ext3 users of data=journal mode with blocksize < pagesize were occasionally
hitting assertion failure in journal_commit_transaction() checking whether the
transaction has at least as many credits reserved as buffers attached. The
core of the problem is that when a file gets truncated, buffers that still need
checkpointing or that are attached to the committing transaction are left with
buffer_mapped set. When this happens to buffers beyond i_size attached to a
page stradding i_size, subsequent write extending the file will see these
buffers and as they are mapped (but underlying blocks were freed) things go
awry from here.
The assertion failure just coincidentally (and in this case luckily as we would
start corrupting filesystem) triggers due to journal_head not being properly
cleaned up as well.
Under some rare circumstances this bug could even hit data=ordered mode users.
There the assertion won't trigger and we would end up corrupting the
filesystem.
We fix the problem by unmapping buffers if possible (in lots of cases we just
need a buffer attached to a transaction as a place holder but it must not be
written out anyway). And in one case, we just have to bite the bullet and wait
for transaction commit to finish.
Reviewed-by: Josef Bacik <jbacik@fusionio.com>
Signed-off-by: Jan Kara <jack@suse.cz>
If journal superblock is written only in disk's caches and other transaction
starts reusing space of the transaction cleaned from the log, it can happen
blocks of a new transaction reach the disk before journal superblock. When
power failure happens in such case, subsequent journal replay would still try
to replay the old transaction but some of it's blocks may be already
overwritten by the new transaction. For this reason we must use WRITE_FUA when
updating log tail and we must first write new log tail to disk and update
in-memory information only after that.
Signed-off-by: Jan Kara <jack@suse.cz>
There are some log tail updates that are not protected by j_checkpoint_mutex.
Some of these are harmless because they happen during startup or shutdown but
updates in journal_commit_transaction() and journal_flush() can really race
with other log tail updates (e.g. someone doing journal_flush() with someone
running cleanup_journal_tail()). So protect all log tail updates with
j_checkpoint_mutex.
Signed-off-by: Jan Kara <jack@suse.cz>
There are three case of updating journal superblock. In the first case, we want
to mark journal as empty (setting s_sequence to 0), in the second case we want
to update log tail, in the third case we want to update s_errno. Split these
cases into separate functions. It makes the code slightly more straightforward
and later patches will make the distinction even more important.
Signed-off-by: Jan Kara <jack@suse.cz>
Currently we write out all journal buffers in WRITE_SYNC mode. This improves
performance for fsync heavy workloads but hinders performance when writes
are mostly asynchronous, most noticably it slows down readers and users
complain about slow desktop response etc.
So submit writes as asynchronous in the normal case and only submit writes as
WRITE_SYNC if we detect someone is waiting for current transaction commit.
I've gathered some numbers to back this change. The first is the read latency
test. It measures time to read 1 MB after several seconds of sleeping in
presence of streaming writes.
Top 10 times (out of 90) in us:
Before After
2131586 697473
1709932 557487
1564598 535642
1480462 347573
1478579 323153
1408496 222181
1388960 181273
1329565 181070
1252486 172832
1223265 172278
Average:
619377 82180
So the improvement in both maximum and average latency is massive.
I've measured fsync throughput by:
fs_mark -n 100 -t 1 -s 16384 -d /mnt/fsync/ -S 1 -L 4
in presence of streaming reader. The numbers (fsyncs/s) are:
Before After
9.9 6.3
6.8 6.0
6.3 6.2
5.8 6.1
So fsync performance seems unharmed by this change.
Signed-off-by: Jan Kara <jack@suse.cz>
Currently, we clear revoked flag only when a block is reused. However,
this can tigger a false journal error. Consider a situation when a block
is used as a meta block and is deleted(revoked) in ordered mode, then the
block is allocated as a data block to a file. At this moment, user changes
the file's journal mode from ordered to journaled and truncates the file.
The block will be considered re-revoked by journal because it has revoked
flag still pending from the last transaction and an assertion triggers.
We fix the problem by keeping the revoked status more uptodate - we clear
revoked flag when switching revoke tables to reflect there is no revoked
buffers in current transaction any more.
Signed-off-by: Yongqiang Yang <xiaoqiangnk@gmail.com>
Signed-off-by: Jan Kara <jack@suse.cz>
journal_remove_journal_head() can oops when trying to access journal_head
returned by bh2jh(). This is caused for example by the following race:
TASK1 TASK2
journal_commit_transaction()
...
processing t_forget list
__journal_refile_buffer(jh);
if (!jh->b_transaction) {
jbd_unlock_bh_state(bh);
journal_try_to_free_buffers()
journal_grab_journal_head(bh)
jbd_lock_bh_state(bh)
__journal_try_to_free_buffer()
journal_put_journal_head(jh)
journal_remove_journal_head(bh);
journal_put_journal_head() in TASK2 sees that b_jcount == 0 and buffer is not
part of any transaction and thus frees journal_head before TASK1 gets to doing
so. Note that even buffer_head can be released by try_to_free_buffers() after
journal_put_journal_head() which adds even larger opportunity for oops (but I
didn't see this happen in reality).
Fix the problem by making transactions hold their own journal_head reference
(in b_jcount). That way we don't have to remove journal_head explicitely via
journal_remove_journal_head() and instead just remove journal_head when
b_jcount drops to zero. The result of this is that [__]journal_refile_buffer(),
[__]journal_unfile_buffer(), and __journal_remove_checkpoint() can free
journal_head which needs modification of a few callers. Also we have to be
careful because once journal_head is removed, buffer_head might be freed as
well. So we have to get our own buffer_head reference where it matters.
Signed-off-by: Jan Kara <jack@suse.cz>
This commit adds fixed tracepoint for jbd. It has been based on fixed
tracepoints for jbd2, however there are missing those for collecting
statistics, since I think that it will require more intrusive patch so I
should have its own commit, if someone decide that it is needed. Also
there are new tracepoints in __journal_drop_transaction() and
journal_update_superblock().
The list of jbd tracepoints:
jbd_checkpoint
jbd_start_commit
jbd_commit_locking
jbd_commit_flushing
jbd_commit_logging
jbd_drop_transaction
jbd_end_commit
jbd_do_submit_data
jbd_cleanup_journal_tail
jbd_update_superblock_end
Signed-off-by: Lukas Czerner <lczerner@redhat.com>
Cc: Jan Kara <jack@suse.cz>
Signed-off-by: Jan Kara <jack@suse.cz>
summarise_journal_usage seems to be obsolete for a long time,
so remove it.
Cc: Jan Kara <jack@suse.cz>
Signed-off-by: Tao Ma <boyu.mt@taobao.com>
Signed-off-by: Jan Kara <jack@suse.cz>
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 <boyu.mt@taobao.com>
Signed-off-by: Jan Kara <jack@suse.cz>
'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 <jaxboe@fusionio.com>
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 <jaxboe@fusionio.com>
* 'for_linus' of git://git.kernel.org/pub/scm/linux/kernel/git/jack/linux-fs-2.6: (24 commits)
quota: Fix possible oops in __dquot_initialize()
ext3: Update kernel-doc comments
jbd/2: fixed typos
ext2: fixed typo.
ext3: Fix debug messages in ext3_group_extend()
jbd: Convert atomic_inc() to get_bh()
ext3: Remove misplaced BUFFER_TRACE() in ext3_truncate()
jbd: Fix debug message in do_get_write_access()
jbd: Check return value of __getblk()
ext3: Use DIV_ROUND_UP() on group desc block counting
ext3: Return proper error code on ext3_fill_super()
ext3: Remove unnecessary casts on bh->b_data
ext3: Cleanup ext3_setup_super()
quota: Fix issuing of warnings from dquot_transfer
quota: fix dquot_disable vs dquot_transfer race v2
jbd: Convert bitops to buffer fns
ext3/jbd: Avoid WARN() messages when failing to write the superblock
jbd: Use offset_in_page() instead of manual calculation
jbd: Remove unnecessary goto statement
jbd: Use printk_ratelimited() in journal_alloc_journal_head()
...
Convert atomic_inc(&bh->b_count) to get_bh(bh) for consistency.
Signed-off-by: Namhyung Kim <namhyung@gmail.com>
Signed-off-by: Jan Kara <jack@suse.cz>
Convert set/clear_bit(BH_JWrite, ...) to set/clear_buffer_jwrite()
for consistency.
Signed-off-by: Namhyung Kim <namhyung@gmail.com>
Signed-off-by: Jan Kara <jack@suse.cz>
* 'for-2.6.37/barrier' of git://git.kernel.dk/linux-2.6-block: (46 commits)
xen-blkfront: disable barrier/flush write support
Added blk-lib.c and blk-barrier.c was renamed to blk-flush.c
block: remove BLKDEV_IFL_WAIT
aic7xxx_old: removed unused 'req' variable
block: remove the BH_Eopnotsupp flag
block: remove the BLKDEV_IFL_BARRIER flag
block: remove the WRITE_BARRIER flag
swap: do not send discards as barriers
fat: do not send discards as barriers
ext4: do not send discards as barriers
jbd2: replace barriers with explicit flush / FUA usage
jbd2: Modify ASYNC_COMMIT code to not rely on queue draining on barrier
jbd: replace barriers with explicit flush / FUA usage
nilfs2: replace barriers with explicit flush / FUA usage
reiserfs: replace barriers with explicit flush / FUA usage
gfs2: replace barriers with explicit flush / FUA usage
btrfs: replace barriers with explicit flush / FUA usage
xfs: replace barriers with explicit flush / FUA usage
block: pass gfp_mask and flags to sb_issue_discard
dm: convey that all flushes are processed as empty
...
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 <vgoyal@redhat.com>
Reviewed-by: Jeff Moyer <jmoyer@redhat.com>
Tested-by: Jeff Moyer <jmoyer@redhat.com>
Signed-off-by: Corrado Zoccolo <czoccolo@gmail.com>
Signed-off-by: Jens Axboe <jaxboe@fusionio.com>
Switch to the WRITE_FLUSH_FUA flag for journal commits and remove the
EOPNOTSUPP detection for barriers.
Signed-off-by: Christoph Hellwig <hch@lst.de>
Acked-by: Jan Kara <jack@suse.cz>
Signed-off-by: Tejun Heo <tj@kernel.org>
Signed-off-by: Jens Axboe <jaxboe@fusionio.com>
Instead of abusing a buffer_head flag just add a variant of
sync_dirty_buffer which allows passing the exact type of write
flag required.
Signed-off-by: Christoph Hellwig <hch@lst.de>
Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>
Provide a function which returns whether a transaction with given tid
will send a barrier to the filesystem device. The function will be used
by ext3 to detect whether fsync needs to send a separate barrier or not.
Signed-off-by: Jan Kara <jack@suse.cz>
percpu.h is included by sched.h and module.h and thus ends up being
included when building most .c files. percpu.h includes slab.h which
in turn includes gfp.h making everything defined by the two files
universally available and complicating inclusion dependencies.
percpu.h -> slab.h dependency is about to be removed. Prepare for
this change by updating users of gfp and slab facilities include those
headers directly instead of assuming availability. As this conversion
needs to touch large number of source files, the following script is
used as the basis of conversion.
http://userweb.kernel.org/~tj/misc/slabh-sweep.py
The script does the followings.
* Scan files for gfp and slab usages and update includes such that
only the necessary includes are there. ie. if only gfp is used,
gfp.h, if slab is used, slab.h.
* When the script inserts a new include, it looks at the include
blocks and try to put the new include such that its order conforms
to its surrounding. It's put in the include block which contains
core kernel includes, in the same order that the rest are ordered -
alphabetical, Christmas tree, rev-Xmas-tree or at the end if there
doesn't seem to be any matching order.
* If the script can't find a place to put a new include (mostly
because the file doesn't have fitting include block), it prints out
an error message indicating which .h file needs to be added to the
file.
The conversion was done in the following steps.
1. The initial automatic conversion of all .c files updated slightly
over 4000 files, deleting around 700 includes and adding ~480 gfp.h
and ~3000 slab.h inclusions. The script emitted errors for ~400
files.
2. Each error was manually checked. Some didn't need the inclusion,
some needed manual addition while adding it to implementation .h or
embedding .c file was more appropriate for others. This step added
inclusions to around 150 files.
3. The script was run again and the output was compared to the edits
from #2 to make sure no file was left behind.
4. Several build tests were done and a couple of problems were fixed.
e.g. lib/decompress_*.c used malloc/free() wrappers around slab
APIs requiring slab.h to be added manually.
5. The script was run on all .h files but without automatically
editing them as sprinkling gfp.h and slab.h inclusions around .h
files could easily lead to inclusion dependency hell. Most gfp.h
inclusion directives were ignored as stuff from gfp.h was usually
wildly available and often used in preprocessor macros. Each
slab.h inclusion directive was examined and added manually as
necessary.
6. percpu.h was updated not to include slab.h.
7. Build test were done on the following configurations and failures
were fixed. CONFIG_GCOV_KERNEL was turned off for all tests (as my
distributed build env didn't work with gcov compiles) and a few
more options had to be turned off depending on archs to make things
build (like ipr on powerpc/64 which failed due to missing writeq).
* x86 and x86_64 UP and SMP allmodconfig and a custom test config.
* powerpc and powerpc64 SMP allmodconfig
* sparc and sparc64 SMP allmodconfig
* ia64 SMP allmodconfig
* s390 SMP allmodconfig
* alpha SMP allmodconfig
* um on x86_64 SMP allmodconfig
8. percpu.h modifications were reverted so that it could be applied as
a separate patch and serve as bisection point.
Given the fact that I had only a couple of failures from tests on step
6, I'm fairly confident about the coverage of this conversion patch.
If there is a breakage, it's likely to be something in one of the arch
headers which should be easily discoverable easily on most builds of
the specific arch.
Signed-off-by: Tejun Heo <tj@kernel.org>
Guess-its-ok-by: Christoph Lameter <cl@linux-foundation.org>
Cc: Ingo Molnar <mingo@redhat.com>
Cc: Lee Schermerhorn <Lee.Schermerhorn@hp.com>
Delay discarding buffers in journal_unmap_buffer until
we know that "add to orphan" operation has definitely been
committed, otherwise the log space of committing transation
may be freed and reused before truncate get committed, updates
may get lost if crash happens.
This patch is a backport of JBD2 fix by dingdinghua <dingdinghua@nrchpc.ac.cn>.
Signed-off-by: Jan Kara <jack@suse.cz>
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 <jack@suse.cz>
In commit code, we scan buffers attached to a transaction. During this
scan, we sometimes have to drop j_list_lock and then we recheck whether
the journal buffer head didn't get freed by journal_try_to_free_buffers().
But checking for buffer_jbd(bh) isn't enough because a new journal head
could get attached to our buffer head. So add a check whether the journal
head remained the same and whether it's still at the same transaction and
list.
This is a nasty bug and can cause problems like memory corruption (use after
free) or trigger various assertions in JBD code (observed).
Signed-off-by: Jan Kara <jack@suse.cz>
Cc: <stable@kernel.org>
Cc: <linux-ext4@vger.kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
The revoke records must be written using the same way as the rest of
the blocks during the commit process; that is, either marked as
synchronous writes or as asynchornous writes.
Signed-off-by: "Theodore Ts'o" <tytso@mit.edu>
When you are going to be submitting several sync writes, we want to
give the IO scheduler a chance to merge some of them. Instead of
using the implicitly unplugging WRITE_SYNC variant, use WRITE_SYNC_PLUG
and rely on sync_buffer() doing the unplug when someone does a
wait_on_buffer()/lock_buffer().
Signed-off-by: Jens Axboe <jens.axboe@oracle.com>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
If a commit is triggered by fsync(), set a flag indicating the journal
blocks associated with the transaction should be flushed out using
WRITE_SYNC.
Signed-off-by: "Theodore Ts'o" <tytso@mit.edu>
Acked-by: Jan Kara <jack@suse.cz>
There is a flaw with the way jbd handles fsync batching. If we fsync() a
file and we were not the last person to run fsync() on this fs then we
automatically sleep for 1 jiffie in order to wait for new writers to join
into the transaction before forcing the commit. The problem with this is
that with really fast storage (ie a Clariion) the time it takes to commit
a transaction to disk is way faster than 1 jiffie in most cases, so
sleeping means waiting longer with nothing to do than if we just committed
the transaction and kept going. Ric Wheeler noticed this when using
fs_mark with more than 1 thread, the throughput would plummet as he added
more threads.
This patch attempts to fix this problem by recording the average time in
nanoseconds that it takes to commit a transaction to disk, and what time
we started the transaction. If we run an fsync() and we have been running
for less time than it takes to commit the transaction to disk, we sleep
for the delta amount of time and then commit to disk. We acheive
sub-jiffie sleeping using schedule_hrtimeout. This means that the wait
time is auto-tuned to the speed of the underlying disk, instead of having
this static timeout. I weighted the average according to somebody's
comments (Andreas Dilger I think) in order to help normalize random
outliers where we take way longer or way less time to commit than the
average. I also have a min() check in there to make sure we don't sleep
longer than a jiffie in case our storage is super slow, this was requested
by Andrew.
I unfortunately do not have access to a Clariion, so I had to use a
ramdisk to represent a super fast array. I tested with a SATA drive with
barrier=1 to make sure there was no regression with local disks, I tested
with a 4 way multipathed Apple Xserve RAID array and of course the
ramdisk. I ran the following command
fs_mark -d /mnt/ext3-test -s 4096 -n 2000 -D 64 -t $i
where $i was 2, 4, 8, 16 and 32. I mkfs'ed the fs each time. Here are my
results
type threads with patch without patch
sata 2 24.6 26.3
sata 4 49.2 48.1
sata 8 70.1 67.0
sata 16 104.0 94.1
sata 32 153.6 142.7
xserve 2 246.4 222.0
xserve 4 480.0 440.8
xserve 8 829.5 730.8
xserve 16 1172.7 1026.9
xserve 32 1816.3 1650.5
ramdisk 2 2538.3 1745.6
ramdisk 4 2942.3 661.9
ramdisk 8 2882.5 999.8
ramdisk 16 2738.7 1801.9
ramdisk 32 2541.9 2394.0
Signed-off-by: Josef Bacik <jbacik@redhat.com>
Cc: Andreas Dilger <adilger@sun.com>
Cc: Arjan van de Ven <arjan@infradead.org>
Cc: Ric Wheeler <rwheeler@redhat.com>
Cc: <linux-ext4@vger.kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
If the journal doesn't abort when it gets an IO error in file data blocks,
the file data corruption will spread silently. Because most of
applications and commands do buffered writes without fsync(), they don't
notice the IO error. It's scary for mission critical systems. On the
other hand, if the journal aborts whenever it gets an IO error in file
data blocks, the system will easily become inoperable. So this patch
introduces a filesystem option to determine whether it aborts the journal
or just call printk() when it gets an IO error in file data.
If you mount a ext3 fs with data_err=abort option, it aborts on file data
write error. If you mount it with data_err=ignore, it doesn't abort, just
call printk(). data_err=ignore is the default.
Signed-off-by: Hidehiro Kawai <hidehiro.kawai.ez@hitachi.com>
Cc: Jan Kara <jack@ucw.cz>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Currently, original metadata buffers are dirtied when they are unfiled
whether the journal has aborted or not. Eventually these buffers will be
written-back to the filesystem by pdflush. This means some metadata
buffers are written to the filesystem without journaling if the journal
aborts. So if both journal abort and system crash happen at the same
time, the filesystem would become inconsistent state. Additionally,
replaying journaled metadata can overwrite the latest metadata on the
filesystem partly. Because, if the journal aborts, journaled metadata are
preserved and replayed during the next mount not to lose uncheckpointed
metadata. This would also break the consistency of the filesystem.
This patch prevents original metadata buffers from being dirtied on abort
by clearing BH_JBDDirty flag from those buffers. Thus, no metadata
buffers are written to the filesystem without journaling.
Signed-off-by: Hidehiro Kawai <hidehiro.kawai.ez@hitachi.com>
Acked-by: Jan Kara <jack@suse.cz>
Cc: <linux-ext4@vger.kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
If we failed to write metadata buffers to the journal space and succeeded
to write the commit record, stale data can be written back to the
filesystem as metadata in the recovery phase.
To avoid this, when we failed to write out metadata buffers, abort the
journal before writing the commit record.
Signed-off-by: Hidehiro Kawai <hidehiro.kawai.ez@hitachi.com>
Acked-by: Jan Kara <jack@suse.cz>
Cc: <linux-ext4@vger.kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Like the page lock change, this also requires name change, so convert the
raw test_and_set bitop to a trylock.
Signed-off-by: Nick Piggin <npiggin@suse.de>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Converting page lock to new locking bitops requires a change of page flag
operation naming, so we might as well convert it to something nicer
(!TestSetPageLocked_Lock => trylock_page, SetPageLocked => set_page_locked).
This also facilitates lockdeping of page lock.
Signed-off-by: Nick Piggin <npiggin@suse.de>
Acked-by: KOSAKI Motohiro <kosaki.motohiro@jp.fujitsu.com>
Acked-by: Peter Zijlstra <peterz@infradead.org>
Acked-by: Andrew Morton <akpm@linux-foundation.org>
Acked-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
In ordered mode, the current jbd aborts the journal if a file data buffer
has an error. But this behavior is unintended, and we found that it has
been adopted accidentally.
This patch undoes it and just calls printk() instead of aborting the
journal. Additionally, set AS_EIO into the address_space object of the
failed buffer which is submitted by journal_do_submit_data() so that
fsync() can get -EIO.
Missing error checkings are also added to inform errors on file data
buffers to the user. The following buffers are targeted.
(a) the buffer which has already been written out by pdflush
(b) the buffer which has been unlocked before scanned in the
t_locked_list loop
[akpm@linux-foundation.org: improve grammar in a printk]
Signed-off-by: Hidehiro Kawai <hidehiro.kawai.ez@hitachi.com>
Acked-by: Jan Kara <jack@suse.cz>
Cc: <linux-ext4@vger.kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
After ext3-ordered files are truncated, there is a possibility that the
pages which cannot be estimated still remain. Remaining pages can be
released when the system has really few memory. So, it is not memory
leakage. But the resource management software etc. may not work
correctly.
It is possible that journal_unmap_buffer() cannot release the buffers, and
the pages to which they belong because they are attached to a commiting
transaction and journal_unmap_buffer() cannot release them. To release
such the buffers and the pages later, journal_unmap_buffer() leaves it to
journal_commit_transaction(). (journal_unmap_buffer() puts the mark
'BH_Freed' to the buffers so that journal_commit_transaction() can
identify whether they can be released or not.)
In the journalled mode and the writeback mode, jbd does with only metadata
buffers. But in the ordered mode, jbd does with metadata buffers and also
data buffers.
Actually, journal_commit_transaction() releases only the metadata buffers
of which release is demanded by journal_unmap_buffer(), and also releases
the pages to which they belong if possible.
As a result, the data buffers of which release is demanded by
journal_unmap_buffer() remain after a transaction commits. And also the
pages to which they belong remain.
Such the remained pages don't have mapping any longer. Due to this fact,
there is a possibility that the pages which cannot be estimated remain.
The metadata buffers marked 'BH_Freed' and the pages to which
they belong can be released at 'JBD: commit phase 7'.
Therefore, by applying the same code into 'JBD: commit phase 2' (where the
data buffers are done with), journal_commit_transaction() can also release
the data buffers marked 'BH_Freed' and the pages to which they belong.
As a result, all the buffers marked 'BH_Freed' can be released, and also
all the pages to which these buffers belong can be released at
journal_commit_transaction(). So, the page which cannot be estimated is
lost.
<<Excerpt of code at 'JBD: commit phase 7'>>
> spin_lock(&journal->j_list_lock);
> while (commit_transaction->t_forget) {
> transaction_t *cp_transaction;
> struct buffer_head *bh;
>
> jh = commit_transaction->t_forget;
>...
> if (buffer_freed(bh)) {
> ^^^^^^^^^^^^^^^^^^^^^^^^
> clear_buffer_freed(bh);
> ^^^^^^^^^^^^^^^^^^^^^^^^
> clear_buffer_jbddirty(bh);
> }
>
> if (buffer_jbddirty(bh)) {
> JBUFFER_TRACE(jh, "add to new checkpointing trans");
> __journal_insert_checkpoint(jh, commit_transaction);
> JBUFFER_TRACE(jh, "refile for checkpoint writeback");
> __journal_refile_buffer(jh);
> jbd_unlock_bh_state(bh);
> } else {
> J_ASSERT_BH(bh, !buffer_dirty(bh));
> ...
> JBUFFER_TRACE(jh, "refile or unfile freed buffer");
> __journal_refile_buffer(jh);
> if (!jh->b_transaction) {
> jbd_unlock_bh_state(bh);
> /* needs a brelse */
> journal_remove_journal_head(bh);
> release_buffer_page(bh);
> ^^^^^^^^^^^^^^^^^^^^^^^^
> } else
> }
****************************************************************
* Apply the code of "^^^^^^" lines into 'JBD: commit phase 2' *
****************************************************************
At journal_commit_transaction() code, there is one extra message in the
series of jbd debug messages. ("JBD: commit phase 2") This patch fixes
it, too.
Signed-off-by: Toshiyuki Okajima <toshi.okajima@jp.fujitsu.com>
Acked-by: Jan Kara <jack@suse.cz>
Cc: <linux-ext4@vger.kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Updating the current transaction's t_state is protected by j_state_lock. We
need to do the same when updating the t_state to T_COMMIT.
Signed-off-by: Mingming Cao <cmm@us.ibm.com>
Acked-by: Jan Kara <jack@ucw.cz>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
There are several cases where the running transaction can get buffers added to
its BJ_Metadata list which it never dirtied, which makes its t_nr_buffers
counter end up larger than its t_outstanding_credits counter.
This will cause issues when starting new transactions as while we are logging
buffers we decrement t_outstanding_buffers, so when t_outstanding_buffers goes
negative, we will report that we need less space in the journal than we
actually need, so transactions will be started even though there may not be
enough room for them. In the worst case scenario (which admittedly is almost
impossible to reproduce) this will result in the journal running out of space.
The fix is to only
refile buffers from the committing transaction to the running transactions
BJ_Modified list when b_modified is set on that journal, which is the only way
to be sure if the running transaction has modified that buffer.
This patch also fixes an accounting error in journal_forget, it is possible
that we can call journal_forget on a buffer without having modified it, only
gotten write access to it, so instead of freeing a credit, we only do so if
the buffer was modified. The assert will help catch if this problem occurs.
Without these two patches I could hit this assert within minutes of running
postmark, with them this issue no longer arises. Thank you,
Signed-off-by: Josef Bacik <jbacik@redhat.com>
Cc: <linux-ext4@vger.kernel.org>
Acked-by: Jan Kara <jack@ucw.cz>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Currently at the start of a journal commit we loop through all of the buffers
on the committing transaction and clear the b_modified flag (the flag that is
set when a transaction modifies the buffer) under the j_list_lock.
The problem is that everywhere else this flag is modified only under the jbd
lock buffer flag, so it will race with a running transaction who could
potentially set it, and have it unset by the committing transaction.
This is also a big waste, you can have several thousands of buffers that you
are clearing the modified flag on when you may not need to. This patch
removes this code and instead clears the b_modified flag upon entering
do_get_write_access/journal_get_create_access, so if that transaction does
indeed use the buffer then it will be accounted for properly, and if it does
not then we know we didn't use it.
That will be important for the next patch in this series. Tested thoroughly
by myself using postmark/iozone/bonnie++.
Signed-off-by: Josef Bacik <jbacik@redhat.com>
Cc: <linux-ext4@vger.kernel.org>
Acked-by: Jan Kara <jack@ucw.cz>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Commit block was intended to have several copies of the header. But
due to a bug it never had them and actually, nobody checks that. So
just remove the useless loop.
Signed-off-by: Jan Kara <jack@suse.cz>
Signed-off-by: "Theodore Ts'o" <tytso@mit.edu>
Some devices - notably dm and md - can change their behaviour in response
to BIO_RW_BARRIER requests. They might start out accepting such requests
but on reconfiguration, they find out that they cannot any more.
ext3 (and other filesystems) deal with this by always testing if
BIO_RW_BARRIER requests fail with EOPNOTSUPP, and retrying the write
requests without the barrier (probably after waiting for any pending writes
to complete).
However there is a bug in the handling for this for ext3.
When ext3 (jbd actually) decides to submit a BIO_RW_BARRIER request, it
sets the buffer_ordered flag on the buffer head. If the request completes
successfully, the flag STAYS SET.
Other code might then write the same buffer_head after the device has been
reconfigured to not accept barriers. This write will then fail, but the
"other code" is not ready to handle EOPNOTSUPP errors and the error will be
treated as fatal.
This can be seen without having to reconfigure a device at exactly the
wrong time by putting:
if (buffer_ordered(bh))
printk("OH DEAR, and ordered buffer\n");
in the while loop in "commit phase 5" of journal_commit_transaction.
If it ever prints the "OH DEAR ..." message (as it does sometimes for
me), then that request could (in different circumstances) have failed
with EOPNOTSUPP, but that isn't tested for.
My proposed fix is to clear the buffer_ordered flag after it has been
used, as in the following patch.
Signed-off-by: Neil Brown <neilb@suse.de>
Cc: <linux-ext4@vger.kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
The break_lock data structure and code for spinlocks is quite nasty.
Not only does it double the size of a spinlock but it changes locking to
a potentially less optimal trylock.
Put all of that under CONFIG_GENERIC_LOCKBREAK, and introduce a
__raw_spin_is_contended that uses the lock data itself to determine whether
there are waiters on the lock, to be used if CONFIG_GENERIC_LOCKBREAK is
not set.
Rename need_lockbreak to spin_needbreak, make it use spin_is_contended to
decouple it from the spinlock implementation, and make it typesafe (rwlocks
do not have any need_lockbreak sites -- why do they even get bloated up
with that break_lock then?).
Signed-off-by: Nick Piggin <npiggin@suse.de>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
Before we start committing a transaction, we call
__journal_clean_checkpoint_list() to cleanup transaction's written-back
buffers.
If this call happens to remove all of them (and there were already some
buffers), __journal_remove_checkpoint() will decide to free the transaction
because it isn't (yet) a committing transaction and soon we fail some
assertion - the transaction really isn't ready to be freed :).
We change the check in __journal_remove_checkpoint() to free only a
transaction in T_FINISHED state. The locking there is subtle though (as
everywhere in JBD ;(). We use j_list_lock to protect the check and a
subsequent call to __journal_drop_transaction() and do the same in the end
of journal_commit_transaction() which is the only place where a transaction
can get to T_FINISHED state.
Probably I'm too paranoid here and such locking is not really necessary -
checkpoint lists are processed only from log_do_checkpoint() where a
transaction must be already committed to be processed or from
__journal_clean_checkpoint_list() where kjournald itself calls it and thus
transaction cannot change state either. Better be safe if something
changes in future...
Signed-off-by: Jan Kara <jack@suse.cz>
Cc: <linux-ext4@vger.kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
We should really call journal_abort() and not __journal_abort_hard() in
case of errors. The latter call does not record the error in the journal
superblock and thus filesystem won't be marked as with errors later (and
user could happily mount it without any warning).
Signed-off-by: Jan Kara <jack@suse.cz>
Cc: <linux-ext4@vger.kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
JBD: Replace slab allocations with page allocations
JBD allocate memory for committed_data and frozen_data from slab. However
JBD should not pass slab pages down to the block layer. Use page allocator pages instead. This will also prepare JBD for the large blocksize patchset.
Signed-off-by: Christoph Lameter <clameter@sgi.com>
Signed-off-by: Mingming Cao <cmm@us.ibm.com>
We have to check that also the second checkpoint list is non-empty before
dropping the transaction.
Signed-off-by: Jan Kara <jack@suse.cz>
Cc: Chuck Ebbert <cebbert@redhat.com>
Cc: Kirill Korotaev <dev@openvz.org>
Cc: <linux-ext4@vger.kernel.org>
Cc: <stable@kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Remove includes of <linux/smp_lock.h> where it is not used/needed.
Suggested by Al Viro.
Builds cleanly on x86_64, i386, alpha, ia64, powerpc, sparc,
sparc64, and arm (all 59 defconfigs).
Signed-off-by: Randy Dunlap <randy.dunlap@oracle.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
In the current jbd code, if a buffer on BJ_SyncData list is dirty and not
locked, the buffer is refiled to BJ_Locked list, submitted to the IO and
waited for IO completion.
But the fsstress test showed the case that when a buffer was already
submitted to the IO just before the buffer_dirty(bh) check, the buffer was
not waited for IO completion.
Following patch solves this problem. If it is assumed that a buffer is
submitted to the IO before the buffer_dirty(bh) check and still being
written to disk, this buffer is refiled to BJ_Locked list.
Signed-off-by: Hisashi Hifumi <hifumi.hisashi@oss.ntt.co.jp>
Cc: Jan Kara <jack@ucw.cz>
Cc: "Stephen C. Tweedie" <sct@redhat.com>
Cc: <linux-ext4@vger.kernel.org>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
Many files include the filename at the beginning, serveral used a wrong one.
Signed-off-by: Uwe Zeisberger <Uwe_Zeisberger@digi.com>
Signed-off-by: Adrian Bunk <bunk@stusta.de>
Original commit code assumes, that when a buffer on BJ_SyncData list is
locked, it is being written to disk. But this is not true and hence it can
lead to a potential data loss on crash. Also the code didn't count with
the fact that journal_dirty_data() can steal buffers from committing
transaction and hence could write buffers that no longer belong to the
committing transaction. Finally it could possibly happen that we tried
writing out one buffer several times.
The patch below tries to solve these problems by a complete rewrite of the
data commit code. We go through buffers on t_sync_datalist, lock buffers
needing write out and store them in an array. Buffers are also immediately
refiled to BJ_Locked list or unfiled (if the write out is completed). When
the array is full or we have to block on buffer lock, we submit all
accumulated buffers for IO.
[suitable for 2.6.18.x around the 2.6.19-rc2 timeframe]
Signed-off-by: Jan Kara <jack@suse.cz>
Cc: Badari Pulavarty <pbadari@us.ibm.com>
Cc: <stable@kernel.org>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
