diff options
Diffstat (limited to 'glusterfsd/src')
| -rw-r--r-- | glusterfsd/src/Makefile.am | 23 | ||||
| -rw-r--r-- | glusterfsd/src/fetch-spec.c | 266 | ||||
| -rw-r--r-- | glusterfsd/src/glusterfsd-mem-types.h | 27 | ||||
| -rw-r--r-- | glusterfsd/src/glusterfsd-mgmt.c | 2105 | ||||
| -rw-r--r-- | glusterfsd/src/glusterfsd.c | 2775 | ||||
| -rw-r--r-- | glusterfsd/src/glusterfsd.h | 125 |
6 files changed, 4031 insertions, 1290 deletions
diff --git a/glusterfsd/src/Makefile.am b/glusterfsd/src/Makefile.am index 060917930..05a10dee3 100644 --- a/glusterfsd/src/Makefile.am +++ b/glusterfsd/src/Makefile.am @@ -1,13 +1,19 @@ sbin_PROGRAMS = glusterfsd -glusterfsd_SOURCES = glusterfsd.c fetch-spec.c -glusterfsd_LDADD = $(top_builddir)/libglusterfs/src/libglusterfs.la $(GF_LDADD) -glusterfsd_LDFLAGS = $(GF_LDFLAGS) $(GF_GLUSTERFS_LDFLAGS) -noinst_HEADERS = glusterfsd.h +glusterfsd_SOURCES = glusterfsd.c glusterfsd-mgmt.c +glusterfsd_LDADD = $(top_builddir)/libglusterfs/src/libglusterfs.la \ + $(top_builddir)/rpc/rpc-lib/src/libgfrpc.la \ + $(top_builddir)/rpc/xdr/src/libgfxdr.la \ + $(GF_LDADD) $(GF_GLUSTERFS_LIBS) +glusterfsd_LDFLAGS = $(GF_LDFLAGS) +noinst_HEADERS = glusterfsd.h glusterfsd-mem-types.h -AM_CFLAGS = -fPIC -Wall -D_FILE_OFFSET_BITS=64 -D_GNU_SOURCE -D$(GF_HOST_OS)\ +AM_CPPFLAGS = $(GF_CPPFLAGS) \ -I$(top_srcdir)/libglusterfs/src -DDATADIR=\"$(localstatedir)\" \ - -DCONFDIR=\"$(sysconfdir)/glusterfs\" $(GF_GLUSTERFS_CFLAGS) + -DCONFDIR=\"$(sysconfdir)/glusterfs\" \ + -I$(top_srcdir)/rpc/rpc-lib/src -I$(top_srcdir)/rpc/xdr/src + +AM_CFLAGS = -Wall $(GF_GLUSTERFS_CFLAGS) CLEANFILES = @@ -16,9 +22,14 @@ $(top_builddir)/libglusterfs/src/libglusterfs.la: uninstall-local: rm -f $(DESTDIR)$(sbindir)/glusterfs + rm -f $(DESTDIR)$(sbindir)/glusterd install-data-local: $(INSTALL) -d -m 755 $(DESTDIR)$(localstatedir)/run + $(INSTALL) -d -m 755 $(DESTDIR)$(localstatedir)/run/gluster $(INSTALL) -d -m 755 $(DESTDIR)$(localstatedir)/log/glusterfs + $(INSTALL) -d -m 755 $(DESTDIR)$(sbindir) rm -f $(DESTDIR)$(sbindir)/glusterfs + rm -f $(DESTDIR)$(sbindir)/glusterd ln -s glusterfsd $(DESTDIR)$(sbindir)/glusterfs + ln -s glusterfsd $(DESTDIR)$(sbindir)/glusterd diff --git a/glusterfsd/src/fetch-spec.c b/glusterfsd/src/fetch-spec.c deleted file mode 100644 index 6f2b991bc..000000000 --- a/glusterfsd/src/fetch-spec.c +++ /dev/null @@ -1,266 +0,0 @@ -/* - Copyright (c) 2007-2009 Z RESEARCH, Inc. <http://www.zresearch.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/>. -*/ - -#include <stdio.h> -#include <sys/types.h> -#include <sys/wait.h> -#include <stdlib.h> - -#ifndef _CONFIG_H -#define _CONFIG_H -#include "config.h" -#endif /* _CONFIG_H */ - -#include "glusterfs.h" -#include "stack.h" -#include "dict.h" -#include "transport.h" -#include "event.h" -#include "defaults.h" - - -static int -fetch_cbk (call_frame_t *frame, - void *cookie, - xlator_t *this, - int32_t op_ret, - int32_t op_errno, - char *spec_data) -{ - FILE *spec_fp = NULL; - - spec_fp = frame->local; - - if (op_ret >= 0) { - fwrite (spec_data, strlen (spec_data), 1, spec_fp); - fflush (spec_fp); - fclose (spec_fp); - } - else { - gf_log (frame->this->name, GF_LOG_ERROR, - "GETSPEC from server returned -1 (%s)", - strerror (op_errno)); - } - - frame->local = NULL; - STACK_DESTROY (frame->root); - - /* exit the child process */ - exit (op_ret); -} - - -static int -fetch_notify (xlator_t *this_xl, int event, void *data, ...) -{ - int ret = 0; - call_frame_t *frame = NULL; - - switch (event) - { - case GF_EVENT_CHILD_UP: - frame = create_frame (this_xl, this_xl->ctx->pool); - frame->local = this_xl->private; - - STACK_WIND (frame, fetch_cbk, - this_xl->children->xlator, - this_xl->children->xlator->mops->getspec, - this_xl->ctx->cmd_args.volfile_id, - 0); - break; - case GF_EVENT_CHILD_DOWN: - break; - default: - ret = default_notify (this_xl, event, data); - break; - } - - return ret; -} - - -static int -fetch_init (xlator_t *xl) -{ - return 0; -} - -static xlator_t * -get_shrub (glusterfs_ctx_t *ctx, - const char *remote_host, - const char *transport, - uint32_t remote_port) -{ - int ret = 0; - xlator_t *top = NULL; - xlator_t *trans = NULL; - xlator_list_t *parent = NULL, *tmp = NULL; - - top = CALLOC (1, sizeof (*top)); - ERR_ABORT (top); - trans = CALLOC (1, sizeof (*trans)); - ERR_ABORT (trans); - - top->name = "top"; - top->ctx = ctx; - top->next = trans; - top->init = fetch_init; - top->notify = fetch_notify; - top->children = (void *) CALLOC (1, sizeof (*top->children)); - ERR_ABORT (top->children); - top->children->xlator = trans; - - trans->name = "trans"; - trans->ctx = ctx; - trans->prev = top; - trans->init = fetch_init; - trans->notify = default_notify; - trans->options = get_new_dict (); - - parent = CALLOC (1, sizeof(*parent)); - parent->xlator = top; - if (trans->parents == NULL) - trans->parents = parent; - else { - tmp = trans->parents; - while (tmp->next) - tmp = tmp->next; - tmp->next = parent; - } - - /* TODO: log on failure to set dict */ - if (remote_host) - ret = dict_set_static_ptr (trans->options, "remote-host", - (char *)remote_host); - - if (remote_port) - ret = dict_set_uint32 (trans->options, "remote-port", - remote_port); - - /* 'option remote-subvolume <x>' is needed here even though - * its not used - */ - ret = dict_set_static_ptr (trans->options, "remote-subvolume", - "brick"); - ret = dict_set_static_ptr (trans->options, "disable-handshake", "on"); - ret = dict_set_static_ptr (trans->options, "non-blocking-io", "off"); - - if (transport) { - char *transport_type = CALLOC (1, strlen (transport) + 10); - ERR_ABORT (transport_type); - strcpy(transport_type, transport); - - if (strchr (transport_type, ':')) - *(strchr (transport_type, ':')) = '\0'; - - ret = dict_set_dynstr (trans->options, "transport-type", - transport_type); - } - - xlator_set_type (trans, "protocol/client"); - - if (xlator_tree_init (top) != 0) - return NULL; - - return top; -} - - -static int -_fetch (glusterfs_ctx_t *ctx, - FILE *spec_fp, - const char *remote_host, - const char *transport, - uint32_t remote_port) -{ - xlator_t *this = NULL; - - this = get_shrub (ctx, remote_host, transport, remote_port); - if (this == NULL) - return -1; - - this->private = spec_fp; - - event_dispatch (ctx->event_pool); - - return 0; -} - - -static int -_fork_and_fetch (glusterfs_ctx_t *ctx, - FILE *spec_fp, - const char *remote_host, - const char *transport, - uint32_t remote_port) -{ - int ret; - - ret = fork (); - switch (ret) { - case -1: - perror ("fork()"); - break; - case 0: - /* child */ - ret = _fetch (ctx, spec_fp, remote_host, - transport, remote_port); - if (ret == -1) - exit (ret); - default: - /* parent */ - wait (&ret); - ret = WEXITSTATUS (ret); - } - return ret; -} - -FILE * -fetch_spec (glusterfs_ctx_t *ctx) -{ - char *remote_host = NULL; - char *transport = NULL; - FILE *spec_fp; - int32_t ret; - - spec_fp = tmpfile (); - - if (!spec_fp) { - perror ("tmpfile ()"); - return NULL; - } - - remote_host = ctx->cmd_args.volfile_server; - transport = ctx->cmd_args.volfile_server_transport; - if (!transport) - transport = "tcp"; - - ret = _fork_and_fetch (ctx, spec_fp, remote_host, transport, - ctx->cmd_args.volfile_server_port); - - if (!ret) { - fseek (spec_fp, 0, SEEK_SET); - } - else { - fclose (spec_fp); - spec_fp = NULL; - } - - return spec_fp; -} diff --git a/glusterfsd/src/glusterfsd-mem-types.h b/glusterfsd/src/glusterfsd-mem-types.h new file mode 100644 index 000000000..7135c0ada --- /dev/null +++ b/glusterfsd/src/glusterfsd-mem-types.h @@ -0,0 +1,27 @@ +/* + Copyright (c) 2006-2012 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 __GLUSTERFSD_MEM_TYPES_H__ +#define __GLUSTERFSD_MEM_TYPES_H__ + +#include "mem-types.h" + +#define GF_MEM_TYPE_START (gf_common_mt_end + 1) + +enum gfd_mem_types_ { + gfd_mt_xlator_list_t = GF_MEM_TYPE_START, + gfd_mt_xlator_t, + gfd_mt_server_cmdline_t, + gfd_mt_xlator_cmdline_option_t, + gfd_mt_char, + gfd_mt_call_pool_t, + gfd_mt_end + +}; +#endif diff --git a/glusterfsd/src/glusterfsd-mgmt.c b/glusterfsd/src/glusterfsd-mgmt.c new file mode 100644 index 000000000..1c9220927 --- /dev/null +++ b/glusterfsd/src/glusterfsd-mgmt.c @@ -0,0 +1,2105 @@ +/* + Copyright (c) 2007-2012 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. +*/ +#include <stdio.h> +#include <sys/types.h> +#include <sys/wait.h> +#include <stdlib.h> +#include <signal.h> + +#ifndef _CONFIG_H +#define _CONFIG_H +#include "config.h" +#endif /* _CONFIG_H */ + +#include "glusterfs.h" +#include "stack.h" +#include "dict.h" +#include "event.h" +#include "defaults.h" + +#include "rpc-clnt.h" +#include "protocol-common.h" +#include "glusterfs3.h" +#include "portmap-xdr.h" +#include "xdr-generic.h" + +#include "glusterfsd.h" +#include "rpcsvc.h" +#include "cli1-xdr.h" +#include "statedump.h" +#include "syncop.h" +#include "xlator.h" + +static gf_boolean_t is_mgmt_rpc_reconnect = _gf_false; + +int glusterfs_mgmt_pmap_signin (glusterfs_ctx_t *ctx); +int glusterfs_volfile_fetch (glusterfs_ctx_t *ctx); +int glusterfs_process_volfp (glusterfs_ctx_t *ctx, FILE *fp); +int glusterfs_graph_unknown_options (glusterfs_graph_t *graph); +int emancipate(glusterfs_ctx_t *ctx, int ret); + +int +mgmt_cbk_spec (struct rpc_clnt *rpc, void *mydata, void *data) +{ + glusterfs_ctx_t *ctx = NULL; + xlator_t *this = NULL; + + this = mydata; + ctx = glusterfsd_ctx; + gf_log ("mgmt", GF_LOG_INFO, "Volume file changed"); + + glusterfs_volfile_fetch (ctx); + return 0; +} + + +int +mgmt_cbk_event (struct rpc_clnt *rpc, void *mydata, void *data) +{ + return 0; +} + +struct iobuf * +glusterfs_serialize_reply (rpcsvc_request_t *req, void *arg, + struct iovec *outmsg, xdrproc_t xdrproc) +{ + struct iobuf *iob = NULL; + ssize_t retlen = -1; + ssize_t xdr_size = 0; + + /* First, get the io buffer into which the reply in arg will + * be serialized. + */ + xdr_size = xdr_sizeof (xdrproc, arg); + iob = iobuf_get2 (req->svc->ctx->iobuf_pool, xdr_size); + if (!iob) { + gf_log (THIS->name, GF_LOG_ERROR, "Failed to get iobuf"); + goto ret; + } + + iobuf_to_iovec (iob, outmsg); + /* Use the given serializer to translate the give C structure in arg + * to XDR format which will be written into the buffer in outmsg. + */ + /* retlen is used to received the error since size_t is unsigned and we + * need -1 for error notification during encoding. + */ + retlen = xdr_serialize_generic (*outmsg, arg, xdrproc); + if (retlen == -1) { + gf_log (THIS->name, GF_LOG_ERROR, "Failed to encode message"); + goto ret; + } + + outmsg->iov_len = retlen; +ret: + if (retlen == -1) { + iob = NULL; + } + + return iob; +} + +int +glusterfs_submit_reply (rpcsvc_request_t *req, void *arg, + struct iovec *payload, int payloadcount, + struct iobref *iobref, xdrproc_t xdrproc) +{ + struct iobuf *iob = NULL; + int ret = -1; + struct iovec rsp = {0,}; + char new_iobref = 0; + + if (!req) { + GF_ASSERT (req); + goto out; + } + + if (!iobref) { + iobref = iobref_new (); + if (!iobref) { + gf_log (THIS->name, GF_LOG_ERROR, "out of memory"); + goto out; + } + + new_iobref = 1; + } + + iob = glusterfs_serialize_reply (req, arg, &rsp, xdrproc); + if (!iob) { + gf_log_callingfn (THIS->name, GF_LOG_ERROR, "Failed to serialize reply"); + } else { + iobref_add (iobref, iob); + } + + ret = rpcsvc_submit_generic (req, &rsp, 1, payload, payloadcount, + 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. + */ + if (ret == -1) { + gf_log (THIS->name, GF_LOG_ERROR, "Reply submission failed"); + goto out; + } + + ret = 0; +out: + if (iob) + iobuf_unref (iob); + + if (new_iobref && iobref) + iobref_unref (iobref); + + return ret; +} + +int +glusterfs_terminate_response_send (rpcsvc_request_t *req, int op_ret) +{ + gd1_mgmt_brick_op_rsp rsp = {0,}; + dict_t *dict = NULL; + int ret = 0; + + rsp.op_ret = op_ret; + rsp.op_errno = 0; + rsp.op_errstr = ""; + dict = dict_new (); + + if (dict) + ret = dict_allocate_and_serialize (dict, &rsp.output.output_val, + &rsp.output.output_len); + + + if (ret == 0) + ret = glusterfs_submit_reply (req, &rsp, NULL, 0, NULL, + (xdrproc_t)xdr_gd1_mgmt_brick_op_rsp); + + GF_FREE (rsp.output.output_val); + if (dict) + dict_unref (dict); + return ret; +} + +int +glusterfs_handle_terminate (rpcsvc_request_t *req) +{ + + glusterfs_terminate_response_send (req, 0); + cleanup_and_exit (SIGTERM); + return 0; +} + +int +glusterfs_translator_info_response_send (rpcsvc_request_t *req, int ret, + char *msg, dict_t *output) +{ + gd1_mgmt_brick_op_rsp rsp = {0,}; + gf_boolean_t free_ptr = _gf_false; + GF_ASSERT (req); + + rsp.op_ret = ret; + rsp.op_errno = 0; + if (ret && msg && msg[0]) + rsp.op_errstr = msg; + else + rsp.op_errstr = ""; + + ret = -1; + if (output) { + ret = dict_allocate_and_serialize (output, + &rsp.output.output_val, + &rsp.output.output_len); + } + if (!ret) + free_ptr = _gf_true; + + glusterfs_submit_reply (req, &rsp, NULL, 0, NULL, + (xdrproc_t)xdr_gd1_mgmt_brick_op_rsp); + ret = 0; + if (free_ptr) + GF_FREE (rsp.output.output_val); + return ret; +} + +int +glusterfs_xlator_op_response_send (rpcsvc_request_t *req, int op_ret, + char *msg, dict_t *output) +{ + gd1_mgmt_brick_op_rsp rsp = {0,}; + int ret = -1; + gf_boolean_t free_ptr = _gf_false; + GF_ASSERT (req); + + rsp.op_ret = op_ret; + rsp.op_errno = 0; + if (op_ret && msg && msg[0]) + rsp.op_errstr = msg; + else + rsp.op_errstr = ""; + + if (output) { + ret = dict_allocate_and_serialize (output, + &rsp.output.output_val, + &rsp.output.output_len); + } + if (!ret) + free_ptr = _gf_true; + + ret = glusterfs_submit_reply (req, &rsp, NULL, 0, NULL, + (xdrproc_t)xdr_gd1_mgmt_brick_op_rsp); + + if (free_ptr) + GF_FREE (rsp.output.output_val); + + return ret; +} + +int +glusterfs_handle_translator_info_get (rpcsvc_request_t *req) +{ + int32_t ret = -1; + gd1_mgmt_brick_op_req xlator_req = {0,}; + dict_t *dict = NULL; + xlator_t *this = NULL; + gf1_cli_top_op top_op = 0; + uint32_t blk_size = 0; + uint32_t blk_count = 0; + double time = 0; + double throughput = 0; + xlator_t *any = NULL; + xlator_t *xlator = NULL; + glusterfs_graph_t *active = NULL; + glusterfs_ctx_t *ctx = NULL; + char msg[2048] = {0,}; + dict_t *output = NULL; + + GF_ASSERT (req); + this = THIS; + GF_ASSERT (this); + + ret = xdr_to_generic (req->msg[0], &xlator_req, + (xdrproc_t)xdr_gd1_mgmt_brick_op_req); + if (ret < 0) { + //failed to decode msg; + req->rpc_err = GARBAGE_ARGS; + goto out; + } + + dict = dict_new (); + ret = dict_unserialize (xlator_req.input.input_val, + xlator_req.input.input_len, + &dict); + if (ret < 0) { + gf_log (this->name, GF_LOG_ERROR, + "failed to " + "unserialize req-buffer to dictionary"); + goto out; + } + + ret = dict_get_int32 (dict, "top-op", (int32_t *)&top_op); + if ((!ret) && (GF_CLI_TOP_READ_PERF == top_op || + GF_CLI_TOP_WRITE_PERF == top_op)) { + ret = dict_get_uint32 (dict, "blk-size", &blk_size); + if (ret) + goto cont; + ret = dict_get_uint32 (dict, "blk-cnt", &blk_count); + if (ret) + goto cont; + + if (GF_CLI_TOP_READ_PERF == top_op) { + ret = glusterfs_volume_top_read_perf + (blk_size, blk_count, xlator_req.name, + &throughput, &time); + } else if ( GF_CLI_TOP_WRITE_PERF == top_op) { + ret = glusterfs_volume_top_write_perf + (blk_size, blk_count, xlator_req.name, + &throughput, &time); + } + ret = dict_set_double (dict, "time", time); + if (ret) + goto cont; + ret = dict_set_double (dict, "throughput", throughput); + if (ret) + goto cont; + } +cont: + ctx = glusterfsd_ctx; + GF_ASSERT (ctx); + active = ctx->active; + any = active->first; + + xlator = xlator_search_by_name (any, xlator_req.name); + if (!xlator) { + snprintf (msg, sizeof (msg), "xlator %s is not loaded", + xlator_req.name); + goto out; + } + + output = dict_new (); + ret = xlator->notify (xlator, GF_EVENT_TRANSLATOR_INFO, dict, output); + +out: + ret = glusterfs_translator_info_response_send (req, ret, msg, output); + + free (xlator_req.name); + free (xlator_req.input.input_val); + if (output) + dict_unref (output); + if (dict) + dict_unref (dict); + return ret; +} + +int +glusterfs_volume_top_write_perf (uint32_t blk_size, uint32_t blk_count, + char *brick_path, double *throughput, + double *time) +{ + int32_t fd = -1; + int32_t input_fd = -1; + char export_path[PATH_MAX]; + char *buf = NULL; + int32_t iter = 0; + int32_t ret = -1; + uint64_t total_blks = 0; + struct timeval begin, end = {0,}; + + GF_ASSERT (brick_path); + GF_ASSERT (throughput); + GF_ASSERT (time); + if (!(blk_size > 0) || ! (blk_count > 0)) + goto out; + + snprintf (export_path, sizeof (export_path), "%s/%s", + brick_path, ".gf-tmp-stats-perf"); + + fd = open (export_path, O_CREAT|O_RDWR, S_IRWXU); + if (-1 == fd) { + ret = -1; + gf_log ("glusterd", GF_LOG_ERROR, "Could not open tmp file"); + goto out; + } + + buf = GF_MALLOC (blk_size * sizeof(*buf), gf_common_mt_char); + if (!buf) { + ret = -1; + goto out; + } + + input_fd = open ("/dev/zero", O_RDONLY); + if (-1 == input_fd) { + ret = -1; + gf_log ("glusterd",GF_LOG_ERROR, "Unable to open input file"); + goto out; + } + + gettimeofday (&begin, NULL); + for (iter = 0; iter < blk_count; iter++) { + ret = read (input_fd, buf, blk_size); + if (ret != blk_size) { + ret = -1; + goto out; + } + ret = write (fd, buf, blk_size); + if (ret != blk_size) { + ret = -1; + goto out; + } + total_blks += ret; + } + ret = 0; + if (total_blks != ((uint64_t)blk_size * blk_count)) { + gf_log ("glusterd", GF_LOG_WARNING, "Error in write"); + ret = -1; + goto out; + } + + gettimeofday (&end, NULL); + *time = (end.tv_sec - begin.tv_sec) * 1e6 + + (end.tv_usec - begin.tv_usec); + *throughput = total_blks / *time; + gf_log ("glusterd", GF_LOG_INFO, "Throughput %.2f Mbps time %.2f secs " + "bytes written %"PRId64, *throughput, *time, total_blks); + +out: + if (fd >= 0) + close (fd); + if (input_fd >= 0) + close (input_fd); + GF_FREE (buf); + unlink (export_path); + + return ret; +} + +int +glusterfs_volume_top_read_perf (uint32_t blk_size, uint32_t blk_count, + char *brick_path, double *throughput, + double *time) +{ + int32_t fd = -1; + int32_t input_fd = -1; + int32_t output_fd = -1; + char export_path[PATH_MAX]; + char *buf = NULL; + int32_t iter = 0; + int32_t ret = -1; + uint64_t total_blks = 0; + struct timeval begin, end = {0,}; + + GF_ASSERT (brick_path); + GF_ASSERT (throughput); + GF_ASSERT (time); + if (!(blk_size > 0) || ! (blk_count > 0)) + goto out; + + snprintf (export_path, sizeof (export_path), "%s/%s", + brick_path, ".gf-tmp-stats-perf"); + fd = open (export_path, O_CREAT|O_RDWR, S_IRWXU); + if (-1 == fd) { + ret = -1; + gf_log ("glusterd", GF_LOG_ERROR, "Could not open tmp file"); + goto out; + } + + buf = GF_MALLOC (blk_size * sizeof(*buf), gf_common_mt_char); + if (!buf) { + ret = -1; + gf_log ("glusterd", GF_LOG_ERROR, "Could not allocate memory"); + goto out; + } + + input_fd = open ("/dev/zero", O_RDONLY); + if (-1 == input_fd) { + ret = -1; + gf_log ("glusterd", GF_LOG_ERROR, "Could not open input file"); + goto out; + } + + output_fd = open ("/dev/null", O_RDWR); + if (-1 == output_fd) { + ret = -1; + gf_log ("glusterd", GF_LOG_ERROR, "Could not open output file"); + goto out; + } + + for (iter = 0; iter < blk_count; iter++) { + ret = read (input_fd, buf, blk_size); + if (ret != blk_size) { + ret = -1; + goto out; + } + ret = write (fd, buf, blk_size); + if (ret != blk_size) { + ret = -1; + goto out; + } + } + + ret = fsync (fd); + if (ret) { + gf_log ("glusterd", GF_LOG_ERROR, "could not flush cache"); + goto out; + } + ret = lseek (fd, 0L, 0); + if (ret != 0) { + gf_log ("glusterd", GF_LOG_ERROR, + "could not seek back to start"); + ret = -1; + goto out; + } + gettimeofday (&begin, NULL); + for (iter = 0; iter < blk_count; iter++) { + ret = read (fd, buf, blk_size); + if (ret != blk_size) { + ret = -1; + goto out; + } + ret = write (output_fd, buf, blk_size); + if (ret != blk_size) { + ret = -1; + goto out; + } + total_blks += ret; + } + ret = 0; + if (total_blks != ((uint64_t)blk_size * blk_count)) { + ret = -1; + gf_log ("glusterd", GF_LOG_WARNING, "Error in read"); + goto out; + } + + gettimeofday (&end, NULL); + *time = (end.tv_sec - begin.tv_sec) * 1e6 + + (end.tv_usec - begin.tv_usec); + *throughput = total_blks / *time; + gf_log ("glusterd", GF_LOG_INFO, "Throughput %.2f Mbps time %.2f secs " + "bytes read %"PRId64, *throughput, *time, total_blks); + +out: + if (fd >= 0) + close (fd); + if (input_fd >= 0) + close (input_fd); + if (output_fd >= 0) + close (output_fd); + GF_FREE (buf); + unlink (export_path); + + return ret; +} + +int +glusterfs_handle_translator_op (rpcsvc_request_t *req) +{ + int32_t ret = -1; + gd1_mgmt_brick_op_req xlator_req = {0,}; + dict_t *input = NULL; + xlator_t *xlator = NULL; + xlator_t *any = NULL; + dict_t *output = NULL; + char key[2048] = {0}; + char *xname = NULL; + glusterfs_ctx_t *ctx = NULL; + glusterfs_graph_t *active = NULL; + xlator_t *this = NULL; + int i = 0; + int count = 0; + + GF_ASSERT (req); + this = THIS; + GF_ASSERT (this); + + ret = xdr_to_generic (req->msg[0], &xlator_req, + (xdrproc_t)xdr_gd1_mgmt_brick_op_req); + if (ret < 0) { + //failed to decode msg; + req->rpc_err = GARBAGE_ARGS; + goto out; + } + + ctx = glusterfsd_ctx; + active = ctx->active; + any = active->first; + input = dict_new (); + ret = dict_unserialize (xlator_req.input.input_val, + xlator_req.input.input_len, + &input); + if (ret < 0) { + gf_log (this->name, GF_LOG_ERROR, + "failed to " + "unserialize req-buffer to dictionary"); + goto out; + } else { + input->extra_stdfree = xlator_req.input.input_val; + } + + ret = dict_get_int32 (input, "count", &count); + + output = dict_new (); + if (!output) { + ret = -1; + goto out; + } + + for (i = 0; i < count; i++) { + snprintf (key, sizeof (key), "xl-%d", i); + ret = dict_get_str (input, key, &xname); + if (ret) { + gf_log (this->name, GF_LOG_ERROR, "Couldn't get " + "xlator %s ", key); + goto out; + } + xlator = xlator_search_by_name (any, xname); + if (!xlator) { + gf_log (this->name, GF_LOG_ERROR, "xlator %s is not " + "loaded", xname); + goto out; + } + } + for (i = 0; i < count; i++) { + snprintf (key, sizeof (key), "xl-%d", i); + ret = dict_get_str (input, key, &xname); + xlator = xlator_search_by_name (any, xname); + XLATOR_NOTIFY (xlator, GF_EVENT_TRANSLATOR_OP, input, output); + if (ret) + break; + } +out: + glusterfs_xlator_op_response_send (req, ret, "", output); + if (input) + dict_unref (input); + if (output) + dict_unref (output); + free (xlator_req.name); //malloced by xdr + + return 0; +} + + +int +glusterfs_handle_defrag (rpcsvc_request_t *req) +{ + int32_t ret = -1; + gd1_mgmt_brick_op_req xlator_req = {0,}; + dict_t *dict = NULL; + xlator_t *xlator = NULL; + xlator_t *any = NULL; + dict_t *output = NULL; + char msg[2048] = {0}; + glusterfs_ctx_t *ctx = NULL; + glusterfs_graph_t *active = NULL; + xlator_t *this = NULL; + + GF_ASSERT (req); + this = THIS; + GF_ASSERT (this); + + ctx = glusterfsd_ctx; + GF_ASSERT (ctx); + + active = ctx->active; + if (!active) { + req->rpc_err = GARBAGE_ARGS; + goto out; + } + + any = active->first; + ret = xdr_to_generic (req->msg[0], &xlator_req, + (xdrproc_t)xdr_gd1_mgmt_brick_op_req); + if (ret < 0) { + //failed to decode msg; + req->rpc_err = GARBAGE_ARGS; + goto out; + } + dict = dict_new (); + if (!dict) + goto out; + + ret = dict_unserialize (xlator_req.input.input_val, + xlator_req.input.input_len, + &dict); + if (ret < 0) { + gf_log (this->name, GF_LOG_ERROR, + "failed to " + "unserialize req-buffer to dictionary"); + goto out; + } + xlator = xlator_search_by_name (any, xlator_req.name); + if (!xlator) { + snprintf (msg, sizeof (msg), "xlator %s is not loaded", + xlator_req.name); + goto out; + } + + output = dict_new (); + if (!output) { + ret = -1; + goto out; + } + + ret = xlator->notify (xlator, GF_EVENT_VOLUME_DEFRAG, dict, output); + + ret = glusterfs_translator_info_response_send (req, ret, + msg, output); +out: + if (dict) + dict_unref (dict); + free (xlator_req.input.input_val); // malloced by xdr + if (output) + dict_unref (output); + free (xlator_req.name); //malloced by xdr + + return ret; + +} +int +glusterfs_handle_brick_status (rpcsvc_request_t *req) +{ + int ret = -1; + gd1_mgmt_brick_op_req brick_req = {0,}; + gd1_mgmt_brick_op_rsp rsp = {0,}; + glusterfs_ctx_t *ctx = NULL; + glusterfs_graph_t *active = NULL; + xlator_t *this = NULL; + xlator_t *any = NULL; + xlator_t *xlator = NULL; + dict_t *dict = NULL; + dict_t *output = NULL; + char *volname = NULL; + char *xname = NULL; + uint32_t cmd = 0; + char *msg = NULL; + + GF_ASSERT (req); + this = THIS; + GF_ASSERT (this); + + ret = xdr_to_generic (req->msg[0], &brick_req, + (xdrproc_t)xdr_gd1_mgmt_brick_op_req); + if (ret < 0) { + req->rpc_err = GARBAGE_ARGS; + goto out; + } + + dict = dict_new (); + ret = dict_unserialize (brick_req.input.input_val, + brick_req.input.input_len, &dict); + if (ret < 0) { + gf_log (this->name, GF_LOG_ERROR, "Failed to unserialize " + "req-buffer to dictionary"); + goto out; + } + + ret = dict_get_uint32 (dict, "cmd", &cmd); + if (ret) { + gf_log (this->name, GF_LOG_ERROR, "Couldn't get status op"); + goto out; + } + + ret = dict_get_str (dict, "volname", &volname); + if (ret) { + gf_log (this->name, GF_LOG_ERROR, "Couldn't get volname"); + goto out; + } + + ctx = glusterfsd_ctx; + GF_ASSERT (ctx); + active = ctx->active; + any = active->first; + + ret = gf_asprintf (&xname, "%s-server", volname); + if (-1 == ret) { + gf_log (this->name, GF_LOG_ERROR, "Out of memory"); + goto out; + } + + xlator = xlator_search_by_name (any, xname); + if (!xlator) { + gf_log (this->name, GF_LOG_ERROR, "xlator %s is not loaded", + xname); + ret = -1; + goto out; + } + + + output = dict_new (); + switch (cmd & GF_CLI_STATUS_MASK) { + case GF_CLI_STATUS_MEM: + ret = 0; + gf_proc_dump_mem_info_to_dict (output); + gf_proc_dump_mempool_info_to_dict (ctx, output); + break; + + case GF_CLI_STATUS_CLIENTS: + ret = xlator->dumpops->priv_to_dict (xlator, output); + break; + + case GF_CLI_STATUS_INODE: + ret = xlator->dumpops->inode_to_dict (xlator, output); + break; + + case GF_CLI_STATUS_FD: + ret = xlator->dumpops->fd_to_dict (xlator, output); + break; + + case GF_CLI_STATUS_CALLPOOL: + ret = 0; + gf_proc_dump_pending_frames_to_dict (ctx->pool, output); + break; + + default: + ret = -1; + msg = gf_strdup ("Unknown status op"); + break; + } + rsp.op_ret = ret; + rsp.op_errno = 0; + if (ret && msg) + rsp.op_errstr = msg; + else + rsp.op_errstr = ""; + + ret = dict_allocate_and_serialize (output, &rsp.output.output_val, + &rsp.output.output_len); + if (ret) { + gf_log (this->name, GF_LOG_ERROR, + "Failed to serialize output dict to rsp"); + goto out; + } + + glusterfs_submit_reply (req, &rsp, NULL, 0, NULL, + (xdrproc_t)xdr_gd1_mgmt_brick_op_rsp); + ret = 0; + +out: + if (dict) + dict_unref (dict); + if (output) + dict_unref (output); + free (brick_req.input.input_val); + GF_FREE (xname); + GF_FREE (msg); + GF_FREE (rsp.output.output_val); + + return ret; +} + + +int +glusterfs_handle_node_status (rpcsvc_request_t *req) +{ + int ret = -1; + gd1_mgmt_brick_op_req node_req = {0,}; + gd1_mgmt_brick_op_rsp rsp = {0,}; + glusterfs_ctx_t *ctx = NULL; + glusterfs_graph_t *active = NULL; + xlator_t *any = NULL; + xlator_t *node = NULL; + xlator_t *subvol = NULL; + dict_t *dict = NULL; + dict_t *output = NULL; + char *volname = NULL; + char *node_name = NULL; + char *subvol_name = NULL; + uint32_t cmd = 0; + char *msg = NULL; + + GF_ASSERT (req); + + ret = xdr_to_generic (req->msg[0], &node_req, + (xdrproc_t)xdr_gd1_mgmt_brick_op_req); + if (ret < 0) { + req->rpc_err = GARBAGE_ARGS; + goto out; + } + + dict = dict_new (); + ret = dict_unserialize (node_req.input.input_val, + node_req.input.input_len, &dict); + if (ret < 0) { + gf_log (THIS->name, GF_LOG_ERROR, "Failed to unserialize " + "req buffer to dictionary"); + goto out; + } + + ret = dict_get_uint32 (dict, "cmd", &cmd); + if (ret) { + gf_log (THIS->name, GF_LOG_ERROR, "Couldn't get status op"); + goto out; + } + + ret = dict_get_str (dict, "volname", &volname); + if (ret) { + gf_log (THIS->name, GF_LOG_ERROR, "Couldn't get volname"); + goto out; + } + + ctx = glusterfsd_ctx; + GF_ASSERT (ctx); + active = ctx->active; + any = active->first; + + if ((cmd & GF_CLI_STATUS_NFS) != 0) + ret = gf_asprintf (&node_name, "%s", "nfs-server"); + else if ((cmd & GF_CLI_STATUS_SHD) != 0) + ret = gf_asprintf (&node_name, "%s", "glustershd"); + else { + ret = -1; + goto out; + } + if (ret == -1) { + gf_log (THIS->name, GF_LOG_ERROR, + "Failed to set node xlator name"); + goto out; + } + + node = xlator_search_by_name (any, node_name); + if (!node) { + ret = -1; + gf_log (THIS->name, GF_LOG_ERROR, "%s xlator is not loaded", + node_name); + goto out; + } + + if ((cmd & GF_CLI_STATUS_NFS) != 0) + ret = gf_asprintf (&subvol_name, "%s", volname); + else if ((cmd & GF_CLI_STATUS_SHD) != 0) + ret = gf_asprintf (&subvol_name, "%s-replicate-0", volname); + else { + ret = -1; + goto out; + } + if (ret == -1) { + gf_log (THIS->name, GF_LOG_ERROR, + "Failed to set node xlator name"); + goto out; + } + + subvol = xlator_search_by_name (node, subvol_name); + if (!subvol) { + ret = -1; + gf_log (THIS->name, GF_LOG_ERROR, "%s xlator is not loaded", + subvol_name); + goto out; + } + + output = dict_new (); + switch (cmd & GF_CLI_STATUS_MASK) { + case GF_CLI_STATUS_MEM: + ret = 0; + gf_proc_dump_mem_info_to_dict (output); + gf_proc_dump_mempool_info_to_dict (ctx, output); + break; + + case GF_CLI_STATUS_CLIENTS: + // clients not availbale for SHD + if ((cmd & GF_CLI_STATUS_SHD) != 0) + break; + + ret = dict_set_str (output, "volname", volname); + if (ret) { + gf_log (THIS->name, GF_LOG_ERROR, + "Error setting volname to dict"); + goto out; + } + ret = node->dumpops->priv_to_dict (node, output); + break; + + case GF_CLI_STATUS_INODE: + ret = 0; + inode_table_dump_to_dict (subvol->itable, "conn0", + output); + ret = dict_set_int32 (output, "conncount", 1); + break; + + case GF_CLI_STATUS_FD: + // cannot find fd-tables in nfs-server graph + // TODO: finish once found + break; + + case GF_CLI_STATUS_CALLPOOL: + ret = 0; + gf_proc_dump_pending_frames_to_dict (ctx->pool, output); + break; + + default: + ret = -1; + msg = gf_strdup ("Unknown status op"); + gf_log (THIS->name, GF_LOG_ERROR, "%s", msg); + break; + } + rsp.op_ret = ret; + rsp.op_errno = 0; + if (ret && msg) + rsp.op_errstr = msg; + else + rsp.op_errstr = ""; + + ret = dict_allocate_and_serialize (output, &rsp.output.output_val, + &rsp.output.output_len); + if (ret) { + gf_log (THIS->name, GF_LOG_ERROR, + "Failed to serialize output dict to rsp"); + goto out; + } + + glusterfs_submit_reply (req, &rsp, NULL, 0, NULL, + (xdrproc_t)xdr_gd1_mgmt_brick_op_rsp); + ret = 0; + +out: + if (dict) + dict_unref (dict); + free (node_req.input.input_val); + GF_FREE (msg); + GF_FREE (rsp.output.output_val); + GF_FREE (node_name); + GF_FREE (subvol_name); + + gf_log (THIS->name, GF_LOG_DEBUG, "Returning %d", ret); + return ret; +} + +int +glusterfs_handle_nfs_profile (rpcsvc_request_t *req) +{ + int ret = -1; + gd1_mgmt_brick_op_req nfs_req = {0,}; + gd1_mgmt_brick_op_rsp rsp = {0,}; + dict_t *dict = NULL; + glusterfs_ctx_t *ctx = NULL; + glusterfs_graph_t *active = NULL; + xlator_t *any = NULL; + xlator_t *nfs = NULL; + xlator_t *subvol = NULL; + char *volname = NULL; + dict_t *output = NULL; + + GF_ASSERT (req); + + ret = xdr_to_generic (req->msg[0], &nfs_req, + (xdrproc_t)xdr_gd1_mgmt_brick_op_req); + if (ret < 0) { + req->rpc_err = GARBAGE_ARGS; + goto out; + } + + dict = dict_new (); + ret = dict_unserialize (nfs_req.input.input_val, + nfs_req.input.input_len, &dict); + if (ret < 0) { + gf_log (THIS->name, GF_LOG_ERROR, "Failed to " + "unserialize req-buffer to dict"); + goto out; + } + + ret = dict_get_str (dict, "volname", &volname); + if (ret) { + gf_log (THIS->name, GF_LOG_ERROR, "Couldn't get volname"); + goto out; + } + + ctx = glusterfsd_ctx; + GF_ASSERT (ctx); + + active = ctx->active; + any = active->first; + + // is this needed? + // are problems possible by searching for subvol directly from "any"? + nfs = xlator_search_by_name (any, "nfs-server"); + if (!nfs) { + ret = -1; + gf_log (THIS->name, GF_LOG_ERROR, "xlator nfs-server is " + "not loaded"); + goto out; + } + + subvol = xlator_search_by_name (nfs, volname); + if (!subvol) { + ret = -1; + gf_log (THIS->name, GF_LOG_ERROR, "xlator %s is no loaded", + volname); + goto out; + } + + output = dict_new (); + ret = subvol->notify (subvol, GF_EVENT_TRANSLATOR_INFO, dict, output); + + rsp.op_ret = ret; + rsp.op_errno = 0; + rsp.op_errstr = ""; + + ret = dict_allocate_and_serialize (output, &rsp.output.output_val, + &rsp.output.output_len); + if (ret) { + gf_log (THIS->name, GF_LOG_ERROR, + "Failed to serialize output dict to rsp"); + goto out; + } + + glusterfs_submit_reply (req, &rsp, NULL, 0, NULL, + (xdrproc_t)xdr_gd1_mgmt_brick_op_rsp); + ret = 0; + +out: + free (nfs_req.input.input_val); + if (dict) + dict_unref (dict); + if (output) + dict_unref (output); + GF_FREE (rsp.output.output_val); + + gf_log (THIS->name, GF_LOG_DEBUG, "Returning %d", ret); + return ret; +} + +int +glusterfs_handle_volume_barrier_op (rpcsvc_request_t *req) +{ + int32_t ret = -1; + gd1_mgmt_brick_op_req xlator_req = {0,}; + dict_t *dict = NULL; + xlator_t *xlator = NULL; + xlator_t *any = NULL; + dict_t *output = NULL; + char msg[2048] = {0}; + glusterfs_ctx_t *ctx = NULL; + glusterfs_graph_t *active = NULL; + xlator_t *this = NULL; + + GF_ASSERT (req); + this = THIS; + GF_ASSERT (this); + + ctx = glusterfsd_ctx; + GF_ASSERT (ctx); + + active = ctx->active; + if (!active) { + req->rpc_err = GARBAGE_ARGS; + goto out; + } + + any = active->first; + ret = xdr_to_generic (req->msg[0], &xlator_req, + (xdrproc_t)xdr_gd1_mgmt_brick_op_req); + if (ret < 0) { + //failed to decode msg; + req->rpc_err = GARBAGE_ARGS; + goto out; + } + dict = dict_new (); + if (!dict) + goto out; + + ret = dict_unserialize (xlator_req.input.input_val, + xlator_req.input.input_len, + &dict); + if (ret < 0) { + gf_log (this->name, GF_LOG_ERROR, + "failed to " + "unserialize req-buffer to dictionary"); + goto out; + } + xlator = xlator_search_by_name (any, xlator_req.name); + if (!xlator) { + snprintf (msg, sizeof (msg), "xlator %s is not loaded", + xlator_req.name); + goto out; + } + + output = dict_new (); + if (!output) { + ret = -1; + goto out; + } + + ret = xlator->notify (xlator, GF_EVENT_VOLUME_BARRIER_OP, + dict, output); + + ret = glusterfs_translator_info_response_send (req, ret, + msg, output); +out: + if (dict) + dict_unref (dict); + free (xlator_req.input.input_val); // malloced by xdr + if (output) + dict_unref (output); + free (xlator_req.name); //malloced by xdr + + return ret; + +} +int +glusterfs_handle_rpc_msg (rpcsvc_request_t *req) +{ + int ret = -1; + /* for now, nothing */ + return ret; +} + +rpcclnt_cb_actor_t mgmt_cbk_actors[] = { + [GF_CBK_FETCHSPEC] = {"FETCHSPEC", GF_CBK_FETCHSPEC, mgmt_cbk_spec }, + [GF_CBK_EVENT_NOTIFY] = {"EVENTNOTIFY", GF_CBK_EVENT_NOTIFY, + mgmt_cbk_event}, +}; + + +struct rpcclnt_cb_program mgmt_cbk_prog = { + .progname = "GlusterFS Callback", + .prognum = GLUSTER_CBK_PROGRAM, + .progver = GLUSTER_CBK_VERSION, + .actors = mgmt_cbk_actors, + .numactors = GF_CBK_MAXVALUE, +}; + +char *clnt_pmap_procs[GF_PMAP_MAXVALUE] = { + [GF_PMAP_NULL] = "NULL", + [GF_PMAP_PORTBYBRICK] = "PORTBYBRICK", + [GF_PMAP_BRICKBYPORT] = "BRICKBYPORT", + [GF_PMAP_SIGNIN] = "SIGNIN", + [GF_PMAP_SIGNOUT] = "SIGNOUT", + [GF_PMAP_SIGNUP] = "SIGNUP", +}; + + +rpc_clnt_prog_t clnt_pmap_prog = { + .progname = "Gluster Portmap", + .prognum = GLUSTER_PMAP_PROGRAM, + .progver = GLUSTER_PMAP_VERSION, + .procnames = clnt_pmap_procs, +}; + +char *clnt_handshake_procs[GF_HNDSK_MAXVALUE] = { + [GF_HNDSK_NULL] = "NULL", + [GF_HNDSK_SETVOLUME] = "SETVOLUME", + [GF_HNDSK_GETSPEC] = "GETSPEC", + [GF_HNDSK_PING] = "PING", + [GF_HNDSK_EVENT_NOTIFY] = "EVENTNOTIFY", +}; + +rpc_clnt_prog_t clnt_handshake_prog = { + .progname = "GlusterFS Handshake", + .prognum = GLUSTER_HNDSK_PROGRAM, + .progver = GLUSTER_HNDSK_VERSION, + .procnames = clnt_handshake_procs, +}; + +rpcsvc_actor_t glusterfs_actors[] = { + [GLUSTERD_BRICK_NULL] = {"NULL", GLUSTERD_BRICK_NULL, glusterfs_handle_rpc_msg, NULL, 0, DRC_NA}, + [GLUSTERD_BRICK_TERMINATE] = {"TERMINATE", GLUSTERD_BRICK_TERMINATE, glusterfs_handle_terminate, NULL, 0, DRC_NA}, + [GLUSTERD_BRICK_XLATOR_INFO] = {"TRANSLATOR INFO", GLUSTERD_BRICK_XLATOR_INFO, glusterfs_handle_translator_info_get, NULL, 0, DRC_NA}, + [GLUSTERD_BRICK_XLATOR_OP] = {"TRANSLATOR OP", GLUSTERD_BRICK_XLATOR_OP, glusterfs_handle_translator_op, NULL, 0, DRC_NA}, + [GLUSTERD_BRICK_STATUS] = {"STATUS", GLUSTERD_BRICK_STATUS, glusterfs_handle_brick_status, NULL, 0, DRC_NA}, + [GLUSTERD_BRICK_XLATOR_DEFRAG] = {"TRANSLATOR DEFRAG", GLUSTERD_BRICK_XLATOR_DEFRAG, glusterfs_handle_defrag, NULL, 0, DRC_NA}, + [GLUSTERD_NODE_PROFILE] = {"NFS PROFILE", GLUSTERD_NODE_PROFILE, glusterfs_handle_nfs_profile, NULL, 0, DRC_NA}, + [GLUSTERD_NODE_STATUS] = {"NFS STATUS", GLUSTERD_NODE_STATUS, glusterfs_handle_node_status, NULL, 0, DRC_NA}, + [GLUSTERD_VOLUME_BARRIER_OP] = {"VOLUME BARRIER OP", GLUSTERD_VOLUME_BARRIER_OP, glusterfs_handle_volume_barrier_op, NULL, 0, DRC_NA}, +}; + +struct rpcsvc_program glusterfs_mop_prog = { + .progname = "Gluster Brick operations", + .prognum = GD_BRICK_PROGRAM, + .progver = GD_BRICK_VERSION, + .actors = glusterfs_actors, + .numactors = GLUSTERD_BRICK_MAXVALUE, + .synctask = _gf_true, +}; + +int +mgmt_submit_request (void *req, call_frame_t *frame, + glusterfs_ctx_t *ctx, + rpc_clnt_prog_t *prog, int procnum, + fop_cbk_fn_t cbkfn, xdrproc_t xdrproc) +{ + int ret = -1; + int count = 0; + struct iovec iov = {0, }; + struct iobuf *iobuf = NULL; + struct iobref *iobref = NULL; + ssize_t xdr_size = 0; + + iobref = iobref_new (); + if (!iobref) { + goto out; + } + + if (req) { + xdr_size = xdr_sizeof (xdrproc, req); + + iobuf = iobuf_get2 (ctx->iobuf_pool, xdr_size); + if (!iobuf) { + goto out; + }; + + iobref_add (iobref, iobuf); + + iov.iov_base = iobuf->ptr; + iov.iov_len = iobuf_pagesize (iobuf); + + /* Create the xdr payload */ + ret = xdr_serialize_generic (iov, req, xdrproc); + if (ret == -1) { + gf_log (THIS->name, GF_LOG_WARNING, "failed to create XDR payload"); + goto out; + } + iov.iov_len = ret; + count = 1; + } + + /* Send the msg */ + ret = rpc_clnt_submit (ctx->mgmt, prog, procnum, cbkfn, + &iov, count, + NULL, 0, iobref, frame, NULL, 0, NULL, 0, NULL); + +out: + if (iobref) + iobref_unref (iobref); + + if (iobuf) + iobuf_unref (iobuf); + return ret; +} + + +/* XXX: move these into @ctx */ +static char *oldvolfile = NULL; +static int oldvollen = 0; + + + +int +mgmt_getspec_cbk (struct rpc_req *req, struct iovec *iov, int count, + void *myframe) +{ + gf_getspec_rsp rsp = {0,}; + call_frame_t *frame = NULL; + glusterfs_ctx_t *ctx = NULL; + int ret = 0; + ssize_t size = 0; + FILE *tmpfp = NULL; + char *volfilebuf = NULL; + + frame = myframe; + ctx = frame->this->ctx; + + if (-1 == req->rpc_status) { + ret = -1; + goto out; + } + + ret = xdr_to_generic (*iov, &rsp, (xdrproc_t)xdr_gf_getspec_rsp); + if (ret < 0) { + gf_log (frame->this->name, GF_LOG_ERROR, "XDR decoding error"); + ret = -1; + goto out; + } + + if (-1 == rsp.op_ret) { + gf_log (frame->this->name, GF_LOG_ERROR, + "failed to get the 'volume file' from server"); + ret = rsp.op_errno; + goto out; + } + + ret = 0; + size = rsp.op_ret; + + if (size == oldvollen && (memcmp (oldvolfile, rsp.spec, size) == 0)) { + gf_log (frame->this->name, GF_LOG_INFO, + "No change in volfile, continuing"); + goto out; + } + + tmpfp = tmpfile (); + if (!tmpfp) { + ret = -1; + goto out; + } + + fwrite (rsp.spec, size, 1, tmpfp); + fflush (tmpfp); + if (ferror (tmpfp)) { + ret = -1; + goto out; + } + + /* Check if only options have changed. No need to reload the + * volfile if topology hasn't changed. + * glusterfs_volfile_reconfigure returns 3 possible return states + * return 0 =======> reconfiguration of options has succeeded + * return 1 =======> the graph has to be reconstructed and all the xlators should be inited + * return -1(or -ve) =======> Some Internal Error occurred during the operation + */ + + ret = glusterfs_volfile_reconfigure (oldvollen, tmpfp, ctx, oldvolfile); + if (ret == 0) { + gf_log ("glusterfsd-mgmt", GF_LOG_DEBUG, + "No need to re-load volfile, reconfigure done"); + if (oldvolfile) + volfilebuf = GF_REALLOC (oldvolfile, size); + else + volfilebuf = GF_CALLOC (1, size, gf_common_mt_char); + if (!volfilebuf) { + ret = -1; + goto out; + } + oldvolfile = volfilebuf; + oldvollen = size; + memcpy (oldvolfile, rsp.spec, size); + goto out; + } + + if (ret < 0) { + gf_log ("glusterfsd-mgmt", GF_LOG_DEBUG, "Reconfigure failed !!"); + goto out; + } + + ret = glusterfs_process_volfp (ctx, tmpfp); + /* tmpfp closed */ + tmpfp = NULL; + if (ret) + goto out; + + if (oldvolfile) + volfilebuf = GF_REALLOC (oldvolfile, size); + else + volfilebuf = GF_CALLOC (1, size, gf_common_mt_char); + + if (!volfilebuf) { + ret = -1; + goto out; + } + oldvolfile = volfilebuf; + oldvollen = size; + memcpy (oldvolfile, rsp.spec, size); + if (!is_mgmt_rpc_reconnect) { + glusterfs_mgmt_pmap_signin (ctx); + is_mgmt_rpc_reconnect = _gf_true; + } + +out: + STACK_DESTROY (frame->root); + + free (rsp.spec); + + emancipate (ctx, ret); + + // Stop if server is running at an unsupported op-version + if (ENOTSUP == ret) { + gf_log ("mgmt", GF_LOG_ERROR, "Server is operating at an " + "op-version which is not supported"); + cleanup_and_exit (0); + } + + if (ret && ctx && !ctx->active) { + /* Do it only for the first time */ + /* Failed to get the volume file, something wrong, + restart the process */ + gf_log ("mgmt", GF_LOG_ERROR, + "failed to fetch volume file (key:%s)", + ctx->cmd_args.volfile_id); + cleanup_and_exit (0); + } + + + if (tmpfp) + fclose (tmpfp); + + return 0; +} + + +int +glusterfs_volfile_fetch (glusterfs_ctx_t *ctx) +{ + cmd_args_t *cmd_args = NULL; + gf_getspec_req req = {0, }; + int ret = 0; + call_frame_t *frame = NULL; + dict_t *dict = NULL; + + cmd_args = &ctx->cmd_args; + + frame = create_frame (THIS, ctx->pool); + + req.key = cmd_args->volfile_id; + req.flags = 0; + + dict = dict_new (); + if (!dict) { + ret = -1; + goto out; + } + + // Set the supported min and max op-versions, so glusterd can make a + // decision + ret = dict_set_int32 (dict, "min-op-version", GD_OP_VERSION_MIN); + if (ret) { + gf_log (THIS->name, GF_LOG_ERROR, "Failed to set min-op-version" + " in request dict"); + goto out; + } + + ret = dict_set_int32 (dict, "max-op-version", GD_OP_VERSION_MAX); + if (ret) { + gf_log (THIS->name, GF_LOG_ERROR, "Failed to set max-op-version" + " in request dict"); + goto out; + } + + ret = dict_allocate_and_serialize (dict, &req.xdata.xdata_val, + &req.xdata.xdata_len); + if (ret < 0) { + gf_log (THIS->name, GF_LOG_ERROR, + "Failed to serialize dictionary"); + goto out; + } + + ret = mgmt_submit_request (&req, frame, ctx, &clnt_handshake_prog, + GF_HNDSK_GETSPEC, mgmt_getspec_cbk, + (xdrproc_t)xdr_gf_getspec_req); +out: + return ret; +} + +int32_t +mgmt_event_notify_cbk (struct rpc_req *req, struct iovec *iov, int count, + void *myframe) +{ + gf_event_notify_rsp rsp = {0,}; + call_frame_t *frame = NULL; + glusterfs_ctx_t *ctx = NULL; + int ret = 0; + + frame = myframe; + ctx = frame->this->ctx; + + if (-1 == req->rpc_status) { + ret = -1; + goto out; + } + + ret = xdr_to_generic (*iov, &rsp, (xdrproc_t)xdr_gf_event_notify_rsp); + if (ret < 0) { + gf_log (frame->this->name, GF_LOG_ERROR, "XDR decoding error"); + ret = -1; + goto out; + } + + if (-1 == rsp.op_ret) { + gf_log (frame->this->name, GF_LOG_ERROR, + "failed to get the rsp from server"); + ret = -1; + goto out; + } +out: + free (rsp.dict.dict_val); //malloced by xdr + return ret; + +} + +int32_t +glusterfs_rebalance_event_notify_cbk (struct rpc_req *req, struct iovec *iov, + int count, void *myframe) +{ + gf_event_notify_rsp rsp = {0,}; + call_frame_t *frame = NULL; + glusterfs_ctx_t *ctx = NULL; + int ret = 0; + + frame = myframe; + ctx = frame->this->ctx; + + if (-1 == req->rpc_status) { + gf_log (frame->this->name, GF_LOG_ERROR, + "failed to get the rsp from server"); + ret = -1; + goto out; + } + + ret = xdr_to_generic (*iov, &rsp, (xdrproc_t)xdr_gf_event_notify_rsp); + if (ret < 0) { + gf_log (frame->this->name, GF_LOG_ERROR, "XDR decoding error"); + ret = -1; + goto out; + } + + if (-1 == rsp.op_ret) { + gf_log (frame->this->name, GF_LOG_ERROR, + "Received error (%s) from server", + strerror (rsp.op_errno)); + ret = -1; + goto out; + } +out: + free (rsp.dict.dict_val); //malloced by xdr + return ret; + +} + +int32_t +glusterfs_rebalance_event_notify (dict_t *dict) +{ + glusterfs_ctx_t *ctx = NULL; + gf_event_notify_req req = {0,}; + int32_t ret = -1; + cmd_args_t *cmd_args = NULL; + call_frame_t *frame = NULL; + + ctx = glusterfsd_ctx; + cmd_args = &ctx->cmd_args; + + frame = create_frame (THIS, ctx->pool); + + req.op = GF_EN_DEFRAG_STATUS; + + if (dict) { + ret = dict_set_str (dict, "volname", cmd_args->volfile_id); + if (ret) + gf_log ("", GF_LOG_ERROR, "failed to set volname"); + + ret = dict_allocate_and_serialize (dict, &req.dict.dict_val, + &req.dict.dict_len); + } + + ret = mgmt_submit_request (&req, frame, ctx, &clnt_handshake_prog, + GF_HNDSK_EVENT_NOTIFY, + glusterfs_rebalance_event_notify_cbk, + (xdrproc_t)xdr_gf_event_notify_req); + + GF_FREE (req.dict.dict_val); + + STACK_DESTROY (frame->root); + return ret; +} + +static int +mgmt_rpc_notify (struct rpc_clnt *rpc, void *mydata, rpc_clnt_event_t event, + void *data) +{ + xlator_t *this = NULL; + glusterfs_ctx_t *ctx = NULL; + int ret = 0; + server_cmdline_t *server = NULL; + rpc_transport_t *rpc_trans = NULL; + int need_term = 0; + int emval = 0; + + this = mydata; + rpc_trans = rpc->conn.trans; + ctx = this->ctx; + + switch (event) { + case RPC_CLNT_DISCONNECT: + if (!ctx->active) { + gf_log ("glusterfsd-mgmt", GF_LOG_ERROR, + "failed to connect with remote-host: %s (%s)", + ctx->cmd_args.volfile_server, + strerror (errno)); + server = ctx->cmd_args.curr_server; + if (server->list.next == &ctx->cmd_args.volfile_servers) { + need_term = 1; + emval = ENOTCONN; + gf_log("glusterfsd-mgmt", GF_LOG_INFO, + "Exhausted all volfile servers"); + break; + } + server = list_entry (server->list.next, typeof(*server), + list); + ctx->cmd_args.curr_server = server; + ctx->cmd_args.volfile_server = server->volfile_server; + + ret = dict_set_str (rpc_trans->options, + "remote-host", + server->volfile_server); + if (ret != 0) { + gf_log ("glusterfsd-mgmt", GF_LOG_ERROR, + "failed to set remote-host: %s", + server->volfile_server); + need_term = 1; + emval = ENOTCONN; + break; + } + gf_log ("glusterfsd-mgmt", GF_LOG_INFO, + "connecting to next volfile server %s", + server->volfile_server); + } + break; + case RPC_CLNT_CONNECT: + rpc_clnt_set_connected (&((struct rpc_clnt*)ctx->mgmt)->conn); + + ret = glusterfs_volfile_fetch (ctx); + if (ret) { + emval = ret; + if (!ctx->active) { + need_term = 1; + gf_log ("glusterfsd-mgmt", GF_LOG_ERROR, + "failed to fetch volume file (key:%s)", + ctx->cmd_args.volfile_id); + break; + + } + } + + if (is_mgmt_rpc_reconnect) + glusterfs_mgmt_pmap_signin (ctx); + + break; + default: + break; + } + + if (need_term) { + emancipate (ctx, emval); + cleanup_and_exit (1); + } + + return 0; +} + +int +glusterfs_rpcsvc_notify (rpcsvc_t *rpc, void *xl, rpcsvc_event_t event, + void *data) +{ + if (!xl || !data) { + goto out; + } + + switch (event) { + case RPCSVC_EVENT_ACCEPT: + { + break; + } + case RPCSVC_EVENT_DISCONNECT: + { + break; + } + + default: + break; + } + +out: + return 0; +} + +int +glusterfs_listener_init (glusterfs_ctx_t *ctx) +{ + cmd_args_t *cmd_args = NULL; + rpcsvc_t *rpc = NULL; + dict_t *options = NULL; + int ret = -1; + + cmd_args = &ctx->cmd_args; + + if (ctx->listener) + return 0; + + if (!cmd_args->sock_file) + return 0; + + ret = rpcsvc_transport_unix_options_build (&options, + cmd_args->sock_file); + if (ret) + goto out; + + rpc = rpcsvc_init (THIS, ctx, options, 8); + if (rpc == NULL) { + goto out; + } + + ret = rpcsvc_register_notify (rpc, glusterfs_rpcsvc_notify, THIS); + if (ret) { + goto out; + } + + ret = rpcsvc_create_listeners (rpc, options, "glusterfsd"); + if (ret < 1) { + ret = -1; + goto out; + } + + ret = rpcsvc_program_register (rpc, &glusterfs_mop_prog); + if (ret) { + goto out; + } + + ctx->listener = rpc; + +out: + return ret; +} + +int +glusterfs_listener_stop (glusterfs_ctx_t *ctx) +{ + cmd_args_t *cmd_args = NULL; + rpcsvc_t *rpc = NULL; + rpcsvc_listener_t *listener = NULL; + rpcsvc_listener_t *next = NULL; + int ret = 0; + xlator_t *this = NULL; + + GF_ASSERT (ctx); + + rpc = ctx->listener; + ctx->listener = NULL; + + (void) rpcsvc_program_unregister(rpc, &glusterfs_mop_prog); + + list_for_each_entry_safe (listener, next, &rpc->listeners, list) { + rpcsvc_listener_destroy (listener); + } + + (void) rpcsvc_unregister_notify (rpc, glusterfs_rpcsvc_notify, THIS); + + GF_FREE (rpc); + + cmd_args = &ctx->cmd_args; + if (cmd_args->sock_file) { + ret = unlink (cmd_args->sock_file); + if (ret && (ENOENT == errno)) { + ret = 0; + } + } + + if (ret) { + this = THIS; + gf_log (this->name, GF_LOG_ERROR, "Failed to unlink listener " + "socket %s, error: %s", cmd_args->sock_file, + strerror (errno)); + } + return ret; +} + +int +glusterfs_mgmt_notify (int32_t op, void *data, ...) +{ + int ret = 0; + switch (op) + { + case GF_EN_DEFRAG_STATUS: + ret = glusterfs_rebalance_event_notify ((dict_t*) data); + break; + + default: + gf_log ("", GF_LOG_ERROR, "Invalid op"); + break; + } + + return ret; +} + +int +glusterfs_mgmt_init (glusterfs_ctx_t *ctx) +{ + cmd_args_t *cmd_args = NULL; + struct rpc_clnt *rpc = NULL; + dict_t *options = NULL; + int ret = -1; + int port = GF_DEFAULT_BASE_PORT; + char *host = NULL; + + cmd_args = &ctx->cmd_args; + + if (ctx->mgmt) + return 0; + + if (cmd_args->volfile_server_port) + port = cmd_args->volfile_server_port; + + host = "localhost"; + if (cmd_args->volfile_server) + host = cmd_args->volfile_server; + + ret = rpc_transport_inet_options_build (&options, host, port); + if (ret) + goto out; + + rpc = rpc_clnt_new (options, THIS->ctx, THIS->name, 8); + if (!rpc) { + ret = -1; + gf_log (THIS->name, GF_LOG_WARNING, "failed to create rpc clnt"); + goto out; + } + + ret = rpc_clnt_register_notify (rpc, mgmt_rpc_notify, THIS); + if (ret) { + gf_log (THIS->name, GF_LOG_WARNING, + "failed to register notify function"); + goto out; + } + + ret = rpcclnt_cbk_program_register (rpc, &mgmt_cbk_prog, THIS); + if (ret) { + gf_log (THIS->name, GF_LOG_WARNING, + "failed to register callback function"); + goto out; + } + + ctx->notify = glusterfs_mgmt_notify; + + /* This value should be set before doing the 'rpc_clnt_start()' as + the notify function uses this variable */ + ctx->mgmt = rpc; + + ret = rpc_clnt_start (rpc); +out: + return ret; +} + +static int +mgmt_pmap_signin2_cbk (struct rpc_req *req, struct iovec *iov, int count, + void *myframe) +{ + pmap_signin_rsp rsp = {0,}; + call_frame_t *frame = NULL; + int ret = 0; + + frame = myframe; + + if (-1 == req->rpc_status) { + rsp.op_ret = -1; + rsp.op_errno = EINVAL; + goto out; + } + + ret = xdr_to_generic (*iov, &rsp, (xdrproc_t)xdr_pmap_signin_rsp); + if (ret < 0) { + gf_log (frame->this->name, GF_LOG_ERROR, "XDR decode error"); + rsp.op_ret = -1; + rsp.op_errno = EINVAL; + goto out; + } + + if (-1 == rsp.op_ret) { + gf_log (frame->this->name, GF_LOG_ERROR, + "failed to register the port with glusterd"); + goto out; + } +out: + STACK_DESTROY (frame->root); + return 0; + +} + +static int +mgmt_pmap_signin_cbk (struct rpc_req *req, struct iovec *iov, int count, + void *myframe) +{ + pmap_signin_rsp rsp = {0,}; + call_frame_t *frame = NULL; + int ret = 0; + pmap_signin_req pmap_req = {0, }; + cmd_args_t *cmd_args = NULL; + glusterfs_ctx_t *ctx = NULL; + char brick_name[PATH_MAX] = {0,}; + + frame = myframe; + + if (-1 == req->rpc_status) { + rsp.op_ret = -1; + rsp.op_errno = EINVAL; + goto out; + } + + ret = xdr_to_generic (*iov, &rsp, (xdrproc_t)xdr_pmap_signin_rsp); + if (ret < 0) { + gf_log (frame->this->name, GF_LOG_ERROR, "XDR decode error"); + rsp.op_ret = -1; + rsp.op_errno = EINVAL; + goto out; + } + + if (-1 == rsp.op_ret) { + gf_log (frame->this->name, GF_LOG_ERROR, + "failed to register the port with glusterd"); + goto out; + } + + ctx = glusterfsd_ctx; + cmd_args = &ctx->cmd_args; + + if (!cmd_args->brick_port2) { + /* We are done with signin process */ + goto out; + } + + snprintf (brick_name, PATH_MAX, "%s.rdma", cmd_args->brick_name); + pmap_req.port = cmd_args->brick_port2; + pmap_req.brick = brick_name; + + ret = mgmt_submit_request (&pmap_req, frame, ctx, &clnt_pmap_prog, + GF_PMAP_SIGNIN, mgmt_pmap_signin2_cbk, + (xdrproc_t)xdr_pmap_signin_req); + if (ret) + goto out; + + return 0; + +out: + + STACK_DESTROY (frame->root); + return 0; +} + +int +glusterfs_mgmt_pmap_signin (glusterfs_ctx_t *ctx) +{ + call_frame_t *frame = NULL; + pmap_signin_req req = {0, }; + int ret = -1; + cmd_args_t *cmd_args = NULL; + + frame = create_frame (THIS, ctx->pool); + cmd_args = &ctx->cmd_args; + + if (!cmd_args->brick_port || !cmd_args->brick_name) { + gf_log ("fsd-mgmt", GF_LOG_DEBUG, + "portmapper signin arguments not given"); + goto out; + } + + req.port = cmd_args->brick_port; + req.brick = cmd_args->brick_name; + + ret = mgmt_submit_request (&req, frame, ctx, &clnt_pmap_prog, + GF_PMAP_SIGNIN, mgmt_pmap_signin_cbk, + (xdrproc_t)xdr_pmap_signin_req); + +out: + return ret; +} + + +static int +mgmt_pmap_signout_cbk (struct rpc_req *req, struct iovec *iov, int count, + void *myframe) +{ + pmap_signout_rsp rsp = {0,}; + int ret = 0; + glusterfs_ctx_t *ctx = NULL; + + if (-1 == req->rpc_status) { + rsp.op_ret = -1; + rsp.op_errno = EINVAL; + goto out; + } + + ctx = glusterfsd_ctx; + ret = xdr_to_generic (*iov, &rsp, (xdrproc_t)xdr_pmap_signout_rsp); + if (ret < 0) { + gf_log (THIS->name, GF_LOG_ERROR, "XDR decoding failed"); + rsp.op_ret = -1; + rsp.op_errno = EINVAL; + goto out; + } + + if (-1 == rsp.op_ret) { + gf_log (THIS->name, GF_LOG_ERROR, + "failed to register the port with glusterd"); + goto out; + } +out: + return 0; +} + + +int +glusterfs_mgmt_pmap_signout (glusterfs_ctx_t *ctx) +{ + int ret = 0; + pmap_signout_req req = {0, }; + call_frame_t *frame = NULL; + cmd_args_t *cmd_args = NULL; + + frame = create_frame (THIS, ctx->pool); + cmd_args = &ctx->cmd_args; + + if (!cmd_args->brick_port || !cmd_args->brick_name) { + gf_log ("fsd-mgmt", GF_LOG_DEBUG, + "portmapper signout arguments not given"); + goto out; + } + + req.port = cmd_args->brick_port; + req.brick = cmd_args->brick_name; + + ret = mgmt_submit_request (&req, frame, ctx, &clnt_pmap_prog, + GF_PMAP_SIGNOUT, mgmt_pmap_signout_cbk, + (xdrproc_t)xdr_pmap_signout_req); +out: + return ret; +} diff --git a/glusterfsd/src/glusterfsd.c b/glusterfsd/src/glusterfsd.c index 37eefa01b..3cb8f0f51 100644 --- a/glusterfsd/src/glusterfsd.c +++ b/glusterfsd/src/glusterfsd.c @@ -1,22 +1,12 @@ /* - Copyright (c) 2006-2009 Z RESEARCH, Inc. <http://www.zresearch.com> + Copyright (c) 2006-2013 Red Hat, Inc. <http://www.redhat.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. */ - #include <stdio.h> #include <string.h> #include <stdlib.h> @@ -24,6 +14,8 @@ #include <sys/socket.h> #include <sys/types.h> #include <sys/resource.h> +#include <sys/file.h> +#include <sys/wait.h> #include <netdb.h> #include <signal.h> #include <libgen.h> @@ -32,6 +24,12 @@ #include <stdint.h> #include <pthread.h> +#include <sys/stat.h> +#include <fcntl.h> +#include <time.h> +#include <semaphore.h> +#include <errno.h> +#include <pwd.h> #ifndef _CONFIG_H #define _CONFIG_H @@ -53,7 +51,6 @@ #include "compat.h" #include "logging.h" #include "dict.h" -#include "protocol.h" #include "list.h" #include "timer.h" #include "glusterfsd.h" @@ -61,1079 +58,1913 @@ #include "revision.h" #include "common-utils.h" #include "event.h" - +#include "globals.h" +#include "statedump.h" +#include "latency.h" +#include "glusterfsd-mem-types.h" +#include "syscall.h" +#include "call-stub.h" #include <fnmatch.h> +#include "rpc-clnt.h" +#include "syncop.h" +#include "client_t.h" + +#include "daemon.h" + +/* process mode definitions */ +#define GF_SERVER_PROCESS 0 +#define GF_CLIENT_PROCESS 1 +#define GF_GLUSTERD_PROCESS 2 /* using argp for command line parsing */ static char gf_doc[] = ""; static char argp_doc[] = "--volfile-server=SERVER [MOUNT-POINT]\n" \ - "--volfile=VOLFILE [MOUNT-POINT]"; -const char *argp_program_version = "" \ - PACKAGE_NAME" "PACKAGE_VERSION" built on "__DATE__" "__TIME__ \ - "\nRepository revision: " GLUSTERFS_REPOSITORY_REVISION "\n" \ - "Copyright (c) 2006-2009 Z RESEARCH Inc. " \ - "<http://www.zresearch.com>\n" \ - "GlusterFS comes with ABSOLUTELY NO WARRANTY.\n" \ - "You may redistribute copies of GlusterFS under the terms of "\ - "the GNU General Public License."; + "--volfile=VOLFILE [MOUNT-POINT]"; +const char *argp_program_version = "" + PACKAGE_NAME" "PACKAGE_VERSION" built on "__DATE__" "__TIME__ + "\nRepository revision: " GLUSTERFS_REPOSITORY_REVISION "\n" + "Copyright (c) 2006-2013 Red Hat, Inc. <http://www.redhat.com/>\n" + "GlusterFS comes with ABSOLUTELY NO WARRANTY.\n" + "It is licensed to you under your choice of the GNU Lesser\n" + "General Public License, version 3 or any later version (LGPLv3\n" + "or later), or the GNU General Public License, version 2 (GPLv2),\n" + "in all cases as published by the Free Software Foundation."; const char *argp_program_bug_address = "<" PACKAGE_BUGREPORT ">"; -error_t parse_opts (int32_t key, char *arg, struct argp_state *_state); +static error_t parse_opts (int32_t key, char *arg, struct argp_state *_state); static struct argp_option gf_options[] = { - {0, 0, 0, 0, "Basic options:"}, - {"volfile-server", ARGP_VOLFILE_SERVER_KEY, "SERVER", 0, - "Server to get the volume file from. This option overrides " - "--volfile option"}, - {"volfile", ARGP_VOLUME_FILE_KEY, "VOLFILE", 0, - "File to use as VOLUME_FILE [default: "DEFAULT_CLIENT_VOLUME_FILE" or " - DEFAULT_SERVER_VOLUME_FILE"]"}, - {"spec-file", ARGP_VOLUME_FILE_KEY, "VOLFILE", OPTION_HIDDEN, - "File to use as VOLFILE [default : "DEFAULT_CLIENT_VOLUME_FILE" or " - DEFAULT_SERVER_VOLUME_FILE"]"}, - {"log-level", ARGP_LOG_LEVEL_KEY, "LOGLEVEL", 0, - "Logging severity. Valid options are DEBUG, NORMAL, WARNING, ERROR, " - "CRITICAL and NONE [default: NORMAL]"}, - {"log-file", ARGP_LOG_FILE_KEY, "LOGFILE", 0, - "File to use for logging [default: " - DEFAULT_LOG_FILE_DIRECTORY "/" PACKAGE_NAME ".log" "]"}, - - {0, 0, 0, 0, "Advanced Options:"}, - {"volfile-server-port", ARGP_VOLFILE_SERVER_PORT_KEY, "PORT", 0, - "Listening port number of volfile server"}, - {"volfile-server-transport", ARGP_VOLFILE_SERVER_TRANSPORT_KEY, - "TRANSPORT", 0, - "Transport type to get volfile from server [default: socket]"}, - {"volfile-id", ARGP_VOLFILE_ID_KEY, "KEY", 0, - "'key' of the volfile to be fetched from server"}, - {"pid-file", ARGP_PID_FILE_KEY, "PIDFILE", 0, - "File to use as pid file"}, - {"no-daemon", ARGP_NO_DAEMON_KEY, 0, 0, - "Run in foreground"}, - {"run-id", ARGP_RUN_ID_KEY, "RUN-ID", OPTION_HIDDEN, - "Run ID for the process, used by scripts to keep track of process " - "they started, defaults to none"}, - {"debug", ARGP_DEBUG_KEY, 0, 0, - "Run in debug mode. This option sets --no-daemon, --log-level " - "to DEBUG and --log-file to console"}, - {"volume-name", ARGP_VOLUME_NAME_KEY, "VOLUME-NAME", 0, - "Volume name to be used for MOUNT-POINT [default: top most volume " - "in VOLFILE]"}, - {"xlator-option", ARGP_XLATOR_OPTION_KEY,"VOLUME-NAME.OPTION=VALUE", 0, - "Add/override a translator option for a volume with specified value"}, - - {0, 0, 0, 0, "Fuse options:"}, - {"disable-direct-io-mode", ARGP_DISABLE_DIRECT_IO_MODE_KEY, 0, 0, - "Disable direct I/O mode in fuse kernel module"}, - {"entry-timeout", ARGP_ENTRY_TIMEOUT_KEY, "SECONDS", 0, - "Set entry timeout to SECONDS in fuse kernel module [default: 1]"}, - {"attribute-timeout", ARGP_ATTRIBUTE_TIMEOUT_KEY, "SECONDS", 0, - "Set attribute timeout to SECONDS for inodes in fuse kernel module " - "[default: 1]"}, - {"volfile-check", ARGP_VOLFILE_CHECK_KEY, 0, 0, - "enable strict volume file check in fuse notify"}, + {0, 0, 0, 0, "Basic options:"}, + {"volfile-server", ARGP_VOLFILE_SERVER_KEY, "SERVER", 0, + "Server to get the volume file from. This option overrides " + "--volfile option"}, + {"volfile", ARGP_VOLUME_FILE_KEY, "VOLFILE", 0, + "File to use as VOLUME_FILE"}, + {"spec-file", ARGP_VOLUME_FILE_KEY, "VOLFILE", OPTION_HIDDEN, + "File to use as VOLUME FILE"}, + + {"log-level", ARGP_LOG_LEVEL_KEY, "LOGLEVEL", 0, + "Logging severity. Valid options are DEBUG, INFO, WARNING, ERROR, " + "CRITICAL, TRACE and NONE [default: INFO]"}, + {"log-file", ARGP_LOG_FILE_KEY, "LOGFILE", 0, + "File to use for logging [default: " + DEFAULT_LOG_FILE_DIRECTORY "/" PACKAGE_NAME ".log" "]"}, + + {0, 0, 0, 0, "Advanced Options:"}, + {"volfile-server-port", ARGP_VOLFILE_SERVER_PORT_KEY, "PORT", 0, + "Listening port number of volfile server"}, + {"volfile-server-transport", ARGP_VOLFILE_SERVER_TRANSPORT_KEY, + "TRANSPORT", 0, + "Transport type to get volfile from server [default: socket]"}, + {"volfile-id", ARGP_VOLFILE_ID_KEY, "KEY", 0, + "'key' of the volfile to be fetched from server"}, + {"pid-file", ARGP_PID_FILE_KEY, "PIDFILE", 0, + "File to use as pid file"}, + {"socket-file", ARGP_SOCK_FILE_KEY, "SOCKFILE", 0, + "File to use as unix-socket"}, + {"no-daemon", ARGP_NO_DAEMON_KEY, 0, 0, + "Run in foreground"}, + {"run-id", ARGP_RUN_ID_KEY, "RUN-ID", OPTION_HIDDEN, + "Run ID for the process, used by scripts to keep track of process " + "they started, defaults to none"}, + {"debug", ARGP_DEBUG_KEY, 0, 0, + "Run in debug mode. This option sets --no-daemon, --log-level " + "to DEBUG and --log-file to console"}, + {"volume-name", ARGP_VOLUME_NAME_KEY, "XLATOR-NAME", 0, + "Translator name to be used for MOUNT-POINT [default: top most volume " + "definition in VOLFILE]"}, + {"xlator-option", ARGP_XLATOR_OPTION_KEY,"XLATOR-NAME.OPTION=VALUE", 0, + "Add/override an option for a translator in volume file with specified" + " value"}, + {"read-only", ARGP_READ_ONLY_KEY, 0, 0, + "Mount the filesystem in 'read-only' mode"}, + {"acl", ARGP_ACL_KEY, 0, 0, + "Mount the filesystem with POSIX ACL support"}, + {"selinux", ARGP_SELINUX_KEY, 0, 0, + "Enable SELinux label (extened attributes) support on inodes"}, +#ifdef GF_LINUX_HOST_OS + {"aux-gfid-mount", ARGP_AUX_GFID_MOUNT_KEY, 0, 0, + "Enable access to filesystem through gfid directly"}, +#endif + {"enable-ino32", ARGP_INODE32_KEY, "BOOL", OPTION_ARG_OPTIONAL, + "Use 32-bit inodes when mounting to workaround broken applications" + "that don't support 64-bit inodes"}, + {"worm", ARGP_WORM_KEY, 0, 0, + "Mount the filesystem in 'worm' mode"}, + {"mac-compat", ARGP_MAC_COMPAT_KEY, "BOOL", OPTION_ARG_OPTIONAL, + "Provide stubs for attributes needed for seamless operation on Macs " #ifdef GF_DARWIN_HOST_OS - {"non-local", ARGP_NON_LOCAL_KEY, 0, 0, - "Mount the macfuse volume without '-o local' option"}, + "[default: \"on\" on client side, else \"off\"]" +#else + "[default: \"off\"]" #endif - {0, 0, 0, 0, "Miscellaneous Options:"}, - {0, } + }, + {"brick-name", ARGP_BRICK_NAME_KEY, "BRICK-NAME", OPTION_HIDDEN, + "Brick name to be registered with Gluster portmapper" }, + {"brick-port", ARGP_BRICK_PORT_KEY, "BRICK-PORT", OPTION_HIDDEN, + "Brick Port to be registered with Gluster portmapper" }, + {"fopen-keep-cache", ARGP_FOPEN_KEEP_CACHE_KEY, "BOOL", OPTION_ARG_OPTIONAL, + "Do not purge the cache on file open"}, + + {0, 0, 0, 0, "Fuse options:"}, + {"direct-io-mode", ARGP_DIRECT_IO_MODE_KEY, "BOOL", OPTION_ARG_OPTIONAL, + "Use direct I/O mode in fuse kernel module" + " [default: \"off\" if big writes are supported, else " + "\"on\" for fds not opened with O_RDONLY]"}, + {"entry-timeout", ARGP_ENTRY_TIMEOUT_KEY, "SECONDS", 0, + "Set entry timeout to SECONDS in fuse kernel module [default: 1]"}, + {"negative-timeout", ARGP_NEGATIVE_TIMEOUT_KEY, "SECONDS", 0, + "Set negative timeout to SECONDS in fuse kernel module [default: 0]"}, + {"attribute-timeout", ARGP_ATTRIBUTE_TIMEOUT_KEY, "SECONDS", 0, + "Set attribute timeout to SECONDS for inodes in fuse kernel module " + "[default: 1]"}, + {"gid-timeout", ARGP_GID_TIMEOUT_KEY, "SECONDS", 0, + "Set auxilary group list timeout to SECONDS for fuse translator " + "[default: 0]"}, + {"background-qlen", ARGP_FUSE_BACKGROUND_QLEN_KEY, "N", 0, + "Set fuse module's background queue length to N " + "[default: 64]"}, + {"congestion-threshold", ARGP_FUSE_CONGESTION_THRESHOLD_KEY, "N", 0, + "Set fuse module's congestion threshold to N " + "[default: 48]"}, + {"client-pid", ARGP_CLIENT_PID_KEY, "PID", OPTION_HIDDEN, + "client will authenticate itself with process id PID to server"}, + {"user-map-root", ARGP_USER_MAP_ROOT_KEY, "USER", OPTION_HIDDEN, + "replace USER with root in messages"}, + {"dump-fuse", ARGP_DUMP_FUSE_KEY, "PATH", 0, + "Dump fuse traffic to PATH"}, + {"volfile-check", ARGP_VOLFILE_CHECK_KEY, 0, 0, + "Enable strict volume file checking"}, + {"mem-accounting", ARGP_MEM_ACCOUNTING_KEY, 0, OPTION_HIDDEN, + "Enable internal memory accounting"}, + {"fuse-mountopts", ARGP_FUSE_MOUNTOPTS_KEY, "OPTIONS", OPTION_HIDDEN, + "Extra mount options to pass to FUSE"}, + {"use-readdirp", ARGP_FUSE_USE_READDIRP_KEY, "BOOL", OPTION_ARG_OPTIONAL, + "Use readdirp mode in fuse kernel module" + " [default: \"off\"]"}, + {0, 0, 0, 0, "Miscellaneous Options:"}, + {0, } }; static struct argp argp = { gf_options, parse_opts, argp_doc, gf_doc }; +int glusterfs_pidfile_cleanup (glusterfs_ctx_t *ctx); +int glusterfs_volumes_init (glusterfs_ctx_t *ctx); +int glusterfs_mgmt_init (glusterfs_ctx_t *ctx); +int glusterfs_listener_init (glusterfs_ctx_t *ctx); +int glusterfs_listener_stop (glusterfs_ctx_t *ctx); + -static void -_gf_dump_details (int argc, char **argv) +static int +set_fuse_mount_options (glusterfs_ctx_t *ctx, dict_t *options) { - extern FILE *gf_log_logfile; - int i = 0; - char timestr[256]; - time_t utime = 0; - struct tm *tm = NULL; - pid_t mypid = 0; - struct utsname uname_buf = {{0, }, }; - int uname_ret = -1; - - utime = time (NULL); - tm = localtime (&utime); - mypid = getpid (); - uname_ret = uname (&uname_buf); - - /* Which TLA? What time? */ - strftime (timestr, 256, "%Y-%m-%d %H:%M:%S", tm); - fprintf (gf_log_logfile, - "========================================" - "========================================\n"); - fprintf (gf_log_logfile, "Version : %s %s built on %s %s\n", - PACKAGE_NAME, PACKAGE_VERSION, __DATE__, __TIME__); - fprintf (gf_log_logfile, "TLA Revision : %s\n", - GLUSTERFS_REPOSITORY_REVISION); - fprintf (gf_log_logfile, "Starting Time: %s\n", timestr); - fprintf (gf_log_logfile, "Command line : "); - for (i = 0; i < argc; i++) { - fprintf (gf_log_logfile, "%s ", argv[i]); + int ret = 0; + cmd_args_t *cmd_args = NULL; + char *mount_point = NULL; + char cwd[PATH_MAX] = {0,}; + + cmd_args = &ctx->cmd_args; + + /* Check if mount-point is absolute path, + * if not convert to absolute path by concating with CWD + */ + if (cmd_args->mount_point[0] != '/') { + if (getcwd (cwd, PATH_MAX) != NULL) { + ret = gf_asprintf (&mount_point, "%s/%s", cwd, + cmd_args->mount_point); + if (ret == -1) { + gf_log ("glusterfsd", GF_LOG_ERROR, + "Could not create absolute mountpoint " + "path"); + goto err; + } + } else { + gf_log ("glusterfsd", GF_LOG_ERROR, + "Could not get current working directory"); + goto err; + } + } else + mount_point = gf_strdup (cmd_args->mount_point); + + ret = dict_set_dynstr (options, ZR_MOUNTPOINT_OPT, mount_point); + if (ret < 0) { + gf_log ("glusterfsd", GF_LOG_ERROR, + "failed to set mount-point to options dictionary"); + goto err; } - fprintf (gf_log_logfile, "\nPID : %d\n", mypid); - - if (uname_ret == 0) { - fprintf (gf_log_logfile, "System name : %s\n", uname_buf.sysname); - fprintf (gf_log_logfile, "Nodename : %s\n", uname_buf.nodename); - fprintf (gf_log_logfile, "Kernel Release : %s\n", uname_buf.release); - fprintf (gf_log_logfile, "Hardware Identifier: %s\n", uname_buf.machine); - } + if (cmd_args->fuse_attribute_timeout >= 0) { + ret = dict_set_double (options, ZR_ATTR_TIMEOUT_OPT, + cmd_args->fuse_attribute_timeout); + if (ret < 0) { + gf_log ("glusterfsd", GF_LOG_ERROR, + "failed to set dict value for key %s", + ZR_ATTR_TIMEOUT_OPT); + goto err; + } + } - fprintf (gf_log_logfile, "\n"); - fflush (gf_log_logfile); -} + if (cmd_args->fuse_entry_timeout >= 0) { + ret = dict_set_double (options, ZR_ENTRY_TIMEOUT_OPT, + cmd_args->fuse_entry_timeout); + if (ret < 0) { + gf_log ("glusterfsd", GF_LOG_ERROR, + "failed to set dict value for key %s", + ZR_ENTRY_TIMEOUT_OPT); + goto err; + } + } + + if (cmd_args->fuse_negative_timeout >= 0) { + ret = dict_set_double (options, ZR_NEGATIVE_TIMEOUT_OPT, + cmd_args->fuse_negative_timeout); + if (ret < 0) { + gf_log ("glusterfsd", GF_LOG_ERROR, + "failed to set dict value for key %s", + ZR_NEGATIVE_TIMEOUT_OPT); + goto err; + } + } + if (cmd_args->client_pid_set) { + ret = dict_set_int32 (options, "client-pid", + cmd_args->client_pid); + if (ret < 0) { + gf_log ("glusterfsd", GF_LOG_ERROR, + "failed to set dict value for key %s", + "client-pid"); + goto err; + } + } + if (cmd_args->uid_map_root) { + ret = dict_set_int32 (options, "uid-map-root", + cmd_args->uid_map_root); + if (ret < 0) { + gf_log ("glusterfsd", GF_LOG_ERROR, + "failed to set dict value for key %s", + "uid-map-root"); + goto err; + } + } -static xlator_t * -_add_fuse_mount (xlator_t *graph) -{ - int ret = 0; - cmd_args_t *cmd_args = NULL; - xlator_t *top = NULL; - glusterfs_ctx_t *ctx = NULL; - xlator_list_t *xlchild = NULL; - - ctx = graph->ctx; - cmd_args = &ctx->cmd_args; - - xlchild = CALLOC (sizeof (*xlchild), 1); - ERR_ABORT (xlchild); - xlchild->xlator = graph; - - top = CALLOC (1, sizeof (*top)); - top->name = strdup ("fuse"); - if (xlator_set_type (top, ZR_XLATOR_FUSE) == -1) { - fprintf (stderr, - "MOUNT-POINT %s initialization failed", - cmd_args->mount_point); - gf_log ("glusterfs", GF_LOG_ERROR, - "MOUNT-POINT %s initialization failed", - cmd_args->mount_point); - return NULL; - } - top->children = xlchild; - top->ctx = graph->ctx; - top->next = graph; - top->options = get_new_dict (); - - ret = dict_set_static_ptr (top->options, ZR_MOUNTPOINT_OPT, - cmd_args->mount_point); - if (ret < 0) { - gf_log ("glusterfs", GF_LOG_DEBUG, - "failed to set mount-point to options dictionary"); - } - - if (cmd_args->fuse_attribute_timeout) - ret = dict_set_double (top->options, ZR_ATTR_TIMEOUT_OPT, - cmd_args->fuse_attribute_timeout); - if (cmd_args->fuse_entry_timeout) - ret = dict_set_double (top->options, ZR_ENTRY_TIMEOUT_OPT, - cmd_args->fuse_entry_timeout); - - if (cmd_args->volfile_check) - ret = dict_set_int32 (top->options, ZR_STRICT_VOLFILE_CHECK, + if (cmd_args->volfile_check) { + ret = dict_set_int32 (options, ZR_STRICT_VOLFILE_CHECK, cmd_args->volfile_check); + if (ret < 0) { + gf_log ("glusterfsd", GF_LOG_ERROR, + "failed to set dict value for key %s", + ZR_STRICT_VOLFILE_CHECK); + goto err; + } + } -#ifdef GF_DARWIN_HOST_OS - /* On Darwin machines, O_APPEND is not handled, - * which may corrupt the data - */ - if (cmd_args->fuse_direct_io_mode_flag == _gf_true) { - gf_log ("glusterfs", GF_LOG_DEBUG, - "'direct-io-mode' in fuse causes data corruption " - "if O_APPEND is used. disabling 'direct-io-mode'"); - } - ret = dict_set_static_ptr (top->options, ZR_DIRECT_IO_OPT, "disable"); - - if (cmd_args->non_local) - ret = dict_set_uint32 (top->options, "macfuse-local", - cmd_args->non_local); - -#else /* ! DARWIN HOST OS */ - if (cmd_args->fuse_direct_io_mode_flag == _gf_true) { - ret = dict_set_static_ptr (top->options, ZR_DIRECT_IO_OPT, - "enable"); - } else { - ret = dict_set_static_ptr (top->options, ZR_DIRECT_IO_OPT, - "disable"); - } + if (cmd_args->dump_fuse) { + ret = dict_set_static_ptr (options, ZR_DUMP_FUSE, + cmd_args->dump_fuse); + if (ret < 0) { + gf_log ("glusterfsd", GF_LOG_ERROR, + "failed to set dict value for key %s", + ZR_DUMP_FUSE); + goto err; + } + } -#endif /* GF_DARWIN_HOST_OS */ - - graph->parents = CALLOC (1, sizeof (xlator_list_t)); - graph->parents->xlator = top; - - return top; -} + if (cmd_args->acl) { + ret = dict_set_static_ptr (options, "acl", "on"); + if (ret < 0) { + gf_log ("glusterfsd", GF_LOG_ERROR, + "failed to set dict value for key acl"); + goto err; + } + } + if (cmd_args->selinux) { + ret = dict_set_static_ptr (options, "selinux", "on"); + if (ret < 0) { + gf_log ("glusterfsd", GF_LOG_ERROR, + "failed to set dict value for key selinux"); + goto err; + } + } -static FILE * -_get_specfp (glusterfs_ctx_t *ctx) -{ - int ret = 0; - cmd_args_t *cmd_args = NULL; - FILE *specfp = NULL; - struct stat statbuf; - - cmd_args = &ctx->cmd_args; - - if (cmd_args->volfile_server) { - specfp = fetch_spec (ctx); - - if (specfp == NULL) { - fprintf (stderr, - "error while getting volume file from " - "server %s\n", cmd_args->volfile_server); - gf_log ("glusterfs", GF_LOG_ERROR, - "error while getting volume file from " - "server %s", cmd_args->volfile_server); + if (cmd_args->aux_gfid_mount) { + ret = dict_set_static_ptr (options, "virtual-gfid-access", + "on"); + if (ret < 0) { + gf_log ("glusterfsd", GF_LOG_ERROR, + "failed to set dict value for key " + "aux-gfid-mount"); + goto err; + } + } + + if (cmd_args->enable_ino32) { + ret = dict_set_static_ptr (options, "enable-ino32", "on"); + if (ret < 0) { + gf_log ("glusterfsd", GF_LOG_ERROR, + "failed to set dict value for key enable-ino32"); + goto err; + } + } + + if (cmd_args->read_only) { + ret = dict_set_static_ptr (options, "read-only", "on"); + if (ret < 0) { + gf_log ("glusterfsd", GF_LOG_ERROR, + "failed to set dict value for key read-only"); + goto err; + } + } + + switch (cmd_args->fopen_keep_cache) { + case GF_OPTION_ENABLE: + ret = dict_set_static_ptr(options, "fopen-keep-cache", + "on"); + if (ret < 0) { + gf_log("glusterfsd", GF_LOG_ERROR, + "failed to set dict value for key " + "fopen-keep-cache"); + goto err; } - else { - gf_log ("glusterfs", GF_LOG_DEBUG, - "loading volume file from server %s", - cmd_args->volfile_server); + break; + case GF_OPTION_DISABLE: + ret = dict_set_static_ptr(options, "fopen-keep-cache", + "off"); + if (ret < 0) { + gf_log("glusterfsd", GF_LOG_ERROR, + "failed to set dict value for key " + "fopen-keep-cache"); + goto err; } - - return specfp; - } - - ret = stat (cmd_args->volume_file, &statbuf); - if (ret == -1) { - fprintf (stderr, "%s: %s\n", - cmd_args->volume_file, strerror (errno)); - gf_log ("glusterfs", GF_LOG_ERROR, - "%s: %s", cmd_args->volume_file, strerror (errno)); - return NULL; - } - if (!(S_ISREG (statbuf.st_mode) || S_ISLNK (statbuf.st_mode))) { - fprintf (stderr, - "provide a valid volume file\n"); - gf_log ("glusterfs", GF_LOG_ERROR, - "provide a valid volume file"); - return NULL; - } - if ((specfp = fopen (cmd_args->volume_file, "r")) == NULL) { - fprintf (stderr, "volume file %s: %s\n", - cmd_args->volume_file, - strerror (errno)); - gf_log ("glusterfs", GF_LOG_ERROR, - "volume file %s: %s", - cmd_args->volume_file, - strerror (errno)); - return NULL; + break; + case GF_OPTION_DEFERRED: /* default */ + default: + gf_log ("glusterfsd", GF_LOG_DEBUG, + "fopen-keep-cache mode %d", + cmd_args->fopen_keep_cache); + break; } - - gf_log ("glusterfs", GF_LOG_DEBUG, - "loading volume file %s", cmd_args->volume_file); - - return specfp; -} -static xlator_t * -_parse_specfp (glusterfs_ctx_t *ctx, - FILE *specfp) -{ - int spec_fd = 0; - cmd_args_t *cmd_args = NULL; - xlator_t *tree = NULL, *trav = NULL, *new_tree = NULL; - - cmd_args = &ctx->cmd_args; - - fseek (specfp, 0L, SEEK_SET); - - tree = file_to_xlator_tree (ctx, specfp); - trav = tree; - - if (tree == NULL) { - if (cmd_args->volfile_server) { - fprintf (stderr, - "error in parsing volume file given by " - "server %s\n", cmd_args->volfile_server); - gf_log ("glusterfs", GF_LOG_ERROR, - "error in parsing volume file given by " - "server %s", cmd_args->volfile_server); - } - else { - fprintf (stderr, - "error in parsing volume file %s\n", - cmd_args->volume_file); - gf_log ("glusterfs", GF_LOG_ERROR, - "error in parsing volume file %s", - cmd_args->volume_file); + if (cmd_args->gid_timeout) { + ret = dict_set_int32(options, "gid-timeout", + cmd_args->gid_timeout); + if (ret < 0) { + gf_log("glusterfsd", GF_LOG_ERROR, "failed to set dict " + "value for key gid-timeout"); + goto err; } - return NULL; } - - spec_fd = fileno (specfp); - get_checksum_for_file (spec_fd, &ctx->volfile_checksum); - - /* if volume_name is given, then we attach to it */ - if (cmd_args->volume_name) { - while (trav) { - if (strcmp (trav->name, cmd_args->volume_name) == 0) { - new_tree = trav; - break; - } - trav = trav->next; + if (cmd_args->background_qlen) { + ret = dict_set_int32 (options, "background-qlen", + cmd_args->background_qlen); + if (ret < 0) { + gf_log("glusterfsd", GF_LOG_ERROR, "failed to set dict " + "value for key background-qlen"); + goto err; } - - if (!trav) { - if (cmd_args->volfile_server) { - fprintf (stderr, - "volume %s not found in volume " - "file given by server %s\n", - cmd_args->volume_name, - cmd_args->volfile_server); - gf_log ("glusterfs", GF_LOG_ERROR, - "volume %s not found in volume " - "file given by server %s", - cmd_args->volume_name, - cmd_args->volfile_server); - } else { - fprintf (stderr, - "volume %s not found in volume " - "file %s\n", - cmd_args->volume_name, - cmd_args->volume_file); - gf_log ("glusterfs", GF_LOG_ERROR, - "volume %s not found in volume " - "file %s", cmd_args->volume_name, - cmd_args->volume_file); - } - return NULL; + } + if (cmd_args->congestion_threshold) { + ret = dict_set_int32 (options, "congestion-threshold", + cmd_args->congestion_threshold); + if (ret < 0) { + gf_log("glusterfsd", GF_LOG_ERROR, "failed to set dict " + "value for key congestion-threshold"); + goto err; } - tree = trav; } - return tree; -} -static int -_log_if_option_is_invalid (xlator_t *xl, data_pair_t *pair) -{ - volume_opt_list_t *vol_opt = NULL; - volume_option_t *opt = NULL; - int i = 0; - int index = 0; - int found = 0; - - /* Get the first volume_option */ - list_for_each_entry (vol_opt, &xl->volume_options, list) { - /* Warn for extra option */ - if (!vol_opt->given_opt) - break; + switch (cmd_args->fuse_direct_io_mode) { + case GF_OPTION_DISABLE: /* disable */ + ret = dict_set_static_ptr (options, ZR_DIRECT_IO_OPT, + "disable"); + if (ret < 0) { + gf_log ("glusterfsd", GF_LOG_ERROR, + "failed to set 'disable' for key %s", + ZR_DIRECT_IO_OPT); + goto err; + } + break; + case GF_OPTION_ENABLE: /* enable */ + ret = dict_set_static_ptr (options, ZR_DIRECT_IO_OPT, + "enable"); + if (ret < 0) { + gf_log ("glusterfsd", GF_LOG_ERROR, + "failed to set 'enable' for key %s", + ZR_DIRECT_IO_OPT); + goto err; + } + break; + case GF_OPTION_DEFERRED: /* default */ + default: + gf_log ("", GF_LOG_DEBUG, "fuse direct io type %d", + cmd_args->fuse_direct_io_mode); + break; + } - opt = vol_opt->given_opt; - for (index = 0; - ((index < ZR_OPTION_MAX_ARRAY_SIZE) && - (opt[index].key && opt[index].key[0])); index++) - for (i = 0; (i < ZR_VOLUME_MAX_NUM_KEY) && - opt[index].key[i]; i++) { - if (fnmatch (opt[index].key[i], - pair->key, - FNM_NOESCAPE) == 0) { - found = 1; - break; - } - } - } + if (!cmd_args->no_daemon_mode) { + ret = dict_set_static_ptr (options, "sync-to-mount", + "enable"); + if (ret < 0) { + gf_log ("glusterfsd", GF_LOG_ERROR, + "failed to set dict value for key sync-mtab"); + goto err; + } + } - if (!found) { - gf_log (xl->name, GF_LOG_WARNING, - "option '%s' is not recognized", - pair->key); - } - return 0; + if (cmd_args->use_readdirp) { + ret = dict_set_str (options, "use-readdirp", + cmd_args->use_readdirp); + if (ret < 0) { + gf_log ("glusterfsd", GF_LOG_ERROR, "failed to set dict" + " value for key use-readdirp"); + goto err; + } + } + ret = 0; +err: + return ret; } -static int -_xlator_graph_init (xlator_t *xl) +int +create_fuse_mount (glusterfs_ctx_t *ctx) { - volume_opt_list_t *vol_opt = NULL; - data_pair_t *pair = NULL; - xlator_t *trav = NULL; - int ret = -1; - - trav = xl; - - while (trav->prev) - trav = trav->prev; - - /* Validate phase */ - while (trav) { - /* Get the first volume_option */ - list_for_each_entry (vol_opt, - &trav->volume_options, list) - break; - if ((ret = - validate_xlator_volume_options (trav, - vol_opt->given_opt)) < 0) { - gf_log (trav->name, GF_LOG_ERROR, - "validating translator failed"); - return ret; - } - trav = trav->next; - } + int ret = 0; + cmd_args_t *cmd_args = NULL; + xlator_t *master = NULL; - - trav = xl; - while (trav->prev) - trav = trav->prev; - /* Initialization phase */ - while (trav) { - if (!trav->ready) { - if ((ret = xlator_tree_init (trav)) < 0) { - gf_log ("glusterfs", GF_LOG_ERROR, - "initializing translator failed"); - return ret; - } - } - trav = trav->next; - } - - /* No error in this phase, just bunch of warning if at all */ - trav = xl; - - while (trav->prev) - trav = trav->prev; - - /* Validate again phase */ - while (trav) { - pair = trav->options->members_list; - while (pair) { - _log_if_option_is_invalid (trav, pair); - pair = pair->next; - } - trav = trav->next; - } + cmd_args = &ctx->cmd_args; + + if (!cmd_args->mount_point) { + gf_log ("", GF_LOG_TRACE, + "mount point not found, not a client process"); + return 0; + } + + if (ctx->process_mode != GF_CLIENT_PROCESS) { + gf_log("glusterfsd", GF_LOG_ERROR, + "Not a client process, not performing mount operation"); + return -1; + } + + master = GF_CALLOC (1, sizeof (*master), + gfd_mt_xlator_t); + if (!master) + goto err; + + master->name = gf_strdup ("fuse"); + if (!master->name) + goto err; + + if (xlator_set_type (master, "mount/fuse") == -1) { + gf_log ("glusterfsd", GF_LOG_ERROR, + "MOUNT-POINT %s initialization failed", + cmd_args->mount_point); + goto err; + } + + master->ctx = ctx; + master->options = get_new_dict (); + if (!master->options) + goto err; + + ret = set_fuse_mount_options (ctx, master->options); + if (ret) + goto err; + + if (cmd_args->fuse_mountopts) { + ret = dict_set_static_ptr (master->options, ZR_FUSE_MOUNTOPTS, + cmd_args->fuse_mountopts); + if (ret < 0) { + gf_log ("glusterfsd", GF_LOG_ERROR, + "failed to set dict value for key %s", + ZR_FUSE_MOUNTOPTS); + goto err; + } + } - return ret; + ret = xlator_init (master); + if (ret) { + gf_log ("", GF_LOG_DEBUG, "failed to initialize fuse translator"); + goto err; + } + + ctx->master = master; + + return 0; + +err: + if (master) { + xlator_destroy (master); + } + + return 1; } -int -glusterfs_graph_init (xlator_t *graph, int fuse) -{ - volume_opt_list_t *vol_opt = NULL; - if (fuse) { - /* FUSE needs to be initialized earlier than the - other translators */ - list_for_each_entry (vol_opt, - &graph->volume_options, list) - break; - if (validate_xlator_volume_options (graph, - vol_opt->given_opt) == -1) { - gf_log (graph->name, GF_LOG_ERROR, - "validating translator failed"); - return -1; - } - if (graph->init (graph) != 0) - return -1; - - graph->ready = 1; - } - if (_xlator_graph_init (graph) == -1) - return -1; - - /* check server or fuse is given */ - if (graph->ctx->top == NULL) { - fprintf (stderr, "no valid translator loaded at the top, or" - "no mount point given. exiting\n"); - gf_log ("glusterfs", GF_LOG_ERROR, - "no valid translator loaded at the top or " - "no mount point given. exiting"); - return -1; - } +static FILE * +get_volfp (glusterfs_ctx_t *ctx) +{ + int ret = 0; + cmd_args_t *cmd_args = NULL; + FILE *specfp = NULL; + struct stat statbuf; + + cmd_args = &ctx->cmd_args; + + ret = sys_lstat (cmd_args->volfile, &statbuf); + if (ret == -1) { + gf_log ("glusterfsd", GF_LOG_ERROR, + "%s: %s", cmd_args->volfile, strerror (errno)); + return NULL; + } + + if ((specfp = fopen (cmd_args->volfile, "r")) == NULL) { + gf_log ("glusterfsd", GF_LOG_ERROR, + "volume file %s: %s", + cmd_args->volfile, + strerror (errno)); + return NULL; + } - return 0; + gf_log ("glusterfsd", GF_LOG_DEBUG, + "loading volume file %s", cmd_args->volfile); + + return specfp; } static int -gf_remember_xlator_option (struct list_head *options, char *arg) +gf_remember_backup_volfile_server (char *arg) { - glusterfs_ctx_t *ctx = NULL; - cmd_args_t *cmd_args = NULL; - xlator_cmdline_option_t *option = NULL; - int ret = -1; - char *dot = NULL; - char *equals = NULL; + glusterfs_ctx_t *ctx = NULL; + cmd_args_t *cmd_args = NULL; + int ret = -1; + server_cmdline_t *server = NULL; - ctx = get_global_ctx_ptr (); - cmd_args = &ctx->cmd_args; + ctx = glusterfsd_ctx; + if (!ctx) + goto out; + cmd_args = &ctx->cmd_args; - option = CALLOC (1, sizeof (xlator_cmdline_option_t)); - INIT_LIST_HEAD (&option->cmd_args); + if(!cmd_args) + goto out; - dot = strchr (arg, '.'); - if (!dot) - goto out; + server = GF_CALLOC (1, sizeof (server_cmdline_t), + gfd_mt_server_cmdline_t); + if (!server) + goto out; - option->volume = CALLOC ((dot - arg), sizeof (char)); - strncpy (option->volume, arg, (dot - arg)); + INIT_LIST_HEAD(&server->list); - equals = strchr (arg, '='); - if (!equals) - goto out; + server->volfile_server = gf_strdup(arg); - option->key = CALLOC ((equals - dot), sizeof (char)); - strncpy (option->key, dot + 1, (equals - dot - 1)); + if (!cmd_args->volfile_server) { + cmd_args->volfile_server = server->volfile_server; + cmd_args->curr_server = server; + } - if (!*(equals + 1)) - goto out; + if (!server->volfile_server) { + gf_log ("", GF_LOG_WARNING, + "xlator option %s is invalid", arg); + goto out; + } - option->value = strdup (equals + 1); - - list_add (&option->cmd_args, &cmd_args->xlator_options); + list_add_tail (&server->list, &cmd_args->volfile_servers); - ret = 0; + ret = 0; out: - if (ret == -1) { - if (option) { - if (option->volume) - FREE (option->volume); - if (option->key) - FREE (option->key); - if (option->value) - FREE (option->value); - - FREE (option); - } - } + if (ret == -1) { + if (server) { + GF_FREE (server->volfile_server); + GF_FREE (server); + } + } - return ret; -} + return ret; +} -static void -gf_add_cmdline_options (xlator_t *graph, cmd_args_t *cmd_args) +static int +gf_remember_xlator_option (char *arg) { - int ret = 0; - xlator_t *trav = graph; - xlator_cmdline_option_t *cmd_option = NULL; - - while (trav) { - list_for_each_entry (cmd_option, - &cmd_args->xlator_options, cmd_args) { - if (!fnmatch (cmd_option->volume, - trav->name, FNM_NOESCAPE)) { - ret = dict_set_str (trav->options, - cmd_option->key, - cmd_option->value); - if (ret == 0) { - gf_log ("glusterfs", GF_LOG_WARNING, - "adding option '%s' for " - "volume '%s' with value '%s'", - cmd_option->key, trav->name, - cmd_option->value); - } else { - gf_log ("glusterfs", GF_LOG_WARNING, - "adding option '%s' for " - "volume '%s' failed: %s", - cmd_option->key, trav->name, - strerror (-ret)); - } - } - } - trav = trav->next; - } + glusterfs_ctx_t *ctx = NULL; + cmd_args_t *cmd_args = NULL; + xlator_cmdline_option_t *option = NULL; + int ret = -1; + char *dot = NULL; + char *equals = NULL; + + ctx = glusterfsd_ctx; + cmd_args = &ctx->cmd_args; + + option = GF_CALLOC (1, sizeof (xlator_cmdline_option_t), + gfd_mt_xlator_cmdline_option_t); + if (!option) + goto out; + + INIT_LIST_HEAD (&option->cmd_args); + + dot = strchr (arg, '.'); + if (!dot) { + gf_log ("", GF_LOG_WARNING, + "xlator option %s is invalid", arg); + goto out; + } + + option->volume = GF_CALLOC ((dot - arg) + 1, sizeof (char), + gfd_mt_char); + if (!option->volume) + goto out; + + strncpy (option->volume, arg, (dot - arg)); + + equals = strchr (arg, '='); + if (!equals) { + gf_log ("", GF_LOG_WARNING, + "xlator option %s is invalid", arg); + goto out; + } + + option->key = GF_CALLOC ((equals - dot) + 1, sizeof (char), + gfd_mt_char); + if (!option->key) + goto out; + + strncpy (option->key, dot + 1, (equals - dot - 1)); + + if (!*(equals + 1)) { + gf_log ("", GF_LOG_WARNING, + "xlator option %s is invalid", arg); + goto out; + } + + option->value = gf_strdup (equals + 1); + + list_add (&option->cmd_args, &cmd_args->xlator_options); + + ret = 0; +out: + if (ret == -1) { + if (option) { + GF_FREE (option->volume); + GF_FREE (option->key); + GF_FREE (option->value); + + GF_FREE (option); + } + } + + return ret; } -error_t + +static error_t parse_opts (int key, char *arg, struct argp_state *state) { - cmd_args_t *cmd_args = NULL; - uint32_t n = 0; - double d = 0.0; - - cmd_args = state->input; - - switch (key) { - case ARGP_VOLFILE_SERVER_KEY: - cmd_args->volfile_server = strdup (arg); - break; - - case ARGP_VOLUME_FILE_KEY: - cmd_args->volume_file = strdup (arg); - break; - - case ARGP_LOG_LEVEL_KEY: - if (strcasecmp (arg, ARGP_LOG_LEVEL_NONE_OPTION) == 0) { - cmd_args->log_level = GF_LOG_NONE; - break; - } - if (strcasecmp (arg, ARGP_LOG_LEVEL_TRACE_OPTION) == 0) { - cmd_args->log_level = GF_LOG_TRACE; - break; - } - if (strcasecmp (arg, ARGP_LOG_LEVEL_CRITICAL_OPTION) == 0) { - cmd_args->log_level = GF_LOG_CRITICAL; - break; - } - if (strcasecmp (arg, ARGP_LOG_LEVEL_ERROR_OPTION) == 0) { - cmd_args->log_level = GF_LOG_ERROR; - break; - } - if (strcasecmp (arg, ARGP_LOG_LEVEL_WARNING_OPTION) == 0) { - cmd_args->log_level = GF_LOG_WARNING; - break; - } - if (strcasecmp (arg, ARGP_LOG_LEVEL_NORMAL_OPTION) == 0) { - cmd_args->log_level = GF_LOG_NORMAL; - break; - } - if (strcasecmp (arg, ARGP_LOG_LEVEL_DEBUG_OPTION) == 0) { - cmd_args->log_level = GF_LOG_DEBUG; - break; - } - - argp_failure (state, -1, 0, "unknown log level %s", arg); - break; - - case ARGP_LOG_FILE_KEY: - cmd_args->log_file = strdup (arg); - break; - - case ARGP_VOLFILE_SERVER_PORT_KEY: - n = 0; - - if (gf_string2uint_base10 (arg, &n) == 0) { - cmd_args->volfile_server_port = n; - break; - } - - argp_failure (state, -1, 0, - "unknown volfile server port %s", arg); - break; - - case ARGP_VOLFILE_SERVER_TRANSPORT_KEY: - cmd_args->volfile_server_transport = strdup (arg); - break; - - case ARGP_VOLFILE_ID_KEY: - cmd_args->volfile_id = strdup (arg); - break; + cmd_args_t *cmd_args = NULL; + uint32_t n = 0; + double d = 0.0; + gf_boolean_t b = _gf_false; + char *pwd = NULL; + char tmp_buf[2048] = {0,}; + char *tmp_str = NULL; + char *port_str = NULL; + struct passwd *pw = NULL; + + cmd_args = state->input; + + switch (key) { + case ARGP_VOLFILE_SERVER_KEY: + gf_remember_backup_volfile_server (arg); - case ARGP_PID_FILE_KEY: - cmd_args->pid_file = strdup (arg); - break; - - case ARGP_NO_DAEMON_KEY: - cmd_args->no_daemon_mode = ENABLE_NO_DAEMON_MODE; - break; - - case ARGP_RUN_ID_KEY: - cmd_args->run_id = strdup (arg); - break; - - case ARGP_DEBUG_KEY: - cmd_args->debug_mode = ENABLE_DEBUG_MODE; - break; - - case ARGP_DISABLE_DIRECT_IO_MODE_KEY: - cmd_args->fuse_direct_io_mode_flag = _gf_false; - break; - - case ARGP_ENTRY_TIMEOUT_KEY: - d = 0.0; - - gf_string2double (arg, &d); - if (!(d < 0.0)) { - cmd_args->fuse_entry_timeout = d; - break; - } - - argp_failure (state, -1, 0, "unknown entry timeout %s", arg); - break; - - case ARGP_ATTRIBUTE_TIMEOUT_KEY: - d = 0.0; + break; - gf_string2double (arg, &d); - if (!(d < 0.0)) { - cmd_args->fuse_attribute_timeout = d; - break; - } - - argp_failure (state, -1, 0, - "unknown attribute timeout %s", arg); - break; - - case ARGP_VOLFILE_CHECK_KEY: + case ARGP_READ_ONLY_KEY: + cmd_args->read_only = 1; + break; + + case ARGP_ACL_KEY: + cmd_args->acl = 1; + gf_remember_xlator_option ("*-md-cache.cache-posix-acl=true"); + break; + + case ARGP_SELINUX_KEY: + cmd_args->selinux = 1; + gf_remember_xlator_option ("*-md-cache.cache-selinux=true"); + break; + + case ARGP_AUX_GFID_MOUNT_KEY: + cmd_args->aux_gfid_mount = 1; + break; + + case ARGP_INODE32_KEY: + cmd_args->enable_ino32 = 1; + break; + + case ARGP_WORM_KEY: + cmd_args->worm = 1; + break; + + case ARGP_MAC_COMPAT_KEY: + if (!arg) + arg = "on"; + + if (gf_string2boolean (arg, &b) == 0) { + cmd_args->mac_compat = b; + + break; + } + + argp_failure (state, -1, 0, + "invalid value \"%s\" for mac-compat", arg); + break; + + case ARGP_VOLUME_FILE_KEY: + GF_FREE (cmd_args->volfile); + + if (arg[0] != '/') { + pwd = getcwd (NULL, PATH_MAX); + if (!pwd) { + argp_failure (state, -1, errno, + "getcwd failed with error no %d", + errno); + break; + } + snprintf (tmp_buf, 1024, "%s/%s", pwd, arg); + cmd_args->volfile = gf_strdup (tmp_buf); + free (pwd); + } else { + cmd_args->volfile = gf_strdup (arg); + } + + break; + + case ARGP_LOG_LEVEL_KEY: + if (strcasecmp (arg, ARGP_LOG_LEVEL_NONE_OPTION) == 0) { + cmd_args->log_level = GF_LOG_NONE; + break; + } + if (strcasecmp (arg, ARGP_LOG_LEVEL_CRITICAL_OPTION) == 0) { + cmd_args->log_level = GF_LOG_CRITICAL; + break; + } + if (strcasecmp (arg, ARGP_LOG_LEVEL_ERROR_OPTION) == 0) { + cmd_args->log_level = GF_LOG_ERROR; + break; + } + if (strcasecmp (arg, ARGP_LOG_LEVEL_WARNING_OPTION) == 0) { + cmd_args->log_level = GF_LOG_WARNING; + break; + } + if (strcasecmp (arg, ARGP_LOG_LEVEL_INFO_OPTION) == 0) { + cmd_args->log_level = GF_LOG_INFO; + break; + } + if (strcasecmp (arg, ARGP_LOG_LEVEL_DEBUG_OPTION) == 0) { + cmd_args->log_level = GF_LOG_DEBUG; + break; + } + if (strcasecmp (arg, ARGP_LOG_LEVEL_TRACE_OPTION) == 0) { + cmd_args->log_level = GF_LOG_TRACE; + break; + } + + argp_failure (state, -1, 0, "unknown log level %s", arg); + break; + + case ARGP_LOG_FILE_KEY: + cmd_args->log_file = gf_strdup (arg); + break; + + case ARGP_VOLFILE_SERVER_PORT_KEY: + n = 0; + + if (gf_string2uint_base10 (arg, &n) == 0) { + cmd_args->volfile_server_port = n; + break; + } + + argp_failure (state, -1, 0, + "unknown volfile server port %s", arg); + break; + + case ARGP_VOLFILE_SERVER_TRANSPORT_KEY: + cmd_args->volfile_server_transport = gf_strdup (arg); + break; + + case ARGP_VOLFILE_ID_KEY: + cmd_args->volfile_id = gf_strdup (arg); + break; + + case ARGP_PID_FILE_KEY: + cmd_args->pid_file = gf_strdup (arg); + break; + + case ARGP_SOCK_FILE_KEY: + cmd_args->sock_file = gf_strdup (arg); + break; + + case ARGP_NO_DAEMON_KEY: + cmd_args->no_daemon_mode = ENABLE_NO_DAEMON_MODE; + break; + + case ARGP_RUN_ID_KEY: + cmd_args->run_id = gf_strdup (arg); + break; + + case ARGP_DEBUG_KEY: + cmd_args->debug_mode = ENABLE_DEBUG_MODE; + break; + + case ARGP_DIRECT_IO_MODE_KEY: + if (!arg) + arg = "on"; + + if (gf_string2boolean (arg, &b) == 0) { + cmd_args->fuse_direct_io_mode = b; + + break; + } + + argp_failure (state, -1, 0, + "unknown direct I/O mode setting \"%s\"", arg); + break; + + case ARGP_ENTRY_TIMEOUT_KEY: + d = 0.0; + + gf_string2double (arg, &d); + if (!(d < 0.0)) { + cmd_args->fuse_entry_timeout = d; + break; + } + + argp_failure (state, -1, 0, "unknown entry timeout %s", arg); + break; + + case ARGP_NEGATIVE_TIMEOUT_KEY: + d = 0.0; + + gf_string2double (arg, &d); + if (!(d < 0.0)) { + cmd_args->fuse_negative_timeout = d; + break; + } + + argp_failure (state, -1, 0, "unknown negative timeout %s", arg); + break; + + case ARGP_ATTRIBUTE_TIMEOUT_KEY: + d = 0.0; + + gf_string2double (arg, &d); + if (!(d < 0.0)) { + cmd_args->fuse_attribute_timeout = d; + break; + } + + argp_failure (state, -1, 0, + "unknown attribute timeout %s", arg); + break; + + case ARGP_CLIENT_PID_KEY: + if (gf_string2int (arg, &cmd_args->client_pid) == 0) { + cmd_args->client_pid_set = 1; + break; + } + + argp_failure (state, -1, 0, + "unknown client pid %s", arg); + break; + + case ARGP_USER_MAP_ROOT_KEY: + pw = getpwnam (arg); + if (pw) + cmd_args->uid_map_root = pw->pw_uid; + else + argp_failure (state, -1, 0, + "user %s does not exist", arg); + break; + + case ARGP_VOLFILE_CHECK_KEY: cmd_args->volfile_check = 1; break; - case ARGP_VOLUME_NAME_KEY: - cmd_args->volume_name = strdup (arg); - break; + case ARGP_VOLUME_NAME_KEY: + cmd_args->volume_name = gf_strdup (arg); + break; - case ARGP_XLATOR_OPTION_KEY: - gf_remember_xlator_option (&cmd_args->xlator_options, arg); - break; + case ARGP_XLATOR_OPTION_KEY: + if (gf_remember_xlator_option (arg)) + argp_failure (state, -1, 0, "invalid xlator option %s", + arg); -#ifdef GF_DARWIN_HOST_OS - case ARGP_NON_LOCAL_KEY: - cmd_args->non_local = _gf_true; - break; + break; -#endif /* DARWIN */ + case ARGP_KEY_NO_ARGS: + break; + + case ARGP_KEY_ARG: + if (state->arg_num >= 1) + argp_usage (state); + + cmd_args->mount_point = gf_strdup (arg); + break; + + case ARGP_DUMP_FUSE_KEY: + cmd_args->dump_fuse = gf_strdup (arg); + break; + case ARGP_BRICK_NAME_KEY: + cmd_args->brick_name = gf_strdup (arg); + break; + case ARGP_BRICK_PORT_KEY: + n = 0; + + port_str = strtok_r (arg, ",", &tmp_str); + if (gf_string2uint_base10 (port_str, &n) == 0) { + cmd_args->brick_port = n; + port_str = strtok_r (NULL, ",", &tmp_str); + if (port_str) { + if (gf_string2uint_base10 (port_str, &n) == 0) + cmd_args->brick_port2 = n; + break; + + argp_failure (state, -1, 0, + "wrong brick (listen) port %s", arg); + } + break; + } + + argp_failure (state, -1, 0, + "unknown brick (listen) port %s", arg); + break; + + case ARGP_MEM_ACCOUNTING_KEY: + /* TODO: it should have got handled much earlier */ + //gf_mem_acct_enable_set (THIS->ctx); + break; + + case ARGP_FOPEN_KEEP_CACHE_KEY: + if (!arg) + arg = "on"; + + if (gf_string2boolean (arg, &b) == 0) { + cmd_args->fopen_keep_cache = b; + + break; + } + + argp_failure (state, -1, 0, + "unknown cache setting \"%s\"", arg); - case ARGP_KEY_NO_ARGS: break; - - case ARGP_KEY_ARG: - if (state->arg_num >= 1) - argp_usage (state); - - cmd_args->mount_point = strdup (arg); + + case ARGP_GID_TIMEOUT_KEY: + if (!gf_string2int(arg, &cmd_args->gid_timeout)) + break; + + argp_failure(state, -1, 0, "unknown group list timeout %s", arg); break; + case ARGP_FUSE_BACKGROUND_QLEN_KEY: + if (!gf_string2int (arg, &cmd_args->background_qlen)) + break; + + argp_failure (state, -1, 0, + "unknown background qlen option %s", arg); + break; + case ARGP_FUSE_CONGESTION_THRESHOLD_KEY: + if (!gf_string2int (arg, &cmd_args->congestion_threshold)) + break; + + argp_failure (state, -1, 0, + "unknown congestion threshold option %s", arg); + break; + + case ARGP_FUSE_MOUNTOPTS_KEY: + cmd_args->fuse_mountopts = gf_strdup (arg); + break; + + case ARGP_FUSE_USE_READDIRP_KEY: + if (!arg) + arg = "yes"; + + if (gf_string2boolean (arg, &b) == 0) { + if (b) { + cmd_args->use_readdirp = "yes"; + } else { + cmd_args->use_readdirp = "no"; + } + + break; + } + + argp_failure (state, -1, 0, + "unknown use-readdirp setting \"%s\"", arg); + break; + } - return 0; + return 0; } -void +void cleanup_and_exit (int signum) { - glusterfs_ctx_t *ctx = NULL; - xlator_t *trav = NULL; - - ctx = get_global_ctx_ptr (); - - gf_log ("glusterfs", GF_LOG_WARNING, "shutting down"); + glusterfs_ctx_t *ctx = NULL; + xlator_t *trav = NULL; + + ctx = glusterfsd_ctx; + + if (!ctx) + return; + + gf_log_callingfn ("", GF_LOG_WARNING, + "received signum (%d), shutting down", signum); + + if (ctx->cleanup_started) + return; + + ctx->cleanup_started = 1; + glusterfs_mgmt_pmap_signout (ctx); + + /* below part is a racy code where the rpcsvc object is freed. + * But in another thread (epoll thread), upon poll error in the + * socket the transports are cleaned up where again rpcsvc object + * is accessed (which is already freed by the below function). + * Since the process is about to be killed dont execute the function + * below. + */ + /* if (ctx->listener) { */ + /* (void) glusterfs_listener_stop (ctx); */ + /* } */ + + /* Call fini() of FUSE xlator first: + * so there are no more requests coming and + * 'umount' of mount point is done properly */ + trav = ctx->master; + if (trav && trav->fini) { + THIS = trav; + trav->fini (trav); + } - if (ctx->pidfp) { - gf_unlockfd (fileno (ctx->pidfp)); - fclose (ctx->pidfp); - ctx->pidfp = NULL; - } + glusterfs_pidfile_cleanup (ctx); - if (ctx->specfp) { - fclose (ctx->specfp); - ctx->specfp = NULL; - } + exit (0); +#if 0 + /* TODO: Properly do cleanup_and_exit(), with synchronization */ + if (ctx->mgmt) { + /* cleanup the saved-frames before last unref */ + rpc_clnt_connection_cleanup (&ctx->mgmt->conn); + rpc_clnt_unref (ctx->mgmt); + } - if (ctx->cmd_args.pid_file) { - unlink (ctx->cmd_args.pid_file); - ctx->cmd_args.pid_file = NULL; - } - - if (ctx->graph) { - trav = ctx->graph; - ctx->graph = NULL; - while (trav) { - trav->fini (trav); - trav = trav->next; - } - exit (0); - } else { - gf_log ("glusterfs", GF_LOG_DEBUG, "no graph present"); - } + /* call fini() of each xlator */ + trav = NULL; + if (ctx->active) + trav = ctx->active->top; + while (trav) { + if (trav->fini) { + THIS = trav; + trav->fini (trav); + } + trav = trav->next; + } +#endif } -static char * -zr_build_process_uuid () +static void +reincarnate (int signum) { - char tmp_str[1024] = {0,}; - char hostname[256] = {0,}; - struct timeval tv = {0,}; - struct tm now = {0, }; - char now_str[32]; - - if (-1 == gettimeofday(&tv, NULL)) { - gf_log ("", GF_LOG_ERROR, - "gettimeofday: failed %s", - strerror (errno)); - } + int ret = 0; + glusterfs_ctx_t *ctx = NULL; + cmd_args_t *cmd_args = NULL; + + ctx = glusterfsd_ctx; + cmd_args = &ctx->cmd_args; + + if (cmd_args->volfile_server) { + gf_log ("glusterfsd", GF_LOG_INFO, + "Fetching the volume file from server..."); + ret = glusterfs_volfile_fetch (ctx); + } else { + gf_log ("glusterfsd", GF_LOG_DEBUG, + "Not reloading volume specification file on SIGHUP"); + } - if (-1 == gethostname (hostname, 256)) { - gf_log ("", GF_LOG_ERROR, - "gethostname: failed %s", - strerror (errno)); - } + /* Also, SIGHUP should do logrotate */ + gf_log_logrotate (1); + + if (ret < 0) + gf_log ("glusterfsd", GF_LOG_ERROR, + "volume initialization failed."); - localtime_r (&tv.tv_sec, &now); - strftime (now_str, 32, "%Y/%m/%d-%H:%M:%S", &now); - snprintf (tmp_str, 1024, "%s-%d-%s:%ld", - hostname, getpid(), now_str, tv.tv_usec); - - return strdup (tmp_str); + return; } -#define GF_SERVER_PROCESS 0 -#define GF_CLIENT_PROCESS 1 +void +emancipate (glusterfs_ctx_t *ctx, int ret) +{ + /* break free from the parent */ + if (ctx->daemon_pipe[1] != -1) { + write (ctx->daemon_pipe[1], (void *) &ret, sizeof (ret)); + close (ctx->daemon_pipe[1]); + ctx->daemon_pipe[1] = -1; + } +} static uint8_t gf_get_process_mode (char *exec_name) { - char *dup_execname = NULL, *base = NULL; - uint8_t ret = 0; - - dup_execname = strdup (exec_name); - base = basename (dup_execname); - - if (!strncmp (base, "glusterfsd", 10)) { - ret = GF_SERVER_PROCESS; - } else { - ret = GF_CLIENT_PROCESS; - } - - free (dup_execname); + char *dup_execname = NULL, *base = NULL; + uint8_t ret = 0; + + dup_execname = gf_strdup (exec_name); + base = basename (dup_execname); + + if (!strncmp (base, "glusterfsd", 10)) { + ret = GF_SERVER_PROCESS; + } else if (!strncmp (base, "glusterd", 8)) { + ret = GF_GLUSTERD_PROCESS; + } else { + ret = GF_CLIENT_PROCESS; + } - return ret; + GF_FREE (dup_execname); + + return ret; } -int -main (int argc, char *argv[]) + +static int +glusterfs_ctx_defaults_init (glusterfs_ctx_t *ctx) { - glusterfs_ctx_t *ctx = NULL; - cmd_args_t *cmd_args = NULL; - call_pool_t *pool = NULL; - struct stat stbuf; - char tmp_logfile[1024] = { 0 }; - char timestr[256] = { 0 }; - char *base_exec_name = NULL; - time_t utime; - struct tm *tm = NULL; - int ret = 0; - struct rlimit lim; - FILE *specfp = NULL; - xlator_t *graph = NULL; - xlator_t *trav = NULL; - int fuse_volume_found = 0; - int xl_count = 0; - uint8_t process_mode = 0; - - utime = time (NULL); - ctx = CALLOC (1, sizeof (glusterfs_ctx_t)); - ERR_ABORT (ctx); - base_exec_name = strdup (argv[0]); - process_mode = gf_get_process_mode (base_exec_name); - set_global_ctx_ptr (ctx); - ctx->process_uuid = zr_build_process_uuid (); - cmd_args = &ctx->cmd_args; - - /* parsing command line arguments */ - cmd_args->log_level = DEFAULT_LOG_LEVEL; - cmd_args->fuse_direct_io_mode_flag = _gf_true; - - INIT_LIST_HEAD (&cmd_args->xlator_options); - - argp_parse (&argp, argc, argv, ARGP_IN_ORDER, NULL, cmd_args); - - if (ENABLE_DEBUG_MODE == cmd_args->debug_mode) { - cmd_args->log_level = GF_LOG_DEBUG; - cmd_args->log_file = "/dev/stdout"; - cmd_args->no_daemon_mode = ENABLE_NO_DAEMON_MODE; - } + cmd_args_t *cmd_args = NULL; + struct rlimit lim = {0, }; + int ret = -1; - if ((cmd_args->volfile_server == NULL) - && (cmd_args->volume_file == NULL)) { - if (process_mode == GF_SERVER_PROCESS) - cmd_args->volume_file = strdup (DEFAULT_SERVER_VOLUME_FILE); - else - cmd_args->volume_file = strdup (DEFAULT_CLIENT_VOLUME_FILE); - } + xlator_mem_acct_init (THIS, gfd_mt_end); - if (cmd_args->log_file == NULL) - asprintf (&cmd_args->log_file, - DEFAULT_LOG_FILE_DIRECTORY "/%s.log", - basename (base_exec_name)); - - free (base_exec_name); - - ctx->event_pool = event_pool_new (DEFAULT_EVENT_POOL_SIZE); - pthread_mutex_init (&(ctx->lock), NULL); - pool = ctx->pool = CALLOC (1, sizeof (call_pool_t)); - ERR_ABORT (ctx->pool); - LOCK_INIT (&pool->lock); - INIT_LIST_HEAD (&pool->all_frames); - - if (cmd_args->pid_file != NULL) { - ctx->pidfp = fopen (cmd_args->pid_file, "a+"); - if (ctx->pidfp == NULL) { - fprintf (stderr, - "unable to open pid file %s. %s. exiting\n", - cmd_args->pid_file, strerror (errno)); - /* do cleanup and exit ?! */ - return -1; - } - ret = gf_lockfd (fileno (ctx->pidfp)); - if (ret == -1) { - fprintf (stderr, "unable to lock pid file %s. %s. " - "Is another instance of %s running?!\n" - "exiting\n", cmd_args->pid_file, - strerror (errno), argv[0]); - fclose (ctx->pidfp); - return -1; - } - ret = ftruncate (fileno (ctx->pidfp), 0); - if (ret == -1) { - fprintf (stderr, - "unable to truncate file %s. %s. exiting\n", - cmd_args->pid_file, strerror (errno)); - gf_unlockfd (fileno (ctx->pidfp)); - fclose (ctx->pidfp); - return -1; - } - } - - /* initializing logs */ - if (cmd_args->run_id) { - ret = stat (cmd_args->log_file, &stbuf); - /* If its /dev/null, or /dev/stdout, /dev/stderr, - * let it use the same, no need to alter - */ - if (((ret == 0) && - (S_ISREG (stbuf.st_mode) || S_ISLNK (stbuf.st_mode))) || - (ret == -1)) { - /* Have seperate logfile per run */ - tm = localtime (&utime); - strftime (timestr, 256, "%Y%m%d.%H%M%S", tm); - sprintf (tmp_logfile, "%s.%s.%d", - cmd_args->log_file, timestr, getpid ()); - - /* Create symlink to actual log file */ - unlink (cmd_args->log_file); - symlink (tmp_logfile, cmd_args->log_file); - - FREE (cmd_args->log_file); - cmd_args->log_file = strdup (tmp_logfile); - } - } - - gf_global_variable_init (); - - if (gf_log_init (cmd_args->log_file) == -1) { - fprintf (stderr, - "failed to open logfile %s. exiting\n", - cmd_args->log_file); - return -1; - } - gf_log_set_loglevel (cmd_args->log_level); - - /* setting up environment */ - lim.rlim_cur = RLIM_INFINITY; - lim.rlim_max = RLIM_INFINITY; - if (setrlimit (RLIMIT_CORE, &lim) == -1) { - fprintf (stderr, "ignoring %s\n", - strerror (errno)); - } -#ifdef HAVE_MALLOC_STATS -#ifdef DEBUG - mtrace (); + ctx->process_uuid = generate_glusterfs_ctx_id (); + if (!ctx->process_uuid) { + gf_log ("", GF_LOG_CRITICAL, + "ERROR: glusterfs uuid generation failed"); + goto out; + } + + ctx->page_size = 128 * GF_UNIT_KB; + + ctx->iobuf_pool = iobuf_pool_new (); + if (!ctx->iobuf_pool) { + gf_log ("", GF_LOG_CRITICAL, + "ERROR: glusterfs iobuf pool creation failed"); + goto out; + } + + ctx->event_pool = event_pool_new (DEFAULT_EVENT_POOL_SIZE); + if (!ctx->event_pool) { + gf_log ("", GF_LOG_CRITICAL, + "ERROR: glusterfs event pool creation failed"); + goto out; + } + + ctx->pool = GF_CALLOC (1, sizeof (call_pool_t), gfd_mt_call_pool_t); + if (!ctx->pool) { + gf_log ("", GF_LOG_CRITICAL, + "ERROR: glusterfs call pool creation failed"); + goto out; + } + + INIT_LIST_HEAD (&ctx->pool->all_frames); + LOCK_INIT (&ctx->pool->lock); + + /* frame_mem_pool size 112 * 4k */ + ctx->pool->frame_mem_pool = mem_pool_new (call_frame_t, 4096); + if (!ctx->pool->frame_mem_pool) { + gf_log ("", GF_LOG_CRITICAL, + "ERROR: glusterfs frame pool creation failed"); + goto out; + } + /* stack_mem_pool size 256 * 1024 */ + ctx->pool->stack_mem_pool = mem_pool_new (call_stack_t, 1024); + if (!ctx->pool->stack_mem_pool) { + gf_log ("", GF_LOG_CRITICAL, + "ERROR: glusterfs stack pool creation failed"); + goto out; + } + + ctx->stub_mem_pool = mem_pool_new (call_stub_t, 1024); + if (!ctx->stub_mem_pool) { + gf_log ("", GF_LOG_CRITICAL, + "ERROR: glusterfs stub pool creation failed"); + goto out; + } + + ctx->dict_pool = mem_pool_new (dict_t, GF_MEMPOOL_COUNT_OF_DICT_T); + if (!ctx->dict_pool) + goto out; + + ctx->dict_pair_pool = mem_pool_new (data_pair_t, + GF_MEMPOOL_COUNT_OF_DATA_PAIR_T); + if (!ctx->dict_pair_pool) + goto out; + + ctx->dict_data_pool = mem_pool_new (data_t, GF_MEMPOOL_COUNT_OF_DATA_T); + if (!ctx->dict_data_pool) + goto out; + + pthread_mutex_init (&(ctx->lock), NULL); + + ctx->clienttable = gf_clienttable_alloc(); + if (!ctx->clienttable) + goto out; + + cmd_args = &ctx->cmd_args; + + /* parsing command line arguments */ + cmd_args->log_level = DEFAULT_LOG_LEVEL; + + cmd_args->mac_compat = GF_OPTION_DISABLE; +#ifdef GF_DARWIN_HOST_OS + /* On Darwin machines, O_APPEND is not handled, + * which may corrupt the data + */ + cmd_args->fuse_direct_io_mode = GF_OPTION_DISABLE; +#else + cmd_args->fuse_direct_io_mode = GF_OPTION_DEFERRED; #endif - signal (SIGUSR1, (sighandler_t) malloc_stats); + cmd_args->fuse_attribute_timeout = -1; + cmd_args->fuse_entry_timeout = -1; + cmd_args->fopen_keep_cache = GF_OPTION_DEFERRED; + + INIT_LIST_HEAD (&cmd_args->xlator_options); + INIT_LIST_HEAD (&cmd_args->volfile_servers); + + lim.rlim_cur = RLIM_INFINITY; + lim.rlim_max = RLIM_INFINITY; + setrlimit (RLIMIT_CORE, &lim); + + ret = 0; +out: + + if (ret && ctx) { + if (ctx->pool) { + mem_pool_destroy (ctx->pool->frame_mem_pool); + mem_pool_destroy (ctx->pool->stack_mem_pool); + } + GF_FREE (ctx->pool); + mem_pool_destroy (ctx->stub_mem_pool); + mem_pool_destroy (ctx->dict_pool); + mem_pool_destroy (ctx->dict_data_pool); + mem_pool_destroy (ctx->dict_pair_pool); + } + + return ret; +} + +static int +logging_init (glusterfs_ctx_t *ctx, const char *progpath) +{ + cmd_args_t *cmd_args = NULL; + int ret = 0; + char ident[1024] = {0,}; + char *progname = NULL; + char *ptr = NULL; + + cmd_args = &ctx->cmd_args; + + if (cmd_args->log_file == NULL) { + ret = gf_set_log_file_path (cmd_args); + if (ret == -1) { + fprintf (stderr, "ERROR: failed to set the log file path\n"); + return -1; + } + } + +#ifdef GF_USE_SYSLOG + progname = gf_strdup (progpath); + snprintf (ident, 1024, "%s_%s", basename(progname), + basename(cmd_args->log_file)); + GF_FREE (progname); + /* remove .log suffix */ + if (NULL != (ptr = strrchr(ident, '.'))) { + if (strcmp(ptr, ".log") == 0) { + /* note: ptr points to location in ident only */ + ptr[0] = '\0'; + } + } + ptr = ident; #endif - signal (SIGSEGV, gf_print_trace); - signal (SIGABRT, gf_print_trace); - signal (SIGPIPE, SIG_IGN); - signal (SIGHUP, gf_log_logrotate); - signal (SIGTERM, cleanup_and_exit); - /* This is used to dump details */ - /* signal (SIGUSR2, (sighandler_t) glusterfs_stats); */ - - /* getting and parsing volume file */ - if ((specfp = _get_specfp (ctx)) == NULL) { - /* _get_specfp() prints necessary error message */ - gf_log ("glusterfs", GF_LOG_ERROR, "exiting\n"); - argp_help (&argp, stderr, ARGP_HELP_SEE, (char *) argv[0]); - return -1; - } - _gf_dump_details (argc, argv); - if ((graph = _parse_specfp (ctx, specfp)) == NULL) { - /* _parse_specfp() prints necessary error message */ - fprintf (stderr, "exiting\n"); - gf_log ("glusterfs", GF_LOG_ERROR, "exiting"); - return -1; - } - ctx->specfp = specfp; - - /* check whether MOUNT-POINT argument and fuse volume are given - * at same time or not. If not, add argument MOUNT-POINT to graph - * as top volume if given - */ - trav = graph; - fuse_volume_found = 0; - - while (trav) { - if (strcmp (trav->type, ZR_XLATOR_FUSE) == 0) { - if (dict_get (trav->options, - ZR_MOUNTPOINT_OPT) != NULL) { - trav->ctx = graph->ctx; - fuse_volume_found = 1; - } - } - xl_count++; /* Getting this value right is very important */ - trav = trav->next; - } - - ctx->xl_count = xl_count + 1; - - if (!fuse_volume_found && (cmd_args->mount_point != NULL)) { - if ((graph = _add_fuse_mount (graph)) == NULL) { - /* _add_fuse_mount() prints necessary - * error message - */ - fprintf (stderr, "exiting\n"); - gf_log ("glusterfs", GF_LOG_ERROR, "exiting"); - return -1; - } - } - - /* daemonize now */ - if (!cmd_args->no_daemon_mode) { - if (daemon (0, 0) == -1) { - fprintf (stderr, "unable to run in daemon mode: %s", - strerror (errno)); - gf_log ("glusterfs", GF_LOG_ERROR, - "unable to run in daemon mode: %s", - strerror (errno)); - return -1; - } - - /* we are daemon now */ - /* update pid file, if given */ - if (cmd_args->pid_file != NULL) { - fprintf (ctx->pidfp, "%d\n", getpid ()); - fflush (ctx->pidfp); - /* we close pid file on exit */ - } - } - - gf_log ("glusterfs", GF_LOG_DEBUG, - "running in pid %d", getpid ()); - - gf_timer_registry_init (ctx); - - /* override xlator options with command line options - * where applicable - */ - gf_add_cmdline_options (graph, cmd_args); - - ctx->graph = graph; - if (glusterfs_graph_init (graph, fuse_volume_found) != 0) { - gf_log ("glusterfs", GF_LOG_ERROR, - "translator initialization failed. exiting"); - return -1; - } + if (gf_log_init (ctx, cmd_args->log_file, ptr) == -1) { + fprintf (stderr, "ERROR: failed to open logfile %s\n", + cmd_args->log_file); + return -1; + } + + gf_log_set_loglevel (cmd_args->log_level); + + return 0; +} + +void +gf_check_and_set_mem_acct (glusterfs_ctx_t *ctx, int argc, char *argv[]) +{ + int i = 0; + for (i = 0; i < argc; i++) { + if (strcmp (argv[i], "--mem-accounting") == 0) { + gf_mem_acct_enable_set (ctx); + break; + } + } +} + +int +parse_cmdline (int argc, char *argv[], glusterfs_ctx_t *ctx) +{ + int process_mode = 0; + int ret = 0; + struct stat stbuf = {0, }; + char timestr[32]; + char tmp_logfile[1024] = { 0 }; + char *tmp_logfile_dyn = NULL; + char *tmp_logfilebase = NULL; + cmd_args_t *cmd_args = NULL; + + cmd_args = &ctx->cmd_args; + + argp_parse (&argp, argc, argv, ARGP_IN_ORDER, NULL, cmd_args); + + if (ENABLE_DEBUG_MODE == cmd_args->debug_mode) { + cmd_args->log_level = GF_LOG_DEBUG; + cmd_args->log_file = "/dev/stderr"; + cmd_args->no_daemon_mode = ENABLE_NO_DAEMON_MODE; + } + + process_mode = gf_get_process_mode (argv[0]); + ctx->process_mode = process_mode; + + /* Make sure after the parsing cli, if '--volfile-server' option is + given, then '--volfile-id' is mandatory */ + if (cmd_args->volfile_server && !cmd_args->volfile_id) { + gf_log ("glusterfs", GF_LOG_CRITICAL, + "ERROR: '--volfile-id' is mandatory if '-s' OR " + "'--volfile-server' option is given"); + ret = -1; + goto out; + } + + if ((cmd_args->volfile_server == NULL) + && (cmd_args->volfile == NULL)) { + if (process_mode == GF_SERVER_PROCESS) + cmd_args->volfile = gf_strdup (DEFAULT_SERVER_VOLFILE); + else if (process_mode == GF_GLUSTERD_PROCESS) + cmd_args->volfile = gf_strdup (DEFAULT_GLUSTERD_VOLFILE); + else + cmd_args->volfile = gf_strdup (DEFAULT_CLIENT_VOLFILE); + + /* Check if the volfile exists, if not give usage output + and exit */ + ret = stat (cmd_args->volfile, &stbuf); + if (ret) { + gf_log ("glusterfs", GF_LOG_CRITICAL, + "ERROR: parsing the volfile failed (%s)\n", + strerror (errno)); + /* argp_usage (argp.) */ + fprintf (stderr, "USAGE: %s [options] [mountpoint]\n", + argv[0]); + goto out; + } + } + + if (cmd_args->run_id) { + ret = sys_lstat (cmd_args->log_file, &stbuf); + /* If its /dev/null, or /dev/stdout, /dev/stderr, + * let it use the same, no need to alter + */ + if (((ret == 0) && + (S_ISREG (stbuf.st_mode) || S_ISLNK (stbuf.st_mode))) || + (ret == -1)) { + /* Have separate logfile per run */ + gf_time_fmt (timestr, sizeof timestr, time (NULL), + gf_timefmt_FT); + sprintf (tmp_logfile, "%s.%s.%d", + cmd_args->log_file, timestr, getpid ()); + + /* Create symlink to actual log file */ + sys_unlink (cmd_args->log_file); + + tmp_logfile_dyn = gf_strdup (tmp_logfile); + tmp_logfilebase = basename (tmp_logfile_dyn); + ret = sys_symlink (tmp_logfilebase, + cmd_args->log_file); + if (ret == -1) { + fprintf (stderr, "ERROR: symlink of logfile failed\n"); + goto out; + } + + GF_FREE (cmd_args->log_file); + cmd_args->log_file = gf_strdup (tmp_logfile); + + GF_FREE (tmp_logfile_dyn); + } + } + +#ifdef GF_DARWIN_HOST_OS + if (cmd_args->mount_point) + cmd_args->mac_compat = GF_OPTION_DEFERRED; +#endif + + ret = 0; +out: + return ret; +} + + +int +glusterfs_pidfile_setup (glusterfs_ctx_t *ctx) +{ + cmd_args_t *cmd_args = NULL; + int ret = -1; + FILE *pidfp = NULL; + + cmd_args = &ctx->cmd_args; + + if (!cmd_args->pid_file) + return 0; + + pidfp = fopen (cmd_args->pid_file, "a+"); + if (!pidfp) { + gf_log ("glusterfsd", GF_LOG_ERROR, + "pidfile %s error (%s)", + cmd_args->pid_file, strerror (errno)); + goto out; + } + + ret = lockf (fileno (pidfp), F_TLOCK, 0); + if (ret) { + gf_log ("glusterfsd", GF_LOG_ERROR, + "pidfile %s lock error (%s)", + cmd_args->pid_file, strerror (errno)); + goto out; + } + + gf_log ("glusterfsd", GF_LOG_TRACE, + "pidfile %s lock acquired", + cmd_args->pid_file); + + ret = lockf (fileno (pidfp), F_ULOCK, 0); + if (ret) { + gf_log ("glusterfsd", GF_LOG_ERROR, + "pidfile %s unlock error (%s)", + cmd_args->pid_file, strerror (errno)); + goto out; + } + + ctx->pidfp = pidfp; + + ret = 0; +out: + if (ret && pidfp) + fclose (pidfp); + + return ret; +} + + +int +glusterfs_pidfile_cleanup (glusterfs_ctx_t *ctx) +{ + cmd_args_t *cmd_args = NULL; + + cmd_args = &ctx->cmd_args; + + if (!ctx->pidfp) + return 0; + + gf_log ("glusterfsd", GF_LOG_TRACE, + "pidfile %s cleanup", + cmd_args->pid_file); + + if (ctx->cmd_args.pid_file) { + unlink (ctx->cmd_args.pid_file); + ctx->cmd_args.pid_file = NULL; + } + + lockf (fileno (ctx->pidfp), F_ULOCK, 0); + fclose (ctx->pidfp); + ctx->pidfp = NULL; + + return 0; +} + +int +glusterfs_pidfile_update (glusterfs_ctx_t *ctx) +{ + cmd_args_t *cmd_args = NULL; + int ret = 0; + FILE *pidfp = NULL; + + cmd_args = &ctx->cmd_args; + + pidfp = ctx->pidfp; + if (!pidfp) + return 0; + + ret = lockf (fileno (pidfp), F_TLOCK, 0); + if (ret) { + gf_log ("glusterfsd", GF_LOG_ERROR, + "pidfile %s lock failed", + cmd_args->pid_file); + return ret; + } + + ret = ftruncate (fileno (pidfp), 0); + if (ret) { + gf_log ("glusterfsd", GF_LOG_ERROR, + "pidfile %s truncation failed", + cmd_args->pid_file); + return ret; + } + + ret = fprintf (pidfp, "%d\n", getpid ()); + if (ret <= 0) { + gf_log ("glusterfsd", GF_LOG_ERROR, + "pidfile %s write failed", + cmd_args->pid_file); + return ret; + } + + ret = fflush (pidfp); + if (ret) { + gf_log ("glusterfsd", GF_LOG_ERROR, + "pidfile %s write failed", + cmd_args->pid_file); + return ret; + } + + gf_log ("glusterfsd", GF_LOG_DEBUG, + "pidfile %s updated with pid %d", + cmd_args->pid_file, getpid ()); + + return 0; +} + + +void * +glusterfs_sigwaiter (void *arg) +{ + sigset_t set; + int ret = 0; + int sig = 0; + + + sigemptyset (&set); + sigaddset (&set, SIGINT); /* cleanup_and_exit */ + sigaddset (&set, SIGTERM); /* cleanup_and_exit */ + sigaddset (&set, SIGHUP); /* reincarnate */ + sigaddset (&set, SIGUSR1); /* gf_proc_dump_info */ + sigaddset (&set, SIGUSR2); /* gf_latency_toggle */ + + for (;;) { + ret = sigwait (&set, &sig); + if (ret) + continue; + + + switch (sig) { + case SIGINT: + case SIGTERM: + cleanup_and_exit (sig); + break; + case SIGHUP: + reincarnate (sig); + break; + case SIGUSR1: + gf_proc_dump_info (sig, glusterfsd_ctx); + break; + case SIGUSR2: + gf_latency_toggle (sig, glusterfsd_ctx); + break; + default: + + break; + } + } + + return NULL; +} + + +void +glusterfsd_print_trace (int signum) +{ + gf_print_trace (signum, glusterfsd_ctx); +} + + +int +glusterfs_signals_setup (glusterfs_ctx_t *ctx) +{ + sigset_t set; + int ret = 0; + + sigemptyset (&set); + + /* common setting for all threads */ + signal (SIGSEGV, glusterfsd_print_trace); + signal (SIGABRT, glusterfsd_print_trace); + signal (SIGILL, glusterfsd_print_trace); + signal (SIGTRAP, glusterfsd_print_trace); + signal (SIGFPE, glusterfsd_print_trace); + signal (SIGBUS, glusterfsd_print_trace); + signal (SIGINT, cleanup_and_exit); + signal (SIGPIPE, SIG_IGN); + + /* block these signals from non-sigwaiter threads */ + sigaddset (&set, SIGTERM); /* cleanup_and_exit */ + sigaddset (&set, SIGHUP); /* reincarnate */ + sigaddset (&set, SIGUSR1); /* gf_proc_dump_info */ + sigaddset (&set, SIGUSR2); /* gf_latency_toggle */ + + ret = pthread_sigmask (SIG_BLOCK, &set, NULL); + if (ret) { + gf_log ("glusterfsd", GF_LOG_WARNING, + "failed to execute pthread_signmask %s", + strerror (errno)); + return ret; + } + + ret = pthread_create (&ctx->sigwaiter, NULL, glusterfs_sigwaiter, + (void *) &set); + if (ret) { + /* + TODO: + fallback to signals getting handled by other threads. + setup the signal handlers + */ + gf_log ("glusterfsd", GF_LOG_WARNING, + "failed to create pthread %s", + strerror (errno)); + return ret; + } + + return ret; +} + + +int +daemonize (glusterfs_ctx_t *ctx) +{ + int ret = -1; + cmd_args_t *cmd_args = NULL; + int cstatus = 0; + int err = 0; + + cmd_args = &ctx->cmd_args; + + ret = glusterfs_pidfile_setup (ctx); + if (ret) + goto out; + + if (cmd_args->no_daemon_mode) + goto postfork; + + if (cmd_args->debug_mode) + goto postfork; + + ret = pipe (ctx->daemon_pipe); + if (ret) { + /* If pipe() fails, retain daemon_pipe[] = {-1, -1} + and parent will just not wait for child status + */ + ctx->daemon_pipe[0] = -1; + ctx->daemon_pipe[1] = -1; + } + + ret = os_daemon_return (0, 0); + switch (ret) { + case -1: + if (ctx->daemon_pipe[0] != -1) { + close (ctx->daemon_pipe[0]); + close (ctx->daemon_pipe[1]); + } + + gf_log ("daemonize", GF_LOG_ERROR, + "Daemonization failed: %s", strerror(errno)); + goto out; + case 0: + /* child */ + /* close read */ + close (ctx->daemon_pipe[0]); + break; + default: + /* parent */ + /* close write */ + close (ctx->daemon_pipe[1]); + + if (ctx->mnt_pid > 0) { + ret = waitpid (ctx->mnt_pid, &cstatus, 0); + if (!(ret == ctx->mnt_pid && cstatus == 0)) { + gf_log ("daemonize", GF_LOG_ERROR, + "mount failed"); + exit (1); + } + } + + err = 1; + read (ctx->daemon_pipe[0], (void *)&err, sizeof (err)); + _exit (err); + } + +postfork: + ret = glusterfs_pidfile_update (ctx); + if (ret) + goto out; + + glusterfs_signals_setup (ctx); +out: + return ret; +} + + +int +glusterfs_process_volfp (glusterfs_ctx_t *ctx, FILE *fp) +{ + glusterfs_graph_t *graph = NULL; + int ret = -1; + xlator_t *trav = NULL; + + graph = glusterfs_graph_construct (fp); + if (!graph) { + gf_log ("", GF_LOG_ERROR, "failed to construct the graph"); + goto out; + } + + for (trav = graph->first; trav; trav = trav->next) { + if (strcmp (trav->type, "mount/fuse") == 0) { + gf_log ("glusterfsd", GF_LOG_ERROR, + "fuse xlator cannot be specified " + "in volume file"); + goto out; + } + } + + ret = glusterfs_graph_prepare (graph, ctx); + if (ret) { + glusterfs_graph_destroy (graph); + goto out; + } + + ret = glusterfs_graph_activate (graph, ctx); + + if (ret) { + glusterfs_graph_destroy (graph); + goto out; + } - /* Send PARENT_UP notify to all the translators now */ - graph->notify (graph, GF_EVENT_PARENT_UP, ctx->graph); + gf_log_dump_graph (fp, graph); + + ret = 0; +out: + if (fp) + fclose (fp); + + if (ret && !ctx->active) { + /* there is some error in setting up the first graph itself */ + cleanup_and_exit (0); + } - gf_log ("glusterfs", GF_LOG_NORMAL, "Successfully started"); - - event_dispatch (ctx->event_pool); + return ret; +} + + +int +glusterfs_volumes_init (glusterfs_ctx_t *ctx) +{ + FILE *fp = NULL; + cmd_args_t *cmd_args = NULL; + int ret = 0; + + cmd_args = &ctx->cmd_args; + + if (cmd_args->sock_file) { + ret = glusterfs_listener_init (ctx); + if (ret) + goto out; + } + + if (cmd_args->volfile_server) { + ret = glusterfs_mgmt_init (ctx); + /* return, do not emancipate() yet */ + return ret; + } + + fp = get_volfp (ctx); + + if (!fp) { + gf_log ("glusterfsd", GF_LOG_ERROR, + "Cannot reach volume specification file"); + ret = -1; + goto out; + } + + ret = glusterfs_process_volfp (ctx, fp); + if (ret) + goto out; + +out: + emancipate (ctx, ret); + return ret; +} + + +/* This is the only legal global pointer */ +glusterfs_ctx_t *glusterfsd_ctx; + +int +main (int argc, char *argv[]) +{ + glusterfs_ctx_t *ctx = NULL; + int ret = -1; + char cmdlinestr[PATH_MAX] = {0,}; + + ctx = glusterfs_ctx_new (); + if (!ctx) { + gf_log ("glusterfs", GF_LOG_CRITICAL, + "ERROR: glusterfs context not initialized"); + return ENOMEM; + } + glusterfsd_ctx = ctx; + +#ifdef DEBUG + gf_mem_acct_enable_set (ctx); +#else + /* Enable memory accounting on the fly based on argument */ + gf_check_and_set_mem_acct (ctx, argc, argv); +#endif + + ret = glusterfs_globals_init (ctx); + if (ret) + return ret; + + THIS->ctx = ctx; + + ret = glusterfs_ctx_defaults_init (ctx); + if (ret) + goto out; + + ret = parse_cmdline (argc, argv, ctx); + if (ret) + goto out; + + ret = logging_init (ctx, argv[0]); + if (ret) + goto out; + + /* log the version of glusterfs running here along with the actual + command line options. */ + { + int i = 0; + strcpy (cmdlinestr, argv[0]); + for (i = 1; i < argc; i++) { + strcat (cmdlinestr, " "); + strcat (cmdlinestr, argv[i]); + } + gf_log (argv[0], GF_LOG_INFO, + "Started running %s version %s (%s)", + argv[0], PACKAGE_VERSION, cmdlinestr); + } + + gf_proc_dump_init(); + + ret = create_fuse_mount (ctx); + if (ret) + goto out; + + ret = daemonize (ctx); + if (ret) + goto out; + + ctx->env = syncenv_new (0, 0, 0); + if (!ctx->env) { + gf_log ("", GF_LOG_ERROR, + "Could not create new sync-environment"); + goto out; + } + + ret = glusterfs_volumes_init (ctx); + if (ret) + goto out; + + ret = event_dispatch (ctx->event_pool); + +out: +// glusterfs_ctx_destroy (ctx); - return 0; + return ret; } diff --git a/glusterfsd/src/glusterfsd.h b/glusterfsd/src/glusterfsd.h index 0d6423ab4..9e2a0e56e 100644 --- a/glusterfsd/src/glusterfsd.h +++ b/glusterfsd/src/glusterfsd.h @@ -1,22 +1,12 @@ /* - Copyright (c) 2006-2009 Z RESEARCH, Inc. <http://www.zresearch.com> + Copyright (c) 2006-2012 Red Hat, Inc. <http://www.redhat.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 __GLUSTERFSD_H__ #define __GLUSTERFSD_H__ @@ -24,11 +14,12 @@ #define _CONFIG_H #include "config.h" #endif +#include "rpcsvc.h" +#include "glusterd1-xdr.h" -#define DEFAULT_CLIENT_VOLUME_FILE CONFDIR "/glusterfs.vol" -#define DEFAULT_SERVER_VOLUME_FILE CONFDIR "/glusterfsd.vol" -#define DEFAULT_LOG_FILE_DIRECTORY DATADIR "/log/glusterfs" -#define DEFAULT_LOG_LEVEL GF_LOG_NORMAL +#define DEFAULT_GLUSTERD_VOLFILE CONFDIR "/glusterd.vol" +#define DEFAULT_CLIENT_VOLFILE CONFDIR "/glusterfs.vol" +#define DEFAULT_SERVER_VOLFILE CONFDIR "/glusterfsd.vol" #define DEFAULT_EVENT_POOL_SIZE 16384 @@ -37,44 +28,86 @@ #define ARGP_LOG_LEVEL_CRITICAL_OPTION "CRITICAL" #define ARGP_LOG_LEVEL_ERROR_OPTION "ERROR" #define ARGP_LOG_LEVEL_WARNING_OPTION "WARNING" -#define ARGP_LOG_LEVEL_NORMAL_OPTION "NORMAL" +#define ARGP_LOG_LEVEL_INFO_OPTION "INFO" #define ARGP_LOG_LEVEL_DEBUG_OPTION "DEBUG" #define ENABLE_NO_DAEMON_MODE 1 #define ENABLE_DEBUG_MODE 1 -#define ZR_XLATOR_FUSE "mount/fuse" -#define ZR_MOUNTPOINT_OPT "mountpoint" -#define ZR_ATTR_TIMEOUT_OPT "attribute-timeout" -#define ZR_ENTRY_TIMEOUT_OPT "entry-timeout" -#define ZR_DIRECT_IO_OPT "direct-io-mode" -#define ZR_STRICT_VOLFILE_CHECK "strict-volfile-check" +#define GF_MEMPOOL_COUNT_OF_DICT_T 4096 +/* Considering 4 key/value pairs in a dictionary on an average */ +#define GF_MEMPOOL_COUNT_OF_DATA_T (GF_MEMPOOL_COUNT_OF_DICT_T * 4) +#define GF_MEMPOOL_COUNT_OF_DATA_PAIR_T (GF_MEMPOOL_COUNT_OF_DICT_T * 4) enum argp_option_keys { - ARGP_VOLFILE_SERVER_KEY = 's', - ARGP_VOLUME_FILE_KEY = 'f', - ARGP_LOG_LEVEL_KEY = 'L', - ARGP_LOG_FILE_KEY = 'l', - ARGP_VOLFILE_SERVER_PORT_KEY = 131, - ARGP_VOLFILE_SERVER_TRANSPORT_KEY = 132, - ARGP_PID_FILE_KEY = 'p', - ARGP_NO_DAEMON_KEY = 'N', - ARGP_RUN_ID_KEY = 'r', - ARGP_DEBUG_KEY = 133, - ARGP_DISABLE_DIRECT_IO_MODE_KEY = 134, - ARGP_ENTRY_TIMEOUT_KEY = 135, - ARGP_ATTRIBUTE_TIMEOUT_KEY = 136, - ARGP_VOLUME_NAME_KEY = 137, - ARGP_XLATOR_OPTION_KEY = 138, + ARGP_VOLFILE_SERVER_KEY = 's', + ARGP_VOLUME_FILE_KEY = 'f', + ARGP_LOG_LEVEL_KEY = 'L', + ARGP_LOG_FILE_KEY = 'l', + ARGP_VOLFILE_SERVER_PORT_KEY = 131, + ARGP_VOLFILE_SERVER_TRANSPORT_KEY = 132, + ARGP_PID_FILE_KEY = 'p', + ARGP_SOCK_FILE_KEY = 'S', + ARGP_NO_DAEMON_KEY = 'N', + ARGP_RUN_ID_KEY = 'r', + ARGP_DEBUG_KEY = 133, + ARGP_NEGATIVE_TIMEOUT_KEY = 134, + ARGP_ENTRY_TIMEOUT_KEY = 135, + ARGP_ATTRIBUTE_TIMEOUT_KEY = 136, + ARGP_VOLUME_NAME_KEY = 137, + ARGP_XLATOR_OPTION_KEY = 138, + ARGP_DIRECT_IO_MODE_KEY = 139, #ifdef GF_DARWIN_HOST_OS - ARGP_NON_LOCAL_KEY = 139, + ARGP_NON_LOCAL_KEY = 140, #endif /* DARWIN */ - ARGP_VOLFILE_ID_KEY = 143, - ARGP_VOLFILE_CHECK_KEY = 144, + ARGP_VOLFILE_ID_KEY = 143, + ARGP_VOLFILE_CHECK_KEY = 144, + ARGP_VOLFILE_MAX_FETCH_ATTEMPTS = 145, + ARGP_LOG_SERVER_KEY = 146, + ARGP_LOG_SERVER_PORT_KEY = 147, + ARGP_READ_ONLY_KEY = 148, + ARGP_MAC_COMPAT_KEY = 149, + ARGP_DUMP_FUSE_KEY = 150, + ARGP_BRICK_NAME_KEY = 151, + ARGP_BRICK_PORT_KEY = 152, + ARGP_CLIENT_PID_KEY = 153, + ARGP_ACL_KEY = 154, + ARGP_WORM_KEY = 155, + ARGP_USER_MAP_ROOT_KEY = 156, + ARGP_MEM_ACCOUNTING_KEY = 157, + ARGP_SELINUX_KEY = 158, + ARGP_FOPEN_KEEP_CACHE_KEY = 159, + ARGP_GID_TIMEOUT_KEY = 160, + ARGP_FUSE_BACKGROUND_QLEN_KEY = 161, + ARGP_FUSE_CONGESTION_THRESHOLD_KEY = 162, + ARGP_INODE32_KEY = 163, + ARGP_FUSE_MOUNTOPTS_KEY = 164, + ARGP_FUSE_USE_READDIRP_KEY = 165, + ARGP_AUX_GFID_MOUNT_KEY = 166, +}; + +struct _gfd_vol_top_priv_t { + rpcsvc_request_t *req; + gd1_mgmt_brick_op_req xlator_req; + uint32_t blk_count; + uint32_t blk_size; + double throughput; + double time; + int32_t ret; }; +typedef struct _gfd_vol_top_priv_t gfd_vol_top_priv_t; -/* Moved here from fetch-spec.h */ -FILE *fetch_spec (glusterfs_ctx_t *ctx); +int glusterfs_mgmt_pmap_signout (glusterfs_ctx_t *ctx); +int glusterfs_mgmt_pmap_signin (glusterfs_ctx_t *ctx); +int glusterfs_volfile_fetch (glusterfs_ctx_t *ctx); +void cleanup_and_exit (int signum); +int glusterfs_volume_top_write_perf (uint32_t blk_size, uint32_t blk_count, + char *brick_path, double *throughput, + double *time); +int glusterfs_volume_top_read_perf (uint32_t blk_size, uint32_t blk_count, + char *brick_path, double *throughput, + double *time); +extern glusterfs_ctx_t *glusterfsd_ctx; #endif /* __GLUSTERFSD_H__ */ |