From 4524451ef7e88c64a868a8f5a0b49bda73beb2a3 Mon Sep 17 00:00:00 2001 From: Christoph Hellwig Date: Wed, 18 Aug 2010 05:29:16 -0400 Subject: jbd: replace barriers with explicit flush / FUA usage Switch to the WRITE_FLUSH_FUA flag for journal commits and remove the EOPNOTSUPP detection for barriers. Signed-off-by: Christoph Hellwig Acked-by: Jan Kara Signed-off-by: Tejun Heo Signed-off-by: Jens Axboe --- fs/jbd/commit.c | 30 +++--------------------------- 1 file changed, 3 insertions(+), 27 deletions(-) (limited to 'fs/jbd/commit.c') diff --git a/fs/jbd/commit.c b/fs/jbd/commit.c index 95d8c11c929e..484c5e5fa8af 100644 --- a/fs/jbd/commit.c +++ b/fs/jbd/commit.c @@ -137,34 +137,10 @@ static int journal_write_commit_record(journal_t *journal, JBUFFER_TRACE(descriptor, "write commit block"); set_buffer_dirty(bh); - if (journal->j_flags & JFS_BARRIER) { - ret = __sync_dirty_buffer(bh, WRITE_SYNC | WRITE_BARRIER); - - /* - * Is it possible for another commit to fail at roughly - * the same time as this one? If so, we don't want to - * trust the barrier flag in the super, but instead want - * to remember if we sent a barrier request - */ - if (ret == -EOPNOTSUPP) { - char b[BDEVNAME_SIZE]; - - printk(KERN_WARNING - "JBD: barrier-based sync failed on %s - " - "disabling barriers\n", - bdevname(journal->j_dev, b)); - spin_lock(&journal->j_state_lock); - journal->j_flags &= ~JFS_BARRIER; - spin_unlock(&journal->j_state_lock); - - /* And try again, without the barrier */ - set_buffer_uptodate(bh); - set_buffer_dirty(bh); - ret = sync_dirty_buffer(bh); - } - } else { + if (journal->j_flags & JFS_BARRIER) + ret = __sync_dirty_buffer(bh, WRITE_SYNC | WRITE_FLUSH_FUA); + else ret = sync_dirty_buffer(bh); - } put_bh(bh); /* One for getblk() */ journal_put_journal_head(descriptor); -- cgit v1.2.2 From 749ef9f8423054e326f3a246327ed2db4b6d395f Mon Sep 17 00:00:00 2001 From: Corrado Zoccolo Date: Mon, 20 Sep 2010 15:24:50 +0200 Subject: cfq: improve fsync performance for small files Fsync performance for small files achieved by cfq on high-end disks is lower than what deadline can achieve, due to idling introduced between the sync write happening in process context and the journal commit. Moreover, when competing with a sequential reader, a process writing small files and fsync-ing them is starved. This patch fixes the two problems by: - marking journal commits as WRITE_SYNC, so that they get the REQ_NOIDLE flag set, - force all queues that have REQ_NOIDLE requests to be put in the noidle tree. Having the queue associated to the fsync-ing process and the one associated to journal commits in the noidle tree allows: - switching between them without idling, - fairness vs. competing idling queues, since they will be serviced only after the noidle tree expires its slice. Acked-by: Vivek Goyal Reviewed-by: Jeff Moyer Tested-by: Jeff Moyer Signed-off-by: Corrado Zoccolo Signed-off-by: Jens Axboe --- fs/jbd/commit.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) (limited to 'fs/jbd/commit.c') diff --git a/fs/jbd/commit.c b/fs/jbd/commit.c index 95d8c11c929e..3f030e9efea6 100644 --- a/fs/jbd/commit.c +++ b/fs/jbd/commit.c @@ -318,7 +318,7 @@ void journal_commit_transaction(journal_t *journal) int first_tag = 0; int tag_flag; int i; - int write_op = WRITE; + int write_op = WRITE_SYNC; /* * First job: lock down the current transaction and wait for -- cgit v1.2.2 From a910eefa511f9d1118effc13fba6773163502c4f Mon Sep 17 00:00:00 2001 From: Namhyung Kim Date: Fri, 8 Oct 2010 20:05:06 +0900 Subject: jbd: Convert bitops to buffer fns Convert set/clear_bit(BH_JWrite, ...) to set/clear_buffer_jwrite() for consistency. Signed-off-by: Namhyung Kim Signed-off-by: Jan Kara --- fs/jbd/commit.c | 6 +++--- 1 file changed, 3 insertions(+), 3 deletions(-) (limited to 'fs/jbd/commit.c') diff --git a/fs/jbd/commit.c b/fs/jbd/commit.c index 95d8c11c929e..c8428323167f 100644 --- a/fs/jbd/commit.c +++ b/fs/jbd/commit.c @@ -617,7 +617,7 @@ void journal_commit_transaction(journal_t *journal) (this will requeue both the metadata buffer and the temporary IO buffer). new_bh goes on BJ_IO*/ - set_bit(BH_JWrite, &jh2bh(jh)->b_state); + set_buffer_jwrite(jh2bh(jh)); /* * akpm: journal_write_metadata_buffer() sets * new_bh->b_transaction to commit_transaction. @@ -627,7 +627,7 @@ void journal_commit_transaction(journal_t *journal) JBUFFER_TRACE(jh, "ph3: write metadata"); flags = journal_write_metadata_buffer(commit_transaction, jh, &new_jh, blocknr); - set_bit(BH_JWrite, &jh2bh(new_jh)->b_state); + set_buffer_jwrite(jh2bh(new_jh)); wbuf[bufs++] = jh2bh(new_jh); /* Record the new block's tag in the current descriptor @@ -737,7 +737,7 @@ wait_for_iobuf: shadowed buffer */ jh = commit_transaction->t_shadow_list->b_tprev; bh = jh2bh(jh); - clear_bit(BH_JWrite, &bh->b_state); + clear_buffer_jwrite(bh); J_ASSERT_BH(bh, buffer_jbddirty(bh)); /* The metadata is now released for reuse, but we need -- cgit v1.2.2 From e4d5e3a497e159be7c2dbe4c61cfb185d60cfde2 Mon Sep 17 00:00:00 2001 From: Namhyung Kim Date: Sat, 16 Oct 2010 17:11:02 +0900 Subject: jbd: Convert atomic_inc() to get_bh() Convert atomic_inc(&bh->b_count) to get_bh(bh) for consistency. Signed-off-by: Namhyung Kim Signed-off-by: Jan Kara --- fs/jbd/commit.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) (limited to 'fs/jbd/commit.c') diff --git a/fs/jbd/commit.c b/fs/jbd/commit.c index c8428323167f..a89c4630b1ed 100644 --- a/fs/jbd/commit.c +++ b/fs/jbd/commit.c @@ -611,7 +611,7 @@ void journal_commit_transaction(journal_t *journal) /* Bump b_count to prevent truncate from stumbling over the shadowed buffer! @@@ This can go if we ever get rid of the BJ_IO/BJ_Shadow pairing of buffers. */ - atomic_inc(&jh2bh(jh)->b_count); + get_bh(jh2bh(jh)); /* Make a temporary IO buffer with which to write it out (this will requeue both the metadata buffer and the -- cgit v1.2.2 From 721a9602e6607417c6bc15b18e97a2f35266c690 Mon Sep 17 00:00:00 2001 From: Jens Axboe Date: Wed, 9 Mar 2011 11:56:30 +0100 Subject: block: kill off REQ_UNPLUG With the plugging now being explicitly controlled by the submitter, callers need not pass down unplugging hints to the block layer. If they want to unplug, it's because they manually plugged on their own - in which case, they should just unplug at will. Signed-off-by: Jens Axboe --- fs/jbd/commit.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) (limited to 'fs/jbd/commit.c') diff --git a/fs/jbd/commit.c b/fs/jbd/commit.c index 34a4861c14b8..66be299acb1b 100644 --- a/fs/jbd/commit.c +++ b/fs/jbd/commit.c @@ -333,7 +333,7 @@ void journal_commit_transaction(journal_t *journal) * instead we rely on sync_buffer() doing the unplug for us. */ if (commit_transaction->t_synchronous_commit) - write_op = WRITE_SYNC_PLUG; + write_op = WRITE_SYNC; spin_lock(&commit_transaction->t_handle_lock); while (commit_transaction->t_updates) { DEFINE_WAIT(wait); -- cgit v1.2.2 From 65ab80279d7c323ce151e858d951e2684df72a97 Mon Sep 17 00:00:00 2001 From: Jens Axboe Date: Thu, 17 Mar 2011 10:56:45 +0100 Subject: jbd: finish conversion from WRITE_SYNC_PLUG to WRITE_SYNC and explicit plugging 'write_op' was still used, even though it was always WRITE_SYNC now. Add plugging around the cases where it submits IO, and flush them before we end up waiting for that IO. Signed-off-by: Jens Axboe --- fs/jbd/commit.c | 22 +++++++++++----------- 1 file changed, 11 insertions(+), 11 deletions(-) (limited to 'fs/jbd/commit.c') diff --git a/fs/jbd/commit.c b/fs/jbd/commit.c index 66be299acb1b..da871ee084d3 100644 --- a/fs/jbd/commit.c +++ b/fs/jbd/commit.c @@ -20,6 +20,7 @@ #include #include #include +#include /* * Default IO end handler for temporary BJ_IO buffer_heads. @@ -294,7 +295,7 @@ void journal_commit_transaction(journal_t *journal) int first_tag = 0; int tag_flag; int i; - int write_op = WRITE_SYNC; + struct blk_plug plug; /* * First job: lock down the current transaction and wait for @@ -327,13 +328,6 @@ void journal_commit_transaction(journal_t *journal) spin_lock(&journal->j_state_lock); commit_transaction->t_state = T_LOCKED; - /* - * Use plugged writes here, since we want to submit several before - * we unplug the device. We don't do explicit unplugging in here, - * instead we rely on sync_buffer() doing the unplug for us. - */ - if (commit_transaction->t_synchronous_commit) - write_op = WRITE_SYNC; spin_lock(&commit_transaction->t_handle_lock); while (commit_transaction->t_updates) { DEFINE_WAIT(wait); @@ -418,8 +412,10 @@ void journal_commit_transaction(journal_t *journal) * Now start flushing things to disk, in the order they appear * on the transaction lists. Data blocks go first. */ + blk_start_plug(&plug); err = journal_submit_data_buffers(journal, commit_transaction, - write_op); + WRITE_SYNC); + blk_finish_plug(&plug); /* * Wait for all previously submitted IO to complete. @@ -480,7 +476,9 @@ void journal_commit_transaction(journal_t *journal) err = 0; } - journal_write_revoke_records(journal, commit_transaction, write_op); + blk_start_plug(&plug); + + journal_write_revoke_records(journal, commit_transaction, WRITE_SYNC); /* * If we found any dirty or locked buffers, then we should have @@ -650,7 +648,7 @@ start_journal_io: clear_buffer_dirty(bh); set_buffer_uptodate(bh); bh->b_end_io = journal_end_buffer_io_sync; - submit_bh(write_op, bh); + submit_bh(WRITE_SYNC, bh); } cond_resched(); @@ -661,6 +659,8 @@ start_journal_io: } } + blk_finish_plug(&plug); + /* Lo and behold: we have just managed to send a transaction to the log. Before we can commit it, wait for the IO so far to complete. Control buffers being written are on the -- cgit v1.2.2 From 25985edcedea6396277003854657b5f3cb31a628 Mon Sep 17 00:00:00 2001 From: Lucas De Marchi Date: Wed, 30 Mar 2011 22:57:33 -0300 Subject: Fix common misspellings Fixes generated by 'codespell' and manually reviewed. Signed-off-by: Lucas De Marchi --- fs/jbd/commit.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) (limited to 'fs/jbd/commit.c') diff --git a/fs/jbd/commit.c b/fs/jbd/commit.c index da871ee084d3..69b180459463 100644 --- a/fs/jbd/commit.c +++ b/fs/jbd/commit.c @@ -362,7 +362,7 @@ void journal_commit_transaction(journal_t *journal) * we do not require it to remember exactly which old buffers it * has reserved. This is consistent with the existing behaviour * that multiple journal_get_write_access() calls to the same - * buffer are perfectly permissable. + * buffer are perfectly permissible. */ while (commit_transaction->t_reserved_list) { jh = commit_transaction->t_reserved_list; -- cgit v1.2.2 From 2842bb20eed2e25cde5114298edc62c8883a1d9a Mon Sep 17 00:00:00 2001 From: Jan Kara Date: Thu, 5 May 2011 13:59:35 +0200 Subject: jbd: Fix forever sleeping process in do_get_write_access() In do_get_write_access() we wait on BH_Unshadow bit for buffer to get from shadow state. The waking code in journal_commit_transaction() has a bug because it does not issue a memory barrier after the buffer is moved from the shadow state and before wake_up_bit() is called. Thus a waitqueue check can happen before the buffer is actually moved from the shadow state and waiting process may never be woken. Fix the problem by issuing proper barrier. CC: stable@kernel.org Reported-by: Tao Ma Signed-off-by: Jan Kara --- fs/jbd/commit.c | 9 +++++++-- 1 file changed, 7 insertions(+), 2 deletions(-) (limited to 'fs/jbd/commit.c') diff --git a/fs/jbd/commit.c b/fs/jbd/commit.c index 69b180459463..f486ff697781 100644 --- a/fs/jbd/commit.c +++ b/fs/jbd/commit.c @@ -722,8 +722,13 @@ wait_for_iobuf: required. */ JBUFFER_TRACE(jh, "file as BJ_Forget"); journal_file_buffer(jh, commit_transaction, BJ_Forget); - /* Wake up any transactions which were waiting for this - IO to complete */ + /* + * Wake up any transactions which were waiting for this + * IO to complete. The barrier must be here so that changes + * by journal_file_buffer() take effect before wake_up_bit() + * does the waitqueue check. + */ + smp_mb(); wake_up_bit(&bh->b_state, BH_Unshadow); JBUFFER_TRACE(jh, "brelse shadowed buffer"); __brelse(bh); -- cgit v1.2.2 From 9199e66528f61a06abe09f0589bbe1eecaa301a7 Mon Sep 17 00:00:00 2001 From: Tao Ma Date: Thu, 5 May 2011 23:54:19 +0800 Subject: jbd/jbd2: remove obsolete summarise_journal_usage. summarise_journal_usage seems to be obsolete for a long time, so remove it. Cc: Jan Kara Signed-off-by: Tao Ma Signed-off-by: Jan Kara --- fs/jbd/commit.c | 6 ------ 1 file changed, 6 deletions(-) (limited to 'fs/jbd/commit.c') diff --git a/fs/jbd/commit.c b/fs/jbd/commit.c index f486ff697781..72ffa974b0b8 100644 --- a/fs/jbd/commit.c +++ b/fs/jbd/commit.c @@ -302,12 +302,6 @@ void journal_commit_transaction(journal_t *journal) * all outstanding updates to complete. */ -#ifdef COMMIT_STATS - spin_lock(&journal->j_list_lock); - summarise_journal_usage(journal); - spin_unlock(&journal->j_list_lock); -#endif - /* Do we need to erase the effects of a prior journal_flush? */ if (journal->j_flags & JFS_FLUSHED) { jbd_debug(3, "super block updated\n"); -- cgit v1.2.2