diff options
Diffstat (limited to 'xlators/features/locks/src')
| -rw-r--r-- | xlators/features/locks/src/Makefile.am | 9 | ||||
| -rw-r--r-- | xlators/features/locks/src/clear.c | 117 | ||||
| -rw-r--r-- | xlators/features/locks/src/clear.h | 22 | ||||
| -rw-r--r-- | xlators/features/locks/src/common.c | 293 | ||||
| -rw-r--r-- | xlators/features/locks/src/common.h | 93 | ||||
| -rw-r--r-- | xlators/features/locks/src/entrylk.c | 207 | ||||
| -rw-r--r-- | xlators/features/locks/src/inodelk.c | 347 | ||||
| -rw-r--r-- | xlators/features/locks/src/locks-mem-types.h | 21 | ||||
| -rw-r--r-- | xlators/features/locks/src/locks.h | 65 | ||||
| -rw-r--r-- | xlators/features/locks/src/posix.c | 1135 | ||||
| -rw-r--r-- | xlators/features/locks/src/reservelk.c | 31 |
11 files changed, 1659 insertions, 681 deletions
diff --git a/xlators/features/locks/src/Makefile.am b/xlators/features/locks/src/Makefile.am index e39676826..0f79731b4 100644 --- a/xlators/features/locks/src/Makefile.am +++ b/xlators/features/locks/src/Makefile.am @@ -1,7 +1,7 @@ xlator_LTLIBRARIES = locks.la xlatordir = $(libdir)/glusterfs/$(PACKAGE_VERSION)/xlator/features -locks_la_LDFLAGS = -module -avoidversion +locks_la_LDFLAGS = -module -avoid-version locks_la_SOURCES = common.c posix.c entrylk.c inodelk.c reservelk.c \ clear.c @@ -9,9 +9,10 @@ locks_la_LIBADD = $(top_builddir)/libglusterfs/src/libglusterfs.la noinst_HEADERS = locks.h common.h locks-mem-types.h clear.h -AM_CFLAGS = -fPIC -D_FILE_OFFSET_BITS=64 -D_GNU_SOURCE -Wall \ - -fno-strict-aliasing -D$(GF_HOST_OS) \ - -I$(top_srcdir)/libglusterfs/src $(GF_CFLAGS) -shared -nostartfiles +AM_CPPFLAGS = $(GF_CPPFLAGS) -I$(top_srcdir)/libglusterfs/src + + +AM_CFLAGS = -Wall -fno-strict-aliasing $(GF_CFLAGS) CLEANFILES = diff --git a/xlators/features/locks/src/clear.c b/xlators/features/locks/src/clear.c index a7b2455e3..124b9ad0f 100644 --- a/xlators/features/locks/src/clear.c +++ b/xlators/features/locks/src/clear.c @@ -1,22 +1,12 @@ /* - Copyright (c) 2006-2011 Gluster, Inc. <http://www.gluster.com> - This file is part of GlusterFS. - - GlusterFS is free software; you can redistribute it and/or modify - it under the terms of the GNU General Public License as published - by the Free Software Foundation; either version 3 of the License, - or (at your option) any later version. - - GlusterFS is distributed in the hope that it will be useful, but - WITHOUT ANY WARRANTY; without even the implied warranty of - MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU - General Public License for more details. - - You should have received a copy of the GNU General Public License - along with this program. If not, see - <http://www.gnu.org/licenses/>. -*/ + 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. +*/ #include <unistd.h> #include <fcntl.h> #include <limits.h> @@ -141,15 +131,15 @@ clrlk_parse_args (const char* cmd, clrlk_args *args) if ((args->type == CLRLK_TYPE_MAX) || (args->kind == CLRLK_KIND_MAX)) goto out; - /*optional args*/ - tok = strtok_r (NULL, ".", &sptr); + /*optional args, neither range nor basename can 'legally' contain + * "/" in them*/ + tok = strtok_r (NULL, "/", &sptr); if (tok) args->opts = gf_strdup (tok); ret = 0; out: - if (free_ptr) - GF_FREE (free_ptr); + GF_FREE (free_ptr); return ret; } @@ -193,8 +183,8 @@ clrlk_clear_posixlk (xlator_t *this, pl_inode_t *pl_inode, clrlk_args *args, F_SETLKW, &plock->user_flock, -1, EAGAIN, NULL); - STACK_UNWIND (plock->frame, -1, EAGAIN, - &plock->user_flock); + STACK_UNWIND_STRICT (lk, plock->frame, -1, EAGAIN, + &plock->user_flock, NULL); } else { gcount++; @@ -204,6 +194,7 @@ clrlk_clear_posixlk (xlator_t *this, pl_inode_t *pl_inode, clrlk_args *args, } pthread_mutex_unlock (&pl_inode->mutex); grant_blocked_locks (this, pl_inode); + ret = 0; out: *blkd = bcount; *granted = gcount; @@ -222,7 +213,9 @@ clrlk_clear_inodelk (xlator_t *this, pl_inode_t *pl_inode, pl_dom_list_t *dom, int bcount = 0; int gcount = 0; gf_boolean_t chk_range = _gf_false; + struct list_head released; + INIT_LIST_HEAD (&released); if (clrlk_get_lock_range (args->opts, &ulock, &chk_range)) { *op_errno = EINVAL; goto out; @@ -246,17 +239,23 @@ blkd: continue; bcount++; - list_del_init (&ilock->list); - pl_trace_out (this, ilock->frame, NULL, NULL, F_SETLKW, - &ilock->user_flock, -1, EAGAIN, - ilock->volume); - STACK_UNWIND_STRICT (inodelk, ilock->frame, -1, - EAGAIN); - GF_FREE (ilock); + list_del_init (&ilock->blocked_locks); + list_add (&ilock->blocked_locks, &released); } } pthread_mutex_unlock (&pl_inode->mutex); + list_for_each_entry_safe (ilock, tmp, &released, blocked_locks) { + list_del_init (&ilock->blocked_locks); + pl_trace_out (this, ilock->frame, NULL, NULL, F_SETLKW, + &ilock->user_flock, -1, EAGAIN, + ilock->volume); + STACK_UNWIND_STRICT (inodelk, ilock->frame, -1, + EAGAIN, NULL); + //No need to take lock as the locks are only in one list + __pl_inodelk_unref (ilock); + } + if (!(args->kind & CLRLK_GRANTED)) { ret = 0; goto out; @@ -275,14 +274,20 @@ granted: gcount++; list_del_init (&ilock->list); - GF_FREE (ilock); + list_add (&ilock->list, &released); } } pthread_mutex_unlock (&pl_inode->mutex); - grant_blocked_inode_locks (this, pl_inode, dom); + list_for_each_entry_safe (ilock, tmp, &released, list) { + list_del_init (&ilock->list); + //No need to take lock as the locks are only in one list + __pl_inodelk_unref (ilock); + } + ret = 0; out: + grant_blocked_inode_locks (this, pl_inode, dom); *blkd = bcount; *granted = gcount; return ret; @@ -295,11 +300,13 @@ clrlk_clear_entrylk (xlator_t *this, pl_inode_t *pl_inode, pl_dom_list_t *dom, { pl_entry_lock_t *elock = NULL; pl_entry_lock_t *tmp = NULL; - struct list_head removed = {0}; int bcount = 0; int gcount = 0; int ret = -1; + struct list_head removed; + struct list_head released; + INIT_LIST_HEAD (&released); if (args->kind & CLRLK_BLOCKED) goto blkd; @@ -311,21 +318,31 @@ blkd: { list_for_each_entry_safe (elock, tmp, &dom->blocked_entrylks, blocked_locks) { - if (args->opts && - strncmp (elock->basename, args->opts, - strlen (elock->basename))) - continue; + if (args->opts) { + if (!elock->basename || + strcmp (elock->basename, args->opts)) + continue; + } bcount++; - list_del_init (&elock->domain_list); - STACK_UNWIND_STRICT (entrylk, elock->frame, -1, - EAGAIN); - GF_FREE ((char *) elock->basename); - GF_FREE (elock); + + list_del_init (&elock->blocked_locks); + list_add_tail (&elock->blocked_locks, &released); } } pthread_mutex_unlock (&pl_inode->mutex); + list_for_each_entry_safe (elock, tmp, &released, blocked_locks) { + list_del_init (&elock->blocked_locks); + entrylk_trace_out (this, elock->frame, elock->volume, NULL, NULL, + elock->basename, ENTRYLK_LOCK, elock->type, + -1, EAGAIN); + STACK_UNWIND_STRICT (entrylk, elock->frame, -1, EAGAIN, NULL); + GF_FREE ((char *) elock->basename); + GF_FREE (elock->connection_id); + GF_FREE (elock); + } + if (!(args->kind & CLRLK_GRANTED)) { ret = 0; goto out; @@ -337,13 +354,11 @@ granted: { list_for_each_entry_safe (elock, tmp, &dom->entrylk_list, domain_list) { - if (!elock->basename) - continue; - - if (args->opts && - strncmp (elock->basename, args->opts, - strlen (elock->basename))) - continue; + if (args->opts) { + if (!elock->basename || + strcmp (elock->basename, args->opts)) + continue; + } gcount++; list_del_init (&elock->domain_list); @@ -365,8 +380,8 @@ out: int clrlk_clear_lks_in_all_domains (xlator_t *this, pl_inode_t *pl_inode, - clrlk_args *args, int *blkd, int *granted, - int *op_errno) + clrlk_args *args, int *blkd, int *granted, + int *op_errno) { pl_dom_list_t *dom = NULL; int ret = -1; diff --git a/xlators/features/locks/src/clear.h b/xlators/features/locks/src/clear.h index 95572a971..511f3f74a 100644 --- a/xlators/features/locks/src/clear.h +++ b/xlators/features/locks/src/clear.h @@ -1,22 +1,12 @@ /* - Copyright (c) 2006-2011 Gluster, Inc. <http://www.gluster.com> - This file is part of GlusterFS. + 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 __CLEAR_H__ #define __CLEAR_H__ diff --git a/xlators/features/locks/src/common.c b/xlators/features/locks/src/common.c index 0bc7baa30..b3309580d 100644 --- a/xlators/features/locks/src/common.c +++ b/xlators/features/locks/src/common.c @@ -1,22 +1,12 @@ /* - Copyright (c) 2006-2011 Gluster, Inc. <http://www.gluster.com> - This file is part of GlusterFS. - - GlusterFS is free software; you can redistribute it and/or modify - it under the terms of the GNU General Public License as published - by the Free Software Foundation; either version 3 of the License, - or (at your option) any later version. - - GlusterFS is distributed in the hope that it will be useful, but - WITHOUT ANY WARRANTY; without even the implied warranty of - MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU - General Public License for more details. - - You should have received a copy of the GNU General Public License - along with this program. If not, see - <http://www.gnu.org/licenses/>. -*/ + 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. +*/ #include <unistd.h> #include <fcntl.h> #include <limits.h> @@ -45,8 +35,9 @@ __insert_and_merge (pl_inode_t *pl_inode, posix_lock_t *lock); static int pl_send_prelock_unlock (xlator_t *this, pl_inode_t *pl_inode, posix_lock_t *old_lock); + static pl_dom_list_t * -allocate_domain (const char *volume) +__allocate_domain (const char *volume) { pl_dom_list_t *dom = NULL; @@ -85,24 +76,26 @@ get_domain (pl_inode_t *pl_inode, const char *volume) { pl_dom_list_t *dom = NULL; - GF_VALIDATE_OR_GOTO (POSIX_LOCKS, pl_inode, out); - GF_VALIDATE_OR_GOTO (POSIX_LOCKS, volume, out); - - list_for_each_entry (dom, &pl_inode->dom_list, inode_list) { - if (strcmp (dom->domain, volume) == 0) - goto found; + GF_VALIDATE_OR_GOTO ("posix-locks", pl_inode, out); + GF_VALIDATE_OR_GOTO ("posix-locks", volume, out); + pthread_mutex_lock (&pl_inode->mutex); + { + list_for_each_entry (dom, &pl_inode->dom_list, inode_list) { + if (strcmp (dom->domain, volume) == 0) + goto unlock; + } + dom = __allocate_domain (volume); + if (dom) + list_add (&dom->inode_list, &pl_inode->dom_list); } - - dom = allocate_domain (volume); - if (dom) - list_add (&dom->inode_list, &pl_inode->dom_list); -found: +unlock: + pthread_mutex_unlock (&pl_inode->mutex); if (dom) { - gf_log (POSIX_LOCKS, GF_LOG_TRACE, "Domain %s found", volume); + gf_log ("posix-locks", GF_LOG_TRACE, "Domain %s found", volume); } else { - gf_log (POSIX_LOCKS, GF_LOG_TRACE, "Domain %s not found", volume); + gf_log ("posix-locks", GF_LOG_TRACE, "Domain %s not found", volume); } out: return dom; @@ -143,10 +136,10 @@ __pl_inode_is_empty (pl_inode_t *pl_inode) void pl_print_locker (char *str, int size, xlator_t *this, call_frame_t *frame) { - snprintf (str, size, "Pid=%llu, lk-owner=%s, Transport=%p, Frame=%llu", + snprintf (str, size, "Pid=%llu, lk-owner=%s, Client=%p, Frame=%llu", (unsigned long long) frame->root->pid, lkowner_utoa (&frame->root->lk_owner), - (void *)frame->root->trans, + frame->root->client, (unsigned long long) frame->root->unique); } @@ -180,8 +173,7 @@ pl_print_lockee (char *str, int size, fd_t *fd, loc_t *loc) uuid_utoa (inode->gfid), fd, ipath ? ipath : "<nul>"); - if (ipath) - GF_FREE (ipath); + GF_FREE (ipath); } @@ -435,46 +427,50 @@ pl_inode_get (xlator_t *this, inode_t *inode) pl_inode_t *pl_inode = NULL; int ret = 0; - ret = inode_ctx_get (inode, this,&tmp_pl_inode); - if (ret == 0) { - pl_inode = (pl_inode_t *)(long)tmp_pl_inode; - goto out; - } - pl_inode = GF_CALLOC (1, sizeof (*pl_inode), - gf_locks_mt_pl_inode_t); - if (!pl_inode) { - goto out; - } + LOCK (&inode->lock); + { + ret = __inode_ctx_get (inode, this, &tmp_pl_inode); + if (ret == 0) { + pl_inode = (pl_inode_t *)(long)tmp_pl_inode; + goto unlock; + } + pl_inode = GF_CALLOC (1, sizeof (*pl_inode), + gf_locks_mt_pl_inode_t); + if (!pl_inode) { + goto unlock; + } - gf_log (this->name, GF_LOG_TRACE, - "Allocating new pl inode"); + gf_log (this->name, GF_LOG_TRACE, + "Allocating new pl inode"); - pthread_mutex_init (&pl_inode->mutex, NULL); + pthread_mutex_init (&pl_inode->mutex, NULL); - INIT_LIST_HEAD (&pl_inode->dom_list); - INIT_LIST_HEAD (&pl_inode->ext_list); - INIT_LIST_HEAD (&pl_inode->rw_list); - INIT_LIST_HEAD (&pl_inode->reservelk_list); - INIT_LIST_HEAD (&pl_inode->blocked_reservelks); - INIT_LIST_HEAD (&pl_inode->blocked_calls); + INIT_LIST_HEAD (&pl_inode->dom_list); + INIT_LIST_HEAD (&pl_inode->ext_list); + INIT_LIST_HEAD (&pl_inode->rw_list); + INIT_LIST_HEAD (&pl_inode->reservelk_list); + INIT_LIST_HEAD (&pl_inode->blocked_reservelks); + INIT_LIST_HEAD (&pl_inode->blocked_calls); - inode_ctx_put (inode, this, (uint64_t)(long)(pl_inode)); + __inode_ctx_put (inode, this, (uint64_t)(long)(pl_inode)); + } +unlock: + UNLOCK (&inode->lock); -out: return pl_inode; } /* Create a new posix_lock_t */ posix_lock_t * -new_posix_lock (struct gf_flock *flock, void *transport, pid_t client_pid, +new_posix_lock (struct gf_flock *flock, client_t *client, pid_t client_pid, gf_lkowner_t *owner, fd_t *fd) { posix_lock_t *lock = NULL; - GF_VALIDATE_OR_GOTO (POSIX_LOCKS, flock, out); - GF_VALIDATE_OR_GOTO (POSIX_LOCKS, transport, out); - GF_VALIDATE_OR_GOTO (POSIX_LOCKS, fd, out); + GF_VALIDATE_OR_GOTO ("posix-locks", flock, out); + GF_VALIDATE_OR_GOTO ("posix-locks", client, out); + GF_VALIDATE_OR_GOTO ("posix-locks", fd, out); lock = GF_CALLOC (1, sizeof (posix_lock_t), gf_locks_mt_posix_lock_t); @@ -490,7 +486,7 @@ new_posix_lock (struct gf_flock *flock, void *transport, pid_t client_pid, else lock->fl_end = flock->l_start + flock->l_len - 1; - lock->transport = transport; + lock->client = client; lock->fd_num = fd_to_fdnum (fd); lock->fd = fd; lock->client_pid = client_pid; @@ -570,7 +566,7 @@ same_owner (posix_lock_t *l1, posix_lock_t *l2) { return (is_same_lkowner (&l1->owner, &l2->owner) && - (l1->transport == l2->transport)); + (l1->client == l2->client)); } @@ -699,7 +695,7 @@ subtract_locks (posix_lock_t *big, posix_lock_t *small) } GF_ASSERT (0); - gf_log (POSIX_LOCKS, GF_LOG_ERROR, "Unexpected case in subtract_locks"); + gf_log ("posix-locks", GF_LOG_ERROR, "Unexpected case in subtract_locks"); out: if (v.locks[0]) { @@ -719,6 +715,36 @@ done: return v; } +static posix_lock_t * +first_conflicting_overlap (pl_inode_t *pl_inode, posix_lock_t *lock) +{ + posix_lock_t *l = NULL; + posix_lock_t *conf = NULL; + + pthread_mutex_lock (&pl_inode->mutex); + { + list_for_each_entry (l, &pl_inode->ext_list, list) { + if (l->blocked) + continue; + + if (locks_overlap (l, lock)) { + if (same_owner (l, lock)) + continue; + + if ((l->fl_type == F_WRLCK) || + (lock->fl_type == F_WRLCK)) { + conf = l; + goto unlock; + } + } + } + } +unlock: + pthread_mutex_unlock (&pl_inode->mutex); + + return conf; +} + /* Start searching from {begin}, and return the first lock that conflicts, NULL if no conflict @@ -777,6 +803,8 @@ __insert_and_merge (pl_inode_t *pl_inode, posix_lock_t *lock) struct _values v = { .locks = {0, 0, 0} }; list_for_each_entry_safe (conf, t, &pl_inode->ext_list, list) { + if (conf->blocked) + continue; if (!locks_overlap (conf, lock)) continue; @@ -785,7 +813,7 @@ __insert_and_merge (pl_inode_t *pl_inode, posix_lock_t *lock) sum = add_locks (lock, conf); sum->fl_type = lock->fl_type; - sum->transport = lock->transport; + sum->client = lock->client; sum->fd_num = lock->fd_num; sum->client_pid = lock->client_pid; sum->owner = lock->owner; @@ -794,6 +822,8 @@ __insert_and_merge (pl_inode_t *pl_inode, posix_lock_t *lock) __destroy_lock (conf); __destroy_lock (lock); + INIT_LIST_HEAD (&sum->list); + posix_lock_to_flock (sum, &sum->user_flock); __insert_and_merge (pl_inode, sum); return; @@ -801,7 +831,7 @@ __insert_and_merge (pl_inode_t *pl_inode, posix_lock_t *lock) sum = add_locks (lock, conf); sum->fl_type = conf->fl_type; - sum->transport = conf->transport; + sum->client = conf->client; sum->fd_num = conf->fd_num; sum->client_pid = conf->client_pid; sum->owner = conf->owner; @@ -821,6 +851,8 @@ __insert_and_merge (pl_inode_t *pl_inode, posix_lock_t *lock) continue; INIT_LIST_HEAD (&v.locks[i]->list); + posix_lock_to_flock (v.locks[i], + &v.locks[i]->user_flock); __insert_and_merge (pl_inode, v.locks[i]); } @@ -927,7 +959,8 @@ grant_blocked_locks (xlator_t *this, pl_inode_t *pl_inode) pl_trace_out (this, lock->frame, NULL, NULL, F_SETLKW, &lock->user_flock, 0, 0, NULL); - STACK_UNWIND (lock->frame, 0, 0, &lock->user_flock); + STACK_UNWIND_STRICT (lk, lock->frame, 0, 0, + &lock->user_flock, NULL); GF_FREE (lock); } @@ -956,7 +989,7 @@ pl_send_prelock_unlock (xlator_t *this, pl_inode_t *pl_inode, flock.l_len = old_lock->user_flock.l_len; - unlock_lock = new_posix_lock (&flock, old_lock->transport, + unlock_lock = new_posix_lock (&flock, old_lock->client, old_lock->client_pid, &old_lock->owner, old_lock->fd); GF_VALIDATE_OR_GOTO (this->name, unlock_lock, out); @@ -972,7 +1005,8 @@ pl_send_prelock_unlock (xlator_t *this, pl_inode_t *pl_inode, pl_trace_out (this, lock->frame, NULL, NULL, F_SETLKW, &lock->user_flock, 0, 0, NULL); - STACK_UNWIND (lock->frame, 0, 0, &lock->user_flock); + STACK_UNWIND_STRICT (lk, lock->frame, 0, 0, + &lock->user_flock, NULL); GF_FREE (lock); } @@ -1055,7 +1089,7 @@ pl_getlk (pl_inode_t *pl_inode, posix_lock_t *lock) { posix_lock_t *conf = NULL; - conf = first_overlap (pl_inode, lock); + conf = first_conflicting_overlap (pl_inode, lock); if (conf == NULL) { lock->fl_type = F_UNLCK; @@ -1064,3 +1098,124 @@ pl_getlk (pl_inode_t *pl_inode, posix_lock_t *lock) return conf; } + + +struct _lock_table * +pl_lock_table_new (void) +{ + struct _lock_table *new = NULL; + + new = GF_CALLOC (1, sizeof (struct _lock_table), gf_common_mt_lock_table); + if (new == NULL) { + goto out; + } + INIT_LIST_HEAD (&new->entrylk_lockers); + INIT_LIST_HEAD (&new->inodelk_lockers); + LOCK_INIT (&new->lock); +out: + return new; +} + + +int +pl_add_locker (struct _lock_table *table, const char *volume, + loc_t *loc, fd_t *fd, pid_t pid, gf_lkowner_t *owner, + glusterfs_fop_t type) +{ + int32_t ret = -1; + struct _locker *new = NULL; + + GF_VALIDATE_OR_GOTO ("lock-table", table, out); + GF_VALIDATE_OR_GOTO ("lock-table", volume, out); + + new = GF_CALLOC (1, sizeof (struct _locker), gf_common_mt_locker); + if (new == NULL) { + goto out; + } + INIT_LIST_HEAD (&new->lockers); + + new->volume = gf_strdup (volume); + + if (fd == NULL) { + loc_copy (&new->loc, loc); + } else { + new->fd = fd_ref (fd); + } + + new->pid = pid; + new->owner = *owner; + + LOCK (&table->lock); + { + if (type == GF_FOP_ENTRYLK) + list_add_tail (&new->lockers, &table->entrylk_lockers); + else + list_add_tail (&new->lockers, &table->inodelk_lockers); + } + UNLOCK (&table->lock); +out: + return ret; +} + +int +pl_del_locker (struct _lock_table *table, const char *volume, + loc_t *loc, fd_t *fd, gf_lkowner_t *owner, glusterfs_fop_t type) +{ + struct _locker *locker = NULL; + struct _locker *tmp = NULL; + int32_t ret = -1; + struct list_head *head = NULL; + struct list_head del; + + GF_VALIDATE_OR_GOTO ("lock-table", table, out); + GF_VALIDATE_OR_GOTO ("lock-table", volume, out); + + INIT_LIST_HEAD (&del); + + LOCK (&table->lock); + { + if (type == GF_FOP_ENTRYLK) { + head = &table->entrylk_lockers; + } else { + head = &table->inodelk_lockers; + } + + list_for_each_entry_safe (locker, tmp, head, lockers) { + if (!is_same_lkowner (&locker->owner, owner) || + strcmp (locker->volume, volume)) + continue; + + /* + * It is possible for inodelk lock to come on anon-fd + * and inodelk unlock to come on normal fd in case of + * client re-opens. So don't check for fds to be equal. + */ + if (locker->fd && fd) + list_move_tail (&locker->lockers, &del); + else if (locker->loc.inode && loc && + (locker->loc.inode == loc->inode)) + list_move_tail (&locker->lockers, &del); + } + } + UNLOCK (&table->lock); + + tmp = NULL; + locker = NULL; + + list_for_each_entry_safe (locker, tmp, &del, lockers) { + list_del_init (&locker->lockers); + if (locker->fd) + fd_unref (locker->fd); + else + loc_wipe (&locker->loc); + + GF_FREE (locker->volume); + GF_FREE (locker); + } + + ret = 0; +out: + return ret; + +} + diff --git a/xlators/features/locks/src/common.h b/xlators/features/locks/src/common.h index 0c0dd22e7..db19ec978 100644 --- a/xlators/features/locks/src/common.h +++ b/xlators/features/locks/src/common.h @@ -1,30 +1,54 @@ /* - Copyright (c) 2006-2011 Gluster, Inc. <http://www.gluster.com> - This file is part of GlusterFS. - - GlusterFS is free software; you can redistribute it and/or modify - it under the terms of the GNU General Public License as published - by the Free Software Foundation; either version 3 of the License, - or (at your option) any later version. - - GlusterFS is distributed in the hope that it will be useful, but - WITHOUT ANY WARRANTY; without even the implied warranty of - MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU - General Public License for more details. - - You should have received a copy of the GNU General Public License - along with this program. If not, see - <http://www.gnu.org/licenses/>. -*/ + 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 __COMMON_H__ #define __COMMON_H__ #include "lkowner.h" +/*dump locks format strings */ +#define RANGE_FMT "type=%s, whence=%hd, start=%llu, len=%llu" +#define ENTRY_FMT "type=%s on basename=%s" +#define DUMP_GEN_FMT "pid = %llu, owner=%s, client=%p" +#define GRNTD_AT "granted at %s" +#define BLKD_AT "blocked at %s" +#define CONN_ID "connection-id=%s" +#define DUMP_BLKD_FMT DUMP_GEN_FMT", "CONN_ID", "BLKD_AT +#define DUMP_GRNTD_FMT DUMP_GEN_FMT", "CONN_ID", "GRNTD_AT +#define DUMP_BLKD_GRNTD_FMT DUMP_GEN_FMT", "CONN_ID", "BLKD_AT", "GRNTD_AT + +#define ENTRY_BLKD_FMT ENTRY_FMT", "DUMP_BLKD_FMT +#define ENTRY_GRNTD_FMT ENTRY_FMT", "DUMP_GRNTD_FMT +#define ENTRY_BLKD_GRNTD_FMT ENTRY_FMT", "DUMP_BLKD_GRNTD_FMT + +#define RANGE_BLKD_FMT RANGE_FMT", "DUMP_BLKD_FMT +#define RANGE_GRNTD_FMT RANGE_FMT", "DUMP_GRNTD_FMT +#define RANGE_BLKD_GRNTD_FMT RANGE_FMT", "DUMP_BLKD_GRNTD_FMT #define SET_FLOCK_PID(flock, lock) ((flock)->l_pid = lock->client_pid) + +struct _locker { + struct list_head lockers; + char *volume; + loc_t loc; + fd_t *fd; + gf_lkowner_t owner; + pid_t pid; +}; + +struct _lock_table { + struct list_head inodelk_lockers; + struct list_head entrylk_lockers; + gf_lock_t lock; +}; + posix_lock_t * -new_posix_lock (struct gf_flock *flock, void *transport, pid_t client_pid, +new_posix_lock (struct gf_flock *flock, client_t *client, pid_t client_pid, gf_lkowner_t *owner, fd_t *fd); pl_inode_t * @@ -57,13 +81,14 @@ pl_dom_list_t * get_domain (pl_inode_t *pl_inode, const char *volume); void -grant_blocked_inode_locks (xlator_t *this, pl_inode_t *pl_inode, pl_dom_list_t *dom); +grant_blocked_inode_locks (xlator_t *this, pl_inode_t *pl_inode, + pl_dom_list_t *dom); void __delete_inode_lock (pl_inode_lock_t *lock); void -__destroy_inode_lock (pl_inode_lock_t *lock); +__pl_inodelk_unref (pl_inode_lock_t *lock); void grant_blocked_entry_locks (xlator_t *this, pl_inode_t *pl_inode, @@ -72,9 +97,13 @@ grant_blocked_entry_locks (xlator_t *this, pl_inode_t *pl_inode, void pl_update_refkeeper (xlator_t *this, inode_t *inode); int32_t -get_inodelk_count (xlator_t *this, inode_t *inode); +__get_inodelk_count (xlator_t *this, pl_inode_t *pl_inode, char *domname); +int32_t +get_inodelk_count (xlator_t *this, inode_t *inode, char *domname); int32_t +__get_entrylk_count (xlator_t *this, pl_inode_t *pl_inode); +int32_t get_entrylk_count (xlator_t *this, inode_t *inode); void pl_trace_in (xlator_t *this, call_frame_t *frame, fd_t *fd, loc_t *loc, @@ -133,4 +162,26 @@ pl_verify_reservelk (xlator_t *this, pl_inode_t *pl_inode, posix_lock_t *lock, int can_block); int pl_reserve_unlock (xlator_t *this, pl_inode_t *pl_inode, posix_lock_t *reqlock); + +uint32_t +check_entrylk_on_basename (xlator_t *this, inode_t *parent, char *basename); + +int32_t +pl_add_locker (struct _lock_table *table, const char *volume, + loc_t *loc, + fd_t *fd, + pid_t pid, + gf_lkowner_t *owner, + glusterfs_fop_t type); + +int32_t +pl_del_locker (struct _lock_table *table, const char *volume, + loc_t *loc, + fd_t *fd, + gf_lkowner_t *owner, + glusterfs_fop_t type); + +struct _lock_table * +pl_lock_table_new (void); + #endif /* __COMMON_H__ */ diff --git a/xlators/features/locks/src/entrylk.c b/xlators/features/locks/src/entrylk.c index 3e91b9d94..0785dc547 100644 --- a/xlators/features/locks/src/entrylk.c +++ b/xlators/features/locks/src/entrylk.c @@ -1,22 +1,12 @@ /* - Copyright (c) 2006-2011 Gluster, Inc. <http://www.gluster.com> - This file is part of GlusterFS. - - GlusterFS is free software; you can redistribute it and/or modify - it under the terms of the GNU General Public License as published - by the Free Software Foundation; either version 3 of the License, - or (at your option) any later version. - - GlusterFS is distributed in the hope that it will be useful, but - WITHOUT ANY WARRANTY; without even the implied warranty of - MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU - General Public License for more details. - - You should have received a copy of the GNU General Public License - along with this program. If not, see - <http://www.gnu.org/licenses/>. -*/ + 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 _CONFIG_H #define _CONFIG_H #include "config.h" @@ -35,7 +25,7 @@ static pl_entry_lock_t * new_entrylk_lock (pl_inode_t *pinode, const char *basename, entrylk_type type, - void *trans, pid_t client_pid, gf_lkowner_t *owner, + client_t *client, pid_t client_pid, gf_lkowner_t *owner, const char *volume) { @@ -49,7 +39,7 @@ new_entrylk_lock (pl_inode_t *pinode, const char *basename, entrylk_type type, newlock->basename = basename ? gf_strdup (basename) : NULL; newlock->type = type; - newlock->trans = trans; + newlock->trans = client; newlock->volume = volume; newlock->client_pid = client_pid; newlock->owner = *owner; @@ -315,18 +305,15 @@ __find_most_matching_lock (pl_dom_list_t *dom, const char *basename) int __lock_name (pl_inode_t *pinode, const char *basename, entrylk_type type, - call_frame_t *frame, pl_dom_list_t *dom, xlator_t *this, int nonblock) + call_frame_t *frame, pl_dom_list_t *dom, xlator_t *this, + int nonblock, char *conn_id) { pl_entry_lock_t *lock = NULL; pl_entry_lock_t *conf = NULL; - void *trans = NULL; - pid_t client_pid = 0; int ret = -EINVAL; - trans = frame->root->trans; - client_pid = frame->root->pid; - - lock = new_entrylk_lock (pinode, basename, type, trans, client_pid, + lock = new_entrylk_lock (pinode, basename, type, + frame->root->client, frame->root->pid, &frame->root->lk_owner, dom->domain); if (!lock) { ret = -ENOMEM; @@ -335,14 +322,18 @@ __lock_name (pl_inode_t *pinode, const char *basename, entrylk_type type, lock->frame = frame; lock->this = this; - lock->trans = trans; + lock->trans = frame->root->client; + + if (conn_id) { + lock->connection_id = gf_strdup (conn_id); + } conf = __lock_grantable (dom, basename, type); if (conf) { ret = -EAGAIN; if (nonblock){ - if (lock->basename) - GF_FREE ((char *)lock->basename); + GF_FREE (lock->connection_id); + GF_FREE ((char *)lock->basename); GF_FREE (lock); goto out; @@ -361,8 +352,8 @@ __lock_name (pl_inode_t *pinode, const char *basename, entrylk_type type, if ( __blocked_lock_conflict (dom, basename, type) && !(__owner_has_lock (dom, lock))) { ret = -EAGAIN; if (nonblock) { - if (lock->basename) - GF_FREE ((char *) lock->basename); + GF_FREE (lock->connection_id); + GF_FREE ((char *) lock->basename); GF_FREE (lock); goto out; @@ -441,6 +432,32 @@ out: return ret_lock; } +uint32_t +check_entrylk_on_basename (xlator_t *this, inode_t *parent, char *basename) +{ + uint32_t entrylk = 0; + pl_inode_t *pinode = 0; + pl_dom_list_t *dom = NULL; + pl_entry_lock_t *conf = NULL; + + pinode = pl_inode_get (this, parent); + if (!pinode) + goto out; + pthread_mutex_lock (&pinode->mutex); + { + list_for_each_entry (dom, &pinode->dom_list, inode_list) { + conf = __lock_grantable (dom, basename, ENTRYLK_WRLCK); + if (conf && conf->basename) { + entrylk = 1; + break; + } + } + } + pthread_mutex_unlock (&pinode->mutex); + +out: + return entrylk; +} void __grant_blocked_entry_locks (xlator_t *this, pl_inode_t *pl_inode, @@ -466,15 +483,16 @@ __grant_blocked_entry_locks (xlator_t *this, pl_inode_t *pl_inode, pl_inode, bl->basename); bl_ret = __lock_name (pl_inode, bl->basename, bl->type, - bl->frame, dom, bl->this, 0); + bl->frame, dom, bl->this, 0, + bl->connection_id); if (bl_ret == 0) { list_add (&bl->blocked_locks, granted); } else { gf_log (this->name, GF_LOG_DEBUG, "should never happen"); - if (bl->basename) - GF_FREE ((char *)bl->basename); + GF_FREE (bl->connection_id); + GF_FREE ((char *)bl->basename); GF_FREE (bl); } } @@ -494,7 +512,8 @@ grant_blocked_entry_locks (xlator_t *this, pl_inode_t *pl_inode, pthread_mutex_lock (&pl_inode->mutex); { - __grant_blocked_entry_locks (this, pl_inode, dom, &granted_list); + __grant_blocked_entry_locks (this, pl_inode, dom, + &granted_list); } pthread_mutex_unlock (&pl_inode->mutex); @@ -505,24 +524,28 @@ grant_blocked_entry_locks (xlator_t *this, pl_inode_t *pl_inode, lock->basename, ENTRYLK_LOCK, lock->type, 0, 0); - STACK_UNWIND_STRICT (entrylk, lock->frame, 0, 0); + STACK_UNWIND_STRICT (entrylk, lock->frame, 0, 0, NULL); + GF_FREE (lock->connection_id); + GF_FREE ((char *)lock->basename); + GF_FREE (lock); } GF_FREE ((char *)unlocked->basename); + GF_FREE (unlocked->connection_id); GF_FREE (unlocked); return; } /** - * release_entry_locks_for_transport: release all entry locks from this - * transport for this loc_t + * release_entry_locks_for_client: release all entry locks from this + * client for this loc_t */ static int -release_entry_locks_for_transport (xlator_t *this, pl_inode_t *pinode, - pl_dom_list_t *dom, void *trans) +release_entry_locks_for_client (xlator_t *this, pl_inode_t *pinode, + pl_dom_list_t *dom, client_t *client) { pl_entry_lock_t *lock = NULL; pl_entry_lock_t *tmp = NULL; @@ -536,14 +559,14 @@ release_entry_locks_for_transport (xlator_t *this, pl_inode_t *pinode, { list_for_each_entry_safe (lock, tmp, &dom->blocked_entrylks, blocked_locks) { - if (lock->trans != trans) + if (lock->trans != client) continue; list_del_init (&lock->blocked_locks); gf_log (this->name, GF_LOG_TRACE, "releasing lock on held by " - "{transport=%p}",trans); + "{client=%p}", client); list_add (&lock->blocked_locks, &released); @@ -551,16 +574,17 @@ release_entry_locks_for_transport (xlator_t *this, pl_inode_t *pinode, list_for_each_entry_safe (lock, tmp, &dom->entrylk_list, domain_list) { - if (lock->trans != trans) + if (lock->trans != client) continue; list_del_init (&lock->domain_list); gf_log (this->name, GF_LOG_TRACE, "releasing lock on held by " - "{transport=%p}",trans); + "{client=%p}", client); GF_FREE ((char *)lock->basename); + GF_FREE (lock->connection_id); GF_FREE (lock); } @@ -573,10 +597,10 @@ release_entry_locks_for_transport (xlator_t *this, pl_inode_t *pinode, list_for_each_entry_safe (lock, tmp, &released, blocked_locks) { list_del_init (&lock->blocked_locks); - STACK_UNWIND_STRICT (entrylk, lock->frame, -1, EAGAIN); + STACK_UNWIND_STRICT (entrylk, lock->frame, -1, EAGAIN, NULL); - if (lock->basename) - GF_FREE ((char *)lock->basename); + GF_FREE ((char *)lock->basename); + GF_FREE (lock->connection_id); GF_FREE (lock); } @@ -584,10 +608,10 @@ release_entry_locks_for_transport (xlator_t *this, pl_inode_t *pinode, list_for_each_entry_safe (lock, tmp, &granted, blocked_locks) { list_del_init (&lock->blocked_locks); - STACK_UNWIND_STRICT (entrylk, lock->frame, 0, 0); + STACK_UNWIND_STRICT (entrylk, lock->frame, 0, 0, NULL); - if (lock->basename) - GF_FREE ((char *)lock->basename); + GF_FREE ((char *)lock->basename); + GF_FREE (lock->connection_id); GF_FREE (lock); } @@ -598,19 +622,23 @@ release_entry_locks_for_transport (xlator_t *this, pl_inode_t *pinode, int pl_common_entrylk (call_frame_t *frame, xlator_t *this, const char *volume, inode_t *inode, const char *basename, - entrylk_cmd cmd, entrylk_type type, loc_t *loc, fd_t *fd) -{ - int32_t op_ret = -1; - int32_t op_errno = 0; + entrylk_cmd cmd, entrylk_type type, loc_t *loc, fd_t *fd, + dict_t *xdata) - void * transport = NULL; - - pl_inode_t * pinode = NULL; +{ + int32_t op_ret = -1; + int32_t op_errno = 0; int ret = -1; - pl_entry_lock_t *unlocked = NULL; char unwind = 1; + GF_UNUSED int dict_ret = -1; + pl_inode_t *pinode = NULL; + pl_entry_lock_t *unlocked = NULL; + pl_dom_list_t *dom = NULL; + char *conn_id = NULL; + pl_ctx_t *ctx = NULL; - pl_dom_list_t *dom = NULL; + if (xdata) + dict_ret = dict_get_str (xdata, "connection-id", &conn_id); pinode = pl_inode_get (this, inode); if (!pinode) { @@ -626,18 +654,17 @@ pl_common_entrylk (call_frame_t *frame, xlator_t *this, entrylk_trace_in (this, frame, volume, fd, loc, basename, cmd, type); - transport = frame->root->trans; - if (frame->root->lk_owner.len == 0) { /* this is a special case that means release - all locks from this transport + all locks from this client */ gf_log (this->name, GF_LOG_TRACE, - "Releasing locks for transport %p", transport); + "Releasing locks for client %p", frame->root->client); - release_entry_locks_for_transport (this, pinode, dom, transport); + release_entry_locks_for_client (this, pinode, dom, + frame->root->client); op_ret = 0; goto out; @@ -648,7 +675,7 @@ pl_common_entrylk (call_frame_t *frame, xlator_t *this, pthread_mutex_lock (&pinode->mutex); { ret = __lock_name (pinode, basename, type, - frame, dom, this, 0); + frame, dom, this, 0, conn_id); } pthread_mutex_unlock (&pinode->mutex); @@ -673,7 +700,7 @@ pl_common_entrylk (call_frame_t *frame, xlator_t *this, pthread_mutex_lock (&pinode->mutex); { ret = __lock_name (pinode, basename, type, - frame, dom, this, 1); + frame, dom, this, 1, conn_id); } pthread_mutex_unlock (&pinode->mutex); @@ -710,7 +737,25 @@ out: entrylk_trace_out (this, frame, volume, fd, loc, basename, cmd, type, op_ret, op_errno); - STACK_UNWIND_STRICT (entrylk, frame, op_ret, op_errno); + ctx = pl_ctx_get (frame->root->client, this); + + if (ctx == NULL) { + gf_log (this->name, GF_LOG_INFO, "pl_ctx_get() failed"); + goto unwind; + } + + if (cmd == ENTRYLK_UNLOCK) + pl_del_locker (ctx->ltable, volume, loc, fd, + &frame->root->lk_owner, + GF_FOP_ENTRYLK); + else + pl_add_locker (ctx->ltable, volume, loc, fd, + frame->root->pid, + &frame->root->lk_owner, + GF_FOP_ENTRYLK); + +unwind: + STACK_UNWIND_STRICT (entrylk, frame, op_ret, op_errno, NULL); } else { entrylk_trace_block (this, frame, volume, fd, loc, basename, cmd, type); @@ -729,10 +774,10 @@ out: int pl_entrylk (call_frame_t *frame, xlator_t *this, const char *volume, loc_t *loc, const char *basename, - entrylk_cmd cmd, entrylk_type type) + entrylk_cmd cmd, entrylk_type type, dict_t *xdata) { - - pl_common_entrylk (frame, this, volume, loc->inode, basename, cmd, type, loc, NULL); + pl_common_entrylk (frame, this, volume, loc->inode, basename, cmd, + type, loc, NULL, xdata); return 0; } @@ -747,16 +792,16 @@ pl_entrylk (call_frame_t *frame, xlator_t *this, int pl_fentrylk (call_frame_t *frame, xlator_t *this, const char *volume, fd_t *fd, const char *basename, - entrylk_cmd cmd, entrylk_type type) + entrylk_cmd cmd, entrylk_type type, dict_t *xdata) { - - pl_common_entrylk (frame, this, volume, fd->inode, basename, cmd, type, NULL, fd); + pl_common_entrylk (frame, this, volume, fd->inode, basename, cmd, + type, NULL, fd, xdata); return 0; } -static int32_t +int32_t __get_entrylk_count (xlator_t *this, pl_inode_t *pl_inode) { int32_t count = 0; @@ -765,24 +810,10 @@ __get_entrylk_count (xlator_t *this, pl_inode_t *pl_inode) list_for_each_entry (dom, &pl_inode->dom_list, inode_list) { list_for_each_entry (lock, &dom->entrylk_list, domain_list) { - - gf_log (this->name, GF_LOG_DEBUG, - " XATTR DEBUG" - " domain: %s %s on %s state = Active", - dom->domain, - lock->type == ENTRYLK_RDLCK ? "ENTRYLK_RDLCK" : - "ENTRYLK_WRLCK", lock->basename); count++; } list_for_each_entry (lock, &dom->blocked_entrylks, blocked_locks) { - - gf_log (this->name, GF_LOG_DEBUG, - " XATTR DEBUG" - " domain: %s %s on %s state = Blocked", - dom->domain, - lock->type == ENTRYLK_RDLCK ? "ENTRYLK_RDLCK" : - "ENTRYLK_WRLCK", lock->basename); count++; } diff --git a/xlators/features/locks/src/inodelk.c b/xlators/features/locks/src/inodelk.c index 5785f1f8e..508523e11 100644 --- a/xlators/features/locks/src/inodelk.c +++ b/xlators/features/locks/src/inodelk.c @@ -1,22 +1,12 @@ /* - Copyright (c) 2006-2011 Gluster, Inc. <http://www.gluster.com> - This file is part of GlusterFS. - - GlusterFS is free software; you can redistribute it and/or modify - it under the terms of the GNU General Public License as published - by the Free Software Foundation; either version 3 of the License, - or (at your option) any later version. - - GlusterFS is distributed in the hope that it will be useful, but - WITHOUT ANY WARRANTY; without even the implied warranty of - MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU - General Public License for more details. - - You should have received a copy of the GNU General Public License - along with this program. If not, see - <http://www.gnu.org/licenses/>. -*/ + 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 _CONFIG_H #define _CONFIG_H #include "config.h" @@ -33,20 +23,30 @@ #include "locks.h" #include "common.h" -void +inline void __delete_inode_lock (pl_inode_lock_t *lock) { list_del (&lock->list); } -void -__destroy_inode_lock (pl_inode_lock_t *lock) +static inline void +__pl_inodelk_ref (pl_inode_lock_t *lock) +{ + lock->ref++; +} + +inline void +__pl_inodelk_unref (pl_inode_lock_t *lock) { - GF_FREE (lock); + lock->ref--; + if (!lock->ref) { + GF_FREE (lock->connection_id); + GF_FREE (lock); + } } /* Check if 2 inodelks are conflicting on type. Only 2 shared locks don't conflict */ -static int +static inline int inodelk_type_conflict (pl_inode_lock_t *l1, pl_inode_lock_t *l2) { if (l2->fl_type == F_WRLCK || l1->fl_type == F_WRLCK) @@ -124,7 +124,7 @@ static inline int same_inodelk_owner (pl_inode_lock_t *l1, pl_inode_lock_t *l2) { return (is_same_lkowner (&l1->owner, &l2->owner) && - (l1->transport == l2->transport)); + (l1->client == l2->client)); } /* Returns true if the 2 inodelks conflict with each other */ @@ -245,6 +245,7 @@ __lock_inodelk (xlator_t *this, pl_inode_t *pl_inode, pl_inode_lock_t *lock, goto out; } + __pl_inodelk_ref (lock); gettimeofday (&lock->granted_time, NULL); list_add (&lock->list, &dom->inodelk_list); @@ -289,20 +290,22 @@ __inode_unlock_lock (xlator_t *this, pl_inode_lock_t *lock, pl_dom_list_t *dom) conf = find_matching_inodelk (lock, dom); if (!conf) { - gf_log (this->name, GF_LOG_DEBUG, - " Matching lock not found for unlock"); + gf_log (this->name, GF_LOG_ERROR, + " Matching lock not found for unlock %llu-%llu, by %s " + "on %p", (unsigned long long)lock->fl_start, + (unsigned long long)lock->fl_end, + lkowner_utoa (&lock->owner), lock->client); goto out; } __delete_inode_lock (conf); gf_log (this->name, GF_LOG_DEBUG, - " Matching lock found for unlock"); - __destroy_inode_lock (lock); - + " Matching lock found for unlock %llu-%llu, by %s on %p", + (unsigned long long)lock->fl_start, + (unsigned long long)lock->fl_end, lkowner_utoa (&lock->owner), + lock->client); out: return conf; - - } static void __grant_blocked_inode_locks (xlator_t *this, pl_inode_t *pl_inode, @@ -332,7 +335,8 @@ __grant_blocked_inode_locks (xlator_t *this, pl_inode_t *pl_inode, /* Grant all inodelks blocked on a lock */ void -grant_blocked_inode_locks (xlator_t *this, pl_inode_t *pl_inode, pl_dom_list_t *dom) +grant_blocked_inode_locks (xlator_t *this, pl_inode_t *pl_inode, + pl_dom_list_t *dom) { struct list_head granted; pl_inode_lock_t *lock; @@ -340,11 +344,6 @@ grant_blocked_inode_locks (xlator_t *this, pl_inode_t *pl_inode, pl_dom_list_t * INIT_LIST_HEAD (&granted); - if (list_empty (&dom->blocked_inodelks)) { - gf_log (this->name, GF_LOG_TRACE, - "No blocked locks to be granted for domain: %s", dom->domain); - } - pthread_mutex_lock (&pl_inode->mutex); { __grant_blocked_inode_locks (this, pl_inode, &granted, dom); @@ -363,28 +362,34 @@ grant_blocked_inode_locks (xlator_t *this, pl_inode_t *pl_inode, pl_dom_list_t * pl_trace_out (this, lock->frame, NULL, NULL, F_SETLKW, &lock->user_flock, 0, 0, lock->volume); - STACK_UNWIND_STRICT (inodelk, lock->frame, 0, 0); + STACK_UNWIND_STRICT (inodelk, lock->frame, 0, 0, NULL); } + pthread_mutex_lock (&pl_inode->mutex); + { + list_for_each_entry_safe (lock, tmp, &granted, blocked_locks) { + list_del_init (&lock->blocked_locks); + __pl_inodelk_unref (lock); + } + } + pthread_mutex_unlock (&pl_inode->mutex); } -/* Release all inodelks from this transport */ +/* Release all inodelks from this client */ static int -release_inode_locks_of_transport (xlator_t *this, pl_dom_list_t *dom, - inode_t *inode, void *trans) +release_inode_locks_of_client (xlator_t *this, pl_dom_list_t *dom, + inode_t *inode, client_t *client) { pl_inode_lock_t *tmp = NULL; pl_inode_lock_t *l = NULL; pl_inode_t * pinode = NULL; - struct list_head granted; struct list_head released; char *path = NULL; char *file = NULL; - INIT_LIST_HEAD (&granted); INIT_LIST_HEAD (&released); pinode = pl_inode_get (this, inode); @@ -393,7 +398,7 @@ release_inode_locks_of_transport (xlator_t *this, pl_dom_list_t *dom, { list_for_each_entry_safe (l, tmp, &dom->blocked_inodelks, blocked_locks) { - if (l->transport != trans) + if (l->client != client) continue; list_del_init (&l->blocked_locks); @@ -406,8 +411,8 @@ release_inode_locks_of_transport (xlator_t *this, pl_dom_list_t *dom, gf_log (this->name, GF_LOG_DEBUG, "releasing blocking lock on %s held by " - "{transport=%p, pid=%"PRId64" lk-owner=%s}", - file, trans, (uint64_t) l->client_pid, + "{client=%p, pid=%"PRId64" lk-owner=%s}", + file, client, (uint64_t) l->client_pid, lkowner_utoa (&l->owner)); list_add (&l->blocked_locks, &released); @@ -418,12 +423,9 @@ release_inode_locks_of_transport (xlator_t *this, pl_dom_list_t *dom, } list_for_each_entry_safe (l, tmp, &dom->inodelk_list, list) { - if (l->transport != trans) + if (l->client != client) continue; - __delete_inode_lock (l); - __destroy_inode_lock (l); - inode_path (inode, NULL, &path); if (path) file = path; @@ -432,25 +434,29 @@ release_inode_locks_of_transport (xlator_t *this, pl_dom_list_t *dom, gf_log (this->name, GF_LOG_DEBUG, "releasing granted lock on %s held by " - "{transport=%p, pid=%"PRId64" lk-owner=%s}", - file, trans, (uint64_t) l->client_pid, + "{client=%p, pid=%"PRId64" lk-owner=%s}", + file, client, (uint64_t) l->client_pid, lkowner_utoa (&l->owner)); + if (path) { GF_FREE (path); path = NULL; } + + __delete_inode_lock (l); + __pl_inodelk_unref (l); } } - if (path) - GF_FREE (path); + GF_FREE (path); pthread_mutex_unlock (&pinode->mutex); list_for_each_entry_safe (l, tmp, &released, blocked_locks) { list_del_init (&l->blocked_locks); - STACK_UNWIND_STRICT (inodelk, l->frame, -1, EAGAIN); - GF_FREE (l); + STACK_UNWIND_STRICT (inodelk, l->frame, -1, EAGAIN, NULL); + //No need to take lock as the locks are only in one list + __pl_inodelk_unref (l); } grant_blocked_inode_locks (this, pinode, dom); @@ -464,12 +470,13 @@ pl_inode_setlk (xlator_t *this, pl_inode_t *pl_inode, pl_inode_lock_t *lock, { int ret = -EINVAL; pl_inode_lock_t *retlock = NULL; + gf_boolean_t unref = _gf_true; pthread_mutex_lock (&pl_inode->mutex); { if (lock->fl_type != F_UNLCK) { ret = __lock_inodelk (this, pl_inode, lock, can_block, dom); - if (ret == 0) + if (ret == 0) { gf_log (this->name, GF_LOG_TRACE, "%s (pid=%d) (lk-owner=%s) %"PRId64" - %"PRId64" => OK", lock->fl_type == F_UNLCK ? "Unlock" : "Lock", @@ -477,8 +484,7 @@ pl_inode_setlk (xlator_t *this, pl_inode_t *pl_inode, pl_inode_lock_t *lock, lkowner_utoa (&lock->owner), lock->fl_start, lock->fl_end); - - if (ret == -EAGAIN) + } else if (ret == -EAGAIN) { gf_log (this->name, GF_LOG_TRACE, "%s (pid=%d) (lk-owner=%s) %"PRId64" - %"PRId64" => NOK", lock->fl_type == F_UNLCK ? "Unlock" : "Lock", @@ -486,25 +492,25 @@ pl_inode_setlk (xlator_t *this, pl_inode_t *pl_inode, pl_inode_lock_t *lock, lkowner_utoa (&lock->owner), lock->user_flock.l_start, lock->user_flock.l_len); + if (can_block) + unref = _gf_false; + } + } else { + retlock = __inode_unlock_lock (this, lock, dom); + if (!retlock) { + gf_log (this->name, GF_LOG_DEBUG, + "Bad Unlock issued on Inode lock"); + ret = -EINVAL; + goto out; + } + __pl_inodelk_unref (retlock); - goto out; - } - - - retlock = __inode_unlock_lock (this, lock, dom); - if (!retlock) { - gf_log (this->name, GF_LOG_DEBUG, - "Bad Unlock issued on Inode lock"); - ret = -EINVAL; - goto out; + ret = 0; } - __destroy_inode_lock (retlock); - - ret = 0; - - } out: + if (unref) + __pl_inodelk_unref (lock); pthread_mutex_unlock (&pl_inode->mutex); grant_blocked_inode_locks (this, pl_inode, dom); return ret; @@ -512,8 +518,9 @@ out: /* Create a new inode_lock_t */ pl_inode_lock_t * -new_inode_lock (struct gf_flock *flock, void *transport, pid_t client_pid, - gf_lkowner_t *owner, const char *volume) +new_inode_lock (struct gf_flock *flock, client_t *client, pid_t client_pid, + call_frame_t *frame, xlator_t *this, const char *volume, + char *conn_id) { pl_inode_lock_t *lock = NULL; @@ -532,32 +539,76 @@ new_inode_lock (struct gf_flock *flock, void *transport, pid_t client_pid, else lock->fl_end = flock->l_start + flock->l_len - 1; - lock->transport = transport; + lock->client = client; lock->client_pid = client_pid; lock->volume = volume; - lock->owner = *owner; + lock->owner = frame->root->lk_owner; + lock->frame = frame; + lock->this = this; + + if (conn_id) { + lock->connection_id = gf_strdup (conn_id); + } INIT_LIST_HEAD (&lock->list); INIT_LIST_HEAD (&lock->blocked_locks); + __pl_inodelk_ref (lock); return lock; } +int32_t +_pl_convert_volume (const char *volume, char **res) +{ + char *mdata_vol = NULL; + int ret = 0; + + mdata_vol = strrchr (volume, ':'); + //if the volume already ends with :metadata don't bother + if (mdata_vol && (strcmp (mdata_vol, ":metadata") == 0)) + return 0; + + ret = gf_asprintf (res, "%s:metadata", volume); + if (ret <= 0) + return ENOMEM; + return 0; +} + +int32_t +_pl_convert_volume_for_special_range (struct gf_flock *flock, + const char *volume, char **res) +{ + int32_t ret = 0; + + if ((flock->l_start == LLONG_MAX -1) && + (flock->l_len == 0)) { + ret = _pl_convert_volume (volume, res); + } + + return ret; +} + /* Common inodelk code called from pl_inodelk and pl_finodelk */ int pl_common_inodelk (call_frame_t *frame, xlator_t *this, const char *volume, inode_t *inode, int32_t cmd, - struct gf_flock *flock, loc_t *loc, fd_t *fd) + struct gf_flock *flock, loc_t *loc, fd_t *fd, dict_t *xdata) { int32_t op_ret = -1; int32_t op_errno = 0; int ret = -1; + GF_UNUSED int dict_ret = -1; int can_block = 0; - pid_t client_pid = -1; - void * transport = NULL; pl_inode_t * pinode = NULL; pl_inode_lock_t * reqlock = NULL; pl_dom_list_t * dom = NULL; + char *res = NULL; + char *res1 = NULL; + char *conn_id = NULL; + pl_ctx_t *ctx = NULL; + + if (xdata) + dict_ret = dict_get_str (xdata, "connection-id", &conn_id); VALIDATE_OR_GOTO (frame, out); VALIDATE_OR_GOTO (inode, unwind); @@ -568,10 +619,13 @@ pl_common_inodelk (call_frame_t *frame, xlator_t *this, goto unwind; } - pl_trace_in (this, frame, fd, loc, cmd, flock, volume); + op_errno = _pl_convert_volume_for_special_range (flock, volume, &res); + if (op_errno) + goto unwind; + if (res) + volume = res; - transport = frame->root->trans; - client_pid = frame->root->pid; + pl_trace_in (this, frame, fd, loc, cmd, flock, volume); pinode = pl_inode_get (this, inode); if (!pinode) { @@ -580,23 +634,34 @@ pl_common_inodelk (call_frame_t *frame, xlator_t *this, } dom = get_domain (pinode, volume); + if (!dom) { + op_errno = ENOMEM; + goto unwind; + } if (frame->root->lk_owner.len == 0) { /* special case: this means release all locks - from this transport + from this client */ gf_log (this->name, GF_LOG_TRACE, - "Releasing all locks from transport %p", transport); - - release_inode_locks_of_transport (this, dom, inode, transport); + "Releasing all locks from client %p", frame->root->client); + + release_inode_locks_of_client (this, dom, inode, frame->root->client); + _pl_convert_volume (volume, &res1); + if (res1) { + dom = get_domain (pinode, res1); + if (dom) + release_inode_locks_of_client (this, dom, + inode, frame->root->client); + } op_ret = 0; goto unwind; } - reqlock = new_inode_lock (flock, transport, client_pid, - &frame->root->lk_owner, volume); + reqlock = new_inode_lock (flock, frame->root->client, frame->root->pid, + frame, this, volume, conn_id); if (!reqlock) { op_ret = -1; @@ -604,14 +669,10 @@ pl_common_inodelk (call_frame_t *frame, xlator_t *this, goto unwind; } - reqlock->frame = frame; - reqlock->this = this; switch (cmd) { case F_SETLKW: can_block = 1; - reqlock->frame = frame; - reqlock->this = this; /* fall through */ @@ -628,7 +689,6 @@ pl_common_inodelk (call_frame_t *frame, xlator_t *this, } gf_log (this->name, GF_LOG_TRACE, "returning EAGAIN"); op_errno = -ret; - __destroy_inode_lock (reqlock); goto unwind; } break; @@ -644,86 +704,103 @@ pl_common_inodelk (call_frame_t *frame, xlator_t *this, op_ret = 0; + ctx = pl_ctx_get (frame->root->client, this); + + if (ctx == NULL) { + gf_log (this->name, GF_LOG_INFO, "pl_ctx_get() failed"); + goto unwind; + } + + if (flock->l_type == F_UNLCK) + pl_del_locker (ctx->ltable, volume, loc, fd, + &frame->root->lk_owner, + GF_FOP_INODELK); + else + pl_add_locker (ctx->ltable, volume, loc, fd, + frame->root->pid, + &frame->root->lk_owner, + GF_FOP_INODELK); + unwind: if ((inode != NULL) && (flock !=NULL)) { pl_update_refkeeper (this, inode); pl_trace_out (this, frame, fd, loc, cmd, flock, op_ret, op_errno, volume); } - STACK_UNWIND_STRICT (inodelk, frame, op_ret, op_errno); + STACK_UNWIND_STRICT (inodelk, frame, op_ret, op_errno, NULL); out: + GF_FREE (res); + GF_FREE (res1); return 0; } int pl_inodelk (call_frame_t *frame, xlator_t *this, - const char *volume, loc_t *loc, int32_t cmd, struct gf_flock *flock) + const char *volume, loc_t *loc, int32_t cmd, struct gf_flock *flock, + dict_t *xdata) { - - pl_common_inodelk (frame, this, volume, loc->inode, cmd, flock, loc, NULL); + pl_common_inodelk (frame, this, volume, loc->inode, cmd, flock, + loc, NULL, xdata); return 0; } int pl_finodelk (call_frame_t *frame, xlator_t *this, - const char *volume, fd_t *fd, int32_t cmd, struct gf_flock *flock) + const char *volume, fd_t *fd, int32_t cmd, struct gf_flock *flock, + dict_t *xdata) { - - pl_common_inodelk (frame, this, volume, fd->inode, cmd, flock, NULL, fd); + pl_common_inodelk (frame, this, volume, fd->inode, cmd, flock, + NULL, fd, xdata); return 0; } +static inline int32_t +__get_inodelk_dom_count (pl_dom_list_t *dom) +{ + pl_inode_lock_t *lock = NULL; + int32_t count = 0; -static int32_t -__get_inodelk_count (xlator_t *this, pl_inode_t *pl_inode) + list_for_each_entry (lock, &dom->inodelk_list, list) { + count++; + } + list_for_each_entry (lock, &dom->blocked_inodelks, blocked_locks) { + count++; + } + return count; +} + +/* Returns the no. of locks (blocked/granted) held on a given domain name + * If @domname is NULL, returns the no. of locks in all the domains present. + * If @domname is non-NULL and non-existent, returns 0 */ +int32_t +__get_inodelk_count (xlator_t *this, pl_inode_t *pl_inode, char *domname) { int32_t count = 0; - pl_inode_lock_t *lock = NULL; pl_dom_list_t *dom = NULL; list_for_each_entry (dom, &pl_inode->dom_list, inode_list) { - list_for_each_entry (lock, &dom->inodelk_list, list) { - - gf_log (this->name, GF_LOG_DEBUG, - " XATTR DEBUG" - " domain: %s %s (pid=%d) (lk-owner=%s) %"PRId64" - %"PRId64" " - "state = Active", - dom->domain, - lock->fl_type == F_UNLCK ? "Unlock" : "Lock", - lock->client_pid, - lkowner_utoa (&lock->owner), - lock->user_flock.l_start, - lock->user_flock.l_len); - - count++; - } + if (domname) { + if (strcmp (domname, dom->domain) == 0) { + count = __get_inodelk_dom_count (dom); + goto out; + } - list_for_each_entry (lock, &dom->blocked_inodelks, blocked_locks) { + } else { + /* Counting locks from all domains */ + count += __get_inodelk_dom_count (dom); - gf_log (this->name, GF_LOG_DEBUG, - " XATTR DEBUG" - " domain: %s %s (pid=%d) (lk-owner=%s) %"PRId64" - %"PRId64" " - "state = Blocked", - dom->domain, - lock->fl_type == F_UNLCK ? "Unlock" : "Lock", - lock->client_pid, - lkowner_utoa (&lock->owner), - lock->user_flock.l_start, - lock->user_flock.l_len); - - count++; } - } +out: return count; } int32_t -get_inodelk_count (xlator_t *this, inode_t *inode) +get_inodelk_count (xlator_t *this, inode_t *inode, char *domname) { pl_inode_t *pl_inode = NULL; uint64_t tmp_pl_inode = 0; @@ -739,7 +816,7 @@ get_inodelk_count (xlator_t *this, inode_t *inode) pthread_mutex_lock (&pl_inode->mutex); { - count = __get_inodelk_count (this, pl_inode); + count = __get_inodelk_count (this, pl_inode, domname); } pthread_mutex_unlock (&pl_inode->mutex); diff --git a/xlators/features/locks/src/locks-mem-types.h b/xlators/features/locks/src/locks-mem-types.h index 9d44f0eba..08aeb0a79 100644 --- a/xlators/features/locks/src/locks-mem-types.h +++ b/xlators/features/locks/src/locks-mem-types.h @@ -1,23 +1,13 @@ /* - Copyright (c) 2008-2011 Gluster, Inc. <http://www.gluster.com> + Copyright (c) 2008-2012 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 __LOCKS_MEM_TYPES_H__ #define __LOCKS_MEM_TYPES_H__ @@ -32,7 +22,6 @@ enum gf_locks_mem_types_ { gf_locks_mt_truncate_ops, gf_locks_mt_pl_rw_req_t, gf_locks_mt_posix_locks_private_t, - gf_locks_mt_pl_local_t, gf_locks_mt_pl_fdctx_t, gf_locks_mt_end }; diff --git a/xlators/features/locks/src/locks.h b/xlators/features/locks/src/locks.h index 0dcbdf979..76fc941d7 100644 --- a/xlators/features/locks/src/locks.h +++ b/xlators/features/locks/src/locks.h @@ -1,22 +1,12 @@ /* - Copyright (c) 2006-2011 Gluster, Inc. <http://www.gluster.com> - This file is part of GlusterFS. - - GlusterFS is free software; you can redistribute it and/or modify - it under the terms of the GNU General Public License as published - by the Free Software Foundation; either version 3 of the License, - or (at your option) any later version. - - GlusterFS is distributed in the hope that it will be useful, but - WITHOUT ANY WARRANTY; without even the implied warranty of - MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU - General Public License for more details. - - You should have received a copy of the GNU General Public License - along with this program. If not, see - <http://www.gnu.org/licenses/>. -*/ + 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 __POSIX_LOCKS_H__ #define __POSIX_LOCKS_H__ @@ -29,10 +19,10 @@ #include "stack.h" #include "call-stub.h" #include "locks-mem-types.h" +#include "client_t.h" #include "lkowner.h" -#define POSIX_LOCKS "posix-locks" struct __pl_fd; struct __posix_lock { @@ -43,7 +33,7 @@ struct __posix_lock { off_t fl_end; short blocked; /* waiting to acquire */ - struct gf_flock user_flock; /* the flock supplied by the user */ + struct gf_flock user_flock; /* the flock supplied by the user */ xlator_t *this; /* required for blocked locks */ unsigned long fd_num; @@ -56,7 +46,7 @@ struct __posix_lock { /* These two together serve to uniquely identify each process across nodes */ - void *transport; /* to identify client node */ + void *client; /* to identify client node */ gf_lkowner_t owner; pid_t client_pid; /* pid of client process */ }; @@ -65,6 +55,7 @@ typedef struct __posix_lock posix_lock_t; struct __pl_inode_lock { struct list_head list; struct list_head blocked_locks; /* list_head pointing to blocked_inodelks */ + int ref; short fl_type; off_t fl_start; @@ -72,7 +63,7 @@ struct __pl_inode_lock { const char *volume; - struct gf_flock user_flock; /* the flock supplied by the user */ + struct gf_flock user_flock; /* the flock supplied by the user */ xlator_t *this; /* required for blocked locks */ fd_t *fd; @@ -84,9 +75,11 @@ struct __pl_inode_lock { /* These two together serve to uniquely identify each process across nodes */ - void *transport; /* to identify client node */ + void *client; /* to identify client node */ gf_lkowner_t owner; pid_t client_pid; /* pid of client process */ + + char *connection_id; /* stores the client connection id */ }; typedef struct __pl_inode_lock pl_inode_lock_t; @@ -124,7 +117,9 @@ struct __entry_lock { void *trans; gf_lkowner_t owner; - pid_t client_pid; /* pid of client process */ + pid_t client_pid; /* pid of client process */ + + char *connection_id; /* stores the client connection id */ }; typedef struct __entry_lock pl_entry_lock_t; @@ -158,16 +153,40 @@ typedef struct __pl_fd pl_fd_t; typedef struct { gf_boolean_t mandatory; /* if mandatory locking is enabled */ gf_boolean_t trace; /* trace lock requests in and out */ + char *brickname; } posix_locks_private_t; + typedef struct { gf_boolean_t entrylk_count_req; gf_boolean_t inodelk_count_req; + gf_boolean_t inodelk_dom_count_req; gf_boolean_t posixlk_count_req; + gf_boolean_t parent_entrylk_req; + + /* used by {f,}truncate */ + loc_t loc; + fd_t *fd; + off_t offset; + dict_t *xdata; + enum {TRUNCATE, FTRUNCATE} op; } pl_local_t; + typedef struct { struct list_head locks_list; } pl_fdctx_t; + +typedef struct _locks_ctx { + gf_lock_t ltable_lock; /* only for replace, + ltable has its own internal + lock for operations */ + struct _lock_table *ltable; +} pl_ctx_t; + + +pl_ctx_t * +pl_ctx_get (client_t *client, xlator_t *xlator); + #endif /* __POSIX_LOCKS_H__ */ diff --git a/xlators/features/locks/src/posix.c b/xlators/features/locks/src/posix.c index bd30ca493..7bfb38a51 100644 --- a/xlators/features/locks/src/posix.c +++ b/xlators/features/locks/src/posix.c @@ -1,22 +1,12 @@ /* - Copyright (c) 2006-2011 Gluster, Inc. <http://www.gluster.com> - This file is part of GlusterFS. - - GlusterFS is free software; you can redistribute it and/or modify - it under the terms of the GNU General Public License as published - by the Free Software Foundation; either version 3 of the License, - or (at your option) any later version. - - GlusterFS is distributed in the hope that it will be useful, but - WITHOUT ANY WARRANTY; without even the implied warranty of - MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU - General Public License for more details. - - You should have received a copy of the GNU General Public License - along with this program. If not, see - <http://www.gnu.org/licenses/>. -*/ + 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. +*/ #include <unistd.h> #include <fcntl.h> #include <limits.h> @@ -38,6 +28,8 @@ #include "common.h" #include "statedump.h" #include "clear.h" +#include "defaults.h" +#include "syncop.h" #ifndef LLONG_MAX #define LLONG_MAX LONG_LONG_MAX /* compat with old gcc */ @@ -48,13 +40,9 @@ void do_blocked_rw (pl_inode_t *); static int __rw_allowable (pl_inode_t *, posix_lock_t *, glusterfs_fop_t); - -struct _truncate_ops { - loc_t loc; - fd_t *fd; - off_t offset; - enum {TRUNCATE, FTRUNCATE} op; -}; +static int format_brickname(char *); +int pl_lockinfo_get_brickname (xlator_t *, inode_t *, int32_t *); +static int fetch_pathinfo(xlator_t *, inode_t *, int32_t *, char **); static pl_fdctx_t * pl_new_fdctx () @@ -63,7 +51,7 @@ pl_new_fdctx () fdctx = GF_CALLOC (1, sizeof (*fdctx), gf_locks_mt_pl_fdctx_t); - GF_VALIDATE_OR_GOTO (POSIX_LOCKS, fdctx, out); + GF_VALIDATE_OR_GOTO ("posix-locks", fdctx, out); INIT_LIST_HEAD (&fdctx->locks_list); @@ -78,7 +66,7 @@ pl_check_n_create_fdctx (xlator_t *this, fd_t *fd) uint64_t tmp = 0; pl_fdctx_t *fdctx = NULL; - GF_VALIDATE_OR_GOTO (POSIX_LOCKS, this, out); + GF_VALIDATE_OR_GOTO ("posix-locks", this, out); GF_VALIDATE_OR_GOTO (this->name, fd, out); LOCK (&fd->lock); @@ -109,24 +97,29 @@ out: int pl_truncate_cbk (call_frame_t *frame, void *cookie, xlator_t *this, int32_t op_ret, int32_t op_errno, struct iatt *prebuf, - struct iatt *postbuf) + struct iatt *postbuf, dict_t *xdata) { - struct _truncate_ops *local = NULL; + pl_local_t *local = NULL; local = frame->local; if (local->op == TRUNCATE) loc_wipe (&local->loc); + if (local->xdata) + dict_unref (local->xdata); + if (local->fd) + fd_unref (local->fd); + STACK_UNWIND_STRICT (truncate, frame, op_ret, op_errno, - prebuf, postbuf); + prebuf, postbuf, xdata); return 0; } static int truncate_allowed (pl_inode_t *pl_inode, - void *transport, pid_t client_pid, + client_t *client, pid_t client_pid, gf_lkowner_t *owner, off_t offset) { posix_lock_t *l = NULL; @@ -135,7 +128,7 @@ truncate_allowed (pl_inode_t *pl_inode, region.fl_start = offset; region.fl_end = LLONG_MAX; - region.transport = transport; + region.client = client; region.client_pid = client_pid; region.owner = *owner; @@ -146,7 +139,7 @@ truncate_allowed (pl_inode_t *pl_inode, && locks_overlap (®ion, l) && !same_owner (®ion, l)) { ret = 0; - gf_log (POSIX_LOCKS, GF_LOG_TRACE, "Truncate " + gf_log ("posix-locks", GF_LOG_TRACE, "Truncate " "allowed"); break; } @@ -160,10 +153,11 @@ truncate_allowed (pl_inode_t *pl_inode, static int truncate_stat_cbk (call_frame_t *frame, void *cookie, xlator_t *this, - int32_t op_ret, int32_t op_errno, struct iatt *buf) + int32_t op_ret, int32_t op_errno, struct iatt *buf, + dict_t *xdata) { posix_locks_private_t *priv = NULL; - struct _truncate_ops *local = NULL; + pl_local_t *local = NULL; inode_t *inode = NULL; pl_inode_t *pl_inode = NULL; @@ -192,7 +186,7 @@ truncate_stat_cbk (call_frame_t *frame, void *cookie, xlator_t *this, if (priv->mandatory && pl_inode->mandatory - && !truncate_allowed (pl_inode, frame->root->trans, + && !truncate_allowed (pl_inode, frame->root->client, frame->root->pid, &frame->root->lk_owner, local->offset)) { op_ret = -1; @@ -204,12 +198,12 @@ truncate_stat_cbk (call_frame_t *frame, void *cookie, xlator_t *this, case TRUNCATE: STACK_WIND (frame, pl_truncate_cbk, FIRST_CHILD (this), FIRST_CHILD (this)->fops->truncate, - &local->loc, local->offset); + &local->loc, local->offset, local->xdata); break; case FTRUNCATE: STACK_WIND (frame, pl_truncate_cbk, FIRST_CHILD (this), FIRST_CHILD (this)->fops->ftruncate, - local->fd, local->offset); + local->fd, local->offset, local->xdata); break; } @@ -220,37 +214,42 @@ unwind: "error: %s", op_ret, strerror (op_errno)); if (local->op == TRUNCATE) loc_wipe (&local->loc); + if (local->xdata) + dict_unref (local->xdata); + if (local->fd) + fd_unref (local->fd); - STACK_UNWIND_STRICT (truncate, frame, op_ret, op_errno, buf, NULL); + STACK_UNWIND_STRICT (truncate, frame, op_ret, op_errno, buf, NULL, xdata); return 0; } int pl_truncate (call_frame_t *frame, xlator_t *this, - loc_t *loc, off_t offset) + loc_t *loc, off_t offset, dict_t *xdata) { - struct _truncate_ops *local = NULL; + pl_local_t *local = NULL; - local = GF_CALLOC (1, sizeof (struct _truncate_ops), - gf_locks_mt_truncate_ops); + local = mem_get0 (this->local_pool); GF_VALIDATE_OR_GOTO (this->name, local, unwind); local->op = TRUNCATE; local->offset = offset; loc_copy (&local->loc, loc); + if (xdata) + local->xdata = dict_ref (xdata); frame->local = local; STACK_WIND (frame, truncate_stat_cbk, FIRST_CHILD (this), - FIRST_CHILD (this)->fops->stat, loc); + FIRST_CHILD (this)->fops->stat, loc, NULL); return 0; unwind: gf_log (this->name, GF_LOG_ERROR, "truncate for %s failed with ret: %d, " "error: %s", loc->path, -1, strerror (ENOMEM)); - STACK_UNWIND_STRICT (truncate, frame, -1, ENOMEM, NULL, NULL); + STACK_UNWIND_STRICT (truncate, frame, -1, ENOMEM, NULL, NULL, NULL); return 0; } @@ -258,32 +257,54 @@ unwind: int pl_ftruncate (call_frame_t *frame, xlator_t *this, - fd_t *fd, off_t offset) + fd_t *fd, off_t offset, dict_t *xdata) { - struct _truncate_ops *local = NULL; + pl_local_t *local = NULL; - local = GF_CALLOC (1, sizeof (struct _truncate_ops), - gf_locks_mt_truncate_ops); + local = mem_get0 (this->local_pool); GF_VALIDATE_OR_GOTO (this->name, local, unwind); local->op = FTRUNCATE; local->offset = offset; - local->fd = fd; + local->fd = fd_ref (fd); + if (xdata) + local->xdata = dict_ref (xdata); frame->local = local; STACK_WIND (frame, truncate_stat_cbk, FIRST_CHILD(this), - FIRST_CHILD(this)->fops->fstat, fd); + FIRST_CHILD(this)->fops->fstat, fd, xdata); return 0; unwind: gf_log (this->name, GF_LOG_ERROR, "ftruncate failed with ret: %d, " "error: %s", -1, strerror (ENOMEM)); - STACK_UNWIND_STRICT (ftruncate, frame, -1, ENOMEM, NULL, NULL); + STACK_UNWIND_STRICT (ftruncate, frame, -1, ENOMEM, NULL, NULL, NULL); return 0; } +int +pl_locks_by_fd (pl_inode_t *pl_inode, fd_t *fd) +{ + posix_lock_t *l = NULL; + int found = 0; + + pthread_mutex_lock (&pl_inode->mutex); + { + + list_for_each_entry (l, &pl_inode->ext_list, list) { + if ((l->fd_num == fd_to_fdnum(fd))) { + found = 1; + break; + } + } + + } + pthread_mutex_unlock (&pl_inode->mutex); + return found; +} + static void delete_locks_of_fd (xlator_t *this, pl_inode_t *pl_inode, fd_t *fd) { @@ -313,7 +334,8 @@ delete_locks_of_fd (xlator_t *this, pl_inode_t *pl_inode, fd_t *fd) list_for_each_entry_safe (l, tmp, &blocked_list, list) { list_del_init(&l->list); - STACK_UNWIND_STRICT (lk, l->frame, -1, EAGAIN, &l->user_flock); + STACK_UNWIND_STRICT (lk, l->frame, -1, EAGAIN, &l->user_flock, + NULL); __destroy_lock (l); } @@ -325,7 +347,7 @@ delete_locks_of_fd (xlator_t *this, pl_inode_t *pl_inode, fd_t *fd) static void __delete_locks_of_owner (pl_inode_t *pl_inode, - void *transport, gf_lkowner_t *owner) + client_t *client, gf_lkowner_t *owner) { posix_lock_t *tmp = NULL; posix_lock_t *l = NULL; @@ -333,7 +355,9 @@ __delete_locks_of_owner (pl_inode_t *pl_inode, /* TODO: what if it is a blocked lock with pending l->frame */ list_for_each_entry_safe (l, tmp, &pl_inode->ext_list, list) { - if ((l->transport == transport) && + if (l->blocked) + continue; + if ((l->client == client) && is_same_lkowner (&l->owner, owner)) { gf_log ("posix-locks", GF_LOG_TRACE, " Flushing lock" @@ -356,18 +380,18 @@ __delete_locks_of_owner (pl_inode_t *pl_inode, int32_t pl_getxattr_cbk (call_frame_t *frame, void *cookie, xlator_t *this, - int32_t op_ret, int32_t op_errno, dict_t *dict) + int32_t op_ret, int32_t op_errno, dict_t *dict, dict_t *xdata) { - STACK_UNWIND_STRICT (getxattr, frame, op_ret, op_errno, dict); + STACK_UNWIND_STRICT (getxattr, frame, op_ret, op_errno, dict, xdata); return 0; } int32_t pl_getxattr (call_frame_t *frame, xlator_t *this, loc_t *loc, - const char *name) + const char *name, dict_t *xdata) { - int op_errno = EINVAL; + int32_t op_errno = EINVAL; int op_ret = -1; int32_t bcount = 0; int32_t gcount = 0; @@ -375,7 +399,8 @@ pl_getxattr (call_frame_t *frame, xlator_t *this, loc_t *loc, char *lk_summary = NULL; pl_inode_t *pl_inode = NULL; dict_t *dict = NULL; - clrlk_args args = {0,}; + clrlk_args args = {0,}; + char *brickname = NULL; if (!name) goto usual; @@ -422,33 +447,51 @@ pl_getxattr (call_frame_t *frame, xlator_t *this, loc_t *loc, goto out; } + op_ret = fetch_pathinfo (this, loc->inode, &op_errno, &brickname); + if (op_ret) { + gf_log (this->name, GF_LOG_WARNING, + "Couldn't get brickname"); + } else { + op_ret = format_brickname(brickname); + if (op_ret) { + gf_log (this->name, GF_LOG_WARNING, + "Couldn't format brickname"); + GF_FREE(brickname); + brickname = NULL; + } + } + if (!gcount && !bcount) { if (gf_asprintf (&lk_summary, "No locks cleared.") == -1) { + op_ret = -1; op_errno = ENOMEM; goto out; } } else if (gf_asprintf (&lk_summary, "%s: %s blocked locks=%d " - "granted locks=%d", this->name, + "granted locks=%d", + (brickname == NULL)? this->name : brickname, (args.type == CLRLK_INODE)? "inode": (args.type == CLRLK_ENTRY)? "entry": (args.type == CLRLK_POSIX)? "posix": " ", bcount, gcount) == -1) { + op_ret = -1; op_errno = ENOMEM; goto out; } strncpy (key, name, strlen (name)); if (dict_set_dynstr (dict, key, lk_summary)) { + op_ret = -1; op_errno = ENOMEM; goto out; } op_ret = 0; out: - STACK_UNWIND_STRICT (getxattr, frame, op_ret, op_errno, dict); + GF_FREE(brickname); + STACK_UNWIND_STRICT (getxattr, frame, op_ret, op_errno, dict, xdata); - if (args.opts) - GF_FREE (args.opts); + GF_FREE (args.opts); if (op_ret && lk_summary) GF_FREE (lk_summary); if (dict) @@ -457,17 +500,415 @@ out: usual: STACK_WIND (frame, pl_getxattr_cbk, FIRST_CHILD(this), - FIRST_CHILD(this)->fops->getxattr, loc, name); + FIRST_CHILD(this)->fops->getxattr, loc, name, xdata); + return 0; +} + +static int +format_brickname(char *brickname) +{ + int ret = -1; + char *hostname = NULL; + char *volume = NULL; + char *saveptr = NULL; + + if (!brickname) + goto out; + + strtok_r(brickname, ":", &saveptr); + hostname = gf_strdup(strtok_r(NULL, ":", &saveptr)); + if (hostname == NULL) + goto out; + volume = gf_strdup(strtok_r(NULL, ".", &saveptr)); + if (volume == NULL) + goto out; + + sprintf(brickname, "%s:%s", hostname, volume); + + ret = 0; +out: + GF_FREE(hostname); + GF_FREE(volume); + return ret; +} + +static int +fetch_pathinfo (xlator_t *this, inode_t *inode, int32_t *op_errno, + char **brickname) +{ + int ret = -1; + loc_t loc = {0, }; + dict_t *dict = NULL; + + if (!brickname) + goto out; + + if (!op_errno) + goto out; + + uuid_copy (loc.gfid, inode->gfid); + loc.inode = inode_ref (inode); + + ret = syncop_getxattr (FIRST_CHILD(this), &loc, &dict, + GF_XATTR_PATHINFO_KEY); + if (ret < 0) { + *op_errno = errno; + goto out; + } + + ret = dict_get_str (dict, GF_XATTR_PATHINFO_KEY, brickname); + if (ret) + goto out; + + *brickname = gf_strdup(*brickname); + if (*brickname == NULL) { + ret = -1; + goto out; + } + + ret = 0; +out: + if (dict != NULL) { + dict_unref (dict); + } + loc_wipe(&loc); + + return ret; +} + + +int +pl_lockinfo_get_brickname (xlator_t *this, inode_t *inode, int32_t *op_errno) +{ + int ret = -1; + posix_locks_private_t *priv = NULL; + char *brickname = NULL; + char *end = NULL; + char *tmp = NULL; + + priv = this->private; + + ret = fetch_pathinfo (this, inode, op_errno, &brickname); + if (ret) + goto out; + + end = strrchr (brickname, ':'); + if (!end) { + GF_FREE(brickname); + ret = -1; + goto out; + } + + tmp = brickname; + brickname = gf_strndup (brickname, (end - brickname)); + if (brickname == NULL) { + ret = -1; + goto out; + } + + priv->brickname = brickname; + ret = 0; +out: + GF_FREE(tmp); + return ret; +} + +char * +pl_lockinfo_key (xlator_t *this, inode_t *inode, int32_t *op_errno) +{ + posix_locks_private_t *priv = NULL; + char *key = NULL; + int ret = 0; + + priv = this->private; + + if (priv->brickname == NULL) { + ret = pl_lockinfo_get_brickname (this, inode, op_errno); + if (ret < 0) { + gf_log (this->name, GF_LOG_WARNING, + "cannot get brickname"); + goto out; + } + } + + key = priv->brickname; +out: + return key; +} + +int32_t +pl_fgetxattr_handle_lockinfo (xlator_t *this, fd_t *fd, + dict_t *dict, int32_t *op_errno) +{ + pl_inode_t *pl_inode = NULL; + char *key = NULL, *buf = NULL; + int32_t op_ret = 0; + unsigned long fdnum = 0, len = 0; + dict_t *tmp = NULL; + + pl_inode = pl_inode_get (this, fd->inode); + + if (!pl_inode) { + gf_log (this->name, GF_LOG_DEBUG, "Could not get inode."); + *op_errno = EBADFD; + op_ret = -1; + goto out; + } + + if (!pl_locks_by_fd (pl_inode, fd)) { + op_ret = 0; + goto out; + } + + fdnum = fd_to_fdnum (fd); + + key = pl_lockinfo_key (this, fd->inode, op_errno); + if (key == NULL) { + op_ret = -1; + goto out; + } + + tmp = dict_new (); + if (tmp == NULL) { + op_ret = -1; + *op_errno = ENOMEM; + goto out; + } + + op_ret = dict_set_uint64 (tmp, key, fdnum); + if (op_ret < 0) { + *op_errno = -op_ret; + op_ret = -1; + gf_log (this->name, GF_LOG_WARNING, "setting lockinfo value " + "(%lu) for fd (ptr:%p inode-gfid:%s) failed (%s)", + fdnum, fd, uuid_utoa (fd->inode->gfid), + strerror (*op_errno)); + goto out; + } + + len = dict_serialized_length (tmp); + if (len < 0) { + *op_errno = -op_ret; + op_ret = -1; + gf_log (this->name, GF_LOG_WARNING, + "dict_serialized_length failed (%s) while handling " + "lockinfo for fd (ptr:%p inode-gfid:%s)", + strerror (*op_errno), fd, uuid_utoa (fd->inode->gfid)); + goto out; + } + + buf = GF_CALLOC (1, len, gf_common_mt_char); + if (buf == NULL) { + op_ret = -1; + *op_errno = ENOMEM; + goto out; + } + + op_ret = dict_serialize (tmp, buf); + if (op_ret < 0) { + *op_errno = -op_ret; + op_ret = -1; + gf_log (this->name, GF_LOG_WARNING, + "dict_serialize failed (%s) while handling lockinfo " + "for fd (ptr: %p inode-gfid:%s)", strerror (*op_errno), + fd, uuid_utoa (fd->inode->gfid)); + goto out; + } + + op_ret = dict_set_dynptr (dict, GF_XATTR_LOCKINFO_KEY, buf, len); + if (op_ret < 0) { + *op_errno = -op_ret; + op_ret = -1; + gf_log (this->name, GF_LOG_WARNING, "setting lockinfo value " + "(%lu) for fd (ptr:%p inode-gfid:%s) failed (%s)", + fdnum, fd, uuid_utoa (fd->inode->gfid), + strerror (*op_errno)); + goto out; + } + + buf = NULL; +out: + if (tmp != NULL) { + dict_unref (tmp); + } + + if (buf != NULL) { + GF_FREE (buf); + } + + return op_ret; +} + + +int32_t +pl_fgetxattr (call_frame_t *frame, xlator_t *this, fd_t *fd, + const char *name, dict_t *xdata) +{ + int32_t op_ret = 0, op_errno = 0; + dict_t *dict = NULL; + + if (!name) { + goto usual; + } + + if (strcmp (name, GF_XATTR_LOCKINFO_KEY) == 0) { + dict = dict_new (); + if (dict == NULL) { + op_ret = -1; + op_errno = ENOMEM; + goto unwind; + } + + op_ret = pl_fgetxattr_handle_lockinfo (this, fd, dict, + &op_errno); + if (op_ret < 0) { + gf_log (this->name, GF_LOG_WARNING, + "getting lockinfo on fd (ptr:%p inode-gfid:%s) " + "failed (%s)", fd, uuid_utoa (fd->inode->gfid), + strerror (op_errno)); + } + + goto unwind; + } else { + goto usual; + } + +unwind: + STACK_UNWIND_STRICT (fgetxattr, frame, op_ret, op_errno, dict, NULL); + if (dict != NULL) { + dict_unref (dict); + } + + return 0; + +usual: + STACK_WIND (frame, default_fgetxattr_cbk, FIRST_CHILD(this), + FIRST_CHILD(this)->fops->fgetxattr, fd, name, xdata); + return 0; +} + +int32_t +pl_migrate_locks (call_frame_t *frame, fd_t *newfd, uint64_t oldfd_num, + int32_t *op_errno) +{ + pl_inode_t *pl_inode = NULL; + uint64_t newfd_num = 0; + posix_lock_t *l = NULL; + int32_t op_ret = 0; + + newfd_num = fd_to_fdnum (newfd); + + pl_inode = pl_inode_get (frame->this, newfd->inode); + if (pl_inode == NULL) { + op_ret = -1; + *op_errno = EBADFD; + goto out; + } + + pthread_mutex_lock (&pl_inode->mutex); + { + list_for_each_entry (l, &pl_inode->ext_list, list) { + if (l->fd_num == oldfd_num) { + l->fd_num = newfd_num; + l->client = frame->root->client; + } + } + } + pthread_mutex_unlock (&pl_inode->mutex); + + op_ret = 0; +out: + return op_ret; +} + +int32_t +pl_fsetxattr_handle_lockinfo (call_frame_t *frame, fd_t *fd, char *lockinfo_buf, + int len, int32_t *op_errno) +{ + int32_t op_ret = -1; + dict_t *lockinfo = NULL; + uint64_t oldfd_num = 0; + char *key = NULL; + + lockinfo = dict_new (); + if (lockinfo == NULL) { + op_ret = -1; + *op_errno = ENOMEM; + goto out; + } + + op_ret = dict_unserialize (lockinfo_buf, len, &lockinfo); + if (op_ret < 0) { + *op_errno = -op_ret; + op_ret = -1; + goto out; + } + + key = pl_lockinfo_key (frame->this, fd->inode, op_errno); + if (key == NULL) { + op_ret = -1; + goto out; + } + + op_ret = dict_get_uint64 (lockinfo, key, &oldfd_num); + + if (oldfd_num == 0) { + op_ret = 0; + goto out; + } + + op_ret = pl_migrate_locks (frame, fd, oldfd_num, op_errno); + if (op_ret < 0) { + gf_log (frame->this->name, GF_LOG_WARNING, + "migration of locks from oldfd (ptr:%p) to newfd " + "(ptr:%p) (inode-gfid:%s)", (void *)oldfd_num, fd, + uuid_utoa (fd->inode->gfid)); + goto out; + } + +out: + dict_unref (lockinfo); + + return op_ret; +} + +int32_t +pl_fsetxattr (call_frame_t *frame, xlator_t *this, fd_t *fd, dict_t *dict, + int32_t flags, dict_t *xdata) +{ + int32_t op_ret = 0, op_errno = 0; + void *lockinfo_buf = NULL; + int len = 0; + + op_ret = dict_get_ptr_and_len (dict, GF_XATTR_LOCKINFO_KEY, + &lockinfo_buf, &len); + if (lockinfo_buf == NULL) { + goto usual; + } + + op_ret = pl_fsetxattr_handle_lockinfo (frame, fd, lockinfo_buf, len, + &op_errno); + if (op_ret < 0) { + goto unwind; + } + +usual: + STACK_WIND (frame, default_fsetxattr_cbk, FIRST_CHILD(this), + FIRST_CHILD(this)->fops->fsetxattr, fd, dict, flags, xdata); + return 0; + +unwind: + STACK_UNWIND_STRICT (fsetxattr, frame, op_ret, op_errno, NULL); return 0; } int32_t pl_opendir_cbk (call_frame_t *frame, - void *cookie, - xlator_t *this, - int32_t op_ret, - int32_t op_errno, - fd_t *fd) + void *cookie, + xlator_t *this, + int32_t op_ret, + int32_t op_errno, + fd_t *fd, dict_t *xdata) { pl_fdctx_t *fdctx = NULL; @@ -486,28 +927,28 @@ unwind: frame, op_ret, op_errno, - fd); + fd, xdata); return 0; } int32_t pl_opendir (call_frame_t *frame, xlator_t *this, - loc_t *loc, fd_t *fd) + loc_t *loc, fd_t *fd, dict_t *xdata) { STACK_WIND (frame, pl_opendir_cbk, FIRST_CHILD(this), FIRST_CHILD(this)->fops->opendir, - loc, fd); + loc, fd, xdata); return 0; } int pl_flush_cbk (call_frame_t *frame, void *cookie, xlator_t *this, - int32_t op_ret, int32_t op_errno) + int32_t op_ret, int32_t op_errno, dict_t *xdata) { - STACK_UNWIND_STRICT (flush, frame, op_ret, op_errno); + STACK_UNWIND_STRICT (flush, frame, op_ret, op_errno, xdata); return 0; } @@ -515,7 +956,7 @@ pl_flush_cbk (call_frame_t *frame, void *cookie, xlator_t *this, int pl_flush (call_frame_t *frame, xlator_t *this, - fd_t *fd) + fd_t *fd, dict_t *xdata) { pl_inode_t *pl_inode = NULL; @@ -523,7 +964,7 @@ pl_flush (call_frame_t *frame, xlator_t *this, if (!pl_inode) { gf_log (this->name, GF_LOG_DEBUG, "Could not get inode."); - STACK_UNWIND_STRICT (flush, frame, -1, EBADFD); + STACK_UNWIND_STRICT (flush, frame, -1, EBADFD, NULL); return 0; } @@ -542,7 +983,7 @@ pl_flush (call_frame_t *frame, xlator_t *this, } pthread_mutex_lock (&pl_inode->mutex); { - __delete_locks_of_owner (pl_inode, frame->root->trans, + __delete_locks_of_owner (pl_inode, frame->root->client, &frame->root->lk_owner); } pthread_mutex_unlock (&pl_inode->mutex); @@ -553,14 +994,14 @@ pl_flush (call_frame_t *frame, xlator_t *this, wind: STACK_WIND (frame, pl_flush_cbk, FIRST_CHILD(this), - FIRST_CHILD(this)->fops->flush, fd); + FIRST_CHILD(this)->fops->flush, fd, xdata); return 0; } int pl_open_cbk (call_frame_t *frame, void *cookie, xlator_t *this, - int32_t op_ret, int32_t op_errno, fd_t *fd) + int32_t op_ret, int32_t op_errno, fd_t *fd, dict_t *xdata) { pl_fdctx_t *fdctx = NULL; @@ -575,7 +1016,7 @@ pl_open_cbk (call_frame_t *frame, void *cookie, xlator_t *this, } unwind: - STACK_UNWIND_STRICT (open, frame, op_ret, op_errno, fd); + STACK_UNWIND_STRICT (open, frame, op_ret, op_errno, fd, xdata); return 0; } @@ -583,12 +1024,11 @@ unwind: int pl_open (call_frame_t *frame, xlator_t *this, loc_t *loc, int32_t flags, - fd_t *fd, int32_t wbflags) + fd_t *fd, dict_t *xdata) { - /* why isn't O_TRUNC being handled ? */ STACK_WIND (frame, pl_open_cbk, FIRST_CHILD(this), FIRST_CHILD(this)->fops->open, - loc, flags & ~O_TRUNC, fd, wbflags); + loc, flags, fd, xdata); return 0; } @@ -598,7 +1038,7 @@ int pl_create_cbk (call_frame_t *frame, void *cookie, xlator_t *this, int32_t op_ret, int32_t op_errno, fd_t *fd, inode_t *inode, struct iatt *buf, - struct iatt *preparent, struct iatt *postparent) + struct iatt *preparent, struct iatt *postparent, dict_t *xdata) { pl_fdctx_t *fdctx = NULL; @@ -614,7 +1054,7 @@ pl_create_cbk (call_frame_t *frame, void *cookie, unwind: STACK_UNWIND_STRICT (create, frame, op_ret, op_errno, fd, inode, buf, - preparent, postparent); + preparent, postparent, xdata); return 0; } @@ -622,11 +1062,12 @@ unwind: int pl_create (call_frame_t *frame, xlator_t *this, - loc_t *loc, int32_t flags, mode_t mode, fd_t *fd, dict_t *params) + loc_t *loc, int32_t flags, mode_t mode, mode_t umask, fd_t *fd, + dict_t *xdata) { STACK_WIND (frame, pl_create_cbk, FIRST_CHILD (this), FIRST_CHILD (this)->fops->create, - loc, flags, mode, fd, params); + loc, flags, mode, umask, fd, xdata); return 0; } @@ -635,10 +1076,10 @@ int pl_readv_cbk (call_frame_t *frame, void *cookie, xlator_t *this, int32_t op_ret, int32_t op_errno, struct iovec *vector, int32_t count, struct iatt *stbuf, - struct iobref *iobref) + struct iobref *iobref, dict_t *xdata) { STACK_UNWIND_STRICT (readv, frame, op_ret, op_errno, - vector, count, stbuf, iobref); + vector, count, stbuf, iobref, xdata); return 0; } @@ -646,9 +1087,10 @@ pl_readv_cbk (call_frame_t *frame, void *cookie, xlator_t *this, int pl_writev_cbk (call_frame_t *frame, void *cookie, xlator_t *this, int32_t op_ret, int32_t op_errno, struct iatt *prebuf, - struct iatt *postbuf) + struct iatt *postbuf, dict_t *xdata) { - STACK_UNWIND_STRICT (writev, frame, op_ret, op_errno, prebuf, postbuf); + STACK_UNWIND_STRICT (writev, frame, op_ret, op_errno, prebuf, postbuf, + xdata); return 0; } @@ -706,12 +1148,12 @@ __rw_allowable (pl_inode_t *pl_inode, posix_lock_t *region, int -pl_readv_cont (call_frame_t *frame, xlator_t *this, - fd_t *fd, size_t size, off_t offset, uint32_t flags) +pl_readv_cont (call_frame_t *frame, xlator_t *this, fd_t *fd, size_t size, + off_t offset, uint32_t flags, dict_t *xdata) { STACK_WIND (frame, pl_readv_cbk, FIRST_CHILD (this), FIRST_CHILD (this)->fops->readv, - fd, size, offset, flags); + fd, size, offset, flags, xdata); return 0; } @@ -719,7 +1161,7 @@ pl_readv_cont (call_frame_t *frame, xlator_t *this, int pl_readv (call_frame_t *frame, xlator_t *this, - fd_t *fd, size_t size, off_t offset, uint32_t flags) + fd_t *fd, size_t size, off_t offset, uint32_t flags, dict_t *xdata) { posix_locks_private_t *priv = NULL; pl_inode_t *pl_inode = NULL; @@ -736,7 +1178,7 @@ pl_readv (call_frame_t *frame, xlator_t *this, if (priv->mandatory && pl_inode->mandatory) { region.fl_start = offset; region.fl_end = offset + size - 1; - region.transport = frame->root->trans; + region.client = frame->root->client; region.fd_num = fd_to_fdnum(fd); region.client_pid = frame->root->pid; region.owner = frame->root->lk_owner; @@ -766,7 +1208,8 @@ pl_readv (call_frame_t *frame, xlator_t *this, } rw->stub = fop_readv_stub (frame, pl_readv_cont, - fd, size, offset, flags); + fd, size, offset, flags, + xdata); if (!rw->stub) { op_errno = ENOMEM; op_ret = -1; @@ -786,12 +1229,12 @@ pl_readv (call_frame_t *frame, xlator_t *this, if (wind_needed) { STACK_WIND (frame, pl_readv_cbk, FIRST_CHILD (this), FIRST_CHILD (this)->fops->readv, - fd, size, offset, flags); + fd, size, offset, flags, xdata); } if (op_ret == -1) STACK_UNWIND_STRICT (readv, frame, -1, op_errno, - NULL, 0, NULL, NULL); + NULL, 0, NULL, NULL, NULL); return 0; } @@ -800,11 +1243,11 @@ pl_readv (call_frame_t *frame, xlator_t *this, int pl_writev_cont (call_frame_t *frame, xlator_t *this, fd_t *fd, struct iovec *vector, int count, off_t offset, - uint32_t flags, struct iobref *iobref) + uint32_t flags, struct iobref *iobref, dict_t *xdata) { STACK_WIND (frame, pl_writev_cbk, FIRST_CHILD (this), FIRST_CHILD (this)->fops->writev, - fd, vector, count, offset, flags, iobref); + fd, vector, count, offset, flags, iobref, xdata); return 0; } @@ -813,7 +1256,7 @@ pl_writev_cont (call_frame_t *frame, xlator_t *this, fd_t *fd, int pl_writev (call_frame_t *frame, xlator_t *this, fd_t *fd, struct iovec *vector, int32_t count, off_t offset, - uint32_t flags, struct iobref *iobref) + uint32_t flags, struct iobref *iobref, dict_t *xdata) { posix_locks_private_t *priv = NULL; pl_inode_t *pl_inode = NULL; @@ -829,7 +1272,7 @@ pl_writev (call_frame_t *frame, xlator_t *this, fd_t *fd, if (priv->mandatory && pl_inode->mandatory) { region.fl_start = offset; region.fl_end = offset + iov_length (vector, count) - 1; - region.transport = frame->root->trans; + region.client = frame->root->client; region.fd_num = fd_to_fdnum(fd); region.client_pid = frame->root->pid; region.owner = frame->root->lk_owner; @@ -860,7 +1303,7 @@ pl_writev (call_frame_t *frame, xlator_t *this, fd_t *fd, rw->stub = fop_writev_stub (frame, pl_writev_cont, fd, vector, count, offset, - flags, iobref); + flags, iobref, xdata); if (!rw->stub) { op_errno = ENOMEM; op_ret = -1; @@ -880,10 +1323,11 @@ pl_writev (call_frame_t *frame, xlator_t *this, fd_t *fd, if (wind_needed) STACK_WIND (frame, pl_writev_cbk, FIRST_CHILD (this), FIRST_CHILD (this)->fops->writev, - fd, vector, count, offset, flags, iobref); + fd, vector, count, offset, flags, iobref, xdata); if (op_ret == -1) - STACK_UNWIND_STRICT (writev, frame, -1, op_errno, NULL, NULL); + STACK_UNWIND_STRICT (writev, frame, -1, op_errno, NULL, NULL, + NULL); return 0; } @@ -909,7 +1353,7 @@ lock_dup (posix_lock_t *lock) { posix_lock_t *new_lock = NULL; - new_lock = new_posix_lock (&lock->user_flock, lock->transport, + new_lock = new_posix_lock (&lock->user_flock, lock->client, lock->client_pid, &lock->owner, (fd_t *)lock->fd_num); return new_lock; @@ -1067,10 +1511,8 @@ unlock: int pl_lk (call_frame_t *frame, xlator_t *this, - fd_t *fd, int32_t cmd, struct gf_flock *flock) + fd_t *fd, int32_t cmd, struct gf_flock *flock, dict_t *xdata) { - void *transport = NULL; - pid_t client_pid = 0; pl_inode_t *pl_inode = NULL; int op_ret = 0; int op_errno = 0; @@ -1079,9 +1521,6 @@ pl_lk (call_frame_t *frame, xlator_t *this, posix_lock_t *conf = NULL; int ret = 0; - transport = frame->root->trans; - client_pid = frame->root->pid; - if ((flock->l_start < 0) || (flock->l_len < 0)) { op_ret = -1; op_errno = EINVAL; @@ -1095,7 +1534,7 @@ pl_lk (call_frame_t *frame, xlator_t *this, goto unwind; } - reqlock = new_posix_lock (flock, transport, client_pid, + reqlock = new_posix_lock (flock, frame->root->client, frame->root->pid, &frame->root->lk_owner, fd); if (!reqlock) { @@ -1213,13 +1652,22 @@ pl_lk (call_frame_t *frame, xlator_t *this, op_ret = -1; op_errno = EAGAIN; __destroy_lock (reqlock); + + } else if ((0 == ret) && (F_UNLCK == flock->l_type)) { + /* For NLM's last "unlock on fd" detection */ + if (pl_locks_by_fd (pl_inode, fd)) + flock->l_type = F_RDLCK; + else + flock->l_type = F_UNLCK; } } unwind: pl_trace_out (this, frame, fd, NULL, cmd, flock, op_ret, op_errno, NULL); pl_update_refkeeper (this, fd->inode); - STACK_UNWIND_STRICT (lk, frame, op_ret, op_errno, flock); + + + STACK_UNWIND_STRICT (lk, frame, op_ret, op_errno, flock, xdata); out: return 0; } @@ -1296,7 +1744,7 @@ pl_forget (xlator_t *this, list_for_each_entry_safe (ino_l, ino_tmp, &dom->inodelk_list, list) { __delete_inode_lock (ino_l); - __destroy_inode_lock (ino_l); + __pl_inodelk_unref (ino_l); } list_splice_init (&dom->blocked_inodelks, &inodelks_released); @@ -1310,8 +1758,8 @@ pl_forget (xlator_t *this, list_for_each_entry_safe (entry_l, entry_tmp, &dom->entrylk_list, domain_list) { list_del_init (&entry_l->domain_list); - if (entry_l->basename) - GF_FREE ((char *)entry_l->basename); + GF_FREE ((char *)entry_l->basename); + GF_FREE (entry_l->connection_id); GF_FREE (entry_l); } @@ -1330,21 +1778,22 @@ pl_forget (xlator_t *this, list_for_each_entry_safe (ext_l, ext_tmp, &posixlks_released, list) { - STACK_UNWIND_STRICT (lk, ext_l->frame, -1, 0, &ext_l->user_flock); + STACK_UNWIND_STRICT (lk, ext_l->frame, -1, 0, + &ext_l->user_flock, NULL); __destroy_lock (ext_l); } list_for_each_entry_safe (ino_l, ino_tmp, &inodelks_released, blocked_locks) { - STACK_UNWIND_STRICT (inodelk, ino_l->frame, -1, 0); - __destroy_inode_lock (ino_l); + STACK_UNWIND_STRICT (inodelk, ino_l->frame, -1, 0, NULL); + __pl_inodelk_unref (ino_l); } list_for_each_entry_safe (entry_l, entry_tmp, &entrylks_released, blocked_locks) { - STACK_UNWIND_STRICT (entrylk, entry_l->frame, -1, 0); - if (entry_l->basename) - GF_FREE ((char *)entry_l->basename); + STACK_UNWIND_STRICT (entrylk, entry_l->frame, -1, 0, NULL); + GF_FREE ((char *)entry_l->basename); + GF_FREE (entry_l->connection_id); GF_FREE (entry_l); } @@ -1420,7 +1869,7 @@ out: return ret; } -static int32_t +int32_t __get_posixlk_count (xlator_t *this, pl_inode_t *pl_inode) { posix_lock_t *lock = NULL; @@ -1428,16 +1877,6 @@ __get_posixlk_count (xlator_t *this, pl_inode_t *pl_inode) list_for_each_entry (lock, &pl_inode->ext_list, list) { - gf_log (this->name, GF_LOG_DEBUG, - " XATTR DEBUG" - "%s (pid=%d) (lk-owner=%s) %"PRId64" - %"PRId64" state: %s", - lock->fl_type == F_UNLCK ? "Unlock" : "Lock", - lock->client_pid, - lkowner_utoa (&lock->owner), - lock->user_flock.l_start, - lock->user_flock.l_len, - lock->blocked == 1 ? "Blocked" : "Active"); - count++; } @@ -1470,6 +1909,24 @@ out: } void +pl_parent_entrylk_xattr_fill (xlator_t *this, inode_t *parent, + char *basename, dict_t *dict) +{ + uint32_t entrylk = 0; + int ret = -1; + + if (!parent || !basename || !strlen (basename)) + goto out; + entrylk = check_entrylk_on_basename (this, parent, basename); +out: + ret = dict_set_uint32 (dict, GLUSTERFS_PARENT_ENTRYLK, entrylk); + if (ret < 0) { + gf_log (this->name, GF_LOG_DEBUG, + " dict_set failed on key %s", GLUSTERFS_PARENT_ENTRYLK); + } +} + +void pl_entrylk_xattr_fill (xlator_t *this, inode_t *inode, dict_t *dict) { @@ -1486,19 +1943,34 @@ pl_entrylk_xattr_fill (xlator_t *this, inode_t *inode, } void -pl_inodelk_xattr_fill (xlator_t *this, inode_t *inode, - dict_t *dict) +pl_inodelk_xattr_fill (xlator_t *this, inode_t *inode, dict_t *dict, + gf_boolean_t per_dom) { int32_t count = 0; int ret = -1; + char *domname = NULL; + + + if (per_dom){ + ret = dict_get_str (dict, GLUSTERFS_INODELK_DOM_COUNT, + &domname); + if (ret) { + gf_log (this->name, GF_LOG_ERROR, "Failed to get " + "value for key %s",GLUSTERFS_INODELK_DOM_COUNT); + goto out; + } + } + + count = get_inodelk_count (this, inode, domname); - count = get_inodelk_count (this, inode); ret = dict_set_int32 (dict, GLUSTERFS_INODELK_COUNT, count); if (ret < 0) { - gf_log (this->name, GF_LOG_DEBUG, - " dict_set failed on key %s", GLUSTERFS_INODELK_COUNT); + gf_log (this->name, GF_LOG_DEBUG, "Failed to set count for " + "key %s", GLUSTERFS_INODELK_COUNT); } +out: + return; } void @@ -1525,50 +1997,57 @@ pl_lookup_cbk (call_frame_t *frame, int32_t op_errno, inode_t *inode, struct iatt *buf, - dict_t *dict, + dict_t *xdata, struct iatt *postparent) { pl_local_t *local = NULL; GF_VALIDATE_OR_GOTO (this->name, frame->local, out); - if (op_ret) { + if (op_ret) goto out; - } local = frame->local; + if (local->parent_entrylk_req) + pl_parent_entrylk_xattr_fill (this, local->loc.parent, + (char*)local->loc.name, xdata); if (local->entrylk_count_req) - pl_entrylk_xattr_fill (this, inode, dict); + pl_entrylk_xattr_fill (this, inode, xdata); if (local->inodelk_count_req) - pl_inodelk_xattr_fill (this, inode, dict); + pl_inodelk_xattr_fill (this, inode, xdata, _gf_false); + if (local->inodelk_dom_count_req) + pl_inodelk_xattr_fill (this, inode, xdata, _gf_true); if (local->posixlk_count_req) - pl_posixlk_xattr_fill (this, inode, dict); + pl_posixlk_xattr_fill (this, inode, xdata); +out: + local = frame->local; frame->local = NULL; - if (local != NULL) - GF_FREE (local); + if (local != NULL) { + loc_wipe (&local->loc); + mem_put (local); + } -out: STACK_UNWIND_STRICT ( lookup, frame, - op_ret, - op_errno, - inode, - buf, - dict, - postparent); + op_ret, + op_errno, + inode, + buf, + xdata, + postparent); return 0; } int32_t pl_lookup (call_frame_t *frame, - xlator_t *this, - loc_t *loc, - dict_t *xattr_req) + xlator_t *this, + loc_t *loc, + dict_t *xdata) { pl_local_t *local = NULL; int ret = -1; @@ -1577,36 +2056,41 @@ pl_lookup (call_frame_t *frame, VALIDATE_OR_GOTO (this, out); VALIDATE_OR_GOTO (loc, out); - local = GF_CALLOC (1, sizeof (*local), gf_locks_mt_pl_local_t); + local = mem_get0 (this->local_pool); GF_VALIDATE_OR_GOTO (this->name, local, out); - if (xattr_req) { - if (dict_get (xattr_req, GLUSTERFS_ENTRYLK_COUNT)) + if (xdata) { + if (dict_get (xdata, GLUSTERFS_ENTRYLK_COUNT)) local->entrylk_count_req = 1; - if (dict_get (xattr_req, GLUSTERFS_INODELK_COUNT)) + if (dict_get (xdata, GLUSTERFS_INODELK_COUNT)) local->inodelk_count_req = 1; - if (dict_get (xattr_req, GLUSTERFS_POSIXLK_COUNT)) + if (dict_get (xdata, GLUSTERFS_INODELK_DOM_COUNT)) + local->inodelk_dom_count_req = 1; + if (dict_get (xdata, GLUSTERFS_POSIXLK_COUNT)) local->posixlk_count_req = 1; + if (dict_get (xdata, GLUSTERFS_PARENT_ENTRYLK)) + local->parent_entrylk_req = 1; } frame->local = local; + loc_copy (&local->loc, loc); STACK_WIND (frame, pl_lookup_cbk, FIRST_CHILD(this), FIRST_CHILD(this)->fops->lookup, - loc, - xattr_req); + loc, xdata); ret = 0; out: if (ret == -1) - STACK_UNWIND_STRICT (lookup, frame, -1, 0, NULL, NULL, NULL, NULL); + STACK_UNWIND_STRICT (lookup, frame, -1, 0, NULL, + NULL, NULL, NULL); return 0; } int pl_readdirp_cbk (call_frame_t *frame, void *cookie, xlator_t *this, - int op_ret, int op_errno, gf_dirent_t *entries) + int op_ret, int op_errno, gf_dirent_t *entries, dict_t *xdata) { pl_local_t *local = NULL; gf_dirent_t *entry = NULL; @@ -1620,17 +2104,21 @@ pl_readdirp_cbk (call_frame_t *frame, void *cookie, xlator_t *this, if (local->entrylk_count_req) pl_entrylk_xattr_fill (this, entry->inode, entry->dict); if (local->inodelk_count_req) - pl_inodelk_xattr_fill (this, entry->inode, entry->dict); + pl_inodelk_xattr_fill (this, entry->inode, entry->dict, + _gf_false); + if (local->inodelk_dom_count_req) + pl_inodelk_xattr_fill (this, entry->inode, entry->dict, + _gf_true); if (local->posixlk_count_req) pl_posixlk_xattr_fill (this, entry->inode, entry->dict); } unwind: frame->local = NULL; - STACK_UNWIND_STRICT (readdirp, frame, op_ret, op_errno, entries); + STACK_UNWIND_STRICT (readdirp, frame, op_ret, op_errno, entries, xdata); if (local) - GF_FREE (local); + mem_put (local); return 0; } @@ -1641,7 +2129,7 @@ pl_readdirp (call_frame_t *frame, xlator_t *this, fd_t *fd, size_t size, { pl_local_t *local = NULL; - local = GF_CALLOC (1, sizeof (*local), gf_locks_mt_pl_local_t); + local = mem_get0 (this->local_pool); GF_VALIDATE_OR_GOTO (this->name, local, out); if (dict) { @@ -1649,6 +2137,8 @@ pl_readdirp (call_frame_t *frame, xlator_t *this, fd_t *fd, size_t size, local->entrylk_count_req = 1; if (dict_get (dict, GLUSTERFS_INODELK_COUNT)) local->inodelk_count_req = 1; + if (dict_get (dict, GLUSTERFS_INODELK_DOM_COUNT)) + local->inodelk_dom_count_req = 1; if (dict_get (dict, GLUSTERFS_POSIXLK_COUNT)) local->posixlk_count_req = 1; } @@ -1661,17 +2151,19 @@ pl_readdirp (call_frame_t *frame, xlator_t *this, fd_t *fd, size_t size, return 0; out: - STACK_UNWIND_STRICT (readdirp, frame, -1, ENOMEM, NULL); + STACK_UNWIND_STRICT (readdirp, frame, -1, ENOMEM, NULL, NULL); return 0; } void pl_dump_lock (char *str, int size, struct gf_flock *flock, - gf_lkowner_t *owner, void *trans, time_t *granted_time, - time_t *blkd_time, gf_boolean_t active) + gf_lkowner_t *owner, void *trans, char *conn_id, + time_t *granted_time, time_t *blkd_time, gf_boolean_t active) { - char *type_str = NULL; + char *type_str = NULL; + char granted[32] = {0,}; + char blocked[32] = {0,}; switch (flock->l_type) { case F_RDLCK: @@ -1690,33 +2182,32 @@ pl_dump_lock (char *str, int size, struct gf_flock *flock, if (active) { if (blkd_time && *blkd_time == 0) { - snprintf (str, size, "type=%s, start=%llu, len=%llu, " - "pid=%llu, lk-owner=%s, transport=%p, " - "granted at %s", - type_str, (unsigned long long) flock->l_start, + snprintf (str, size, RANGE_GRNTD_FMT, + type_str, flock->l_whence, + (unsigned long long) flock->l_start, (unsigned long long) flock->l_len, (unsigned long long) flock->l_pid, - lkowner_utoa (owner), - trans, ctime (granted_time)); + lkowner_utoa (owner), trans, conn_id, + ctime_r (granted_time, granted)); } else { - snprintf (str, size, "type=%s, start=%llu, len=%llu, " - "pid=%llu, lk-owner=%s, transport=%p, " - "blocked at %s, granted at %s", - type_str, (unsigned long long) flock->l_start, + snprintf (str, size, RANGE_BLKD_GRNTD_FMT, + type_str, flock->l_whence, + (unsigned long long) flock->l_start, (unsigned long long) flock->l_len, (unsigned long long) flock->l_pid, - lkowner_utoa (owner), - trans, ctime (blkd_time), ctime (granted_time)); + lkowner_utoa (owner), trans, conn_id, + ctime_r (blkd_time, blocked), + ctime_r (granted_time, granted)); } } else { - snprintf (str, size, "type=%s, start=%llu, len=%llu, pid=%llu, " - "lk-owner=%s, transport=%p, blocked at %s", - type_str, (unsigned long long) flock->l_start, + snprintf (str, size, RANGE_BLKD_FMT, + type_str, flock->l_whence, + (unsigned long long) flock->l_start, (unsigned long long) flock->l_len, (unsigned long long) flock->l_pid, - lkowner_utoa (owner), - trans, ctime (blkd_time)); + lkowner_utoa (owner), trans, conn_id, + ctime_r (blkd_time, blocked)); } } @@ -1726,8 +2217,10 @@ __dump_entrylks (pl_inode_t *pl_inode) { pl_dom_list_t *dom = NULL; pl_entry_lock_t *lock = NULL; - int count = 0; - char key[GF_DUMP_MAX_BUF_LEN]; + char blocked[32] = {0,}; + char granted[32] = {0,}; + int count = 0; + char key[GF_DUMP_MAX_BUF_LEN] = {0,}; char tmp[256]; @@ -1746,22 +2239,22 @@ __dump_entrylks (pl_inode_t *pl_inode) "xlator.feature.locks.lock-dump.domain.entrylk", "entrylk[%d](ACTIVE)", count ); if (lock->blkd_time.tv_sec == 0 && lock->blkd_time.tv_usec == 0) { - snprintf (tmp, 256," %s on %s pid = %llu, owner=%s, transport=%p," - " granted at %s", + snprintf (tmp, 256, ENTRY_GRNTD_FMT, lock->type == ENTRYLK_RDLCK ? "ENTRYLK_RDLCK" : "ENTRYLK_WRLCK", lock->basename, (unsigned long long) lock->client_pid, lkowner_utoa (&lock->owner), lock->trans, - ctime (&lock->granted_time.tv_sec)); + lock->connection_id, + ctime_r (&lock->granted_time.tv_sec, granted)); } else { - snprintf (tmp, 256," %s on %s pid = %llu, owner=%s, transport=%p," - " blocked at %s, granted at %s", + snprintf (tmp, 256, ENTRY_BLKD_GRNTD_FMT, lock->type == ENTRYLK_RDLCK ? "ENTRYLK_RDLCK" : "ENTRYLK_WRLCK", lock->basename, (unsigned long long) lock->client_pid, lkowner_utoa (&lock->owner), lock->trans, - ctime (&lock->blkd_time.tv_sec), - ctime (&lock->granted_time.tv_sec)); + lock->connection_id, + ctime_r (&lock->blkd_time.tv_sec, blocked), + ctime_r (&lock->granted_time.tv_sec, granted)); } gf_proc_dump_write(key, tmp); @@ -1774,13 +2267,13 @@ __dump_entrylks (pl_inode_t *pl_inode) gf_proc_dump_build_key(key, "xlator.feature.locks.lock-dump.domain.entrylk", "entrylk[%d](BLOCKED)", count ); - snprintf (tmp, 256," %s on %s pid = %llu, owner=%s, transport=%p," - " blocked at %s", + snprintf (tmp, 256, ENTRY_BLKD_FMT, lock->type == ENTRYLK_RDLCK ? "ENTRYLK_RDLCK" : "ENTRYLK_WRLCK", lock->basename, (unsigned long long) lock->client_pid, lkowner_utoa (&lock->owner), lock->trans, - ctime (&lock->blkd_time.tv_sec)); + lock->connection_id, + ctime_r (&lock->blkd_time.tv_sec, blocked)); gf_proc_dump_write(key, tmp); @@ -1830,7 +2323,7 @@ __dump_inodelks (pl_inode_t *pl_inode) SET_FLOCK_PID (&lock->user_flock, lock); pl_dump_lock (tmp, 256, &lock->user_flock, &lock->owner, - lock->transport, + lock->client, lock->connection_id, &lock->granted_time.tv_sec, &lock->blkd_time.tv_sec, _gf_true); @@ -1847,7 +2340,7 @@ __dump_inodelks (pl_inode_t *pl_inode) SET_FLOCK_PID (&lock->user_flock, lock); pl_dump_lock (tmp, 256, &lock->user_flock, &lock->owner, - lock->transport, + lock->client, lock->connection_id, 0, &lock->blkd_time.tv_sec, _gf_false); gf_proc_dump_write(key, tmp); @@ -1888,16 +2381,13 @@ __dump_posixlks (pl_inode_t *pl_inode) count, lock->blocked ? "BLOCKED" : "ACTIVE"); pl_dump_lock (tmp, 256, &lock->user_flock, - &lock->owner, lock->transport, + &lock->owner, lock->client, NULL, &lock->granted_time.tv_sec, &lock->blkd_time.tv_sec, (lock->blocked)? _gf_false: _gf_true); gf_proc_dump_write(key, tmp); count++; } - - - } void @@ -1918,46 +2408,81 @@ pl_dump_inode_priv (xlator_t *this, inode_t *inode) int ret = -1; uint64_t tmp_pl_inode = 0; pl_inode_t *pl_inode = NULL; + char *pathname = NULL; + gf_boolean_t section_added = _gf_false; int count = 0; - GF_VALIDATE_OR_GOTO (this->name, inode, out); - - ret = inode_ctx_get (inode, this, &tmp_pl_inode); + if (!inode) { + errno = EINVAL; + goto out; + } - if (ret != 0) + ret = TRY_LOCK (&inode->lock); + if (ret) + goto out; + { + ret = __inode_ctx_get (inode, this, &tmp_pl_inode); + if (ret) + goto unlock; + } +unlock: + UNLOCK (&inode->lock); + if (ret) goto out; pl_inode = (pl_inode_t *)(long)tmp_pl_inode; - if (!pl_inode) { ret = -1; goto out; } gf_proc_dump_add_section("xlator.features.locks.%s.inode", this->name); + section_added = _gf_true; + + /*We are safe to call __inode_path since we have the + * inode->table->lock */ + __inode_path (inode, NULL, &pathname); + if (pathname) + gf_proc_dump_write ("path", "%s", pathname); gf_proc_dump_write("mandatory", "%d", pl_inode->mandatory); - count = get_entrylk_count (this, inode); - if (count) { - gf_proc_dump_write("entrylk-count", "%d", count); - dump_entrylks(pl_inode); - } + ret = pthread_mutex_trylock (&pl_inode->mutex); + if (ret) + goto out; + { + count = __get_entrylk_count (this, pl_inode); + if (count) { + gf_proc_dump_write("entrylk-count", "%d", count); + __dump_entrylks (pl_inode); + } - count = get_inodelk_count (this, inode); - if (count) { - gf_proc_dump_write("inodlk-count", "%d", count); - dump_inodelks(pl_inode); - } + count = __get_inodelk_count (this, pl_inode, NULL); + if (count) { + gf_proc_dump_write("inodelk-count", "%d", count); + __dump_inodelks (pl_inode); + } - count = get_posixlk_count (this, inode); - if (count) { - gf_proc_dump_write("posixlk-count", "%d", count); - dump_posixlks(pl_inode); + count = __get_posixlk_count (this, pl_inode); + if (count) { + gf_proc_dump_write("posixlk-count", "%d", count); + __dump_posixlks (pl_inode); + } } + pthread_mutex_unlock (&pl_inode->mutex); out: + GF_FREE (pathname); + + if (ret && inode) { + if (!section_added) + gf_proc_dump_add_section ("xlator.features.locks.%s." + "inode", this->name); + gf_proc_dump_write ("Unable to print lock state", "(Lock " + "acquisition failure) %s", + uuid_utoa (inode->gfid)); + } return ret; } @@ -1980,6 +2505,124 @@ mem_acct_init (xlator_t *this) return ret; } + +pl_ctx_t* +pl_ctx_get (client_t *client, xlator_t *xlator) +{ + void *tmp = NULL; + pl_ctx_t *ctx = NULL; + + client_ctx_get (client, xlator, &tmp); + + ctx = tmp; + + if (ctx != NULL) + goto out; + + ctx = GF_CALLOC (1, sizeof (pl_ctx_t), gf_locks_mt_posix_lock_t); + + if (ctx == NULL) + goto out; + + ctx->ltable = pl_lock_table_new(); + + if (ctx->ltable == NULL) { + GF_FREE (ctx); + ctx = NULL; + goto out; + } + + LOCK_INIT (&ctx->ltable_lock); + + if (client_ctx_set (client, xlator, ctx) != 0) { + LOCK_DESTROY (&ctx->ltable_lock); + GF_FREE (ctx->ltable); + GF_FREE (ctx); + ctx = NULL; + } +out: + return ctx; +} + +static void +ltable_delete_locks (struct _lock_table *ltable) +{ + struct _locker *locker = NULL; + struct _locker *tmp = NULL; + + list_for_each_entry_safe (locker, tmp, <able->inodelk_lockers, lockers) { + if (locker->fd) + pl_del_locker (ltable, locker->volume, &locker->loc, + locker->fd, &locker->owner, + GF_FOP_INODELK); + GF_FREE (locker->volume); + GF_FREE (locker); + } + + list_for_each_entry_safe (locker, tmp, <able->entrylk_lockers, lockers) { + if (locker->fd) + pl_del_locker (ltable, locker->volume, &locker->loc, + locker->fd, &locker->owner, + GF_FOP_ENTRYLK); + GF_FREE (locker->volume); + GF_FREE (locker); + } + GF_FREE (ltable); +} + + +static int32_t +destroy_cbk (xlator_t *this, client_t *client) +{ + void *tmp = NULL; + pl_ctx_t *locks_ctx = NULL; + + client_ctx_del (client, this, &tmp); + + if (tmp == NULL) + return 0 +; + locks_ctx = tmp; + if (locks_ctx->ltable) + ltable_delete_locks (locks_ctx->ltable); + + LOCK_DESTROY (&locks_ctx->ltable_lock); + GF_FREE (locks_ctx); + + return 0; +} + + +static int32_t +disconnect_cbk (xlator_t *this, client_t *client) +{ + int32_t ret = 0; + pl_ctx_t *locks_ctx = NULL; + struct _lock_table *ltable = NULL; + + locks_ctx = pl_ctx_get (client, this); + if (locks_ctx == NULL) { + gf_log (this->name, GF_LOG_INFO, "pl_ctx_get() failed"); + goto out; + } + + LOCK (&locks_ctx->ltable_lock); + { + if (locks_ctx->ltable) { + ltable = locks_ctx->ltable; + locks_ctx->ltable = pl_lock_table_new (); + } + } + UNLOCK (&locks_ctx->ltable_lock); + + if (ltable) + ltable_delete_locks (ltable); + +out: + return ret; +} + + int init (xlator_t *this) { @@ -2008,7 +2651,7 @@ init (xlator_t *this) gf_log (this->name, GF_LOG_CRITICAL, "'locks' translator is not loaded over a storage " "translator"); - goto out;; + goto out; } priv = GF_CALLOC (1, sizeof (*priv), @@ -2029,13 +2672,20 @@ init (xlator_t *this) } } + this->local_pool = mem_pool_new (pl_local_t, 32); + if (!this->local_pool) { + ret = -1; + gf_log (this->name, GF_LOG_ERROR, + "failed to create local_t's memory pool"); + goto out; + } + this->private = priv; ret = 0; out: if (ret) { - if (priv) - GF_FREE (priv); + GF_FREE (priv); } return ret; } @@ -2050,6 +2700,7 @@ fini (xlator_t *this) if (!priv) return 0; this->private = NULL; + GF_FREE (priv->brickname); GF_FREE (priv); return 0; @@ -2058,21 +2709,23 @@ fini (xlator_t *this) int pl_inodelk (call_frame_t *frame, xlator_t *this, - const char *volume, loc_t *loc, int32_t cmd, struct gf_flock *flock); + const char *volume, loc_t *loc, int32_t cmd, struct gf_flock *flock, + dict_t *xdata); int pl_finodelk (call_frame_t *frame, xlator_t *this, - const char *volume, fd_t *fd, int32_t cmd, struct gf_flock *flock); + const char *volume, fd_t *fd, int32_t cmd, struct gf_flock *flock, + dict_t *xdata); int pl_entrylk (call_frame_t *frame, xlator_t *this, const char *volume, loc_t *loc, const char *basename, - entrylk_cmd cmd, entrylk_type type); + entrylk_cmd cmd, entrylk_type type, dict_t *xdata); int pl_fentrylk (call_frame_t *frame, xlator_t *this, const char *volume, fd_t *fd, const char *basename, - entrylk_cmd cmd, entrylk_type type); + entrylk_cmd cmd, entrylk_type type, dict_t *xdata); struct xlator_fops fops = { .lookup = pl_lookup, @@ -2091,6 +2744,8 @@ struct xlator_fops fops = { .opendir = pl_opendir, .readdirp = pl_readdirp, .getxattr = pl_getxattr, + .fgetxattr = pl_fgetxattr, + .fsetxattr = pl_fsetxattr, }; struct xlator_dumpops dumpops = { @@ -2098,9 +2753,11 @@ struct xlator_dumpops dumpops = { }; struct xlator_cbks cbks = { - .forget = pl_forget, - .release = pl_release, - .releasedir = pl_releasedir, + .forget = pl_forget, + .release = pl_release, + .releasedir = pl_releasedir, + .client_destroy = destroy_cbk, + .client_disconnect = disconnect_cbk, }; diff --git a/xlators/features/locks/src/reservelk.c b/xlators/features/locks/src/reservelk.c index 7a75cdea1..11abd26d8 100644 --- a/xlators/features/locks/src/reservelk.c +++ b/xlators/features/locks/src/reservelk.c @@ -1,22 +1,12 @@ /* - Copyright (c) 2006-2011 Gluster, Inc. <http://www.gluster.com> - This file is part of GlusterFS. - - GlusterFS is free software; you can redistribute it and/or modify - it under the terms of the GNU General Public License as published - by the Free Software Foundation; either version 3 of the License, - or (at your option) any later version. - - GlusterFS is distributed in the hope that it will be useful, but - WITHOUT ANY WARRANTY; without even the implied warranty of - MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU - General Public License for more details. - - You should have received a copy of the GNU General Public License - along with this program. If not, see - <http://www.gnu.org/licenses/>. -*/ + 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 _CONFIG_H #define _CONFIG_H #include "config.h" @@ -299,7 +289,8 @@ grant_blocked_reserve_locks (xlator_t *this, pl_inode_t *pl_inode) lock->user_flock.l_start, lock->user_flock.l_len); - STACK_UNWIND_STRICT (lk, lock->frame, 0, 0, &lock->user_flock); + STACK_UNWIND_STRICT (lk, lock->frame, 0, 0, &lock->user_flock, + NULL); } } @@ -376,7 +367,9 @@ grant_blocked_lock_calls (xlator_t *this, pl_inode_t *pl_inode) pl_trace_out (this, lock->frame, fd, NULL, cmd, &lock->user_flock, -1, EAGAIN, NULL); pl_update_refkeeper (this, fd->inode); - STACK_UNWIND_STRICT (lk, lock->frame, -1, EAGAIN, &lock->user_flock); + STACK_UNWIND_STRICT (lk, lock->frame, -1, + EAGAIN, &lock->user_flock, + NULL); __destroy_lock (lock); } } |