + buf->ia_mtime, buf->ia_atime); + stat = NFS3ERR_EXIST; + } + +nfs3err: + if (ret < 0) { + nfs3_log_common_res (rpcsvc_request_xid (cs->req), NFS3_CREATE, + stat, op_errno); + nfs3_create_reply (cs->req, stat, &cs->fh, buf, NULL, NULL); + nfs3_call_state_wipe (cs); + } + + return 0; +} + + +int +nfs3_create_exclusive (nfs3_call_state_t *cs) +{ + int ret = -EFAULT; + nfs_user_t nfu = {0, }; + + if (!cs) + return ret; + + /* Storing verifier as a mtime and atime attribute, to store it + * in stable storage */ + memcpy (&cs->stbuf.ia_atime, &cs->cookieverf, + sizeof (cs->stbuf.ia_atime)); + memcpy (&cs->stbuf.ia_mtime, + ((char *) &cs->cookieverf) + sizeof (cs->stbuf.ia_atime), + sizeof (cs->stbuf.ia_mtime)); + cs->setattr_valid |= GF_SET_ATTR_ATIME; + cs->setattr_valid |= GF_SET_ATTR_MTIME; + nfs_request_user_init (&nfu, cs->req); + + /* If the file already existed we need to get that attributes so we can + * compare and check whether a previous create operation was + * interrupted due to server failure or dropped packets. + */ + if ((cs->resolve_ret == 0) || + ((cs->resolve_ret == -1) && (cs->resolve_errno != ENOENT))) { + ret = nfs_stat (cs->nfsx, cs->vol, &nfu, &cs->resolvedloc, + nfs3svc_create_stat_cbk, cs); + goto nfs3err; + } + + ret = nfs3_create_common (cs); +nfs3err: + return ret; +} + + +int +nfs3_create_resume (void *carg) +{ + nfsstat3 stat = NFS3ERR_SERVERFAULT; + int ret = -EFAULT; + nfs3_call_state_t *cs = NULL; + + if (!carg) + return ret; + + cs = (nfs3_call_state_t *)carg; + nfs3_check_new_fh_resolve_status (cs, stat, nfs3err); + if (cs->createmode == EXCLUSIVE) + ret = nfs3_create_exclusive (cs); + else + ret = nfs3_create_common (cs); + + /* Handle a failure return from either of the create functions above. */ + if (ret < 0) + stat = nfs3_errno_to_nfsstat3 (-ret); + +nfs3err: + if (ret < 0) { + nfs3_log_common_res (rpcsvc_request_xid (cs->req), NFS3_CREATE, + stat, -ret); + nfs3_create_reply (cs->req, stat, NULL, NULL, NULL, NULL); + nfs3_call_state_wipe (cs); + } + + return ret; +} + +int +nfs3_create (rpcsvc_request_t *req, struct nfs3_fh *dirfh, char *name, + createmode3 mode, sattr3 *sattr, uint64_t cverf) +{ + xlator_t *vol = NULL; + nfsstat3 stat = NFS3ERR_SERVERFAULT; + int ret = -EFAULT; + struct nfs3_state *nfs3 = NULL; + nfs3_call_state_t *cs = NULL; + + if ((!req) || (!dirfh) || (!name) || (!sattr)) + return -1; + + nfs3_log_create_call (rpcsvc_request_xid (req), dirfh, name, mode); + nfs3_validate_gluster_fh (dirfh, stat, nfs3err); + nfs3_validate_nfs3_state (req, nfs3, stat, nfs3err, ret); + nfs3_validate_strlen_or_goto (name, NFS_NAME_MAX, nfs3err, stat, ret); + nfs3_map_fh_to_volume (nfs3, dirfh, req, vol, stat, nfs3err); + nfs3_volume_started_check (nfs3, vol, ret, out); + nfs3_check_rw_volaccess (nfs3, dirfh->exportid, stat, nfs3err); + nfs3_handle_call_state_init (nfs3, cs, req, vol, stat, nfs3err); + + cs->cookieverf = cverf; + /*In Exclusive create client is supposed to send cverf instead of + * sattr*/ + if (mode != EXCLUSIVE) + cs->setattr_valid = nfs3_sattr3_to_setattr_valid (sattr, + &cs->stbuf, + &cs->mode); + cs->createmode = mode; + cs->parent = *dirfh; + + ret = nfs3_fh_resolve_and_resume (cs, dirfh, name, nfs3_create_resume); + if (ret < 0) + stat = nfs3_errno_to_nfsstat3 (-ret); + +nfs3err: + if (ret < 0) { + nfs3_log_common_res (rpcsvc_request_xid (req), NFS3_CREATE, + stat, -ret); + nfs3_create_reply (req, stat, NULL, NULL, NULL, NULL); + nfs3_call_state_wipe (cs); + ret = 0; + } +out: + return ret; +} + + +int +nfs3svc_create (rpcsvc_request_t *req) +{ + char name[NFS_PATH_MAX]; + struct nfs3_fh dirfh = {{0}, }; + create3args args; + int ret = RPCSVC_ACTOR_ERROR; + uint64_t cverf = 0; + uint64_t *cval; + + if (!req) + return ret; + + nfs3_prep_create3args (&args, &dirfh, name); + if (xdr_to_create3args (req->msg[0], &args) <= 0) { + gf_log (GF_NFS3, GF_LOG_ERROR, "Error decoding args"); + rpcsvc_request_seterr (req, GARBAGE_ARGS); + goto rpcerr; + } + + cval = (uint64_t *)args.how.createhow3_u.verf; + cverf = *cval; + + ret = nfs3_create (req, &dirfh, name, args.how.mode, + &args.how.createhow3_u.obj_attributes, cverf); + if ((ret < 0) && (ret != RPCSVC_ACTOR_IGNORE)) { + gf_log (GF_NFS3, GF_LOG_ERROR, "CREATE procedure failed"); + rpcsvc_request_seterr (req, SYSTEM_ERR); + ret = RPCSVC_ACTOR_ERROR; + } + +rpcerr: + return ret; +} + + +int +nfs3_mkdir_reply (rpcsvc_request_t *req, nfsstat3 stat, struct nfs3_fh *fh, + struct iatt *buf, struct iatt *preparent, + struct iatt *postparent) +{ + mkdir3res res = {0, }; + uint64_t deviceid = 0; + + deviceid = nfs3_request_xlator_deviceid (req); + nfs3_fill_mkdir3res (&res, stat, fh, buf, preparent, postparent, + deviceid); + nfs3svc_submit_reply (req, &res, + (nfs3_serializer)xdr_serialize_mkdir3res); + return 0; +} + +int32_t +nfs3svc_mkdir_setattr_cbk (call_frame_t *frame, void *cookie, xlator_t *this, + int32_t op_ret, int32_t op_errno, + struct iatt *preop, struct iatt *postop, dict_t *xdata) +{ + nfsstat3 stat = NFS3ERR_SERVERFAULT; + nfs3_call_state_t *cs = NULL; + + cs = frame->local; + if (op_ret == -1) { + gf_log (GF_NFS, GF_LOG_WARNING, + "%x: %s => -1 (%s)", rpcsvc_request_xid (cs->req), + cs->resolvedloc.path, strerror (op_errno)); + stat = nfs3_cbk_errno_status (op_ret, op_errno); + goto nfs3err; + } + + stat = NFS3_OK; +nfs3err: + nfs3_log_newfh_res (rpcsvc_request_xid (cs->req), NFS3_MKDIR, stat, + op_errno, &cs->fh); + nfs3_mkdir_reply (cs->req, stat, &cs->fh, postop, &cs->preparent, + &cs->postparent); + nfs3_call_state_wipe (cs); + + return 0; +} + + +int32_t +nfs3svc_mkdir_cbk (call_frame_t *frame, void *cookie, xlator_t *this, + int32_t op_ret, int32_t op_errno, inode_t *inode, + struct iatt *buf, struct iatt *preparent, + struct iatt *postparent, dict_t *xdata) +{ + nfsstat3 stat = NFS3ERR_SERVERFAULT; + int ret = -EFAULT; + nfs_user_t nfu = {0, }; + nfs3_call_state_t *cs = NULL; + + cs = frame->local; + if (op_ret == -1) { + gf_log (GF_NFS, GF_LOG_WARNING, + "%x: %s => -1 (%s)", rpcsvc_request_xid (cs->req), + cs->resolvedloc.path, strerror (op_errno)); + stat = nfs3_cbk_errno_status (op_ret, op_errno); + goto nfs3err; + } + + nfs3_fh_build_child_fh (&cs->parent, buf, &cs->fh); + + /* Means no attributes were required to be set. */ + if (!cs->setattr_valid) { + stat = NFS3_OK; + goto nfs3err; + } + + cs->preparent = *preparent; + cs->postparent = *postparent; + nfs_request_user_init (&nfu, cs->req); + ret = nfs_setattr (cs->nfsx, cs->vol, &nfu, &cs->resolvedloc,&cs->stbuf, + cs->setattr_valid, nfs3svc_mkdir_setattr_cbk, cs); + if (ret < 0) + stat = nfs3_errno_to_nfsstat3 (-ret); + +nfs3err: + if (ret < 0) { + nfs3_log_newfh_res (rpcsvc_request_xid (cs->req), NFS3_MKDIR, + stat, op_errno, &cs->fh); + nfs3_mkdir_reply (cs->req, stat, &cs->fh, buf, preparent, + postparent); + nfs3_call_state_wipe (cs); + } + + return 0; +} + + +int +nfs3_mkdir_resume (void *carg) +{ + nfsstat3 stat = NFS3ERR_SERVERFAULT; + int ret = -EFAULT; + nfs_user_t nfu = {0, }; + nfs3_call_state_t *cs = NULL; + + if (!carg) + return ret; + + cs = (nfs3_call_state_t *)carg; + nfs3_check_new_fh_resolve_status (cs, stat, nfs3err); + nfs_request_user_init (&nfu, cs->req); + + if (gf_attr_mode_set (cs->setattr_valid)) { + cs->setattr_valid &= ~GF_SET_ATTR_MODE; + ret = nfs_mkdir (cs->nfsx, cs->vol, &nfu, &cs->resolvedloc, + cs->mode, nfs3svc_mkdir_cbk, cs); + } else + ret = nfs_mkdir (cs->nfsx, cs->vol, &nfu, &cs->resolvedloc, + cs->mode, nfs3svc_mkdir_cbk, cs); + + if (ret < 0) + stat = nfs3_errno_to_nfsstat3 (-ret); + +nfs3err: + if (ret < 0) { + nfs3_log_common_res (rpcsvc_request_xid (cs->req), NFS3_MKDIR, + stat, -ret); + nfs3_mkdir_reply (cs->req, stat, NULL, NULL, NULL, NULL); + nfs3_call_state_wipe (cs); + } + + return 0; +} + + + +int +nfs3_mkdir (rpcsvc_request_t *req, struct nfs3_fh *dirfh, char *name, + sattr3 *sattr) +{ + xlator_t *vol = NULL; + nfsstat3 stat = NFS3ERR_SERVERFAULT; + int ret = -EFAULT; + struct nfs3_state *nfs3 = NULL; + nfs3_call_state_t *cs = NULL; + + if ((!req) || (!dirfh) || (!name) || (!sattr)) { + gf_log (GF_NFS3, GF_LOG_ERROR, "Bad arguments"); + return -1; + } + + nfs3_log_fh_entry_call (rpcsvc_request_xid (req), "MKDIR", dirfh, + name); + nfs3_validate_gluster_fh (dirfh, stat, nfs3err); + nfs3_validate_nfs3_state (req, nfs3, stat, nfs3err, ret); + nfs3_validate_strlen_or_goto (name, NFS_NAME_MAX, nfs3err, stat, ret); + nfs3_map_fh_to_volume (nfs3, dirfh, req, vol, stat, nfs3err); + nfs3_volume_started_check (nfs3, vol, ret, out); + nfs3_check_rw_volaccess (nfs3, dirfh->exportid, stat, nfs3err); + nfs3_handle_call_state_init (nfs3, cs, req, vol, stat, nfs3err); + + cs->parent = *dirfh; + cs->setattr_valid = nfs3_sattr3_to_setattr_valid (sattr, &cs->stbuf, + &cs->mode); + ret = nfs3_fh_resolve_and_resume (cs, dirfh, name, nfs3_mkdir_resume); + if (ret < 0) + stat = nfs3_errno_to_nfsstat3 (-ret); + +nfs3err: + if (ret < 0) { + nfs3_log_common_res (rpcsvc_request_xid (req), NFS3_MKDIR, + stat, -ret); + nfs3_mkdir_reply (req, stat, NULL, NULL, NULL, NULL); + nfs3_call_state_wipe (cs); + ret = 0; + } +out: + return ret; +} + + +int +nfs3svc_mkdir (rpcsvc_request_t *req) +{ + char name[NFS_PATH_MAX]; + struct nfs3_fh dirfh = {{0}, }; + mkdir3args args; + int ret = RPCSVC_ACTOR_ERROR; + + if (!req) + return ret; + nfs3_prep_mkdir3args (&args, &dirfh, name); + if (xdr_to_mkdir3args (req->msg[0], &args) <= 0) { + gf_log (GF_NFS3, GF_LOG_ERROR, "Error decoding args"); + rpcsvc_request_seterr (req, GARBAGE_ARGS); + goto rpcerr; + } + + ret = nfs3_mkdir (req, &dirfh, name, &args.attributes); + if ((ret < 0) && (ret != RPCSVC_ACTOR_IGNORE)) { + gf_log (GF_NFS3, GF_LOG_ERROR, "MKDIR procedure failed"); + rpcsvc_request_seterr (req, SYSTEM_ERR); + ret = RPCSVC_ACTOR_ERROR; + } + +rpcerr: + return ret; +} + + +int +nfs3_symlink_reply (rpcsvc_request_t *req, nfsstat3 stat, struct nfs3_fh *fh, + struct iatt *buf, struct iatt *preparent, + struct iatt *postparent) +{ + symlink3res res = {0, }; + uint64_t deviceid = 0; + + deviceid = nfs3_request_xlator_deviceid (req); + nfs3_fill_symlink3res (&res, stat, fh, buf, preparent, postparent, + deviceid); + nfs3svc_submit_reply (req, (void *)&res, + (nfs3_serializer)xdr_serialize_symlink3res); + + return 0; +} + + +int32_t +nfs3svc_symlink_cbk (call_frame_t *frame, void *cookie, xlator_t *this, + int32_t op_ret, int32_t op_errno, inode_t *inode, + struct iatt *buf, struct iatt *preparent, + struct iatt *postparent, dict_t *xdata) +{ + nfsstat3 stat = NFS3ERR_SERVERFAULT; + nfs3_call_state_t *cs = NULL; + + cs = frame->local; + if (op_ret == -1) { + gf_log (GF_NFS, GF_LOG_WARNING, + "%x: %s => -1 (%s)", rpcsvc_request_xid (cs->req), + cs->resolvedloc.path, strerror (op_errno)); + stat = nfs3_cbk_errno_status (op_ret, op_errno); + goto nfs3err; + } + + nfs3_fh_build_child_fh (&cs->parent, buf, &cs->fh); + stat = NFS3_OK; + +nfs3err: + nfs3_log_newfh_res (rpcsvc_request_xid (cs->req), NFS3_SYMLINK, stat, + op_errno, &cs->fh); + nfs3_symlink_reply (cs->req, stat, &cs->fh, buf, preparent, + postparent); + nfs3_call_state_wipe (cs); + return 0; +} + + +int +nfs3_symlink_resume (void *carg) +{ + nfsstat3 stat = NFS3ERR_SERVERFAULT; + int ret = -EFAULT; + nfs3_call_state_t *cs = NULL; + nfs_user_t nfu = {0, }; + + if (!carg) + return ret; + + cs = (nfs3_call_state_t *)carg; + nfs3_check_new_fh_resolve_status (cs, stat, nfs3err); + nfs_request_user_init (&nfu, cs->req); + ret = nfs_symlink (cs->nfsx, cs->vol, &nfu, cs->pathname, + &cs->resolvedloc, nfs3svc_symlink_cbk, cs); + if (ret < 0) + stat = nfs3_errno_to_nfsstat3 (-ret); + +nfs3err: + if (ret < 0) { + nfs3_log_common_res (rpcsvc_request_xid (cs->req), + NFS3_SYMLINK, stat, -ret); + nfs3_symlink_reply (cs->req, stat, NULL, NULL, NULL, NULL); + nfs3_call_state_wipe (cs); + } + + return ret; +} + + +int +nfs3_symlink (rpcsvc_request_t *req, struct nfs3_fh *dirfh, char *name, + char *target, sattr3 *sattr) +{ + xlator_t *vol = NULL; + nfsstat3 stat = NFS3ERR_SERVERFAULT; + int ret = -EFAULT; + struct nfs3_state *nfs3 = NULL; + nfs3_call_state_t *cs = NULL; + + if ((!req) || (!dirfh) || (!name) || (!target) || (!sattr)) { + gf_log (GF_NFS3, GF_LOG_ERROR, "Bad arguments"); + return -1; + } + + nfs3_log_symlink_call (rpcsvc_request_xid (req), dirfh, name, + target); + nfs3_validate_gluster_fh (dirfh, stat, nfs3err); + nfs3_validate_nfs3_state (req, nfs3, stat, nfs3err, ret); + nfs3_validate_strlen_or_goto (name, NFS_NAME_MAX, nfs3err, stat, ret); + nfs3_map_fh_to_volume (nfs3, dirfh, req, vol, stat, nfs3err); + nfs3_volume_started_check (nfs3, vol, ret, out); + nfs3_check_rw_volaccess (nfs3, dirfh->exportid, stat, nfs3err); + nfs3_handle_call_state_init (nfs3, cs, req, vol, stat, nfs3err); + + cs->parent = *dirfh; + cs->pathname = gf_strdup (target); + if (!cs->pathname) { + ret = -1; + stat = NFS3ERR_SERVERFAULT; + goto nfs3err; + } + + ret = nfs3_fh_resolve_and_resume (cs, dirfh, name, nfs3_symlink_resume); + if (ret < 0) + stat = nfs3_errno_to_nfsstat3 (-ret); + +nfs3err: + if (ret < 0) { + nfs3_log_common_res (rpcsvc_request_xid (req), NFS3_SYMLINK, + stat, -ret); + nfs3_symlink_reply (req, stat, NULL, NULL, NULL, NULL); + nfs3_call_state_wipe (cs); + /* Ret must be 0 after this so that the caller does not + * also send an RPC reply. + */ + ret = 0; + } +out: + return ret; +} + + +int +nfs3svc_symlink (rpcsvc_request_t *req) +{ + char name[NFS_PATH_MAX]; + struct nfs3_fh dirfh = {{0}, }; + char target[NFS_PATH_MAX]; + symlink3args args; + int ret = RPCSVC_ACTOR_ERROR; + + if (!req) + return ret; + nfs3_prep_symlink3args (&args, &dirfh, name, target); + if (xdr_to_symlink3args (req->msg[0], &args) <= 0) { + gf_log (GF_NFS3, GF_LOG_ERROR, "Error decoding args"); + rpcsvc_request_seterr (req, GARBAGE_ARGS); + goto rpcerr; + } + + ret = nfs3_symlink (req, &dirfh, name, target, + &args.symlink.symlink_attributes); + if ((ret < 0) && (ret != RPCSVC_ACTOR_IGNORE)) { + gf_log (GF_NFS3, GF_LOG_ERROR, "SYMLINK procedure failed"); + rpcsvc_request_seterr (req, SYSTEM_ERR); + ret = RPCSVC_ACTOR_ERROR; + } + +rpcerr: + return ret; +} + + +int +nfs3_mknod_reply (rpcsvc_request_t *req, nfsstat3 stat, struct nfs3_fh *fh, + struct iatt *buf, struct iatt *preparent, + struct iatt *postparent) +{ + mknod3res res = {0, }; + uint64_t deviceid = 0; + + deviceid = nfs3_request_xlator_deviceid (req); + nfs3_fill_mknod3res (&res, stat, fh, buf, preparent, postparent, + deviceid); + nfs3svc_submit_reply (req, (void *)&res, + (nfs3_serializer)xdr_serialize_mknod3res); + + return 0; +} + +int32_t +nfs3svc_mknod_setattr_cbk (call_frame_t *frame, void *cookie, xlator_t *this, + int32_t op_ret, int32_t op_errno, + struct iatt *preop, struct iatt *postop, dict_t *xdata) +{ + nfsstat3 stat = NFS3ERR_SERVERFAULT; + nfs3_call_state_t *cs = NULL; + + cs = frame->local; + if (op_ret == -1) { + gf_log (GF_NFS, GF_LOG_WARNING, + "%x: %s => -1 (%s)", rpcsvc_request_xid (cs->req), + cs->resolvedloc.path, strerror (op_errno)); + stat = nfs3_cbk_errno_status (op_ret, op_errno); + goto nfs3err; + } + + stat = NFS3_OK; +nfs3err: + nfs3_log_newfh_res (rpcsvc_request_xid (cs->req), NFS3_MKNOD, stat, + op_errno, &cs->fh); + nfs3_mknod_reply (cs->req, stat, &cs->fh, postop, &cs->preparent, + &cs->postparent); + nfs3_call_state_wipe (cs); + return 0; +} + + + +int32_t +nfs3svc_mknod_cbk (call_frame_t *frame, void *cookie, xlator_t *this, + int32_t op_ret, int32_t op_errno, inode_t *inode, + struct iatt *buf, struct iatt *preparent, + struct iatt *postparent, dict_t *xdata) +{ + nfsstat3 stat = NFS3ERR_SERVERFAULT; + int ret = -1; + nfs_user_t nfu = {0, }; + nfs3_call_state_t *cs = NULL; + + cs = frame->local; + if (op_ret == -1) { + gf_log (GF_NFS, GF_LOG_WARNING, + "%x: %s => -1 (%s)", rpcsvc_request_xid (cs->req), + cs->resolvedloc.path, strerror (op_errno)); + stat = nfs3_cbk_errno_status (op_ret, op_errno); + goto nfs3err; + } + + nfs3_fh_build_child_fh (&cs->parent, buf, &cs->fh); + + /* Means no attributes were required to be set. */ + if (!cs->setattr_valid) { + stat = NFS3_OK; + ret = -1; + goto nfs3err; + } + + cs->preparent = *preparent; + cs->postparent = *postparent; + nfs_request_user_init (&nfu, cs->req); + ret = nfs_setattr (cs->nfsx, cs->vol, &nfu, &cs->resolvedloc,&cs->stbuf, + cs->setattr_valid, nfs3svc_mknod_setattr_cbk, cs); + if (ret < 0) + stat = nfs3_errno_to_nfsstat3 (-ret); +nfs3err: + if (ret < 0) { + nfs3_log_newfh_res (rpcsvc_request_xid (cs->req), NFS3_MKNOD, + stat, + op_errno, &cs->fh); + nfs3_mknod_reply (cs->req, stat, &cs->fh, buf, preparent, + postparent); + nfs3_call_state_wipe (cs); + } + + return 0; +} + + +int +nfs3_mknod_device (nfs3_call_state_t *cs) +{ + int ret = -EFAULT; + dev_t devnum = 0; + mode_t mode = 0; + nfs_user_t nfu = {0, }; + + if (!cs) + return ret; + + devnum = makedev (cs->devnums.specdata1, cs->devnums.specdata2); + if (cs->mknodtype == NF3CHR) + mode = S_IFCHR; + else + mode = S_IFBLK; + + nfs_request_user_init (&nfu, cs->req); + if (gf_attr_mode_set (cs->setattr_valid)) { + cs->setattr_valid &= ~GF_SET_ATTR_MODE; + mode |= cs->mode; + ret = nfs_mknod (cs->nfsx, cs->vol, &nfu, &cs->resolvedloc, + mode, devnum, nfs3svc_mknod_cbk, cs); + } else + ret = nfs_mknod (cs->nfsx, cs->vol, &nfu, &cs->resolvedloc, + mode, devnum, nfs3svc_mknod_cbk, cs); + + return ret; +} + + +int +nfs3_mknod_fifo (nfs3_call_state_t *cs) +{ + int ret = -EFAULT; + nfs_user_t nfu = {0, }; + mode_t mode = S_IFIFO; + + if (!cs) + return ret; + + nfs_request_user_init (&nfu, cs->req); + if (gf_attr_mode_set (cs->setattr_valid)) { + cs->setattr_valid &= ~GF_SET_ATTR_MODE; + mode |= cs->mode; + ret = nfs_mknod (cs->nfsx, cs->vol, &nfu, &cs->resolvedloc, + mode, 0, nfs3svc_mknod_cbk, cs); + } else + ret = nfs_mknod (cs->nfsx, cs->vol, &nfu, &cs->resolvedloc, + mode, 0, nfs3svc_mknod_cbk, cs); + + return ret; +} + + +int +nfs3_mknod_resume (void *carg) +{ + nfsstat3 stat = NFS3ERR_SERVERFAULT; + int ret = -EFAULT; + nfs3_call_state_t *cs = NULL; + + if (!carg) + return ret; + + cs = (nfs3_call_state_t *)carg; + nfs3_check_new_fh_resolve_status (cs, stat, nfs3err); + switch (cs->mknodtype) { + + case NF3CHR: + case NF3BLK: + ret = nfs3_mknod_device (cs); + break; + case NF3SOCK: + case NF3FIFO: + ret = nfs3_mknod_fifo (cs); + break; + default: + ret = -EBADF; + break; + } + + if (ret < 0) + stat = nfs3_errno_to_nfsstat3 (-ret); + +nfs3err: + if (ret < 0) { + nfs3_log_common_res (rpcsvc_request_xid (cs->req), NFS3_MKNOD, + stat, -ret); + nfs3_mknod_reply (cs->req, stat, NULL, NULL, NULL, NULL); + nfs3_call_state_wipe (cs); + } + + return ret; +} + + + +int +nfs3_mknod (rpcsvc_request_t *req, struct nfs3_fh *fh, char *name, + mknoddata3 *nodedata) +{ + xlator_t *vol = NULL; + nfsstat3 stat = NFS3ERR_SERVERFAULT; + int ret = -EFAULT; + struct nfs3_state *nfs3 = NULL; + nfs3_call_state_t *cs = NULL; + sattr3 *sattr = NULL; + + if ((!req) || (!fh) || (!name) || (!nodedata)) { + gf_log (GF_NFS3, GF_LOG_ERROR, "Bad arguments"); + return -1; + } + + nfs3_log_mknod_call (rpcsvc_request_xid (req), fh, name, + nodedata->type); + nfs3_validate_gluster_fh (fh, stat, nfs3err); + nfs3_validate_nfs3_state (req, nfs3, stat, nfs3err, ret); + nfs3_validate_strlen_or_goto (name, NFS_NAME_MAX, nfs3err, stat, ret); + nfs3_map_fh_to_volume (nfs3, fh, req, vol, stat, nfs3err); + nfs3_volume_started_check (nfs3, vol, ret, out); + nfs3_check_rw_volaccess (nfs3, fh->exportid, stat, nfs3err); + nfs3_handle_call_state_init (nfs3, cs, req, vol, stat, nfs3err); + + cs->mknodtype = nodedata->type; + switch (nodedata->type) { + case NF3CHR: + case NF3BLK: + cs->devnums = nodedata->mknoddata3_u.device.spec; + sattr = &nodedata->mknoddata3_u.device.dev_attributes; + cs->setattr_valid = nfs3_sattr3_to_setattr_valid (sattr, + &cs->stbuf, + &cs->mode); + break; + case NF3SOCK: + case NF3FIFO: + sattr = &nodedata->mknoddata3_u.pipe_attributes; + cs->setattr_valid = nfs3_sattr3_to_setattr_valid (sattr, + &cs->stbuf, + &cs->mode); + break; + default: + ret = -EBADF; + break; + } + + cs->parent = *fh; + ret = nfs3_fh_resolve_and_resume (cs, fh, name, nfs3_mknod_resume); + if (ret < 0) + stat = nfs3_errno_to_nfsstat3 (-ret); + +nfs3err: + if (ret < 0) { + nfs3_log_common_res (rpcsvc_request_xid (req), NFS3_MKNOD, + stat, -ret); + nfs3_mknod_reply (req, stat, NULL, NULL, NULL, NULL); + /* Ret must be 0 after this so that the caller does not + * also send an RPC reply. + */ + nfs3_call_state_wipe (cs); + ret = 0; + } +out: + return ret; +} + + +int +nfs3svc_mknod (rpcsvc_request_t *req) +{ + char name[NFS_PATH_MAX]; + struct nfs3_fh fh = {{0}, }; + mknod3args args; + int ret = RPCSVC_ACTOR_ERROR; + + if (!req) + return ret; + nfs3_prep_mknod3args (&args, &fh, name); + if (xdr_to_mknod3args (req->msg[0], &args) <= 0) { + gf_log (GF_NFS3, GF_LOG_ERROR, "Error decoding args"); + rpcsvc_request_seterr (req, GARBAGE_ARGS); + goto rpcerr; + } + + ret = nfs3_mknod (req, &fh, name, &args.what); + if ((ret < 0) && (ret != RPCSVC_ACTOR_IGNORE)) { + gf_log (GF_NFS3, GF_LOG_ERROR, "MKNOD procedure failed"); + rpcsvc_request_seterr (req, SYSTEM_ERR); + ret = RPCSVC_ACTOR_ERROR; + } + +rpcerr: + return ret; +} + +int +nfs3_remove_reply (rpcsvc_request_t *req, nfsstat3 stat, struct iatt *preparent + , struct iatt *postparent) +{ + remove3res res = {0, }; + uint64_t deviceid = 0; + + deviceid = nfs3_request_xlator_deviceid (req); + nfs3_fill_remove3res (&res, stat, preparent, postparent, deviceid); + nfs3svc_submit_reply (req, (void *)&res, + (nfs3_serializer)xdr_serialize_remove3res); + return 0; +} + + + +int32_t +nfs3svc_remove_cbk (call_frame_t *frame, void *cookie, xlator_t *this, + int32_t op_ret, int32_t op_errno, struct iatt *preparent, + struct iatt *postparent, dict_t *xdata) +{ + nfsstat3 stat = NFS3ERR_SERVERFAULT; + nfs3_call_state_t *cs = NULL; + + cs = frame->local; + if (op_ret == -1) { + gf_log (GF_NFS, GF_LOG_WARNING, + "%x: %s => -1 (%s)", rpcsvc_request_xid (cs->req), + cs->resolvedloc.path, strerror (op_errno)); + stat = nfs3_cbk_errno_status (op_ret, op_errno); + } + + if (op_ret == 0) + stat = NFS3_OK; + + nfs3_log_common_res (rpcsvc_request_xid (cs->req), NFS3_REMOVE, stat, + op_errno); + nfs3_remove_reply (cs->req, stat, preparent, postparent); + nfs3_call_state_wipe (cs); + + return 0; +} + + +int +__nfs3_remove (nfs3_call_state_t *cs) +{ + int ret = -EFAULT; + nfs_user_t nfu = {0, }; + ia_type_t type = 0; + + if (!cs) + return ret; + type = cs->resolvedloc.inode->ia_type; + nfs_request_user_init (&nfu, cs->req); + if (IA_ISDIR (type)) + ret = nfs_rmdir (cs->nfsx, cs->vol, &nfu, &cs->resolvedloc, + nfs3svc_remove_cbk, cs); + else + ret = nfs_unlink (cs->nfsx, cs->vol, &nfu, &cs->resolvedloc, + nfs3svc_remove_cbk, cs); + + return ret; +} + + +int +nfs3_remove_resume (void *carg) +{ + nfsstat3 stat = NFS3ERR_SERVERFAULT; + int ret = -EFAULT; + nfs3_call_state_t *cs = NULL; + + if (!carg) + return ret; + + cs = (nfs3_call_state_t *)carg; + nfs3_check_fh_resolve_status (cs, stat, nfs3err); + ret = __nfs3_remove (cs); + if (ret < 0) + stat = nfs3_errno_to_nfsstat3 (-ret); + +nfs3err: + if (ret < 0) { + nfs3_log_common_res (rpcsvc_request_xid (cs->req), NFS3_REMOVE, + stat, -ret); + nfs3_remove_reply (cs->req, stat, NULL, NULL); + nfs3_call_state_wipe (cs); + } + + return ret; +} + + +int +nfs3_remove (rpcsvc_request_t *req, struct nfs3_fh *fh, char *name) +{ + xlator_t *vol = NULL; + nfsstat3 stat = NFS3ERR_SERVERFAULT; + int ret = -EFAULT; + struct nfs3_state *nfs3 = NULL; + nfs3_call_state_t *cs = NULL; + + if ((!req) || (!fh) || (!name)) { + gf_log (GF_NFS3, GF_LOG_ERROR, "Bad arguments"); + return -1; + } + + nfs3_log_fh_entry_call (rpcsvc_request_xid (req), "REMOVE", fh, + name); + nfs3_validate_gluster_fh (fh, stat, nfs3err); + nfs3_validate_nfs3_state (req, nfs3, stat, nfs3err, ret); + nfs3_validate_strlen_or_goto (name, NFS_NAME_MAX, nfs3err, stat, ret); + nfs3_map_fh_to_volume (nfs3, fh, req, vol, stat, nfs3err); + nfs3_volume_started_check (nfs3, vol, ret, out); + nfs3_check_rw_volaccess (nfs3, fh->exportid, stat, nfs3err); + nfs3_handle_call_state_init (nfs3, cs, req, vol, stat, nfs3err); + + ret = nfs3_fh_resolve_and_resume (cs, fh, name, nfs3_remove_resume); + if (ret < 0) + stat = nfs3_errno_to_nfsstat3 (-ret); + +nfs3err: + if (ret < 0) { + nfs3_log_common_res (rpcsvc_request_xid (req), NFS3_REMOVE, + stat, -ret); + nfs3_remove_reply (req, stat, NULL, NULL); + nfs3_call_state_wipe (cs); + /* Ret must be 0 after this so that the caller does not + * also send an RPC reply. + */ + ret = 0; + } +out: + return ret; +} + + +int +nfs3svc_remove (rpcsvc_request_t *req) +{ + char name[NFS_PATH_MAX]; + struct nfs3_fh fh = {{0}, }; + remove3args args; + int ret = RPCSVC_ACTOR_ERROR; + + if (!req) + return ret; + nfs3_prep_remove3args (&args, &fh, name); + if (xdr_to_remove3args (req->msg[0], &args) <= 0) { + gf_log (GF_NFS3, GF_LOG_ERROR, "Error decoding args"); + rpcsvc_request_seterr (req, GARBAGE_ARGS); + goto rpcerr; + } + + ret = nfs3_remove (req, &fh, name); + if ((ret < 0) && (ret != RPCSVC_ACTOR_IGNORE)) { + gf_log (GF_NFS3, GF_LOG_ERROR, "REMOVE procedure failed"); + rpcsvc_request_seterr (req, SYSTEM_ERR); + ret = RPCSVC_ACTOR_ERROR; + } + +rpcerr: + return ret; +} + + +int +nfs3_rmdir_reply (rpcsvc_request_t *req, nfsstat3 stat, struct iatt *preparent, + struct iatt *postparent) +{ + rmdir3res res = {0, }; + uint64_t deviceid = 0; + + deviceid = nfs3_request_xlator_deviceid (req); + nfs3_fill_rmdir3res (&res, stat, preparent, postparent, deviceid); + nfs3svc_submit_reply (req, (void *)&res, + (nfs3_serializer)xdr_serialize_rmdir3res); + return 0; +} + + +int32_t +nfs3svc_rmdir_cbk (call_frame_t *frame, void *cookie, xlator_t *this, + int32_t op_ret, int32_t op_errno, struct iatt *preparent, + struct iatt *postparent, dict_t *xdata) +{ + nfsstat3 stat = NFS3ERR_SERVERFAULT; + nfs3_call_state_t *cs = NULL; + + cs = frame->local; + if (op_ret == -1) { + gf_log (GF_NFS, GF_LOG_WARNING, + "%x: %s => -1 (%s)", rpcsvc_request_xid (cs->req), + cs->resolvedloc.path, strerror (op_errno)); + stat = nfs3_cbk_errno_status (op_ret, op_errno); + } else { + stat = NFS3_OK; + } + + nfs3_log_common_res (rpcsvc_request_xid (cs->req), NFS3_RMDIR, stat, + op_errno); + nfs3_rmdir_reply (cs->req, stat, preparent, postparent); + nfs3_call_state_wipe (cs); + + return 0; +} + +int +nfs3_rmdir_resume (void *carg) +{ + nfsstat3 stat = NFS3ERR_SERVERFAULT; + int ret = -EFAULT; + nfs3_call_state_t *cs = NULL; + nfs_user_t nfu = {0, }; + + if (!carg) + return ret; + + cs = (nfs3_call_state_t *)carg; + nfs3_check_fh_resolve_status (cs, stat, nfs3err); + nfs_request_user_init (&nfu, cs->req); + ret = nfs_rmdir (cs->nfsx, cs->vol, &nfu, &cs->resolvedloc, + nfs3svc_rmdir_cbk, cs); + if (ret < 0) + stat = nfs3_errno_to_nfsstat3 (-ret); + +nfs3err: + if (ret < 0) { + nfs3_log_common_res (rpcsvc_request_xid (cs->req), NFS3_RMDIR, + stat, -ret); + nfs3_rmdir_reply (cs->req, stat, NULL, NULL); + nfs3_call_state_wipe (cs); + } + + return ret; +} + + + +int +nfs3_rmdir (rpcsvc_request_t *req, struct nfs3_fh *fh, char *name) +{ + xlator_t *vol = NULL; + nfsstat3 stat = NFS3ERR_SERVERFAULT; + int ret = -EFAULT; + struct nfs3_state *nfs3 = NULL; + nfs3_call_state_t *cs = NULL; + + if ((!req) || (!fh) || (!name)) { + gf_log (GF_NFS3, GF_LOG_ERROR, "Bad arguments"); + return -1; + } + + nfs3_log_fh_entry_call (rpcsvc_request_xid (req), "RMDIR", fh, + name); + nfs3_validate_gluster_fh (fh, stat, nfs3err); + nfs3_validate_nfs3_state (req, nfs3, stat, nfs3err, ret); + nfs3_validate_strlen_or_goto (name, NFS_NAME_MAX, nfs3err, stat, ret); + nfs3_map_fh_to_volume (nfs3, fh, req, vol, stat, nfs3err); + nfs3_volume_started_check (nfs3, vol, ret, out); + nfs3_check_rw_volaccess (nfs3, fh->exportid, stat, nfs3err); + nfs3_handle_call_state_init (nfs3, cs, req, vol, stat, nfs3err); + + ret = nfs3_fh_resolve_and_resume (cs, fh, name, nfs3_rmdir_resume); + if (ret < 0) + stat = nfs3_errno_to_nfsstat3 (-ret); + +nfs3err: + if (ret < 0) { + nfs3_log_common_res (rpcsvc_request_xid (req), NFS3_RMDIR, + stat, -ret); + nfs3_rmdir_reply (req, stat, NULL, NULL); + nfs3_call_state_wipe (cs); + /* Ret must be 0 after this so that the caller does not + * also send an RPC reply. + */ + ret = 0; + } +out: + return ret; +} + + +int +nfs3svc_rmdir (rpcsvc_request_t *req) +{ + char name[NFS_PATH_MAX]; + struct nfs3_fh fh = {{0}, }; + rmdir3args args; + int ret = RPCSVC_ACTOR_ERROR; + + if (!req) + return ret; + nfs3_prep_rmdir3args (&args, &fh, name); + if (xdr_to_rmdir3args (req->msg[0], &args) <= 0) { + gf_log (GF_NFS3, GF_LOG_ERROR, "Error decoding args"); + rpcsvc_request_seterr (req, GARBAGE_ARGS); + goto rpcerr; + } + + ret = nfs3_rmdir (req, &fh, name); + if ((ret < 0) && (ret != RPCSVC_ACTOR_IGNORE)) { + gf_log (GF_NFS3, GF_LOG_ERROR, "RMDIR procedure failed"); + rpcsvc_request_seterr (req, SYSTEM_ERR); + ret = RPCSVC_ACTOR_ERROR; + } + +rpcerr: + return ret; +} + + +int +nfs3_rename_reply (rpcsvc_request_t *req, nfsstat3 stat, struct iatt *buf, + struct iatt *preoldparent, struct iatt *postoldparent, + struct iatt *prenewparent, struct iatt *postnewparent) +{ + rename3res res = {0, }; + uint64_t deviceid = 0; + + deviceid = nfs3_request_xlator_deviceid (req); + nfs3_fill_rename3res (&res, stat, buf, preoldparent, postoldparent, + prenewparent, postnewparent, deviceid); + + nfs3svc_submit_reply (req, (void *)&res, + (nfs3_serializer) xdr_serialize_rename3res); + + return 0; +} + + + +int32_t +nfs3svc_rename_cbk (call_frame_t *frame, void *cookie, xlator_t *this, + int32_t op_ret, int32_t op_errno, struct iatt *buf, + struct iatt *preoldparent, struct iatt *postoldparent, + struct iatt *prenewparent, struct iatt *postnewparent, + dict_t *xdata) +{ + int ret = -EFAULT; + nfsstat3 stat = NFS3ERR_SERVERFAULT; + nfs3_call_state_t *cs = NULL; + + cs = frame->local; + if (op_ret == -1) { + gf_log (GF_NFS, GF_LOG_WARNING, + "%x: rename %s -> %s => -1 (%s)", + rpcsvc_request_xid (cs->req), cs->oploc.path, + cs->resolvedloc.path, strerror (op_errno)); + stat = nfs3_cbk_errno_status (op_ret, op_errno); + goto nfs3err; + } + + stat = NFS3_OK; +nfs3err: + nfs3_log_common_res (rpcsvc_request_xid (cs->req), NFS3_RENAME, stat, + -ret); + nfs3_rename_reply (cs->req, stat, buf, preoldparent, postoldparent, + prenewparent, postnewparent); + nfs3_call_state_wipe (cs); + return 0; +} + + +int +nfs3_rename_resume_dst (void *carg) +{ + nfsstat3 stat = NFS3ERR_SERVERFAULT; + int ret = -EFAULT; + nfs3_call_state_t *cs = NULL; + nfs_user_t nfu = {0, }; + + if (!carg) + return ret; + + cs = (nfs3_call_state_t *)carg; + nfs3_check_new_fh_resolve_status (cs, stat, nfs3err); + cs->parent = cs->resolvefh; + nfs_request_user_init (&nfu, cs->req); + ret = nfs_rename (cs->nfsx, cs->vol, &nfu, &cs->oploc, &cs->resolvedloc, + nfs3svc_rename_cbk, cs); + if (ret < 0) + stat = nfs3_errno_to_nfsstat3 (-ret); + +nfs3err: + if (ret < 0) { + nfs3_log_common_res (rpcsvc_request_xid (cs->req), NFS3_RENAME, + stat, -ret); + nfs3_rename_reply (cs->req, stat, NULL, NULL, NULL, NULL, NULL); + nfs3_call_state_wipe (cs); + } + + return ret; +} + + + +int +nfs3_rename_resume_src (void *carg) +{ + nfsstat3 stat = NFS3ERR_SERVERFAULT; + int ret = -EFAULT; + nfs3_call_state_t *cs = NULL; + + if (!carg) + return ret; + + cs = (nfs3_call_state_t *)carg; + nfs3_check_fh_resolve_status (cs, stat, nfs3err); + /* Copy the resolved loc for the source file into another loc + * for safekeeping till we resolve the dest loc. + */ + nfs_loc_copy (&cs->oploc, &cs->resolvedloc); + nfs_loc_wipe (&cs->resolvedloc); + GF_FREE (cs->resolventry); + + ret = nfs3_fh_resolve_and_resume (cs, &cs->fh, cs->pathname, + nfs3_rename_resume_dst); + if (ret < 0) + stat = nfs3_errno_to_nfsstat3 (-ret); + +nfs3err: + if (ret < 0) { + nfs3_log_common_res (rpcsvc_request_xid (cs->req), NFS3_RENAME, + stat, -ret); + nfs3_rename_reply (cs->req, stat, NULL, NULL, NULL, NULL, NULL); + nfs3_call_state_wipe (cs); + } + + return ret; +} + + +int +nfs3_rename (rpcsvc_request_t *req, struct nfs3_fh *olddirfh, char *oldname, + struct nfs3_fh *newdirfh, char *newname) +{ + xlator_t *vol = NULL; + nfsstat3 stat = NFS3ERR_SERVERFAULT; + int ret = -EFAULT; + struct nfs3_state *nfs3 = NULL; + nfs3_call_state_t *cs = NULL; + + if ((!req) || (!olddirfh) || (!oldname) || (!newdirfh) || (!newname)) { + gf_log (GF_NFS3, GF_LOG_ERROR, "Bad arguments"); + return -1; + } + + nfs3_log_rename_call (rpcsvc_request_xid (req), olddirfh, oldname, + newdirfh, newname); + nfs3_validate_gluster_fh (olddirfh, stat, nfs3err); + nfs3_validate_gluster_fh (newdirfh, stat, nfs3err); + nfs3_validate_nfs3_state (req, nfs3, stat, nfs3err, ret); + nfs3_validate_strlen_or_goto(oldname, NFS_NAME_MAX, nfs3err, stat, ret); + nfs3_validate_strlen_or_goto(newname, NFS_NAME_MAX, nfs3err, stat, ret); + nfs3_map_fh_to_volume (nfs3, olddirfh, req, vol, stat, nfs3err); + nfs3_volume_started_check (nfs3, vol, ret, out); + nfs3_check_rw_volaccess (nfs3, olddirfh->exportid, stat, nfs3err); + nfs3_handle_call_state_init (nfs3, cs, req, vol, stat, nfs3err); + + /* While we resolve the source (fh, name) pair, we need to keep a copy + * of the dest (fh,name) pair. + */ + cs->fh = *newdirfh; + cs->pathname = gf_strdup (newname); + if (!cs->pathname) { + stat = NFS3ERR_SERVERFAULT; + ret = -1; + goto nfs3err; + } + + ret = nfs3_fh_resolve_and_resume (cs, olddirfh, oldname, + nfs3_rename_resume_src); + if (ret < 0) + stat = nfs3_errno_to_nfsstat3 (-ret); + +nfs3err: + if (ret < 0) { + nfs3_log_common_res (rpcsvc_request_xid (req), NFS3_RENAME, + stat, -ret); + nfs3_rename_reply (req, stat, NULL, NULL, NULL, NULL, NULL); + nfs3_call_state_wipe (cs); + /* Ret must be 0 after this so that the caller does not + * also send an RPC reply. + */ + ret = 0; + } +out: + return ret; +} + + +int +nfs3svc_rename (rpcsvc_request_t *req) +{ + char newname[NFS_PATH_MAX]; + char oldname[NFS_PATH_MAX]; + struct nfs3_fh olddirfh = {{0}, }; + struct nfs3_fh newdirfh = {{0}, }; + rename3args args; + int ret = RPCSVC_ACTOR_ERROR; + + if (!req) + return ret; + nfs3_prep_rename3args (&args, &olddirfh, oldname, &newdirfh, newname); + if (xdr_to_rename3args (req->msg[0], &args) <= 0) { + gf_log (GF_NFS3, GF_LOG_ERROR, "Error decoding args"); + rpcsvc_request_seterr (req, GARBAGE_ARGS); + goto rpcerr; + } + + ret = nfs3_rename (req, &olddirfh, oldname, &newdirfh, newname); + if ((ret < 0) && (ret != RPCSVC_ACTOR_IGNORE)) { + gf_log (GF_NFS3, GF_LOG_ERROR, "RENAME procedure failed"); + rpcsvc_request_seterr (req, SYSTEM_ERR); + ret = RPCSVC_ACTOR_ERROR; + } + +rpcerr: + return ret; +} + + +int +nfs3_link_reply (rpcsvc_request_t *req, nfsstat3 stat, struct iatt *buf, + struct iatt *preparent, struct iatt *postparent) +{ + link3res res = {0, }; + uint64_t deviceid = 0; + + deviceid = nfs3_request_xlator_deviceid (req); + nfs3_fill_link3res (&res, stat, buf, preparent, postparent, deviceid); + nfs3svc_submit_reply (req, (void *)&res, + (nfs3_serializer)xdr_serialize_link3res); + + return 0; +} + + +int32_t +nfs3svc_link_cbk (call_frame_t *frame, void *cookie, xlator_t *this, + int32_t op_ret, int32_t op_errno, inode_t *inode, + struct iatt *buf, struct iatt *preparent, + struct iatt *postparent, dict_t *xdata) +{ + nfsstat3 stat = NFS3ERR_SERVERFAULT; + nfs3_call_state_t *cs = NULL; + + cs = frame->local; + if (op_ret == -1) { + gf_log (GF_NFS, GF_LOG_WARNING, + "%x: link %s <- %s => -1 (%s)", + rpcsvc_request_xid (cs->req), cs->oploc.path, + cs->resolvedloc.path, strerror (op_errno)); + stat = nfs3_cbk_errno_status (op_ret, op_errno); + } else + stat = NFS3_OK; + + nfs3_log_common_res (rpcsvc_request_xid (cs->req), NFS3_LINK, stat, + op_errno); + nfs3_link_reply (cs->req, stat, buf, preparent, postparent); + nfs3_call_state_wipe (cs); + + return 0; +} + + +int +nfs3_link_resume_lnk (void *carg) +{ + nfsstat3 stat = NFS3ERR_SERVERFAULT; + int ret = -EFAULT; + nfs3_call_state_t *cs = NULL; + nfs_user_t nfu = {0, }; + + if (!carg) + return ret; + + cs = (nfs3_call_state_t *)carg; + nfs3_check_new_fh_resolve_status (cs, stat, nfs3err); + + nfs_request_user_init (&nfu, cs->req); + ret = nfs_link (cs->nfsx, cs->vol, &nfu, &cs->oploc, &cs->resolvedloc, + nfs3svc_link_cbk, cs); + if (ret < 0) + stat = nfs3_errno_to_nfsstat3 (-ret); + +nfs3err: + if (ret < 0) { + nfs3_log_common_res (rpcsvc_request_xid (cs->req), NFS3_LINK, + stat, -ret); + nfs3_link_reply (cs->req, stat, NULL, NULL, NULL); + nfs3_call_state_wipe (cs); + } + return ret; +} + + +int +nfs3_link_resume_tgt (void *carg) +{ + nfsstat3 stat = NFS3ERR_SERVERFAULT; + int ret = -EFAULT; + nfs3_call_state_t *cs = NULL; + + if (!carg) + return ret; + + cs = (nfs3_call_state_t *)carg; + nfs3_check_fh_resolve_status (cs, stat, nfs3err); + nfs_loc_copy (&cs->oploc, &cs->resolvedloc); + nfs_loc_wipe (&cs->resolvedloc); + + ret = nfs3_fh_resolve_and_resume (cs, &cs->fh, cs->pathname, + nfs3_link_resume_lnk); + if (ret < 0) + stat = nfs3_errno_to_nfsstat3 (-ret); + +nfs3err: + if (ret < 0) { + nfs3_log_common_res (rpcsvc_request_xid (cs->req), NFS3_LINK, + stat, -ret); + nfs3_link_reply (cs->req, stat, NULL, NULL, NULL); + nfs3_call_state_wipe (cs); + } + + return ret; +} + + +int +nfs3_link (rpcsvc_request_t *req, struct nfs3_fh *targetfh, + struct nfs3_fh *dirfh, char *newname) +{ + xlator_t *vol = NULL; + nfsstat3 stat = NFS3ERR_SERVERFAULT; + int ret = -EFAULT; + struct nfs3_state *nfs3 = NULL; + nfs3_call_state_t *cs = NULL; + + if ((!req) || (!targetfh) || (!dirfh) || (!newname)) { + gf_log (GF_NFS3, GF_LOG_ERROR, "Bad arguments"); + return -1; + } + + nfs3_validate_gluster_fh (dirfh, stat, nfs3err); + nfs3_validate_gluster_fh (targetfh, stat, nfs3err); + nfs3_validate_nfs3_state (req, nfs3, stat, nfs3err, ret); + nfs3_validate_strlen_or_goto(newname, NFS_NAME_MAX, nfs3err, stat, ret); + nfs3_map_fh_to_volume (nfs3, dirfh, req, vol, stat, nfs3err); + nfs3_volume_started_check (nfs3, vol, ret, out); + nfs3_check_rw_volaccess (nfs3, dirfh->exportid, stat, nfs3err); + nfs3_handle_call_state_init (nfs3, cs, req, vol, stat, nfs3err); + + cs->fh = *dirfh; + cs->pathname = gf_strdup (newname); + if (!cs->pathname) { + stat = NFS3ERR_SERVERFAULT; + ret = -1; + goto nfs3err; + } + + ret = nfs3_fh_resolve_and_resume (cs, targetfh, NULL, + nfs3_link_resume_tgt); + if (ret < 0) + stat = nfs3_errno_to_nfsstat3 (-ret); + +nfs3err: + if (ret < 0) { + nfs3_log_common_res (rpcsvc_request_xid (req), NFS3_LINK, stat, + -ret); + nfs3_link_reply (req, stat, NULL, NULL, NULL); + nfs3_call_state_wipe (cs); + /* Ret must be 0 after this so that the caller does not + * also send an RPC reply. + */ + ret = 0; + } +out: + return ret; +} + +int +nfs3svc_link (rpcsvc_request_t *req) +{ + char newpath[NFS_PATH_MAX]; + struct nfs3_fh dirfh = {{0}, }; + struct nfs3_fh targetfh = {{0}, }; + link3args args; + int ret = RPCSVC_ACTOR_ERROR; + + if (!req) + return ret; + nfs3_prep_link3args (&args, &targetfh, &dirfh, newpath); + if (xdr_to_link3args (req->msg[0], &args) <= 0) { + gf_log (GF_NFS3, GF_LOG_ERROR, "Error decoding args"); + rpcsvc_request_seterr (req, GARBAGE_ARGS); + goto rpcerr; + } + + ret = nfs3_link (req, &targetfh, &dirfh, newpath); + if ((ret < 0) && (ret != RPCSVC_ACTOR_IGNORE)) { + gf_log (GF_NFS3, GF_LOG_ERROR, "LINK procedure failed"); + rpcsvc_request_seterr (req, SYSTEM_ERR); + ret = RPCSVC_ACTOR_ERROR; + } + +rpcerr: + return ret; +} + + +int +nfs3_readdirp_reply (rpcsvc_request_t *req, nfsstat3 stat,struct nfs3_fh *dirfh, + uint64_t cverf, struct iatt *dirstat, gf_dirent_t *entries, + count3 dircount, count3 maxcount, int is_eof) +{ + readdirp3res res = {0, }; + uint64_t deviceid = 0; + + deviceid = nfs3_request_xlator_deviceid (req); + nfs3_fill_readdirp3res (&res, stat, dirfh, cverf, dirstat, entries, + dircount, maxcount, is_eof, deviceid); + nfs3svc_submit_reply (req, (void *)&res, + (nfs3_serializer) xdr_serialize_readdirp3res); + nfs3_free_readdirp3res (&res); + + return 0; +} + + +int +nfs3_readdir_reply (rpcsvc_request_t *req, nfsstat3 stat, struct nfs3_fh *dirfh, + uint64_t cverf, struct iatt *dirstat, gf_dirent_t *entries, + count3 count, int is_eof) +{ + readdir3res res = {0, }; + uint64_t deviceid = 0; + + deviceid = nfs3_request_xlator_deviceid (req); + nfs3_fill_readdir3res (&res, stat, dirfh, cverf, dirstat, entries, count + , is_eof, deviceid); + nfs3svc_submit_reply (req, (void *)&res, + (nfs3_serializer) xdr_serialize_readdir3res); + nfs3_free_readdir3res (&res); + + return 0; +} + + +int32_t +nfs3svc_readdir_fstat_cbk (call_frame_t *frame, void *cookie, xlator_t *this, + int32_t op_ret, int32_t op_errno, struct iatt *buf, + dict_t *xdata) +{ + nfsstat3 stat = NFS3ERR_SERVERFAULT; + int is_eof = 0; + nfs3_call_state_t *cs = NULL; + + cs = frame->local; + if (op_ret == -1) { + gf_log (GF_NFS, GF_LOG_WARNING, + "%x: %s => -1 (%s)", rpcsvc_request_xid (cs->req), + cs->resolvedloc.path, strerror (op_errno)); + stat = nfs3_cbk_errno_status (op_ret, op_errno); + goto nfs3err; + } + + /* Check whether we encountered a end of directory stream while + * readdir'ing. + */ + if (cs->operrno == ENOENT) { + gf_log (GF_NFS3, GF_LOG_TRACE, "Reached end-of-directory"); + is_eof = 1; + } + + stat = NFS3_OK; +nfs3err: + if (cs->maxcount == 0) { + nfs3_log_readdir_res (rpcsvc_request_xid (cs->req), stat, + op_errno, (uintptr_t)cs->fd, + cs->dircount, is_eof); + nfs3_readdir_reply (cs->req, stat, &cs->parent, + (uintptr_t)cs->fd, buf, &cs->entries, + cs->dircount, is_eof); + } else { + nfs3_log_readdirp_res (rpcsvc_request_xid (cs->req), stat, + op_errno, (uintptr_t)cs->fd, + cs->dircount, cs->maxcount, is_eof); + nfs3_readdirp_reply (cs->req, stat, &cs->parent, + (uintptr_t)cs->fd, buf, + &cs->entries, cs->dircount, + cs->maxcount, is_eof); + } + + if (is_eof) { + /* do nothing */ + } + + nfs3_call_state_wipe (cs); + return 0; +} + + +int32_t +nfs3svc_readdir_cbk (call_frame_t *frame, void *cookie, xlator_t *this, + int32_t op_ret, int32_t op_errno, gf_dirent_t *entries, + dict_t *xdata) +{ + nfsstat3 stat = NFS3ERR_SERVERFAULT; + int ret = -EFAULT; + nfs_user_t nfu = {0, }; + nfs3_call_state_t *cs = NULL; + + cs = frame->local; + if (op_ret == -1) { + gf_log (GF_NFS, GF_LOG_WARNING, + "%x: %s => -1 (%s)", rpcsvc_request_xid (cs->req), + cs->resolvedloc.path, strerror (op_errno)); + stat = nfs3_cbk_errno_status (op_ret, op_errno); + goto err; + } + + cs->operrno = op_errno; + list_splice_init (&entries->list, &cs->entries.list); + nfs_request_user_init (&nfu, cs->req); + ret = nfs_fstat (cs->nfsx, cs->vol, &nfu, cs->fd, + nfs3svc_readdir_fstat_cbk, cs); + if (ret < 0) { + op_ret = -1; + stat = nfs3_errno_to_nfsstat3 (-ret); + op_errno = -ret; + } + +err: + if (op_ret >= 0) + goto ret; + + if (cs->maxcount == 0) { + nfs3_log_common_res (rpcsvc_request_xid (cs->req), + NFS3_READDIR, stat, op_errno); + nfs3_readdir_reply (cs->req, stat, NULL, 0, NULL, NULL, 0, 0); + } else { + nfs3_log_common_res (rpcsvc_request_xid (cs->req), + NFS3_READDIRP, stat, op_errno); + nfs3_readdirp_reply (cs->req, stat, NULL, 0, NULL, NULL, + 0, 0, 0); + } + + /* For directories, we force a purge from the fd cache on close + * so that next time the dir is read, we'll get any changed directory + * entries. + */ + nfs3_call_state_wipe (cs); +ret: + return 0; +} + +int +nfs3_readdir_process (nfs3_call_state_t *cs) +{ + int ret = -EFAULT; + nfs_user_t nfu = {0, }; + + if (!cs) + return ret; + + nfs_request_user_init (&nfu, cs->req); + ret = nfs_readdirp (cs->nfsx, cs->vol, &nfu, cs->fd, cs->dircount, + cs->cookie, nfs3svc_readdir_cbk, cs); + return ret; +} + + +int +nfs3_readdir_read_resume (void *carg) +{ + nfsstat3 stat = NFS3ERR_SERVERFAULT; + int ret = -EFAULT; + nfs3_call_state_t *cs = NULL; + struct nfs3_state *nfs3 = NULL; + + if (!carg) + return ret; + + cs = (nfs3_call_state_t *)carg; + nfs3_check_fh_resolve_status (cs, stat, nfs3err); + nfs3 = rpcsvc_request_program_private (cs->req); + ret = nfs3_verify_dircookie (nfs3, cs->fd, cs->cookie, cs->cookieverf, + &stat); + if (ret < 0) /* Stat already set by verifier function above. */ + goto nfs3err; + + ret = nfs3_readdir_process (cs); + if (ret < 0) + stat = nfs3_errno_to_nfsstat3 (-ret); +nfs3err: + if (ret < 0) { + if (cs->maxcount == 0) { + nfs3_log_common_res (rpcsvc_request_xid (cs->req), + NFS3_READDIR, stat, -ret); + nfs3_readdir_reply (cs->req, stat, NULL, 0, NULL, NULL, + 0, 0); + } else { + nfs3_log_common_res (rpcsvc_request_xid (cs->req), + NFS3_READDIRP, stat, -ret); + nfs3_readdirp_reply (cs->req, stat, NULL, 0, NULL, NULL, + 0, 0, 0); + } + nfs3_call_state_wipe (cs); + } + + return 0; +} + + +int32_t +nfs3svc_readdir_opendir_cbk (call_frame_t *frame, void *cookie, xlator_t *this, + int32_t op_ret, int32_t op_errno, fd_t *fd, + dict_t *xdata) +{ + /* + * We don't really need this, it's just an artifact of forcing the + * opendir to happen. + */ + if (fd) { + fd_unref(fd); + } + + return 0; +} + + +int +nfs3_readdir_open_resume (void *carg) +{ + nfsstat3 stat = NFS3ERR_SERVERFAULT; + int ret = -EFAULT; + nfs3_call_state_t *cs = NULL; + nfs_user_t nfu = {0, }; + + if (!carg) + return ret; + + cs = (nfs3_call_state_t *)carg; + nfs3_check_fh_resolve_status (cs, stat, nfs3err); + cs->fd = fd_anonymous (cs->resolvedloc.inode); + if (!cs->fd) { + gf_log (GF_NFS3, GF_LOG_ERROR, "Faile to create anonymous fd"); + goto nfs3err; + } + + /* + * NFS client will usually send us a readdirp without an opendir, + * which would cause us to skip our usual self-heal checks which occur + * in opendir for native protocol. To make sure those checks do happen, + * our most reliable option is to do our own opendir for any readdirp + * at the beginning of the directory. + */ + if (cs->cookie == 0) { + nfs_request_user_init (&nfu, cs->req); + ret = nfs_opendir (cs->nfsx, cs->vol, &nfu, &cs->resolvedloc, + nfs3svc_readdir_opendir_cbk, cs); + if (ret < 0) { + gf_log (GF_NFS3, GF_LOG_ERROR, "auto-opendir failed"); + } + } + + ret = nfs3_readdir_read_resume (cs); + if (ret < 0) + stat = nfs3_errno_to_nfsstat3 (-ret); + +nfs3err: + if (ret < 0) { + if (cs->maxcount == 0) { + nfs3_log_common_res (rpcsvc_request_xid (cs->req), + NFS3_READDIR, stat, -ret); + nfs3_readdir_reply (cs->req, stat, NULL, 0, NULL, NULL, + 0, 0); + } else { + nfs3_log_common_res (rpcsvc_request_xid (cs->req), + NFS3_READDIRP, stat, -ret); + nfs3_readdirp_reply (cs->req, stat, NULL, 0, NULL, NULL, + 0, 0, 0); + } + nfs3_call_state_wipe (cs); + } + + return ret; +} + + + +int +nfs3_readdir (rpcsvc_request_t *req, struct nfs3_fh *fh, cookie3 cookie, + uint64_t cverf, count3 dircount, count3 maxcount) +{ + xlator_t *vol = NULL; + nfsstat3 stat = NFS3ERR_SERVERFAULT; + int ret = -EFAULT; + struct nfs3_state *nfs3 = NULL; + nfs3_call_state_t *cs = NULL; + + if ((!req) || (!fh)) { + gf_log (GF_NFS3, GF_LOG_ERROR, "Bad arguments"); + return -1; + } + + nfs3_log_readdir_call (rpcsvc_request_xid (req), fh, dircount, + maxcount); + nfs3_validate_gluster_fh (fh, stat, nfs3err); + nfs3_validate_nfs3_state (req, nfs3, stat, nfs3err, ret); + nfs3_map_fh_to_volume (nfs3, fh, req, vol, stat, nfs3err); + nfs3_volume_started_check (nfs3, vol, ret, out); + nfs3_handle_call_state_init (nfs3, cs, req, vol, stat, nfs3err); + + cs->cookieverf = cverf; + cs->dircount = dircount; + cs->maxcount = maxcount; + cs->cookie = cookie; + cs->parent = *fh; + ret = nfs3_fh_resolve_and_resume (cs, fh, NULL, + nfs3_readdir_open_resume); + if (ret < 0) + stat = nfs3_errno_to_nfsstat3 (-ret); + +nfs3err: + if (ret < 0) { + if (maxcount == 0) { + nfs3_log_common_res (rpcsvc_request_xid (req), + NFS3_READDIR, stat, -ret); + nfs3_readdir_reply (req, stat, NULL, 0, NULL, NULL, 0, + 0); + } else { + nfs3_log_common_res (rpcsvc_request_xid (req), + NFS3_READDIRP, stat, -ret); + nfs3_readdirp_reply (req, stat, NULL, 0, NULL, NULL, 0, + 0, 0); + } + /* Ret must be NULL after this so that the caller does not + * also send an RPC reply. + */ + ret = 0; + nfs3_call_state_wipe (cs); + } +out: + return ret; +} + + +int +nfs3svc_readdir (rpcsvc_request_t *req) +{ + readdir3args ra; + struct nfs3_fh fh = {{0},}; + int ret = RPCSVC_ACTOR_ERROR; + uint64_t verf = 0; + uint64_t *cval; + + if (!req) + return ret; + nfs3_prep_readdir3args (&ra, &fh); + if (xdr_to_readdir3args (req->msg[0], &ra) <= 0) { + gf_log (GF_NFS3, GF_LOG_ERROR, "Error decoding args"); + rpcsvc_request_seterr (req, GARBAGE_ARGS); + goto rpcerr; + } + cval = (uint64_t *) ra.cookieverf; + verf = *cval; + + ret = nfs3_readdir (req, &fh, ra.cookie, verf, ra.count, 0); + if ((ret < 0) && (ret != RPCSVC_ACTOR_IGNORE)) { + gf_log (GF_NFS3, GF_LOG_ERROR, "READDIR procedure failed"); + rpcsvc_request_seterr (req, SYSTEM_ERR); + ret = RPCSVC_ACTOR_ERROR; + } + +rpcerr: + return ret; +} + + +int +nfs3svc_readdirp (rpcsvc_request_t *req) +{ + readdirp3args ra; + struct nfs3_fh fh = {{0},}; + int ret = RPCSVC_ACTOR_ERROR; + uint64_t cverf = 0; + uint64_t *cval; + + if (!req) + return ret; + nfs3_prep_readdirp3args (&ra, &fh); + if (xdr_to_readdirp3args (req->msg[0], &ra) <= 0) { + gf_log (GF_NFS3, GF_LOG_ERROR, "Error decoding args"); + rpcsvc_request_seterr (req, GARBAGE_ARGS); + goto rpcerr; + } + cval = (uint64_t *) ra.cookieverf; + cverf = *cval; + + ret = nfs3_readdir (req, &fh, ra.cookie, cverf, ra.dircount, + ra.maxcount); + if ((ret < 0) && (ret != RPCSVC_ACTOR_IGNORE)) { + gf_log (GF_NFS3, GF_LOG_ERROR, "READDIRP procedure failed"); + rpcsvc_request_seterr (req, SYSTEM_ERR); + ret = RPCSVC_ACTOR_ERROR; + } + +rpcerr: + return ret; +} + + +int +nfs3_fsstat_reply (rpcsvc_request_t *req, nfsstat3 stat, struct statvfs *fsbuf, + struct iatt *postbuf) +{ + fsstat3res res = {0, }; + uint64_t deviceid = 0; + + deviceid = nfs3_request_xlator_deviceid (req); + nfs3_fill_fsstat3res (&res, stat, fsbuf, postbuf, deviceid); + return nfs3svc_submit_reply (req, &res, + (nfs3_serializer)xdr_serialize_fsstat3res); + +} + + +int32_t +nfs3_fsstat_stat_cbk (call_frame_t *frame, void *cookie, xlator_t *this, + int32_t op_ret, int32_t op_errno, struct iatt *buf, + dict_t *xdata) +{ + nfsstat3 stat = NFS3ERR_SERVERFAULT; + nfs3_call_state_t *cs = NULL; + + cs = frame->local; + if (op_ret == -1) { + gf_log (GF_NFS, GF_LOG_WARNING, + "%x: %s => -1 (%s)", rpcsvc_request_xid (cs->req), + cs->resolvedloc.path, strerror (op_errno)); + stat = nfs3_cbk_errno_status (op_ret, op_errno); + } else + stat = NFS3_OK; + + nfs3_log_common_res (rpcsvc_request_xid (cs->req), NFS3_FSSTAT, stat, + op_errno); + nfs3_fsstat_reply (cs->req, stat, &cs->fsstat, buf); + nfs3_call_state_wipe (cs); + return 0; +} + + +int32_t +nfs3_fsstat_statfs_cbk (call_frame_t *frame, void *cookie, xlator_t *this, + int32_t op_ret, int32_t op_errno, struct statvfs *buf, + dict_t *xdata) +{ + nfs_user_t nfu = {0, }; + int ret = -EFAULT; + nfsstat3 stat = NFS3ERR_SERVERFAULT; + nfs3_call_state_t *cs = NULL; + + cs = frame->local; + if (op_ret == -1) { + gf_log (GF_NFS, GF_LOG_WARNING, + "%x: %s => -1 (%s)", rpcsvc_request_xid (cs->req), + cs->resolvedloc.path, strerror (op_errno)); + ret = -op_errno; + stat = nfs3_cbk_errno_status (op_ret, op_errno); + goto err; + } + + /* Then get the stat for the fs root in order to fill in the + * post_op_attr. + */ + cs->fsstat = *buf; + nfs_request_user_init (&nfu, cs->req); + ret = nfs_stat (cs->nfsx, cs->vol, &nfu, &cs->resolvedloc, + nfs3_fsstat_stat_cbk, cs); + if (ret < 0) + stat = nfs3_errno_to_nfsstat3 (-ret); + +err: + if (ret < 0) { + nfs3_log_common_res (rpcsvc_request_xid (cs->req), NFS3_FSSTAT, + stat, -ret); + nfs3_fsstat_reply (cs->req, stat, NULL, NULL); + nfs3_call_state_wipe (cs); + } + + return 0; +} + + +int +nfs3_fsstat_resume (void *carg) +{ + nfsstat3 stat = NFS3ERR_SERVERFAULT; + int ret = -EFAULT; + nfs3_call_state_t *cs = NULL; + nfs_user_t nfu = {0, }; + + if (!carg) + return ret; + + cs = (nfs3_call_state_t *)carg; + nfs3_check_fh_resolve_status (cs, stat, nfs3err); + nfs_request_user_init (&nfu, cs->req); + /* First, we need to get the statvfs for the subvol */ + ret = nfs_statfs (cs->nfsx, cs->vol, &nfu, &cs->resolvedloc, + nfs3_fsstat_statfs_cbk, cs); + if (ret < 0) + stat = nfs3_errno_to_nfsstat3 (-ret); + +nfs3err: + if (ret < 0) { + nfs3_log_common_res (rpcsvc_request_xid (cs->req), NFS3_FSSTAT, + stat, -ret); + nfs3_fsstat_reply (cs->req, stat, NULL, NULL); + nfs3_call_state_wipe (cs); + } + + return ret; +} + + + +int +nfs3_fsstat (rpcsvc_request_t *req, struct nfs3_fh *fh) +{ + xlator_t *vol = NULL; + nfsstat3 stat = NFS3ERR_SERVERFAULT; + int ret = -EFAULT; + struct nfs3_state *nfs3 = NULL; + nfs3_call_state_t *cs = NULL; + + if ((!req) || (!fh)) { + gf_log (GF_NFS3, GF_LOG_ERROR, "Bad arguments"); + return -1; + } + + nfs3_log_common_call (rpcsvc_request_xid (req), "FSSTAT", fh); + nfs3_validate_gluster_fh (fh, stat, nfs3err); + nfs3_validate_nfs3_state (req, nfs3, stat, nfs3err, ret); + nfs3_map_fh_to_volume (nfs3, fh, req, vol, stat, nfs3err); + nfs3_volume_started_check (nfs3, vol, ret, out); + nfs3_handle_call_state_init (nfs3, cs, req, vol, stat, nfs3err); + + ret = nfs3_fh_resolve_and_resume (cs, fh, NULL, nfs3_fsstat_resume); + if (ret < 0) + stat = nfs3_errno_to_nfsstat3 (-ret); + +nfs3err: + if (ret < 0) { + nfs3_log_common_res (rpcsvc_request_xid (req), NFS3_FSSTAT, + stat, -ret); + nfs3_fsstat_reply (req, stat, NULL, NULL); + nfs3_call_state_wipe (cs); + /* Ret must be 0 after this so that the caller does not + * also send an RPC reply. + */ + ret = 0; + } +out: + return ret; +} + + +int +nfs3svc_fsstat (rpcsvc_request_t *req) +{ + struct nfs3_fh fh = {{0}, }; + fsstat3args args; + int ret = RPCSVC_ACTOR_ERROR; + + if (!req) + return ret; + nfs3_prep_fsstat3args (&args, &fh); + if (xdr_to_fsstat3args (req->msg[0], &args) <= 0) { + gf_log (GF_NFS3, GF_LOG_ERROR, "Error decoding args"); + rpcsvc_request_seterr (req, GARBAGE_ARGS); + goto rpcerr; + } + + ret = nfs3_fsstat (req, &fh); + if ((ret < 0) && (ret != RPCSVC_ACTOR_IGNORE)) { + gf_log (GF_NFS3, GF_LOG_ERROR, "FSTAT procedure failed"); + rpcsvc_request_seterr (req, SYSTEM_ERR); + ret = RPCSVC_ACTOR_ERROR; + } + +rpcerr: + return ret; +} + + +int +nfs3_fsinfo_reply (rpcsvc_request_t *req, nfsstat3 status, struct iatt *fsroot) +{ + fsinfo3res res; + struct nfs3_state *nfs3 = NULL; + uint64_t deviceid = 0; + + deviceid = nfs3_request_xlator_deviceid (req); + nfs3 = rpcsvc_request_program_private (req); + nfs3_fill_fsinfo3res (nfs3, &res, status, fsroot, deviceid); + + nfs3svc_submit_reply (req, &res, + (nfs3_serializer)xdr_serialize_fsinfo3res); + return 0; +} + + +int32_t +nfs3svc_fsinfo_cbk (call_frame_t *frame, void *cookie, xlator_t *this, + int32_t op_ret, int32_t op_errno, struct iatt *buf, + dict_t *xdata) +{ + nfsstat3 status = NFS3ERR_SERVERFAULT; + nfs3_call_state_t *cs = NULL; + + cs = frame->local; + + if (op_ret == -1) { + gf_log (GF_NFS, GF_LOG_WARNING, + "%x: %s => -1 (%s)", rpcsvc_request_xid (cs->req), + cs->resolvedloc.path, strerror (op_errno)); + status = nfs3_cbk_errno_status (op_ret, op_errno); + }else + status = NFS3_OK; + + nfs3_log_common_res (rpcsvc_request_xid (cs->req), NFS3_FSINFO, status, + op_errno); + + nfs3_fsinfo_reply (cs->req, status, buf); + nfs3_call_state_wipe (cs); + + return 0; +} + + +int +nfs3_fsinfo_resume (void *carg) +{ + nfsstat3 stat = NFS3ERR_SERVERFAULT; + int ret = -EFAULT; + nfs_user_t nfu = {0, }; + nfs3_call_state_t *cs = NULL; + + + if (!carg) + return ret; + + cs = (nfs3_call_state_t *)carg; + nfs3_check_fh_resolve_status (cs, stat, nfs3err); + nfs_request_user_init (&nfu, cs->req); + + ret = nfs_stat (cs->nfsx, cs->vol, &nfu, &cs->resolvedloc, + nfs3svc_fsinfo_cbk, cs); + if (ret < 0) + stat = nfs3_errno_to_nfsstat3 (-ret); + +nfs3err: + if (ret < 0) { + nfs3_log_common_res (rpcsvc_request_xid (cs->req), NFS3_FSINFO, + stat, -ret); + nfs3_fsinfo_reply (cs->req, stat, NULL); + nfs3_call_state_wipe (cs); + } + + return ret; +} + + +int +nfs3_fsinfo (rpcsvc_request_t *req, struct nfs3_fh *fh) +{ + xlator_t *vol = NULL; + nfsstat3 stat = NFS3ERR_SERVERFAULT; + int ret = -EFAULT; + struct nfs3_state *nfs3 = NULL; + nfs3_call_state_t *cs = NULL; + + if ((!req) || (!fh)) { + gf_log (GF_NFS3, GF_LOG_ERROR, "Bad arguments"); + return -1; + } + + nfs3_log_common_call (rpcsvc_request_xid (req), "FSINFO", fh); + nfs3_validate_gluster_fh (fh, stat, nfs3err); + nfs3_validate_nfs3_state (req, nfs3, stat, nfs3err, ret); + nfs3_map_fh_to_volume (nfs3, fh, req, vol, stat, nfs3err); + nfs3_volume_started_check (nfs3, vol, ret, out); + nfs3_handle_call_state_init (nfs3, cs, req, vol, stat, nfs3err); + + ret = nfs3_fh_resolve_and_resume (cs, fh, NULL, nfs3_fsinfo_resume); + if (ret < 0) + stat = nfs3_errno_to_nfsstat3 (-ret); + +nfs3err: + if (ret < 0) { + nfs3_log_common_res (rpcsvc_request_xid (req), NFS3_FSINFO, + stat, -ret); + nfs3_fsinfo_reply (req, stat, NULL); + nfs3_call_state_wipe (cs); + ret = 0; + } +out: + return ret; +} + + +int +nfs3svc_fsinfo (rpcsvc_request_t *req) +{ + int ret = RPCSVC_ACTOR_ERROR; + fsinfo3args args; + struct nfs3_fh root = {{0}, }; + + if (!req) + return ret; + + nfs3_prep_fsinfo3args (&args, &root); + if (xdr_to_fsinfo3args (req->msg[0], &args) <= 0) { + gf_log (GF_NFS3, GF_LOG_ERROR, "Error decoding arguments"); + rpcsvc_request_seterr (req, GARBAGE_ARGS); + goto rpcerr; + } + + ret = nfs3_fsinfo (req, &root); + if ((ret < 0) && (ret != RPCSVC_ACTOR_IGNORE)) { + gf_log (GF_NFS3, GF_LOG_ERROR, "FSINFO procedure failed"); + rpcsvc_request_seterr (req, SYSTEM_ERR); + ret = RPCSVC_ACTOR_ERROR; + } + +rpcerr: + return ret; +} + + +int +nfs3_pathconf_reply (rpcsvc_request_t *req, nfsstat3 stat, struct iatt *buf) +{ + pathconf3res res = {0, }; + uint64_t deviceid = 0; + + deviceid = nfs3_request_xlator_deviceid (req); + nfs3_fill_pathconf3res (&res, stat, buf, deviceid); + nfs3svc_submit_reply (req, (void *)&res, + (nfs3_serializer)xdr_serialize_pathconf3res); + return 0; +} + + +int32_t +nfs3svc_pathconf_cbk (call_frame_t *frame, void *cookie, xlator_t *this, + int32_t op_ret, int32_t op_errno, struct iatt *buf, + dict_t *xdata) +{ + struct iatt *sbuf = NULL; + nfs3_call_state_t *cs = NULL; + nfsstat3 stat = NFS3ERR_SERVERFAULT; + + cs = frame->local; + if (op_ret == -1) { + gf_log (GF_NFS, GF_LOG_WARNING, + "%x: %s => -1 (%s)", rpcsvc_request_xid (cs->req), + cs->resolvedloc.path, strerror (op_errno)); + stat = nfs3_cbk_errno_status (op_ret, op_errno); + } else { + /* If stat fop failed, we can still send the other components + * in a pathconf reply. + */ + sbuf = buf; + stat = NFS3_OK; + } + + nfs3_log_common_res (rpcsvc_request_xid (cs->req), NFS3_PATHCONF, stat, + op_errno); + nfs3_pathconf_reply (cs->req, stat, sbuf); + nfs3_call_state_wipe (cs); + + return 0; +} + + +int +nfs3_pathconf_resume (void *carg) +{ + nfsstat3 stat = NFS3ERR_SERVERFAULT; + int ret = -EFAULT; + nfs_user_t nfu = {0, }; + nfs3_call_state_t *cs = NULL; + + if (!carg) + return ret; + + cs = (nfs3_call_state_t *)carg; + nfs3_check_fh_resolve_status (cs, stat, nfs3err); + nfs_request_user_init (&nfu, cs->req); + ret = nfs_stat (cs->nfsx, cs->vol, &nfu, &cs->resolvedloc, + nfs3svc_pathconf_cbk, cs); + if (ret < 0) + stat = nfs3_errno_to_nfsstat3 (-ret); +nfs3err: + if (ret < 0) { + nfs3_log_common_res (rpcsvc_request_xid (cs->req), + NFS3_PATHCONF, stat, -ret); + nfs3_pathconf_reply (cs->req, stat, NULL); + nfs3_call_state_wipe (cs); + } + + return ret; +} + +int +nfs3_pathconf (rpcsvc_request_t *req, struct nfs3_fh *fh) +{ + xlator_t *vol = NULL; + nfsstat3 stat = NFS3ERR_SERVERFAULT; + int ret = -EFAULT; + struct nfs3_state *nfs3 = NULL; + nfs3_call_state_t *cs = NULL; + + if ((!req) || (!fh)) { + gf_log (GF_NFS3, GF_LOG_ERROR, "Bad arguments"); + return -1; + } + + nfs3_log_common_call (rpcsvc_request_xid (req), "PATHCONF", fh); + nfs3_validate_gluster_fh (fh, stat, nfs3err); + nfs3_validate_nfs3_state (req, nfs3, stat, nfs3err, ret); + nfs3_map_fh_to_volume (nfs3, fh, req, vol, stat, nfs3err); + nfs3_volume_started_check (nfs3, vol, ret, out); + nfs3_handle_call_state_init (nfs3, cs, req, vol, stat, nfs3err); + + ret = nfs3_fh_resolve_and_resume (cs, fh, NULL, nfs3_pathconf_resume); + if (ret < 0) + stat = nfs3_errno_to_nfsstat3 (-ret); + +nfs3err: + if (ret < 0) { + nfs3_log_common_res (rpcsvc_request_xid (req), NFS3_PATHCONF, + stat, -ret); + nfs3_pathconf_reply (req, stat, NULL); + nfs3_call_state_wipe (cs); + /* Ret must be 0 after this so that the caller does not + * also send an RPC reply. + */ + ret = 0; + } +out: + return ret; +} + + +int +nfs3svc_pathconf (rpcsvc_request_t *req) +{ + struct nfs3_fh fh = {{0}, }; + pathconf3args args; + int ret = RPCSVC_ACTOR_ERROR; + + if (!req) + return ret; + nfs3_prep_pathconf3args (&args, &fh); + if (xdr_to_pathconf3args (req->msg[0], &args) <= 0) { + gf_log (GF_NFS3, GF_LOG_ERROR, "Error decoding args"); + rpcsvc_request_seterr (req, GARBAGE_ARGS); + goto rpcerr; + } + + ret = nfs3_pathconf (req, &fh); + if ((ret < 0) && (ret != RPCSVC_ACTOR_IGNORE)) { + gf_log (GF_NFS3, GF_LOG_ERROR, "PATHCONF procedure failed"); + rpcsvc_request_seterr (req, SYSTEM_ERR); + ret = RPCSVC_ACTOR_ERROR; + } + +rpcerr: + return ret; +} + +int +nfs3_commit_reply (rpcsvc_request_t *req, nfsstat3 stat, uint64_t wverf, + struct iatt *prestat, struct iatt *poststat) +{ + commit3res res = {0, }; + uint64_t deviceid = 0; + + deviceid = nfs3_request_xlator_deviceid (req); + nfs3_fill_commit3res (&res, stat, wverf, prestat, poststat, deviceid); + nfs3svc_submit_reply (req, (void *)&res, + (nfs3_serializer)xdr_serialize_commit3res); + + return 0; +} + + +int32_t +nfs3svc_commit_cbk (call_frame_t *frame, void *cookie, xlator_t *this, + int32_t op_ret, int32_t op_errno, dict_t *xdata) +{ + nfsstat3 stat = NFS3ERR_SERVERFAULT; + nfs3_call_state_t *cs = NULL; + struct nfs3_state *nfs3 = NULL; + + cs = frame->local; + if (op_ret == -1) { + gf_log (GF_NFS, GF_LOG_WARNING, + "%x: %s => -1 (%s)", rpcsvc_request_xid (cs->req), + cs->resolvedloc.path, strerror (op_errno)); + stat = nfs3_cbk_errno_status (op_ret, op_errno); + } else + stat = NFS3_OK; + + nfs3 = rpcsvc_request_program_private (cs->req); + nfs3_log_commit_res (rpcsvc_request_xid (cs->req), stat, op_errno, + nfs3->serverstart); + nfs3_commit_reply (cs->req, stat, nfs3->serverstart, NULL, NULL); + nfs3_call_state_wipe (cs); + + return 0; +} + +int +nfs3_commit_resume (void *carg) +{ + nfsstat3 stat = NFS3ERR_SERVERFAULT; + int ret = -EFAULT; + nfs_user_t nfu = {0, }; + nfs3_call_state_t *cs = NULL; + + if (!carg) + return ret; + + cs = (nfs3_call_state_t *)carg; + nfs3_check_fh_resolve_status (cs, stat, nfs3err); + + if (nfs3_export_sync_trusted (cs->nfs3state, cs->resolvefh.exportid)) { + ret = -1; + stat = NFS3_OK; + goto nfs3err; + } + + nfs_request_user_init (&nfu, cs->req); + ret = nfs_flush (cs->nfsx, cs->vol, &nfu, cs->fd, + nfs3svc_commit_cbk, cs); + if (ret < 0) + stat = nfs3_errno_to_nfsstat3 (-ret); + +nfs3err: + if (ret < 0) { + nfs3_log_common_res (rpcsvc_request_xid (cs->req), NFS3_COMMIT, + stat, -ret); + nfs3_commit_reply (cs->req, stat, cs->nfs3state->serverstart, + NULL, NULL); + nfs3_call_state_wipe (cs); + ret = 0; + } + + return 0; +} + + +int +nfs3_commit_open_resume (void *carg) +{ + nfsstat3 stat = NFS3ERR_SERVERFAULT; + int ret = -EFAULT; + nfs3_call_state_t *cs = NULL; + + if (!carg) + return ret; + + cs = (nfs3_call_state_t *)carg; + nfs3_check_fh_resolve_status (cs, stat, nfs3err); + cs->fd = fd_anonymous (cs->resolvedloc.inode); + if (!cs->fd) { + gf_log (GF_NFS3, GF_LOG_ERROR, "Failed to create anonymous fd."); + goto nfs3err; + } + + ret = nfs3_commit_resume (cs); + if (ret < 0) + stat = nfs3_errno_to_nfsstat3 (-ret); +nfs3err: + if (ret < 0) { + nfs3_log_common_res (rpcsvc_request_xid (cs->req), NFS3_COMMIT, + stat, -ret); + nfs3_commit_reply (cs->req, stat, 0, NULL, NULL); + nfs3_call_state_wipe (cs); + } + + return ret; +} + + + +int +nfs3_commit (rpcsvc_request_t *req, struct nfs3_fh *fh, offset3 offset, + count3 count) +{ + xlator_t *vol = NULL; + nfsstat3 stat = NFS3ERR_SERVERFAULT; + int ret = -EFAULT; + struct nfs3_state *nfs3 = NULL; + nfs3_call_state_t *cs = NULL; + + if ((!req) || (!fh)) { + gf_log (GF_NFS3, GF_LOG_ERROR, "Bad arguments"); + return -1; + } + + nfs3_log_rw_call (rpcsvc_request_xid (req), "COMMIT", fh, offset, + count, -1); + nfs3_validate_gluster_fh (fh, stat, nfs3err); + nfs3_validate_nfs3_state (req, nfs3, stat, nfs3err, ret); + nfs3_map_fh_to_volume (nfs3, fh, req, vol, stat, nfs3err); + nfs3_volume_started_check (nfs3, vol, ret, out); + nfs3_check_rw_volaccess (nfs3, fh->exportid, stat, nfs3err); + nfs3_handle_call_state_init (nfs3, cs, req, vol, stat, nfs3err); + + cs->datacount = count; + cs->dataoffset = offset; + ret = nfs3_fh_resolve_and_resume (cs, fh, NULL, + nfs3_commit_open_resume); + if (ret < 0) + stat = nfs3_errno_to_nfsstat3 (-ret); + +nfs3err: + if (ret < 0) { + nfs3_log_common_res (rpcsvc_request_xid (req), NFS3_COMMIT, + stat, -ret); + nfs3_commit_reply (req, stat, 0, NULL, NULL); + nfs3_call_state_wipe (cs); + ret = 0; + } +out: + return ret; +} + + + +int +nfs3svc_commit (rpcsvc_request_t *req) +{ + struct nfs3_fh fh = {{0}, }; + commit3args args; + int ret = RPCSVC_ACTOR_ERROR; + + if (!req) + return ret; + nfs3_prep_commit3args (&args, &fh); + if (xdr_to_commit3args (req->msg[0], &args) <= 0) { + gf_log (GF_NFS3, GF_LOG_ERROR, "Error decoding args"); + rpcsvc_request_seterr (req, GARBAGE_ARGS); + goto rpcerr; + } + + ret = nfs3_commit (req, &fh, args.offset, args.count); + if ((ret < 0) && (ret != RPCSVC_ACTOR_IGNORE)) { + gf_log (GF_NFS3, GF_LOG_ERROR, "COMMIT procedure failed"); + rpcsvc_request_seterr (req, SYSTEM_ERR); + ret = RPCSVC_ACTOR_ERROR; + } + +rpcerr: + return ret; +} + + +rpcsvc_actor_t nfs3svc_actors[NFS3_PROC_COUNT] = { + {"NULL", NFS3_NULL, nfs3svc_null, NULL, 0, DRC_IDEMPOTENT}, + {"GETATTR", NFS3_GETATTR, nfs3svc_getattr, NULL, 0, DRC_IDEMPOTENT}, + {"SETATTR", NFS3_SETATTR, nfs3svc_setattr, NULL, 0, DRC_NON_IDEMPOTENT}, + {"LOOKUP", NFS3_LOOKUP, nfs3svc_lookup, NULL, 0, DRC_IDEMPOTENT}, + {"ACCESS", NFS3_ACCESS, nfs3svc_access, NULL, 0, DRC_IDEMPOTENT}, + {"READLINK", NFS3_READLINK, nfs3svc_readlink, NULL, 0, DRC_IDEMPOTENT}, + {"READ", NFS3_READ, nfs3svc_read, NULL, 0, DRC_IDEMPOTENT}, + {"WRITE", NFS3_WRITE, nfs3svc_write, nfs3svc_write_vecsizer, 0, DRC_NON_IDEMPOTENT}, + {"CREATE", NFS3_CREATE, nfs3svc_create, NULL, 0, DRC_NON_IDEMPOTENT}, + {"MKDIR", NFS3_MKDIR, nfs3svc_mkdir, NULL, 0, DRC_NON_IDEMPOTENT}, + {"SYMLINK", NFS3_SYMLINK, nfs3svc_symlink, NULL, 0, DRC_NON_IDEMPOTENT}, + {"MKNOD", NFS3_MKNOD, nfs3svc_mknod, NULL, 0, DRC_NON_IDEMPOTENT}, + {"REMOVE", NFS3_REMOVE, nfs3svc_remove, NULL, 0, DRC_NON_IDEMPOTENT}, + {"RMDIR", NFS3_RMDIR, nfs3svc_rmdir, NULL, 0, DRC_NON_IDEMPOTENT}, + {"RENAME", NFS3_RENAME, nfs3svc_rename, NULL, 0, DRC_NON_IDEMPOTENT}, + {"LINK", NFS3_LINK, nfs3svc_link, NULL, 0, DRC_NON_IDEMPOTENT}, + {"READDIR", NFS3_READDIR, nfs3svc_readdir, NULL, 0, DRC_IDEMPOTENT}, + {"READDIRPLUS", NFS3_READDIRP, nfs3svc_readdirp, NULL, 0, DRC_IDEMPOTENT}, + {"FSSTAT", NFS3_FSSTAT, nfs3svc_fsstat, NULL, 0, DRC_IDEMPOTENT}, + {"FSINFO", NFS3_FSINFO, nfs3svc_fsinfo, NULL, 0, DRC_IDEMPOTENT}, + {"PATHCONF", NFS3_PATHCONF, nfs3svc_pathconf, NULL, 0, DRC_IDEMPOTENT}, + {"COMMIT", NFS3_COMMIT, nfs3svc_commit, NULL, 0, DRC_IDEMPOTENT} +}; + + +rpcsvc_program_t nfs3prog = { + .progname = "NFS3", + .prognum = NFS_PROGRAM, + .progver = NFS_V3, + .progport = GF_NFS3_PORT, + .actors = nfs3svc_actors, + .numactors = NFS3_PROC_COUNT, + + /* Requests like FSINFO are sent before an auth scheme + * is inited by client. See RFC 2623, Section 2.3.2. */ + .min_auth = AUTH_NULL, +}; + +/* + * This function rounds up the input value to multiple of 4096. Min and Max + * supported I/O size limits are 4KB (GF_NFS3_FILE_IO_SIZE_MIN) and + * 1MB (GF_NFS3_FILE_IO_SIZE_MAX). + */ +void +nfs3_iosize_roundup_4KB (uint64_t *ioszptr) +{ + uint64_t iosize; + uint64_t iopages; + + if (!ioszptr) + return; + + iosize = *ioszptr; + iopages = (iosize + GF_NFS3_IO_SIZE -1) >> GF_NFS3_IO_SHIFT; + iosize = (iopages * GF_NFS3_IO_SIZE); + + /* Double check - boundary conditions */ + if (iosize < GF_NFS3_FILE_IO_SIZE_MIN) { + iosize = GF_NFS3_FILE_IO_SIZE_MIN; + } else if (iosize > GF_NFS3_FILE_IO_SIZE_MAX) { + iosize = GF_NFS3_FILE_IO_SIZE_MAX; + } + + *ioszptr = iosize; +} + +int +nfs3_init_options (struct nfs3_state *nfs3, dict_t *options) +{ + int ret = -1; + char *optstr = NULL; + uint64_t size64 = 0; + + if ((!nfs3) || (!options)) + return -1; + + /* nfs3.read-size */ + nfs3->readsize = GF_NFS3_RTPREF; + if (dict_get (options, "nfs3.read-size")) { + ret = dict_get_str (options, "nfs3.read-size", &optstr); + if (ret < 0) { + gf_log (GF_NFS3, GF_LOG_ERROR, "Failed to read " + " option: nfs3.read-size"); + ret = -1; + goto err; + } + + ret = gf_string2bytesize (optstr, &size64); + if (ret == -1) { + gf_log (GF_NFS3, GF_LOG_ERROR, "Failed to format" + " option: nfs3.read-size"); + ret = -1; + goto err; + } + + nfs3_iosize_roundup_4KB (&size64); + nfs3->readsize = size64; + } + + /* nfs3.write-size */ + nfs3->writesize = GF_NFS3_WTPREF; + if (dict_get (options, "nfs3.write-size")) { + ret = dict_get_str (options, "nfs3.write-size", &optstr); + if (ret < 0) { + gf_log (GF_NFS3, GF_LOG_ERROR, "Failed to read " + " option: nfs3.write-size"); + ret = -1; + goto err; + } + + ret = gf_string2bytesize (optstr, &size64); + if (ret == -1) { + gf_log (GF_NFS3, GF_LOG_ERROR, "Failed to format" + " option: nfs3.write-size"); + ret = -1; + goto err; + } + + nfs3_iosize_roundup_4KB (&size64); + nfs3->writesize = size64; + } + + /* nfs3.readdir.size */ + nfs3->readdirsize = GF_NFS3_DTPREF; + if (dict_get (options, "nfs3.readdir-size")) { + ret = dict_get_str (options,"nfs3.readdir-size", &optstr); + if (ret < 0) { + gf_log (GF_NFS3, GF_LOG_ERROR, "Failed to read" + " option: nfs3.readdir-size"); + ret = -1; + goto err; + } + + ret = gf_string2bytesize (optstr, &size64); + if (ret == -1) { + gf_log (GF_NFS3, GF_LOG_ERROR, "Failed to format" + " option: nfs3.readdir-size"); + ret = -1; + goto err; + } + + nfs3_iosize_roundup_4KB (&size64); + nfs3->readdirsize = size64; + } + + /* We want to use the size of the biggest param for the io buffer size. + */ + nfs3->iobsize = nfs3->readsize; + if (nfs3->iobsize < nfs3->writesize) + nfs3->iobsize = nfs3->writesize; + if (nfs3->iobsize < nfs3->readdirsize) + nfs3->iobsize = nfs3->readdirsize; + + /* But this is the true size of each iobuf. We need this size to + * accommodate the NFS headers also in the same buffer. */ + nfs3->iobsize = nfs3->iobsize * 2; + + /* mem-factor */ + nfs3->memfactor = GF_NFS3_DEFAULT_MEMFACTOR; + ret = 0; +err: + return ret; +} + +int +nfs3_init_subvolume_options (xlator_t *nfsx, + struct nfs3_export *exp, + dict_t *options) +{ + int ret = -1; + char *optstr = NULL; + char searchkey[1024]; + char *name = NULL; + gf_boolean_t boolt = _gf_false; + uuid_t volumeid = {0, }; + + if ((!nfsx) || (!exp)) + return -1; + + /* For init, fetch options from xlator but for + * reconfigure, take the parameter */ + if (!options) + options = nfsx->options; + + if (!options) + return (-1); + + uuid_clear (volumeid); + if (gf_nfs_dvm_off (nfs_state (nfsx))) + goto no_dvm; + + ret = snprintf (searchkey, 1024, "nfs3.%s.volume-id",exp->subvol->name); + if (ret < 0) { + gf_log (GF_MNT, GF_LOG_ERROR, "snprintf failed"); + ret = -1; + goto err; + } + + if (dict_get (options, searchkey)) { + ret = dict_get_str (options, searchkey, &optstr); + if (ret < 0) { + gf_log (GF_MNT, GF_LOG_ERROR, "Failed to read option" + ": %s", searchkey); + ret = -1; + goto err; + } + } else { + gf_log (GF_MNT, GF_LOG_ERROR, "DVM is on but volume-id not " + "given for volume: %s", exp->subvol->name); + ret = -1; + goto err; + } + + if (optstr) { + ret = uuid_parse (optstr, volumeid); + if (ret < 0) { + gf_log (GF_MNT, GF_LOG_ERROR, "Failed to parse volume " + "UUID"); + ret = -1; + goto err; + } + uuid_copy (exp->volumeid, volumeid); + } + +no_dvm: + /* Volume Access */ + name = exp->subvol->name; + ret = snprintf (searchkey, 1024, "nfs3.%s.volume-access", name); + if (ret < 0) { + gf_log (GF_NFS3, GF_LOG_ERROR, "snprintf failed"); + ret = -1; + goto err; + } + + exp->access = GF_NFS3_DEFAULT_VOLACCESS; + if (dict_get (options, searchkey)) { + ret = dict_get_str (options, searchkey, &optstr); + if (ret < 0) { + gf_log (GF_NFS3, GF_LOG_ERROR, "Failed to read " + " option: %s", searchkey); + ret = -1; + goto err; + } + + if (strcmp (optstr, "read-only") == 0) + exp->access = GF_NFS3_VOLACCESS_RO; + } + + ret = snprintf (searchkey, 1024, "rpc-auth.%s.unix", name); + if (ret < 0) { + gf_log (GF_NFS3, GF_LOG_ERROR, "snprintf failed"); + ret = -1; + goto err; + } + + if (dict_get (options, searchkey)) { + ret = dict_get_str (options, searchkey, &optstr); + if (ret < 0) { + gf_log (GF_NFS3, GF_LOG_ERROR, "Failed to read " + " option: %s", searchkey); + ret = -1; + goto err; + } + } + + exp->trusted_sync = 0; + ret = snprintf (searchkey, 1024, "nfs3.%s.trusted-sync", name); + if (ret < 0) { + gf_log (GF_NFS3, GF_LOG_ERROR, "snprintf failed"); + ret = -1; + goto err; + } + + if (dict_get (options, searchkey)) { + ret = dict_get_str (options, searchkey, &optstr); + if (ret < 0) { + gf_log (GF_NFS3, GF_LOG_ERROR, "Failed to read " + " option: %s", searchkey); + ret = -1; + goto err; + } + + ret = gf_string2boolean (optstr, &boolt); + if (ret < 0) { + gf_log (GF_NFS3, GF_LOG_ERROR, "Failed to convert str " + "to gf_boolean_t"); + ret = -1; + goto err; + } + + if (boolt == _gf_true) + exp->trusted_sync = 1; + } + + exp->trusted_write = 0; + ret = snprintf (searchkey, 1024, "nfs3.%s.trusted-write", name); + if (ret < 0) { + gf_log (GF_NFS3, GF_LOG_ERROR, "snprintf failed"); + ret = -1; + goto err; + } + + if (dict_get (options, searchkey)) { + ret = dict_get_str (options, searchkey, &optstr); + if (ret < 0) { + gf_log (GF_NFS3, GF_LOG_ERROR, "Failed to read " + " option: %s", searchkey); + ret = -1; + goto err; + } + + ret = gf_string2boolean (optstr, &boolt); + if (ret < 0) { + gf_log (GF_NFS3, GF_LOG_ERROR, "Failed to convert str " + "to gf_boolean_t"); + ret = -1; + goto err; + } + + if (boolt == _gf_true) + exp->trusted_write = 1; + } + + /* If trusted-sync is on, then we also switch on trusted-write because + * tw is included in ts. In write logic, we're then only checking for + * tw. + */ + if (exp->trusted_sync) + exp->trusted_write = 1; + + gf_log (GF_NFS3, GF_LOG_TRACE, "%s: %s, %s, %s", exp->subvol->name, + (exp->access == GF_NFS3_VOLACCESS_RO)?"read-only":"read-write", + (exp->trusted_sync == 0)?"no trusted_sync":"trusted_sync", + (exp->trusted_write == 0)?"no trusted_write":"trusted_write"); + ret = 0; +err: + return ret; +} + + +struct nfs3_export * +nfs3_init_subvolume (struct nfs3_state *nfs3, xlator_t *subvol) +{ + int ret = -1; + struct nfs3_export *exp = NULL; + + if ((!nfs3) || (!subvol)) + return NULL; + + exp = GF_CALLOC (1, sizeof (*exp), gf_nfs_mt_nfs3_export); + exp->subvol = subvol; + INIT_LIST_HEAD (&exp->explist); + gf_log (GF_NFS3, GF_LOG_TRACE, "Initing state: %s", exp->subvol->name); + + ret = nfs3_init_subvolume_options (nfs3->nfsx, exp, NULL); + if (ret == -1) { + gf_log (GF_NFS3, GF_LOG_ERROR, "Failed to init subvol"); + goto exp_free; + } + + ret = 0; +exp_free: + if (ret < 0) { + GF_FREE (exp); + exp = NULL; + } + + return exp; +} + + +int +nfs3_init_subvolumes (struct nfs3_state *nfs3) +{ + int ret = -1; + struct xlator_list *xl_list = NULL; + struct nfs3_export *exp = NULL; + + if (!nfs3) + return -1; + + xl_list = nfs3->nfsx->children; + + while (xl_list) { + exp = nfs3_init_subvolume (nfs3, xl_list->xlator); + if (!exp) { + gf_log (GF_NFS3, GF_LOG_ERROR, "Failed to init subvol: " + "%s", xl_list->xlator->name); + goto err; + } + list_add_tail (&exp->explist, &nfs3->exports); + xl_list = xl_list->next; + } + + ret = 0; +err: + return ret; +} + + +struct nfs3_state * +nfs3_init_state (xlator_t *nfsx) +{ + struct nfs3_state *nfs3 = NULL; + int ret = -1; + unsigned int localpool = 0; + struct nfs_state *nfs = NULL; + + if (!nfsx) + return NULL; + + nfs3 = (struct nfs3_state *)GF_CALLOC (1, sizeof (*nfs3), + gf_nfs_mt_nfs3_state); + if (!nfs3) { + gf_log (GF_NFS3, GF_LOG_ERROR, "Memory allocation failed"); + return NULL; + } + + nfs = nfsx->private; + ret = nfs3_init_options (nfs3, nfsx->options); + if (ret == -1) { + gf_log (GF_NFS3, GF_LOG_ERROR, "Failed to init options"); + goto ret; + } + + nfs3->iobpool = nfsx->ctx->iobuf_pool; + + localpool = nfs3->memfactor * GF_NFS_CONCURRENT_OPS_MULT; + gf_log (GF_NFS3, GF_LOG_TRACE, "local pool: %d", localpool); + nfs3->localpool = mem_pool_new (nfs3_call_state_t, localpool); + if (!nfs3->localpool) { + gf_log (GF_NFS3, GF_LOG_ERROR, "local mempool creation failed"); + ret = -1; + goto ret; + } + + nfs3->nfsx = nfsx; + nfs3->exportslist = nfsx->children; + INIT_LIST_HEAD (&nfs3->exports); + ret = nfs3_init_subvolumes (nfs3); + if (ret == -1) { + gf_log (GF_NFS3, GF_LOG_ERROR, "Failed to init per-subvolume " + "state"); + goto free_localpool; + } + + nfs3->serverstart = (uint64_t)time (NULL); + INIT_LIST_HEAD (&nfs3->fdlru); + LOCK_INIT (&nfs3->fdlrulock); + nfs3->fdcount = 0; + + ret = rpcsvc_create_listeners (nfs->rpcsvc, nfsx->options, nfsx->name); + if (ret == -1) { + gf_log (GF_NFS, GF_LOG_ERROR, "Unable to create listeners"); + goto free_localpool; + } + + nfs->nfs3state = nfs3; + ret = 0; + +free_localpool: + if (ret == -1) + mem_pool_destroy (nfs3->localpool); + +ret: + if (ret == -1) { + GF_FREE (nfs3); + nfs3 = NULL; + } + + return nfs3; +} + + +rpcsvc_program_t * +nfs3svc_init (xlator_t *nfsx) +{ + struct nfs3_state *nfs3 = NULL; + + if (!nfsx) + return NULL; + + nfs3 = nfs3_init_state (nfsx); + if (!nfs3) { + gf_log (GF_NFS3, GF_LOG_ERROR, "NFSv3 state init failed"); + return NULL; + } + + nfs3prog.private = nfs3; + + return &nfs3prog; +} + +int +nfs3_reconfigure_state (xlator_t *nfsx, dict_t *options) +{ + int ret = -1; + struct nfs3_export *exp = NULL; + struct nfs_state *nfs = NULL; + struct nfs3_state *nfs3 = NULL; + + if ((!nfsx) || (!nfsx->private) || (!options)) + goto out; + + nfs = (struct nfs_state *)nfsx->private; + nfs3 = nfs->nfs3state; + if (!nfs3) + goto out; + + ret = nfs3_init_options (nfs3, options); + if (ret) { + gf_log (GF_NFS3, GF_LOG_ERROR, + "Failed to reconfigure options"); + goto out; + } + + list_for_each_entry (exp, &nfs3->exports, explist) { + ret = nfs3_init_subvolume_options (nfsx, exp, options); + if (ret) { + gf_log (GF_NFS3, GF_LOG_ERROR, + "Failed to reconfigure subvol options"); + goto out; + } + } + + ret = 0; +out: + return ret; +} diff --git a/xlators/nfs/server/src/nfs3.h b/xlators/nfs/server/src/nfs3.h new file mode 100644 index 000000000..0c35445a4 --- /dev/null +++ b/xlators/nfs/server/src/nfs3.h @@ -0,0 +1,285 @@ +/* + Copyright (c) 2010-2011 Gluster, Inc. <http://www.gluster.com> + This file is part of GlusterFS. + + This file is licensed to you under your choice of the GNU Lesser + General Public License, version 3 or any later version (LGPLv3 or + later), or the GNU General Public License, version 2 (GPLv2), in all + cases as published by the Free Software Foundation. +*/ + +#ifndef _NFS3_H_ +#define _NFS3_H_ + +#ifndef _CONFIG_H +#define _CONFIG_H +#include "config.h" +#endif + +#include "rpcsvc.h" +#include "dict.h" +#include "xlator.h" +#include "iobuf.h" +#include "nfs.h" +#include "nfs3-fh.h" +#include "nfs-common.h" +#include "xdr-nfs3.h" +#include "mem-pool.h" +#include "nlm4.h" +#include "acl3-xdr.h" +#include "acl3.h" +#include <sys/statvfs.h> + +#define GF_NFS3 GF_NFS"-nfsv3" + +#define GF_NFS3_DEFAULT_MEMFACTOR 15 +#define GF_NFS3_IOBPOOL_MULT GF_NFS_CONCURRENT_OPS_MULT +#define GF_NFS3_CLTABLE_BUCKETS_MULT 2 +#define GF_NFS3_FDTABLE_BUCKETS_MULT 2 + + +/* Static values used for FSINFO + * To change the maximum rsize and wsize supported by the NFS client, adjust + * GF_NFS3_FILE_IO_SIZE_MAX. The Gluster NFS server defaults to 1MB(1048576) + * (same as kernel NFS server). For slower network, rsize/wsize can be trimmed + * to 16/32/64-KB. rsize and wsize can be tuned through nfs.read-size and + * nfs.write-size respectively. + * + * NB: For Kernel-NFS, NFS_MAX_FILE_IO_SIZE is 1048576U (1MB). + */ +#define GF_NFS3_FILE_IO_SIZE_MAX (1 * GF_UNIT_MB) /* 1048576 */ +#define GF_NFS3_FILE_IO_SIZE_MIN (4 * GF_UNIT_KB) /* 4096 */ + +#define GF_NFS3_FILE_IO_SIZE_DEF GF_NFS3_FILE_IO_SIZE_MAX + +#define GF_NFS3_RTMAX GF_NFS3_FILE_IO_SIZE_MAX +#define GF_NFS3_RTMIN GF_NFS3_FILE_IO_SIZE_MIN +#define GF_NFS3_RTPREF GF_NFS3_FILE_IO_SIZE_DEF +#define GF_NFS3_RTMULT GF_NFS3_FILE_IO_SIZE_MIN + +#define GF_NFS3_WTMAX GF_NFS3_FILE_IO_SIZE_MAX +#define GF_NFS3_WTMIN GF_NFS3_FILE_IO_SIZE_MIN +#define GF_NFS3_WTPREF GF_NFS3_FILE_IO_SIZE_DEF +#define GF_NFS3_WTMULT GF_NFS3_FILE_IO_SIZE_MIN + +/* This can be tuned through nfs.readdir-size */ +#define GF_NFS3_DTMAX GF_NFS3_FILE_IO_SIZE_MAX +#define GF_NFS3_DTMIN GF_NFS3_FILE_IO_SIZE_MIN +#define GF_NFS3_DTPREF GF_NFS3_FILE_IO_SIZE_DEF + +#define GF_NFS3_MAXFILESIZE (1 * GF_UNIT_PB) + +#define GF_NFS3_IO_SIZE 4096 /* 4-KB */ +#define GF_NFS3_IO_SHIFT 12 /* 2^12 = 4KB */ + +/* FIXME: Handle time resolutions */ +#define GF_NFS3_TIMEDELTA_SECS {1,0} +#define GF_NFS3_TIMEDELTA_NSECS {0,1} +#define GF_NFS3_TIMEDELTA_MSECS {0,1000000} + +#define GF_NFS3_FS_PROP (FSF3_LINK | FSF3_SYMLINK | FSF3_HOMOGENEOUS | FSF3_CANSETTIME) + +#define GF_NFS3_DIRFD_VALID 1 +#define GF_NFS3_DIRFD_INVALID 0 + +#define GF_NFS3_VOLACCESS_RW 1 +#define GF_NFS3_VOLACCESS_RO 2 + + +#define GF_NFS3_FDCACHE_SIZE 512 +/* This should probably be moved to a more generic layer so that if needed + * different versions of NFS protocol can use the same thing. + */ +struct nfs3_fd_entry { + fd_t *cachedfd; + struct list_head list; +}; + +/* Per subvolume nfs3 specific state */ +struct nfs3_export { + struct list_head explist; + xlator_t *subvol; + uuid_t volumeid; + int access; + int trusted_sync; + int trusted_write; + int rootlookedup; +}; + +#define GF_NFS3_DEFAULT_VOLACCESS (GF_NFS3_VOLACCESS_RW) + +/* The NFSv3 protocol state */ +typedef struct nfs3_state { + + /* The NFS xlator pointer. The NFS xlator can be running + * multiple versions of the NFS protocol. + */ + xlator_t *nfsx; + + /* The iob pool from which memory allocations are made for receiving + * and sending network messages. + */ + struct iobuf_pool *iobpool; + + /* List of child subvolumes for the NFSv3 protocol. + * Right now, is simply referring to the list of children in nfsx above. + */ + xlator_list_t *exportslist; + + struct list_head exports; + /* Mempool for allocations of struct nfs3_local */ + struct mem_pool *localpool; + + /* Server start-up timestamp, currently used for write verifier. */ + uint64_t serverstart; + + /* NFSv3 Protocol configurables */ + uint64_t readsize; + uint64_t writesize; + uint64_t readdirsize; + + /* Size of the iobufs used, depends on the sizes of the three params + * above. + */ + uint64_t iobsize; + + unsigned int memfactor; + + struct list_head fdlru; + gf_lock_t fdlrulock; + int fdcount; + uint32_t occ_logger; +} nfs3_state_t; + +typedef enum nfs3_lookup_type { + GF_NFS3_REVALIDATE = 1, + GF_NFS3_FRESH, +} nfs3_lookup_type_t; + +typedef union args_ { + nlm4_stat nlm4_stat; + nlm4_holder nlm4_holder; + nlm4_lock nlm4_lock; + nlm4_share nlm4_share; + nlm4_testrply nlm4_testrply; + nlm4_testres nlm4_testres; + nlm4_testargs nlm4_testargs; + nlm4_res nlm4_res; + nlm4_lockargs nlm4_lockargs; + nlm4_cancargs nlm4_cancargs; + nlm4_unlockargs nlm4_unlockargs; + nlm4_shareargs nlm4_shareargs; + nlm4_shareres nlm4_shareres; + nlm4_freeallargs nlm4_freeallargs; + getaclargs getaclargs; + setaclargs setaclargs; + getaclreply getaclreply; + setaclreply setaclreply; +} args; + + +typedef int (*nfs3_resume_fn_t) (void *cs); +/* Structure used to communicate state between a fop and its callback. + * Not all members are used at all times. Usage is fop and NFS request + * dependent. + * + * I wish we could have a smaller structure for communicating state + * between callers and callbacks. It could be broken into smaller parts + * but I feel that will lead to a proliferation of types/structures and then + * we'll just be tracking down which structure is used by which fop, not + * to mention that having one type allows me to used a single mem-pool. + * Imagine the chaos if we need a mem-pool for each one of those sub-structures. + */ +struct nfs3_local { + rpcsvc_request_t *req; + xlator_t *vol; + nfs3_resume_fn_t resume_fn; + xlator_t *nfsx; + struct nfs3_state *nfs3state; + + /* The list hook to attach this call state to the inode's queue till + * the opening of the fd on the inode completes. + */ + struct list_head openwait_q; + + /* Per-NFSv3 Op state */ + struct nfs3_fh parent; + struct nfs3_fh fh; + fd_t *fd; + uint32_t accessbits; + int operrno; + count3 dircount; + count3 maxcount; + struct statvfs fsstat; + gf_dirent_t entries; + struct iatt stbuf; + struct iatt preparent; + struct iatt postparent; + int32_t setattr_valid; + nfstime3 timestamp; + loc_t oploc; + int writetype; + count3 datacount; + offset3 dataoffset; + struct iobuf *iob; + struct iobref *iobref; + createmode3 createmode; + uint64_t cookieverf; + int sattrguardcheck; + char *pathname; + ftype3 mknodtype; + specdata3 devnums; + cookie3 cookie; + struct iovec datavec; + mode_t mode; + + /* NFSv3 FH resolver state */ + int hardresolved; + struct nfs3_fh resolvefh; + loc_t resolvedloc; + int resolve_ret; + int resolve_errno; + int hashidx; + fd_t *resolve_dir_fd; + char *resolventry; + nfs3_lookup_type_t lookuptype; + gf_dirent_t *hashmatch; + gf_dirent_t *entrymatch; + off_t lastentryoffset; + struct flock flock; + args args; + nlm4_lkowner_t lkowner; + char cookiebytes[1024]; + struct nfs3_fh lockfh; + int monitor; + rpc_transport_t *trans; + call_frame_t *frame; + + /* ACL */ + aclentry aclentry[NFS_ACL_MAX_ENTRIES]; + aclentry daclentry[NFS_ACL_MAX_ENTRIES]; + int aclcount; + char aclxattr[NFS_ACL_MAX_ENTRIES*8 + 4]; + int daclcount; + char daclxattr[NFS_ACL_MAX_ENTRIES*8 + 4]; +}; + +#define nfs3_is_revalidate_lookup(cst) ((cst)->lookuptype == GF_NFS3_REVALIDATE) +#define nfs3_lookup_op(cst) (rpcsvc_request_procnum(cst->req) == NFS3_LOOKUP) +#define nfs3_create_op(cst) (rpcsvc_request_procnum(cst->req) == NFS3_CREATE) +#define nfs3_create_exclusive_op(cst) ((cst)->createmode == EXCLUSIVE) + +typedef struct nfs3_local nfs3_call_state_t; + +/* Queue of ops waiting for open fop to return. */ +struct inode_op_queue { + struct list_head opq; + pthread_mutex_t qlock; +}; + +extern rpcsvc_program_t * +nfs3svc_init (xlator_t *nfsx); + +extern int +nfs3_reconfigure_state (xlator_t *nfsx, dict_t *options); +#endif diff --git a/xlators/nfs/server/src/nlm4.c b/xlators/nfs/server/src/nlm4.c new file mode 100644 index 000000000..5c5d87412 --- /dev/null +++ b/xlators/nfs/server/src/nlm4.c @@ -0,0 +1,2525 @@ +/* + Copyright (c) 2012 Gluster, Inc. <http://www.gluster.com> + This file is part of GlusterFS. + + This file is licensed to you under your choice of the GNU Lesser + General Public License, version 3 or any later version (LGPLv3 or + later), or the GNU General Public License, version 2 (GPLv2), in all + cases as published by the Free Software Foundation. +*/ + +#ifndef _CONFIG_H +#define _CONFIG_H +#include "config.h" +#endif + +#include "defaults.h" +#include "rpcsvc.h" +#include "dict.h" +#include "xlator.h" +#include "nfs.h" +#include "mem-pool.h" +#include "logging.h" +#include "nfs-fops.h" +#include "inode.h" +#include "mount3.h" +#include "nfs3.h" +#include "nfs-mem-types.h" +#include "nfs3-helpers.h" +#include "nfs3-fh.h" +#include "nlm4.h" +#include "nlm4-xdr.h" +#include "msg-nfs3.h" +#include "nfs-generics.h" +#include "rpc-clnt.h" +#include "nsm-xdr.h" +#include "nlmcbk-xdr.h" +#include "run.h" +#include <unistd.h> +#include <rpc/pmap_clnt.h> +#include <rpc/rpc.h> +#include <rpc/xdr.h> +#include <statedump.h> + +/* TODO: + * 1) 2 opens racing .. creating an fd leak. + * 2) use mempool for nlmclnt - destroy if no fd exists, create during 1st call + */ + +typedef ssize_t (*nlm4_serializer) (struct iovec outmsg, void *args); + +extern void nfs3_call_state_wipe (nfs3_call_state_t *cs); + +struct list_head nlm_client_list; +gf_lock_t nlm_client_list_lk; + +/* race on this is harmless */ +int nlm_grace_period = 50; + +#define nlm4_validate_nfs3_state(request, state, status, label, retval) \ + do { \ + state = rpcsvc_request_program_private (request); \ + if (!state) { \ + gf_log (GF_NLM, GF_LOG_ERROR, "NFSv3 state " \ + "missing from RPC request"); \ + rpcsvc_request_seterr (req, SYSTEM_ERR); \ + status = nlm4_failed; \ + goto label; \ + } \ + } while (0); \ + +nfs3_call_state_t * +nfs3_call_state_init (struct nfs3_state *s, rpcsvc_request_t *req, xlator_t *v); + +#define nlm4_handle_call_state_init(nfs3state, calls, rq, opstat, errlabel)\ + do { \ + calls = nlm4_call_state_init ((nfs3state), (rq)); \ + if (!calls) { \ + gf_log (GF_NLM, GF_LOG_ERROR, "Failed to " \ + "init call state"); \ + opstat = nlm4_failed; \ + rpcsvc_request_seterr (req, SYSTEM_ERR); \ + goto errlabel; \ + } \ + } while (0) \ + +#define nlm4_validate_gluster_fh(handle, status, errlabel) \ + do { \ + if (!nfs3_fh_validate (handle)) { \ + status = nlm4_stale_fh; \ + goto errlabel; \ + } \ + } while (0) \ + +xlator_t * +nfs3_fh_to_xlator (struct nfs3_state *nfs3, struct nfs3_fh *fh); + +#define nlm4_map_fh_to_volume(nfs3state, handle, req, volume, status, label) \ + do { \ + char exportid[256], gfid[256]; \ + rpc_transport_t *trans = NULL; \ + volume = nfs3_fh_to_xlator ((nfs3state), &handle); \ + if (!volume) { \ + uuid_unparse (handle.exportid, exportid); \ + uuid_unparse (handle.gfid, gfid); \ + trans = rpcsvc_request_transport (req); \ + gf_log (GF_NLM, GF_LOG_ERROR, "Failed to map " \ + "FH to vol: client=%s, exportid=%s, gfid=%s",\ + trans->peerinfo.identifier, exportid, \ + gfid); \ + gf_log (GF_NLM, GF_LOG_ERROR, \ + "Stale nfs client %s must be trying to "\ + "connect to a deleted volume, please " \ + "unmount it.", trans->peerinfo.identifier);\ + status = nlm4_stale_fh; \ + goto label; \ + } else { \ + gf_log (GF_NLM, GF_LOG_TRACE, "FH to Volume: %s"\ + ,volume->name); \ + rpcsvc_request_set_private (req, volume); \ + } \ + } while (0); \ + +#define nlm4_volume_started_check(nfs3state, vlm, rtval, erlbl) \ + do { \ + if ((!nfs_subvolume_started (nfs_state (nfs3state->nfsx), vlm))){\ + gf_log (GF_NLM, GF_LOG_ERROR, "Volume is disabled: %s",\ + vlm->name); \ + rtval = RPCSVC_ACTOR_IGNORE; \ + goto erlbl; \ + } \ + } while (0) \ + +#define nlm4_check_fh_resolve_status(cst, nfstat, erlabl) \ + do { \ + xlator_t *xlatorp = NULL; \ + char buf[256], gfid[256]; \ + rpc_transport_t *trans = NULL; \ + if ((cst)->resolve_ret < 0) { \ + trans = rpcsvc_request_transport (cst->req); \ + xlatorp = nfs3_fh_to_xlator (cst->nfs3state, \ + &cst->resolvefh); \ + uuid_unparse (cst->resolvefh.gfid, gfid); \ + snprintf (buf, sizeof (buf), "(%s) %s : %s", \ + trans->peerinfo.identifier, \ + xlatorp ? xlatorp->name : "ERR", \ + gfid); \ + gf_log (GF_NLM, GF_LOG_ERROR, "Unable to resolve FH"\ + ": %s", buf); \ + nfstat = nlm4_errno_to_nlm4stat (cst->resolve_errno);\ + goto erlabl; \ + } \ + } while (0) \ + + +void +nlm4_prep_nlm4_testargs (nlm4_testargs *args, struct nfs3_fh *fh, + nlm4_lkowner_t *oh, char *cookiebytes) +{ + memset (args, 0, sizeof (*args)); + args->alock.fh.n_bytes = (void *)fh; + args->alock.oh.n_bytes = (void *)oh; + args->cookie.n_bytes = (void *)cookiebytes; +} + +void +nlm4_prep_nlm4_lockargs (nlm4_lockargs *args, struct nfs3_fh *fh, + nlm4_lkowner_t *oh, char *cookiebytes) +{ + memset (args, 0, sizeof (*args)); + args->alock.fh.n_bytes = (void *)fh; + args->alock.oh.n_bytes = (void *)oh; + args->cookie.n_bytes = (void *)cookiebytes; +} + +void +nlm4_prep_nlm4_cancargs (nlm4_cancargs *args, struct nfs3_fh *fh, + nlm4_lkowner_t *oh, char *cookiebytes) +{ + memset (args, 0, sizeof (*args)); + args->alock.fh.n_bytes = (void *)fh; + args->alock.oh.n_bytes = (void *)oh; + args->cookie.n_bytes = (void *)cookiebytes; +} + +void +nlm4_prep_nlm4_unlockargs (nlm4_unlockargs *args, struct nfs3_fh *fh, + nlm4_lkowner_t *oh, char *cookiebytes) +{ + memset (args, 0, sizeof (*args)); + args->alock.fh.n_bytes = (void *)fh; + args->alock.oh.n_bytes = (void *)oh; + args->cookie.n_bytes = (void *)cookiebytes; +} + +void +nlm4_prep_shareargs (nlm4_shareargs *args, struct nfs3_fh *fh, + nlm4_lkowner_t *oh, char *cookiebytes) +{ + memset (args, 0, sizeof (*args)); + args->share.fh.n_bytes = (void *)fh; + args->share.oh.n_bytes = (void *)oh; + args->cookie.n_bytes = (void *)cookiebytes; +} + +void +nlm4_prep_freeallargs (nlm4_freeallargs *args, nlm4_lkowner_t *oh) +{ + memset (args, 0, sizeof (*args)); + args->name = (void *)oh; +} + +void +nlm_copy_lkowner (gf_lkowner_t *dst, netobj *src) +{ + dst->len = src->n_len; + memcpy (dst->data, src->n_bytes, dst->len); +} + +int +nlm_is_oh_same_lkowner (gf_lkowner_t *a, netobj *b) +{ + if (!a || !b) { + gf_log (GF_NLM, GF_LOG_ERROR, "invalid args"); + return -1; + } + + return (a->len == b->n_len && + !memcmp (a->data, b->n_bytes, a->len)); +} + +nlm4_stats +nlm4_errno_to_nlm4stat (int errnum) +{ + nlm4_stats stat = nlm4_denied; + + switch (errnum) { + case 0: + stat = nlm4_granted; + break; + case EROFS: + stat = nlm4_rofs; + break; + case ESTALE: + stat = nlm4_stale_fh; + break; + case ENOLCK: + stat = nlm4_failed; + break; + default: + stat = nlm4_denied; + break; + } + + return stat; +} + +nfs3_call_state_t * +nlm4_call_state_init (struct nfs3_state *s, rpcsvc_request_t *req) +{ + nfs3_call_state_t *cs = NULL; + + if ((!s) || (!req)) + return NULL; + + cs = (nfs3_call_state_t *) mem_get (s->localpool); + if (!cs) + return NULL; + + memset (cs, 0, sizeof (*cs)); + INIT_LIST_HEAD (&cs->entries.list); + INIT_LIST_HEAD (&cs->openwait_q); + cs->operrno = EINVAL; + cs->req = req; + cs->nfsx = s->nfsx; + cs->nfs3state = s; + cs->monitor = 1; + + return cs; +} + +int +nlm_monitor (char *caller_name) +{ + nlm_client_t *nlmclnt = NULL; + int monitor = -1; + + LOCK (&nlm_client_list_lk); + list_for_each_entry (nlmclnt, &nlm_client_list, nlm_clients) { + if (!strcmp(caller_name, nlmclnt->caller_name)) { + monitor = nlmclnt->nsm_monitor; + nlmclnt->nsm_monitor = 1; + break; + } + } + UNLOCK (&nlm_client_list_lk); + + if (monitor == -1) + gf_log (GF_NLM, GF_LOG_ERROR, "%s was not found in " + "the nlmclnt list", caller_name); + + return monitor; +} + +rpc_clnt_t * +nlm_get_rpc_clnt (char *caller_name) +{ + nlm_client_t *nlmclnt = NULL; + int nlmclnt_found = 0; + rpc_clnt_t *rpc_clnt = NULL; + + LOCK (&nlm_client_list_lk); + list_for_each_entry (nlmclnt, &nlm_client_list, nlm_clients) { + if (!strcmp(caller_name, nlmclnt->caller_name)) { + nlmclnt_found = 1; + break; + } + } + if (!nlmclnt_found) + goto ret; + if (nlmclnt->rpc_clnt) + rpc_clnt = rpc_clnt_ref (nlmclnt->rpc_clnt); +ret: + UNLOCK (&nlm_client_list_lk); + return rpc_clnt; +} + +int +nlm_set_rpc_clnt (rpc_clnt_t *rpc_clnt, char *caller_name) +{ + nlm_client_t *nlmclnt = NULL; + int nlmclnt_found = 0; + int ret = -1; + + LOCK (&nlm_client_list_lk); + list_for_each_entry (nlmclnt, &nlm_client_list, nlm_clients) { + if (!strcmp(caller_name, nlmclnt->caller_name)) { + nlmclnt_found = 1; + break; + } + } + + if (!nlmclnt_found) { + nlmclnt = GF_CALLOC (1, sizeof(*nlmclnt), + gf_nfs_mt_nlm4_nlmclnt); + if (nlmclnt == NULL) + goto ret; + + INIT_LIST_HEAD(&nlmclnt->fdes); + INIT_LIST_HEAD(&nlmclnt->nlm_clients); + INIT_LIST_HEAD(&nlmclnt->shares); + + list_add (&nlmclnt->nlm_clients, &nlm_client_list); + nlmclnt->caller_name = gf_strdup (caller_name); + } + + if (nlmclnt->rpc_clnt == NULL) { + nlmclnt->rpc_clnt = rpc_clnt_ref (rpc_clnt); + } + ret = 0; +ret: + UNLOCK (&nlm_client_list_lk); + return ret; +} + +int +nlm_unset_rpc_clnt (rpc_clnt_t *rpc) +{ + nlm_client_t *nlmclnt = NULL; + rpc_clnt_t *rpc_clnt = NULL; + + LOCK (&nlm_client_list_lk); + list_for_each_entry (nlmclnt, &nlm_client_list, nlm_clients) { + if (rpc == nlmclnt->rpc_clnt) { + rpc_clnt = nlmclnt->rpc_clnt; + nlmclnt->rpc_clnt = NULL; + break; + } + } + UNLOCK (&nlm_client_list_lk); + if (rpc_clnt == NULL) { + return -1; + } + if (rpc_clnt) { + /* cleanup the saved-frames before last unref */ + rpc_clnt_connection_cleanup (&rpc_clnt->conn); + + rpc_clnt_unref (rpc_clnt); + } + return 0; +} + +int +nlm_add_nlmclnt (char *caller_name) +{ + nlm_client_t *nlmclnt = NULL; + int nlmclnt_found = 0; + int ret = -1; + + LOCK (&nlm_client_list_lk); + list_for_each_entry (nlmclnt, &nlm_client_list, nlm_clients) { + if (!strcmp(caller_name, nlmclnt->caller_name)) { + nlmclnt_found = 1; + break; + } + } + if (!nlmclnt_found) { + nlmclnt = GF_CALLOC (1, sizeof(*nlmclnt), + gf_nfs_mt_nlm4_nlmclnt); + if (nlmclnt == NULL) { + gf_log (GF_NLM, GF_LOG_DEBUG, "malloc error"); + goto ret; + } + + INIT_LIST_HEAD(&nlmclnt->fdes); + INIT_LIST_HEAD(&nlmclnt->nlm_clients); + INIT_LIST_HEAD(&nlmclnt->shares); + + list_add (&nlmclnt->nlm_clients, &nlm_client_list); + nlmclnt->caller_name = gf_strdup (caller_name); + } + ret = 0; +ret: + UNLOCK (&nlm_client_list_lk); + return ret; +} + +int +nlm4svc_submit_reply (rpcsvc_request_t *req, void *arg, nlm4_serializer sfunc) +{ + struct iovec outmsg = {0, }; + struct iobuf *iob = NULL; + struct nfs3_state *nfs3 = NULL; + int ret = -1; + ssize_t msglen = 0; + struct iobref *iobref = NULL; + + if (!req) + return -1; + + nfs3 = (struct nfs3_state *)rpcsvc_request_program_private (req); + if (!nfs3) { + gf_log (GF_NLM, GF_LOG_ERROR, "mount state not found"); + goto ret; + } + + /* First, get the io buffer into which the reply in arg will + * be serialized. + */ + iob = iobuf_get (nfs3->iobpool); + if (!iob) { + gf_log (GF_NLM, GF_LOG_ERROR, "Failed to get iobuf"); + goto ret; + } + + iobuf_to_iovec (iob, &outmsg); + /* Use the given serializer to translate the give C structure in arg + * to XDR format which will be written into the buffer in outmsg. + */ + msglen = sfunc (outmsg, arg); + if (msglen < 0) { + gf_log (GF_NLM, GF_LOG_ERROR, "Failed to encode message"); + goto ret; + } + outmsg.iov_len = msglen; + + iobref = iobref_new (); + if (iobref == NULL) { + gf_log (GF_NLM, GF_LOG_ERROR, "Failed to get iobref"); + goto ret; + } + + ret = iobref_add (iobref, iob); + if (ret) { + gf_log (GF_NLM, GF_LOG_ERROR, "Failed to add iob to iobref"); + goto ret; + } + + /* Then, submit the message for transmission. */ + ret = rpcsvc_submit_message (req, &outmsg, 1, NULL, 0, iobref); + if (ret == -1) { + gf_log (GF_NLM, GF_LOG_ERROR, "Reply submission failed"); + goto ret; + } + + ret = 0; +ret: + if (iob) + iobuf_unref (iob); + if (iobref) + iobref_unref (iobref); + + return ret; +} + +typedef int (*nlm4_resume_fn_t) (void *cs); + +int32_t +nlm4_file_open_cbk (call_frame_t *frame, void *cookie, xlator_t *this, + int32_t op_ret, int32_t op_errno, fd_t *fd, dict_t *xdata) +{ + nfs3_call_state_t *cs = frame->local; + + if (op_ret == 0) + fd_bind (cs->fd); + cs->resolve_ret = op_ret; + cs->resume_fn (cs); + frame->local = NULL; + STACK_DESTROY (frame->root); + return 0; +} + +void * +nsm_monitor(void *arg) +{ + CLIENT *clnt = NULL; + enum clnt_stat ret; + struct mon nsm_mon; + struct sm_stat_res res; + struct timeval tout = { 5, 0 }; + char *host = NULL; + + host = arg; + nsm_mon.mon_id.mon_name = gf_strdup(host); + nsm_mon.mon_id.my_id.my_name = gf_strdup("localhost"); + nsm_mon.mon_id.my_id.my_prog = NLMCBK_PROGRAM; + nsm_mon.mon_id.my_id.my_vers = NLMCBK_V1; + nsm_mon.mon_id.my_id.my_proc = NLMCBK_SM_NOTIFY; + /* nothing to put in the private data */ +#define SM_PROG 100024 +#define SM_VERS 1 +#define SM_MON 2 + + /* create a connection to nsm on the localhost */ + clnt = clnt_create("localhost", SM_PROG, SM_VERS, "tcp"); + if(!clnt) + { + gf_log (GF_NLM, GF_LOG_ERROR, "%s", + clnt_spcreateerror ("Clnt_create()")); + goto out; + } + + ret = clnt_call(clnt, SM_MON, + (xdrproc_t) xdr_mon, (caddr_t) & nsm_mon, + (xdrproc_t) xdr_sm_stat_res, (caddr_t) & res, tout); + if(ret != RPC_SUCCESS) + { + gf_log (GF_NLM, GF_LOG_ERROR, "clnt_call(): %s", + clnt_sperrno(ret)); + goto out; + } + if(res.res_stat != STAT_SUCC) + { + gf_log (GF_NLM, GF_LOG_ERROR, "clnt_call(): %s", + clnt_sperrno(ret)); + goto out; + } + +out: + GF_FREE(nsm_mon.mon_id.mon_name); + GF_FREE(nsm_mon.mon_id.my_id.my_name); + if (clnt != NULL) + clnt_destroy(clnt); + return NULL; +} + +nlm_client_t * +__nlm_get_uniq (char *caller_name) +{ + nlm_client_t *nlmclnt = NULL; + + if (!caller_name) + return NULL; + + list_for_each_entry (nlmclnt, &nlm_client_list, nlm_clients) { + if (!strcmp(caller_name, nlmclnt->caller_name)) + return nlmclnt; + } + + return NULL; +} + +nlm_client_t * +nlm_get_uniq (char *caller_name) +{ + nlm_client_t *nlmclnt = NULL; + + LOCK (&nlm_client_list_lk); + nlmclnt = __nlm_get_uniq (caller_name); + UNLOCK (&nlm_client_list_lk); + + return nlmclnt; +} + + +int +nlm4_file_open_and_resume(nfs3_call_state_t *cs, nlm4_resume_fn_t resume) +{ + fd_t *fd = NULL; + int ret = -1; + int flags = 0; + nlm_client_t *nlmclnt = NULL; + call_frame_t *frame = NULL; + + if (cs->args.nlm4_lockargs.exclusive == _gf_false) + flags = O_RDONLY; + else + flags = O_WRONLY; + + nlmclnt = nlm_get_uniq (cs->args.nlm4_lockargs.alock.caller_name); + if (nlmclnt == NULL) { + gf_log (GF_NLM, GF_LOG_ERROR, "nlm_get_uniq() returned NULL"); + ret = -ENOLCK; + goto err; + } + cs->resume_fn = resume; + fd = fd_lookup_uint64 (cs->resolvedloc.inode, (uint64_t)nlmclnt); + if (fd) { + cs->fd = fd; + cs->resolve_ret = 0; + cs->resume_fn(cs); + ret = 0; + goto err; + } + + fd = fd_create_uint64 (cs->resolvedloc.inode, (uint64_t)nlmclnt); + if (fd == NULL) { + gf_log (GF_NLM, GF_LOG_ERROR, "fd_create_uint64() returned NULL"); + ret = -ENOLCK; + goto err; + } + + cs->fd = fd; + + frame = create_frame (cs->nfsx, cs->nfsx->ctx->pool); + if (!frame) { + gf_log (GF_NLM, GF_LOG_ERROR, "unable to create frame"); + ret = -ENOMEM; + goto err; + } + + frame->root->pid = NFS_PID; + frame->root->uid = rpcsvc_request_uid (cs->req); + frame->root->gid = rpcsvc_request_gid (cs->req); + frame->local = cs; + nfs_fix_groups (cs->nfsx, frame->root); + + STACK_WIND_COOKIE (frame, nlm4_file_open_cbk, cs->vol, cs->vol, + cs->vol->fops->open, &cs->resolvedloc, flags, + cs->fd, NULL); + ret = 0; +err: + return ret; +} + +int +nlm4_generic_reply (rpcsvc_request_t *req, netobj cookie, nlm4_stats stat) +{ + nlm4_res res; + + memset (&res, 0, sizeof (res)); + res.cookie = cookie; + res.stat.stat = stat; + + nlm4svc_submit_reply (req, (void *)&res, + (nlm4_serializer)xdr_serialize_nlm4_res); + return 0; +} + +int +nlm4svc_null (rpcsvc_request_t *req) +{ + struct iovec dummyvec = {0, }; + + if (!req) { + gf_log (GF_NLM, GF_LOG_ERROR, "Got NULL request!"); + return 0; + } + rpcsvc_submit_generic (req, &dummyvec, 1, NULL, 0, NULL); + return 0; +} + +int +nlm4_gf_flock_to_holder (nlm4_holder *holder, struct gf_flock *flock) +{ + switch (flock->l_type) { + case GF_LK_F_WRLCK: + holder->exclusive = 1; + break; + } + + holder->svid = flock->l_pid; + holder->l_offset = flock->l_start; + holder->l_len = flock->l_len; + return 0; +} + +int +nlm4_lock_to_gf_flock (struct gf_flock *flock, nlm4_lock *lock, int excl) +{ + flock->l_pid = lock->svid; + flock->l_start = lock->l_offset; + flock->l_len = lock->l_len; + if (excl) + flock->l_type = F_WRLCK; + else + flock->l_type = F_RDLCK; + flock->l_whence = SEEK_SET; + nlm_copy_lkowner (&flock->l_owner, &lock->oh); + return 0; +} + +rpc_clnt_procedure_t nlm4_clnt_actors[NLM4_PROC_COUNT] = { + [NLM4_NULL] = {"NULL", NULL}, + [NLM4_GRANTED] = {"GRANTED", NULL}, +}; + +char *nlm4_clnt_names[NLM4_PROC_COUNT] = { + [NLM4_NULL] = "NULL", + [NLM4_GRANTED] = "GRANTED", +}; + +rpc_clnt_prog_t nlm4clntprog = { + .progname = "NLMv4", + .prognum = NLM_PROGRAM, + .progver = NLM_V4, + .numproc = NLM4_PROC_COUNT, + .proctable = nlm4_clnt_actors, + .procnames = nlm4_clnt_names, +}; + +int +nlm4_test_reply (nfs3_call_state_t *cs, nlm4_stats stat, struct gf_flock *flock) +{ + nlm4_testres res; + + memset (&res, 0, sizeof (res)); + res.cookie = cs->args.nlm4_testargs.cookie; + res.stat.stat = stat; + if (stat == nlm4_denied) + nlm4_gf_flock_to_holder (&res.stat.nlm4_testrply_u.holder, + flock); + + nlm4svc_submit_reply (cs->req, (void *)&res, + (nlm4_serializer)xdr_serialize_nlm4_testres); + return 0; +} + +int +nlm4svc_test_cbk (call_frame_t *frame, void *cookie, xlator_t *this, + int32_t op_ret, int32_t op_errno, struct gf_flock *flock, + dict_t *xdata) +{ + nlm4_stats stat = nlm4_denied; + nfs3_call_state_t *cs = NULL; + + cs = frame->local; + if (op_ret == -1) { + stat = nlm4_errno_to_nlm4stat (op_errno); + goto err; + } else if (flock->l_type == F_UNLCK) + stat = nlm4_granted; + +err: + nlm4_test_reply (cs, stat, flock); + nfs3_call_state_wipe (cs); + return 0; +} + +int +nlm4_test_fd_resume (void *carg) +{ + int ret = -EFAULT; + nfs_user_t nfu = {0, }; + nfs3_call_state_t *cs = NULL; + struct gf_flock flock = {0, }; + + if (!carg) + return ret; + + cs = (nfs3_call_state_t *)carg; + nfs_request_user_init (&nfu, cs->req); + nlm4_lock_to_gf_flock (&flock, &cs->args.nlm4_testargs.alock, + cs->args.nlm4_testargs.exclusive); + nlm_copy_lkowner (&nfu.lk_owner, &cs->args.nlm4_testargs.alock.oh); + ret = nfs_lk (cs->nfsx, cs->vol, &nfu, cs->fd, F_GETLK, &flock, + nlm4svc_test_cbk, cs); + + return ret; +} + + +int +nlm4_test_resume (void *carg) +{ + nlm4_stats stat = nlm4_failed; + int ret = -1; + nfs3_call_state_t *cs = NULL; + fd_t *fd = NULL; + + if (!carg) + return ret; + + cs = (nfs3_call_state_t *)carg; + nlm4_check_fh_resolve_status (cs, stat, nlm4err); + fd = fd_anonymous (cs->resolvedloc.inode); + if (!fd) + goto nlm4err; + cs->fd = fd; + ret = nlm4_test_fd_resume (cs); + +nlm4err: + if (ret < 0) { + gf_log (GF_NLM, GF_LOG_ERROR, "unable to open_and_resume"); + stat = nlm4_errno_to_nlm4stat (-ret); + nlm4_test_reply (cs, stat, NULL); + nfs3_call_state_wipe (cs); + } + + return ret; +} + +int +nlm4svc_test (rpcsvc_request_t *req) +{ + xlator_t *vol = NULL; + nlm4_stats stat = nlm4_failed; + struct nfs_state *nfs = NULL; + nfs3_state_t *nfs3 = NULL; + nfs3_call_state_t *cs = NULL; + int ret = RPCSVC_ACTOR_ERROR; + struct nfs3_fh fh = {{0}, }; + + if (!req) + return ret; + + nlm4_validate_nfs3_state (req, nfs3, stat, rpcerr, ret); + nfs = nfs_state (nfs3->nfsx); + nlm4_handle_call_state_init (nfs->nfs3state, cs, req, + stat, rpcerr); + + nlm4_prep_nlm4_testargs (&cs->args.nlm4_testargs, &fh, &cs->lkowner, + cs->cookiebytes); + if (xdr_to_nlm4_testargs(req->msg[0], &cs->args.nlm4_testargs) <= 0) { + gf_log (GF_NLM, GF_LOG_ERROR, "Error decoding args"); + rpcsvc_request_seterr (req, GARBAGE_ARGS); + goto rpcerr; + } + + nlm4_validate_gluster_fh (&fh, stat, nlm4err); + nlm4_map_fh_to_volume (cs->nfs3state, fh, req, vol, stat, nlm4err); + + if (nlm_grace_period) { + gf_log (GF_NLM, GF_LOG_WARNING, "NLM in grace period"); + stat = nlm4_denied_grace_period; + nlm4_test_reply (cs, stat, NULL); + nfs3_call_state_wipe (cs); + return 0; + } + + cs->vol = vol; + nlm4_volume_started_check (nfs3, vol, ret, rpcerr); + + ret = nfs3_fh_resolve_and_resume (cs, &fh, + NULL, nlm4_test_resume); + +nlm4err: + if (ret < 0) { + gf_log (GF_NLM, GF_LOG_ERROR, "unable to resolve and resume"); + nlm4_test_reply (cs, stat, NULL); + nfs3_call_state_wipe (cs); + return 0; + } + +rpcerr: + if (ret < 0) + nfs3_call_state_wipe (cs); + + return ret; +} + +int +nlm4svc_send_granted_cbk (struct rpc_req *req, struct iovec *iov, int count, + void *myframe) +{ + STACK_DESTROY (((call_frame_t*)myframe)->root); + return 0; +} + +void +nlm4svc_send_granted (nfs3_call_state_t *cs); + +int +nlm_rpcclnt_notify (struct rpc_clnt *rpc_clnt, void *mydata, + rpc_clnt_event_t fn, void *data) +{ + int ret = 0; + char *caller_name = NULL; + nfs3_call_state_t *cs = NULL; + + cs = mydata; + caller_name = cs->args.nlm4_lockargs.alock.caller_name; + + switch (fn) { + case RPC_CLNT_CONNECT: + ret = nlm_set_rpc_clnt (rpc_clnt, caller_name); + if (ret == -1) { + gf_log (GF_NLM, GF_LOG_ERROR, "Failed to set rpc clnt"); + goto err; + } + rpc_clnt_unref (rpc_clnt); + nlm4svc_send_granted (cs); + + break; + + case RPC_CLNT_MSG: + break; + + case RPC_CLNT_DISCONNECT: + nlm_unset_rpc_clnt (rpc_clnt); + break; + } + + err: + return 0; +} + +void * +nlm4_establish_callback (void *csarg) +{ + int ret = -1; + nfs3_call_state_t *cs = NULL; + union gf_sock_union sock_union; + dict_t *options = NULL; + char peerip[INET6_ADDRSTRLEN+1] = {0}; + char *portstr = NULL; + char myip[INET6_ADDRSTRLEN+1] = {0}; + rpc_clnt_t *rpc_clnt = NULL; + int port = -1; + + + cs = (nfs3_call_state_t *) csarg; + glusterfs_this_set (cs->nfsx); + + rpc_transport_get_peeraddr (cs->trans, NULL, 0, &sock_union.storage, + sizeof (sock_union.storage)); + + switch (sock_union.sa.sa_family) { + case AF_INET6: + /* can not come here as NLM listens on IPv4 */ + gf_log (GF_NLM, GF_LOG_ERROR, "NLM is not supported on IPv6 in" + " this release"); + goto err; +/* + inet_ntop (AF_INET6, + &((struct sockaddr_in6 *)sockaddr)->sin6_addr, + peerip, INET6_ADDRSTRLEN+1); + break; +*/ + case AF_INET: + inet_ntop (AF_INET, &sock_union.sin.sin_addr, peerip, + INET6_ADDRSTRLEN+1); + inet_ntop (AF_INET, &(((struct sockaddr_in *)&cs->trans->myinfo.sockaddr)->sin_addr), + myip, INET6_ADDRSTRLEN + 1); + + break; + default: + break; + /* FIXME: handle the error */ + } + + /* looks like libc rpc supports only ipv4 */ + port = pmap_getport (&sock_union.sin, NLM_PROGRAM, + NLM_V4, IPPROTO_TCP); + + if (port == 0) { + gf_log (GF_NLM, GF_LOG_ERROR, "Unable to get NLM port of the " + "client. Is the firewall running on client?"); + goto err; + } + + options = dict_new(); + ret = dict_set_str (options, "transport-type", "socket"); + if (ret == -1) { + gf_log (GF_NLM, GF_LOG_ERROR, "dict_set_str error"); + goto err; + } + + ret = dict_set_dynstr (options, "remote-host", gf_strdup (peerip)); + if (ret == -1) { + gf_log (GF_NLM, GF_LOG_ERROR, "dict_set_str error"); + goto err; + } + + ret = gf_asprintf (&portstr, "%d", port); + if (ret == -1) + goto err; + + ret = dict_set_dynstr (options, "remote-port", + portstr); + if (ret == -1) { + gf_log (GF_NLM, GF_LOG_ERROR, "dict_set_dynstr error"); + goto err; + } + + /* needed in case virtual IP is used */ + ret = dict_set_dynstr (options, "transport.socket.source-addr", + gf_strdup (myip)); + if (ret == -1) + goto err; + + ret = dict_set_str (options, "auth-null", "on"); + if (ret == -1) { + gf_log (GF_NLM, GF_LOG_ERROR, "dict_set_dynstr error"); + goto err; + } + + /* TODO: is 32 frames in transit enough ? */ + rpc_clnt = rpc_clnt_new (options, cs->nfsx->ctx, "NLM-client", 32); + if (rpc_clnt == NULL) { + gf_log (GF_NLM, GF_LOG_ERROR, "rpc_clnt NULL"); + goto err; + } + + ret = rpc_clnt_register_notify (rpc_clnt, nlm_rpcclnt_notify, cs); + if (ret == -1) { + gf_log (GF_NLM, GF_LOG_ERROR,"rpc_clnt_register_connect error"); + goto err; + } + + /* After this connect succeeds, granted msg is sent in notify */ + ret = rpc_transport_connect (rpc_clnt->conn.trans, port); + + if (ret == -1 && EINPROGRESS == errno) + ret = 0; + +err: + if (ret == -1 && rpc_clnt) { + rpc_clnt_unref (rpc_clnt); + } + + return rpc_clnt; +} + +void +nlm4svc_send_granted (nfs3_call_state_t *cs) +{ + int ret = -1; + rpc_clnt_t *rpc_clnt = NULL; + struct iovec outmsg = {0, }; + nlm4_testargs testargs; + struct iobuf *iobuf = NULL; + struct iobref *iobref = NULL; + char peerip[INET6_ADDRSTRLEN+1]; + union gf_sock_union sock_union; + + rpc_clnt = nlm_get_rpc_clnt (cs->args.nlm4_lockargs.alock.caller_name); + if (rpc_clnt == NULL) { + nlm4_establish_callback ((void*)cs); + return; + } + + rpc_transport_get_peeraddr (cs->trans, NULL, 0, &sock_union.storage, + sizeof (sock_union.storage)); + + switch (sock_union.sa.sa_family) { + case AF_INET6: + inet_ntop (AF_INET6, &sock_union.sin6.sin6_addr, peerip, + INET6_ADDRSTRLEN+1); + break; + case AF_INET: + inet_ntop (AF_INET, &sock_union.sin.sin_addr, peerip, + INET6_ADDRSTRLEN+1); + break; + default: + break; + } + + testargs.cookie = cs->args.nlm4_lockargs.cookie; + testargs.exclusive = cs->args.nlm4_lockargs.exclusive; + testargs.alock = cs->args.nlm4_lockargs.alock; + + iobuf = iobuf_get (cs->nfs3state->iobpool); + if (!iobuf) { + gf_log (GF_NLM, GF_LOG_ERROR, "Failed to get iobuf"); + goto ret; + } + + iobuf_to_iovec (iobuf, &outmsg); + /* Use the given serializer to translate the give C structure in arg + * to XDR format which will be written into the buffer in outmsg. + */ + outmsg.iov_len = xdr_serialize_nlm4_testargs (outmsg, &testargs); + + iobref = iobref_new (); + if (iobref == NULL) { + gf_log (GF_NLM, GF_LOG_ERROR, "Failed to get iobref"); + goto ret; + } + + ret = iobref_add (iobref, iobuf); + if (ret) { + gf_log (GF_NLM, GF_LOG_ERROR, "Failed to add iob to iobref"); + goto ret; + } + + ret = rpc_clnt_submit (rpc_clnt, &nlm4clntprog, NLM4_GRANTED, + nlm4svc_send_granted_cbk, &outmsg, 1, + NULL, 0, iobref, cs->frame, NULL, 0, + NULL, 0, NULL); + + if (ret < 0) { + gf_log (GF_NLM, GF_LOG_ERROR, "rpc_clnt_submit error"); + goto ret; + } +ret: + if (iobref) + iobref_unref (iobref); + if (iobuf) + iobuf_unref (iobuf); + + rpc_clnt_unref (rpc_clnt); + nfs3_call_state_wipe (cs); + return; +} + +int +nlm_cleanup_fds (char *caller_name) +{ + int nlmclnt_found = 0; + nlm_fde_t *fde = NULL, *tmp = NULL; + nlm_client_t *nlmclnt = NULL; + + LOCK (&nlm_client_list_lk); + list_for_each_entry (nlmclnt, + &nlm_client_list, nlm_clients) { + if (!strcmp(caller_name, nlmclnt->caller_name)) { + nlmclnt_found = 1; + break; + } + } + + if (!nlmclnt_found) + goto ret; + + if (list_empty (&nlmclnt->fdes)) + goto ret; + + list_for_each_entry_safe (fde, tmp, &nlmclnt->fdes, fde_list) { + fd_unref (fde->fd); + list_del (&fde->fde_list); + GF_FREE (fde); + } + +ret: + UNLOCK (&nlm_client_list_lk); + return 0; +} + +void +nlm_search_and_delete (fd_t *fd, char *caller_name) +{ + nlm_fde_t *fde = NULL; + nlm_client_t *nlmclnt = NULL; + int nlmclnt_found = 0; + int fde_found = 0; + int transit_cnt = 0; + + LOCK (&nlm_client_list_lk); + list_for_each_entry (nlmclnt, + &nlm_client_list, nlm_clients) { + if (!strcmp(caller_name, nlmclnt->caller_name)) { + nlmclnt_found = 1; + break; + } + } + + if (!nlmclnt_found) + goto ret; + + list_for_each_entry (fde, &nlmclnt->fdes, fde_list) { + if (fde->fd == fd) { + fde_found = 1; + break; + } + } + + if (!fde_found) + goto ret; + transit_cnt = fde->transit_cnt; + if (transit_cnt) + goto ret; + list_del (&fde->fde_list); + +ret: + UNLOCK (&nlm_client_list_lk); + + if (fde_found && !transit_cnt) { + fd_unref (fde->fd); + GF_FREE (fde); + } + return; +} + +int +nlm_dec_transit_count (fd_t *fd, char *caller_name) +{ + nlm_fde_t *fde = NULL; + nlm_client_t *nlmclnt = NULL; + int nlmclnt_found = 0; + int fde_found = 0; + int transit_cnt = -1; + + LOCK (&nlm_client_list_lk); + list_for_each_entry (nlmclnt, + &nlm_client_list, nlm_clients) { + if (!strcmp(caller_name, nlmclnt->caller_name)) { + nlmclnt_found = 1; + break; + } + } + + if (!nlmclnt_found) { + gf_log (GF_NLM, GF_LOG_ERROR, "nlmclnt not found"); + nlmclnt = NULL; + goto ret; + } + + list_for_each_entry (fde, &nlmclnt->fdes, fde_list) { + if (fde->fd == fd) { + fde_found = 1; + break; + } + } + + if (fde_found) { + transit_cnt = --fde->transit_cnt; + goto ret; + } +ret: + + UNLOCK (&nlm_client_list_lk); + return transit_cnt; +} + + +nlm_client_t * +nlm_search_and_add (fd_t *fd, char *caller_name) +{ + nlm_fde_t *fde = NULL; + nlm_client_t *nlmclnt = NULL; + int nlmclnt_found = 0; + int fde_found = 0; + + LOCK (&nlm_client_list_lk); + list_for_each_entry (nlmclnt, + &nlm_client_list, nlm_clients) { + if (!strcmp(caller_name, nlmclnt->caller_name)) { + nlmclnt_found = 1; + break; + } + } + + if (!nlmclnt_found) { + gf_log (GF_NLM, GF_LOG_ERROR, "nlmclnt not found"); + nlmclnt = NULL; + goto ret; + } + + list_for_each_entry (fde, &nlmclnt->fdes, fde_list) { + if (fde->fd == fd) { + fde_found = 1; + break; + } + } + + if (fde_found) + goto ret; + + fde = GF_CALLOC (1, sizeof (*fde), gf_nfs_mt_nlm4_fde); + + fde->fd = fd_ref (fd); + list_add (&fde->fde_list, &nlmclnt->fdes); +ret: + if (nlmclnt_found && fde) + fde->transit_cnt++; + UNLOCK (&nlm_client_list_lk); + return nlmclnt; +} + +int +nlm4svc_lock_cbk (call_frame_t *frame, void *cookie, xlator_t *this, + int32_t op_ret, int32_t op_errno, struct gf_flock *flock, + dict_t *xdata) +{ + nlm4_stats stat = nlm4_denied; + int transit_cnt = -1; + char *caller_name = NULL; + nfs3_call_state_t *cs = NULL; + pthread_t thr; + + cs = frame->local; + caller_name = cs->args.nlm4_lockargs.alock.caller_name; + transit_cnt = nlm_dec_transit_count (cs->fd, caller_name); + + if (op_ret == -1) { + if (transit_cnt == 0) + nlm_search_and_delete (cs->fd, caller_name); + stat = nlm4_errno_to_nlm4stat (op_errno); + goto err; + } else { + stat = nlm4_granted; + if (cs->monitor && !nlm_monitor (caller_name)) { + /* FIXME: handle nsm_monitor failure */ + pthread_create (&thr, NULL, nsm_monitor, (void*)caller_name); + } + } + +err: + if (cs->args.nlm4_lockargs.block) { + cs->frame = copy_frame (frame); + frame->local = NULL; + nlm4svc_send_granted (cs); + } else { + nlm4_generic_reply (cs->req, cs->args.nlm4_lockargs.cookie, + stat); + nfs3_call_state_wipe (cs); + } + return 0; +} + +int +nlm4_lock_fd_resume (void *carg) +{ + nlm4_stats stat = nlm4_denied; + int ret = -EFAULT; + nfs_user_t nfu = {0, }; + nfs3_call_state_t *cs = NULL; + struct gf_flock flock = {0, }; + + if (!carg) + return ret; + + cs = (nfs3_call_state_t *)carg; + nlm4_check_fh_resolve_status (cs, stat, nlm4err); + (void) nlm_search_and_add (cs->fd, + cs->args.nlm4_lockargs.alock.caller_name); + nfs_request_user_init (&nfu, cs->req); + nlm4_lock_to_gf_flock (&flock, &cs->args.nlm4_lockargs.alock, + cs->args.nlm4_lockargs.exclusive); + nlm_copy_lkowner (&nfu.lk_owner, &cs->args.nlm4_lockargs.alock.oh); + if (cs->args.nlm4_lockargs.block) { + nlm4_generic_reply (cs->req, cs->args.nlm4_lockargs.cookie, + nlm4_blocked); + ret = nfs_lk (cs->nfsx, cs->vol, &nfu, cs->fd, F_SETLKW, + &flock, nlm4svc_lock_cbk, cs); + /* FIXME: handle error from nfs_lk() specially by just + * cleaning up cs and unblock the client lock request. + */ + ret = 0; + } else + ret = nfs_lk (cs->nfsx, cs->vol, &nfu, cs->fd, F_SETLK, + &flock, nlm4svc_lock_cbk, cs); + +nlm4err: + if (ret < 0) { + stat = nlm4_errno_to_nlm4stat (-ret); + gf_log (GF_NLM, GF_LOG_ERROR, "unable to call lk()"); + nlm4_generic_reply (cs->req, cs->args.nlm4_lockargs.cookie, + stat); + nfs3_call_state_wipe (cs); + } + + return ret; +} + + +int +nlm4_lock_resume (void *carg) +{ + nlm4_stats stat = nlm4_failed; + int ret = -1; + nfs3_call_state_t *cs = NULL; + + if (!carg) + return ret; + + cs = (nfs3_call_state_t *)carg; + nlm4_check_fh_resolve_status (cs, stat, nlm4err); + ret = nlm4_file_open_and_resume (cs, nlm4_lock_fd_resume); + +nlm4err: + if (ret < 0) { + gf_log (GF_NLM, GF_LOG_ERROR, "unable to open and resume"); + stat = nlm4_errno_to_nlm4stat (-ret); + nlm4_generic_reply (cs->req, cs->args.nlm4_lockargs.cookie, + stat); + nfs3_call_state_wipe (cs); + } + + return ret; +} + +int +nlm4svc_lock_common (rpcsvc_request_t *req, int mon) +{ + int ret = RPCSVC_ACTOR_ERROR; + nlm4_stats stat = nlm4_failed; + struct nfs3_fh fh = {{0}, }; + xlator_t *vol = NULL; + nfs3_state_t *nfs3 = NULL; + nfs3_call_state_t *cs = NULL; + struct nfs_state *nfs = NULL; + + if (!req) + return ret; + + nlm4_validate_nfs3_state (req, nfs3, stat, rpcerr, ret); + nfs = nfs_state (nfs3->nfsx); + nlm4_handle_call_state_init (nfs->nfs3state, cs, req, + stat, rpcerr); + + nlm4_prep_nlm4_lockargs (&cs->args.nlm4_lockargs, &cs->lockfh, + &cs->lkowner, cs->cookiebytes); + if (xdr_to_nlm4_lockargs(req->msg[0], &cs->args.nlm4_lockargs) <= 0) { + gf_log (GF_NLM, GF_LOG_ERROR, "Error decoding args"); + rpcsvc_request_seterr (req, GARBAGE_ARGS); + goto rpcerr; + } + + fh = cs->lockfh; + cs->monitor = mon; + nlm4_validate_gluster_fh (&fh, stat, nlm4err); + nlm4_map_fh_to_volume (cs->nfs3state, fh, req, vol, stat, nlm4err); + + if (nlm_grace_period && !cs->args.nlm4_lockargs.reclaim) { + gf_log (GF_NLM, GF_LOG_WARNING, "NLM in grace period"); + stat = nlm4_denied_grace_period; + nlm4_generic_reply (req, cs->args.nlm4_unlockargs.cookie, stat); + nfs3_call_state_wipe (cs); + return 0; + } + + cs->vol = vol; + cs->trans = rpcsvc_request_transport_ref(req); + nlm4_volume_started_check (nfs3, vol, ret, rpcerr); + + ret = nlm_add_nlmclnt (cs->args.nlm4_lockargs.alock.caller_name); + + ret = nfs3_fh_resolve_and_resume (cs, &fh, + NULL, nlm4_lock_resume); + +nlm4err: + if (ret < 0) { + gf_log (GF_NLM, GF_LOG_ERROR, "unable to resolve and resume"); + nlm4_generic_reply (cs->req, cs->args.nlm4_lockargs.cookie, + stat); + nfs3_call_state_wipe (cs); + return 0; + } + +rpcerr: + if (ret < 0) { + nfs3_call_state_wipe (cs); + } + + return ret; +} + +int +nlm4svc_lock (rpcsvc_request_t *req) +{ + return nlm4svc_lock_common (req, 1); +} + +int +nlm4svc_nm_lock (rpcsvc_request_t *req) +{ + return nlm4svc_lock_common (req, 0); +} + +int +nlm4svc_cancel_cbk (call_frame_t *frame, void *cookie, xlator_t *this, + int32_t op_ret, int32_t op_errno, struct gf_flock *flock, + dict_t *xdata) +{ + nlm4_stats stat = nlm4_denied; + nfs3_call_state_t *cs = NULL; + + cs = frame->local; + if (op_ret == -1) { + stat = nlm4_errno_to_nlm4stat (op_errno); + goto err; + } else + stat = nlm4_granted; + +err: + nlm4_generic_reply (cs->req, cs->args.nlm4_cancargs.cookie, + stat); + nfs3_call_state_wipe (cs); + return 0; +} + +int +nlm4_cancel_fd_resume (void *carg) +{ + int ret = -EFAULT; + nfs_user_t nfu = {0, }; + nfs3_call_state_t *cs = NULL; + struct gf_flock flock = {0, }; + + if (!carg) + return ret; + + cs = (nfs3_call_state_t *)carg; + nfs_request_user_init (&nfu, cs->req); + nlm4_lock_to_gf_flock (&flock, &cs->args.nlm4_cancargs.alock, + cs->args.nlm4_cancargs.exclusive); + nlm_copy_lkowner (&nfu.lk_owner, &cs->args.nlm4_cancargs.alock.oh); + flock.l_type = F_UNLCK; + ret = nfs_lk (cs->nfsx, cs->vol, &nfu, cs->fd, F_SETLK, + &flock, nlm4svc_cancel_cbk, cs); + + return ret; +} + +int +nlm4_cancel_resume (void *carg) +{ + nlm4_stats stat = nlm4_failed; + int ret = -EFAULT; + nfs3_call_state_t *cs = NULL; + nlm_client_t *nlmclnt = NULL; + + if (!carg) + return ret; + + cs = (nfs3_call_state_t *)carg; + nlm4_check_fh_resolve_status (cs, stat, nlm4err); + + nlmclnt = nlm_get_uniq (cs->args.nlm4_cancargs.alock.caller_name); + if (nlmclnt == NULL) { + gf_log (GF_NLM, GF_LOG_ERROR, "nlm_get_uniq() returned NULL"); + goto nlm4err; + } + cs->fd = fd_lookup_uint64 (cs->resolvedloc.inode, (uint64_t)nlmclnt); + if (cs->fd == NULL) { + gf_log (GF_NLM, GF_LOG_ERROR, "fd_lookup_uint64 retrned NULL"); + goto nlm4err; + } + ret = nlm4_cancel_fd_resume (cs); + +nlm4err: + if (ret < 0) { + gf_log (GF_NLM, GF_LOG_WARNING, "unable to unlock_fd_resume()"); + stat = nlm4_errno_to_nlm4stat (-ret); + nlm4_generic_reply (cs->req, cs->args.nlm4_cancargs.cookie, + stat); + + nfs3_call_state_wipe (cs); + } + /* clean up is taken care of */ + return 0; +} + +int +nlm4svc_cancel (rpcsvc_request_t *req) +{ + xlator_t *vol = NULL; + nlm4_stats stat = nlm4_failed; + struct nfs_state *nfs = NULL; + nfs3_state_t *nfs3 = NULL; + nfs3_call_state_t *cs = NULL; + int ret = RPCSVC_ACTOR_ERROR; + struct nfs3_fh fh = {{0}, }; + + if (!req) + return ret; + + nlm4_validate_nfs3_state (req, nfs3, stat, rpcerr, ret); + nfs = nfs_state (nfs3->nfsx); + nlm4_handle_call_state_init (nfs->nfs3state, cs, req, + stat, rpcerr); + + nlm4_prep_nlm4_cancargs (&cs->args.nlm4_cancargs, &fh, &cs->lkowner, + cs->cookiebytes); + if (xdr_to_nlm4_cancelargs(req->msg[0], &cs->args.nlm4_cancargs) <= 0) { + gf_log (GF_NLM, GF_LOG_ERROR, "Error decoding args"); + rpcsvc_request_seterr (req, GARBAGE_ARGS); + goto rpcerr; + } + + nlm4_validate_gluster_fh (&fh, stat, nlm4err); + nlm4_map_fh_to_volume (cs->nfs3state, fh, req, vol, stat, nlm4err); + + if (nlm_grace_period) { + gf_log (GF_NLM, GF_LOG_WARNING, "NLM in grace period"); + stat = nlm4_denied_grace_period; + nlm4_generic_reply (req, cs->args.nlm4_unlockargs.cookie, stat); + nfs3_call_state_wipe (cs); + return 0; + } + + cs->vol = vol; + nlm4_volume_started_check (nfs3, vol, ret, rpcerr); + + ret = nfs3_fh_resolve_and_resume (cs, &fh, + NULL, nlm4_cancel_resume); + +nlm4err: + if (ret < 0) { + gf_log (GF_NLM, GF_LOG_ERROR, "unable to resolve and resume"); + nlm4_generic_reply (cs->req, cs->args.nlm4_cancargs.cookie, + stat); + nfs3_call_state_wipe (cs); + return 0; + } + +rpcerr: + if (ret < 0) { + nfs3_call_state_wipe (cs); + } + return ret; +} + +int +nlm4svc_unlock_cbk (call_frame_t *frame, void *cookie, xlator_t *this, + int32_t op_ret, int32_t op_errno, struct gf_flock *flock, + dict_t *xdata) +{ + nlm4_stats stat = nlm4_denied; + nfs3_call_state_t *cs = NULL; + + cs = frame->local; + if (op_ret == -1) { + stat = nlm4_errno_to_nlm4stat (op_errno); + goto err; + } else { + stat = nlm4_granted; + if (flock->l_type == F_UNLCK) + nlm_search_and_delete (cs->fd, + cs->args.nlm4_unlockargs.alock.caller_name); + } + +err: + nlm4_generic_reply (cs->req, cs->args.nlm4_unlockargs.cookie, stat); + nfs3_call_state_wipe (cs); + return 0; +} + +int +nlm4_unlock_fd_resume (void *carg) +{ + int ret = -EFAULT; + nfs_user_t nfu = {0, }; + nfs3_call_state_t *cs = NULL; + struct gf_flock flock = {0, }; + + if (!carg) + return ret; + cs = (nfs3_call_state_t *)carg; + nfs_request_user_init (&nfu, cs->req); + nlm4_lock_to_gf_flock (&flock, &cs->args.nlm4_unlockargs.alock, 0); + nlm_copy_lkowner (&nfu.lk_owner, &cs->args.nlm4_unlockargs.alock.oh); + flock.l_type = F_UNLCK; + ret = nfs_lk (cs->nfsx, cs->vol, &nfu, cs->fd, F_SETLK, + &flock, nlm4svc_unlock_cbk, cs); + + return ret; +} + +int +nlm4_unlock_resume (void *carg) +{ + nlm4_stats stat = nlm4_failed; + int ret = -1; + nfs3_call_state_t *cs = NULL; + nlm_client_t *nlmclnt = NULL; + + if (!carg) + return ret; + + cs = (nfs3_call_state_t *)carg; + nlm4_check_fh_resolve_status (cs, stat, nlm4err); + + nlmclnt = nlm_get_uniq (cs->args.nlm4_unlockargs.alock.caller_name); + if (nlmclnt == NULL) { + stat = nlm4_granted; + gf_log (GF_NLM, GF_LOG_WARNING, "nlm_get_uniq() returned NULL"); + goto nlm4err; + } + cs->fd = fd_lookup_uint64 (cs->resolvedloc.inode, (uint64_t)nlmclnt); + if (cs->fd == NULL) { + stat = nlm4_granted; + gf_log (GF_NLM, GF_LOG_WARNING, "fd_lookup_uint64() returned " + "NULL"); + goto nlm4err; + } + ret = nlm4_unlock_fd_resume (cs); + +nlm4err: + if (ret < 0) { + gf_log (GF_NLM, GF_LOG_WARNING, "unable to unlock_fd_resume"); + stat = nlm4_errno_to_nlm4stat (-ret); + nlm4_generic_reply (cs->req, cs->args.nlm4_unlockargs.cookie, + stat); + + nfs3_call_state_wipe (cs); + } + /* we have already taken care of cleanup */ + return 0; +} + +int +nlm4svc_unlock (rpcsvc_request_t *req) +{ + xlator_t *vol = NULL; + nlm4_stats stat = nlm4_failed; + struct nfs_state *nfs = NULL; + nfs3_state_t *nfs3 = NULL; + nfs3_call_state_t *cs = NULL; + int ret = RPCSVC_ACTOR_ERROR; + struct nfs3_fh fh = {{0}, }; + + if (!req) + return ret; + + nlm4_validate_nfs3_state (req, nfs3, stat, rpcerr, ret); + nfs = nfs_state (nfs3->nfsx); + nlm4_handle_call_state_init (nfs->nfs3state, cs, req, + stat, rpcerr); + + nlm4_prep_nlm4_unlockargs (&cs->args.nlm4_unlockargs, &fh, &cs->lkowner, + cs->cookiebytes); + if (xdr_to_nlm4_unlockargs(req->msg[0], &cs->args.nlm4_unlockargs) <= 0) + { + gf_log (GF_NLM, GF_LOG_ERROR, "Error decoding args"); + rpcsvc_request_seterr (req, GARBAGE_ARGS); + goto rpcerr; + } + + nlm4_validate_gluster_fh (&fh, stat, nlm4err); + nlm4_map_fh_to_volume (cs->nfs3state, fh, req, vol, stat, nlm4err); + + if (nlm_grace_period) { + gf_log (GF_NLM, GF_LOG_WARNING, "NLM in grace period"); + stat = nlm4_denied_grace_period; + nlm4_generic_reply (req, cs->args.nlm4_unlockargs.cookie, stat); + nfs3_call_state_wipe (cs); + return 0; + } + + cs->vol = vol; + /* FIXME: check if trans is being used at all for unlock */ + cs->trans = rpcsvc_request_transport_ref(req); + nlm4_volume_started_check (nfs3, vol, ret, rpcerr); + + ret = nfs3_fh_resolve_and_resume (cs, &fh, + NULL, nlm4_unlock_resume); + +nlm4err: + if (ret < 0) { + gf_log (GF_NLM, GF_LOG_ERROR, "unable to resolve and resume"); + nlm4_generic_reply (req, cs->args.nlm4_unlockargs.cookie, stat); + nfs3_call_state_wipe (cs); + return 0; + } + +rpcerr: + if (ret < 0) { + nfs3_call_state_wipe (cs); + } + return ret; +} + +int +nlm4_share_reply (nfs3_call_state_t *cs, nlm4_stats stat) +{ + nlm4_shareres res = {{0}, 0, 0}; + + if (!cs) + return -1; + + res.cookie = cs->args.nlm4_shareargs.cookie; + res.stat = stat; + res.sequence = 0; + + nlm4svc_submit_reply (cs->req, (void *)&res, + (nlm4_serializer)xdr_serialize_nlm4_shareres); + return 0; +} + +nlm_share_t * +nlm4_share_new () +{ + nlm_share_t *share = NULL; + + share = GF_CALLOC (1, sizeof (nlm_share_t), + gf_nfs_mt_nlm4_share); + if (!share) + goto out; + + INIT_LIST_HEAD (&share->client_list); + INIT_LIST_HEAD (&share->inode_list); + out: + return share; +} + +int +nlm4_add_share_to_inode (nlm_share_t *share) +{ + int ret = -1; + uint64_t ctx = 0; + struct list_head *head = NULL; + xlator_t *this = NULL; + inode_t *inode = NULL; + struct nfs_inode_ctx *ictx = NULL; + struct nfs_state *priv = NULL; + + this = THIS; + priv = this->private; + inode = share->inode; + ret = inode_ctx_get (inode, this, &ctx); + + if (ret == -1) { + ictx = GF_CALLOC (1, sizeof (struct nfs_inode_ctx), + gf_nfs_mt_inode_ctx); + if (!ictx ) { + gf_log (this->name, GF_LOG_ERROR, + "could not allocate nfs inode ctx"); + ret = -1; + goto out; + } + ictx->generation = priv->generation; + + head = &ictx->shares; + INIT_LIST_HEAD (head); + + ret = inode_ctx_put (inode, this, (uint64_t)ictx); + if (ret) { + gf_log (this->name, GF_LOG_ERROR, + "could not store share list"); + goto out; + } + } + else { + ictx = (struct nfs_inode_ctx *)ctx; + head = &ictx->shares; + } + + list_add (&share->inode_list, head); + + out: + if (ret && head) + GF_FREE (head); + + return ret; +} + +int +nlm4_approve_share_reservation (nfs3_call_state_t *cs) +{ + int ret = -1; + uint64_t ctx = 0; + fsh_mode req_mode = 0; + fsh_access req_access = 0; + inode_t *inode = NULL; + nlm_share_t *share = NULL; + struct list_head *head = NULL; + struct nfs_inode_ctx *ictx = NULL; + + if (!cs) + goto out; + + inode = cs->resolvedloc.inode; + + ret = inode_ctx_get (inode, THIS, &ctx); + if (ret) { + ret = 0; + goto out; + } + ictx = (struct nfs_inode_ctx *)ctx; + + head = &ictx->shares; + if (!head || list_empty (head)) + goto out; + + req_mode = cs->args.nlm4_shareargs.share.mode; + req_access = cs->args.nlm4_shareargs.share.access; + + list_for_each_entry (share, head, inode_list) { + ret = (((req_mode & share->access) == 0) && + ((req_access & share->mode) == 0)); + if (!ret) { + ret = -1; + goto out; + } + } + ret = 0; + + out: + return ret; +} + +int +nlm4_create_share_reservation (nfs3_call_state_t *cs) +{ + int ret = -1; + nlm_share_t *share = NULL; + nlm_client_t *client = NULL; + inode_t *inode = NULL; + + LOCK (&nlm_client_list_lk); + + inode = inode_ref (cs->resolvedloc.inode); + if (!inode) { + gf_log (GF_NLM, GF_LOG_ERROR, "inode not found"); + goto out; + } + + client = __nlm_get_uniq (cs->args.nlm4_shareargs.share.caller_name); + if (!client) { + /* DO NOT add client. the client is supposed + to be here, since nlm4svc_share adds it */ + gf_log (GF_NLM, GF_LOG_ERROR, "client not found"); + goto out; + } + + ret = nlm4_approve_share_reservation (cs); + if (ret) + goto out; + + share = nlm4_share_new (); + if (!share) { + ret = -1; + goto out; + } + + share->inode = inode; + share->mode = cs->args.nlm4_shareargs.share.mode; + share->access = cs->args.nlm4_shareargs.share.access; + nlm_copy_lkowner (&share->lkowner, + &cs->args.nlm4_shareargs.share.oh); + + ret = nlm4_add_share_to_inode (share); + if (ret) + goto out; + + list_add (&share->client_list, &client->shares); + + out: + if (ret && inode) { + inode_unref (inode); + GF_FREE (share); + } + + UNLOCK (&nlm_client_list_lk); + return ret; +} + +/* + SHARE and UNSHARE calls DO NOT perform STACK_WIND, + the (non-monitored) share reservations are maintained + at *nfs xlator level only*, in memory +*/ +int +nlm4_share_resume (void *call_state) +{ + int ret = -1; + nlm4_stats stat = nlm4_failed; + nfs3_call_state_t *cs = NULL; + + if (!call_state) + return ret; + + cs = (nfs3_call_state_t *)call_state; + nlm4_check_fh_resolve_status (cs, stat, out); + + ret = nlm4_create_share_reservation (cs); + if (!ret) + stat = nlm4_granted; + + out: + nlm4_share_reply (cs, stat); + nfs3_call_state_wipe (cs); + return 0; +} + +int +nlm4svc_share (rpcsvc_request_t *req) +{ + nlm4_stats stat = nlm4_failed; + xlator_t *vol = NULL; + nfs3_state_t *nfs3 = NULL; + nfs3_call_state_t *cs = NULL; + struct nfs_state *nfs = NULL; + struct nfs3_fh fh = {{0}, }; + int ret = RPCSVC_ACTOR_ERROR; + + if (!req) + return ret; + + nlm4_validate_nfs3_state (req, nfs3, stat, rpcerr, ret); + nfs = nfs_state (nfs3->nfsx); + nlm4_handle_call_state_init (nfs->nfs3state, cs, req, + stat, rpcerr); + + nlm4_prep_shareargs (&cs->args.nlm4_shareargs, &cs->lockfh, + &cs->lkowner, cs->cookiebytes); + + if (xdr_to_nlm4_shareargs (req->msg[0], + &cs->args.nlm4_shareargs) <= 0) { + gf_log (GF_NLM, GF_LOG_ERROR, "Error decoding SHARE args"); + rpcsvc_request_seterr (req, GARBAGE_ARGS); + goto rpcerr; + } + + fh = cs->lockfh; + nlm4_validate_gluster_fh (&fh, stat, nlm4err); + nlm4_map_fh_to_volume (cs->nfs3state, fh, req, + vol, stat, nlm4err); + + if (nlm_grace_period && !cs->args.nlm4_shareargs.reclaim) { + gf_log (GF_NLM, GF_LOG_DEBUG, "NLM in grace period"); + stat = nlm4_denied_grace_period; + nlm4_share_reply (cs, stat); + nfs3_call_state_wipe (cs); + return 0; + } + + cs->vol = vol; + cs->trans = rpcsvc_request_transport_ref(req); + nlm4_volume_started_check (nfs3, vol, ret, rpcerr); + + ret = nlm_add_nlmclnt (cs->args.nlm4_shareargs.share.caller_name); + + ret = nfs3_fh_resolve_and_resume (cs, &fh, NULL, nlm4_share_resume); + + nlm4err: + if (ret < 0) { + gf_log (GF_NLM, GF_LOG_ERROR, "SHARE call failed"); + nlm4_share_reply (cs, stat); + nfs3_call_state_wipe (cs); + return 0; + } + + rpcerr: + if (ret < 0) + nfs3_call_state_wipe (cs); + + return ret; +} + +int +nlm4_remove_share_reservation (nfs3_call_state_t *cs) +{ + int ret = -1; + uint64_t ctx = 0; + fsh_mode req_mode = 0; + fsh_access req_access = 0; + nlm_share_t *share = NULL; + nlm_share_t *tmp = NULL; + nlm_client_t *client = NULL; + char *caller = NULL; + inode_t *inode = NULL; + xlator_t *this = NULL; + struct list_head *head = NULL; + nlm4_shareargs *args = NULL; + struct nfs_inode_ctx *ictx = NULL; + + LOCK (&nlm_client_list_lk); + + args = &cs->args.nlm4_shareargs; + caller = args->share.caller_name; + + client = __nlm_get_uniq (caller); + if (!client) { + gf_log (GF_NLM, GF_LOG_ERROR, + "client not found: %s", caller); + goto out; + } + + inode = cs->resolvedloc.inode; + if (!inode) { + gf_log (GF_NLM, GF_LOG_ERROR, + "inode not found: client: %s", caller); + goto out; + } + + this = THIS; + ret = inode_ctx_get (inode, this, &ctx); + if (ret) { + gf_log (GF_NLM, GF_LOG_ERROR, + "no shares found for inode:" + "gfid: %s; client: %s", + inode->gfid, caller); + goto out; + } + ictx = (struct nfs_inode_ctx *)ctx; + + head = &ictx->shares; + if (list_empty (head)) { + ret = -1; + goto out; + } + + ret = 0; + req_mode = args->share.mode; + req_access = args->share.access; + + list_for_each_entry_safe (share, tmp, head, inode_list) { + ret = ((req_mode == share->mode) && + (req_access == share->access) && + nlm_is_oh_same_lkowner (&share->lkowner, &args->share.oh)); + if (ret) { + list_del (&share->client_list); + list_del (&share->inode_list); + inode_unref (share->inode); + GF_FREE (share); + break; + } + } + + ret = 0; + out: + UNLOCK (&nlm_client_list_lk); + return ret; + +} + +int +nlm4_unshare_resume (void *call_state) +{ + int ret = -1; + nlm4_stats stat = nlm4_failed; + nfs3_call_state_t *cs = NULL; + + if (!call_state) + return ret; + + cs = (nfs3_call_state_t *)call_state; + + nlm4_check_fh_resolve_status (cs, stat, out); + ret = nlm4_remove_share_reservation (cs); + if (!ret) + stat = nlm4_granted; + + out: + nlm4_share_reply (cs, stat); + nfs3_call_state_wipe (cs); + return 0; +} + +int +nlm4svc_unshare (rpcsvc_request_t *req) +{ + nlm4_stats stat = nlm4_failed; + xlator_t *vol = NULL; + nfs3_state_t *nfs3 = NULL; + nfs3_call_state_t *cs = NULL; + struct nfs_state *nfs = NULL; + struct nfs3_fh fh = {{0}, }; + int ret = RPCSVC_ACTOR_ERROR; + + if (!req) + return ret; + + nlm4_validate_nfs3_state (req, nfs3, stat, rpcerr, ret); + nfs = nfs_state (nfs3->nfsx); + nlm4_handle_call_state_init (nfs->nfs3state, cs, req, + stat, rpcerr); + + nlm4_prep_shareargs (&cs->args.nlm4_shareargs, &cs->lockfh, + &cs->lkowner, cs->cookiebytes); + + if (xdr_to_nlm4_shareargs (req->msg[0], + &cs->args.nlm4_shareargs) <= 0) { + gf_log (GF_NLM, GF_LOG_ERROR, "Error decoding UNSHARE args"); + rpcsvc_request_seterr (req, GARBAGE_ARGS); + goto rpcerr; + } + + fh = cs->lockfh; + nlm4_validate_gluster_fh (&fh, stat, nlm4err); + nlm4_map_fh_to_volume (cs->nfs3state, fh, req, + vol, stat, nlm4err); + + if (nlm_grace_period && !cs->args.nlm4_shareargs.reclaim) { + gf_log (GF_NLM, GF_LOG_DEBUG, "NLM in grace period"); + stat = nlm4_denied_grace_period; + nlm4_share_reply (cs, stat); + nfs3_call_state_wipe (cs); + return 0; + } + + cs->vol = vol; + cs->trans = rpcsvc_request_transport_ref(req); + nlm4_volume_started_check (nfs3, vol, ret, rpcerr); + + ret = nfs3_fh_resolve_and_resume (cs, &fh, NULL, + nlm4_unshare_resume); + + nlm4err: + if (ret < 0) { + gf_log (GF_NLM, GF_LOG_ERROR, "UNSHARE call failed"); + nlm4_share_reply (cs, stat); + ret = 0; + return 0; + } + + rpcerr: + if (ret < 0) + nfs3_call_state_wipe (cs); + + return ret; +} + +int +nlm4_free_all_shares (char *caller_name) +{ + nlm_share_t *share = NULL; + nlm_share_t *tmp = NULL; + nlm_client_t *client = NULL; + + LOCK (&nlm_client_list_lk); + + client = __nlm_get_uniq (caller_name); + if (!client) { + gf_log (GF_NLM, GF_LOG_DEBUG, + "client not found: %s", caller_name); + goto out; + } + + list_for_each_entry_safe (share, tmp, &client->shares, client_list) { + list_del (&share->inode_list); + list_del (&share->client_list); + inode_unref (share->inode); + GF_FREE (share); + } + out: + UNLOCK (&nlm_client_list_lk); + return 0; +} + +int +nlm4svc_free_all (rpcsvc_request_t *req) +{ + int ret = RPCSVC_ACTOR_ERROR; + nlm4_stats stat = nlm4_failed; + nfs3_state_t *nfs3 = NULL; + nfs3_call_state_t *cs = NULL; + struct nfs_state *nfs = NULL; + + nlm4_validate_nfs3_state (req, nfs3, stat, err, ret); + nfs = nfs_state (nfs3->nfsx); + nlm4_handle_call_state_init (nfs->nfs3state, cs, + req, stat, err); + + nlm4_prep_freeallargs (&cs->args.nlm4_freeallargs, + &cs->lkowner); + + if (xdr_to_nlm4_freeallargs (req->msg[0], + &cs->args.nlm4_freeallargs) <= 0) { + gf_log (GF_NLM, GF_LOG_ERROR, "Error decoding FREE_ALL args"); + rpcsvc_request_seterr (req, GARBAGE_ARGS); + goto err; + } + + ret = nlm4_free_all_shares (cs->args.nlm4_freeallargs.name); + if (ret) + goto err; + + ret = nlm_cleanup_fds (cs->args.nlm4_freeallargs.name); + if (ret) + goto err; + + err: + nfs3_call_state_wipe (cs); + if (ret) + gf_log (GF_NLM, GF_LOG_DEBUG, + "error in free all; stat: %d", stat); + return ret; + +} + +void +nlm4svc_sm_notify (struct nlm_sm_status *status) +{ + gf_log (GF_NLM, GF_LOG_INFO, "sm_notify: %s, state: %d", + status->mon_name, + status->state); + nlm_cleanup_fds (status->mon_name); +} + +rpcsvc_actor_t nlm4svc_actors[NLM4_PROC_COUNT] = { + /* 0 */ + {"NULL", NLM4_NULL, nlm4svc_null, NULL, 0, DRC_IDEMPOTENT}, + {"TEST", NLM4_TEST, nlm4svc_test, NULL, 0, DRC_IDEMPOTENT}, + {"LOCK", NLM4_LOCK, nlm4svc_lock, NULL, 0, DRC_NON_IDEMPOTENT}, + {"CANCEL", NLM4_CANCEL, nlm4svc_cancel, NULL, 0, DRC_NON_IDEMPOTENT}, + {"UNLOCK", NLM4_UNLOCK, nlm4svc_unlock, NULL, 0, DRC_NON_IDEMPOTENT}, + /* 5 */ + {"GRANTED", NLM4_GRANTED, NULL, NULL, 0, DRC_NA}, + {"TEST", NLM4_TEST_MSG, NULL, NULL, 0, DRC_NA}, + {"LOCK", NLM4_LOCK_MSG, NULL, NULL, 0, DRC_NA}, + {"CANCEL", NLM4_CANCEL_MSG, NULL, NULL, 0, DRC_NA}, + {"UNLOCK", NLM4_UNLOCK_MSG, NULL, NULL, 0, DRC_NA}, + /* 10 */ + {"GRANTED", NLM4_GRANTED_MSG, NULL, NULL, 0, DRC_NA}, + {"TEST", NLM4_TEST_RES, NULL, NULL, 0, DRC_NA}, + {"LOCK", NLM4_LOCK_RES, NULL, NULL, 0, DRC_NA}, + {"CANCEL", NLM4_CANCEL_RES, NULL, NULL, 0, DRC_NA}, + {"UNLOCK", NLM4_UNLOCK_RES, NULL, NULL, 0, DRC_NA}, + /* 15 ; procedures 17,18,19 are not defined by nlm */ + {"GRANTED", NLM4_GRANTED_RES, NULL, NULL, 0, DRC_NA}, + {"SM_NOTIFY", NLM4_SM_NOTIFY, NULL, NULL, 0, DRC_NA}, + {"SEVENTEEN", NLM4_SEVENTEEN, NULL, NULL, 0, DRC_NA}, + {"EIGHTEEN", NLM4_EIGHTEEN, NULL, NULL, 0, DRC_NA}, + {"NINETEEN", NLM4_NINETEEN, NULL, NULL, 0, DRC_NA}, + /* 20 */ + {"SHARE", NLM4_SHARE, nlm4svc_share, NULL, 0, DRC_NON_IDEMPOTENT}, + {"UNSHARE", NLM4_UNSHARE, nlm4svc_unshare, NULL, 0, DRC_NON_IDEMPOTENT}, + {"NM_LOCK", NLM4_NM_LOCK, nlm4svc_nm_lock, NULL, 0, DRC_NON_IDEMPOTENT}, + {"FREE_ALL", NLM4_FREE_ALL, nlm4svc_free_all, NULL, 0, DRC_IDEMPOTENT}, +}; + +rpcsvc_program_t nlm4prog = { + .progname = "NLM4", + .prognum = NLM_PROGRAM, + .progver = NLM_V4, + .progport = GF_NLM4_PORT, + .actors = nlm4svc_actors, + .numactors = NLM4_PROC_COUNT, + .min_auth = AUTH_NULL, +}; + + +int +nlm4_init_state (xlator_t *nfsx) +{ + return 0; +} + +extern void *nsm_thread (void *argv); + +void nlm_grace_period_over(void *arg) +{ + nlm_grace_period = 0; +} + +rpcsvc_program_t * +nlm4svc_init(xlator_t *nfsx) +{ + struct nfs3_state *ns = NULL; + struct nfs_state *nfs = NULL; + dict_t *options = NULL; + int ret = -1; + char *portstr = NULL; + pthread_t thr; + struct timespec timeout = {0,}; + FILE *pidfile = NULL; + pid_t pid = -1; + static gf_boolean_t nlm4_inited = _gf_false; + + /* Already inited */ + if (nlm4_inited) + return &nlm4prog; + + nfs = (struct nfs_state*)nfsx->private; + + ns = nfs->nfs3state; + if (!ns) { + gf_log (GF_NLM, GF_LOG_ERROR, "NLM4 init failed"); + goto err; + } + nlm4prog.private = ns; + + options = dict_new (); + + ret = gf_asprintf (&portstr, "%d", GF_NLM4_PORT); + if (ret == -1) + goto err; + + ret = dict_set_dynstr (options, "transport.socket.listen-port", + portstr); + if (ret == -1) + goto err; + ret = dict_set_str (options, "transport-type", "socket"); + if (ret == -1) { + gf_log (GF_NLM, GF_LOG_ERROR, "dict_set_str error"); + goto err; + } + + if (nfs->allow_insecure) { + ret = dict_set_str (options, "rpc-auth-allow-insecure", "on"); + if (ret == -1) { + gf_log (GF_NLM, GF_LOG_ERROR, "dict_set_str error"); + goto err; + } + ret = dict_set_str (options, "rpc-auth.ports.insecure", "on"); + if (ret == -1) { + gf_log (GF_NLM, GF_LOG_ERROR, "dict_set_str error"); + goto err; + } + } + + ret = dict_set_str (options, "transport.address-family", "inet"); + if (ret == -1) { + gf_log (GF_NLM, GF_LOG_ERROR, "dict_set_str error"); + goto err; + } + + ret = rpcsvc_create_listeners (nfs->rpcsvc, options, "NLM"); + if (ret == -1) { + gf_log (GF_NLM, GF_LOG_ERROR, "Unable to create listeners"); + dict_unref (options); + goto err; + } + INIT_LIST_HEAD(&nlm_client_list); + LOCK_INIT (&nlm_client_list_lk); + + /* unlink sm-notify.pid so that when we restart rpc.statd/sm-notify + * it thinks that the machine has restarted and sends NOTIFY to clients. + */ + ret = unlink ("/var/run/sm-notify.pid"); + if (ret == -1 && errno != ENOENT) { + gf_log (GF_NLM, GF_LOG_ERROR, "unable to unlink sm-notify"); + goto err; + } + /* temporary work around to restart statd, not distro/OS independant. + * Need to figure out a more graceful way + * killall will cause problems on solaris. + */ + + pidfile = fopen ("/var/run/rpc.statd.pid", "r"); + if (pidfile) { + ret = fscanf (pidfile, "%d", &pid); + if (ret <= 0) { + gf_log (GF_NLM, GF_LOG_WARNING, "unable to get pid of " + "rpc.statd"); + ret = runcmd ("killall", "-9", "rpc.statd", NULL); + } else + kill (pid, SIGKILL); + + fclose (pidfile); + } else { + gf_log (GF_NLM, GF_LOG_WARNING, "opening the pid file of " + "rpc.statd failed (%s)", strerror (errno)); + /* if ret == -1, do nothing - case either statd was not + * running or was running in valgrind mode + */ + ret = runcmd ("killall", "-9", "rpc.statd", NULL); + } + + ret = unlink ("/var/run/rpc.statd.pid"); + if (ret == -1 && errno != ENOENT) { + gf_log (GF_NLM, GF_LOG_ERROR, "unable to unlink rpc.statd"); + goto err; + } + + ret = runcmd ("/sbin/rpc.statd", NULL); + if (ret == -1) { + gf_log (GF_NLM, GF_LOG_ERROR, "unable to start rpc.statd"); + goto err; + } + pthread_create (&thr, NULL, nsm_thread, (void*)NULL); + + timeout.tv_sec = nlm_grace_period; + timeout.tv_nsec = 0; + + gf_timer_call_after (nfsx->ctx, timeout, nlm_grace_period_over, NULL); + nlm4_inited = _gf_true; + return &nlm4prog; +err: + return NULL; +} + +int32_t +nlm_priv (xlator_t *this) +{ + int32_t ret = -1; + uint32_t client_count = 0; + uint64_t file_count = 0; + nlm_client_t *client = NULL; + nlm_fde_t *fde = NULL; + char key[GF_DUMP_MAX_BUF_LEN] = {0}; + char gfid_str[64] = {0}; + + gf_proc_dump_add_section("nfs.nlm"); + + if (TRY_LOCK (&nlm_client_list_lk)) + goto out; + + list_for_each_entry (client, &nlm_client_list, nlm_clients) { + + gf_proc_dump_build_key (key, "client", "%d.hostname", client_count); + gf_proc_dump_write (key, "%s\n", client->caller_name); + + file_count = 0; + list_for_each_entry (fde, &client->fdes, fde_list) { + gf_proc_dump_build_key (key, "file", "%ld.gfid", file_count); + memset (gfid_str, 0, 64); + uuid_utoa_r (fde->fd->inode->gfid, gfid_str); + gf_proc_dump_write (key, "%s", gfid_str); + file_count++; + } + + gf_proc_dump_build_key (key, "client", "files-locked"); + gf_proc_dump_write (key, "%ld\n", file_count); + client_count++; + } + + gf_proc_dump_build_key (key, "nlm", "client-count"); + gf_proc_dump_write (key, "%d", client_count); + ret = 0; + UNLOCK (&nlm_client_list_lk); + + out: + if (ret) { + gf_proc_dump_build_key (key, "nlm", "statedump_error"); + gf_proc_dump_write (key, "Unable to dump nlm state because " + "nlm_client_list_lk lock couldn't be acquired"); + } + + return ret; +} diff --git a/xlators/nfs/server/src/nlm4.h b/xlators/nfs/server/src/nlm4.h new file mode 100644 index 000000000..9b5d54081 --- /dev/null +++ b/xlators/nfs/server/src/nlm4.h @@ -0,0 +1,77 @@ +/* + Copyright (c) 2012 Gluster, Inc. <http://www.gluster.com> + This file is part of GlusterFS. + + This file is licensed to you under your choice of the GNU Lesser + General Public License, version 3 or any later version (LGPLv3 or + later), or the GNU General Public License, version 2 (GPLv2), in all + cases as published by the Free Software Foundation. +*/ + +#ifndef _NLM4_H_ +#define _NLM4_H_ + +#ifndef _CONFIG_H +#define _CONFIG_H +#include "config.h" +#endif + +#include <sys/types.h> +#include <signal.h> +#include "rpcsvc.h" +#include "dict.h" +#include "xlator.h" +#include "iobuf.h" +#include "nfs.h" +#include "list.h" +#include "xdr-nfs3.h" +#include "locking.h" +#include "nfs3-fh.h" +#include "uuid.h" +#include "nlm4-xdr.h" +#include "lkowner.h" + +/* Registered with portmap */ +#define GF_NLM4_PORT 38468 +#define GF_NLM GF_NFS"-NLM" + +extern rpcsvc_program_t * +nlm4svc_init (xlator_t *nfsx); + +extern int +nlm4_init_state (xlator_t *nfsx); + +#define NLM_PROGRAM 100021 +#define NLM_V4 4 + +typedef struct nlm4_lwowner { + char temp[1024]; +} nlm4_lkowner_t; + +typedef struct nlm_client { + struct sockaddr_storage sa; + pid_t uniq; + struct list_head nlm_clients; + struct list_head fdes; + struct list_head shares; + struct rpc_clnt *rpc_clnt; + char *caller_name; + int nsm_monitor; +} nlm_client_t; + +typedef struct nlm_share { + struct list_head client_list; + struct list_head inode_list; + gf_lkowner_t lkowner; + inode_t *inode; + fsh_mode mode; + fsh_access access; +} nlm_share_t; + +typedef struct nlm_fde { + struct list_head fde_list; + fd_t *fd; + int transit_cnt; +} nlm_fde_t; + +#endif diff --git a/xlators/nfs/server/src/nlmcbk_svc.c b/xlators/nfs/server/src/nlmcbk_svc.c new file mode 100644 index 000000000..e1b588765 --- /dev/null +++ b/xlators/nfs/server/src/nlmcbk_svc.c @@ -0,0 +1,117 @@ +/* + Copyright (c) 2012 Gluster, Inc. <http://www.gluster.com> + This file is part of GlusterFS. + + This file is licensed to you under your choice of the GNU Lesser + General Public License, version 3 or any later version (LGPLv3 or + later), or the GNU General Public License, version 2 (GPLv2), in all + cases as published by the Free Software Foundation. +*/ + +/* + * Please do not edit this file. + * It was generated using rpcgen. + */ + +#include "nlmcbk-xdr.h" +#include "nlm4.h" +#include "logging.h" +#include <stdio.h> +#include <stdlib.h> +#include <rpc/pmap_clnt.h> +#include <string.h> +#include <memory.h> +#include <sys/socket.h> +#include <netinet/in.h> + +#ifndef SIG_PF +#define SIG_PF void(*)(int) +#endif + +void +nlm4svc_sm_notify (struct nlm_sm_status *status); + +void *nlmcbk_sm_notify_0_svc(struct nlm_sm_status *status, struct svc_req *req) +{ + nlm4svc_sm_notify (status); + return NULL; +} + +static void +nlmcbk_program_0(struct svc_req *rqstp, register SVCXPRT *transp) +{ + union { + struct nlm_sm_status nlmcbk_sm_notify_0_arg; + } argument; + char *result; + xdrproc_t _xdr_argument, _xdr_result; + char *(*local)(char *, struct svc_req *); + + switch (rqstp->rq_proc) { + case NULLPROC: + (void) svc_sendreply (transp, (xdrproc_t) xdr_void, (char *)NULL); + return; + + case NLMCBK_SM_NOTIFY: + _xdr_argument = (xdrproc_t) xdr_nlm_sm_status; + _xdr_result = (xdrproc_t) xdr_void; + local = (char *(*)(char *, struct svc_req *)) nlmcbk_sm_notify_0_svc; + break; + + default: + svcerr_noproc (transp); + return; + } + memset ((char *)&argument, 0, sizeof (argument)); + if (!svc_getargs (transp, (xdrproc_t) _xdr_argument, (caddr_t) &argument)) { + svcerr_decode (transp); + return; + } + result = (*local)((char *)&argument, rqstp); + if (!svc_sendreply(transp, (xdrproc_t) _xdr_result, result)) { + svcerr_systemerr (transp); + } + + if (!svc_freeargs (transp, (xdrproc_t) _xdr_argument, (caddr_t) &argument)) { + gf_log (GF_NLM, GF_LOG_ERROR, "unable to free arguments"); + return; + } + return; +} + +void * +nsm_thread (void *argv) +{ + register SVCXPRT *transp; + int ret = 0; + + ret = pmap_unset (NLMCBK_PROGRAM, NLMCBK_V1); + if (ret == 0) { + gf_log (GF_NLM, GF_LOG_ERROR, "pmap_unset failed"); + return NULL; + } + transp = svcudp_create(RPC_ANYSOCK); + if (transp == NULL) { + gf_log (GF_NLM, GF_LOG_ERROR, "cannot create udp service."); + return NULL; + } + if (!svc_register(transp, NLMCBK_PROGRAM, NLMCBK_V1, nlmcbk_program_0, IPPROTO_UDP)) { + gf_log (GF_NLM, GF_LOG_ERROR, "unable to register (NLMCBK_PROGRAM, NLMCBK_V0, udp)."); + return NULL; + } + + transp = svctcp_create(RPC_ANYSOCK, 0, 0); + if (transp == NULL) { + gf_log (GF_NLM, GF_LOG_ERROR, "cannot create tcp service."); + return NULL; + } + if (!svc_register(transp, NLMCBK_PROGRAM, NLMCBK_V1, nlmcbk_program_0, IPPROTO_TCP)) { + gf_log (GF_NLM, GF_LOG_ERROR, "unable to register (NLMCBK_PROGRAM, NLMCBK_V0, tcp)."); + return NULL; + } + + svc_run (); + gf_log (GF_NLM, GF_LOG_ERROR, "svc_run returned"); + return NULL; + /* NOTREACHED */ +} |