diff options
Diffstat (limited to 'libglusterfs/src/fd.c')
| -rw-r--r-- | libglusterfs/src/fd.c | 1624 |
1 files changed, 1029 insertions, 595 deletions
diff --git a/libglusterfs/src/fd.c b/libglusterfs/src/fd.c index c6b060c61a5..62606e91164 100644 --- a/libglusterfs/src/fd.c +++ b/libglusterfs/src/fd.c @@ -1,769 +1,1203 @@ /* - Copyright (c) 2007-2009 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. */ -#include "fd.h" -#include "glusterfs.h" -#include "inode.h" -#include "dict.h" -#include "statedump.h" - - -#ifndef _CONFIG_H -#define _CONFIG_H -#include "config.h" -#endif +#include "glusterfs/fd.h" +#include <errno.h> // for EINVAL, errno, ENOMEM +#include <inttypes.h> // for PRIu64 +#include <stdint.h> // for UINT32_MAX +#include <string.h> // for NULL, memcpy, memset, size_t +#include "glusterfs/statedump.h" +static int +gf_fd_fdtable_expand(fdtable_t *fdtable, uint32_t nr); -static uint32_t -gf_fd_fdtable_expand (fdtable_t *fdtable, uint32_t nr); - -static fd_t * -_fd_ref (fd_t *fd); +fd_t * +__fd_ref(fd_t *fd); -/* - Allocate in memory chunks of power of 2 starting from 1024B - Assumes fdtable->lock is held -*/ -static inline uint32_t -gf_roundup_power_of_two (uint32_t nr) +static int +gf_fd_chain_fd_entries(fdentry_t *entries, uint32_t startidx, uint32_t endcount) { - uint32_t result = 1; + uint32_t i = 0; - if (nr < 0) { - gf_log ("server-protocol/fd", - GF_LOG_ERROR, - "Negative number passed"); - return -1; - } + if (!entries) { + gf_msg_callingfn("fd", GF_LOG_WARNING, EINVAL, LG_MSG_INVALID_ARG, + "!entries"); + return -1; + } + + /* Chain only till the second to last entry because we want to + * ensure that the last entry has GF_FDTABLE_END. + */ + for (i = startidx; i < (endcount - 1); i++) + entries[i].next_free = i + 1; - while (result <= nr) - result *= 2; + /* i has already been incremented up to the last entry. */ + entries[i].next_free = GF_FDTABLE_END; - return result; + return 0; } static int -gf_fd_chain_fd_entries (fdentry_t *entries, uint32_t startidx, - uint32_t endcount) +gf_fd_fdtable_expand(fdtable_t *fdtable, uint32_t nr) { - uint32_t i = 0; + fdentry_t *oldfds = NULL; + uint32_t oldmax_fds = -1; + int ret = -1; + + if (fdtable == NULL || nr > UINT32_MAX) { + gf_msg_callingfn("fd", GF_LOG_ERROR, EINVAL, LG_MSG_INVALID_ARG, + "invalid argument"); + ret = EINVAL; + goto out; + } + + nr /= (1024 / sizeof(fdentry_t)); + nr = gf_roundup_next_power_of_two(nr + 1); + nr *= (1024 / sizeof(fdentry_t)); + + oldfds = fdtable->fdentries; + oldmax_fds = fdtable->max_fds; + + fdtable->fdentries = GF_CALLOC(nr, sizeof(fdentry_t), + gf_common_mt_fdentry_t); + if (!fdtable->fdentries) { + ret = ENOMEM; + goto out; + } + fdtable->max_fds = nr; + + if (oldfds) { + uint32_t cpy = oldmax_fds * sizeof(fdentry_t); + memcpy(fdtable->fdentries, oldfds, cpy); + } + + gf_fd_chain_fd_entries(fdtable->fdentries, oldmax_fds, fdtable->max_fds); + + /* Now that expansion is done, we must update the fd list + * head pointer so that the fd allocation functions can continue + * using the expanded table. + */ + fdtable->first_free = oldmax_fds; + GF_FREE(oldfds); + ret = 0; +out: + return ret; +} - if (!entries) - return -1; +fdtable_t * +gf_fd_fdtable_alloc(void) +{ + fdtable_t *fdtable = NULL; - /* Chain only till the second to last entry because we want to - * ensure that the last entry has GF_FDTABLE_END. - */ - for (i = startidx; i < (endcount - 1); i++) - entries[i].next_free = i + 1; + fdtable = GF_CALLOC(1, sizeof(*fdtable), gf_common_mt_fdtable_t); + if (!fdtable) + return NULL; - /* i has already been incremented upto the last entry. */ - entries[i].next_free = GF_FDTABLE_END; + pthread_rwlock_init(&fdtable->lock, NULL); - return 0; -} + pthread_rwlock_wrlock(&fdtable->lock); + { + gf_fd_fdtable_expand(fdtable, 0); + } + pthread_rwlock_unlock(&fdtable->lock); + return fdtable; +} -static uint32_t -gf_fd_fdtable_expand (fdtable_t *fdtable, uint32_t nr) +static fdentry_t * +__gf_fd_fdtable_get_all_fds(fdtable_t *fdtable, uint32_t *count) { - fdentry_t *oldfds = NULL; - uint32_t oldmax_fds = -1; - - if (fdtable == NULL || nr < 0) - { - gf_log ("fd", GF_LOG_ERROR, "invalid argument"); - return EINVAL; - } - - nr /= (1024 / sizeof (fdentry_t)); - nr = gf_roundup_power_of_two (nr + 1); - nr *= (1024 / sizeof (fdentry_t)); - - oldfds = fdtable->fdentries; - oldmax_fds = fdtable->max_fds; + fdentry_t *fdentries = NULL; - fdtable->fdentries = CALLOC (nr, sizeof (fdentry_t)); - ERR_ABORT (fdtable->fdentries); - fdtable->max_fds = nr; + if (count == NULL) { + gf_msg_callingfn("fd", GF_LOG_WARNING, EINVAL, LG_MSG_INVALID_ARG, + "!count"); + goto out; + } - if (oldfds) { - uint32_t cpy = oldmax_fds * sizeof (fdentry_t); - memcpy (fdtable->fdentries, oldfds, cpy); - } + fdentries = fdtable->fdentries; + fdtable->fdentries = GF_CALLOC(fdtable->max_fds, sizeof(fdentry_t), + gf_common_mt_fdentry_t); + gf_fd_chain_fd_entries(fdtable->fdentries, 0, fdtable->max_fds); + *count = fdtable->max_fds; - gf_fd_chain_fd_entries (fdtable->fdentries, oldmax_fds, - fdtable->max_fds); - - /* Now that expansion is done, we must update the fd list - * head pointer so that the fd allocation functions can continue - * using the expanded table. - */ - fdtable->first_free = oldmax_fds; - FREE (oldfds); - return 0; +out: + return fdentries; } -fdtable_t * -gf_fd_fdtable_alloc (void) +fdentry_t * +gf_fd_fdtable_get_all_fds(fdtable_t *fdtable, uint32_t *count) { - fdtable_t *fdtable = NULL; + fdentry_t *entries = NULL; - fdtable = CALLOC (1, sizeof (*fdtable)); - if (!fdtable) - return NULL; + if (fdtable) { + pthread_rwlock_wrlock(&fdtable->lock); + { + entries = __gf_fd_fdtable_get_all_fds(fdtable, count); + } + pthread_rwlock_unlock(&fdtable->lock); + } - pthread_mutex_init (&fdtable->lock, NULL); + return entries; +} - pthread_mutex_lock (&fdtable->lock); - { - gf_fd_fdtable_expand (fdtable, 0); - } - pthread_mutex_unlock (&fdtable->lock); +static fdentry_t * +__gf_fd_fdtable_copy_all_fds(fdtable_t *fdtable, uint32_t *count) +{ + fdentry_t *fdentries = NULL; + int i = 0; + + if (count == NULL) { + gf_msg_callingfn("fd", GF_LOG_WARNING, EINVAL, LG_MSG_INVALID_ARG, + "!count"); + goto out; + } + + fdentries = GF_CALLOC(fdtable->max_fds, sizeof(fdentry_t), + gf_common_mt_fdentry_t); + if (fdentries == NULL) { + goto out; + } + + *count = fdtable->max_fds; + + for (i = 0; i < fdtable->max_fds; i++) { + if (fdtable->fdentries[i].fd != NULL) { + fdentries[i].fd = fd_ref(fdtable->fdentries[i].fd); + } + } - return fdtable; +out: + return fdentries; } fdentry_t * -__gf_fd_fdtable_get_all_fds (fdtable_t *fdtable, uint32_t *count) +gf_fd_fdtable_copy_all_fds(fdtable_t *fdtable, uint32_t *count) { - fdentry_t *fdentries = NULL; + fdentry_t *entries = NULL; - if (count == NULL) { - goto out; + if (fdtable) { + pthread_rwlock_rdlock(&fdtable->lock); + { + entries = __gf_fd_fdtable_copy_all_fds(fdtable, count); } + pthread_rwlock_unlock(&fdtable->lock); + } - fdentries = fdtable->fdentries; - fdtable->fdentries = calloc (fdtable->max_fds, sizeof (fdentry_t)); - gf_fd_chain_fd_entries (fdtable->fdentries, 0, fdtable->max_fds); - *count = fdtable->max_fds; - -out: - return fdentries; + return entries; } -fdentry_t * -gf_fd_fdtable_get_all_fds (fdtable_t *fdtable, uint32_t *count) +void +gf_fd_fdtable_destroy(fdtable_t *fdtable) { - fdentry_t *entries = NULL; - if (fdtable) { - pthread_mutex_lock (&fdtable->lock); - { - entries = __gf_fd_fdtable_get_all_fds (fdtable, count); - } - pthread_mutex_unlock (&fdtable->lock); + struct list_head list = { + 0, + }; + fd_t *fd = NULL; + fdentry_t *fdentries = NULL; + uint32_t fd_count = 0; + int32_t i = 0; + + INIT_LIST_HEAD(&list); + + if (!fdtable) { + gf_msg_callingfn("fd", GF_LOG_WARNING, EINVAL, LG_MSG_INVALID_ARG, + "!fdtable"); + return; + } + + pthread_rwlock_wrlock(&fdtable->lock); + { + fdentries = __gf_fd_fdtable_get_all_fds(fdtable, &fd_count); + GF_FREE(fdtable->fdentries); + } + pthread_rwlock_unlock(&fdtable->lock); + + if (fdentries != NULL) { + for (i = 0; i < fd_count; i++) { + fd = fdentries[i].fd; + if (fd != NULL) { + fd_unref(fd); + } } - return entries; + GF_FREE(fdentries); + pthread_rwlock_destroy(&fdtable->lock); + GF_FREE(fdtable); + } } -void -gf_fd_fdtable_destroy (fdtable_t *fdtable) +int +gf_fd_unused_get(fdtable_t *fdtable, fd_t *fdptr) { - struct list_head list = {0, }; - fd_t *fd = NULL; - fdentry_t *fdentries = NULL; - uint32_t fd_count = 0; - int32_t i = 0; + int32_t fd = -1; + fdentry_t *fde = NULL; + int error; + int alloc_attempts = 0; + + if (fdtable == NULL || fdptr == NULL) { + gf_msg_callingfn("fd", GF_LOG_ERROR, EINVAL, LG_MSG_INVALID_ARG, + "invalid argument"); + return EINVAL; + } + + pthread_rwlock_wrlock(&fdtable->lock); + { + fd_alloc_try_again: + if (fdtable->first_free != GF_FDTABLE_END) { + fde = &fdtable->fdentries[fdtable->first_free]; + fd = fdtable->first_free; + fdtable->first_free = fde->next_free; + fde->next_free = GF_FDENTRY_ALLOCATED; + fde->fd = fdptr; + } else { + /* If this is true, there is something + * seriously wrong with our data structures. + */ + if (alloc_attempts >= 2) { + gf_msg("fd", GF_LOG_ERROR, 0, LG_MSG_EXPAND_FD_TABLE_FAILED, + "multiple attempts to expand fd table" + " have failed."); + goto out; + } + error = gf_fd_fdtable_expand(fdtable, fdtable->max_fds + 1); + if (error) { + gf_msg("fd", GF_LOG_ERROR, error, LG_MSG_EXPAND_FD_TABLE_FAILED, + "Cannot expand fdtable"); + goto out; + } + ++alloc_attempts; + /* At this point, the table stands expanded + * with the first_free referring to the first + * free entry in the new set of fdentries that + * have just been allocated. That means, the + * above logic should just work. + */ + goto fd_alloc_try_again; + } + } +out: + pthread_rwlock_unlock(&fdtable->lock); - INIT_LIST_HEAD (&list); + return fd; +} - if (!fdtable) - return; +void +gf_fd_put(fdtable_t *fdtable, int32_t fd) +{ + fd_t *fdptr = NULL; + fdentry_t *fde = NULL; - pthread_mutex_lock (&fdtable->lock); - { - fdentries = __gf_fd_fdtable_get_all_fds (fdtable, &fd_count); - FREE (fdtable->fdentries); - } - pthread_mutex_unlock (&fdtable->lock); + if (fd == GF_ANON_FD_NO) + return; - if (fdentries != NULL) { - for (i = 0; i < fd_count; i++) { - fd = fdentries[i].fd; - if (fd != NULL) { - fd_unref (fd); - } - } + if (fdtable == NULL || fd < 0) { + gf_msg_callingfn("fd", GF_LOG_ERROR, EINVAL, LG_MSG_INVALID_ARG, + "invalid argument"); + return; + } - FREE (fdentries); - pthread_mutex_destroy (&fdtable->lock); - FREE (fdtable); - } -} - -int32_t -gf_fd_unused_get (fdtable_t *fdtable, fd_t *fdptr) -{ - int32_t fd = -1; - fdentry_t *fde = NULL; - int error; - int alloc_attempts = 0; - - if (fdtable == NULL || fdptr == NULL) - { - gf_log ("fd", GF_LOG_ERROR, "invalid argument"); - return EINVAL; - } - - pthread_mutex_lock (&fdtable->lock); - { -fd_alloc_try_again: - if (fdtable->first_free != GF_FDTABLE_END) { - fde = &fdtable->fdentries[fdtable->first_free]; - fd = fdtable->first_free; - fdtable->first_free = fde->next_free; - fde->next_free = GF_FDENTRY_ALLOCATED; - fde->fd = fdptr; - } else { - /* If this is true, there is something - * seriously wrong with our data structures. - */ - if (alloc_attempts >= 2) { - gf_log ("server-protocol.c", GF_LOG_ERROR, - "Multiple attempts to expand fd table" - " have failed."); - goto out; - } - error = gf_fd_fdtable_expand (fdtable, - fdtable->max_fds + 1); - if (error) { - gf_log ("server-protocol.c", - GF_LOG_ERROR, - "Cannot expand fdtable:%s", strerror (error)); - goto out; - } - ++alloc_attempts; - /* At this point, the table stands expanded - * with the first_free referring to the first - * free entry in the new set of fdentries that - * have just been allocated. That means, the - * above logic should just work. - */ - goto fd_alloc_try_again; - } - } -out: - pthread_mutex_unlock (&fdtable->lock); - - return fd; -} - - -inline void -gf_fd_put (fdtable_t *fdtable, int32_t fd) -{ - fd_t *fdptr = NULL; - fdentry_t *fde = NULL; - - if (fdtable == NULL || fd < 0) - { - gf_log ("fd", GF_LOG_ERROR, "invalid argument"); - return; - } - - if (!(fd < fdtable->max_fds)) - { - gf_log ("fd", GF_LOG_ERROR, "invalid argument"); - return; - } - - pthread_mutex_lock (&fdtable->lock); - { - fde = &fdtable->fdentries[fd]; - /* If the entry is not allocated, put operation must return - * without doing anything. - * This has the potential of masking out any bugs in a user of - * fd that ends up calling gf_fd_put twice for the same fd or - * for an unallocated fd, but thats a price we have to pay for - * ensuring sanity of our fd-table. - */ - if (fde->next_free != GF_FDENTRY_ALLOCATED) - goto unlock_out; - fdptr = fde->fd; - fde->fd = NULL; - fde->next_free = fdtable->first_free; - fdtable->first_free = fd; - } + if (!(fd < fdtable->max_fds)) { + gf_msg_callingfn("fd", GF_LOG_ERROR, EINVAL, LG_MSG_INVALID_ARG, + "invalid argument"); + return; + } + + pthread_rwlock_wrlock(&fdtable->lock); + { + fde = &fdtable->fdentries[fd]; + /* If the entry is not allocated, put operation must return + * without doing anything. + * This has the potential of masking out any bugs in a user of + * fd that ends up calling gf_fd_put twice for the same fd or + * for an unallocated fd, but it is a price we have to pay for + * ensuring sanity of our fd-table. + */ + if (fde->next_free != GF_FDENTRY_ALLOCATED) + goto unlock_out; + fdptr = fde->fd; + fde->fd = NULL; + fde->next_free = fdtable->first_free; + fdtable->first_free = fd; + } unlock_out: - pthread_mutex_unlock (&fdtable->lock); + pthread_rwlock_unlock(&fdtable->lock); - if (fdptr) { - fd_unref (fdptr); - } + if (fdptr) { + fd_unref(fdptr); + } } +void +gf_fdptr_put(fdtable_t *fdtable, fd_t *fd) +{ + fdentry_t *fde = NULL; + int32_t i = 0; -fd_t * -gf_fd_fdptr_get (fdtable_t *fdtable, int64_t fd) -{ - fd_t *fdptr = NULL; - - if (fdtable == NULL || fd < 0) - { - gf_log ("fd", GF_LOG_ERROR, "invalid argument"); - errno = EINVAL; - return NULL; - } - - if (!(fd < fdtable->max_fds)) - { - gf_log ("fd", GF_LOG_ERROR, "invalid argument"); - errno = EINVAL; - return NULL; - } - - pthread_mutex_lock (&fdtable->lock); - { - fdptr = fdtable->fdentries[fd].fd; - if (fdptr) { - fd_ref (fdptr); - } - } - pthread_mutex_unlock (&fdtable->lock); - - return fdptr; + if ((fdtable == NULL) || (fd == NULL)) { + gf_msg_callingfn("fd", GF_LOG_ERROR, EINVAL, LG_MSG_INVALID_ARG, + "invalid argument"); + return; + } + + pthread_rwlock_wrlock(&fdtable->lock); + { + for (i = 0; i < fdtable->max_fds; i++) { + if (fdtable->fdentries[i].fd == fd) { + fde = &fdtable->fdentries[i]; + break; + } + } + + if (fde == NULL) { + gf_msg_callingfn("fd", GF_LOG_WARNING, 0, + LG_MSG_FD_NOT_FOUND_IN_FDTABLE, + "fd (%p) is not present in fdtable", fd); + goto unlock_out; + } + + /* If the entry is not allocated, put operation must return + * without doing anything. + * This has the potential of masking out any bugs in a user of + * fd that ends up calling gf_fd_put twice for the same fd or + * for an unallocated fd, but it is a price we have to pay for + * ensuring sanity of our fd-table. + */ + if (fde->next_free != GF_FDENTRY_ALLOCATED) + goto unlock_out; + fde->fd = NULL; + fde->next_free = fdtable->first_free; + fdtable->first_free = i; + } +unlock_out: + pthread_rwlock_unlock(&fdtable->lock); + + if ((fd != NULL) && (fde != NULL)) { + fd_unref(fd); + } } fd_t * -_fd_ref (fd_t *fd) +gf_fd_fdptr_get(fdtable_t *fdtable, int64_t fd) { - ++fd->refcount; - - return fd; + fd_t *fdptr = NULL; + + if (fdtable == NULL || fd < 0) { + gf_msg_callingfn("fd", GF_LOG_ERROR, EINVAL, LG_MSG_INVALID_ARG, + "invalid argument"); + errno = EINVAL; + return NULL; + } + + if (!(fd < fdtable->max_fds)) { + gf_msg_callingfn("fd", GF_LOG_ERROR, EINVAL, LG_MSG_INVALID_ARG, + "invalid argument"); + errno = EINVAL; + return NULL; + } + + pthread_rwlock_rdlock(&fdtable->lock); + { + fdptr = fdtable->fdentries[fd].fd; + if (fdptr) { + fd_ref(fdptr); + } + } + pthread_rwlock_unlock(&fdtable->lock); + + return fdptr; } fd_t * -fd_ref (fd_t *fd) +__fd_ref(fd_t *fd) { - fd_t *refed_fd = NULL; - - if (!fd) { - gf_log ("fd", GF_LOG_ERROR, "@fd=%p", fd); - return NULL; - } + GF_ATOMIC_INC(fd->refcount); - LOCK (&fd->inode->lock); - refed_fd = _fd_ref (fd); - UNLOCK (&fd->inode->lock); - - return refed_fd; + return fd; } fd_t * -_fd_unref (fd_t *fd) +fd_ref(fd_t *fd) { - assert (fd->refcount); + if (!fd) { + gf_msg_callingfn("fd", GF_LOG_ERROR, EINVAL, LG_MSG_INVALID_ARG, + "null fd"); + return NULL; + } - --fd->refcount; + GF_ATOMIC_INC(fd->refcount); - if (fd->refcount == 0){ - list_del_init (&fd->inode_list); - } - - return fd; + return fd; } static void -fd_destroy (fd_t *fd) +fd_destroy(fd_t *fd, gf_boolean_t bound) { - xlator_t *xl = NULL; - int i = 0; - xlator_t *old_THIS = NULL; - - if (fd == NULL){ - gf_log ("xlator", GF_LOG_ERROR, "invalid arugument"); - goto out; + xlator_t *xl = NULL; + int i = 0; + xlator_t *old_THIS = NULL; + + if (fd == NULL) { + gf_msg_callingfn("xlator", GF_LOG_ERROR, EINVAL, LG_MSG_INVALID_ARG, + "invalid argument"); + goto out; + } + + if (fd->inode == NULL) { + gf_msg_callingfn("xlator", GF_LOG_ERROR, 0, LG_MSG_FD_INODE_NULL, + "fd->inode is NULL"); + goto out; + } + if (!fd->_ctx) + goto out; + + if (IA_ISDIR(fd->inode->ia_type)) { + for (i = 0; i < fd->xl_count; i++) { + if (fd->_ctx[i].key) { + xl = fd->_ctx[i].xl_key; + old_THIS = THIS; + THIS = xl; + if (!xl->call_cleanup && xl->cbks->releasedir) + xl->cbks->releasedir(xl, fd); + THIS = old_THIS; + } } - - if (fd->inode == NULL){ - gf_log ("xlator", GF_LOG_ERROR, "fd->inode is NULL"); - goto out; + } else { + for (i = 0; i < fd->xl_count; i++) { + if (fd->_ctx[i].key) { + xl = fd->_ctx[i].xl_key; + old_THIS = THIS; + THIS = xl; + if (!xl->call_cleanup && xl->cbks->release) + xl->cbks->release(xl, fd); + THIS = old_THIS; + } } - if (!fd->_ctx) - goto out; - - if (S_ISDIR (fd->inode->st_mode)) { - for (i = 0; i < fd->inode->table->xl->ctx->xl_count; i++) { - if (fd->_ctx[i].key) { - xl = (xlator_t *)(long)fd->_ctx[i].key; - old_THIS = THIS; - THIS = xl; - if (xl->cbks->releasedir) - xl->cbks->releasedir (xl, fd); - THIS = old_THIS; - } - } - } else { - for (i = 0; i < fd->inode->table->xl->ctx->xl_count; i++) { - if (fd->_ctx[i].key) { - xl = (xlator_t *)(long)fd->_ctx[i].key; - old_THIS = THIS; - THIS = xl; - if (xl->cbks->release) - xl->cbks->release (xl, fd); - THIS = old_THIS; - } - } + } + + LOCK_DESTROY(&fd->lock); + + GF_FREE(fd->_ctx); + if (bound) { + /*Decrease the count only after close happens on file*/ + LOCK(&fd->inode->lock); + { + fd->inode->fd_count--; } - - LOCK_DESTROY (&fd->lock); - - FREE (fd->_ctx); - inode_unref (fd->inode); - fd->inode = (inode_t *)0xaaaaaaaa; - FREE (fd); - + UNLOCK(&fd->inode->lock); + } + inode_unref(fd->inode); + fd->inode = NULL; + fd_lk_ctx_unref(fd->lk_ctx); + mem_put(fd); out: - return; + return; } void -fd_unref (fd_t *fd) +fd_close(fd_t *fd) { - int32_t refcount = 0; + xlator_t *xl, *old_THIS; - if (!fd) { - gf_log ("fd.c", GF_LOG_ERROR, "fd is NULL"); - return; - } - - LOCK (&fd->inode->lock); - { - _fd_unref (fd); - refcount = fd->refcount; + old_THIS = THIS; + + for (xl = fd->inode->table->xl->graph->first; xl != NULL; xl = xl->next) { + if (!xl->call_cleanup) { + THIS = xl; + + if (IA_ISDIR(fd->inode->ia_type)) { + if (xl->cbks->fdclosedir != NULL) { + xl->cbks->fdclosedir(xl, fd); + } + } else { + if (xl->cbks->fdclose != NULL) { + xl->cbks->fdclose(xl, fd); + } + } } - UNLOCK (&fd->inode->lock); - + } + + THIS = old_THIS; +} + +void +fd_unref(fd_t *fd) +{ + int32_t refcount = 0; + gf_boolean_t bound = _gf_false; + + if (!fd) { + gf_msg_callingfn("fd", GF_LOG_ERROR, EINVAL, LG_MSG_INVALID_ARG, + "fd is NULL"); + return; + } + + LOCK(&fd->inode->lock); + { + refcount = GF_ATOMIC_DEC(fd->refcount); if (refcount == 0) { - fd_destroy (fd); + if (!list_empty(&fd->inode_list)) { + list_del_init(&fd->inode_list); + fd->inode->active_fd_count--; + bound = _gf_true; + } } + } + UNLOCK(&fd->inode->lock); - return ; + if (refcount == 0) { + fd_destroy(fd, bound); + } + + return; +} + +static fd_t * +__fd_bind(fd_t *fd) +{ + list_del_init(&fd->inode_list); + list_add(&fd->inode_list, &fd->inode->fd_list); + fd->inode->fd_count++; + fd->inode->active_fd_count++; + + return fd; +} + +fd_t * +fd_bind(fd_t *fd) +{ + if (!fd || !fd->inode) { + gf_msg_callingfn("fd", GF_LOG_ERROR, EINVAL, LG_MSG_INVALID_ARG, + "!fd || !fd->inode"); + return NULL; + } + + LOCK(&fd->inode->lock); + { + fd = __fd_bind(fd); + } + UNLOCK(&fd->inode->lock); + + return fd; +} + +static fd_t * +fd_allocate(inode_t *inode, uint64_t pid) +{ + fd_t *fd; + + if (inode == NULL) { + gf_msg_callingfn("fd", GF_LOG_ERROR, EINVAL, LG_MSG_INVALID_ARG, + "invalid argument"); + return NULL; + } + + fd = mem_get0(inode->table->fd_mem_pool); + if (fd == NULL) { + return NULL; + } + + fd->xl_count = inode->table->xl->graph->xl_count + 1; + + fd->_ctx = GF_CALLOC(1, (sizeof(struct _fd_ctx) * fd->xl_count), + gf_common_mt_fd_ctx); + if (fd->_ctx == NULL) { + goto failed; + } + + fd->lk_ctx = fd_lk_ctx_create(); + if (fd->lk_ctx != NULL) { + /* We need to take a reference from the inode, but we cannot do it + * here because this function can be called with the inode lock taken + * and inode_ref() takes the inode's table lock. This is the reverse + * of the logical lock acquisition order and can cause a deadlock. So + * we simply assign the inode here and we delefate the inode reference + * responsibility to the caller (when this function succeeds and the + * inode lock is released). This is safe because the caller must hold + * a reference of the inode to use it, so it's guaranteed that the + * number of references won't reach 0 before the caller finishes. + * + * TODO: minimize use of locks in favor of atomic operations to avoid + * these dependencies. */ + fd->inode = inode; + fd->pid = pid; + INIT_LIST_HEAD(&fd->inode_list); + LOCK_INIT(&fd->lock); + GF_ATOMIC_INIT(fd->refcount, 1); + return fd; + } + + GF_FREE(fd->_ctx); + +failed: + mem_put(fd); + + return NULL; +} + +fd_t * +fd_create_uint64(inode_t *inode, uint64_t pid) +{ + fd_t *fd; + + fd = fd_allocate(inode, pid); + if (fd != NULL) { + /* fd_allocate() doesn't get a reference from the inode. We need to + * take it here in case of success. */ + inode_ref(inode); + } + + return fd; } fd_t * -fd_bind (fd_t *fd) +fd_create(inode_t *inode, pid_t pid) { - inode_t *inode = fd->inode; + return fd_create_uint64(inode, (uint64_t)pid); +} - if (!fd) { - gf_log ("fd.c", GF_LOG_ERROR, "fd is NULL"); - return NULL; +static fd_t * +__fd_lookup(inode_t *inode, uint64_t pid) +{ + fd_t *iter_fd = NULL; + fd_t *fd = NULL; + + if (list_empty(&inode->fd_list)) + return NULL; + + list_for_each_entry(iter_fd, &inode->fd_list, inode_list) + { + if (iter_fd->anonymous) + /* If someone was interested in getting an + anonymous fd (or was OK getting an anonymous fd), + they can as well call fd_anonymous() directly */ + continue; + + if (!pid || iter_fd->pid == pid) { + fd = __fd_ref(iter_fd); + break; } + } - LOCK (&inode->lock); - { - list_add (&fd->inode_list, &inode->fd_list); + return fd; +} + +fd_t * +fd_lookup(inode_t *inode, pid_t pid) +{ + fd_t *fd = NULL; + + if (!inode) { + gf_msg_callingfn("fd", GF_LOG_WARNING, EINVAL, LG_MSG_INVALID_ARG, + "!inode"); + return NULL; + } + + LOCK(&inode->lock); + { + fd = __fd_lookup(inode, (uint64_t)pid); + } + UNLOCK(&inode->lock); + + return fd; +} + +fd_t * +fd_lookup_uint64(inode_t *inode, uint64_t pid) +{ + fd_t *fd = NULL; + + if (!inode) { + gf_msg_callingfn("fd", GF_LOG_WARNING, EINVAL, LG_MSG_INVALID_ARG, + "!inode"); + return NULL; + } + + LOCK(&inode->lock); + { + fd = __fd_lookup(inode, pid); + } + UNLOCK(&inode->lock); + + return fd; +} + +static fd_t * +__fd_lookup_anonymous(inode_t *inode, int32_t flags) +{ + fd_t *iter_fd = NULL; + fd_t *fd = NULL; + + if (list_empty(&inode->fd_list)) + return NULL; + + list_for_each_entry(iter_fd, &inode->fd_list, inode_list) + { + if ((iter_fd->anonymous) && (flags == iter_fd->flags)) { + fd = __fd_ref(iter_fd); + break; } - UNLOCK (&inode->lock); - - return fd; + } + + return fd; } fd_t * -fd_create (inode_t *inode, pid_t pid) +fd_anonymous_with_flags(inode_t *inode, int32_t flags) { - fd_t *fd = NULL; - - if (inode == NULL) { - gf_log ("fd", GF_LOG_ERROR, "invalid argument"); - return NULL; + fd_t *fd = NULL; + bool ref = false; + + LOCK(&inode->lock); + + fd = __fd_lookup_anonymous(inode, flags); + + /* if (fd); then we already have increased the refcount in + __fd_lookup_anonymous(), so no need of one more fd_ref(). + if (!fd); then both create and bind won't bump up the ref + count, so we have to call fd_ref() after bind. */ + if (fd == NULL) { + fd = fd_allocate(inode, 0); + if (fd != NULL) { + fd->anonymous = _gf_true; + fd->flags = GF_ANON_FD_FLAGS | (flags & O_DIRECT); + + __fd_bind(fd); + + ref = true; } - - fd = CALLOC (1, sizeof (fd_t)); - ERR_ABORT (fd); - - fd->_ctx = CALLOC (1, (sizeof (struct _fd_ctx) * - inode->table->xl->ctx->xl_count)); - fd->inode = inode_ref (inode); - fd->pid = pid; - INIT_LIST_HEAD (&fd->inode_list); - - LOCK_INIT (&fd->lock); + } - LOCK (&inode->lock); - fd = _fd_ref (fd); - UNLOCK (&inode->lock); + UNLOCK(&inode->lock); - return fd; + if (ref) { + /* fd_allocate() doesn't get a reference from the inode. We need to + * take it here in case of success. */ + inode_ref(inode); + } + + return fd; } fd_t * -fd_lookup (inode_t *inode, pid_t pid) +fd_anonymous(inode_t *inode) { - fd_t *fd = NULL; - fd_t *iter_fd = NULL; + return fd_anonymous_with_flags(inode, 0); +} - LOCK (&inode->lock); - { - if (list_empty (&inode->fd_list)) { - fd = NULL; - } else { - list_for_each_entry (iter_fd, &inode->fd_list, inode_list) { - if (pid) { - if (iter_fd->pid == pid) { - fd = _fd_ref (iter_fd); - break; - } - } else { - fd = _fd_ref (iter_fd); - break; - } - } - } - } - UNLOCK (&inode->lock); - - return fd; +fd_t * +fd_lookup_anonymous(inode_t *inode, int32_t flags) +{ + fd_t *fd = NULL; + + if (!inode) { + gf_msg_callingfn("fd", GF_LOG_WARNING, EINVAL, LG_MSG_INVALID_ARG, + "!inode"); + return NULL; + } + + LOCK(&inode->lock); + { + fd = __fd_lookup_anonymous(inode, flags); + } + UNLOCK(&inode->lock); + return fd; +} + +gf_boolean_t +fd_is_anonymous(fd_t *fd) +{ + return (fd && fd->anonymous); } uint8_t -fd_list_empty (inode_t *inode) +fd_list_empty(inode_t *inode) { - uint8_t empty = 0; + uint8_t empty = 0; - LOCK (&inode->lock); - { - empty = list_empty (&inode->fd_list); - } - UNLOCK (&inode->lock); - - return empty; + LOCK(&inode->lock); + { + empty = list_empty(&inode->fd_list); + } + UNLOCK(&inode->lock); + + return empty; } int -__fd_ctx_set (fd_t *fd, xlator_t *xlator, uint64_t value) -{ - int index = 0; - int ret = 0; - int set_idx = -1; - - if (!fd || !xlator) - return -1; - - for (index = 0; index < xlator->ctx->xl_count; index++) { - if (!fd->_ctx[index].key) { - if (set_idx == -1) - set_idx = index; - /* dont break, to check if key already exists - further on */ - } - if (fd->_ctx[index].key == (uint64_t)(long) xlator) { - set_idx = index; - break; - } +__fd_ctx_set(fd_t *fd, xlator_t *xlator, uint64_t value) +{ + int index = 0, new_xl_count = 0; + int ret = 0; + int set_idx = -1; + void *begin = NULL; + size_t diff = 0; + struct _fd_ctx *tmp = NULL; + + if (!fd || !xlator) + return -1; + + for (index = 0; index < fd->xl_count; index++) { + if (!fd->_ctx[index].key) { + if (set_idx == -1) + set_idx = index; + /* don't break, to check if key already exists + further on */ } - - if (set_idx == -1) { - ret = -1; - goto out; + if (fd->_ctx[index].xl_key == xlator) { + set_idx = index; + break; + } + } + + if (set_idx == -1) { + set_idx = fd->xl_count; + + new_xl_count = fd->xl_count + xlator->graph->xl_count; + + tmp = GF_REALLOC(fd->_ctx, (sizeof(struct _fd_ctx) * new_xl_count)); + if (tmp == NULL) { + ret = -1; + goto out; } - - fd->_ctx[set_idx].key = (uint64_t)(long) xlator; - fd->_ctx[set_idx].value = value; + + fd->_ctx = tmp; + + begin = fd->_ctx; + begin += (fd->xl_count * sizeof(struct _fd_ctx)); + + diff = (new_xl_count - fd->xl_count) * sizeof(struct _fd_ctx); + + memset(begin, 0, diff); + + fd->xl_count = new_xl_count; + } + + fd->_ctx[set_idx].xl_key = xlator; + fd->_ctx[set_idx].value1 = value; out: - return ret; + return ret; } - int -fd_ctx_set (fd_t *fd, xlator_t *xlator, uint64_t value) +fd_ctx_set(fd_t *fd, xlator_t *xlator, uint64_t value) { - int ret = 0; + int ret = 0; - if (!fd || !xlator) - return -1; - - LOCK (&fd->lock); - { - ret = __fd_ctx_set (fd, xlator, value); - } - UNLOCK (&fd->lock); + if (!fd || !xlator) { + gf_msg_callingfn("fd", GF_LOG_WARNING, EINVAL, LG_MSG_INVALID_ARG, + "%p %p", fd, xlator); + return -1; + } - return ret; -} + LOCK(&fd->lock); + { + ret = __fd_ctx_set(fd, xlator, value); + } + UNLOCK(&fd->lock); + return ret; +} -int -__fd_ctx_get (fd_t *fd, xlator_t *xlator, uint64_t *value) +int +__fd_ctx_get(fd_t *fd, xlator_t *xlator, uint64_t *value) { - int index = 0; - int ret = 0; + int index = 0; + int ret = 0; - if (!fd || !xlator) - return -1; - - for (index = 0; index < xlator->ctx->xl_count; index++) { - if (fd->_ctx[index].key == (uint64_t)(long)xlator) - break; - } - - if (index == xlator->ctx->xl_count) { - ret = -1; - goto out; - } + if (!fd || !xlator) + return -1; + + for (index = 0; index < fd->xl_count; index++) { + if (fd->_ctx[index].xl_key == xlator) + break; + } + + if (index == fd->xl_count) { + ret = -1; + goto out; + } + + if (value) + *value = fd->_ctx[index].value1; - if (value) - *value = fd->_ctx[index].value; - out: - return ret; + return ret; } - -int -fd_ctx_get (fd_t *fd, xlator_t *xlator, uint64_t *value) +int +fd_ctx_get(fd_t *fd, xlator_t *xlator, uint64_t *value) { - int ret = 0; + int ret = 0; - if (!fd || !xlator) - return -1; + if (!fd || !xlator) + return -1; - LOCK (&fd->lock); - { - ret = __fd_ctx_get (fd, xlator, value); - } - UNLOCK (&fd->lock); + LOCK(&fd->lock); + { + ret = __fd_ctx_get(fd, xlator, value); + } + UNLOCK(&fd->lock); - return ret; + return ret; } - -int -__fd_ctx_del (fd_t *fd, xlator_t *xlator, uint64_t *value) +int +__fd_ctx_del(fd_t *fd, xlator_t *xlator, uint64_t *value) { - int index = 0; - int ret = 0; + int index = 0; + int ret = 0; - if (!fd || !xlator) - return -1; - - for (index = 0; index < xlator->ctx->xl_count; index++) { - if (fd->_ctx[index].key == (uint64_t)(long)xlator) - break; - } - - if (index == xlator->ctx->xl_count) { - ret = -1; - goto out; - } - - if (value) - *value = fd->_ctx[index].value; - - fd->_ctx[index].key = 0; - fd->_ctx[index].value = 0; + if (!fd || !xlator) + return -1; + + for (index = 0; index < fd->xl_count; index++) { + if (fd->_ctx[index].xl_key == xlator) + break; + } + + if (index == fd->xl_count) { + ret = -1; + goto out; + } + + if (value) + *value = fd->_ctx[index].value1; + + fd->_ctx[index].key = 0; + fd->_ctx[index].value1 = 0; out: - return ret; + return ret; } +int +fd_ctx_del(fd_t *fd, xlator_t *xlator, uint64_t *value) +{ + int ret = 0; + + if (!fd || !xlator) + return -1; -int -fd_ctx_del (fd_t *fd, xlator_t *xlator, uint64_t *value) + LOCK(&fd->lock); + { + ret = __fd_ctx_del(fd, xlator, value); + } + UNLOCK(&fd->lock); + + return ret; +} + +void +fd_dump(fd_t *fd, char *prefix) { - int ret = 0; + char key[GF_DUMP_MAX_BUF_LEN]; - if (!fd || !xlator) - return -1; - - LOCK (&fd->lock); - { - ret = __fd_ctx_del (fd, xlator, value); - } - UNLOCK (&fd->lock); + if (!fd) + return; + + gf_proc_dump_write("pid", "%" PRIu64, fd->pid); + gf_proc_dump_write("refcount", "%" GF_PRI_ATOMIC, + GF_ATOMIC_GET(fd->refcount)); + gf_proc_dump_write("flags", "%d", fd->flags); - return ret; + if (fd->inode) { + gf_proc_dump_build_key(key, "inode", NULL); + gf_proc_dump_add_section("%s", key); + inode_dump(fd->inode, key); + } } +void +fdentry_dump(fdentry_t *fdentry, char *prefix) +{ + if (!fdentry) + return; + + if (GF_FDENTRY_ALLOCATED != fdentry->next_free) + return; + + if (fdentry->fd) + fd_dump(fdentry->fd, prefix); +} void -fd_dump (fd_t *fd, char *prefix) -{ - char key[GF_DUMP_MAX_BUF_LEN]; - - if (!fd) - return; - - memset(key, 0, sizeof(key)); - gf_proc_dump_build_key(key, prefix, "pid"); - gf_proc_dump_write(key, "%d", fd->pid); - gf_proc_dump_build_key(key, prefix, "refcount"); - gf_proc_dump_write(key, "%d", fd->refcount); - gf_proc_dump_build_key(key, prefix, "flags"); - gf_proc_dump_write(key, "%d", fd->flags); - if (fd->inode) { - gf_proc_dump_build_key(key, prefix, "inode"); - gf_proc_dump_write(key, "%ld", fd->inode->ino); +fdtable_dump(fdtable_t *fdtable, char *prefix) +{ + char key[GF_DUMP_MAX_BUF_LEN]; + int i = 0; + int ret = -1; + + if (!fdtable) + return; + + ret = pthread_rwlock_tryrdlock(&fdtable->lock); + if (ret) + goto out; + + gf_proc_dump_build_key(key, prefix, "refcount"); + gf_proc_dump_write(key, "%d", fdtable->refcount); + gf_proc_dump_build_key(key, prefix, "maxfds"); + gf_proc_dump_write(key, "%d", fdtable->max_fds); + gf_proc_dump_build_key(key, prefix, "first_free"); + gf_proc_dump_write(key, "%d", fdtable->first_free); + + for (i = 0; i < fdtable->max_fds; i++) { + if (GF_FDENTRY_ALLOCATED == fdtable->fdentries[i].next_free) { + gf_proc_dump_build_key(key, prefix, "fdentry[%d]", i); + gf_proc_dump_add_section("%s", key); + fdentry_dump(&fdtable->fdentries[i], key); } + } + + pthread_rwlock_unlock(&fdtable->lock); + +out: + if (ret != 0) + gf_proc_dump_write("Unable to dump the fdtable", + "(Lock acquistion failed) %p", fdtable); + return; } +void +fd_ctx_dump(fd_t *fd, char *prefix) +{ + struct _fd_ctx *fd_ctx = NULL; + xlator_t *xl = NULL; + int i = 0; + + if ((fd == NULL) || (fd->_ctx == NULL)) { + goto out; + } + + LOCK(&fd->lock); + { + if (fd->_ctx != NULL) { + fd_ctx = GF_CALLOC(fd->xl_count, sizeof(*fd_ctx), + gf_common_mt_fd_ctx); + if (fd_ctx == NULL) { + goto unlock; + } + + for (i = 0; i < fd->xl_count; i++) { + fd_ctx[i] = fd->_ctx[i]; + } + } + } +unlock: + UNLOCK(&fd->lock); + + if (fd_ctx == NULL) { + goto out; + } + + for (i = 0; i < fd->xl_count; i++) { + if (fd_ctx[i].xl_key) { + xl = (xlator_t *)(long)fd_ctx[i].xl_key; + if (xl->dumpops && xl->dumpops->fdctx) + xl->dumpops->fdctx(xl, fd); + } + } + +out: + GF_FREE(fd_ctx); + + return; +} void -fdentry_dump (fdentry_t *fdentry, char *prefix) +fdentry_dump_to_dict(fdentry_t *fdentry, char *prefix, dict_t *dict, + int *openfds) { - if (!fdentry) - return; + char key[GF_DUMP_MAX_BUF_LEN] = { + 0, + }; + int ret = -1; - if (GF_FDENTRY_ALLOCATED != fdentry->next_free) - return; + if (!fdentry) + return; + if (!dict) + return; + + if (GF_FDENTRY_ALLOCATED != fdentry->next_free) + return; - if (fdentry->fd) - fd_dump(fdentry->fd, prefix); + if (fdentry->fd) { + snprintf(key, sizeof(key), "%s.pid", prefix); + ret = dict_set_uint64(dict, key, fdentry->fd->pid); + if (ret) + return; + + snprintf(key, sizeof(key), "%s.refcount", prefix); + ret = dict_set_int32(dict, key, GF_ATOMIC_GET(fdentry->fd->refcount)); + if (ret) + return; + + snprintf(key, sizeof(key), "%s.flags", prefix); + ret = dict_set_int32(dict, key, fdentry->fd->flags); + if (ret) + return; + + (*openfds)++; + } + return; } void -fdtable_dump (fdtable_t *fdtable, char *prefix) +fdtable_dump_to_dict(fdtable_t *fdtable, char *prefix, dict_t *dict) { - char key[GF_DUMP_MAX_BUF_LEN]; - int i = 0; - int ret = -1; + char key[GF_DUMP_MAX_BUF_LEN] = { + 0, + }; + int i = 0; + int openfds = 0; + int ret = -1; + + if (!fdtable) + return; + if (!dict) + return; - if (!fdtable) - return; - - ret = pthread_mutex_trylock (&fdtable->lock); + ret = pthread_rwlock_tryrdlock(&fdtable->lock); + if (ret) + return; - if (ret) { - gf_log ("fd", GF_LOG_WARNING, "Unable to acquire lock"); - return; + snprintf(key, sizeof(key), "%s.fdtable.refcount", prefix); + ret = dict_set_int32(dict, key, fdtable->refcount); + if (ret) + goto out; + + snprintf(key, sizeof(key), "%s.fdtable.maxfds", prefix); + ret = dict_set_uint32(dict, key, fdtable->max_fds); + if (ret) + goto out; + + snprintf(key, sizeof(key), "%s.fdtable.firstfree", prefix); + ret = dict_set_int32(dict, key, fdtable->first_free); + if (ret) + goto out; + + for (i = 0; i < fdtable->max_fds; i++) { + if (GF_FDENTRY_ALLOCATED == fdtable->fdentries[i].next_free) { + snprintf(key, sizeof(key), "%s.fdtable.fdentry%d", prefix, i); + fdentry_dump_to_dict(&fdtable->fdentries[i], key, dict, &openfds); } + } - memset(key, 0, sizeof(key)); - gf_proc_dump_build_key(key, prefix, "refcount"); - gf_proc_dump_write(key, "%d", fdtable->refcount); - gf_proc_dump_build_key(key, prefix, "maxfds"); - gf_proc_dump_write(key, "%d", fdtable->max_fds); - gf_proc_dump_build_key(key, prefix, "first_free"); - gf_proc_dump_write(key, "%d", fdtable->first_free); - - for ( i = 0 ; i < fdtable->max_fds; i++) { - if (GF_FDENTRY_ALLOCATED == - fdtable->fdentries[i].next_free) { - gf_proc_dump_build_key(key, prefix, "fdentry[%d]", i); - gf_proc_dump_add_section(key); - fdentry_dump(&fdtable->fdentries[i], key); - } - } + snprintf(key, sizeof(key), "%s.fdtable.openfds", prefix); + ret = dict_set_int32(dict, key, openfds); + if (ret) + goto out; - pthread_mutex_unlock(&fdtable->lock); +out: + pthread_rwlock_unlock(&fdtable->lock); + return; } - |