From 34d024f84345807bf44163fac84e921513dde323 Mon Sep 17 00:00:00 2001 From: Mark Fasheh Date: Mon, 24 Sep 2007 15:56:19 -0700 Subject: ocfs2: Remove mount/unmount votes The node maps that are set/unset by these votes are no longer relevant, thus we can remove the mount and umount votes. Since those are the last two remaining votes, we can also remove the entire vote infrastructure. The vote thread has been renamed to the downconvert thread, and the small amount of functionality related to managing it has been moved into fs/ocfs2/dlmglue.c. All references to votes have been removed or updated. Signed-off-by: Mark Fasheh --- fs/ocfs2/dlmglue.h | 5 +++-- 1 file changed, 3 insertions(+), 2 deletions(-) (limited to 'fs/ocfs2/dlmglue.h') diff --git a/fs/ocfs2/dlmglue.h b/fs/ocfs2/dlmglue.h index 87a785e41205..931f6ee55146 100644 --- a/fs/ocfs2/dlmglue.h +++ b/fs/ocfs2/dlmglue.h @@ -54,7 +54,7 @@ struct ocfs2_meta_lvb { #define OCFS2_META_LOCK_RECOVERY (0x01) /* Instruct the dlm not to queue ourselves on the other node. */ #define OCFS2_META_LOCK_NOQUEUE (0x02) -/* don't block waiting for the vote thread, instead return -EAGAIN */ +/* don't block waiting for the downconvert thread, instead return -EAGAIN */ #define OCFS2_LOCK_NONBLOCK (0x04) int ocfs2_dlm_init(struct ocfs2_super *osb); @@ -112,9 +112,10 @@ void ocfs2_mark_lockres_freeing(struct ocfs2_lock_res *lockres); void ocfs2_simple_drop_lockres(struct ocfs2_super *osb, struct ocfs2_lock_res *lockres); -/* for the vote thread */ +/* for the downconvert thread */ void ocfs2_process_blocked_lock(struct ocfs2_super *osb, struct ocfs2_lock_res *lockres); +void ocfs2_wake_downconvert_thread(struct ocfs2_super *osb); struct ocfs2_dlm_debug *ocfs2_new_dlm_debug(void); void ocfs2_put_dlm_debug(struct ocfs2_dlm_debug *dlm_debug); -- cgit v1.2.3 From c934a92d05b549dd2f25db72c5fc3cb9dcf1b611 Mon Sep 17 00:00:00 2001 From: Mark Fasheh Date: Thu, 18 Oct 2007 15:23:46 -0700 Subject: ocfs2: Remove data locks The meta lock now covers both meta data and data, so this just removes the now-redundant data lock. Combining locks saves us a round of lock mastery per inode and one less lock to ping between nodes during read/write. We don't lose much - since meta locks were always held before a data lock (and at the same level) ordered writeout mode (the default) ensured that flushing for the meta data lock also pushed out data anyways. Signed-off-by: Mark Fasheh --- fs/ocfs2/aops.c | 44 +---------------- fs/ocfs2/cluster/tcp_internal.h | 5 +- fs/ocfs2/dlmglue.c | 104 ---------------------------------------- fs/ocfs2/dlmglue.h | 11 +---- fs/ocfs2/file.c | 55 ++++++--------------- fs/ocfs2/inode.c | 6 --- fs/ocfs2/inode.h | 1 - fs/ocfs2/mmap.c | 9 ---- fs/ocfs2/super.c | 1 - 9 files changed, 22 insertions(+), 214 deletions(-) (limited to 'fs/ocfs2/dlmglue.h') diff --git a/fs/ocfs2/aops.c b/fs/ocfs2/aops.c index 56f7790cad46..5fc27cfaee50 100644 --- a/fs/ocfs2/aops.c +++ b/fs/ocfs2/aops.c @@ -305,21 +305,12 @@ static int ocfs2_readpage(struct file *file, struct page *page) goto out_alloc; } - ret = ocfs2_data_lock_with_page(inode, 0, page); - if (ret != 0) { - if (ret == AOP_TRUNCATED_PAGE) - unlock = 0; - mlog_errno(ret); - goto out_alloc; - } - if (oi->ip_dyn_features & OCFS2_INLINE_DATA_FL) ret = ocfs2_readpage_inline(inode, page); else ret = block_read_full_page(page, ocfs2_get_block); unlock = 0; - ocfs2_data_unlock(inode, 0); out_alloc: up_read(&OCFS2_I(inode)->ip_alloc_sem); out_meta_unlock: @@ -638,34 +629,12 @@ static ssize_t ocfs2_direct_IO(int rw, if (OCFS2_I(inode)->ip_dyn_features & OCFS2_INLINE_DATA_FL) return 0; - if (!ocfs2_sparse_alloc(OCFS2_SB(inode->i_sb))) { - /* - * We get PR data locks even for O_DIRECT. This - * allows concurrent O_DIRECT I/O but doesn't let - * O_DIRECT with extending and buffered zeroing writes - * race. If they did race then the buffered zeroing - * could be written back after the O_DIRECT I/O. It's - * one thing to tell people not to mix buffered and - * O_DIRECT writes, but expecting them to understand - * that file extension is also an implicit buffered - * write is too much. By getting the PR we force - * writeback of the buffered zeroing before - * proceeding. - */ - ret = ocfs2_data_lock(inode, 0); - if (ret < 0) { - mlog_errno(ret); - goto out; - } - ocfs2_data_unlock(inode, 0); - } - ret = blockdev_direct_IO_no_locking(rw, iocb, inode, inode->i_sb->s_bdev, iov, offset, nr_segs, ocfs2_direct_IO_get_blocks, ocfs2_dio_end_io); -out: + mlog_exit(ret); return ret; } @@ -1769,25 +1738,17 @@ static int ocfs2_write_begin(struct file *file, struct address_space *mapping, */ down_write(&OCFS2_I(inode)->ip_alloc_sem); - ret = ocfs2_data_lock(inode, 1); - if (ret) { - mlog_errno(ret); - goto out_fail; - } - ret = ocfs2_write_begin_nolock(mapping, pos, len, flags, pagep, fsdata, di_bh, NULL); if (ret) { mlog_errno(ret); - goto out_fail_data; + goto out_fail; } brelse(di_bh); return 0; -out_fail_data: - ocfs2_data_unlock(inode, 1); out_fail: up_write(&OCFS2_I(inode)->ip_alloc_sem); @@ -1908,7 +1869,6 @@ static int ocfs2_write_end(struct file *file, struct address_space *mapping, ret = ocfs2_write_end_nolock(mapping, pos, len, copied, page, fsdata); - ocfs2_data_unlock(inode, 1); up_write(&OCFS2_I(inode)->ip_alloc_sem); ocfs2_meta_unlock(inode, 1); diff --git a/fs/ocfs2/cluster/tcp_internal.h b/fs/ocfs2/cluster/tcp_internal.h index 79bd6665b3ca..b2e832aca567 100644 --- a/fs/ocfs2/cluster/tcp_internal.h +++ b/fs/ocfs2/cluster/tcp_internal.h @@ -38,6 +38,9 @@ * locking semantics of the file system using the protocol. It should * be somewhere else, I'm sure, but right now it isn't. * + * New in version 10: + * - Meta/data locks combined + * * New in version 9: * - All votes removed * @@ -63,7 +66,7 @@ * - full 64 bit i_size in the metadata lock lvbs * - introduction of "rw" lock and pushing meta/data locking down */ -#define O2NET_PROTOCOL_VERSION 9ULL +#define O2NET_PROTOCOL_VERSION 10ULL struct o2net_handshake { __be64 protocol_version; __be64 connector_id; diff --git a/fs/ocfs2/dlmglue.c b/fs/ocfs2/dlmglue.c index 7e36abea8f40..ecf58c6e2fa3 100644 --- a/fs/ocfs2/dlmglue.c +++ b/fs/ocfs2/dlmglue.c @@ -232,12 +232,6 @@ static struct ocfs2_lock_res_ops ocfs2_inode_meta_lops = { .flags = LOCK_TYPE_REQUIRES_REFRESH|LOCK_TYPE_USES_LVB, }; -static struct ocfs2_lock_res_ops ocfs2_inode_data_lops = { - .get_osb = ocfs2_get_inode_osb, - .downconvert_worker = ocfs2_data_convert_worker, - .flags = 0, -}; - static struct ocfs2_lock_res_ops ocfs2_super_lops = { .flags = LOCK_TYPE_REQUIRES_REFRESH, }; @@ -261,7 +255,6 @@ static struct ocfs2_lock_res_ops ocfs2_inode_open_lops = { static inline int ocfs2_is_inode_lock(struct ocfs2_lock_res *lockres) { return lockres->l_type == OCFS2_LOCK_TYPE_META || - lockres->l_type == OCFS2_LOCK_TYPE_DATA || lockres->l_type == OCFS2_LOCK_TYPE_RW || lockres->l_type == OCFS2_LOCK_TYPE_OPEN; } @@ -405,9 +398,6 @@ void ocfs2_inode_lock_res_init(struct ocfs2_lock_res *res, case OCFS2_LOCK_TYPE_META: ops = &ocfs2_inode_meta_lops; break; - case OCFS2_LOCK_TYPE_DATA: - ops = &ocfs2_inode_data_lops; - break; case OCFS2_LOCK_TYPE_OPEN: ops = &ocfs2_inode_open_lops; break; @@ -1154,12 +1144,6 @@ int ocfs2_create_new_inode_locks(struct inode *inode) goto bail; } - ret = ocfs2_create_new_lock(osb, &OCFS2_I(inode)->ip_data_lockres, 1, 1); - if (ret) { - mlog_errno(ret); - goto bail; - } - ret = ocfs2_create_new_lock(osb, &OCFS2_I(inode)->ip_open_lockres, 0, 0); if (ret) { mlog_errno(ret); @@ -1312,67 +1296,6 @@ out: mlog_exit_void(); } -int ocfs2_data_lock_full(struct inode *inode, - int write, - int arg_flags) -{ - int status = 0, level; - struct ocfs2_lock_res *lockres; - struct ocfs2_super *osb = OCFS2_SB(inode->i_sb); - - BUG_ON(!inode); - - mlog_entry_void(); - - mlog(0, "inode %llu take %s DATA lock\n", - (unsigned long long)OCFS2_I(inode)->ip_blkno, - write ? "EXMODE" : "PRMODE"); - - /* We'll allow faking a readonly data lock for - * rodevices. */ - if (ocfs2_is_hard_readonly(OCFS2_SB(inode->i_sb))) { - if (write) { - status = -EROFS; - mlog_errno(status); - } - goto out; - } - - if (ocfs2_mount_local(osb)) - goto out; - - lockres = &OCFS2_I(inode)->ip_data_lockres; - - level = write ? LKM_EXMODE : LKM_PRMODE; - - status = ocfs2_cluster_lock(OCFS2_SB(inode->i_sb), lockres, level, - 0, arg_flags); - if (status < 0 && status != -EAGAIN) - mlog_errno(status); - -out: - mlog_exit(status); - return status; -} - -/* see ocfs2_meta_lock_with_page() */ -int ocfs2_data_lock_with_page(struct inode *inode, - int write, - struct page *page) -{ - int ret; - - ret = ocfs2_data_lock_full(inode, write, OCFS2_LOCK_NONBLOCK); - if (ret == -EAGAIN) { - unlock_page(page); - if (ocfs2_data_lock(inode, write) == 0) - ocfs2_data_unlock(inode, write); - ret = AOP_TRUNCATED_PAGE; - } - - return ret; -} - static void ocfs2_downconvert_on_unlock(struct ocfs2_super *osb, struct ocfs2_lock_res *lockres) { @@ -1404,26 +1327,6 @@ static void ocfs2_downconvert_on_unlock(struct ocfs2_super *osb, mlog_exit_void(); } -void ocfs2_data_unlock(struct inode *inode, - int write) -{ - int level = write ? LKM_EXMODE : LKM_PRMODE; - struct ocfs2_lock_res *lockres = &OCFS2_I(inode)->ip_data_lockres; - struct ocfs2_super *osb = OCFS2_SB(inode->i_sb); - - mlog_entry_void(); - - mlog(0, "inode %llu drop %s DATA lock\n", - (unsigned long long)OCFS2_I(inode)->ip_blkno, - write ? "EXMODE" : "PRMODE"); - - if (!ocfs2_is_hard_readonly(OCFS2_SB(inode->i_sb)) && - !ocfs2_mount_local(osb)) - ocfs2_cluster_unlock(OCFS2_SB(inode->i_sb), lockres, level); - - mlog_exit_void(); -} - #define OCFS2_SEC_BITS 34 #define OCFS2_SEC_SHIFT (64 - 34) #define OCFS2_NSEC_MASK ((1ULL << OCFS2_SEC_SHIFT) - 1) @@ -2591,13 +2494,6 @@ int ocfs2_drop_inode_locks(struct inode *inode) status = err; - err = ocfs2_drop_lock(OCFS2_SB(inode->i_sb), - &OCFS2_I(inode)->ip_data_lockres); - if (err < 0) - mlog_errno(err); - if (err < 0 && !status) - status = err; - err = ocfs2_drop_lock(OCFS2_SB(inode->i_sb), &OCFS2_I(inode)->ip_meta_lockres); if (err < 0) diff --git a/fs/ocfs2/dlmglue.h b/fs/ocfs2/dlmglue.h index 931f6ee55146..3fd7729daeef 100644 --- a/fs/ocfs2/dlmglue.h +++ b/fs/ocfs2/dlmglue.h @@ -49,7 +49,7 @@ struct ocfs2_meta_lvb { __be32 lvb_reserved2; }; -/* ocfs2_meta_lock_full() and ocfs2_data_lock_full() 'arg_flags' flags */ +/* ocfs2_meta_lock_full() 'arg_flags' flags */ /* don't wait on recovery. */ #define OCFS2_META_LOCK_RECOVERY (0x01) /* Instruct the dlm not to queue ourselves on the other node. */ @@ -69,15 +69,6 @@ void ocfs2_dentry_lock_res_init(struct ocfs2_dentry_lock *dl, void ocfs2_lock_res_free(struct ocfs2_lock_res *res); int ocfs2_create_new_inode_locks(struct inode *inode); int ocfs2_drop_inode_locks(struct inode *inode); -int ocfs2_data_lock_full(struct inode *inode, - int write, - int arg_flags); -#define ocfs2_data_lock(inode, write) ocfs2_data_lock_full(inode, write, 0) -int ocfs2_data_lock_with_page(struct inode *inode, - int write, - struct page *page); -void ocfs2_data_unlock(struct inode *inode, - int write); int ocfs2_rw_lock(struct inode *inode, int write); void ocfs2_rw_unlock(struct inode *inode, int write); int ocfs2_open_lock(struct inode *inode); diff --git a/fs/ocfs2/file.c b/fs/ocfs2/file.c index b75b2e1f0e42..c5c183ac41fe 100644 --- a/fs/ocfs2/file.c +++ b/fs/ocfs2/file.c @@ -382,18 +382,13 @@ static int ocfs2_truncate_file(struct inode *inode, down_write(&OCFS2_I(inode)->ip_alloc_sem); - /* This forces other nodes to sync and drop their pages. Do - * this even if we have a truncate without allocation change - - * ocfs2 cluster sizes can be much greater than page size, so - * we have to truncate them anyway. */ - status = ocfs2_data_lock(inode, 1); - if (status < 0) { - up_write(&OCFS2_I(inode)->ip_alloc_sem); - - mlog_errno(status); - goto bail; - } - + /* + * The inode lock forced other nodes to sync and drop their + * pages, which (correctly) happens even if we have a truncate + * without allocation change - ocfs2 cluster sizes can be much + * greater than page size, so we have to truncate them + * anyway. + */ unmap_mapping_range(inode->i_mapping, new_i_size + PAGE_SIZE - 1, 0, 1); truncate_inode_pages(inode->i_mapping, new_i_size); @@ -403,7 +398,7 @@ static int ocfs2_truncate_file(struct inode *inode, if (status) mlog_errno(status); - goto bail_unlock_data; + goto bail_unlock_sem; } /* alright, we're going to need to do a full blown alloc size @@ -413,25 +408,23 @@ static int ocfs2_truncate_file(struct inode *inode, status = ocfs2_orphan_for_truncate(osb, inode, di_bh, new_i_size); if (status < 0) { mlog_errno(status); - goto bail_unlock_data; + goto bail_unlock_sem; } status = ocfs2_prepare_truncate(osb, inode, di_bh, &tc); if (status < 0) { mlog_errno(status); - goto bail_unlock_data; + goto bail_unlock_sem; } status = ocfs2_commit_truncate(osb, inode, di_bh, tc); if (status < 0) { mlog_errno(status); - goto bail_unlock_data; + goto bail_unlock_sem; } /* TODO: orphan dir cleanup here. */ -bail_unlock_data: - ocfs2_data_unlock(inode, 1); - +bail_unlock_sem: up_write(&OCFS2_I(inode)->ip_alloc_sem); bail: @@ -917,7 +910,7 @@ static int ocfs2_extend_file(struct inode *inode, struct buffer_head *di_bh, u64 new_i_size) { - int ret = 0, data_locked = 0; + int ret = 0; struct ocfs2_inode_info *oi = OCFS2_I(inode); BUG_ON(!di_bh); @@ -943,20 +936,6 @@ static int ocfs2_extend_file(struct inode *inode, && ocfs2_sparse_alloc(OCFS2_SB(inode->i_sb))) goto out_update_size; - /* - * protect the pages that ocfs2_zero_extend is going to be - * pulling into the page cache.. we do this before the - * metadata extend so that we don't get into the situation - * where we've extended the metadata but can't get the data - * lock to zero. - */ - ret = ocfs2_data_lock(inode, 1); - if (ret < 0) { - mlog_errno(ret); - goto out; - } - data_locked = 1; - /* * The alloc sem blocks people in read/write from reading our * allocation until we're done changing it. We depend on @@ -980,7 +959,7 @@ static int ocfs2_extend_file(struct inode *inode, up_write(&oi->ip_alloc_sem); mlog_errno(ret); - goto out_unlock; + goto out; } } @@ -991,7 +970,7 @@ static int ocfs2_extend_file(struct inode *inode, if (ret < 0) { mlog_errno(ret); - goto out_unlock; + goto out; } out_update_size: @@ -999,10 +978,6 @@ out_update_size: if (ret < 0) mlog_errno(ret); -out_unlock: - if (data_locked) - ocfs2_data_unlock(inode, 1); - out: return ret; } diff --git a/fs/ocfs2/inode.c b/fs/ocfs2/inode.c index 86cf073996b5..8ff201d3705e 100644 --- a/fs/ocfs2/inode.c +++ b/fs/ocfs2/inode.c @@ -332,10 +332,6 @@ int ocfs2_populate_inode(struct inode *inode, struct ocfs2_dinode *fe, OCFS2_LOCK_TYPE_RW, inode->i_generation, inode); - ocfs2_inode_lock_res_init(&OCFS2_I(inode)->ip_data_lockres, - OCFS2_LOCK_TYPE_DATA, inode->i_generation, - inode); - ocfs2_set_inode_flags(inode); status = 0; @@ -1014,7 +1010,6 @@ void ocfs2_clear_inode(struct inode *inode) * the downconvert thread while waiting to destroy the locks. */ ocfs2_mark_lockres_freeing(&oi->ip_rw_lockres); ocfs2_mark_lockres_freeing(&oi->ip_meta_lockres); - ocfs2_mark_lockres_freeing(&oi->ip_data_lockres); ocfs2_mark_lockres_freeing(&oi->ip_open_lockres); /* We very well may get a clear_inode before all an inodes @@ -1038,7 +1033,6 @@ void ocfs2_clear_inode(struct inode *inode) ocfs2_lock_res_free(&oi->ip_rw_lockres); ocfs2_lock_res_free(&oi->ip_meta_lockres); - ocfs2_lock_res_free(&oi->ip_data_lockres); ocfs2_lock_res_free(&oi->ip_open_lockres); ocfs2_metadata_cache_purge(inode); diff --git a/fs/ocfs2/inode.h b/fs/ocfs2/inode.h index 70e881c55536..d1c54da687c9 100644 --- a/fs/ocfs2/inode.h +++ b/fs/ocfs2/inode.h @@ -35,7 +35,6 @@ struct ocfs2_inode_info struct ocfs2_lock_res ip_rw_lockres; struct ocfs2_lock_res ip_meta_lockres; - struct ocfs2_lock_res ip_data_lockres; struct ocfs2_lock_res ip_open_lockres; /* protects allocation changes on this inode. */ diff --git a/fs/ocfs2/mmap.c b/fs/ocfs2/mmap.c index 98756156d298..a7f0ccc6fdd8 100644 --- a/fs/ocfs2/mmap.c +++ b/fs/ocfs2/mmap.c @@ -181,17 +181,8 @@ static int ocfs2_page_mkwrite(struct vm_area_struct *vma, struct page *page) */ down_write(&OCFS2_I(inode)->ip_alloc_sem); - ret = ocfs2_data_lock(inode, 1); - if (ret < 0) { - mlog_errno(ret); - goto out_meta_unlock; - } - ret = __ocfs2_page_mkwrite(inode, di_bh, page); - ocfs2_data_unlock(inode, 1); - -out_meta_unlock: up_write(&OCFS2_I(inode)->ip_alloc_sem); brelse(di_bh); diff --git a/fs/ocfs2/super.c b/fs/ocfs2/super.c index 1996820488cc..064eba074f1e 100644 --- a/fs/ocfs2/super.c +++ b/fs/ocfs2/super.c @@ -1020,7 +1020,6 @@ static void ocfs2_inode_init_once(struct kmem_cache *cachep, void *data) ocfs2_lock_res_init_once(&oi->ip_rw_lockres); ocfs2_lock_res_init_once(&oi->ip_meta_lockres); - ocfs2_lock_res_init_once(&oi->ip_data_lockres); ocfs2_lock_res_init_once(&oi->ip_open_lockres); ocfs2_metadata_cache_init(&oi->vfs_inode); -- cgit v1.2.3 From e63aecb651ba73dffc62f9608ee1b7ae2a0ffd4b Mon Sep 17 00:00:00 2001 From: Mark Fasheh Date: Thu, 18 Oct 2007 15:30:42 -0700 Subject: ocfs2: Rename ocfs2_meta_[un]lock Call this the "inode_lock" now, since it covers both data and meta data. This patch makes no functional changes. Signed-off-by: Mark Fasheh --- fs/ocfs2/alloc.c | 8 ++++---- fs/ocfs2/aops.c | 18 ++++++++--------- fs/ocfs2/dir.c | 8 ++++---- fs/ocfs2/dlmglue.c | 50 +++++++++++++++++++++++------------------------ fs/ocfs2/dlmglue.h | 12 ++++++------ fs/ocfs2/export.c | 4 ++-- fs/ocfs2/file.c | 42 +++++++++++++++++++-------------------- fs/ocfs2/inode.c | 30 ++++++++++++++-------------- fs/ocfs2/inode.h | 2 +- fs/ocfs2/ioctl.c | 8 ++++---- fs/ocfs2/journal.c | 26 ++++++++++++------------- fs/ocfs2/localalloc.c | 8 ++++---- fs/ocfs2/mmap.c | 8 ++++---- fs/ocfs2/namei.c | 54 +++++++++++++++++++++++++-------------------------- fs/ocfs2/suballoc.c | 4 ++-- fs/ocfs2/super.c | 6 +++--- 16 files changed, 144 insertions(+), 144 deletions(-) (limited to 'fs/ocfs2/dlmglue.h') diff --git a/fs/ocfs2/alloc.c b/fs/ocfs2/alloc.c index 23c8cda43f19..e6df06ac6405 100644 --- a/fs/ocfs2/alloc.c +++ b/fs/ocfs2/alloc.c @@ -4731,7 +4731,7 @@ int __ocfs2_flush_truncate_log(struct ocfs2_super *osb) mutex_lock(&data_alloc_inode->i_mutex); - status = ocfs2_meta_lock(data_alloc_inode, &data_alloc_bh, 1); + status = ocfs2_inode_lock(data_alloc_inode, &data_alloc_bh, 1); if (status < 0) { mlog_errno(status); goto out_mutex; @@ -4753,7 +4753,7 @@ int __ocfs2_flush_truncate_log(struct ocfs2_super *osb) out_unlock: brelse(data_alloc_bh); - ocfs2_meta_unlock(data_alloc_inode, 1); + ocfs2_inode_unlock(data_alloc_inode, 1); out_mutex: mutex_unlock(&data_alloc_inode->i_mutex); @@ -5077,7 +5077,7 @@ static int ocfs2_free_cached_items(struct ocfs2_super *osb, mutex_lock(&inode->i_mutex); - ret = ocfs2_meta_lock(inode, &di_bh, 1); + ret = ocfs2_inode_lock(inode, &di_bh, 1); if (ret) { mlog_errno(ret); goto out_mutex; @@ -5118,7 +5118,7 @@ out_journal: ocfs2_commit_trans(osb, handle); out_unlock: - ocfs2_meta_unlock(inode, 1); + ocfs2_inode_unlock(inode, 1); brelse(di_bh); out_mutex: mutex_unlock(&inode->i_mutex); diff --git a/fs/ocfs2/aops.c b/fs/ocfs2/aops.c index 5fc27cfaee50..ac8c39055717 100644 --- a/fs/ocfs2/aops.c +++ b/fs/ocfs2/aops.c @@ -275,7 +275,7 @@ static int ocfs2_readpage(struct file *file, struct page *page) mlog_entry("(0x%p, %lu)\n", file, (page ? page->index : 0)); - ret = ocfs2_meta_lock_with_page(inode, NULL, 0, page); + ret = ocfs2_inode_lock_with_page(inode, NULL, 0, page); if (ret != 0) { if (ret == AOP_TRUNCATED_PAGE) unlock = 0; @@ -285,7 +285,7 @@ static int ocfs2_readpage(struct file *file, struct page *page) if (down_read_trylock(&oi->ip_alloc_sem) == 0) { ret = AOP_TRUNCATED_PAGE; - goto out_meta_unlock; + goto out_inode_unlock; } /* @@ -313,8 +313,8 @@ static int ocfs2_readpage(struct file *file, struct page *page) out_alloc: up_read(&OCFS2_I(inode)->ip_alloc_sem); -out_meta_unlock: - ocfs2_meta_unlock(inode, 0); +out_inode_unlock: + ocfs2_inode_unlock(inode, 0); out: if (unlock) unlock_page(page); @@ -443,7 +443,7 @@ static sector_t ocfs2_bmap(struct address_space *mapping, sector_t block) * accessed concurrently from multiple nodes. */ if (!INODE_JOURNAL(inode)) { - err = ocfs2_meta_lock(inode, NULL, 0); + err = ocfs2_inode_lock(inode, NULL, 0); if (err) { if (err != -ENOENT) mlog_errno(err); @@ -458,7 +458,7 @@ static sector_t ocfs2_bmap(struct address_space *mapping, sector_t block) if (!INODE_JOURNAL(inode)) { up_read(&OCFS2_I(inode)->ip_alloc_sem); - ocfs2_meta_unlock(inode, 0); + ocfs2_inode_unlock(inode, 0); } if (err) { @@ -1723,7 +1723,7 @@ static int ocfs2_write_begin(struct file *file, struct address_space *mapping, struct buffer_head *di_bh = NULL; struct inode *inode = mapping->host; - ret = ocfs2_meta_lock(inode, &di_bh, 1); + ret = ocfs2_inode_lock(inode, &di_bh, 1); if (ret) { mlog_errno(ret); return ret; @@ -1753,7 +1753,7 @@ out_fail: up_write(&OCFS2_I(inode)->ip_alloc_sem); brelse(di_bh); - ocfs2_meta_unlock(inode, 1); + ocfs2_inode_unlock(inode, 1); return ret; } @@ -1870,7 +1870,7 @@ static int ocfs2_write_end(struct file *file, struct address_space *mapping, ret = ocfs2_write_end_nolock(mapping, pos, len, copied, page, fsdata); up_write(&OCFS2_I(inode)->ip_alloc_sem); - ocfs2_meta_unlock(inode, 1); + ocfs2_inode_unlock(inode, 1); return ret; } diff --git a/fs/ocfs2/dir.c b/fs/ocfs2/dir.c index 63b28fdceb4a..6b0107f21344 100644 --- a/fs/ocfs2/dir.c +++ b/fs/ocfs2/dir.c @@ -846,14 +846,14 @@ int ocfs2_readdir(struct file * filp, void * dirent, filldir_t filldir) mlog_entry("dirino=%llu\n", (unsigned long long)OCFS2_I(inode)->ip_blkno); - error = ocfs2_meta_lock_atime(inode, filp->f_vfsmnt, &lock_level); + error = ocfs2_inode_lock_atime(inode, filp->f_vfsmnt, &lock_level); if (lock_level && error >= 0) { /* We release EX lock which used to update atime * and get PR lock again to reduce contention * on commonly accessed directories. */ - ocfs2_meta_unlock(inode, 1); + ocfs2_inode_unlock(inode, 1); lock_level = 0; - error = ocfs2_meta_lock(inode, NULL, 0); + error = ocfs2_inode_lock(inode, NULL, 0); } if (error < 0) { if (error != -ENOENT) @@ -865,7 +865,7 @@ int ocfs2_readdir(struct file * filp, void * dirent, filldir_t filldir) error = ocfs2_dir_foreach_blk(inode, &filp->f_version, &filp->f_pos, dirent, filldir, NULL); - ocfs2_meta_unlock(inode, lock_level); + ocfs2_inode_unlock(inode, lock_level); bail_nolock: mlog_exit(error); diff --git a/fs/ocfs2/dlmglue.c b/fs/ocfs2/dlmglue.c index ecf58c6e2fa3..fa5e3bdc295d 100644 --- a/fs/ocfs2/dlmglue.c +++ b/fs/ocfs2/dlmglue.c @@ -224,7 +224,7 @@ static struct ocfs2_lock_res_ops ocfs2_inode_rw_lops = { .flags = 0, }; -static struct ocfs2_lock_res_ops ocfs2_inode_meta_lops = { +static struct ocfs2_lock_res_ops ocfs2_inode_inode_lops = { .get_osb = ocfs2_get_inode_osb, .check_downconvert = ocfs2_check_meta_downconvert, .set_lvb = ocfs2_set_meta_lvb, @@ -306,7 +306,7 @@ static inline void ocfs2_recover_from_dlm_error(struct ocfs2_lock_res *lockres, static int ocfs2_downconvert_thread(void *arg); static void ocfs2_downconvert_on_unlock(struct ocfs2_super *osb, struct ocfs2_lock_res *lockres); -static int ocfs2_meta_lock_update(struct inode *inode, +static int ocfs2_inode_lock_update(struct inode *inode, struct buffer_head **bh); static void ocfs2_drop_osb_locks(struct ocfs2_super *osb); static inline int ocfs2_highest_compat_lock_level(int level); @@ -396,7 +396,7 @@ void ocfs2_inode_lock_res_init(struct ocfs2_lock_res *res, ops = &ocfs2_inode_rw_lops; break; case OCFS2_LOCK_TYPE_META: - ops = &ocfs2_inode_meta_lops; + ops = &ocfs2_inode_inode_lops; break; case OCFS2_LOCK_TYPE_OPEN: ops = &ocfs2_inode_open_lops; @@ -1138,7 +1138,7 @@ int ocfs2_create_new_inode_locks(struct inode *inode) * We don't want to use LKM_LOCAL on a meta data lock as they * don't use a generation in their lock names. */ - ret = ocfs2_create_new_lock(osb, &OCFS2_I(inode)->ip_meta_lockres, 1, 0); + ret = ocfs2_create_new_lock(osb, &OCFS2_I(inode)->ip_inode_lockres, 1, 0); if (ret) { mlog_errno(ret); goto bail; @@ -1346,11 +1346,11 @@ static u64 ocfs2_pack_timespec(struct timespec *spec) /* Call this with the lockres locked. I am reasonably sure we don't * need ip_lock in this function as anyone who would be changing those - * values is supposed to be blocked in ocfs2_meta_lock right now. */ + * values is supposed to be blocked in ocfs2_inode_lock right now. */ static void __ocfs2_stuff_meta_lvb(struct inode *inode) { struct ocfs2_inode_info *oi = OCFS2_I(inode); - struct ocfs2_lock_res *lockres = &oi->ip_meta_lockres; + struct ocfs2_lock_res *lockres = &oi->ip_inode_lockres; struct ocfs2_meta_lvb *lvb; mlog_entry_void(); @@ -1400,7 +1400,7 @@ static void ocfs2_unpack_timespec(struct timespec *spec, static void ocfs2_refresh_inode_from_lvb(struct inode *inode) { struct ocfs2_inode_info *oi = OCFS2_I(inode); - struct ocfs2_lock_res *lockres = &oi->ip_meta_lockres; + struct ocfs2_lock_res *lockres = &oi->ip_inode_lockres; struct ocfs2_meta_lvb *lvb; mlog_entry_void(); @@ -1508,12 +1508,12 @@ static inline void ocfs2_complete_lock_res_refresh(struct ocfs2_lock_res *lockre } /* may or may not return a bh if it went to disk. */ -static int ocfs2_meta_lock_update(struct inode *inode, +static int ocfs2_inode_lock_update(struct inode *inode, struct buffer_head **bh) { int status = 0; struct ocfs2_inode_info *oi = OCFS2_I(inode); - struct ocfs2_lock_res *lockres = &oi->ip_meta_lockres; + struct ocfs2_lock_res *lockres = &oi->ip_inode_lockres; struct ocfs2_dinode *fe; struct ocfs2_super *osb = OCFS2_SB(inode->i_sb); @@ -1625,7 +1625,7 @@ static int ocfs2_assign_bh(struct inode *inode, * returns < 0 error if the callback will never be called, otherwise * the result of the lock will be communicated via the callback. */ -int ocfs2_meta_lock_full(struct inode *inode, +int ocfs2_inode_lock_full(struct inode *inode, struct buffer_head **ret_bh, int ex, int arg_flags) @@ -1660,7 +1660,7 @@ int ocfs2_meta_lock_full(struct inode *inode, wait_event(osb->recovery_event, ocfs2_node_map_is_empty(osb, &osb->recovery_map)); - lockres = &OCFS2_I(inode)->ip_meta_lockres; + lockres = &OCFS2_I(inode)->ip_inode_lockres; level = ex ? LKM_EXMODE : LKM_PRMODE; dlm_flags = 0; if (arg_flags & OCFS2_META_LOCK_NOQUEUE) @@ -1699,11 +1699,11 @@ local: } /* This is fun. The caller may want a bh back, or it may - * not. ocfs2_meta_lock_update definitely wants one in, but + * not. ocfs2_inode_lock_update definitely wants one in, but * may or may not read one, depending on what's in the * LVB. The result of all of this is that we've *only* gone to * disk if we have to, so the complexity is worthwhile. */ - status = ocfs2_meta_lock_update(inode, &local_bh); + status = ocfs2_inode_lock_update(inode, &local_bh); if (status < 0) { if (status != -ENOENT) mlog_errno(status); @@ -1725,7 +1725,7 @@ bail: *ret_bh = NULL; } if (acquired) - ocfs2_meta_unlock(inode, ex); + ocfs2_inode_unlock(inode, ex); } if (local_bh) @@ -1757,32 +1757,32 @@ bail: * ping locks back and forth, but that's a risk we're willing to take to avoid * the lock inversion simply. */ -int ocfs2_meta_lock_with_page(struct inode *inode, +int ocfs2_inode_lock_with_page(struct inode *inode, struct buffer_head **ret_bh, int ex, struct page *page) { int ret; - ret = ocfs2_meta_lock_full(inode, ret_bh, ex, OCFS2_LOCK_NONBLOCK); + ret = ocfs2_inode_lock_full(inode, ret_bh, ex, OCFS2_LOCK_NONBLOCK); if (ret == -EAGAIN) { unlock_page(page); - if (ocfs2_meta_lock(inode, ret_bh, ex) == 0) - ocfs2_meta_unlock(inode, ex); + if (ocfs2_inode_lock(inode, ret_bh, ex) == 0) + ocfs2_inode_unlock(inode, ex); ret = AOP_TRUNCATED_PAGE; } return ret; } -int ocfs2_meta_lock_atime(struct inode *inode, +int ocfs2_inode_lock_atime(struct inode *inode, struct vfsmount *vfsmnt, int *level) { int ret; mlog_entry_void(); - ret = ocfs2_meta_lock(inode, NULL, 0); + ret = ocfs2_inode_lock(inode, NULL, 0); if (ret < 0) { mlog_errno(ret); return ret; @@ -1795,8 +1795,8 @@ int ocfs2_meta_lock_atime(struct inode *inode, if (ocfs2_should_update_atime(inode, vfsmnt)) { struct buffer_head *bh = NULL; - ocfs2_meta_unlock(inode, 0); - ret = ocfs2_meta_lock(inode, &bh, 1); + ocfs2_inode_unlock(inode, 0); + ret = ocfs2_inode_lock(inode, &bh, 1); if (ret < 0) { mlog_errno(ret); return ret; @@ -1813,11 +1813,11 @@ int ocfs2_meta_lock_atime(struct inode *inode, return ret; } -void ocfs2_meta_unlock(struct inode *inode, +void ocfs2_inode_unlock(struct inode *inode, int ex) { int level = ex ? LKM_EXMODE : LKM_PRMODE; - struct ocfs2_lock_res *lockres = &OCFS2_I(inode)->ip_meta_lockres; + struct ocfs2_lock_res *lockres = &OCFS2_I(inode)->ip_inode_lockres; struct ocfs2_super *osb = OCFS2_SB(inode->i_sb); mlog_entry_void(); @@ -2495,7 +2495,7 @@ int ocfs2_drop_inode_locks(struct inode *inode) status = err; err = ocfs2_drop_lock(OCFS2_SB(inode->i_sb), - &OCFS2_I(inode)->ip_meta_lockres); + &OCFS2_I(inode)->ip_inode_lockres); if (err < 0) mlog_errno(err); if (err < 0 && !status) diff --git a/fs/ocfs2/dlmglue.h b/fs/ocfs2/dlmglue.h index 3fd7729daeef..6dcbc944e8ce 100644 --- a/fs/ocfs2/dlmglue.h +++ b/fs/ocfs2/dlmglue.h @@ -49,7 +49,7 @@ struct ocfs2_meta_lvb { __be32 lvb_reserved2; }; -/* ocfs2_meta_lock_full() 'arg_flags' flags */ +/* ocfs2_inode_lock_full() 'arg_flags' flags */ /* don't wait on recovery. */ #define OCFS2_META_LOCK_RECOVERY (0x01) /* Instruct the dlm not to queue ourselves on the other node. */ @@ -74,21 +74,21 @@ void ocfs2_rw_unlock(struct inode *inode, int write); int ocfs2_open_lock(struct inode *inode); int ocfs2_try_open_lock(struct inode *inode, int write); void ocfs2_open_unlock(struct inode *inode); -int ocfs2_meta_lock_atime(struct inode *inode, +int ocfs2_inode_lock_atime(struct inode *inode, struct vfsmount *vfsmnt, int *level); -int ocfs2_meta_lock_full(struct inode *inode, +int ocfs2_inode_lock_full(struct inode *inode, struct buffer_head **ret_bh, int ex, int arg_flags); -int ocfs2_meta_lock_with_page(struct inode *inode, +int ocfs2_inode_lock_with_page(struct inode *inode, struct buffer_head **ret_bh, int ex, struct page *page); /* 99% of the time we don't want to supply any additional flags -- * those are for very specific cases only. */ -#define ocfs2_meta_lock(i, b, e) ocfs2_meta_lock_full(i, b, e, 0) -void ocfs2_meta_unlock(struct inode *inode, +#define ocfs2_inode_lock(i, b, e) ocfs2_inode_lock_full(i, b, e, 0) +void ocfs2_inode_unlock(struct inode *inode, int ex); int ocfs2_super_lock(struct ocfs2_super *osb, int ex); diff --git a/fs/ocfs2/export.c b/fs/ocfs2/export.c index 535bfa9568a4..1f9e353cac45 100644 --- a/fs/ocfs2/export.c +++ b/fs/ocfs2/export.c @@ -95,7 +95,7 @@ static struct dentry *ocfs2_get_parent(struct dentry *child) mlog(0, "find parent of directory %llu\n", (unsigned long long)OCFS2_I(dir)->ip_blkno); - status = ocfs2_meta_lock(dir, NULL, 0); + status = ocfs2_inode_lock(dir, NULL, 0); if (status < 0) { if (status != -ENOENT) mlog_errno(status); @@ -126,7 +126,7 @@ static struct dentry *ocfs2_get_parent(struct dentry *child) parent->d_op = &ocfs2_dentry_ops; bail_unlock: - ocfs2_meta_unlock(dir, 0); + ocfs2_inode_unlock(dir, 0); bail: mlog_exit_ptr(parent); diff --git a/fs/ocfs2/file.c b/fs/ocfs2/file.c index c5c183ac41fe..432e5f3c4784 100644 --- a/fs/ocfs2/file.c +++ b/fs/ocfs2/file.c @@ -1025,7 +1025,7 @@ int ocfs2_setattr(struct dentry *dentry, struct iattr *attr) } } - status = ocfs2_meta_lock(inode, &bh, 1); + status = ocfs2_inode_lock(inode, &bh, 1); if (status < 0) { if (status != -ENOENT) mlog_errno(status); @@ -1077,7 +1077,7 @@ int ocfs2_setattr(struct dentry *dentry, struct iattr *attr) bail_commit: ocfs2_commit_trans(osb, handle); bail_unlock: - ocfs2_meta_unlock(inode, 1); + ocfs2_inode_unlock(inode, 1); bail_unlock_rw: if (size_change) ocfs2_rw_unlock(inode, 1); @@ -1124,7 +1124,7 @@ int ocfs2_permission(struct inode *inode, int mask, struct nameidata *nd) mlog_entry_void(); - ret = ocfs2_meta_lock(inode, NULL, 0); + ret = ocfs2_inode_lock(inode, NULL, 0); if (ret) { if (ret != -ENOENT) mlog_errno(ret); @@ -1133,7 +1133,7 @@ int ocfs2_permission(struct inode *inode, int mask, struct nameidata *nd) ret = generic_permission(inode, mask, NULL); - ocfs2_meta_unlock(inode, 0); + ocfs2_inode_unlock(inode, 0); out: mlog_exit(ret); return ret; @@ -1605,7 +1605,7 @@ static int __ocfs2_change_file_space(struct file *file, struct inode *inode, goto out; } - ret = ocfs2_meta_lock(inode, &di_bh, 1); + ret = ocfs2_inode_lock(inode, &di_bh, 1); if (ret) { mlog_errno(ret); goto out_rw_unlock; @@ -1613,7 +1613,7 @@ static int __ocfs2_change_file_space(struct file *file, struct inode *inode, if (inode->i_flags & (S_IMMUTABLE|S_APPEND)) { ret = -EPERM; - goto out_meta_unlock; + goto out_inode_unlock; } switch (sr->l_whence) { @@ -1627,7 +1627,7 @@ static int __ocfs2_change_file_space(struct file *file, struct inode *inode, break; default: ret = -EINVAL; - goto out_meta_unlock; + goto out_inode_unlock; } sr->l_whence = 0; @@ -1638,14 +1638,14 @@ static int __ocfs2_change_file_space(struct file *file, struct inode *inode, || (sr->l_start + llen) < 0 || (sr->l_start + llen) > max_off) { ret = -EINVAL; - goto out_meta_unlock; + goto out_inode_unlock; } size = sr->l_start + sr->l_len; if (cmd == OCFS2_IOC_RESVSP || cmd == OCFS2_IOC_RESVSP64) { if (sr->l_len <= 0) { ret = -EINVAL; - goto out_meta_unlock; + goto out_inode_unlock; } } @@ -1653,7 +1653,7 @@ static int __ocfs2_change_file_space(struct file *file, struct inode *inode, ret = __ocfs2_write_remove_suid(inode, di_bh); if (ret) { mlog_errno(ret); - goto out_meta_unlock; + goto out_inode_unlock; } } @@ -1679,7 +1679,7 @@ static int __ocfs2_change_file_space(struct file *file, struct inode *inode, up_write(&OCFS2_I(inode)->ip_alloc_sem); if (ret) { mlog_errno(ret); - goto out_meta_unlock; + goto out_inode_unlock; } /* @@ -1689,7 +1689,7 @@ static int __ocfs2_change_file_space(struct file *file, struct inode *inode, if (IS_ERR(handle)) { ret = PTR_ERR(handle); mlog_errno(ret); - goto out_meta_unlock; + goto out_inode_unlock; } if (change_size && i_size_read(inode) < size) @@ -1702,9 +1702,9 @@ static int __ocfs2_change_file_space(struct file *file, struct inode *inode, ocfs2_commit_trans(osb, handle); -out_meta_unlock: +out_inode_unlock: brelse(di_bh); - ocfs2_meta_unlock(inode, 1); + ocfs2_inode_unlock(inode, 1); out_rw_unlock: ocfs2_rw_unlock(inode, 1); @@ -1774,7 +1774,7 @@ static int ocfs2_prepare_inode_for_write(struct dentry *dentry, * if we need to make modifications here. */ for(;;) { - ret = ocfs2_meta_lock(inode, NULL, meta_level); + ret = ocfs2_inode_lock(inode, NULL, meta_level); if (ret < 0) { meta_level = -1; mlog_errno(ret); @@ -1792,7 +1792,7 @@ static int ocfs2_prepare_inode_for_write(struct dentry *dentry, * set inode->i_size at the end of a write. */ if (should_remove_suid(dentry)) { if (meta_level == 0) { - ocfs2_meta_unlock(inode, meta_level); + ocfs2_inode_unlock(inode, meta_level); meta_level = 1; continue; } @@ -1861,7 +1861,7 @@ static int ocfs2_prepare_inode_for_write(struct dentry *dentry, *ppos = saved_pos; out_unlock: - ocfs2_meta_unlock(inode, meta_level); + ocfs2_inode_unlock(inode, meta_level); out: return ret; @@ -2074,12 +2074,12 @@ static ssize_t ocfs2_file_splice_read(struct file *in, /* * See the comment in ocfs2_file_aio_read() */ - ret = ocfs2_meta_lock(inode, NULL, 0); + ret = ocfs2_inode_lock(inode, NULL, 0); if (ret < 0) { mlog_errno(ret); goto bail; } - ocfs2_meta_unlock(inode, 0); + ocfs2_inode_unlock(inode, 0); ret = generic_file_splice_read(in, ppos, pipe, len, flags); @@ -2135,12 +2135,12 @@ static ssize_t ocfs2_file_aio_read(struct kiocb *iocb, * like i_size. This allows the checks down below * generic_file_aio_read() a chance of actually working. */ - ret = ocfs2_meta_lock_atime(inode, filp->f_vfsmnt, &lock_level); + ret = ocfs2_inode_lock_atime(inode, filp->f_vfsmnt, &lock_level); if (ret < 0) { mlog_errno(ret); goto bail; } - ocfs2_meta_unlock(inode, lock_level); + ocfs2_inode_unlock(inode, lock_level); ret = generic_file_aio_read(iocb, iov, nr_segs, iocb->ki_pos); if (ret == -EINVAL) diff --git a/fs/ocfs2/inode.c b/fs/ocfs2/inode.c index 8ff201d3705e..00cd5b7f3e52 100644 --- a/fs/ocfs2/inode.c +++ b/fs/ocfs2/inode.c @@ -321,7 +321,7 @@ int ocfs2_populate_inode(struct inode *inode, struct ocfs2_dinode *fe, */ BUG_ON(le32_to_cpu(fe->i_flags) & OCFS2_SYSTEM_FL); - ocfs2_inode_lock_res_init(&OCFS2_I(inode)->ip_meta_lockres, + ocfs2_inode_lock_res_init(&OCFS2_I(inode)->ip_inode_lockres, OCFS2_LOCK_TYPE_META, 0, inode); ocfs2_inode_lock_res_init(&OCFS2_I(inode)->ip_open_lockres, @@ -409,7 +409,7 @@ static int ocfs2_read_locked_inode(struct inode *inode, if (args->fi_flags & OCFS2_FI_FLAG_SYSFILE) generation = osb->fs_generation; - ocfs2_inode_lock_res_init(&OCFS2_I(inode)->ip_meta_lockres, + ocfs2_inode_lock_res_init(&OCFS2_I(inode)->ip_inode_lockres, OCFS2_LOCK_TYPE_META, generation, inode); @@ -424,7 +424,7 @@ static int ocfs2_read_locked_inode(struct inode *inode, mlog_errno(status); return status; } - status = ocfs2_meta_lock(inode, NULL, 0); + status = ocfs2_inode_lock(inode, NULL, 0); if (status) { make_bad_inode(inode); mlog_errno(status); @@ -479,7 +479,7 @@ static int ocfs2_read_locked_inode(struct inode *inode, bail: if (can_lock) - ocfs2_meta_unlock(inode, 0); + ocfs2_inode_unlock(inode, 0); if (status < 0) make_bad_inode(inode); @@ -581,7 +581,7 @@ static int ocfs2_remove_inode(struct inode *inode, } mutex_lock(&inode_alloc_inode->i_mutex); - status = ocfs2_meta_lock(inode_alloc_inode, &inode_alloc_bh, 1); + status = ocfs2_inode_lock(inode_alloc_inode, &inode_alloc_bh, 1); if (status < 0) { mutex_unlock(&inode_alloc_inode->i_mutex); @@ -630,7 +630,7 @@ static int ocfs2_remove_inode(struct inode *inode, bail_commit: ocfs2_commit_trans(osb, handle); bail_unlock: - ocfs2_meta_unlock(inode_alloc_inode, 1); + ocfs2_inode_unlock(inode_alloc_inode, 1); mutex_unlock(&inode_alloc_inode->i_mutex); brelse(inode_alloc_bh); bail: @@ -704,7 +704,7 @@ static int ocfs2_wipe_inode(struct inode *inode, * delete_inode operation. We do this now to avoid races with * recovery completion on other nodes. */ mutex_lock(&orphan_dir_inode->i_mutex); - status = ocfs2_meta_lock(orphan_dir_inode, &orphan_dir_bh, 1); + status = ocfs2_inode_lock(orphan_dir_inode, &orphan_dir_bh, 1); if (status < 0) { mutex_unlock(&orphan_dir_inode->i_mutex); @@ -728,7 +728,7 @@ static int ocfs2_wipe_inode(struct inode *inode, mlog_errno(status); bail_unlock_dir: - ocfs2_meta_unlock(orphan_dir_inode, 1); + ocfs2_inode_unlock(orphan_dir_inode, 1); mutex_unlock(&orphan_dir_inode->i_mutex); brelse(orphan_dir_bh); bail: @@ -929,7 +929,7 @@ void ocfs2_delete_inode(struct inode *inode) * allocation lock here as it won't be needed - nobody will * have the file open. */ - status = ocfs2_meta_lock(inode, &di_bh, 1); + status = ocfs2_inode_lock(inode, &di_bh, 1); if (status < 0) { if (status != -ENOENT) mlog_errno(status); @@ -975,7 +975,7 @@ void ocfs2_delete_inode(struct inode *inode) OCFS2_I(inode)->ip_flags |= OCFS2_INODE_DELETED; bail_unlock_inode: - ocfs2_meta_unlock(inode, 1); + ocfs2_inode_unlock(inode, 1); brelse(di_bh); bail_unblock: status = sigprocmask(SIG_SETMASK, &oldset, NULL); @@ -1009,7 +1009,7 @@ void ocfs2_clear_inode(struct inode *inode) /* Do these before all the other work so that we don't bounce * the downconvert thread while waiting to destroy the locks. */ ocfs2_mark_lockres_freeing(&oi->ip_rw_lockres); - ocfs2_mark_lockres_freeing(&oi->ip_meta_lockres); + ocfs2_mark_lockres_freeing(&oi->ip_inode_lockres); ocfs2_mark_lockres_freeing(&oi->ip_open_lockres); /* We very well may get a clear_inode before all an inodes @@ -1032,7 +1032,7 @@ void ocfs2_clear_inode(struct inode *inode) mlog_errno(status); ocfs2_lock_res_free(&oi->ip_rw_lockres); - ocfs2_lock_res_free(&oi->ip_meta_lockres); + ocfs2_lock_res_free(&oi->ip_inode_lockres); ocfs2_lock_res_free(&oi->ip_open_lockres); ocfs2_metadata_cache_purge(inode); @@ -1176,15 +1176,15 @@ int ocfs2_inode_revalidate(struct dentry *dentry) } spin_unlock(&OCFS2_I(inode)->ip_lock); - /* Let ocfs2_meta_lock do the work of updating our struct + /* Let ocfs2_inode_lock do the work of updating our struct * inode for us. */ - status = ocfs2_meta_lock(inode, NULL, 0); + status = ocfs2_inode_lock(inode, NULL, 0); if (status < 0) { if (status != -ENOENT) mlog_errno(status); goto bail; } - ocfs2_meta_unlock(inode, 0); + ocfs2_inode_unlock(inode, 0); bail: mlog_exit(status); diff --git a/fs/ocfs2/inode.h b/fs/ocfs2/inode.h index d1c54da687c9..a61c044eb7da 100644 --- a/fs/ocfs2/inode.h +++ b/fs/ocfs2/inode.h @@ -34,7 +34,7 @@ struct ocfs2_inode_info u64 ip_blkno; struct ocfs2_lock_res ip_rw_lockres; - struct ocfs2_lock_res ip_meta_lockres; + struct ocfs2_lock_res ip_inode_lockres; struct ocfs2_lock_res ip_open_lockres; /* protects allocation changes on this inode. */ diff --git a/fs/ocfs2/ioctl.c b/fs/ocfs2/ioctl.c index 87dcece7e1b5..67c2fb4bae91 100644 --- a/fs/ocfs2/ioctl.c +++ b/fs/ocfs2/ioctl.c @@ -27,14 +27,14 @@ static int ocfs2_get_inode_attr(struct inode *inode, unsigned *flags) { int status; - status = ocfs2_meta_lock(inode, NULL, 0); + status = ocfs2_inode_lock(inode, NULL, 0); if (status < 0) { mlog_errno(status); return status; } ocfs2_get_inode_flags(OCFS2_I(inode)); *flags = OCFS2_I(inode)->ip_attr; - ocfs2_meta_unlock(inode, 0); + ocfs2_inode_unlock(inode, 0); mlog_exit(status); return status; @@ -52,7 +52,7 @@ static int ocfs2_set_inode_attr(struct inode *inode, unsigned flags, mutex_lock(&inode->i_mutex); - status = ocfs2_meta_lock(inode, &bh, 1); + status = ocfs2_inode_lock(inode, &bh, 1); if (status < 0) { mlog_errno(status); goto bail; @@ -100,7 +100,7 @@ static int ocfs2_set_inode_attr(struct inode *inode, unsigned flags, ocfs2_commit_trans(osb, handle); bail_unlock: - ocfs2_meta_unlock(inode, 1); + ocfs2_inode_unlock(inode, 1); bail: mutex_unlock(&inode->i_mutex); diff --git a/fs/ocfs2/journal.c b/fs/ocfs2/journal.c index f2ebe2eb3c21..4f440a88bf53 100644 --- a/fs/ocfs2/journal.c +++ b/fs/ocfs2/journal.c @@ -336,7 +336,7 @@ int ocfs2_journal_init(struct ocfs2_journal *journal, int *dirty) struct ocfs2_dinode *di = NULL; struct buffer_head *bh = NULL; struct ocfs2_super *osb; - int meta_lock = 0; + int inode_lock = 0; mlog_entry_void(); @@ -366,14 +366,14 @@ int ocfs2_journal_init(struct ocfs2_journal *journal, int *dirty) /* Skip recovery waits here - journal inode metadata never * changes in a live cluster so it can be considered an * exception to the rule. */ - status = ocfs2_meta_lock_full(inode, &bh, 1, OCFS2_META_LOCK_RECOVERY); + status = ocfs2_inode_lock_full(inode, &bh, 1, OCFS2_META_LOCK_RECOVERY); if (status < 0) { if (status != -ERESTARTSYS) mlog(ML_ERROR, "Could not get lock on journal!\n"); goto done; } - meta_lock = 1; + inode_lock = 1; di = (struct ocfs2_dinode *)bh->b_data; if (inode->i_size < OCFS2_MIN_JOURNAL_SIZE) { @@ -413,8 +413,8 @@ int ocfs2_journal_init(struct ocfs2_journal *journal, int *dirty) status = 0; done: if (status < 0) { - if (meta_lock) - ocfs2_meta_unlock(inode, 1); + if (inode_lock) + ocfs2_inode_unlock(inode, 1); if (bh != NULL) brelse(bh); if (inode) { @@ -543,7 +543,7 @@ void ocfs2_journal_shutdown(struct ocfs2_super *osb) OCFS2_I(inode)->ip_open_count--; /* unlock our journal */ - ocfs2_meta_unlock(inode, 1); + ocfs2_inode_unlock(inode, 1); brelse(journal->j_bh); journal->j_bh = NULL; @@ -972,9 +972,9 @@ static int ocfs2_replay_journal(struct ocfs2_super *osb, } SET_INODE_JOURNAL(inode); - status = ocfs2_meta_lock_full(inode, &bh, 1, OCFS2_META_LOCK_RECOVERY); + status = ocfs2_inode_lock_full(inode, &bh, 1, OCFS2_META_LOCK_RECOVERY); if (status < 0) { - mlog(0, "status returned from ocfs2_meta_lock=%d\n", status); + mlog(0, "status returned from ocfs2_inode_lock=%d\n", status); if (status != -ERESTARTSYS) mlog(ML_ERROR, "Could not lock journal!\n"); goto done; @@ -1046,7 +1046,7 @@ static int ocfs2_replay_journal(struct ocfs2_super *osb, done: /* drop the lock on this nodes journal */ if (got_lock) - ocfs2_meta_unlock(inode, 1); + ocfs2_inode_unlock(inode, 1); if (inode) iput(inode); @@ -1161,14 +1161,14 @@ static int ocfs2_trylock_journal(struct ocfs2_super *osb, SET_INODE_JOURNAL(inode); flags = OCFS2_META_LOCK_RECOVERY | OCFS2_META_LOCK_NOQUEUE; - status = ocfs2_meta_lock_full(inode, NULL, 1, flags); + status = ocfs2_inode_lock_full(inode, NULL, 1, flags); if (status < 0) { if (status != -EAGAIN) mlog_errno(status); goto bail; } - ocfs2_meta_unlock(inode, 1); + ocfs2_inode_unlock(inode, 1); bail: if (inode) iput(inode); @@ -1276,7 +1276,7 @@ static int ocfs2_queue_orphans(struct ocfs2_super *osb, } mutex_lock(&orphan_dir_inode->i_mutex); - status = ocfs2_meta_lock(orphan_dir_inode, NULL, 0); + status = ocfs2_inode_lock(orphan_dir_inode, NULL, 0); if (status < 0) { mlog_errno(status); goto out; @@ -1292,7 +1292,7 @@ static int ocfs2_queue_orphans(struct ocfs2_super *osb, *head = priv.head; out_cluster: - ocfs2_meta_unlock(orphan_dir_inode, 0); + ocfs2_inode_unlock(orphan_dir_inode, 0); out: mutex_unlock(&orphan_dir_inode->i_mutex); iput(orphan_dir_inode); diff --git a/fs/ocfs2/localalloc.c b/fs/ocfs2/localalloc.c index 58ea88b5af36..0de0792fce7f 100644 --- a/fs/ocfs2/localalloc.c +++ b/fs/ocfs2/localalloc.c @@ -231,7 +231,7 @@ void ocfs2_shutdown_local_alloc(struct ocfs2_super *osb) mutex_lock(&main_bm_inode->i_mutex); - status = ocfs2_meta_lock(main_bm_inode, &main_bm_bh, 1); + status = ocfs2_inode_lock(main_bm_inode, &main_bm_bh, 1); if (status < 0) { mlog_errno(status); goto out_mutex; @@ -286,7 +286,7 @@ out_unlock: if (main_bm_bh) brelse(main_bm_bh); - ocfs2_meta_unlock(main_bm_inode, 1); + ocfs2_inode_unlock(main_bm_inode, 1); out_mutex: mutex_unlock(&main_bm_inode->i_mutex); @@ -399,7 +399,7 @@ int ocfs2_complete_local_alloc_recovery(struct ocfs2_super *osb, mutex_lock(&main_bm_inode->i_mutex); - status = ocfs2_meta_lock(main_bm_inode, &main_bm_bh, 1); + status = ocfs2_inode_lock(main_bm_inode, &main_bm_bh, 1); if (status < 0) { mlog_errno(status); goto out_mutex; @@ -424,7 +424,7 @@ int ocfs2_complete_local_alloc_recovery(struct ocfs2_super *osb, ocfs2_commit_trans(osb, handle); out_unlock: - ocfs2_meta_unlock(main_bm_inode, 1); + ocfs2_inode_unlock(main_bm_inode, 1); out_mutex: mutex_unlock(&main_bm_inode->i_mutex); diff --git a/fs/ocfs2/mmap.c b/fs/ocfs2/mmap.c index a7f0ccc6fdd8..3dc18d67557c 100644 --- a/fs/ocfs2/mmap.c +++ b/fs/ocfs2/mmap.c @@ -168,7 +168,7 @@ static int ocfs2_page_mkwrite(struct vm_area_struct *vma, struct page *page) * node. Taking the data lock will also ensure that we don't * attempt page truncation as part of a downconvert. */ - ret = ocfs2_meta_lock(inode, &di_bh, 1); + ret = ocfs2_inode_lock(inode, &di_bh, 1); if (ret < 0) { mlog_errno(ret); goto out; @@ -186,7 +186,7 @@ static int ocfs2_page_mkwrite(struct vm_area_struct *vma, struct page *page) up_write(&OCFS2_I(inode)->ip_alloc_sem); brelse(di_bh); - ocfs2_meta_unlock(inode, 1); + ocfs2_inode_unlock(inode, 1); out: ret2 = ocfs2_vm_op_unblock_sigs(&oldset); @@ -205,13 +205,13 @@ int ocfs2_mmap(struct file *file, struct vm_area_struct *vma) { int ret = 0, lock_level = 0; - ret = ocfs2_meta_lock_atime(file->f_dentry->d_inode, + ret = ocfs2_inode_lock_atime(file->f_dentry->d_inode, file->f_vfsmnt, &lock_level); if (ret < 0) { mlog_errno(ret); goto out; } - ocfs2_meta_unlock(file->f_dentry->d_inode, lock_level); + ocfs2_inode_unlock(file->f_dentry->d_inode, lock_level); out: vma->vm_ops = &ocfs2_file_vm_ops; vma->vm_flags |= VM_CAN_NONLINEAR; diff --git a/fs/ocfs2/namei.c b/fs/ocfs2/namei.c index 6295fd6ae469..74018caf8053 100644 --- a/fs/ocfs2/namei.c +++ b/fs/ocfs2/namei.c @@ -115,7 +115,7 @@ static struct dentry *ocfs2_lookup(struct inode *dir, struct dentry *dentry, mlog(0, "find name %.*s in directory %llu\n", dentry->d_name.len, dentry->d_name.name, (unsigned long long)OCFS2_I(dir)->ip_blkno); - status = ocfs2_meta_lock(dir, NULL, 0); + status = ocfs2_inode_lock(dir, NULL, 0); if (status < 0) { if (status != -ENOENT) mlog_errno(status); @@ -176,7 +176,7 @@ bail_unlock: * unlink on another node will message us to remove that * dentry under this lock so otherwise we can race this with * the downconvert thread and have a stale dentry. */ - ocfs2_meta_unlock(dir, 0); + ocfs2_inode_unlock(dir, 0); bail: @@ -208,7 +208,7 @@ static int ocfs2_mknod(struct inode *dir, /* get our super block */ osb = OCFS2_SB(dir->i_sb); - status = ocfs2_meta_lock(dir, &parent_fe_bh, 1); + status = ocfs2_inode_lock(dir, &parent_fe_bh, 1); if (status < 0) { if (status != -ENOENT) mlog_errno(status); @@ -322,7 +322,7 @@ leave: if (handle) ocfs2_commit_trans(osb, handle); - ocfs2_meta_unlock(dir, 1); + ocfs2_inode_unlock(dir, 1); if (status == -ENOSPC) mlog(0, "Disk is full\n"); @@ -552,7 +552,7 @@ static int ocfs2_link(struct dentry *old_dentry, if (S_ISDIR(inode->i_mode)) return -EPERM; - err = ocfs2_meta_lock(dir, &parent_fe_bh, 1); + err = ocfs2_inode_lock(dir, &parent_fe_bh, 1); if (err < 0) { if (err != -ENOENT) mlog_errno(err); @@ -577,7 +577,7 @@ static int ocfs2_link(struct dentry *old_dentry, goto out; } - err = ocfs2_meta_lock(inode, &fe_bh, 1); + err = ocfs2_inode_lock(inode, &fe_bh, 1); if (err < 0) { if (err != -ENOENT) mlog_errno(err); @@ -642,10 +642,10 @@ static int ocfs2_link(struct dentry *old_dentry, out_commit: ocfs2_commit_trans(osb, handle); out_unlock_inode: - ocfs2_meta_unlock(inode, 1); + ocfs2_inode_unlock(inode, 1); out: - ocfs2_meta_unlock(dir, 1); + ocfs2_inode_unlock(dir, 1); if (de_bh) brelse(de_bh); @@ -719,7 +719,7 @@ static int ocfs2_unlink(struct inode *dir, return -EPERM; } - status = ocfs2_meta_lock(dir, &parent_node_bh, 1); + status = ocfs2_inode_lock(dir, &parent_node_bh, 1); if (status < 0) { if (status != -ENOENT) mlog_errno(status); @@ -744,7 +744,7 @@ static int ocfs2_unlink(struct inode *dir, goto leave; } - status = ocfs2_meta_lock(inode, &fe_bh, 1); + status = ocfs2_inode_lock(inode, &fe_bh, 1); if (status < 0) { if (status != -ENOENT) mlog_errno(status); @@ -840,13 +840,13 @@ leave: ocfs2_commit_trans(osb, handle); if (child_locked) - ocfs2_meta_unlock(inode, 1); + ocfs2_inode_unlock(inode, 1); - ocfs2_meta_unlock(dir, 1); + ocfs2_inode_unlock(dir, 1); if (orphan_dir) { /* This was locked for us in ocfs2_prepare_orphan_dir() */ - ocfs2_meta_unlock(orphan_dir, 1); + ocfs2_inode_unlock(orphan_dir, 1); mutex_unlock(&orphan_dir->i_mutex); iput(orphan_dir); } @@ -907,7 +907,7 @@ static int ocfs2_double_lock(struct ocfs2_super *osb, inode1 = tmpinode; } /* lock id2 */ - status = ocfs2_meta_lock(inode2, bh2, 1); + status = ocfs2_inode_lock(inode2, bh2, 1); if (status < 0) { if (status != -ENOENT) mlog_errno(status); @@ -916,14 +916,14 @@ static int ocfs2_double_lock(struct ocfs2_super *osb, } /* lock id1 */ - status = ocfs2_meta_lock(inode1, bh1, 1); + status = ocfs2_inode_lock(inode1, bh1, 1); if (status < 0) { /* * An error return must mean that no cluster locks * were held on function exit. */ if (oi1->ip_blkno != oi2->ip_blkno) - ocfs2_meta_unlock(inode2, 1); + ocfs2_inode_unlock(inode2, 1); if (status != -ENOENT) mlog_errno(status); @@ -936,10 +936,10 @@ bail: static void ocfs2_double_unlock(struct inode *inode1, struct inode *inode2) { - ocfs2_meta_unlock(inode1, 1); + ocfs2_inode_unlock(inode1, 1); if (inode1 != inode2) - ocfs2_meta_unlock(inode2, 1); + ocfs2_inode_unlock(inode2, 1); } static int ocfs2_rename(struct inode *old_dir, @@ -1034,7 +1034,7 @@ static int ocfs2_rename(struct inode *old_dir, * won't have to concurrently downconvert the inode and the * dentry locks. */ - status = ocfs2_meta_lock(old_inode, &old_inode_bh, 1); + status = ocfs2_inode_lock(old_inode, &old_inode_bh, 1); if (status < 0) { if (status != -ENOENT) mlog_errno(status); @@ -1143,7 +1143,7 @@ static int ocfs2_rename(struct inode *old_dir, goto bail; } - status = ocfs2_meta_lock(new_inode, &newfe_bh, 1); + status = ocfs2_inode_lock(new_inode, &newfe_bh, 1); if (status < 0) { if (status != -ENOENT) mlog_errno(status); @@ -1355,14 +1355,14 @@ bail: ocfs2_double_unlock(old_dir, new_dir); if (old_child_locked) - ocfs2_meta_unlock(old_inode, 1); + ocfs2_inode_unlock(old_inode, 1); if (new_child_locked) - ocfs2_meta_unlock(new_inode, 1); + ocfs2_inode_unlock(new_inode, 1); if (orphan_dir) { /* This was locked for us in ocfs2_prepare_orphan_dir() */ - ocfs2_meta_unlock(orphan_dir, 1); + ocfs2_inode_unlock(orphan_dir, 1); mutex_unlock(&orphan_dir->i_mutex); iput(orphan_dir); } @@ -1530,7 +1530,7 @@ static int ocfs2_symlink(struct inode *dir, credits = ocfs2_calc_symlink_credits(sb); /* lock the parent directory */ - status = ocfs2_meta_lock(dir, &parent_fe_bh, 1); + status = ocfs2_inode_lock(dir, &parent_fe_bh, 1); if (status < 0) { if (status != -ENOENT) mlog_errno(status); @@ -1657,7 +1657,7 @@ bail: if (handle) ocfs2_commit_trans(osb, handle); - ocfs2_meta_unlock(dir, 1); + ocfs2_inode_unlock(dir, 1); if (new_fe_bh) brelse(new_fe_bh); @@ -1735,7 +1735,7 @@ static int ocfs2_prepare_orphan_dir(struct ocfs2_super *osb, mutex_lock(&orphan_dir_inode->i_mutex); - status = ocfs2_meta_lock(orphan_dir_inode, &orphan_dir_bh, 1); + status = ocfs2_inode_lock(orphan_dir_inode, &orphan_dir_bh, 1); if (status < 0) { mlog_errno(status); goto leave; @@ -1745,7 +1745,7 @@ static int ocfs2_prepare_orphan_dir(struct ocfs2_super *osb, orphan_dir_bh, name, OCFS2_ORPHAN_NAMELEN, de_bh); if (status < 0) { - ocfs2_meta_unlock(orphan_dir_inode, 1); + ocfs2_inode_unlock(orphan_dir_inode, 1); mlog_errno(status); goto leave; diff --git a/fs/ocfs2/suballoc.c b/fs/ocfs2/suballoc.c index 8f09f5235e3a..6df4dbf67d18 100644 --- a/fs/ocfs2/suballoc.c +++ b/fs/ocfs2/suballoc.c @@ -114,7 +114,7 @@ void ocfs2_free_alloc_context(struct ocfs2_alloc_context *ac) if (inode) { if (ac->ac_which != OCFS2_AC_USE_LOCAL) - ocfs2_meta_unlock(inode, 1); + ocfs2_inode_unlock(inode, 1); mutex_unlock(&inode->i_mutex); @@ -412,7 +412,7 @@ static int ocfs2_reserve_suballoc_bits(struct ocfs2_super *osb, mutex_lock(&alloc_inode->i_mutex); - status = ocfs2_meta_lock(alloc_inode, &bh, 1); + status = ocfs2_inode_lock(alloc_inode, &bh, 1); if (status < 0) { mutex_unlock(&alloc_inode->i_mutex); iput(alloc_inode); diff --git a/fs/ocfs2/super.c b/fs/ocfs2/super.c index 064eba074f1e..7708df36e223 100644 --- a/fs/ocfs2/super.c +++ b/fs/ocfs2/super.c @@ -964,7 +964,7 @@ static int ocfs2_statfs(struct dentry *dentry, struct kstatfs *buf) goto bail; } - status = ocfs2_meta_lock(inode, &bh, 0); + status = ocfs2_inode_lock(inode, &bh, 0); if (status < 0) { mlog_errno(status); goto bail; @@ -988,7 +988,7 @@ static int ocfs2_statfs(struct dentry *dentry, struct kstatfs *buf) brelse(bh); - ocfs2_meta_unlock(inode, 0); + ocfs2_inode_unlock(inode, 0); status = 0; bail: if (inode) @@ -1019,7 +1019,7 @@ static void ocfs2_inode_init_once(struct kmem_cache *cachep, void *data) oi->ip_clusters = 0; ocfs2_lock_res_init_once(&oi->ip_rw_lockres); - ocfs2_lock_res_init_once(&oi->ip_meta_lockres); + ocfs2_lock_res_init_once(&oi->ip_inode_lockres); ocfs2_lock_res_init_once(&oi->ip_open_lockres); ocfs2_metadata_cache_init(&oi->vfs_inode); -- cgit v1.2.3 From cf8e06f1a860d8680d6bb4ac8ec7d7724988e46f Mon Sep 17 00:00:00 2001 From: Mark Fasheh Date: Thu, 20 Dec 2007 16:43:10 -0800 Subject: [PATCH 1/2] ocfs2: add flock lock type This adds a new dlmglue lock type which is intended to back flock() requests. Since these locks are driven from userspace, usage rules are much more liberal than the typical Ocfs2 internal cluster lock. As a result, we can't make use of most dlmglue features - lock caching and lock level optimizations in particular. Additionally, userspace is free to deadlock itself, so we have to deal with that in the same way as the rest of the kernel - by allowing a signal to abort a lock request. In order to keep ocfs2_cluster_lock() complexity down, ocfs2_file_lock() does it's own dlm coordination. We still use the same helper functions though, so duplicated code is kept to a minimum. Signed-off-by: Mark Fasheh --- fs/ocfs2/dlmglue.c | 267 ++++++++++++++++++++++++++++++++++++++++++++++++ fs/ocfs2/dlmglue.h | 5 + fs/ocfs2/file.h | 6 ++ fs/ocfs2/ocfs2.h | 1 + fs/ocfs2/ocfs2_lockid.h | 5 + 5 files changed, 284 insertions(+) (limited to 'fs/ocfs2/dlmglue.h') diff --git a/fs/ocfs2/dlmglue.c b/fs/ocfs2/dlmglue.c index fa5e3bdc295d..3867244fb144 100644 --- a/fs/ocfs2/dlmglue.c +++ b/fs/ocfs2/dlmglue.c @@ -68,6 +68,7 @@ struct ocfs2_mask_waiter { static struct ocfs2_super *ocfs2_get_dentry_osb(struct ocfs2_lock_res *lockres); static struct ocfs2_super *ocfs2_get_inode_osb(struct ocfs2_lock_res *lockres); +static struct ocfs2_super *ocfs2_get_file_osb(struct ocfs2_lock_res *lockres); /* * Return value from ->downconvert_worker functions. @@ -252,6 +253,11 @@ static struct ocfs2_lock_res_ops ocfs2_inode_open_lops = { .flags = 0, }; +static struct ocfs2_lock_res_ops ocfs2_flock_lops = { + .get_osb = ocfs2_get_file_osb, + .flags = 0, +}; + static inline int ocfs2_is_inode_lock(struct ocfs2_lock_res *lockres) { return lockres->l_type == OCFS2_LOCK_TYPE_META || @@ -310,6 +316,17 @@ static int ocfs2_inode_lock_update(struct inode *inode, struct buffer_head **bh); static void ocfs2_drop_osb_locks(struct ocfs2_super *osb); static inline int ocfs2_highest_compat_lock_level(int level); +static void ocfs2_prepare_downconvert(struct ocfs2_lock_res *lockres, + int new_level); +static int ocfs2_downconvert_lock(struct ocfs2_super *osb, + struct ocfs2_lock_res *lockres, + int new_level, + int lvb); +static int ocfs2_prepare_cancel_convert(struct ocfs2_super *osb, + struct ocfs2_lock_res *lockres); +static int ocfs2_cancel_convert(struct ocfs2_super *osb, + struct ocfs2_lock_res *lockres); + static void ocfs2_build_lock_name(enum ocfs2_lock_type type, u64 blkno, @@ -419,6 +436,13 @@ static struct ocfs2_super *ocfs2_get_inode_osb(struct ocfs2_lock_res *lockres) return OCFS2_SB(inode->i_sb); } +static struct ocfs2_super *ocfs2_get_file_osb(struct ocfs2_lock_res *lockres) +{ + struct ocfs2_file_private *fp = lockres->l_priv; + + return OCFS2_SB(fp->fp_file->f_mapping->host->i_sb); +} + static __u64 ocfs2_get_dentry_lock_ino(struct ocfs2_lock_res *lockres) { __be64 inode_blkno_be; @@ -499,6 +523,21 @@ static void ocfs2_rename_lock_res_init(struct ocfs2_lock_res *res, &ocfs2_rename_lops, osb); } +void ocfs2_file_lock_res_init(struct ocfs2_lock_res *lockres, + struct ocfs2_file_private *fp) +{ + struct inode *inode = fp->fp_file->f_mapping->host; + struct ocfs2_inode_info *oi = OCFS2_I(inode); + + ocfs2_lock_res_init_once(lockres); + ocfs2_build_lock_name(OCFS2_LOCK_TYPE_FLOCK, oi->ip_blkno, + inode->i_generation, lockres->l_name); + ocfs2_lock_res_init_common(OCFS2_SB(inode->i_sb), lockres, + OCFS2_LOCK_TYPE_FLOCK, &ocfs2_flock_lops, + fp); + lockres->l_flags |= OCFS2_LOCK_NOCACHE; +} + void ocfs2_lock_res_free(struct ocfs2_lock_res *res) { mlog_entry_void(); @@ -715,6 +754,13 @@ static void ocfs2_blocking_ast(void *opaque, int level) lockres->l_name, level, lockres->l_level, ocfs2_lock_type_string(lockres->l_type)); + /* + * We can skip the bast for locks which don't enable caching - + * they'll be dropped at the earliest possible time anyway. + */ + if (lockres->l_flags & OCFS2_LOCK_NOCACHE) + return; + spin_lock_irqsave(&lockres->l_lock, flags); needs_downconvert = ocfs2_generic_handle_bast(lockres, level); if (needs_downconvert) @@ -926,6 +972,21 @@ static int lockres_remove_mask_waiter(struct ocfs2_lock_res *lockres, } +static int ocfs2_wait_for_mask_interruptible(struct ocfs2_mask_waiter *mw, + struct ocfs2_lock_res *lockres) +{ + int ret; + + ret = wait_for_completion_interruptible(&mw->mw_complete); + if (ret) + lockres_remove_mask_waiter(lockres, mw); + else + ret = mw->mw_status; + /* Re-arm the completion in case we want to wait on it again */ + INIT_COMPLETION(mw->mw_complete); + return ret; +} + static int ocfs2_cluster_lock(struct ocfs2_super *osb, struct ocfs2_lock_res *lockres, int level, @@ -1296,6 +1357,212 @@ out: mlog_exit_void(); } +static int ocfs2_flock_handle_signal(struct ocfs2_lock_res *lockres, + int level) +{ + int ret; + struct ocfs2_super *osb = ocfs2_get_lockres_osb(lockres); + unsigned long flags; + struct ocfs2_mask_waiter mw; + + ocfs2_init_mask_waiter(&mw); + +retry_cancel: + spin_lock_irqsave(&lockres->l_lock, flags); + if (lockres->l_flags & OCFS2_LOCK_BUSY) { + ret = ocfs2_prepare_cancel_convert(osb, lockres); + if (ret) { + spin_unlock_irqrestore(&lockres->l_lock, flags); + ret = ocfs2_cancel_convert(osb, lockres); + if (ret < 0) { + mlog_errno(ret); + goto out; + } + goto retry_cancel; + } + lockres_add_mask_waiter(lockres, &mw, OCFS2_LOCK_BUSY, 0); + spin_unlock_irqrestore(&lockres->l_lock, flags); + + ocfs2_wait_for_mask(&mw); + goto retry_cancel; + } + + ret = -ERESTARTSYS; + /* + * We may still have gotten the lock, in which case there's no + * point to restarting the syscall. + */ + if (lockres->l_level == level) + ret = 0; + + mlog(0, "Cancel returning %d. flags: 0x%lx, level: %d, act: %d\n", ret, + lockres->l_flags, lockres->l_level, lockres->l_action); + + spin_unlock_irqrestore(&lockres->l_lock, flags); + +out: + return ret; +} + +/* + * ocfs2_file_lock() and ocfs2_file_unlock() map to a single pair of + * flock() calls. The locking approach this requires is sufficiently + * different from all other cluster lock types that we implement a + * seperate path to the "low-level" dlm calls. In particular: + * + * - No optimization of lock levels is done - we take at exactly + * what's been requested. + * + * - No lock caching is employed. We immediately downconvert to + * no-lock at unlock time. This also means flock locks never go on + * the blocking list). + * + * - Since userspace can trivially deadlock itself with flock, we make + * sure to allow cancellation of a misbehaving applications flock() + * request. + * + * - Access to any flock lockres doesn't require concurrency, so we + * can simplify the code by requiring the caller to guarantee + * serialization of dlmglue flock calls. + */ +int ocfs2_file_lock(struct file *file, int ex, int trylock) +{ + int ret, level = ex ? LKM_EXMODE : LKM_PRMODE; + unsigned int lkm_flags = trylock ? LKM_NOQUEUE : 0; + unsigned long flags; + struct ocfs2_file_private *fp = file->private_data; + struct ocfs2_lock_res *lockres = &fp->fp_flock; + struct ocfs2_super *osb = OCFS2_SB(file->f_mapping->host->i_sb); + struct ocfs2_mask_waiter mw; + + ocfs2_init_mask_waiter(&mw); + + if ((lockres->l_flags & OCFS2_LOCK_BUSY) || + (lockres->l_level > LKM_NLMODE)) { + mlog(ML_ERROR, + "File lock \"%s\" has busy or locked state: flags: 0x%lx, " + "level: %u\n", lockres->l_name, lockres->l_flags, + lockres->l_level); + return -EINVAL; + } + + spin_lock_irqsave(&lockres->l_lock, flags); + if (!(lockres->l_flags & OCFS2_LOCK_ATTACHED)) { + lockres_add_mask_waiter(lockres, &mw, OCFS2_LOCK_BUSY, 0); + spin_unlock_irqrestore(&lockres->l_lock, flags); + + /* + * Get the lock at NLMODE to start - that way we + * can cancel the upconvert request if need be. + */ + ret = ocfs2_lock_create(osb, lockres, LKM_NLMODE, 0); + if (ret < 0) { + mlog_errno(ret); + goto out; + } + + ret = ocfs2_wait_for_mask(&mw); + if (ret) { + mlog_errno(ret); + goto out; + } + spin_lock_irqsave(&lockres->l_lock, flags); + } + + lockres->l_action = OCFS2_AST_CONVERT; + lkm_flags |= LKM_CONVERT; + lockres->l_requested = level; + lockres_or_flags(lockres, OCFS2_LOCK_BUSY); + + lockres_add_mask_waiter(lockres, &mw, OCFS2_LOCK_BUSY, 0); + spin_unlock_irqrestore(&lockres->l_lock, flags); + + ret = dlmlock(osb->dlm, level, &lockres->l_lksb, lkm_flags, + lockres->l_name, OCFS2_LOCK_ID_MAX_LEN - 1, + ocfs2_locking_ast, lockres, ocfs2_blocking_ast); + if (ret != DLM_NORMAL) { + if (trylock && ret == DLM_NOTQUEUED) + ret = -EAGAIN; + else { + ocfs2_log_dlm_error("dlmlock", ret, lockres); + ret = -EINVAL; + } + + ocfs2_recover_from_dlm_error(lockres, 1); + lockres_remove_mask_waiter(lockres, &mw); + goto out; + } + + ret = ocfs2_wait_for_mask_interruptible(&mw, lockres); + if (ret == -ERESTARTSYS) { + /* + * Userspace can cause deadlock itself with + * flock(). Current behavior locally is to allow the + * deadlock, but abort the system call if a signal is + * received. We follow this example, otherwise a + * poorly written program could sit in kernel until + * reboot. + * + * Handling this is a bit more complicated for Ocfs2 + * though. We can't exit this function with an + * outstanding lock request, so a cancel convert is + * required. We intentionally overwrite 'ret' - if the + * cancel fails and the lock was granted, it's easier + * to just bubble sucess back up to the user. + */ + ret = ocfs2_flock_handle_signal(lockres, level); + } + +out: + + mlog(0, "Lock: \"%s\" ex: %d, trylock: %d, returns: %d\n", + lockres->l_name, ex, trylock, ret); + return ret; +} + +void ocfs2_file_unlock(struct file *file) +{ + int ret; + unsigned long flags; + struct ocfs2_file_private *fp = file->private_data; + struct ocfs2_lock_res *lockres = &fp->fp_flock; + struct ocfs2_super *osb = OCFS2_SB(file->f_mapping->host->i_sb); + struct ocfs2_mask_waiter mw; + + ocfs2_init_mask_waiter(&mw); + + if (!(lockres->l_flags & OCFS2_LOCK_ATTACHED)) + return; + + if (lockres->l_level == LKM_NLMODE) + return; + + mlog(0, "Unlock: \"%s\" flags: 0x%lx, level: %d, act: %d\n", + lockres->l_name, lockres->l_flags, lockres->l_level, + lockres->l_action); + + spin_lock_irqsave(&lockres->l_lock, flags); + /* + * Fake a blocking ast for the downconvert code. + */ + lockres_or_flags(lockres, OCFS2_LOCK_BLOCKED); + lockres->l_blocking = LKM_EXMODE; + + ocfs2_prepare_downconvert(lockres, LKM_NLMODE); + lockres_add_mask_waiter(lockres, &mw, OCFS2_LOCK_BUSY, 0); + spin_unlock_irqrestore(&lockres->l_lock, flags); + + ret = ocfs2_downconvert_lock(osb, lockres, LKM_NLMODE, 0); + if (ret) { + mlog_errno(ret); + return; + } + + ret = ocfs2_wait_for_mask(&mw); + if (ret) + mlog_errno(ret); +} + static void ocfs2_downconvert_on_unlock(struct ocfs2_super *osb, struct ocfs2_lock_res *lockres) { diff --git a/fs/ocfs2/dlmglue.h b/fs/ocfs2/dlmglue.h index 6dcbc944e8ce..5f17243ba501 100644 --- a/fs/ocfs2/dlmglue.h +++ b/fs/ocfs2/dlmglue.h @@ -66,6 +66,9 @@ void ocfs2_inode_lock_res_init(struct ocfs2_lock_res *res, struct inode *inode); void ocfs2_dentry_lock_res_init(struct ocfs2_dentry_lock *dl, u64 parent, struct inode *inode); +struct ocfs2_file_private; +void ocfs2_file_lock_res_init(struct ocfs2_lock_res *lockres, + struct ocfs2_file_private *fp); void ocfs2_lock_res_free(struct ocfs2_lock_res *res); int ocfs2_create_new_inode_locks(struct inode *inode); int ocfs2_drop_inode_locks(struct inode *inode); @@ -98,6 +101,8 @@ int ocfs2_rename_lock(struct ocfs2_super *osb); void ocfs2_rename_unlock(struct ocfs2_super *osb); int ocfs2_dentry_lock(struct dentry *dentry, int ex); void ocfs2_dentry_unlock(struct dentry *dentry, int ex); +int ocfs2_file_lock(struct file *file, int ex, int trylock); +void ocfs2_file_unlock(struct file *file); void ocfs2_mark_lockres_freeing(struct ocfs2_lock_res *lockres); void ocfs2_simple_drop_lockres(struct ocfs2_super *osb, diff --git a/fs/ocfs2/file.h b/fs/ocfs2/file.h index 066f14add3a8..048ddcaf5c80 100644 --- a/fs/ocfs2/file.h +++ b/fs/ocfs2/file.h @@ -32,6 +32,12 @@ extern const struct inode_operations ocfs2_file_iops; extern const struct inode_operations ocfs2_special_file_iops; struct ocfs2_alloc_context; +struct ocfs2_file_private { + struct file *fp_file; + struct mutex fp_mutex; + struct ocfs2_lock_res fp_flock; +}; + enum ocfs2_alloc_restarted { RESTART_NONE = 0, RESTART_TRANS, diff --git a/fs/ocfs2/ocfs2.h b/fs/ocfs2/ocfs2.h index d12bd7036da7..63c131e1cc77 100644 --- a/fs/ocfs2/ocfs2.h +++ b/fs/ocfs2/ocfs2.h @@ -101,6 +101,7 @@ enum ocfs2_unlock_action { * about to be * dropped. */ #define OCFS2_LOCK_QUEUED (0x00000100) /* queued for downconvert */ +#define OCFS2_LOCK_NOCACHE (0x00000200) /* don't use a holder count */ struct ocfs2_lock_res_ops; diff --git a/fs/ocfs2/ocfs2_lockid.h b/fs/ocfs2/ocfs2_lockid.h index 4ca02b1c38ac..86f3e3799c2b 100644 --- a/fs/ocfs2/ocfs2_lockid.h +++ b/fs/ocfs2/ocfs2_lockid.h @@ -45,6 +45,7 @@ enum ocfs2_lock_type { OCFS2_LOCK_TYPE_RW, OCFS2_LOCK_TYPE_DENTRY, OCFS2_LOCK_TYPE_OPEN, + OCFS2_LOCK_TYPE_FLOCK, OCFS2_NUM_LOCK_TYPES }; @@ -73,6 +74,9 @@ static inline char ocfs2_lock_type_char(enum ocfs2_lock_type type) case OCFS2_LOCK_TYPE_OPEN: c = 'O'; break; + case OCFS2_LOCK_TYPE_FLOCK: + c = 'F'; + break; default: c = '\0'; } @@ -90,6 +94,7 @@ static char *ocfs2_lock_type_strings[] = { [OCFS2_LOCK_TYPE_RW] = "Write/Read", [OCFS2_LOCK_TYPE_DENTRY] = "Dentry", [OCFS2_LOCK_TYPE_OPEN] = "Open", + [OCFS2_LOCK_TYPE_FLOCK] = "Flock", }; static inline const char *ocfs2_lock_type_string(enum ocfs2_lock_type type) -- cgit v1.2.3