From b449fc6fcc07a392c69f3c1db9a4ad4dda8cbcba Mon Sep 17 00:00:00 2001 From: Andreas Dilger Date: Thu, 30 Jul 2009 20:09:46 +0200 Subject: JBD: round commit timer up to avoid uncommitted transaction Fix jiffie rounding in jbd commit timer setup code. Rounding down could cause the timer to be fired before the corresponding transaction has expired. That transaction can stay not committed forever if no new transaction is created or explicit sync/umount happens. Signed-off-by: Andreas Dilger Signed-off-by: Jan Kara --- fs/jbd/transaction.c | 3 ++- 1 file changed, 2 insertions(+), 1 deletion(-) (limited to 'fs') diff --git a/fs/jbd/transaction.c b/fs/jbd/transaction.c index c03ac11f74be..833c1675a00c 100644 --- a/fs/jbd/transaction.c +++ b/fs/jbd/transaction.c @@ -56,7 +56,8 @@ get_transaction(journal_t *journal, transaction_t *transaction) spin_lock_init(&transaction->t_handle_lock); /* Set up the commit timer for the new transaction. */ - journal->j_commit_timer.expires = round_jiffies(transaction->t_expires); + journal->j_commit_timer.expires = + round_jiffies_up(transaction->t_expires); add_timer(&journal->j_commit_timer); J_ASSERT(journal->j_running_transaction == NULL); -- cgit v1.2.3-59-g8ed1b From 9c28cbccec66a5ca292c6659bf5a0fe0c8459fa7 Mon Sep 17 00:00:00 2001 From: Jan Kara Date: Mon, 3 Aug 2009 19:21:00 +0200 Subject: jbd: Journal block numbers can ever be only 32-bit use unsigned int for them It does not make sense to store block number for journal as unsigned long since they can be only 32-bit (because of on-disk format limitation). So change in-memory structures and variables to use unsigned int instead. Signed-off-by: Jan Kara --- fs/jbd/checkpoint.c | 6 +++--- fs/jbd/commit.c | 2 +- fs/jbd/journal.c | 30 +++++++++++++++--------------- fs/jbd/recovery.c | 18 +++++++++--------- fs/jbd/revoke.c | 16 ++++++++-------- include/linux/jbd.h | 26 +++++++++++++------------- 6 files changed, 49 insertions(+), 49 deletions(-) (limited to 'fs') diff --git a/fs/jbd/checkpoint.c b/fs/jbd/checkpoint.c index 61f32f3868cd..b0435dd0654d 100644 --- a/fs/jbd/checkpoint.c +++ b/fs/jbd/checkpoint.c @@ -456,7 +456,7 @@ int cleanup_journal_tail(journal_t *journal) { transaction_t * transaction; tid_t first_tid; - unsigned long blocknr, freed; + unsigned int blocknr, freed; if (is_journal_aborted(journal)) return 1; @@ -502,8 +502,8 @@ int cleanup_journal_tail(journal_t *journal) freed = freed + journal->j_last - journal->j_first; jbd_debug(1, - "Cleaning journal tail from %d to %d (offset %lu), " - "freeing %lu\n", + "Cleaning journal tail from %d to %d (offset %u), " + "freeing %u\n", journal->j_tail_sequence, first_tid, blocknr, freed); journal->j_free += freed; diff --git a/fs/jbd/commit.c b/fs/jbd/commit.c index 618e21c0b7a3..4bd882548c45 100644 --- a/fs/jbd/commit.c +++ b/fs/jbd/commit.c @@ -308,7 +308,7 @@ void journal_commit_transaction(journal_t *journal) int bufs; int flags; int err; - unsigned long blocknr; + unsigned int blocknr; ktime_t start_time; u64 commit_time; char *tagp = NULL; diff --git a/fs/jbd/journal.c b/fs/jbd/journal.c index f96f85092d1c..bd3c073b485d 100644 --- a/fs/jbd/journal.c +++ b/fs/jbd/journal.c @@ -276,7 +276,7 @@ static void journal_kill_thread(journal_t *journal) int journal_write_metadata_buffer(transaction_t *transaction, struct journal_head *jh_in, struct journal_head **jh_out, - unsigned long blocknr) + unsigned int blocknr) { int need_copy_out = 0; int done_copy_out = 0; @@ -567,9 +567,9 @@ int log_wait_commit(journal_t *journal, tid_t tid) * Log buffer allocation routines: */ -int journal_next_log_block(journal_t *journal, unsigned long *retp) +int journal_next_log_block(journal_t *journal, unsigned int *retp) { - unsigned long blocknr; + unsigned int blocknr; spin_lock(&journal->j_state_lock); J_ASSERT(journal->j_free > 1); @@ -590,11 +590,11 @@ int journal_next_log_block(journal_t *journal, unsigned long *retp) * this is a no-op. If needed, we can use j_blk_offset - everything is * ready. */ -int journal_bmap(journal_t *journal, unsigned long blocknr, - unsigned long *retp) +int journal_bmap(journal_t *journal, unsigned int blocknr, + unsigned int *retp) { int err = 0; - unsigned long ret; + unsigned int ret; if (journal->j_inode) { ret = bmap(journal->j_inode, blocknr); @@ -604,7 +604,7 @@ int journal_bmap(journal_t *journal, unsigned long blocknr, char b[BDEVNAME_SIZE]; printk(KERN_ALERT "%s: journal block not found " - "at offset %lu on %s\n", + "at offset %u on %s\n", __func__, blocknr, bdevname(journal->j_dev, b)); @@ -630,7 +630,7 @@ int journal_bmap(journal_t *journal, unsigned long blocknr, struct journal_head *journal_get_descriptor_buffer(journal_t *journal) { struct buffer_head *bh; - unsigned long blocknr; + unsigned int blocknr; int err; err = journal_next_log_block(journal, &blocknr); @@ -774,7 +774,7 @@ journal_t * journal_init_inode (struct inode *inode) journal_t *journal = journal_init_common(); int err; int n; - unsigned long blocknr; + unsigned int blocknr; if (!journal) return NULL; @@ -846,12 +846,12 @@ static void journal_fail_superblock (journal_t *journal) static int journal_reset(journal_t *journal) { journal_superblock_t *sb = journal->j_superblock; - unsigned long first, last; + unsigned int first, last; first = be32_to_cpu(sb->s_first); last = be32_to_cpu(sb->s_maxlen); if (first + JFS_MIN_JOURNAL_BLOCKS > last + 1) { - printk(KERN_ERR "JBD: Journal too short (blocks %lu-%lu).\n", + printk(KERN_ERR "JBD: Journal too short (blocks %u-%u).\n", first, last); journal_fail_superblock(journal); return -EINVAL; @@ -885,7 +885,7 @@ static int journal_reset(journal_t *journal) **/ int journal_create(journal_t *journal) { - unsigned long blocknr; + unsigned int blocknr; struct buffer_head *bh; journal_superblock_t *sb; int i, err; @@ -969,14 +969,14 @@ void journal_update_superblock(journal_t *journal, int wait) if (sb->s_start == 0 && journal->j_tail_sequence == journal->j_transaction_sequence) { jbd_debug(1,"JBD: Skipping superblock update on recovered sb " - "(start %ld, seq %d, errno %d)\n", + "(start %u, seq %d, errno %d)\n", journal->j_tail, journal->j_tail_sequence, journal->j_errno); goto out; } spin_lock(&journal->j_state_lock); - jbd_debug(1,"JBD: updating superblock (start %ld, seq %d, errno %d)\n", + jbd_debug(1,"JBD: updating superblock (start %u, seq %d, errno %d)\n", journal->j_tail, journal->j_tail_sequence, journal->j_errno); sb->s_sequence = cpu_to_be32(journal->j_tail_sequence); @@ -1371,7 +1371,7 @@ int journal_flush(journal_t *journal) { int err = 0; transaction_t *transaction = NULL; - unsigned long old_tail; + unsigned int old_tail; spin_lock(&journal->j_state_lock); diff --git a/fs/jbd/recovery.c b/fs/jbd/recovery.c index db5e982c5ddf..cb1a49ae605e 100644 --- a/fs/jbd/recovery.c +++ b/fs/jbd/recovery.c @@ -70,7 +70,7 @@ static int do_readahead(journal_t *journal, unsigned int start) { int err; unsigned int max, nbufs, next; - unsigned long blocknr; + unsigned int blocknr; struct buffer_head *bh; struct buffer_head * bufs[MAXBUF]; @@ -132,7 +132,7 @@ static int jread(struct buffer_head **bhp, journal_t *journal, unsigned int offset) { int err; - unsigned long blocknr; + unsigned int blocknr; struct buffer_head *bh; *bhp = NULL; @@ -314,7 +314,7 @@ static int do_one_pass(journal_t *journal, struct recovery_info *info, enum passtype pass) { unsigned int first_commit_ID, next_commit_ID; - unsigned long next_log_block; + unsigned int next_log_block; int err, success = 0; journal_superblock_t * sb; journal_header_t * tmp; @@ -367,14 +367,14 @@ static int do_one_pass(journal_t *journal, if (tid_geq(next_commit_ID, info->end_transaction)) break; - jbd_debug(2, "Scanning for sequence ID %u at %lu/%lu\n", + jbd_debug(2, "Scanning for sequence ID %u at %u/%u\n", next_commit_ID, next_log_block, journal->j_last); /* Skip over each chunk of the transaction looking * either the next descriptor block or the final commit * record. */ - jbd_debug(3, "JBD: checking block %ld\n", next_log_block); + jbd_debug(3, "JBD: checking block %u\n", next_log_block); err = jread(&bh, journal, next_log_block); if (err) goto failed; @@ -429,7 +429,7 @@ static int do_one_pass(journal_t *journal, tagp = &bh->b_data[sizeof(journal_header_t)]; while ((tagp - bh->b_data +sizeof(journal_block_tag_t)) <= journal->j_blocksize) { - unsigned long io_block; + unsigned int io_block; tag = (journal_block_tag_t *) tagp; flags = be32_to_cpu(tag->t_flags); @@ -443,10 +443,10 @@ static int do_one_pass(journal_t *journal, success = err; printk (KERN_ERR "JBD: IO error %d recovering " - "block %ld in log\n", + "block %u in log\n", err, io_block); } else { - unsigned long blocknr; + unsigned int blocknr; J_ASSERT(obh != NULL); blocknr = be32_to_cpu(tag->t_blocknr); @@ -581,7 +581,7 @@ static int scan_revoke_records(journal_t *journal, struct buffer_head *bh, max = be32_to_cpu(header->r_count); while (offset < max) { - unsigned long blocknr; + unsigned int blocknr; int err; blocknr = be32_to_cpu(* ((__be32 *) (bh->b_data+offset))); diff --git a/fs/jbd/revoke.c b/fs/jbd/revoke.c index da6cd9bdaabc..ad717328343a 100644 --- a/fs/jbd/revoke.c +++ b/fs/jbd/revoke.c @@ -101,7 +101,7 @@ struct jbd_revoke_record_s { struct list_head hash; tid_t sequence; /* Used for recovery only */ - unsigned long blocknr; + unsigned int blocknr; }; @@ -126,7 +126,7 @@ static void flush_descriptor(journal_t *, struct journal_head *, int, int); /* Utility functions to maintain the revoke table */ /* Borrowed from buffer.c: this is a tried and tested block hash function */ -static inline int hash(journal_t *journal, unsigned long block) +static inline int hash(journal_t *journal, unsigned int block) { struct jbd_revoke_table_s *table = journal->j_revoke; int hash_shift = table->hash_shift; @@ -136,7 +136,7 @@ static inline int hash(journal_t *journal, unsigned long block) (block << (hash_shift - 12))) & (table->hash_size - 1); } -static int insert_revoke_hash(journal_t *journal, unsigned long blocknr, +static int insert_revoke_hash(journal_t *journal, unsigned int blocknr, tid_t seq) { struct list_head *hash_list; @@ -166,7 +166,7 @@ oom: /* Find a revoke record in the journal's hash table. */ static struct jbd_revoke_record_s *find_revoke_record(journal_t *journal, - unsigned long blocknr) + unsigned int blocknr) { struct list_head *hash_list; struct jbd_revoke_record_s *record; @@ -332,7 +332,7 @@ void journal_destroy_revoke(journal_t *journal) * by one. */ -int journal_revoke(handle_t *handle, unsigned long blocknr, +int journal_revoke(handle_t *handle, unsigned int blocknr, struct buffer_head *bh_in) { struct buffer_head *bh = NULL; @@ -401,7 +401,7 @@ int journal_revoke(handle_t *handle, unsigned long blocknr, } } - jbd_debug(2, "insert revoke for block %lu, bh_in=%p\n", blocknr, bh_in); + jbd_debug(2, "insert revoke for block %u, bh_in=%p\n", blocknr, bh_in); err = insert_revoke_hash(journal, blocknr, handle->h_transaction->t_tid); BUFFER_TRACE(bh_in, "exit"); @@ -644,7 +644,7 @@ static void flush_descriptor(journal_t *journal, */ int journal_set_revoke(journal_t *journal, - unsigned long blocknr, + unsigned int blocknr, tid_t sequence) { struct jbd_revoke_record_s *record; @@ -668,7 +668,7 @@ int journal_set_revoke(journal_t *journal, */ int journal_test_revoke(journal_t *journal, - unsigned long blocknr, + unsigned int blocknr, tid_t sequence) { struct jbd_revoke_record_s *record; diff --git a/include/linux/jbd.h b/include/linux/jbd.h index c2049a04fa0b..a1187a0c99b4 100644 --- a/include/linux/jbd.h +++ b/include/linux/jbd.h @@ -446,7 +446,7 @@ struct transaction_s /* * Where in the log does this transaction's commit start? [no locking] */ - unsigned long t_log_start; + unsigned int t_log_start; /* Number of buffers on the t_buffers list [j_list_lock] */ int t_nr_buffers; @@ -701,26 +701,26 @@ struct journal_s * Journal head: identifies the first unused block in the journal. * [j_state_lock] */ - unsigned long j_head; + unsigned int j_head; /* * Journal tail: identifies the oldest still-used block in the journal. * [j_state_lock] */ - unsigned long j_tail; + unsigned int j_tail; /* * Journal free: how many free blocks are there in the journal? * [j_state_lock] */ - unsigned long j_free; + unsigned int j_free; /* * Journal start and end: the block numbers of the first usable block * and one beyond the last usable block in the journal. [j_state_lock] */ - unsigned long j_first; - unsigned long j_last; + unsigned int j_first; + unsigned int j_last; /* * Device, blocksize and starting block offset for the location where we @@ -728,7 +728,7 @@ struct journal_s */ struct block_device *j_dev; int j_blocksize; - unsigned long j_blk_offset; + unsigned int j_blk_offset; /* * Device which holds the client fs. For internal journal this will be @@ -859,7 +859,7 @@ extern void __journal_clean_data_list(transaction_t *transaction); /* Log buffer allocation */ extern struct journal_head * journal_get_descriptor_buffer(journal_t *); -int journal_next_log_block(journal_t *, unsigned long *); +int journal_next_log_block(journal_t *, unsigned int *); /* Commit management */ extern void journal_commit_transaction(journal_t *); @@ -874,7 +874,7 @@ extern int journal_write_metadata_buffer(transaction_t *transaction, struct journal_head *jh_in, struct journal_head **jh_out, - unsigned long blocknr); + unsigned int blocknr); /* Transaction locking */ extern void __wait_on_journal (journal_t *); @@ -942,7 +942,7 @@ extern void journal_abort (journal_t *, int); extern int journal_errno (journal_t *); extern void journal_ack_err (journal_t *); extern int journal_clear_err (journal_t *); -extern int journal_bmap(journal_t *, unsigned long, unsigned long *); +extern int journal_bmap(journal_t *, unsigned int, unsigned int *); extern int journal_force_commit(journal_t *); /* @@ -976,14 +976,14 @@ extern int journal_init_revoke_caches(void); extern void journal_destroy_revoke(journal_t *); extern int journal_revoke (handle_t *, - unsigned long, struct buffer_head *); + unsigned int, struct buffer_head *); extern int journal_cancel_revoke(handle_t *, struct journal_head *); extern void journal_write_revoke_records(journal_t *, transaction_t *, int); /* Recovery revoke support */ -extern int journal_set_revoke(journal_t *, unsigned long, tid_t); -extern int journal_test_revoke(journal_t *, unsigned long, tid_t); +extern int journal_set_revoke(journal_t *, unsigned int, tid_t); +extern int journal_test_revoke(journal_t *, unsigned int, tid_t); extern void journal_clear_revoke(journal_t *); extern void journal_switch_revoke_table(journal_t *journal); -- cgit v1.2.3-59-g8ed1b From 3adae9da0b35d2ca908039f42a1e90395c335181 Mon Sep 17 00:00:00 2001 From: Jan Kara Date: Tue, 11 Aug 2009 17:27:21 +0200 Subject: jbd: Annotate transaction start also for journal_restart() lockdep annotation for a transaction start has been at the end of journal_start(). But a transaction is also started from journal_restart(). Move the lockdep annotation to start_this_handle() which covers both cases. Signed-off-by: Jan Kara --- fs/jbd/transaction.c | 6 +++--- 1 file changed, 3 insertions(+), 3 deletions(-) (limited to 'fs') diff --git a/fs/jbd/transaction.c b/fs/jbd/transaction.c index 833c1675a00c..006f9ad838a2 100644 --- a/fs/jbd/transaction.c +++ b/fs/jbd/transaction.c @@ -229,6 +229,8 @@ repeat_locked: __log_space_left(journal)); spin_unlock(&transaction->t_handle_lock); spin_unlock(&journal->j_state_lock); + + lock_map_acquire(&handle->h_lockdep_map); out: if (unlikely(new_transaction)) /* It's usually NULL */ kfree(new_transaction); @@ -293,9 +295,6 @@ handle_t *journal_start(journal_t *journal, int nblocks) handle = ERR_PTR(err); goto out; } - - lock_map_acquire(&handle->h_lockdep_map); - out: return handle; } @@ -417,6 +416,7 @@ int journal_restart(handle_t *handle, int nblocks) __log_start_commit(journal, transaction->t_tid); spin_unlock(&journal->j_state_lock); + lock_map_release(&handle->h_lockdep_map); handle->h_buffer_credits = nblocks; ret = start_this_handle(journal, handle); return ret; -- cgit v1.2.3-59-g8ed1b From 00171d3c7e3b738ba582c7a9b37408e796f49046 Mon Sep 17 00:00:00 2001 From: Jan Kara Date: Tue, 11 Aug 2009 19:06:10 +0200 Subject: ext3: Fix possible deadlock between ext3_truncate() and ext3_get_blocks() During truncate we are sometimes forced to start a new transaction as the amount of blocks to be journaled is both quite large and hard to predict. So far we restarted a transaction while holding truncate_mutex and that violates lock ordering because truncate_mutex ranks below transaction start (and it can lead to a real deadlock with ext3_get_blocks() allocating new blocks from ext3_writepage()). Luckily, the problem is easy to fix: We just drop the truncate_mutex before restarting the transaction and acquire it afterwards. We are safe to do this as by the time ext3_truncate() is called, all the page cache for the truncated part of the file is dropped and so writepage() cannot come and allocate new blocks in the part of the file we are truncating. The rest of writers is stopped by us holding i_mutex. Signed-off-by: Jan Kara --- fs/ext3/inode.c | 19 +++++++++++++++---- 1 file changed, 15 insertions(+), 4 deletions(-) (limited to 'fs') diff --git a/fs/ext3/inode.c b/fs/ext3/inode.c index b49908a167ae..eb0b4e0d517b 100644 --- a/fs/ext3/inode.c +++ b/fs/ext3/inode.c @@ -172,10 +172,21 @@ static int try_to_extend_transaction(handle_t *handle, struct inode *inode) * so before we call here everything must be consistently dirtied against * this transaction. */ -static int ext3_journal_test_restart(handle_t *handle, struct inode *inode) +static int truncate_restart_transaction(handle_t *handle, struct inode *inode) { + int ret; + jbd_debug(2, "restarting handle %p\n", handle); - return ext3_journal_restart(handle, blocks_for_truncate(inode)); + /* + * Drop truncate_mutex to avoid deadlock with ext3_get_blocks_handle + * At this moment, get_block can be called only for blocks inside + * i_size since page cache has been already dropped and writes are + * blocked by i_mutex. So we can safely drop the truncate_mutex. + */ + mutex_unlock(&EXT3_I(inode)->truncate_mutex); + ret = ext3_journal_restart(handle, blocks_for_truncate(inode)); + mutex_lock(&EXT3_I(inode)->truncate_mutex); + return ret; } /* @@ -2072,7 +2083,7 @@ static void ext3_clear_blocks(handle_t *handle, struct inode *inode, ext3_journal_dirty_metadata(handle, bh); } ext3_mark_inode_dirty(handle, inode); - ext3_journal_test_restart(handle, inode); + truncate_restart_transaction(handle, inode); if (bh) { BUFFER_TRACE(bh, "retaking write access"); ext3_journal_get_write_access(handle, bh); @@ -2282,7 +2293,7 @@ static void ext3_free_branches(handle_t *handle, struct inode *inode, return; if (try_to_extend_transaction(handle, inode)) { ext3_mark_inode_dirty(handle, inode); - ext3_journal_test_restart(handle, inode); + truncate_restart_transaction(handle, inode); } ext3_free_blocks(handle, inode, nr, 1); -- cgit v1.2.3-59-g8ed1b From 4f003fd32bc54ec438b8691795279844df27ce38 Mon Sep 17 00:00:00 2001 From: Chris Mason Date: Tue, 8 Sep 2009 00:22:14 +0200 Subject: ext3: Add locking to ext3_do_update_inode I've been struggling with this off and on while I've been testing the data=guarded work. The symptom is corrupted orphan lists and inodes with the wrong i_size stored on disk. I was convinced the data=guarded code was just missing a call to ext3_mark_inode_dirty, but tracing showed the i_disksize I was sending to ext3_mark_inode_dirty wasn't actually making it to the drive. ext3_mark_inode_dirty can be called without locks held (atime updates and a few others), so the data=guarded code uses locks while updating the in-memory inode, and then calls ext3_mark_inode_dirty without any locks held. But, ext3_mark_inode_dirty has no internal locking to make sure that only one CPU is updating the buffer head at a time. Generally this works out ok because everyone that changes the inode then calls ext3_mark_inode_dirty themselves. Even though it races, eventually someone updates the buffer heads and things move on. But there is still a risk of the wrong values getting in, and the data=guarded code seems to hit the race very often. Since everyone that changes the inode also logs it, it should be possible to fix this with some memory barriers. I'll leave that as an exercise to the reader and lock the buffer head instead. It it probably a good idea to have a different patch series for lockless bit flipping on the ext3 i_state field. ext3_do_update_inode &= clears EXT3_STATE_NEW without any locks held. Signed-off-by: Chris Mason Signed-off-by: Jan Kara --- fs/ext3/inode.c | 9 +++++++++ 1 file changed, 9 insertions(+) (limited to 'fs') diff --git a/fs/ext3/inode.c b/fs/ext3/inode.c index eb0b4e0d517b..cd098a7b77fc 100644 --- a/fs/ext3/inode.c +++ b/fs/ext3/inode.c @@ -2903,6 +2903,10 @@ static int ext3_do_update_inode(handle_t *handle, struct buffer_head *bh = iloc->bh; int err = 0, rc, block; +again: + /* we can't allow multiple procs in here at once, its a bit racey */ + lock_buffer(bh); + /* For fields not not tracking in the in-memory inode, * initialise them to zero for new inodes. */ if (ei->i_state & EXT3_STATE_NEW) @@ -2962,16 +2966,20 @@ static int ext3_do_update_inode(handle_t *handle, /* If this is the first large file * created, add a flag to the superblock. */ + unlock_buffer(bh); err = ext3_journal_get_write_access(handle, EXT3_SB(sb)->s_sbh); if (err) goto out_brelse; + ext3_update_dynamic_rev(sb); EXT3_SET_RO_COMPAT_FEATURE(sb, EXT3_FEATURE_RO_COMPAT_LARGE_FILE); handle->h_sync = 1; err = ext3_journal_dirty_metadata(handle, EXT3_SB(sb)->s_sbh); + /* get our lock and start over */ + goto again; } } } @@ -2994,6 +3002,7 @@ static int ext3_do_update_inode(handle_t *handle, raw_inode->i_extra_isize = cpu_to_le16(ei->i_extra_isize); BUFFER_TRACE(bh, "call ext3_journal_dirty_metadata"); + unlock_buffer(bh); rc = ext3_journal_dirty_metadata(handle, bh); if (!err) err = rc; -- cgit v1.2.3-59-g8ed1b From 56fcad29d4b3cbcbb2ed47a9d3ceca3f57175417 Mon Sep 17 00:00:00 2001 From: Jan Kara Date: Tue, 8 Sep 2009 14:59:42 +0200 Subject: ext3: Flush disk caches on fsync when needed In case we fsync() a file and inode is not dirty, we don't force a transaction to disk and hence don't flush disk caches. Thus file data could be just in disk caches and not on persistent storage. Fix the problem by flushing disk caches if we didn't force a transaction commit. Signed-off-by: Jan Kara --- fs/ext3/fsync.c | 12 +++++++++++- 1 file changed, 11 insertions(+), 1 deletion(-) (limited to 'fs') diff --git a/fs/ext3/fsync.c b/fs/ext3/fsync.c index d33634119e17..451d166bbe93 100644 --- a/fs/ext3/fsync.c +++ b/fs/ext3/fsync.c @@ -23,6 +23,7 @@ */ #include +#include #include #include #include @@ -73,7 +74,7 @@ int ext3_sync_file(struct file * file, struct dentry *dentry, int datasync) } if (datasync && !(inode->i_state & I_DIRTY_DATASYNC)) - goto out; + goto flush; /* * The VFS has written the file data. If the inode is unaltered @@ -85,7 +86,16 @@ int ext3_sync_file(struct file * file, struct dentry *dentry, int datasync) .nr_to_write = 0, /* sys_fsync did this */ }; ret = sync_inode(inode, &wbc); + goto out; } +flush: + /* + * In case we didn't commit a transaction, we have to flush + * disk caches manually so that data really is on persistent + * storage + */ + if (test_opt(inode->i_sb, BARRIER)) + blkdev_issue_flush(inode->i_sb->s_bdev, NULL); out: return ret; } -- cgit v1.2.3-59-g8ed1b