diff options
Diffstat (limited to 'xlators/nfs')
26 files changed, 4606 insertions, 1282 deletions
diff --git a/xlators/nfs/server/src/Makefile.am b/xlators/nfs/server/src/Makefile.am index b5fd8586d..62fbf6535 100644 --- a/xlators/nfs/server/src/Makefile.am +++ b/xlators/nfs/server/src/Makefile.am @@ -1,19 +1,24 @@ xlator_LTLIBRARIES = server.la xlatordir = $(libdir)/glusterfs/$(PACKAGE_VERSION)/xlator/nfs nfsrpclibdir = $(top_srcdir)/rpc/rpc-lib/src -server_la_LDFLAGS = -module -avoidversion +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 + nlmcbk_svc.c mount3udp_svc.c acl3.c server_la_LIBADD = $(top_builddir)/libglusterfs/src/libglusterfs.la noinst_HEADERS = nfs.h nfs-common.h nfs-fops.h nfs-inodes.h nfs-generics.h \ - mount3.h nfs3-fh.h nfs3.h nfs3-helpers.h nfs-mem-types.h nlm4.h + mount3.h nfs3-fh.h nfs3.h nfs3-helpers.h nfs-mem-types.h nlm4.h \ + acl3.h -AM_CFLAGS = -fPIC -D_FILE_OFFSET_BITS=64 -D_GNU_SOURCE -Wall -D$(GF_HOST_OS)\ +AM_CPPFLAGS = $(GF_CPPFLAGS) \ -DLIBDIR=\"$(libdir)/glusterfs/$(PACKAGE_VERSION)/auth\" \ - -I$(top_srcdir)/libglusterfs/src -shared -nostartfiles $(GF_CFLAGS)\ - -I$(nfsrpclibdir) -L$(xlatordir)/ -I$(CONTRIBDIR)/rbtree\ + -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 index cebdf5270..b0824bf10 100644 --- a/xlators/nfs/server/src/mount3.c +++ b/xlators/nfs/server/src/mount3.c @@ -2,19 +2,10 @@ Copyright (c) 2010-2011 Gluster, Inc. <http://www.gluster.com> This file is part of GlusterFS. - GlusterFS is free software; you can redistribute it and/or modify - it under the terms of the GNU General Public License as published - by the Free Software Foundation; either version 3 of the License, - or (at your option) any later version. - - GlusterFS is distributed in the hope that it will be useful, but - WITHOUT ANY WARRANTY; without even the implied warranty of - MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU - General Public License for more details. - - You should have received a copy of the GNU General Public License - along with this program. If not, see - <http://www.gnu.org/licenses/>. + 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 @@ -39,23 +30,71 @@ #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; + 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) @@ -81,7 +120,12 @@ mnt3svc_submit_reply (rpcsvc_request_t *req, void *arg, mnt3_serializer sfunc) /* 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 = sfunc (outmsg, arg); + 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) { @@ -89,12 +133,14 @@ mnt3svc_submit_reply (rpcsvc_request_t *req, void *arg, mnt3_serializer sfunc) goto ret; } - iobref_add (iobref, iob); + 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); - iobuf_unref (iob); - iobref_unref (iobref); if (ret == -1) { gf_log (GF_MNT, GF_LOG_ERROR, "Reply submission failed"); goto ret; @@ -102,6 +148,11 @@ mnt3svc_submit_reply (rpcsvc_request_t *req, void *arg, mnt3_serializer sfunc) ret = 0; ret: + if (NULL != iob) + iobuf_unref (iob); + if (NULL != iobref) + iobref_unref (iobref); + return ret; } @@ -185,13 +236,278 @@ mnt3svc_set_mountres3 (mountstat3 stat, struct nfs3_fh *fh, int *authflavor, 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; @@ -201,21 +517,62 @@ mnt3svc_update_mountlist (struct mount3_state *ms, rpcsvc_request_t *req, if (!me) return -1; - strcpy (me->exname, expname); + 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_err; + 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); @@ -234,6 +591,7 @@ __mnt3_get_volume_id (struct mount3_state *ms, xlator_t *mntxl, 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); @@ -243,6 +601,7 @@ __mnt3_get_volume_id (struct mount3_state *ms, xlator_t *mntxl, } out: + UNLOCK (&ms->mountlock); return ret; } @@ -263,7 +622,7 @@ mnt3svc_lookup_mount_cbk (call_frame_t *frame, void *cookie, rpcsvc_t *svc = NULL; xlator_t *mntxl = NULL; uuid_t volumeid = {0, }; - char fhstr[1024]; + char fhstr[1024], *path = NULL; req = (rpcsvc_request_t *)frame->local; @@ -285,7 +644,15 @@ mnt3svc_lookup_mount_cbk (call_frame_t *frame, void *cookie, if (status != MNT3_OK) goto xmit_res; - mnt3svc_update_mountlist (ms, req, mntxl->name); + 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; @@ -295,7 +662,7 @@ mnt3svc_lookup_mount_cbk (call_frame_t *frame, void *cookie, fh = nfs3_fh_build_uuid_root_fh (volumeid); xmit_res: - nfs3_fh_to_str (&fh, fhstr); + 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) { @@ -316,7 +683,7 @@ int mnt3_match_dirpath_export (char *expname, char *dirpath) { int ret = 0; - int dlen = 0; + size_t dlen; if ((!expname) || (!dirpath)) return 0; @@ -327,7 +694,7 @@ mnt3_match_dirpath_export (char *expname, char *dirpath) * compare. */ dlen = strlen (dirpath); - if (dirpath [dlen - 1] == '/') + if (dlen && dirpath [dlen - 1] == '/') dirpath [dlen - 1] = '\0'; if (dirpath[0] != '/') @@ -459,8 +826,8 @@ mnt3_resolve_state_wipe (mnt3_resolve_t *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. */ -char * -__setup_next_component (char *path, char *component) +static char * +setup_next_component (char *path, size_t plen, char *component, size_t clen) { char *comp = NULL; char *nextcomp = NULL; @@ -468,7 +835,7 @@ __setup_next_component (char *path, char *component) if ((!path) || (!component)) return NULL; - strcpy (component, path); + strncpy (component, path, clen); comp = index (component, (int)'/'); if (!comp) goto err; @@ -476,7 +843,7 @@ __setup_next_component (char *path, char *component) comp++; nextcomp = index (comp, (int)'/'); if (nextcomp) { - strcpy (path, nextcomp); + strncpy (path, nextcomp, plen); *nextcomp = '\0'; } else path[0] = '\0'; @@ -506,7 +873,9 @@ __mnt3_resolve_export_subdir_comp (mnt3_resolve_t *mres) if (!mres) return ret; - nextcomp = __setup_next_component (mres->remainingdir, dupsubdir); + nextcomp = setup_next_component (mres->remainingdir, + sizeof (mres->remainingdir), + dupsubdir, sizeof (dupsubdir)); if (!nextcomp) goto err; @@ -518,7 +887,7 @@ __mnt3_resolve_export_subdir_comp (mnt3_resolve_t *mres) 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 (mres->resolveloc.inode->gfid), nextcomp); + uuid_utoa (gfid), nextcomp); ret = -EFAULT; goto err; } @@ -546,6 +915,7 @@ mnt3_resolve_subdir_cbk (call_frame_t *frame, void *cookie, xlator_t *this, rpcsvc_t *svc = NULL; mountres3 res = {0, }; xlator_t *mntxl = NULL; + char *path = NULL; mres = frame->local; mntxl = (xlator_t *)cookie; @@ -563,15 +933,23 @@ mnt3_resolve_subdir_cbk (call_frame_t *frame, void *cookie, xlator_t *this, 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, - mres->exp->expname); - goto err; + 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); } - - 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", @@ -614,7 +992,9 @@ __mnt3_resolve_subdir (mnt3_resolve_t *mres) if (!mres) return ret; - firstcomp = __setup_next_component (mres->remainingdir, dupsubdir); + firstcomp = setup_next_component (mres->remainingdir, + sizeof (mres->remainingdir), + dupsubdir, sizeof (dupsubdir)); if (!firstcomp) goto err; @@ -637,6 +1017,132 @@ err: } +/** + * 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) @@ -648,6 +1154,16 @@ mnt3_resolve_subdir (rpcsvc_request_t *req, struct mount3_state *ms, 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"); @@ -657,7 +1173,7 @@ mnt3_resolve_subdir (rpcsvc_request_t *req, struct mount3_state *ms, mres->exp = exp; mres->mstate = ms; mres->req = req; - strcpy (mres->remainingdir, subdir); + 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 @@ -732,6 +1248,7 @@ mnt3_mntpath_to_export (struct mount3_state *ms, char *dirpath) 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 */ @@ -745,6 +1262,7 @@ mnt3_mntpath_to_export (struct mount3_state *ms, char *dirpath) gf_log (GF_MNT, GF_LOG_DEBUG, "Export not found"); foundexp: + UNLOCK (&ms->mountlock); return found; } @@ -754,24 +1272,36 @@ mnt3_check_client_net (struct mount3_state *ms, rpcsvc_request_t *req, xlator_t *targetxl) { - rpcsvc_t *svc = NULL; - int ret = -1; + 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); - ret = rpcsvc_transport_peer_check (svc->options, targetxl->name, - rpcsvc_request_transport (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_TRACE, "Peer not allowed"); + 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_TRACE, "Unprivileged port not allowed"); + gf_log (GF_MNT, GF_LOG_INFO, "Peer %s rejected. Unprivileged " + "port not allowed", peer); goto err; } @@ -956,6 +1486,9 @@ __build_mountlist (struct mount3_state *ms, int *count) 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) { @@ -966,6 +1499,8 @@ __build_mountlist (struct mount3_state *ms, int *count) " failed"); goto free_list; } + if (!first) + first = mlist; mlist->ml_directory = GF_CALLOC (namelen + 2, sizeof (char), gf_nfs_mt_char); @@ -975,8 +1510,7 @@ __build_mountlist (struct mount3_state *ms, int *count) goto free_list; } - strcpy (mlist->ml_directory, "/"); - strcat (mlist->ml_directory, me->exname); + strcpy (mlist->ml_directory, me->exname); namelen = strlen (me->hostname); mlist->ml_hostname = GF_CALLOC (namelen + 2, sizeof (char), @@ -997,9 +1531,6 @@ __build_mountlist (struct mount3_state *ms, int *count) } else prev = mlist; - if (!first) - first = mlist; - (*count)++; } @@ -1076,67 +1607,71 @@ rpcerr: int -__mnt3svc_umount (struct mount3_state *ms, char *dirpath, char *hostname) +mnt3svc_umount (struct mount3_state *ms, char *dirpath, char *hostname) { struct mountentry *me = NULL; - char *exname = NULL; int ret = -1; + gf_store_handle_t *sh = NULL; + struct nfs_state *nfs = NULL; if ((!ms) || (!dirpath) || (!hostname)) return -1; - if (list_empty (&ms->mountlist)) - return 0; - - if (dirpath[0] == '/') - exname = dirpath+1; - else - exname = dirpath; + nfs = (struct nfs_state *)ms->nfsx->private; - list_for_each_entry (me, &ms->mountlist, mlist) { - if ((strcmp (me->exname, exname) == 0) && - (strcmp (me->hostname, hostname) == 0)) { - ret = 0; - break; - } + 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; } - /* 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_DEBUG, "Export not found"); - goto ret; + ret = gf_store_lock (sh); + if (ret) { + goto out_free; } - if (!me) - goto ret; + LOCK (&ms->mountlock); + { + __mount_read_rmtab (sh, &ms->mountlist, _gf_false); + if (list_empty (&ms->mountlist)) { + ret = 0; + 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); - ret = 0; -ret: - return ret; -} + 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; -int -mnt3svc_umount (struct mount3_state *ms, char *dirpath, char *hostname) -{ - int ret = -1; - if ((!ms) || (!dirpath) || (!hostname)) - return -1; + gf_log (GF_MNT, GF_LOG_DEBUG, "Unmounting: dir %s, host: %s", + me->exname, me->hostname); - LOCK (&ms->mountlock); - { - ret = __mnt3svc_umount (ms, dirpath, 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; } @@ -1150,6 +1685,7 @@ mnt3svc_umnt (rpcsvc_request_t *req) int ret = -1; struct mount3_state *ms = NULL; mountstat3 mstat = MNT3_OK; + char *colon = NULL; if (!req) return -1; @@ -1176,37 +1712,24 @@ mnt3svc_umnt (rpcsvc_request_t *req) if (ret != 0) { gf_log (GF_MNT, GF_LOG_ERROR, "Failed to get remote name: %s", gai_strerror (ret)); - goto try_umount_with_addr; - } - - gf_log (GF_MNT, GF_LOG_DEBUG, "dirpath: %s, hostname: %s", dirpath, - hostname); - ret = mnt3svc_umount (ms, dirpath, hostname); - - /* Unmount succeeded with the given hostname. */ - if (ret == 0) - goto snd_reply; - -try_umount_with_addr: - if (ret != 0) - ret = rpcsvc_transport_peeraddr (req->trans, hostname, - MNTPATHLEN, NULL, 0); - - if (ret != 0) { - gf_log (GF_MNT, GF_LOG_ERROR, "Failed to get remote addr: %s", - gai_strerror (ret)); - rpcsvc_request_seterr (req, SYSTEM_ERR); 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) - mstat = MNT3ERR_INVAL; - ret = 0; -snd_reply: + 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); @@ -1296,6 +1819,10 @@ mnt3_xlchildren_to_exports (rpcsvc_t *svc, struct mount3_state *ms) 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 @@ -1311,7 +1838,8 @@ mnt3_xlchildren_to_exports (rpcsvc_t *svc, struct mount3_state *ms) " 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) { @@ -1319,16 +1847,10 @@ mnt3_xlchildren_to_exports (rpcsvc_t *svc, struct mount3_state *ms) " failed"); goto free_list; } - strcpy (elist->ex_dir, ent->expname); addrstr = rpcsvc_volume_allowed (svc->options, ent->vol->name); - if (addrstr) - addrstr = gf_strdup (addrstr); - else - addrstr = gf_strdup ("No Access"); - elist->ex_groups = GF_CALLOC (1, sizeof (struct groupnode), gf_nfs_mt_groupnode); if (!elist->ex_groups) { @@ -1336,21 +1858,29 @@ mnt3_xlchildren_to_exports (rpcsvc_t *svc, struct mount3_state *ms) " 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; - - if (!first) - first = elist; } ret = 0; free_list: + UNLOCK (&ms->mountlock); if (ret == -1) { xdr_free_exports_list (first); first = NULL; @@ -1400,7 +1930,233 @@ 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) @@ -1418,6 +2174,20 @@ mnt3_init_export_ent (struct mount3_state *ms, xlator_t *xl, char *exportpath, 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); @@ -1427,8 +2197,6 @@ mnt3_init_export_ent (struct mount3_state *ms, xlator_t *xl, char *exportpath, exp->expname = GF_CALLOC (alloclen, sizeof (char), gf_nfs_mt_char); if (!exp->expname) { gf_log (GF_MNT, GF_LOG_ERROR, "Memory allocation failed"); - GF_FREE (exp); - exp = NULL; goto err; } @@ -1445,8 +2213,9 @@ mnt3_init_export_ent (struct mount3_state *ms, xlator_t *xl, char *exportpath, ret = snprintf (exp->expname, alloclen, "/%s", xl->name); } if (ret < 0) { - gf_log (xl->name, GF_LOG_WARNING, - "failed to get the export name"); + 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 @@ -1454,7 +2223,16 @@ mnt3_init_export_ent (struct mount3_state *ms, xlator_t *xl, char *exportpath, */ 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; } @@ -1494,8 +2272,7 @@ __mnt3_init_volume_direxports (struct mount3_state *ms, xlator_t *xlator, ret = 0; err: - if (dupopt) - GF_FREE (dupopt); + GF_FREE (dupopt); return ret; } @@ -1616,8 +2393,11 @@ __mnt3_init_volume_export (struct mount3_state *ms, dict_t *opts) goto err; } - gf_string2boolean (optstr, &boolt); - ret = 0; + 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) { @@ -1656,8 +2436,11 @@ __mnt3_init_dir_export (struct mount3_state *ms, dict_t *opts) goto err; } - gf_string2boolean (optstr, &boolt); - ret = 0; + 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) { @@ -1755,18 +2538,20 @@ out: } rpcsvc_actor_t mnt3svc_actors[MOUNT3_PROC_COUNT] = { - {"NULL", MOUNT3_NULL, mnt3svc_null, NULL, NULL, 0}, - {"MNT", MOUNT3_MNT, mnt3svc_mnt, NULL, NULL, 0}, - {"DUMP", MOUNT3_DUMP, mnt3svc_dump, NULL, NULL, 0}, - {"UMNT", MOUNT3_UMNT, mnt3svc_umnt, NULL, NULL, 0}, - {"UMNTALL", MOUNT3_UMNTALL, mnt3svc_umntall, NULL, NULL, 0}, - {"EXPORT", MOUNT3_EXPORT, mnt3svc_export, NULL, NULL, 0} + {"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", @@ -1776,9 +2561,11 @@ rpcsvc_program_t mnt3prog = { .actors = mnt3svc_actors, .numactors = MOUNT3_PROC_COUNT, .min_auth = AUTH_NULL, + .synctask = _gf_true, }; + rpcsvc_program_t * mnt3svc_init (xlator_t *nfsx) { @@ -1787,6 +2574,7 @@ mnt3svc_init (xlator_t *nfsx) dict_t *options = NULL; char *portstr = NULL; int ret = -1; + pthread_t udp_thread; if (!nfsx || !nfsx->private) return NULL; @@ -1829,13 +2617,16 @@ mnt3svc_init (xlator_t *nfsx) } } - rpcsvc_create_listeners (nfs->rpcsvc, options, nfsx->name); + 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; @@ -1843,12 +2634,12 @@ err: rpcsvc_actor_t mnt1svc_actors[MOUNT1_PROC_COUNT] = { - {"NULL", MOUNT1_NULL, mnt3svc_null, NULL, NULL, 0}, - {{0, 0}, }, - {"DUMP", MOUNT1_DUMP, mnt3svc_dump, NULL, NULL, 0}, - {"UMNT", MOUNT1_UMNT, mnt3svc_umnt, NULL, NULL, 0}, - {{0, 0}, }, - {"EXPORT", MOUNT1_EXPORT, mnt3svc_export, NULL, NULL, 0} + {"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 = { @@ -1859,6 +2650,7 @@ rpcsvc_program_t mnt1prog = { .actors = mnt1svc_actors, .numactors = MOUNT1_PROC_COUNT, .min_auth = AUTH_NULL, + .synctask = _gf_true, }; @@ -1913,7 +2705,7 @@ mnt1svc_init (xlator_t *nfsx) } } - rpcsvc_create_listeners (nfs->rpcsvc, options, nfsx->name); + 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); @@ -1925,4 +2717,43 @@ 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 index c0eae3644..7fc16ed57 100644 --- a/xlators/nfs/server/src/mount3.h +++ b/xlators/nfs/server/src/mount3.h @@ -2,19 +2,10 @@ Copyright (c) 2010-2011 Gluster, Inc. <http://www.gluster.com> This file is part of GlusterFS. - GlusterFS is free software; you can redistribute it and/or modify - it under the terms of the GNU General Public License as published - by the Free Software Foundation; either version 3 of the License, - or (at your option) any later version. - - GlusterFS is distributed in the hope that it will be useful, but - WITHOUT ANY WARRANTY; without even the implied warranty of - MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU - General Public License for more details. - - You should have received a copy of the GNU General Public License - along with this program. If not, see - <http://www.gnu.org/licenses/>. + 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_ @@ -53,6 +44,12 @@ 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. */ @@ -68,6 +65,13 @@ struct mountentry { #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; @@ -75,6 +79,11 @@ struct mnt3_export { * 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; @@ -101,8 +110,8 @@ struct mount3_state { gf_lock_t mountlock; /* Set to 0 if exporting full volumes is disabled. On by default. */ - int export_volumes; - int export_dirs; + gf_boolean_t export_volumes; + gf_boolean_t export_dirs; }; #define gf_mnt3_export_dirs(mst) ((mst)->export_dirs) 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 index ec76c294f..f74396ee8 100644 --- a/xlators/nfs/server/src/nfs-common.c +++ b/xlators/nfs/server/src/nfs-common.c @@ -2,19 +2,10 @@ Copyright (c) 2010-2011 Gluster, Inc. <http://www.gluster.com> This file is part of GlusterFS. - GlusterFS is free software; you can redistribute it and/or modify - it under the terms of the GNU General Public License as published - by the Free Software Foundation; either version 3 of the License, - or (at your option) any later version. - - GlusterFS is distributed in the hope that it will be useful, but - WITHOUT ANY WARRANTY; without even the implied warranty of - MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU - General Public License for more details. - - You should have received a copy of the GNU General Public License - along with this program. If not, see - <http://www.gnu.org/licenses/>. + 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 @@ -88,13 +79,13 @@ nfs_mntpath_to_xlator (xlator_list_t *cl, char *path) { char volname[MNTPATHLEN]; char *volptr = NULL; - int pathlen = 0; + size_t pathlen; xlator_t *targetxl = NULL; if ((!cl) || (!path)) return NULL; - strcpy (volname, path); + strncpy (volname, path, MNTPATHLEN); pathlen = strlen (volname); gf_log (GF_NFS, GF_LOG_TRACE, "Subvolume search: %s", path); if (volname[0] == '/') @@ -102,7 +93,7 @@ nfs_mntpath_to_xlator (xlator_list_t *cl, char *path) else volptr = &volname[0]; - if (volname[pathlen - 1] == '/') + if (pathlen && volname[pathlen - 1] == '/') volname[pathlen - 1] = '\0'; while (cl) { @@ -221,6 +212,8 @@ nfs_inode_loc_fill (inode_t *inode, loc_t *loc, int how) 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); @@ -235,8 +228,7 @@ err: if (parent) inode_unref (parent); - if (resolvedpath) - GF_FREE (resolvedpath); + GF_FREE (resolvedpath); return ret; } @@ -397,8 +389,7 @@ err: if (entryinode) inode_unref (entryinode); - if (resolvedpath) - GF_FREE (resolvedpath); + GF_FREE (resolvedpath); return ret; } @@ -417,6 +408,9 @@ nfs_hash_gfid (uuid_t gfid) 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); @@ -434,3 +428,38 @@ nfs_hash_gfid (uuid_t gfid) } +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 index 88fc14961..2e97f1563 100644 --- a/xlators/nfs/server/src/nfs-common.h +++ b/xlators/nfs/server/src/nfs-common.h @@ -2,19 +2,10 @@ Copyright (c) 2010-2011 Gluster, Inc. <http://www.gluster.com> This file is part of GlusterFS. - GlusterFS is free software; you can redistribute it and/or modify - it under the terms of the GNU General Public License as published - by the Free Software Foundation; either version 3 of the License, - or (at your option) any later version. - - GlusterFS is distributed in the hope that it will be useful, but - WITHOUT ANY WARRANTY; without even the implied warranty of - MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU - General Public License for more details. - - You should have received a copy of the GNU General Public License - along with this program. If not, see - <http://www.gnu.org/licenses/>. + 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_ @@ -84,4 +75,7 @@ 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 index 6e2b33484..236b80c76 100644 --- a/xlators/nfs/server/src/nfs-fops.c +++ b/xlators/nfs/server/src/nfs-fops.c @@ -2,19 +2,10 @@ Copyright (c) 2010-2011 Gluster, Inc. <http://www.gluster.com> This file is part of GlusterFS. - GlusterFS is free software; you can redistribute it and/or modify - it under the terms of the GNU General Public License as published - by the Free Software Foundation; either version 3 of the License, - or (at your option) any later version. - - GlusterFS is distributed in the hope that it will be useful, but - WITHOUT ANY WARRANTY; without even the implied warranty of - MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU - General Public License for more details. - - You should have received a copy of the GNU General Public License - along with this program. If not, see - <http://www.gnu.org/licenses/>. + 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 @@ -22,6 +13,9 @@ #include "config.h" #endif +#include <grp.h> +#include <pwd.h> + #include "dict.h" #include "xlator.h" #include "iobuf.h" @@ -32,9 +26,79 @@ #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) { @@ -118,22 +182,34 @@ nfs_create_frame (xlator_t *xl, nfs_user_t *nfu) 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) - goto err; /* Done, we only got primary gid */ - frame->root->ngrps = nfu->ngrps - 1; + 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]; + 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; } @@ -274,9 +350,6 @@ out: } \ } 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: @@ -310,6 +383,10 @@ nfs_fop_lookup_cbk (call_frame_t *frame, void *cookie, xlator_t *this, 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) @@ -680,6 +757,10 @@ nfs_fop_create_cbk (call_frame_t *frame, void *cookie, xlator_t *this, 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); @@ -782,6 +863,10 @@ nfs_fop_mkdir_cbk (call_frame_t *frame, void *cookie, xlator_t *this, 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) @@ -831,6 +916,10 @@ nfs_fop_symlink_cbk (call_frame_t *frame, void *cookie, xlator_t *this, 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) @@ -927,6 +1016,10 @@ nfs_fop_mknod_cbk (call_frame_t *frame, void *cookie, xlator_t *this, 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) @@ -1074,6 +1167,10 @@ nfs_fop_link_cbk (call_frame_t *frame, void *cookie, xlator_t *this, 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); @@ -1266,7 +1363,7 @@ nfs_fop_write (xlator_t *nfsx, xlator_t *xl, nfs_user_t *nfu, fd_t *fd, iobref_add (nfl->iobref, srciob); */ STACK_WIND_COOKIE (frame, nfs_fop_writev_cbk, xl, xl,xl->fops->writev, - fd, vector, count, offset, 0, srciobref, NULL); + fd, vector, count, offset, fd->flags, srciobref, NULL); ret = 0; err: if (ret < 0) { @@ -1426,6 +1523,95 @@ err: 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, diff --git a/xlators/nfs/server/src/nfs-fops.h b/xlators/nfs/server/src/nfs-fops.h index b82805939..44e99c66b 100644 --- a/xlators/nfs/server/src/nfs-fops.h +++ b/xlators/nfs/server/src/nfs-fops.h @@ -2,19 +2,10 @@ Copyright (c) 2010-2011 Gluster, Inc. <http://www.gluster.com> This file is part of GlusterFS. - GlusterFS is free software; you can redistribute it and/or modify - it under the terms of the GNU General Public License as published - by the Free Software Foundation; either version 3 of the License, - or (at your option) any later version. - - GlusterFS is distributed in the hope that it will be useful, but - WITHOUT ANY WARRANTY; without even the implied warranty of - MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU - General Public License for more details. - - You should have received a copy of the GNU General Public License - along with this program. If not, see - <http://www.gnu.org/licenses/>. + 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_ @@ -120,7 +111,7 @@ nfs_fop_local_wipe (xlator_t *xl, struct nfs_fop_local *l); nflocal = nfs_fop_local_init (nf); \ if (nflocal) { \ nflocal->proglocal = plocal; \ - nflocal->progcbk = pcbk; \ + nflocal->progcbk = *VOID(&pcbk); \ nflocal->nfsx = nf; \ if (fram) \ ((call_frame_t *)fram)->local = nflocal;\ @@ -245,4 +236,13 @@ 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 index 553cc7f39..cb32b7f1b 100644 --- a/xlators/nfs/server/src/nfs-generics.c +++ b/xlators/nfs/server/src/nfs-generics.c @@ -2,19 +2,10 @@ Copyright (c) 2010-2011 Gluster, Inc. <http://www.gluster.com> This file is part of GlusterFS. - GlusterFS is free software; you can redistribute it and/or modify - it under the terms of the GNU General Public License as published - by the Free Software Foundation; either version 3 of the License, - or (at your option) any later version. - - GlusterFS is distributed in the hope that it will be useful, but - WITHOUT ANY WARRANTY; without even the implied warranty of - MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU - General Public License for more details. - - You should have received a copy of the GNU General Public License - along with this program. If not, see - <http://www.gnu.org/licenses/>. + 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 @@ -166,6 +157,22 @@ nfs_lk (xlator_t *nfsx, xlator_t *xl, nfs_user_t *nfu, fd_t *fd, } 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) { diff --git a/xlators/nfs/server/src/nfs-generics.h b/xlators/nfs/server/src/nfs-generics.h index 11f191f5f..01876d68e 100644 --- a/xlators/nfs/server/src/nfs-generics.h +++ b/xlators/nfs/server/src/nfs-generics.h @@ -2,19 +2,10 @@ Copyright (c) 2010-2011 Gluster, Inc. <http://www.gluster.com> This file is part of GlusterFS. - GlusterFS is free software; you can redistribute it and/or modify - it under the terms of the GNU General Public License as published - by the Free Software Foundation; either version 3 of the License, - or (at your option) any later version. - - GlusterFS is distributed in the hope that it will be useful, but - WITHOUT ANY WARRANTY; without even the implied warranty of - MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU - General Public License for more details. - - You should have received a copy of the GNU General Public License - along with this program. If not, see - <http://www.gnu.org/licenses/>. + 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_ @@ -164,4 +155,14 @@ nfs_access (xlator_t *nfsx, xlator_t *xl, nfs_user_t *nfu, loc_t *pathloc, 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 index c9bf37206..63d5e8a19 100644 --- a/xlators/nfs/server/src/nfs-inodes.c +++ b/xlators/nfs/server/src/nfs-inodes.c @@ -2,19 +2,10 @@ Copyright (c) 2010-2011 Gluster, Inc. <http://www.gluster.com> This file is part of GlusterFS. - GlusterFS is free software; you can redistribute it and/or modify - it under the terms of the GNU General Public License as published - by the Free Software Foundation; either version 3 of the License, - or (at your option) any later version. - - GlusterFS is distributed in the hope that it will be useful, but - WITHOUT ANY WARRANTY; without even the implied warranty of - MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU - General Public License for more details. - - You should have received a copy of the GNU General Public License - along with this program. If not, see - <http://www.gnu.org/licenses/>. + 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 @@ -36,7 +27,7 @@ do { \ nflocal = fram->local; \ fram->local = nflocal->proglocal; \ - pcbk = nflocal->progcbk; \ + *VOID(&pcbk) = nflocal->progcbk; \ nfs_fop_local_wipe (nflocal->nfsx, nflocal); \ } while (0) \ @@ -57,10 +48,10 @@ nfl_inodes_init (struct nfs_fop_local *nfl, inode_t *inode, inode_t *parent, nfl->newparent = inode_ref (newparent); if (name) - strcpy (nfl->path, name); + strncpy (nfl->path, name, NFS_NAME_MAX); if (newname) - strcpy (nfl->newpath, newname); + strncpy (nfl->newpath, newname, NFS_NAME_MAX); return; } @@ -572,9 +563,7 @@ nfs_inode_opendir_cbk (call_frame_t *frame, void *cookie, xlator_t *this, struct nfs_fop_local *nfl = NULL; fop_open_cbk_t progcbk = NULL; - if ((op_ret == -1) && (fd)) - fd_unref (fd); - else + if (op_ret != -1) fd_bind (fd); inodes_nfl_to_prog_data (nfl, progcbk, frame); diff --git a/xlators/nfs/server/src/nfs-inodes.h b/xlators/nfs/server/src/nfs-inodes.h index 7c962b339..ba7a57124 100644 --- a/xlators/nfs/server/src/nfs-inodes.h +++ b/xlators/nfs/server/src/nfs-inodes.h @@ -2,19 +2,10 @@ Copyright (c) 2010-2011 Gluster, Inc. <http://www.gluster.com> This file is part of GlusterFS. - GlusterFS is free software; you can redistribute it and/or modify - it under the terms of the GNU General Public License as published - by the Free Software Foundation; either version 3 of the License, - or (at your option) any later version. - - GlusterFS is distributed in the hope that it will be useful, but - WITHOUT ANY WARRANTY; without even the implied warranty of - MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU - General Public License for more details. - - You should have received a copy of the GNU General Public License - along with this program. If not, see - <http://www.gnu.org/licenses/>. + 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_ diff --git a/xlators/nfs/server/src/nfs-mem-types.h b/xlators/nfs/server/src/nfs-mem-types.h index e2dae4a34..450b6f2fe 100644 --- a/xlators/nfs/server/src/nfs-mem-types.h +++ b/xlators/nfs/server/src/nfs-mem-types.h @@ -2,19 +2,10 @@ Copyright (c) 2008-2011 Gluster, Inc. <http://www.gluster.com> This file is part of GlusterFS. - GlusterFS is free software; you can redistribute it and/or modify - it under the terms of the GNU General Public License as published - by the Free Software Foundation; either version 3 of the License, - or (at your option) any later version. - - GlusterFS is distributed in the hope that it will be useful, but - WITHOUT ANY WARRANTY; without even the implied warranty of - MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU - General Public License for more details. - - You should have received a copy of the GNU General Public License - along with this program. If not, see - <http://www.gnu.org/licenses/>. + 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. */ @@ -37,16 +28,25 @@ enum gf_nfs_mem_types_ { 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_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 index f2c690994..75c8fe44e 100644 --- a/xlators/nfs/server/src/nfs.c +++ b/xlators/nfs/server/src/nfs.c @@ -2,19 +2,10 @@ Copyright (c) 2010-2011 Gluster, Inc. <http://www.gluster.com> This file is part of GlusterFS. - GlusterFS is free software; you can redistribute it and/or modify - it under the terms of the GNU General Public License as published - by the Free Software Foundation; either version 3 of the License, - or (at your option) any later version. - - GlusterFS is distributed in the hope that it will be useful, but - WITHOUT ANY WARRANTY; without even the implied warranty of - MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU - General Public License for more details. - - You should have received a copy of the GNU General Public License - along with this program. If not, see - <http://www.gnu.org/licenses/>. + 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. @@ -41,6 +32,207 @@ #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. @@ -118,7 +310,7 @@ nfs_init_versions (struct nfs_state *nfs, xlator_t *this) ret = -1; goto err; } -// prog->actorxl = this; + version->program = prog; if (nfs->override_portnum) prog->progport = nfs->override_portnum; @@ -127,17 +319,21 @@ nfs_init_versions (struct nfs_state *nfs, xlator_t *this) ret = rpcsvc_program_register (nfs->rpcsvc, prog); if (ret == -1) { - gf_log (GF_NFS, GF_LOG_ERROR, "Program init failed"); + gf_log (GF_NFS, GF_LOG_ERROR, "Program: %s init failed", + prog->progname); goto err; } - if (rpcsvc_register_portmap_enabled(nfs->rpcsvc)) { + if (nfs->register_portmap) { ret = rpcsvc_program_register_portmap (prog, prog->progport); if (ret == -1) { - gf_log (GF_NFS, GF_LOG_ERROR, "Program registration failed"); + gf_log (GF_NFS, GF_LOG_ERROR, + "Program %s registration failed", + prog->progname); goto err; } } + } ret = 0; @@ -154,30 +350,41 @@ nfs_add_all_initiators (struct nfs_state *nfs) /* 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 protocol" - " initializer"); + 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 protocol" - " initializer"); + 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 protocol" - " initializer"); + gf_log (GF_NFS, GF_LOG_ERROR, "Failed to add " + "NFS3 protocol initializer"); goto ret; } - 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_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; @@ -508,14 +715,16 @@ nfs_init_state (xlator_t *this) unsigned int fopspoolsize = 0; char *optstr = NULL; gf_boolean_t boolt = _gf_false; + struct stat stbuf = {0,}; if (!this) return NULL; - if ((!this->children) || (!this->children->xlator)) { - gf_log (GF_NFS, GF_LOG_ERROR, "nfs must have at least one" - " child subvolume"); - 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); @@ -570,6 +779,20 @@ nfs_init_state (xlator_t *this) 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", @@ -632,6 +855,34 @@ nfs_init_state (xlator_t *this) } } + 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 */ @@ -692,6 +943,22 @@ nfs_init_state (xlator_t *this) } } + 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; @@ -699,8 +966,11 @@ nfs_init_state (xlator_t *this) 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; @@ -719,6 +989,243 @@ free_rpcsvc: 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) { @@ -774,7 +1281,9 @@ init (xlator_t *this) { goto err; } - gf_log (GF_NFS, GF_LOG_INFO, "NFS service started"); + ret = nfs_drc_init (this); + if (ret == 0) + gf_log (GF_NFS, GF_LOG_INFO, "NFS service started"); err: return ret; @@ -785,24 +1294,26 @@ 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_PARENT_UP: - { - default_notify (this, GF_EVENT_PARENT_UP, data); - break; - } + 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; @@ -824,6 +1335,15 @@ fini (xlator_t *this) 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; } @@ -912,13 +1432,39 @@ out: 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_fops fops; struct xlator_dumpops dumpops = { + .priv = nfs_priv, .priv_to_dict = nfs_priv_to_dict, }; @@ -930,29 +1476,53 @@ struct xlator_dumpops dumpops = { struct volume_options options[] = { { .key = {"nfs3.read-size"}, .type = GF_OPTION_TYPE_SIZET, - .description = "Size in which the client should issue read requests" - " to the Gluster NFSv3 server. Must be a multiple of" - " 4KB." + .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, - .description = "Size in which the client should issue write requests" - " to the Gluster NFSv3 server. Must be a multiple of" - " 4KB." + .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." + "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 " @@ -967,6 +1537,7 @@ struct volume_options options[] = { }, { .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 " @@ -976,25 +1547,40 @@ struct volume_options options[] = { }, { .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." + " 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." + "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 " @@ -1002,6 +1588,7 @@ struct volume_options options[] = { }, { .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" @@ -1009,12 +1596,14 @@ struct volume_options options[] = { }, { .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 " @@ -1024,6 +1613,7 @@ struct volume_options options[] = { }, { .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 " @@ -1033,6 +1623,7 @@ struct volume_options options[] = { }, { .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 " @@ -1041,6 +1632,7 @@ struct volume_options options[] = { }, { .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 " @@ -1048,6 +1640,7 @@ struct volume_options options[] = { }, { .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" @@ -1055,6 +1648,7 @@ struct volume_options options[] = { }, { .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 " @@ -1062,6 +1656,7 @@ struct volume_options options[] = { }, { .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" @@ -1069,6 +1664,7 @@ struct volume_options options[] = { }, { .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 " @@ -1076,31 +1672,35 @@ struct volume_options options[] = { }, { .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." + "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 rpc-auth.addr.* filters. In some " + " 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" + "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. " @@ -1109,22 +1709,34 @@ struct volume_options options[] = { }, { .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." + "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." }, @@ -1132,6 +1744,7 @@ struct volume_options options[] = { .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. " @@ -1142,9 +1755,71 @@ struct volume_options options[] = { }, { .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 index 706cba86f..00c7f8046 100644 --- a/xlators/nfs/server/src/nfs.h +++ b/xlators/nfs/server/src/nfs.h @@ -2,19 +2,10 @@ Copyright (c) 2010-2011 Gluster, Inc. <http://www.gluster.com> This file is part of GlusterFS. - GlusterFS is free software; you can redistribute it and/or modify - it under the terms of the GNU General Public License as published - by the Free Software Foundation; either version 3 of the License, - or (at your option) any later version. - - GlusterFS is distributed in the hope that it will be useful, but - WITHOUT ANY WARRANTY; without even the implied warranty of - MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU - General Public License for more details. - - You should have received a copy of the GNU General Public License - along with this program. If not, see - <http://www.gnu.org/licenses/>. + 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__ @@ -29,6 +20,7 @@ #include "dict.h" #include "xlator.h" #include "lkowner.h" +#include "gidcache.h" #define GF_NFS "nfs" @@ -44,7 +36,7 @@ #define GF_NFS_MAX_MEMFACTOR 30 #define GF_NFS_DVM_ON 1 -#define GF_NFS_DVM_OFF 2 +#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 @@ -65,7 +57,6 @@ struct nfs_initer_list { rpcsvc_program_t *program; }; - struct nfs_state { rpcsvc_t *rpcsvc; struct list_head versions; @@ -85,7 +76,21 @@ struct nfs_state { 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) @@ -124,4 +129,7 @@ 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 index 9aea88137..e199c56dc 100644 --- a/xlators/nfs/server/src/nfs3-fh.c +++ b/xlators/nfs/server/src/nfs3-fh.c @@ -2,19 +2,10 @@ Copyright (c) 2010-2011 Gluster, Inc. <http://www.gluster.com> This file is part of GlusterFS. - GlusterFS is free software; you can redistribute it and/or modify - it under the terms of the GNU General Public License as published - by the Free Software Foundation; either version 3 of the License, - or (at your option) any later version. - - GlusterFS is distributed in the hope that it will be useful, but - WITHOUT ANY WARRANTY; without even the implied warranty of - MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU - General Public License for more details. - - You should have received a copy of the GNU General Public License - along with this program. If not, see - <http://www.gnu.org/licenses/>. + 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 @@ -46,6 +37,12 @@ nfs3_fh_validate (struct nfs3_fh *fh) 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; } @@ -58,8 +55,9 @@ nfs3_fh_init (struct nfs3_fh *fh, struct iatt *buf) fh->ident[0] = GF_NFSFH_IDENT0; fh->ident[1] = GF_NFSFH_IDENT1; + fh->ident[2] = GF_NFSFH_IDENT2; + fh->ident[3] = GF_NFSFH_IDENT3; - fh->hashcount = 0; uuid_copy (fh->gfid, buf->ia_gfid); } @@ -112,99 +110,33 @@ nfs3_fh_is_root_fh (struct nfs3_fh *fh) } -nfs3_hash_entry_t -nfs3_fh_hash_entry (uuid_t gfid) -{ - nfs3_hash_entry_t hash = 0; - int shiftsize = 48; - uint64_t ino = 0; - uint64_t gen = 0; - nfs3_hash_entry_t inomsb = 0; - nfs3_hash_entry_t inolsb = 0; - nfs3_hash_entry_t inols23b = 0; - - nfs3_hash_entry_t genmsb = 0; - nfs3_hash_entry_t genlsb = 0; - nfs3_hash_entry_t genls23b = 0; - - memcpy (&ino, &gfid[8], 8); - hash = ino; - while (shiftsize != 0) { - hash ^= (ino >> shiftsize); - shiftsize -= 16; - } -/* - gf_log ("FILEHANDLE", GF_LOG_TRACE, "INO %"PRIu64, ino); - gf_log ("FILEHANDLE",GF_LOG_TRACE, "PRI HASH %d", hash); -*/ - inomsb = (ino >> 56); -// gf_log ("FILEHANDLE", GF_LOG_TRACE, "inomsb %d", inomsb); - - inolsb = ((ino << 56) >> 56); -// gf_log ("FILEHANDLE", GF_LOG_TRACE, "inolsb %d", inolsb); - - inolsb = (inolsb << 8); -// gf_log ("FILEHANDLE", GF_LOG_TRACE, "inolsb to inomsb %d", inolsb); - inols23b = ((ino << 40) >> 48); -// gf_log ("FILEHANDLE", GF_LOG_TRACE, "inols23b %d", inols23b); - - inols23b = (inols23b << 8); -// gf_log ("FILEHDNALE", GF_LOG_TRACE, "inols23b %d", inols23b); - - memcpy (&gen, &gfid[0], 8); - genmsb = (gen >> 56); -// gf_log ("FILEHANDLE", GF_LOG_TRACE, "inomsb %d", inomsb); - - genlsb = ((gen << 56) >> 56); -// gf_log ("FILEHANDLE", GF_LOG_TRACE, "inolsb %d", inolsb); - - genlsb = (genlsb << 8); -// gf_log ("FILEHANDLE", GF_LOG_TRACE, "inolsb to inomsb %d", inolsb); - - genls23b = ((gen << 40) >> 48); -// gf_log ("FILEHANDLE", GF_LOG_TRACE, "inols23b %d", inols23b); - - genls23b = (genls23b << 8); -// gf_log ("FILEHDNALE", GF_LOG_TRACE, "inols23b %d", inols23b); - - hash ^= inolsb ^ inomsb ^ inols23b ^ genmsb ^ genlsb ^ genls23b; - return hash; - -} - void -nfs3_fh_to_str (struct nfs3_fh *fh, char *str) +nfs3_fh_to_str (struct nfs3_fh *fh, char *str, size_t len) { - char gfid[512]; - char exportid[512]; + char gfid[GF_UUID_BUF_SIZE]; + char exportid[GF_UUID_BUF_SIZE]; if ((!fh) || (!str)) return; - sprintf (str, "FH: hashcount %d, exportid %s, gfid %s", - fh->hashcount, uuid_utoa_r (fh->exportid, exportid), - uuid_utoa_r (fh->gfid, gfid)); + 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) { -// int x = 0; char gfidstr[512]; char exportidstr[512]; if (!fh) return; - gf_log ("nfs3-fh", GF_LOG_TRACE, "filehandle: hashcount %d, exportid " - "0x%s, gfid 0x%s", fh->hashcount, + 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)); -/* - for (; x < fh->hashcount; ++x) - gf_log ("FILEHANDLE", GF_LOG_TRACE, "Hash %d: %d", x, - fh->entryhash[x]); -*/ } int @@ -216,53 +148,35 @@ nfs3_fh_build_parent_fh (struct nfs3_fh *child, struct iatt *newstat, nfs3_fh_init (newfh, newstat); uuid_copy (newfh->exportid, child->exportid); - if (newstat->ia_ino == 1) - goto done; - newfh->hashcount = child->hashcount - 1; - memcpy (newfh->entryhash, child->entryhash, - newfh->hashcount * GF_NFSFH_ENTRYHASH_SIZE); + return 0; +} -done: -// nfs3_log_fh (newfh); +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) { - int hashcount = 0; - int entry = 0; - if ((!parent) || (!newstat) || (!newfh)) return -1; nfs3_fh_init (newfh, newstat); uuid_copy (newfh->exportid, parent->exportid); - newfh->hashcount = parent->hashcount + 1; - /* Only copy the hashes that are available in the parent file - * handle. */ - if (parent->hashcount > GF_NFSFH_MAXHASHES) - hashcount = GF_NFSFH_MAXHASHES; - else - hashcount = parent->hashcount; - - memcpy (newfh->entryhash, parent->entryhash, - hashcount * GF_NFSFH_ENTRYHASH_SIZE); - - /* Do not insert parent dir hash if there is no space left in the hash - * array of the child entry. */ - if (newfh->hashcount <= GF_NFSFH_MAXHASHES) { - entry = newfh->hashcount - 1; - newfh->entryhash[entry] = nfs3_fh_hash_entry (parent->gfid); - } - -// nfs3_log_fh (newfh); - return 0; } @@ -270,34 +184,5 @@ nfs3_fh_build_child_fh (struct nfs3_fh *parent, struct iatt *newstat, uint32_t nfs3_fh_compute_size (struct nfs3_fh *fh) { - uint32_t fhlen = 0; - - if (!fh) - return 0; - - if (fh->hashcount <= GF_NFSFH_MAXHASHES) - fhlen = nfs3_fh_hashcounted_size (fh->hashcount); - else - fhlen = nfs3_fh_hashcounted_size (GF_NFSFH_MAXHASHES); - - return fhlen; -} - - -/* There is no point searching at a directory level which is beyond that of - * the hashcount given in the file handle. - */ -int -nfs3_fh_hash_index_is_beyond (struct nfs3_fh *fh, int hashidx) -{ - if (!fh) - return 1; - - if (fh->hashcount >= hashidx) - return 0; - else - return 1; - - return 1; + return GF_NFSFH_STATIC_SIZE; } - diff --git a/xlators/nfs/server/src/nfs3-fh.h b/xlators/nfs/server/src/nfs3-fh.h index e7c0625ee..1049cdc96 100644 --- a/xlators/nfs/server/src/nfs3-fh.h +++ b/xlators/nfs/server/src/nfs3-fh.h @@ -2,19 +2,10 @@ Copyright (c) 2010-2011 Gluster, Inc. <http://www.gluster.com> This file is part of GlusterFS. - GlusterFS is free software; you can redistribute it and/or modify - it under the terms of the GNU General Public License as published - by the Free Software Foundation; either version 3 of the License, - or (at your option) any later version. - - GlusterFS is distributed in the hope that it will be useful, but - WITHOUT ANY WARRANTY; without even the implied warranty of - MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU - General Public License for more details. - - You should have received a copy of the GNU General Public License - along with this program. If not, see - <http://www.gnu.org/licenses/>. + 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_ @@ -35,17 +26,10 @@ * handles for now. This will change if and when we need v4. */ #define GF_NFSFH_IDENT0 ':' #define GF_NFSFH_IDENT1 'O' -#define GF_NFSFH_IDENT_SIZE (sizeof(char) * 2) -#define GF_NFSFH_STATIC_SIZE (GF_NFSFH_IDENT_SIZE + (2*sizeof (uuid_t)) + sizeof (uint16_t)) -#define GF_NFSFH_MAX_HASH_BYTES (NFS3_FHSIZE - GF_NFSFH_STATIC_SIZE) - -/* Each hash element in the file handle is of 2 bytes thus giving - * us theoretically 65536 unique entries in a directory. - */ -typedef uint16_t nfs3_hash_entry_t; -#define GF_NFSFH_ENTRYHASH_SIZE (sizeof (nfs3_hash_entry_t)) -#define GF_NFSFH_MAXHASHES ((int)(GF_NFSFH_MAX_HASH_BYTES / GF_NFSFH_ENTRYHASH_SIZE)) -#define nfs3_fh_hashcounted_size(hcount) (GF_NFSFH_STATIC_SIZE + (hcount * GF_NFSFH_ENTRYHASH_SIZE)) +#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 @@ -54,9 +38,9 @@ typedef uint16_t nfs3_hash_entry_t; struct nfs3_fh { /* Used to ensure that a bunch of bytes are actually a GlusterFS NFS - * file handle. Should contain ":O" + * file handle. Should contain ":OGL" */ - char ident[2]; + 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 @@ -72,11 +56,11 @@ struct nfs3_fh { /* File/dir gfid. */ uuid_t gfid; - - /* Number of file/ino hash elements that follow the ino. */ - uint16_t hashcount; - - nfs3_hash_entry_t entryhash[GF_NFSFH_MAXHASHES]; + /* 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},} @@ -84,9 +68,6 @@ struct nfs3_fh { extern uint32_t nfs3_fh_compute_size (struct nfs3_fh *fh); -extern int -nfs3_fh_hash_index_is_beyond (struct nfs3_fh *fh, int hashidx); - extern uint16_t nfs3_fh_hash_entry (uuid_t gfid); @@ -107,7 +88,7 @@ extern void nfs3_log_fh (struct nfs3_fh *fh); extern void -nfs3_fh_to_str (struct nfs3_fh *fh, char *str); +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, @@ -115,4 +96,8 @@ nfs3_fh_build_parent_fh (struct nfs3_fh *child, struct iatt *newstat, 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 index e2c16ef0e..9059fc341 100644 --- a/xlators/nfs/server/src/nfs3-helpers.c +++ b/xlators/nfs/server/src/nfs3-helpers.c @@ -2,19 +2,10 @@ Copyright (c) 2010-2011 Gluster, Inc. <http://www.gluster.com> This file is part of GlusterFS. - GlusterFS is free software; you can redistribute it and/or modify - it under the terms of the GNU General Public License as published - by the Free Software Foundation; either version 3 of the License, - or (at your option) any later version. - - GlusterFS is distributed in the hope that it will be useful, but - WITHOUT ANY WARRANTY; without even the implied warranty of - MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU - General Public License for more details. - - You should have received a copy of the GNU General Public License - along with this program. If not, see - <http://www.gnu.org/licenses/>. + 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 @@ -265,6 +256,20 @@ nfs3_errno_to_nfsstat3 (int errnum) 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, @@ -284,6 +289,9 @@ 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)) @@ -353,6 +361,7 @@ nfs3_stat_to_fattr3 (struct iatt *buf) fa.mtime.seconds = buf->ia_mtime; fa.mtime.nseconds = buf->ia_mtime_nsec; +out: return fa; } @@ -496,7 +505,7 @@ nfs3_fill_fsinfo3res (struct nfs3_state *nfs3, fsinfo3res *res, resok.wtpref = nfs3->writesize; resok.wtmult = GF_NFS3_WTMULT; resok.dtpref = nfs3->readdirsize; - resok.maxfilesize = GF_NFS3_MAXFILE; + resok.maxfilesize = GF_NFS3_MAXFILESIZE; resok.time_delta = tdelta; resok.properties = GF_NFS3_FS_PROP; @@ -591,7 +600,8 @@ nfs3_request_to_accessbits (int32_t accbits) return acc_request; } void -nfs3_fill_access3res (access3res *res, nfsstat3 status, int32_t accbits) +nfs3_fill_access3res (access3res *res, nfsstat3 status, int32_t accbits, + int32_t reqaccbits) { uint32_t accres = 0; @@ -602,7 +612,8 @@ nfs3_fill_access3res (access3res *res, nfsstat3 status, int32_t accbits) accres = nfs3_accessbits (accbits); - res->access3res_u.resok.access = accres; + /* do not answer what was not asked */ + res->access3res_u.resok.access = accres & reqaccbits; } void @@ -1605,13 +1616,14 @@ err: void nfs3_stat_to_errstr (uint32_t xid, char *op, nfsstat3 stat, int pstat, - char *errstr) + char *errstr, size_t len) { if ((!op) || (!errstr)) return; - sprintf (errstr, "XID: %x, %s: NFS: %d(%s), POSIX: %d(%s)", xid, op, - stat, nfsstat3_strerror (stat), pstat, strerror (pstat)); + snprintf (errstr, len, "XID: %x, %s: NFS: %d(%s), POSIX: %d(%s)", + xid, op,stat, nfsstat3_strerror (stat), pstat, + strerror (pstat)); } void @@ -1619,10 +1631,10 @@ nfs3_log_common_call (uint32_t xid, char *op, struct nfs3_fh *fh) { char fhstr[1024]; - if (gf_log_loglevel < GF_LOG_DEBUG) - return; + if (THIS->ctx->log.loglevel < GF_LOG_DEBUG) + return; - nfs3_fh_to_str (fh, fhstr); + nfs3_fh_to_str (fh, fhstr, sizeof (fhstr)); gf_log (GF_NFS3, GF_LOG_DEBUG, "XID: %x, %s: args: %s", xid, op, fhstr); } @@ -1634,9 +1646,9 @@ nfs3_log_fh_entry_call (uint32_t xid, char *op, struct nfs3_fh *fh, { char fhstr[1024]; - if (gf_log_loglevel < GF_LOG_DEBUG) - return; - nfs3_fh_to_str (fh, fhstr); + 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); } @@ -1649,10 +1661,10 @@ nfs3_log_rename_call (uint32_t xid, struct nfs3_fh *src, char *sname, char sfhstr[1024]; char dfhstr[1024]; - if (gf_log_loglevel < GF_LOG_DEBUG) - return; - nfs3_fh_to_str (src, sfhstr); - nfs3_fh_to_str (dst, dfhstr); + 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); @@ -1670,9 +1682,9 @@ nfs3_log_create_call (uint32_t xid, struct nfs3_fh *fh, char *name, char unchkd[] = "UNCHECKED"; char guarded[] = "GUARDED"; - if (gf_log_loglevel < GF_LOG_DEBUG) - return; - nfs3_fh_to_str (fh, fhstr); + 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) @@ -1695,9 +1707,9 @@ nfs3_log_mknod_call (uint32_t xid, struct nfs3_fh *fh, char *name, int type) char sock[] = "SOCK"; char fifo[] = "FIFO"; - if (gf_log_loglevel < GF_LOG_DEBUG) - return; - nfs3_fh_to_str (fh, fhstr); + 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) @@ -1718,9 +1730,9 @@ nfs3_log_symlink_call (uint32_t xid, struct nfs3_fh *fh, char *name, char *tgt) { char fhstr[1024]; - if (gf_log_loglevel < GF_LOG_DEBUG) - return; - nfs3_fh_to_str (fh, fhstr); + 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); } @@ -1733,10 +1745,10 @@ nfs3_log_link_call (uint32_t xid, struct nfs3_fh *fh, char *name, char dfhstr[1024]; char tfhstr[1024]; - if (gf_log_loglevel < GF_LOG_DEBUG) - return; - nfs3_fh_to_str (fh, dfhstr); - nfs3_fh_to_str (tgt, tfhstr); + 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); } @@ -1748,9 +1760,9 @@ nfs3_log_rw_call (uint32_t xid, char *op, struct nfs3_fh *fh, offset3 offt, { char fhstr[1024]; - if (gf_log_loglevel < GF_LOG_DEBUG) - return; - nfs3_fh_to_str (fh, fhstr); + 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, @@ -3381,11 +3393,11 @@ void nfs3_log_common_res (uint32_t xid, int op, nfsstat3 stat, int pstat) { char errstr[1024]; - int ll = nfs3_loglevel (op, stat); + int ll = nfs3_loglevel (op, stat); - if (gf_log_loglevel < ll) - return; - nfs3_stat_to_errstr (xid, nfs3op_strings[op].str, stat, pstat, errstr); + 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); } @@ -3393,14 +3405,14 @@ void nfs3_log_readlink_res (uint32_t xid, nfsstat3 stat, int pstat, char *linkpath) { char errstr[1024]; - int ll = nfs3_loglevel (NFS3_READLINK, stat); + int ll = nfs3_loglevel (NFS3_READLINK, stat); - if (gf_log_loglevel < ll) - return; + if (THIS->ctx->log.loglevel < ll) + return; - nfs3_stat_to_errstr (xid, "READLINK", stat, pstat, errstr); + nfs3_stat_to_errstr (xid, "READLINK", stat, pstat, errstr, sizeof (errstr)); gf_log (GF_NFS3, ll, "%s, target: %s", - errstr, linkpath); + errstr, linkpath); } @@ -3409,12 +3421,12 @@ 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; + int ll = GF_LOG_DEBUG; ll = nfs3_loglevel (NFS3_READ, stat); - if (gf_log_loglevel < ll) - return; - nfs3_stat_to_errstr (xid, "READ", stat, pstat, errstr); + 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, @@ -3430,12 +3442,12 @@ 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); + int ll = nfs3_loglevel (NFS3_WRITE, stat); - if (gf_log_loglevel < ll) - return; + if (THIS->ctx->log.loglevel < ll) + return; - nfs3_stat_to_errstr (xid, "WRITE", stat, pstat, errstr); + 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); @@ -3448,12 +3460,12 @@ nfs3_log_newfh_res (uint32_t xid, int op, nfsstat3 stat, int pstat, { char errstr[1024]; char fhstr[1024]; - int ll = nfs3_loglevel (op, stat); + int ll = nfs3_loglevel (op, stat); - if (gf_log_loglevel < ll) - return; - nfs3_stat_to_errstr (xid, nfs3op_strings[op].str, stat, pstat, errstr); - nfs3_fh_to_str (newfh, fhstr); + 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); } @@ -3464,11 +3476,11 @@ 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); + int ll = nfs3_loglevel (NFS3_READDIR, stat); - if (gf_log_loglevel < ll) - return; - nfs3_stat_to_errstr (xid, "READDIR", stat, pstat, errstr); + 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); } @@ -3481,9 +3493,9 @@ nfs3_log_readdirp_res (uint32_t xid, nfsstat3 stat, int pstat, uint64_t cverf, char errstr[1024]; int ll = nfs3_loglevel (NFS3_READDIRP, stat); - if (gf_log_loglevel < ll) - return; - nfs3_stat_to_errstr (xid, "READDIRPLUS", stat, pstat, errstr); + 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); @@ -3496,9 +3508,9 @@ 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 (gf_log_loglevel < ll) - return; - nfs3_stat_to_errstr (xid, "COMMIT", stat, pstat, errstr); + 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); } @@ -3509,10 +3521,10 @@ nfs3_log_readdir_call (uint32_t xid, struct nfs3_fh *fh, count3 dircount, { char fhstr[1024]; - if (gf_log_loglevel < GF_LOG_DEBUG) - return; + if (THIS->ctx->log.loglevel < GF_LOG_DEBUG) + return; - nfs3_fh_to_str (fh, fhstr); + nfs3_fh_to_str (fh, fhstr, sizeof (fhstr)); if (maxcount == 0) gf_log (GF_NFS3, GF_LOG_DEBUG, "XID: %x, READDIR: args: %s," @@ -3575,7 +3587,8 @@ nfs3_fh_resolve_entry_lookup_cbk (call_frame_t *frame, void *cookie, cs->resolvedloc.name, buf); if (linked_inode) { inode_lookup (linked_inode); - inode_unref (linked_inode); + inode_unref (cs->resolvedloc.inode); + cs->resolvedloc.inode = linked_inode; } err: nfs3_call_resume (cs); diff --git a/xlators/nfs/server/src/nfs3-helpers.h b/xlators/nfs/server/src/nfs3-helpers.h index 67935d143..4de1d5623 100644 --- a/xlators/nfs/server/src/nfs3-helpers.h +++ b/xlators/nfs/server/src/nfs3-helpers.h @@ -2,19 +2,10 @@ Copyright (c) 2010-2011 Gluster, Inc. <http://www.gluster.com> This file is part of GlusterFS. - GlusterFS is free software; you can redistribute it and/or modify - it under the terms of the GNU General Public License as published - by the Free Software Foundation; either version 3 of the License, - or (at your option) any later version. - - GlusterFS is distributed in the hope that it will be useful, but - WITHOUT ANY WARRANTY; without even the implied warranty of - MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU - General Public License for more details. - - You should have received a copy of the GNU General Public License - along with this program. If not, see - <http://www.gnu.org/licenses/>. + 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_ @@ -44,6 +35,9 @@ 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, @@ -99,7 +93,8 @@ extern void nfs3_prep_access3args (access3args *args, struct nfs3_fh *fh); extern void -nfs3_fill_access3res (access3res *res, nfsstat3 status, int32_t accbits); +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); diff --git a/xlators/nfs/server/src/nfs3.c b/xlators/nfs/server/src/nfs3.c index 75375f369..f914c3193 100644 --- a/xlators/nfs/server/src/nfs3.c +++ b/xlators/nfs/server/src/nfs3.c @@ -2,19 +2,10 @@ Copyright (c) 2010-2011 Gluster, Inc. <http://www.gluster.com> This file is part of GlusterFS. - GlusterFS is free software; you can redistribute it and/or modify - it under the terms of the GNU General Public License as published - by the Free Software Foundation; either version 3 of the License, - or (at your option) any later version. - - GlusterFS is distributed in the hope that it will be useful, but - WITHOUT ANY WARRANTY; without even the implied warranty of - MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU - General Public License for more details. - - You should have received a copy of the GNU General Public License - along with this program. If not, see - <http://www.gnu.org/licenses/>. + 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 @@ -32,6 +23,7 @@ #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" @@ -225,32 +217,32 @@ out: uuid_unparse (handle->exportid, exportid); \ uuid_unparse (handle->gfid, gfid); \ trans = rpcsvc_request_transport (req); \ - gf_log (GF_NFS3, GF_LOG_ERROR, "Failed to map " \ - "FH to vol: client=%s, exportid=%s, gfid=%s",\ - trans->peerinfo.identifier, exportid, \ - gfid); \ - gf_log (GF_NFS3, GF_LOG_ERROR, \ - "Stale nfs client %s must be trying to "\ - "connect to a deleted volume, please " \ - "unmount it.", trans->peerinfo.identifier);\ + 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); \ + 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 ((handle)) { \ - if (!nfs3_fh_validate (handle)) { \ - gf_log (GF_NFS3, GF_LOG_ERROR, "Bad Handle");\ - status = NFS3ERR_BADHANDLE; \ - goto errlabel; \ - } \ + if (!nfs3_fh_validate (handle)) { \ + gf_log (GF_NFS3, GF_LOG_ERROR, "Bad Handle"); \ + status = NFS3ERR_BADHANDLE; \ + goto errlabel; \ } \ } while (0) \ @@ -264,10 +256,12 @@ out: xlatorp = nfs3_fh_to_xlator (cst->nfs3state, \ &cst->resolvefh); \ uuid_unparse (cst->resolvefh.gfid, gfid); \ - sprintf (buf, "(%s) %s : %s", trans->peerinfo.identifier,\ - xlatorp ? xlatorp->name : "ERR", gfid); \ - gf_log (GF_NFS3, GF_LOG_ERROR, "Unable to resolve FH"\ - ": %s", buf); \ + 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; \ } \ @@ -284,10 +278,12 @@ out: xlatorp = nfs3_fh_to_xlator (cst->nfs3state, \ &cst->resolvefh); \ uuid_unparse (cst->resolvefh.gfid, gfid); \ - sprintf (buf, "(%s) %s : %s", trans->peerinfo.identifier,\ - xlatorp ? xlatorp->name : "ERR", gfid); \ - gf_log (GF_NFS3, GF_LOG_ERROR, "Unable to resolve FH"\ - ": %s", buf); \ + 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; \ } \ @@ -448,11 +444,9 @@ nfs3_call_state_wipe (nfs3_call_state_t *cs) fd_unref (cs->fd); } - if (cs->resolventry) - GF_FREE (cs->resolventry); + GF_FREE (cs->resolventry); - if (cs->pathname) - GF_FREE (cs->pathname); + GF_FREE (cs->pathname); if (!list_empty (&cs->entries.list)) gf_dirent_free (&cs->entries); @@ -555,23 +549,18 @@ nfs3svc_submit_reply (rpcsvc_request_t *req, void *arg, nfs3_serializer sfunc) iobref = iobref_new (); if (!iobref) { - iobuf_unref (iob); gf_log (GF_NFS3, GF_LOG_ERROR, "failed on iobref_new()"); goto ret; } - iobref_add (iobref, iob); + 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); - - /* 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. - */ - iobuf_unref (iob); - iobref_unref (iobref); - if (ret == -1) { gf_log (GF_NFS3, GF_LOG_ERROR, "Reply submission failed"); goto ret; @@ -579,6 +568,14 @@ nfs3svc_submit_reply (rpcsvc_request_t *req, void *arg, nfs3_serializer sfunc) 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; } @@ -610,19 +607,14 @@ nfs3svc_submit_vector_reply (rpcsvc_request_t *req, void *arg, new_iobref = 1; } - iobref_add (iobref, iob); + 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); - - /* 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. - */ - iobuf_unref (iob); - if (new_iobref) - iobref_unref (iobref); - if (ret == -1) { gf_log (GF_NFS3, GF_LOG_ERROR, "Reply submission failed"); goto ret; @@ -630,6 +622,14 @@ nfs3svc_submit_vector_reply (rpcsvc_request_t *req, void *arg, 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; } @@ -694,11 +694,20 @@ nfs3svc_getattr_lookup_cbk (call_frame_t *frame, void *cookie, xlator_t *this, 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_errno_to_nfsstat3 (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, @@ -725,7 +734,7 @@ nfs3svc_getattr_stat_cbk (call_frame_t *frame, void *cookie, xlator_t *this, gf_log (GF_NFS, GF_LOG_WARNING, "%x: %s => -1 (%s)", rpcsvc_request_xid (cs->req), cs->resolvedloc.path, strerror (op_errno)); - status = nfs3_errno_to_nfsstat3 (op_errno); + status = nfs3_cbk_errno_status (op_ret, op_errno); } nfs3_log_common_res (rpcsvc_request_xid (cs->req), NFS3_GETATTR, @@ -745,6 +754,9 @@ nfs3_getattr_resume (void *carg) 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; @@ -771,9 +783,28 @@ nfs3_getattr_resume (void *carg) 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)); @@ -887,7 +918,7 @@ nfs3svc_truncate_cbk (call_frame_t *frame, void *cookie, xlator_t *this, gf_log (GF_NFS, GF_LOG_WARNING, "%x: %s => -1 (%s)", rpcsvc_request_xid (cs->req), cs->resolvedloc.path, strerror (op_errno)); - stat = nfs3_errno_to_nfsstat3 (op_errno); + stat = nfs3_cbk_errno_status (op_ret, op_errno); goto nfs3err; } @@ -927,7 +958,7 @@ nfs3svc_setattr_cbk (call_frame_t *frame, void *cookie, xlator_t *this, gf_log (GF_NFS, GF_LOG_WARNING, "%x: %s => -1 (%s)", rpcsvc_request_xid (cs->req), cs->resolvedloc.path, strerror (op_errno)); - stat = nfs3_errno_to_nfsstat3 (op_errno); + stat = nfs3_cbk_errno_status (op_ret, op_errno); goto nfs3err; } @@ -941,7 +972,8 @@ nfs3svc_setattr_cbk (call_frame_t *frame, void *cookie, xlator_t *this, * truncation and also only if this is not a directory. */ if ((gf_attr_size_set (cs->setattr_valid)) && - (!IA_ISDIR (postop->ia_type))) { + (!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); @@ -982,7 +1014,7 @@ nfs3svc_setattr_stat_cbk (call_frame_t *frame, void *cookie, xlator_t *this, gf_log (GF_NFS, GF_LOG_WARNING, "%x: %s => -1 (%s)", rpcsvc_request_xid (cs->req), cs->resolvedloc.path, strerror (op_errno)); - stat = nfs3_errno_to_nfsstat3 (op_errno); + stat = nfs3_cbk_errno_status (op_ret, op_errno); goto nfs3err; } @@ -1196,7 +1228,7 @@ nfs3svc_lookup_cbk (call_frame_t *frame, void *cookie, xlator_t *this, (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_errno_to_nfsstat3 (op_errno); + status = nfs3_cbk_errno_status (op_ret, op_errno); goto xmit_res; } @@ -1240,7 +1272,7 @@ nfs3svc_lookup_parentdir_cbk (call_frame_t *frame, void *cookie, xlator_t *this, gf_log (GF_NFS, GF_LOG_WARNING, "%x: %s => -1 (%s)", rpcsvc_request_xid (cs->req), cs->resolvedloc.path, strerror (op_errno)); - status = nfs3_errno_to_nfsstat3 (op_errno); + status = nfs3_cbk_errno_status (op_ret, op_errno); goto xmit_res; } @@ -1282,7 +1314,11 @@ nfs3_lookup_parentdir_resume (void *carg) nfs3_call_state_t *cs = NULL; inode_t *parent = NULL; - GF_VALIDATE_OR_GOTO (GF_NFS3, carg, nfs3err); + 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); @@ -1353,10 +1389,16 @@ nfs3_lookup_resume (void *carg) nfs3_call_state_t *cs = NULL; struct nfs3_fh newfh = {{0},}; - GF_VALIDATE_OR_GOTO (GF_NFS3, carg, nfs3err); + 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); @@ -1364,7 +1406,6 @@ nfs3_lookup_resume (void *carg) } nfs_request_user_init (&nfu, cs->req); - cs->parent = cs->resolvefh; ret = nfs_lookup (cs->nfsx, cs->vol, &nfu, &cs->resolvedloc, nfs3svc_lookup_cbk, cs); if (ret < 0) @@ -1409,12 +1450,8 @@ nfs3_lookup (rpcsvc_request_t *req, struct nfs3_fh *fh, int fhlen, char *name) nfs3_handle_call_state_init (nfs3, cs, req, vol, stat, nfs3err); cs->lookuptype = GF_NFS3_REVALIDATE; - if (!nfs3_is_parentdir_entry (name)) - ret = nfs3_fh_resolve_and_resume (cs, fh, name, - nfs3_lookup_resume); - else - ret = nfs3_fh_resolve_and_resume (cs, fh, NULL, - nfs3_lookup_parentdir_resume); + 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"); @@ -1468,11 +1505,12 @@ rpcerr: int -nfs3_access_reply (rpcsvc_request_t *req, nfsstat3 status, int32_t accbits) +nfs3_access_reply (rpcsvc_request_t *req, nfsstat3 status, int32_t accbits, + int32_t reqaccbits) { access3res res; - nfs3_fill_access3res (&res, status, accbits); + nfs3_fill_access3res (&res, status, accbits, reqaccbits); nfs3svc_submit_reply (req, &res, (nfs3_serializer)xdr_serialize_access3res); return 0; @@ -1492,11 +1530,11 @@ nfs3svc_access_cbk (call_frame_t *frame, void *cookie, xlator_t *this, gf_log (GF_NFS, GF_LOG_WARNING, "%x: %s => -1 (%s)", rpcsvc_request_xid (cs->req), cs->resolvedloc.path, strerror (op_errno)); - status = nfs3_errno_to_nfsstat3 (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); + nfs3_access_reply (cs->req, status, op_errno, cs->accessbits); nfs3_call_state_wipe (cs); return 0; @@ -1510,7 +1548,11 @@ nfs3_access_resume (void *carg) nfs_user_t nfu = {0, }; nfs3_call_state_t *cs = NULL; - GF_VALIDATE_OR_GOTO (GF_NFS3, carg, nfs3err); + 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); @@ -1525,7 +1567,7 @@ nfs3err: if (ret < 0) { nfs3_log_common_res (rpcsvc_request_xid (cs->req), NFS3_ACCESS, stat, -ret); - nfs3_access_reply (cs->req, stat, 0); + nfs3_access_reply (cs->req, stat, 0, 0); nfs3_call_state_wipe (cs); ret = 0; } @@ -1561,7 +1603,7 @@ nfs3err: if (ret < 0) { nfs3_log_common_res (rpcsvc_request_xid (req), NFS3_ACCESS, stat, -ret); - nfs3_access_reply (req, stat, 0); + nfs3_access_reply (req, stat, 0, 0); nfs3_call_state_wipe (cs); ret = 0; } @@ -1628,7 +1670,7 @@ nfs3svc_readlink_cbk (call_frame_t *frame, void *cookie, xlator_t *this, gf_log (GF_NFS, GF_LOG_WARNING, "%x: %s => -1 (%s)", rpcsvc_request_xid (cs->req), cs->resolvedloc.path, strerror (op_errno)); - stat = nfs3_errno_to_nfsstat3 (op_errno); + stat = nfs3_cbk_errno_status (op_ret, op_errno); goto nfs3err; } @@ -1795,7 +1837,7 @@ nfs3svc_read_cbk (call_frame_t *frame, void *cookie, xlator_t *this, gf_log (GF_NFS, GF_LOG_WARNING, "%x: %s => -1 (%s)", rpcsvc_request_xid (cs->req), cs->resolvedloc.path, strerror (op_errno)); - stat = nfs3_errno_to_nfsstat3 (op_errno); + stat = nfs3_cbk_errno_status (op_ret, op_errno); goto err; } else stat = NFS3_OK; @@ -1981,7 +2023,7 @@ nfs3svc_write_fsync_cbk (call_frame_t *frame, void *cookie, xlator_t *this, gf_log (GF_NFS, GF_LOG_WARNING, "%x: %s => -1 (%s)", rpcsvc_request_xid (cs->req), cs->resolvedloc.path, strerror (op_errno)); - stat = nfs3_errno_to_nfsstat3 (op_errno); + stat = nfs3_cbk_errno_status (op_ret, op_errno); } else stat = NFS3_OK; @@ -1994,41 +2036,6 @@ nfs3svc_write_fsync_cbk (call_frame_t *frame, void *cookie, xlator_t *this, } -/* - * If this logic determines that the write should return a reply to the client - * after this function, the return value is -1 and the writetype is reset to - * the type of write we want to signify to the client. - * - * In case the write should continue to serve the request according to the type - * of stable write, a 0 is returned and writetype is left as it is. - */ -int -nfs3_write_how (int *writetype, int write_trusted, int sync_trusted) -{ - int ret = -1; - - if (*writetype == UNSTABLE) { - /* On an UNSTABLE write, only return STABLE when trusted-write - * is set. TW is also set when trusted-sync is set. - */ - if (write_trusted) - *writetype = FILE_SYNC; - - goto err; - } else if ((*writetype == DATA_SYNC) || (*writetype == FILE_SYNC)) { - - /* On a STABLE write, if sync-trusted is on, only then, return - * without syncing. - */ - if (sync_trusted) - goto err; - } - - ret = 0; -err: - return ret; -} - /* * Before going into the write reply logic, here is a matrix that shows the @@ -2067,12 +2074,8 @@ nfs3svc_write_cbk (call_frame_t *frame, void *cookie, xlator_t *this, struct iatt *postbuf, dict_t *xdata) { nfsstat3 stat = NFS3ERR_SERVERFAULT; - int ret = -EFAULT; - nfs_user_t nfu = {0, }; nfs3_call_state_t *cs = NULL; struct nfs3_state *nfs3 = NULL; - int write_trusted = 0; - int sync_trusted = 0; cs = frame->local; nfs3 = rpcsvc_request_program_private (cs->req); @@ -2080,42 +2083,21 @@ nfs3svc_write_cbk (call_frame_t *frame, void *cookie, xlator_t *this, gf_log (GF_NFS, GF_LOG_WARNING, "%x: %s => -1 (%s)", rpcsvc_request_xid (cs->req), cs->resolvedloc.path, strerror (op_errno)); - stat = nfs3_errno_to_nfsstat3 (op_errno); + stat = nfs3_cbk_errno_status (op_ret, op_errno); goto err; } stat = NFS3_OK; cs->maxcount = op_ret; - write_trusted = nfs3_export_write_trusted (cs->nfs3state, - cs->resolvefh.exportid); - sync_trusted = nfs3_export_sync_trusted (cs->nfs3state, - cs->resolvefh.exportid); - ret = nfs3_write_how (&cs->writetype, write_trusted, sync_trusted); - if (ret == -1) - goto err; - - nfs_request_user_init (&nfu, cs->req); - /* Store the current preattr so that this can be used as the pre attr - * when fsync returns. We dont want to use the preattr in fsync because - * the write fop happened before the fsync. - */ - cs->stbuf = *prebuf; - ret = nfs_fsync (cs->nfsx, cs->vol, &nfu, cs->fd, 0, - nfs3svc_write_fsync_cbk, cs); - if (ret < 0) - stat = nfs3_errno_to_nfsstat3 (-ret); - err: - if (ret < 0) { - 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); - } + 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; } @@ -2171,6 +2153,25 @@ nfs3_write_resume (void *carg) } 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); @@ -2241,17 +2242,18 @@ out: int -nfs3svc_write_vecsizer (int state, ssize_t *readsize, char *addr) +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; + 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 *)(addr - 4); + fhlen_n = *(uint32_t *)(curr_addr - 4); fhlen = ntohl (fhlen_n); *readsize = xdr_length_round_up (fhlen, NFS3_FHSIZE); ret = NFS3_VECWRITE_READFH; @@ -2302,12 +2304,6 @@ rpcerr: return ret; } -int -nfs3svc_write_vec (rpcsvc_request_t *req, struct iovec *payload, - int payload_count, struct iobref *iobref) -{ - return nfs3svc_write (req); -} int nfs3_create_reply (rpcsvc_request_t *req, nfsstat3 stat, struct nfs3_fh *newfh, @@ -2339,7 +2335,7 @@ nfs3svc_create_setattr_cbk (call_frame_t *frame, void *cookie, xlator_t *this, gf_log (GF_NFS, GF_LOG_WARNING, "%x: %s => -1 (%s)", rpcsvc_request_xid (cs->req), cs->resolvedloc.path, strerror (op_errno)); - stat = nfs3_errno_to_nfsstat3 (op_errno); + stat = nfs3_cbk_errno_status (op_ret, op_errno); goto nfs3err; } @@ -2372,7 +2368,7 @@ nfs3svc_create_cbk (call_frame_t *frame, void *cookie, xlator_t *this, gf_log (GF_NFS, GF_LOG_WARNING, "%x: %s => -1 (%s)", rpcsvc_request_xid (cs->req), cs->resolvedloc.path, strerror (op_errno)); - stat = nfs3_errno_to_nfsstat3 (op_errno); + stat = nfs3_cbk_errno_status (op_ret, op_errno); goto nfs3err; } @@ -2477,7 +2473,7 @@ nfs3svc_create_stat_cbk (call_frame_t *frame, void *cookie, xlator_t *this, "%x: %s => -1 (%s)", rpcsvc_request_xid (cs->req), cs->resolvedloc.path, strerror (op_errno)); ret = -op_errno; - stat = nfs3_errno_to_nfsstat3 (op_errno); + stat = nfs3_cbk_errno_status (op_ret, op_errno); goto nfs3err; } @@ -2648,13 +2644,7 @@ nfs3svc_create (rpcsvc_request_t *req) } cval = (uint64_t *)args.how.createhow3_u.verf; - if (cval) - cverf = *cval; - else { - gf_log(GF_NFS3, GF_LOG_ERROR, - "Error getting createverf3 from args"); - goto rpcerr; - } + cverf = *cval; ret = nfs3_create (req, &dirfh, name, args.how.mode, &args.how.createhow3_u.obj_attributes, cverf); @@ -2698,7 +2688,7 @@ nfs3svc_mkdir_setattr_cbk (call_frame_t *frame, void *cookie, xlator_t *this, gf_log (GF_NFS, GF_LOG_WARNING, "%x: %s => -1 (%s)", rpcsvc_request_xid (cs->req), cs->resolvedloc.path, strerror (op_errno)); - stat = nfs3_errno_to_nfsstat3 (op_errno); + stat = nfs3_cbk_errno_status (op_ret, op_errno); goto nfs3err; } @@ -2730,7 +2720,7 @@ nfs3svc_mkdir_cbk (call_frame_t *frame, void *cookie, xlator_t *this, gf_log (GF_NFS, GF_LOG_WARNING, "%x: %s => -1 (%s)", rpcsvc_request_xid (cs->req), cs->resolvedloc.path, strerror (op_errno)); - stat = nfs3_errno_to_nfsstat3 (op_errno); + stat = nfs3_cbk_errno_status (op_ret, op_errno); goto nfs3err; } @@ -2908,7 +2898,7 @@ nfs3svc_symlink_cbk (call_frame_t *frame, void *cookie, xlator_t *this, gf_log (GF_NFS, GF_LOG_WARNING, "%x: %s => -1 (%s)", rpcsvc_request_xid (cs->req), cs->resolvedloc.path, strerror (op_errno)); - stat = nfs3_errno_to_nfsstat3 (op_errno); + stat = nfs3_cbk_errno_status (op_ret, op_errno); goto nfs3err; } @@ -3070,7 +3060,7 @@ nfs3svc_mknod_setattr_cbk (call_frame_t *frame, void *cookie, xlator_t *this, gf_log (GF_NFS, GF_LOG_WARNING, "%x: %s => -1 (%s)", rpcsvc_request_xid (cs->req), cs->resolvedloc.path, strerror (op_errno)); - stat = nfs3_errno_to_nfsstat3 (op_errno); + stat = nfs3_cbk_errno_status (op_ret, op_errno); goto nfs3err; } @@ -3102,7 +3092,7 @@ nfs3svc_mknod_cbk (call_frame_t *frame, void *cookie, xlator_t *this, gf_log (GF_NFS, GF_LOG_WARNING, "%x: %s => -1 (%s)", rpcsvc_request_xid (cs->req), cs->resolvedloc.path, strerror (op_errno)); - stat = nfs3_errno_to_nfsstat3 (op_errno); + stat = nfs3_cbk_errno_status (op_ret, op_errno); goto nfs3err; } @@ -3360,7 +3350,7 @@ nfs3svc_remove_cbk (call_frame_t *frame, void *cookie, xlator_t *this, gf_log (GF_NFS, GF_LOG_WARNING, "%x: %s => -1 (%s)", rpcsvc_request_xid (cs->req), cs->resolvedloc.path, strerror (op_errno)); - stat = nfs3_errno_to_nfsstat3 (op_errno); + stat = nfs3_cbk_errno_status (op_ret, op_errno); } if (op_ret == 0) @@ -3526,7 +3516,7 @@ nfs3svc_rmdir_cbk (call_frame_t *frame, void *cookie, xlator_t *this, gf_log (GF_NFS, GF_LOG_WARNING, "%x: %s => -1 (%s)", rpcsvc_request_xid (cs->req), cs->resolvedloc.path, strerror (op_errno)); - stat = nfs3_errno_to_nfsstat3 (op_errno); + stat = nfs3_cbk_errno_status (op_ret, op_errno); } else { stat = NFS3_OK; } @@ -3681,7 +3671,7 @@ nfs3svc_rename_cbk (call_frame_t *frame, void *cookie, xlator_t *this, "%x: rename %s -> %s => -1 (%s)", rpcsvc_request_xid (cs->req), cs->oploc.path, cs->resolvedloc.path, strerror (op_errno)); - stat = nfs3_errno_to_nfsstat3 (op_errno); + stat = nfs3_cbk_errno_status (op_ret, op_errno); goto nfs3err; } @@ -3886,7 +3876,7 @@ nfs3svc_link_cbk (call_frame_t *frame, void *cookie, xlator_t *this, "%x: link %s <- %s => -1 (%s)", rpcsvc_request_xid (cs->req), cs->oploc.path, cs->resolvedloc.path, strerror (op_errno)); - stat = nfs3_errno_to_nfsstat3 (op_errno); + stat = nfs3_cbk_errno_status (op_ret, op_errno); } else stat = NFS3_OK; @@ -4096,7 +4086,7 @@ nfs3svc_readdir_fstat_cbk (call_frame_t *frame, void *cookie, xlator_t *this, gf_log (GF_NFS, GF_LOG_WARNING, "%x: %s => -1 (%s)", rpcsvc_request_xid (cs->req), cs->resolvedloc.path, strerror (op_errno)); - stat = nfs3_errno_to_nfsstat3 (op_errno); + stat = nfs3_cbk_errno_status (op_ret, op_errno); goto nfs3err; } @@ -4151,7 +4141,7 @@ nfs3svc_readdir_cbk (call_frame_t *frame, void *cookie, xlator_t *this, gf_log (GF_NFS, GF_LOG_WARNING, "%x: %s => -1 (%s)", rpcsvc_request_xid (cs->req), cs->resolvedloc.path, strerror (op_errno)); - stat = nfs3_errno_to_nfsstat3 (op_errno); + stat = nfs3_cbk_errno_status (op_ret, op_errno); goto err; } @@ -4248,12 +4238,30 @@ nfs3err: } +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; @@ -4266,6 +4274,22 @@ nfs3_readdir_open_resume (void *carg) 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); @@ -4365,16 +4389,8 @@ nfs3svc_readdir (rpcsvc_request_t *req) rpcsvc_request_seterr (req, GARBAGE_ARGS); goto rpcerr; } - cval = (uint64_t *) ra.cookieverf; - - if (cval) - verf = *cval; - else { - gf_log(GF_NFS3, GF_LOG_ERROR, - "Error getting cookieverf from readdir args"); - goto rpcerr; - } + verf = *cval; ret = nfs3_readdir (req, &fh, ra.cookie, verf, ra.count, 0); if ((ret < 0) && (ret != RPCSVC_ACTOR_IGNORE)) { @@ -4405,16 +4421,8 @@ nfs3svc_readdirp (rpcsvc_request_t *req) rpcsvc_request_seterr (req, GARBAGE_ARGS); goto rpcerr; } - cval = (uint64_t *) ra.cookieverf; - - if (cval) - cverf = *cval; - else { - gf_log (GF_NFS3, GF_LOG_ERROR, - "Error getting cookieverf from readdirp args"); - goto rpcerr; - } + cverf = *cval; ret = nfs3_readdir (req, &fh, ra.cookie, cverf, ra.dircount, ra.maxcount); @@ -4457,7 +4465,7 @@ nfs3_fsstat_stat_cbk (call_frame_t *frame, void *cookie, xlator_t *this, gf_log (GF_NFS, GF_LOG_WARNING, "%x: %s => -1 (%s)", rpcsvc_request_xid (cs->req), cs->resolvedloc.path, strerror (op_errno)); - stat = nfs3_errno_to_nfsstat3 (op_errno); + stat = nfs3_cbk_errno_status (op_ret, op_errno); } else stat = NFS3_OK; @@ -4485,7 +4493,7 @@ nfs3_fsstat_statfs_cbk (call_frame_t *frame, void *cookie, xlator_t *this, "%x: %s => -1 (%s)", rpcsvc_request_xid (cs->req), cs->resolvedloc.path, strerror (op_errno)); ret = -op_errno; - stat = nfs3_errno_to_nfsstat3 (op_errno); + stat = nfs3_cbk_errno_status (op_ret, op_errno); goto err; } @@ -4644,7 +4652,7 @@ nfs3svc_fsinfo_cbk (call_frame_t *frame, void *cookie, xlator_t *this, gf_log (GF_NFS, GF_LOG_WARNING, "%x: %s => -1 (%s)", rpcsvc_request_xid (cs->req), cs->resolvedloc.path, strerror (op_errno)); - status = nfs3_errno_to_nfsstat3 (op_errno); + status = nfs3_cbk_errno_status (op_ret, op_errno); }else status = NFS3_OK; @@ -4786,7 +4794,7 @@ nfs3svc_pathconf_cbk (call_frame_t *frame, void *cookie, xlator_t *this, gf_log (GF_NFS, GF_LOG_WARNING, "%x: %s => -1 (%s)", rpcsvc_request_xid (cs->req), cs->resolvedloc.path, strerror (op_errno)); - stat = nfs3_errno_to_nfsstat3 (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. @@ -4930,7 +4938,7 @@ nfs3svc_commit_cbk (call_frame_t *frame, void *cookie, xlator_t *this, gf_log (GF_NFS, GF_LOG_WARNING, "%x: %s => -1 (%s)", rpcsvc_request_xid (cs->req), cs->resolvedloc.path, strerror (op_errno)); - stat = nfs3_errno_to_nfsstat3 (op_errno); + stat = nfs3_cbk_errno_status (op_ret, op_errno); } else stat = NFS3_OK; @@ -5091,28 +5099,28 @@ rpcerr: rpcsvc_actor_t nfs3svc_actors[NFS3_PROC_COUNT] = { - {"NULL", NFS3_NULL, nfs3svc_null, NULL, NULL, 0}, - {"GETATTR", NFS3_GETATTR, nfs3svc_getattr,NULL, NULL, 0}, - {"SETATTR", NFS3_SETATTR, nfs3svc_setattr,NULL, NULL, 0}, - {"LOOKUP", NFS3_LOOKUP, nfs3svc_lookup, NULL, NULL, 0}, - {"ACCESS", NFS3_ACCESS, nfs3svc_access, NULL, NULL, 0}, - {"READLINK", NFS3_READLINK, nfs3svc_readlink,NULL, NULL, 0}, - {"READ", NFS3_READ, nfs3svc_read, NULL, NULL, 0}, - {"WRITE", NFS3_WRITE, nfs3svc_write, nfs3svc_write_vec, nfs3svc_write_vecsizer, 0}, - {"CREATE", NFS3_CREATE, nfs3svc_create, NULL, NULL, 0}, - {"MKDIR", NFS3_MKDIR, nfs3svc_mkdir, NULL, NULL, 0}, - {"SYMLINK", NFS3_SYMLINK, nfs3svc_symlink,NULL, NULL, 0}, - {"MKNOD", NFS3_MKNOD, nfs3svc_mknod, NULL, NULL, 0}, - {"REMOVE", NFS3_REMOVE, nfs3svc_remove, NULL, NULL, 0}, - {"RMDIR", NFS3_RMDIR, nfs3svc_rmdir, NULL, NULL, 0}, - {"RENAME", NFS3_RENAME, nfs3svc_rename, NULL, NULL, 0}, - {"LINK", NFS3_LINK, nfs3svc_link, NULL, NULL, 0}, - {"READDIR", NFS3_READDIR, nfs3svc_readdir,NULL, NULL, 0}, - {"READDIRPLUS", NFS3_READDIRP, nfs3svc_readdirp,NULL, NULL, 0}, - {"FSSTAT", NFS3_FSSTAT, nfs3svc_fsstat, NULL, NULL, 0}, - {"FSINFO", NFS3_FSINFO, nfs3svc_fsinfo, NULL, NULL, 0}, - {"PATHCONF", NFS3_PATHCONF, nfs3svc_pathconf,NULL, NULL, 0}, - {"COMMIT", NFS3_COMMIT, nfs3svc_commit, NULL, NULL, 0} + {"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} }; @@ -5129,21 +5137,48 @@ rpcsvc_program_t nfs3prog = { .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, xlator_t *nfsx) +nfs3_init_options (struct nfs3_state *nfs3, dict_t *options) { int ret = -1; char *optstr = NULL; uint64_t size64 = 0; - if ((!nfs3) || (!nfsx)) + if ((!nfs3) || (!options)) return -1; /* nfs3.read-size */ nfs3->readsize = GF_NFS3_RTPREF; - if (dict_get (nfsx->options, "nfs3.read-size")) { - ret = dict_get_str (nfsx->options, "nfs3.read-size", &optstr); + 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"); @@ -5152,19 +5187,21 @@ nfs3_init_options (struct nfs3_state *nfs3, xlator_t *nfsx) } ret = gf_string2bytesize (optstr, &size64); - nfs3->readsize = 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 (nfsx->options, "nfs3.write-size")) { - ret = dict_get_str (nfsx->options, "nfs3.write-size", &optstr); + 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"); @@ -5173,19 +5210,21 @@ nfs3_init_options (struct nfs3_state *nfs3, xlator_t *nfsx) } ret = gf_string2bytesize (optstr, &size64); - nfs3->writesize = 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 (nfsx->options, "nfs3.readdir-size")) { - ret = dict_get_str (nfsx->options,"nfs3.readdir-size", &optstr); + 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"); @@ -5194,15 +5233,16 @@ nfs3_init_options (struct nfs3_state *nfs3, xlator_t *nfsx) } ret = gf_string2bytesize (optstr, &size64); - nfs3->readdirsize = 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. */ @@ -5223,9 +5263,10 @@ err: return ret; } - int -nfs3_init_subvolume_options (struct nfs3_state *nfs3, struct nfs3_export *exp) +nfs3_init_subvolume_options (xlator_t *nfsx, + struct nfs3_export *exp, + dict_t *options) { int ret = -1; char *optstr = NULL; @@ -5233,14 +5274,20 @@ nfs3_init_subvolume_options (struct nfs3_state *nfs3, struct nfs3_export *exp) char *name = NULL; gf_boolean_t boolt = _gf_false; uuid_t volumeid = {0, }; - dict_t *options = NULL; - if ((!exp) || (!nfs3)) + if ((!nfsx) || (!exp)) return -1; - options = nfs3->nfsx->options; + /* 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 (nfs3->nfsx))) + if (gf_nfs_dvm_off (nfs_state (nfsx))) goto no_dvm; ret = snprintf (searchkey, 1024, "nfs3.%s.volume-id",exp->subvol->name); @@ -5406,7 +5453,7 @@ nfs3_init_subvolume (struct nfs3_state *nfs3, xlator_t *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, exp); + 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; @@ -5471,7 +5518,7 @@ nfs3_init_state (xlator_t *nfsx) } nfs = nfsx->private; - ret = nfs3_init_options (nfs3, nfsx); + ret = nfs3_init_options (nfs3, nfsx->options); if (ret == -1) { gf_log (GF_NFS3, GF_LOG_ERROR, "Failed to init options"); goto ret; @@ -5503,7 +5550,7 @@ nfs3_init_state (xlator_t *nfsx) LOCK_INIT (&nfs3->fdlrulock); nfs3->fdcount = 0; - rpcsvc_create_listeners (nfs->rpcsvc, nfsx->options, nfsx->name); + 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; @@ -5545,4 +5592,39 @@ nfs3svc_init (xlator_t *nfsx) 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 index e6b9edee5..0c35445a4 100644 --- a/xlators/nfs/server/src/nfs3.h +++ b/xlators/nfs/server/src/nfs3.h @@ -2,19 +2,10 @@ Copyright (c) 2010-2011 Gluster, Inc. <http://www.gluster.com> This file is part of GlusterFS. - GlusterFS is free software; you can redistribute it and/or modify - it under the terms of the GNU General Public License as published - by the Free Software Foundation; either version 3 of the License, - or (at your option) any later version. - - GlusterFS is distributed in the hope that it will be useful, but - WITHOUT ANY WARRANTY; without even the implied warranty of - MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU - General Public License for more details. - - You should have received a copy of the GNU General Public License - along with this program. If not, see - <http://www.gnu.org/licenses/>. + 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_ @@ -35,7 +26,8 @@ #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" @@ -47,16 +39,39 @@ /* Static values used for FSINFO -FIXME: This should be configurable */ -#define GF_NFS3_RTMAX (64 * GF_UNIT_KB) -#define GF_NFS3_RTPREF (64 * GF_UNIT_KB) -#define GF_NFS3_RTMULT (4 * GF_UNIT_KB) -#define GF_NFS3_WTMAX (64 * GF_UNIT_KB) -#define GF_NFS3_WTPREF (64 * GF_UNIT_KB) -#define GF_NFS3_WTMULT (4 * GF_UNIT_KB) -#define GF_NFS3_DTMIN (4 * GF_UNIT_KB) -#define GF_NFS3_DTPREF (64 * GF_UNIT_KB) -#define GF_NFS3_MAXFILE (1 * GF_UNIT_PB) + * 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} @@ -119,20 +134,21 @@ typedef struct nfs3_state { uint64_t serverstart; /* NFSv3 Protocol configurables */ - size_t readsize; - size_t writesize; - size_t readdirsize; + uint64_t readsize; + uint64_t writesize; + uint64_t readdirsize; /* Size of the iobufs used, depends on the sizes of the three params * above. */ - size_t iobsize; + 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 { @@ -154,6 +170,11 @@ typedef union args_ { 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; @@ -230,8 +251,17 @@ struct nfs3_local { 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) @@ -247,9 +277,9 @@ struct inode_op_queue { 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 index b0ab510ef..5c5d87412 100644 --- a/xlators/nfs/server/src/nlm4.c +++ b/xlators/nfs/server/src/nlm4.c @@ -2,19 +2,10 @@ Copyright (c) 2012 Gluster, Inc. <http://www.gluster.com> This file is part of GlusterFS. - GlusterFS is free software; you can redistribute it and/or modify - it under the terms of the GNU General Public License as published - by the Free Software Foundation; either version 3 of the License, - or (at your option) any later version. - - GlusterFS is distributed in the hope that it will be useful, but - WITHOUT ANY WARRANTY; without even the implied warranty of - MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU - General Public License for more details. - - You should have received a copy of the GNU General Public License - along with this program. If not, see - <http://www.gnu.org/licenses/>. + 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 @@ -43,10 +34,12 @@ #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. @@ -147,8 +140,10 @@ nfs3_fh_to_xlator (struct nfs3_state *nfs3, struct nfs3_fh *fh); xlatorp = nfs3_fh_to_xlator (cst->nfs3state, \ &cst->resolvefh); \ uuid_unparse (cst->resolvefh.gfid, gfid); \ - sprintf (buf, "(%s) %s : %s", trans->peerinfo.identifier,\ - xlatorp ? xlatorp->name : "ERR", 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);\ @@ -198,14 +193,42 @@ nlm4_prep_nlm4_unlockargs (nlm4_unlockargs *args, struct nfs3_fh *fh, } 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)); +} -nfsstat3 +nlm4_stats nlm4_errno_to_nlm4stat (int errnum) { nlm4_stats stat = nlm4_denied; @@ -250,10 +273,34 @@ nlm4_call_state_init (struct nfs3_state *s, rpcsvc_request_t *req) 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) { @@ -283,8 +330,6 @@ nlm_set_rpc_clnt (rpc_clnt_t *rpc_clnt, char *caller_name) nlm_client_t *nlmclnt = NULL; int nlmclnt_found = 0; int ret = -1; - rpc_clnt_t *rpc_clnt_old = NULL; - char *old_name = NULL; LOCK (&nlm_client_list_lk); list_for_each_entry (nlmclnt, &nlm_client_list, nlm_clients) { @@ -293,38 +338,27 @@ nlm_set_rpc_clnt (rpc_clnt_t *rpc_clnt, char *caller_name) 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"); + 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); } - rpc_clnt_old = nlmclnt->rpc_clnt; - old_name = nlmclnt->caller_name; - if (rpc_clnt) - nlmclnt->rpc_clnt = rpc_clnt_ref (rpc_clnt); - 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); - - if (rpc_clnt_old) { - /* cleanup the saved-frames before last unref */ - rpc_clnt_connection_cleanup (&rpc_clnt_old->conn); - - rpc_clnt_unref (rpc_clnt_old); - } - - if (old_name) - GF_FREE (old_name); return ret; } @@ -379,6 +413,7 @@ nlm_add_nlmclnt (char *caller_name) 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); @@ -392,10 +427,11 @@ 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; + 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) @@ -420,7 +456,12 @@ nlm4svc_submit_reply (rpcsvc_request_t *req, void *arg, nlm4_serializer sfunc) /* 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 = sfunc (outmsg, arg); + 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) { @@ -428,7 +469,11 @@ nlm4svc_submit_reply (rpcsvc_request_t *req, void *arg, nlm4_serializer sfunc) goto ret; } - iobref_add (iobref, iob); + 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); @@ -464,15 +509,17 @@ nlm4_file_open_cbk (call_frame_t *frame, void *cookie, xlator_t *this, return 0; } -int nsm_monitor(char *host) +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 }; - int retstat = -1; + 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; @@ -507,33 +554,41 @@ int nsm_monitor(char *host) clnt_sperrno(ret)); goto out; } - retstat = 0; + 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 retstat; + return NULL; } nlm_client_t * -nlm_get_uniq (char *caller_name) +__nlm_get_uniq (char *caller_name) { nlm_client_t *nlmclnt = NULL; - int nlmclnt_found = 0; - LOCK (&nlm_client_list_lk); + if (!caller_name) + return NULL; + list_for_each_entry (nlmclnt, &nlm_client_list, nlm_clients) { - if (!strcmp(caller_name, nlmclnt->caller_name)) { - nlmclnt_found = 1; - break; - } + 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); - if (nlmclnt_found) - return nlmclnt; - else - return NULL; + + return nlmclnt; } @@ -542,15 +597,19 @@ 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"); - cs->resolve_ret = -1; - cs->resume_fn(cs); - ret = -1; + ret = -ENOLCK; goto err; } cs->resume_fn = resume; @@ -566,21 +625,27 @@ nlm4_file_open_and_resume(nfs3_call_state_t *cs, nlm4_resume_fn_t resume) 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"); - cs->resolve_ret = -1; - cs->resume_fn(cs); - ret = -1; + 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 = 0; - frame->root->gid = 0; + frame->root->uid = rpcsvc_request_uid (cs->req); + frame->root->gid = rpcsvc_request_gid (cs->req); frame->local = cs; - STACK_WIND_COOKIE (frame, nlm4_file_open_cbk, cs->nfsx, cs->nfsx, - cs->nfsx->fops->open, &cs->resolvedloc, O_RDWR, + 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: @@ -704,7 +769,6 @@ err: int nlm4_test_fd_resume (void *carg) { - nlm4_stats stat = nlm4_denied; int ret = -EFAULT; nfs_user_t nfu = {0, }; nfs3_call_state_t *cs = NULL; @@ -714,21 +778,12 @@ nlm4_test_fd_resume (void *carg) return ret; cs = (nfs3_call_state_t *)carg; - nlm4_check_fh_resolve_status (cs, stat, nlm4err); 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); - if (ret < 0) - stat = nlm4_errno_to_nlm4stat (-ret); -nlm4err: - if (ret < 0) { - gf_log (GF_NLM, GF_LOG_ERROR, "Unable to call lk()"); - nlm4_test_reply (cs, stat, &flock); - nfs3_call_state_wipe (cs); - } return ret; } @@ -748,14 +803,15 @@ nlm4_test_resume (void *carg) 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); - if (ret < 0) - stat = nlm4_errno_to_nlm4stat (-ret); 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); } @@ -816,9 +872,9 @@ nlm4err: } rpcerr: - if (ret < 0) { + if (ret < 0) nfs3_call_state_wipe (cs); - } + return ret; } @@ -830,53 +886,65 @@ nlm4svc_send_granted_cbk (struct rpc_req *req, struct iovec *iov, int count, return 0; } -int nlm_rpcclnt_notify (struct rpc_clnt *rpc, void *mydata, - rpc_clnt_event_t fn, void *data) +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) { - nlm_condmutex_t *cm = NULL; - int ret; - cm = mydata; + 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 = pthread_cond_broadcast (&cm->cond); - if (ret!=0) - gf_log (GF_NLM, GF_LOG_ERROR, "cond_broadcast error %s", - strerror (errno)); + 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); + nlm_unset_rpc_clnt (rpc_clnt); break; } + + err: return 0; } -void -nlm4svc_send_granted (nfs3_call_state_t *cs); - void * nlm4_establish_callback (void *csarg) { - nfs3_call_state_t *cs = NULL; - struct sockaddr_storage sa; - struct sockaddr *sockaddr = NULL; - dict_t *options = NULL; - char peerip[INET6_ADDRSTRLEN+1], *portstr = NULL; - rpc_clnt_t *rpc_clnt = NULL; - int port = -1; - int ret = -1; - char *caller_name = NULL; + 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); - caller_name = cs->args.nlm4_lockargs.alock.caller_name; - nsm_monitor (caller_name); + rpc_transport_get_peeraddr (cs->trans, NULL, 0, &sock_union.storage, + sizeof (sock_union.storage)); - rpc_transport_get_peeraddr (cs->trans, NULL, 0, &sa, sizeof (sa)); - sockaddr = (struct sockaddr*) &sa; - switch (sockaddr->sa_family) { + 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" @@ -889,9 +957,11 @@ nlm4_establish_callback (void *csarg) break; */ case AF_INET: - inet_ntop (AF_INET, - &((struct sockaddr_in *)sockaddr)->sin_addr, - peerip, INET6_ADDRSTRLEN+1); + 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; @@ -899,7 +969,7 @@ nlm4_establish_callback (void *csarg) } /* looks like libc rpc supports only ipv4 */ - port = pmap_getport ((struct sockaddr_in*)sockaddr, NLM_PROGRAM, + port = pmap_getport (&sock_union.sin, NLM_PROGRAM, NLM_V4, IPPROTO_TCP); if (port == 0) { @@ -931,6 +1001,13 @@ nlm4_establish_callback (void *csarg) 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"); @@ -943,32 +1020,25 @@ nlm4_establish_callback (void *csarg) gf_log (GF_NLM, GF_LOG_ERROR, "rpc_clnt NULL"); goto err; } - nlm_condmutex_t *cm; - cm = GF_CALLOC (1, sizeof(*cm), gf_nfs_mt_nlm4_cm); - pthread_mutex_init (&cm->mutex, NULL); - pthread_cond_init (&cm->cond, NULL); - ret = rpc_clnt_register_notify (rpc_clnt, nlm_rpcclnt_notify, - cm); + + 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); - pthread_cond_wait (&cm->cond, &cm->mutex); - pthread_mutex_destroy (&cm->mutex); - GF_FREE (cm); - rpc_clnt_set_connected (&rpc_clnt->conn); - ret = nlm_set_rpc_clnt (rpc_clnt, caller_name); - if (ret == -1) { - gf_log (GF_NLM, GF_LOG_ERROR, "dict_set_ptr error"); - goto err; - } - nlm4svc_send_granted (cs); + + if (ret == -1 && EINPROGRESS == errno) + ret = 0; + err: - if (rpc_clnt) { + if (ret == -1 && rpc_clnt) { rpc_clnt_unref (rpc_clnt); } - return NULL; + + return rpc_clnt; } void @@ -981,32 +1051,28 @@ nlm4svc_send_granted (nfs3_call_state_t *cs) struct iobuf *iobuf = NULL; struct iobref *iobref = NULL; char peerip[INET6_ADDRSTRLEN+1]; - pthread_t thr; - struct sockaddr_storage sa; - struct sockaddr *sockaddr = NULL; + 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)); - rpc_transport_get_peeraddr (cs->trans, NULL, 0, &sa, sizeof (sa)); - sockaddr = (struct sockaddr*) &sa; - switch (sockaddr->sa_family) { + switch (sock_union.sa.sa_family) { case AF_INET6: - inet_ntop (AF_INET6, - &((struct sockaddr_in6 *)sockaddr)->sin6_addr, - peerip, INET6_ADDRSTRLEN+1); + inet_ntop (AF_INET6, &sock_union.sin6.sin6_addr, peerip, + INET6_ADDRSTRLEN+1); break; case AF_INET: - inet_ntop (AF_INET, - &((struct sockaddr_in *)sockaddr)->sin_addr, - peerip, INET6_ADDRSTRLEN+1); + inet_ntop (AF_INET, &sock_union.sin.sin_addr, peerip, + INET6_ADDRSTRLEN+1); + break; default: break; - /* FIXME: handle the error */ - } - - rpc_clnt = nlm_get_rpc_clnt (cs->args.nlm4_lockargs.alock.caller_name); - if (rpc_clnt == NULL) { - pthread_create (&thr, NULL, nlm4_establish_callback, (void*)cs); - return; } testargs.cookie = cs->args.nlm4_lockargs.cookie; @@ -1031,13 +1097,15 @@ nlm4svc_send_granted (nfs3_call_state_t *cs) goto ret; } - iobref_add (iobref, iobuf); + 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, + nlm4svc_send_granted_cbk, &outmsg, 1, + NULL, 0, iobref, cs->frame, NULL, 0, NULL, 0, NULL); if (ret < 0) { @@ -1224,26 +1292,33 @@ 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; - nfs3_call_state_t *cs = NULL; - int transit_cnt = -1; + 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); - transit_cnt = nlm_dec_transit_count (cs->fd, - cs->args.nlm4_lockargs.alock.caller_name); if (op_ret == -1) { if (transit_cnt == 0) - nlm_search_and_delete (cs->fd, - cs->args.nlm4_lockargs.alock.caller_name); + nlm_search_and_delete (cs->fd, caller_name); stat = nlm4_errno_to_nlm4stat (op_errno); goto err; - } else + } 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, @@ -1267,7 +1342,6 @@ nlm4_lock_fd_resume (void *carg) 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); @@ -1279,14 +1353,17 @@ nlm4_lock_fd_resume (void *carg) 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); - if (ret < 0) - stat = nlm4_errno_to_nlm4stat (-ret); 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); @@ -1301,8 +1378,8 @@ int nlm4_lock_resume (void *carg) { nlm4_stats stat = nlm4_failed; - int ret = -1; - nfs3_call_state_t *cs = NULL; + int ret = -1; + nfs3_call_state_t *cs = NULL; if (!carg) return ret; @@ -1310,12 +1387,11 @@ nlm4_lock_resume (void *carg) 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); - if (ret < 0) - stat = nlm4_errno_to_nlm4stat (-ret); 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); @@ -1324,17 +1400,16 @@ nlm4err: return ret; } - int -nlm4svc_lock (rpcsvc_request_t *req) +nlm4svc_lock_common (rpcsvc_request_t *req, int mon) { - 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}, }; + 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; @@ -1351,7 +1426,9 @@ nlm4svc_lock (rpcsvc_request_t *req) 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); @@ -1385,10 +1462,23 @@ 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) @@ -1413,7 +1503,6 @@ err: int nlm4_cancel_fd_resume (void *carg) { - nlm4_stats stat = nlm4_denied; int ret = -EFAULT; nfs_user_t nfu = {0, }; nfs3_call_state_t *cs = NULL; @@ -1423,7 +1512,6 @@ nlm4_cancel_fd_resume (void *carg) return ret; cs = (nfs3_call_state_t *)carg; - nlm4_check_fh_resolve_status (cs, stat, nlm4err); nfs_request_user_init (&nfu, cs->req); nlm4_lock_to_gf_flock (&flock, &cs->args.nlm4_cancargs.alock, cs->args.nlm4_cancargs.exclusive); @@ -1432,16 +1520,105 @@ nlm4_cancel_fd_resume (void *carg) ret = nfs_lk (cs->nfsx, cs->vol, &nfu, cs->fd, F_SETLK, &flock, nlm4svc_cancel_cbk, cs); - if (ret < 0) + 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 call lk()"); + 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; } @@ -1473,7 +1650,6 @@ err: int nlm4_unlock_fd_resume (void *carg) { - nlm4_stats stat = nlm4_denied; int ret = -EFAULT; nfs_user_t nfu = {0, }; nfs3_call_state_t *cs = NULL; @@ -1482,7 +1658,6 @@ nlm4_unlock_fd_resume (void *carg) if (!carg) return ret; cs = (nfs3_call_state_t *)carg; - nlm4_check_fh_resolve_status (cs, stat, nlm4err); 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); @@ -1490,21 +1665,11 @@ nlm4_unlock_fd_resume (void *carg) ret = nfs_lk (cs->nfsx, cs->vol, &nfu, cs->fd, F_SETLK, &flock, nlm4svc_unlock_cbk, cs); - if (ret < 0) - stat = nlm4_errno_to_nlm4stat (-ret); -nlm4err: - if (ret < 0) { - gf_log (GF_NLM, GF_LOG_ERROR, "unable to call lk()"); - nlm4_generic_reply (cs->req, cs->args.nlm4_unlockargs.cookie, - stat); - nfs3_call_state_wipe (cs); - } - return ret; } int -nlm4_cancel_resume (void *carg) +nlm4_unlock_resume (void *carg) { nlm4_stats stat = nlm4_failed; int ret = -1; @@ -1517,31 +1682,36 @@ nlm4_cancel_resume (void *carg) 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); + nlmclnt = nlm_get_uniq (cs->args.nlm4_unlockargs.alock.caller_name); if (nlmclnt == NULL) { - gf_log (GF_NLM, GF_LOG_ERROR, "nlm_get_uniq() returned 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) { - gf_log (GF_NLM, GF_LOG_ERROR, "nlm_get_uniq() returned NULL"); + stat = nlm4_granted; + gf_log (GF_NLM, GF_LOG_WARNING, "fd_lookup_uint64() returned " + "NULL"); goto nlm4err; } - ret = nlm4_cancel_fd_resume (cs); + ret = nlm4_unlock_fd_resume (cs); nlm4err: if (ret < 0) { - gf_log (GF_NLM, GF_LOG_ERROR, "unable to unlock_fd_resume()"); - nlm4_generic_reply (cs->req, cs->args.nlm4_cancargs.cookie, + 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); } - - return ret; + /* we have already taken care of cleanup */ + return 0; } int -nlm4svc_cancel (rpcsvc_request_t *req) +nlm4svc_unlock (rpcsvc_request_t *req) { xlator_t *vol = NULL; nlm4_stats stat = nlm4_failed; @@ -1559,9 +1729,10 @@ nlm4svc_cancel (rpcsvc_request_t *req) 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) { + 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; @@ -1579,16 +1750,17 @@ nlm4svc_cancel (rpcsvc_request_t *req) } 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_cancel_resume); + NULL, nlm4_unlock_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); + nlm4_generic_reply (req, cs->args.nlm4_unlockargs.cookie, stat); nfs3_call_state_wipe (cs); return 0; } @@ -1601,53 +1773,393 @@ rpcerr: } int -nlm4_unlock_resume (void *carg) +nlm4_share_reply (nfs3_call_state_t *cs, nlm4_stats stat) { - nlm4_stats stat = nlm4_failed; - int ret = -1; - nfs3_call_state_t *cs = NULL; - nlm_client_t *nlmclnt = NULL; + nlm4_shareres res = {{0}, 0, 0}; - if (!carg) + 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 *)carg; - nlm4_check_fh_resolve_status (cs, stat, nlm4err); + cs = (nfs3_call_state_t *)call_state; + nlm4_check_fh_resolve_status (cs, stat, out); - nlmclnt = nlm_get_uniq (cs->args.nlm4_unlockargs.alock.caller_name); - if (nlmclnt == NULL) { - gf_log (GF_NLM, GF_LOG_ERROR, "nlm_get_uniq() returned NULL"); - goto nlm4err; + 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; } - 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() returned NULL"); - goto nlm4err; + + 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; } - ret = nlm4_unlock_fd_resume (cs); -nlm4err: + 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, "unable to unlock_fd_resume"); - nlm4_generic_reply (cs->req, cs->args.nlm4_unlockargs.cookie, - stat); + 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 -nlm4svc_unlock (rpcsvc_request_t *req) +nlm4_unshare_resume (void *call_state) { - 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}, }; + 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; @@ -1657,22 +2169,25 @@ nlm4svc_unlock (rpcsvc_request_t *req) 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"); + 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); + 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"); + 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_generic_reply (req, cs->args.nlm4_unlockargs.cookie, stat); + nlm4_share_reply (cs, stat); nfs3_call_state_wipe (cs); return 0; } @@ -1681,22 +2196,90 @@ nlm4svc_unlock (rpcsvc_request_t *req) 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); + ret = nfs3_fh_resolve_and_resume (cs, &fh, NULL, + nlm4_unshare_resume); -nlm4err: + 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); + gf_log (GF_NLM, GF_LOG_ERROR, "UNSHARE call failed"); + nlm4_share_reply (cs, stat); + ret = 0; return 0; } -rpcerr: - if (ret < 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 @@ -1710,34 +2293,34 @@ nlm4svc_sm_notify (struct nlm_sm_status *status) rpcsvc_actor_t nlm4svc_actors[NLM4_PROC_COUNT] = { /* 0 */ - {"NULL", NLM4_NULL, nlm4svc_null, NULL, NULL}, - {"TEST", NLM4_TEST, nlm4svc_test, NULL, NULL}, - {"LOCK", NLM4_LOCK, nlm4svc_lock, NULL, NULL}, - {"CANCEL", NLM4_CANCEL, nlm4svc_cancel, NULL, NULL}, - {"UNLOCK", NLM4_UNLOCK, nlm4svc_unlock, NULL, NULL}, + {"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, NULL}, - {"TEST", NLM4_TEST_MSG, NULL, NULL, NULL}, - {"LOCK", NLM4_LOCK_MSG, NULL, NULL, NULL}, - {"CANCEL", NLM4_CANCEL_MSG, NULL, NULL, NULL}, - {"UNLOCK", NLM4_UNLOCK_MSG, NULL, NULL, NULL}, + {"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, NULL}, - {"TEST", NLM4_TEST_RES, NULL, NULL, NULL}, - {"LOCK", NLM4_LOCK_RES, NULL, NULL, NULL}, - {"CANCEL", NLM4_CANCEL_RES, NULL, NULL, NULL}, - {"UNLOCK", NLM4_UNLOCK_RES, NULL, NULL, NULL}, - /* 15 ; 17,18,19 are dummy actors */ - {"GRANTED", NLM4_GRANTED_RES, NULL, NULL, NULL}, - {"SM_NOTIFY", NLM4_SM_NOTIFY, NULL, NULL, NULL}, - {"SEVENTEEN", NLM4_SEVENTEEN, NULL, NULL, NULL}, - {"EIGHTEEN", NLM4_EIGHTEEN, NULL, NULL, NULL}, - {"NINETEEN", NLM4_NINETEEN, NULL, NULL, NULL}, + {"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, NULL, NULL, NULL}, - {"UNSHARE", NLM4_UNSHARE, NULL, NULL, NULL}, - {"NM_LOCK", NLM4_NM_LOCK, NULL, NULL, NULL}, - {"FREE_ALL", NLM4_FREE_ALL, nlm4svc_null, NULL, NULL}, + {"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 = { @@ -1773,7 +2356,14 @@ nlm4svc_init(xlator_t *nfsx) int ret = -1; char *portstr = NULL; pthread_t thr; - struct timeval timeout = {0,}; + 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; @@ -1819,7 +2409,7 @@ nlm4svc_init(xlator_t *nfsx) goto err; } - rpcsvc_create_listeners (nfs->rpcsvc, options, "NLM"); + 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); @@ -1828,11 +2418,108 @@ nlm4svc_init(xlator_t *nfsx) 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 index 3c00186d9..9b5d54081 100644 --- a/xlators/nfs/server/src/nlm4.h +++ b/xlators/nfs/server/src/nlm4.h @@ -2,19 +2,10 @@ Copyright (c) 2012 Gluster, Inc. <http://www.gluster.com> This file is part of GlusterFS. - GlusterFS is free software; you can redistribute it and/or modify - it under the terms of the GNU General Public License as published - by the Free Software Foundation; either version 3 of the License, - or (at your option) any later version. - - GlusterFS is distributed in the hope that it will be useful, but - WITHOUT ANY WARRANTY; without even the implied warranty of - MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU - General Public License for more details. - - You should have received a copy of the GNU General Public License - along with this program. If not, see - <http://www.gnu.org/licenses/>. + 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_ @@ -25,6 +16,8 @@ #include "config.h" #endif +#include <sys/types.h> +#include <signal.h> #include "rpcsvc.h" #include "dict.h" #include "xlator.h" @@ -60,19 +53,25 @@ typedef struct nlm_client { 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; -typedef struct { - pthread_cond_t cond; - pthread_mutex_t mutex; -} nlm_condmutex_t; - #endif diff --git a/xlators/nfs/server/src/nlmcbk_svc.c b/xlators/nfs/server/src/nlmcbk_svc.c index 5401dc39b..e1b588765 100644 --- a/xlators/nfs/server/src/nlmcbk_svc.c +++ b/xlators/nfs/server/src/nlmcbk_svc.c @@ -2,19 +2,10 @@ Copyright (c) 2012 Gluster, Inc. <http://www.gluster.com> This file is part of GlusterFS. - GlusterFS is free software; you can redistribute it and/or modify - it under the terms of the GNU General Public License as published - by the Free Software Foundation; either version 3 of the License, - or (at your option) any later version. - - GlusterFS is distributed in the hope that it will be useful, but - WITHOUT ANY WARRANTY; without even the implied warranty of - MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU - General Public License for more details. - - You should have received a copy of the GNU General Public License - along with this program. If not, see - <http://www.gnu.org/licenses/>. + 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. */ /* @@ -92,10 +83,14 @@ void * nsm_thread (void *argv) { register SVCXPRT *transp; + int ret = 0; - pmap_unset (NLMCBK_PROGRAM, NLMCBK_V1); - - transp = svcudp_create(RPC_ANYSOCK); + 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; |