Lockstat reports have shown that j_state_lock is a major source of
lock contention, especially on systems with more than 4 CPU cores.  So
change it to be a read/write spinlock.
Signed-off-by: N"Theodore Ts'o" <tytso@mit.edu>

By using an atomic_t for t_updates and t_outstanding credits, this
should allow us to not need to take transaction t_handle_lock in
jbd2_journal_stop().
Signed-off-by: N"Theodore Ts'o" <tytso@mit.edu>

Saying things like "sync failed" when a device does
not support barriers makes users slightly more worried than
they need to be; rather than talking about sync failures,
let's just state the barrier-based facts.
Signed-off-by: NEric Sandeen <sandeen@redhat.com>
Signed-off-by: N"Theodore Ts'o" <tytso@mit.edu>

The patch just convert all blkdev_issue_xxx function to common
set of flags. Wait/allocation semantics preserved.
Signed-off-by: NDmitry Monakhov <dmonakhov@openvz.org>
Signed-off-by: NJens Axboe <jens.axboe@oracle.com>

commit_transaction has the same value as journal->j_running_transaction,
so we can simplify the assert statement.
Signed-off-by: Ndingdinghua <dingdinghua@nrchpc.ac.cn>
Signed-off-by: N"Theodore Ts'o" <tytso@mit.edu>

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.
Signed-off-by: Ndingdinghua <dingdinghua@nrchpc.ac.cn>
Signed-off-by: N"Theodore Ts'o" <tytso@mit.edu>

This is a bit complicated because we are trying to optimize when we
send barriers to the fs data disk.  We could just throw in an extra
barrier to the data disk whenever we send a barrier to the journal
disk, but that's not always strictly necessary.

We only need to send a barrier during a commit when there are data
blocks which are must be written out due to an inode written in
ordered mode, or if fsync() depends on the commit to force data blocks
to disk.  Finally, before we drop transactions from the beginning of
the journal during a checkpoint operation, we need to guarantee that
any blocks that were flushed out to the data disk are firmly on the
rust platter before we drop the transaction from the journal.

Thanks to Oleg Drokin for pointing out this flaw in ext3/ext4.
Signed-off-by: N"Theodore Ts'o" <tytso@mit.edu>

All callers really want the more logical filemap_fdatawait_range interface,
so convert them to use it and merge wait_on_page_writeback_range into
filemap_fdatawait_range.
Signed-off-by: NChristoph Hellwig <hch@lst.de>
Signed-off-by: NJan Kara <jack@suse.cz>

OOM happens.
Signed-off-by: N"Theodore Ts'o" <tytso@mit.edu>

The /proc/fs/jbd2/<dev>/history was maintained manually; by using
tracepoints, we can get all of the existing functionality of the /proc
file plus extra capabilities thanks to the ftrace infrastructure.  We
save memory as a bonus.
Signed-off-by: N"Theodore Ts'o" <tytso@mit.edu>

It's only set, it's never checked. Kill it.
Acked-by: NJan Kara <jack@suse.cz>
Signed-off-by: NJens Axboe <jens.axboe@oracle.com>

Previously the journal_async_commit mount option was equivalent to
using barrier=0 (and just as unsafe).  This patch fixes it so that we
eliminate the barrier before the commit block (by not using ordered
mode), and explicitly issuing an empty barrier bio after writing the
commit block.  Because of the journal checksum, it is safe to do this;
if the journal blocks are not all written before a power failure, the
checksum in the commit block will prevent the last transaction from
being replayed.

Using the fs_mark benchmark, using journal_async_commit shows a 50%
improvement:

FSUse%        Count         Size    Files/sec     App Overhead
     8         1000        10240         30.5            28242

vs.

FSUse%        Count         Size    Files/sec     App Overhead
     8         1000        10240         45.8            28620
Signed-off-by: N"Theodore Ts'o" <tytso@mit.edu>

Signed-off-by: N"Theodore Ts'o" <tytso@mit.edu>

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: N"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: NJens Axboe <jens.axboe@oracle.com>
Signed-off-by: NLinus 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: N"Theodore Ts'o" <tytso@mit.edu>

Filesystems often to do compute intensive operation on some
metadata.  If this operation is repeated many times, it can be very
expensive.  It would be much nicer if the operation could be performed
once before a buffer goes to disk.

This adds triggers to jbd2 buffer heads.  Just before writing a metadata
buffer to the journal, jbd2 will optionally call a commit trigger associated
with the buffer.  If the journal is aborted, an abort trigger will be
called on any dirty buffers as they are dropped from pending
transactions.

ocfs2 will use this feature.

Initially I tried to come up with a more generic trigger that could be
used for non-buffer-related events like transaction completion.  It
doesn't tie nicely, because the information a buffer trigger needs
(specific to a journal_head) isn't the same as what a transaction
trigger needs (specific to a tranaction_t or perhaps journal_t).  So I
implemented a buffer set, with the understanding that
journal/transaction wide triggers should be implemented separately.

There is only one trigger set allowed per buffer.  I can't think of any
reason to attach more than one set.  Contrast this with a journal or
transaction in which multiple places may want to watch the entire
transaction separately.

The trigger sets are considered static allocation from the jbd2
perspective.  ocfs2 will just have one trigger set per block type,
setting the same set on every bh of the same type.
Signed-off-by: NJoel Becker <joel.becker@oracle.com>
Cc: "Theodore Ts'o" <tytso@mit.edu>
Cc: <linux-ext4@vger.kernel.org>
Signed-off-by: NMark Fasheh <mfasheh@suse.com>

Since we will be waiting the write of the commit record to the journal
to complete in journal_submit_commit_record(), submit it using
WRITE_SYNC.
Signed-off-by: N"Theodore Ts'o" <tytso@mit.edu>

Avoid freeing the transaction in __jbd2_journal_drop_transaction() so
the journal commit callback can run without holding j_list_lock, to
avoid lock contention on this spinlock.
Signed-off-by: NAneesh Kumar K.V <aneesh.kumar@linux.vnet.ibm.com>
Signed-off-by: N"Theodore Ts'o" <tytso@mit.edu>

This patch removes the static sleep time in favor of a more self
optimizing approach where we measure the average amount of time it
takes to commit a transaction to disk and the ammount of time a
transaction has been running.  If somebody does a sync write or an
fsync() traditionally we would sleep for 1 jiffies, which depending on
the value of HZ could be a significant amount of time compared to how
long it takes to commit a transaction to the underlying storage.  With
this patch instead of sleeping for a jiffie, we check to see if the
amount of time this transaction has been running is less than the
average commit time, and if it is we sleep for the delta using
schedule_hrtimeout to give us a higher precision sleep time.  This
greatly benefits high end storage where you could end up sleeping for
longer than it takes to commit the transaction and therefore sitting
idle instead of allowing the transaction to be committed by keeping
the sleep time to a minimum so you are sure to always be doing
something.
Signed-off-by: NJosef Bacik <jbacik@redhat.com>
Signed-off-by: N"Theodore Ts'o" <tytso@mit.edu>

Xen doesn't report that barriers are not supported until buffer I/O is
reported as completed, instead of when the buffer I/O is submitted.
Add a check and a fallback codepath to journal_wait_on_commit_record()
to detect this case, so that attempts to mount ext4 filesystems on
LVM/devicemapper devices on Xen guests don't blow up with an "Aborting
journal on device XXX"; "Remounting filesystem read-only" error.

Thanks to Andreas Sundstrom for reporting this issue.
Signed-off-by: N"Theodore Ts'o" <tytso@mit.edu>
Cc: stable@kernel.org

The transaction can potentially get dropped if there are no buffers
that need to be written.  Make sure we call the commit callback before
potentially deciding to drop the transaction.  Also avoid
dereferencing the commit_transaction pointer in the marker for the
same reason.

This patch fixes the bug reported by Eric Paris at:
http://bugzilla.kernel.org/show_bug.cgi?id=11838Signed-off-by: N"Theodore Ts'o" <tytso@mit.edu>
Acked-by: NEric Sandeen <sandeen@redhat.com>
Tested-by: NEric Paris <eparis@redhat.com>

The multiblock allocator needs to be able to release blocks (and issue
a blkdev discard request) when the transaction which freed those
blocks is committed.  Previously this was done via a polling mechanism
when blocks are allocated or freed.  A much better way of doing things
is to create a jbd2 callback function and attaching the list of blocks
to be freed directly to the transaction structure.
Signed-off-by: N"Theodore Ts'o" <tytso@mit.edu>

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 an ext4 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.

Here is the corresponding patch of the ext3 version:
http://kerneltrap.org/mailarchive/linux-kernel/2008/9/9/3239374Signed-off-by: NHidehiro Kawai <hidehiro.kawai.ez@hitachi.com>
Signed-off-by: NTheodore Ts'o <tytso@mit.edu>

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 gets aborted, 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: NHidehiro Kawai <hidehiro.kawai.ez@hitachi.com>
Signed-off-by: NTheodore Ts'o <tytso@mit.edu>

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.

We can also avoid this kind of corruption by using the journal
checksum feature because it can detect invalid metadata blocks in the
journal and avoid them from being replayed.  So we don't need to care
about asynchronous commit record writeout with a checksum.
Signed-off-by: NHidehiro Kawai <hidehiro.kawai.ez@hitachi.com>
Signed-off-by: NTheodore Ts'o <tytso@mit.edu>

Also make sure the buffer heads are marked clean before submitting bh
for writing.  The previous code was marking the buffer head dirty,
which would have forced an unneeded write (and seek) to the journal
for no good reason.
Signed-off-by: N"Theodore Ts'o" <tytso@mit.edu>

This debugging markers are designed to debug problems such as the
random filesystem latency problems reported by Arjan.
Signed-off-by: N"Theodore Ts'o" <tytso@mit.edu>

Calculate the journal device name once and stash it away in the
journal_s structure.  This avoids needing to call bdevname()
everywhere and reduces stack usage by not needing to allocate an
on-stack buffer.  In addition, we eliminate the '/' that can appear in
device names (e.g. "cciss/c0d0p9" --- see kernel bugzilla #11321) that
can cause problems when creating proc directory names, and include the
inode number to support ocfs2 which creates multiple journals with
different inode numbers.
Signed-off-by: N"Theodore Ts'o" <tytso@mit.edu>

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: NNick Piggin <npiggin@suse.de>
Acked-by: NKOSAKI Motohiro <kosaki.motohiro@jp.fujitsu.com>
Acked-by: NPeter Zijlstra <peterz@infradead.org>
Acked-by: NAndrew Morton <akpm@linux-foundation.org>
Acked-by: NBenjamin Herrenschmidt <benh@kernel.crashing.org>
Signed-off-by: NLinus Torvalds <torvalds@linux-foundation.org>

In ordered mode, the current jbd2 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.  Unlike a similar patch for ext3/jbd, file data buffers are
written via generic_writepages().  But we also need to set AS_EIO
into their mappings because wait_on_page_writeback_range() clears
AS_EIO before a user process sees it.
Signed-off-by: NHidehiro Kawai <hidehiro.kawai.ez@hitachi.com>
Signed-off-by: N"Theodore Ts'o" <tytso@mit.edu>

This provides a new ordered mode implementation which gets rid of using
buffer heads to enforce the ordering between metadata change with the
related data chage.  Instead, in the new ordering mode, it keeps track
of all of the inodes touched by each transaction on a list, and when
that transaction is committed, it flushes all of the dirty pages for
those inodes.  In addition, the new ordered mode reverses the lock
ordering of the page lock and transaction lock, which provides easier
support for delayed allocation.
Signed-off-by: NAneesh Kumar K.V <aneesh.kumar@linux.vnet.ibm.com>
Signed-off-by: NMingming Cao <cmm@us.ibm.com>
Signed-off-by: N"Theodore Ts'o" <tytso@mit.edu>

Signed-off-by: NJan Kara <jack@suse.cz>

This patch adds necessary framework into JBD2 to be able to track inodes
with each transaction and write-out their dirty data during transaction
commit time.

This new ordered mode brings all sorts of advantages such as possibility 
to get rid of journal heads and buffer heads for data buffers in ordered 
mode, better ordering of writes on transaction commit, simplification of
 some JBD code, no more anonymous pages when truncate of data being 
committed happens.  Also with this new ordered mode, delayed allocation 
on ordered mode is much simpler.
Signed-off-by: NJan Kara <jack@suse.cz>

Carlo Wood has demonstrated that it's possible to recover deleted
files from the journal.  Something that will make this easier is if we
can put the time of the commit into commit block.
Signed-off-by: N"Theodore Ts'o" <tytso@mit.edu>

If the device doesn't support write barriers, the write is retried
without ordered mode.  But the buffer head needs to be re-locked or
submit_bh will fail with on BUG(!buffer_locked(bh)).
Signed-off-by: N"Theodore Ts'o" <tytso@mit.edu>

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.
Acked-by: NJan Kara <jack@suse.cz>
Signed-off-by: NMingming Cao <cmm@us.ibm.com>
Signed-off-by: N"Theodore Ts'o" <tytso@mit.edu>
Signed-off-by: NAndrew Morton <akpm@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.

Cc: <linux-ext4@vger.kernel.org>
Cc: Jan Kara <jack@ucw.cz>
Signed-off-by: NJosef Bacik <jbacik@redhat.com>
Signed-off-by: NAndrew Morton <akpm@linux-foundation.org>
Signed-off-by: N"Theodore Ts'o" <tytso@mit.edu>

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 jbd2
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++.

Cc: <linux-ext4@vger.kernel.org>
Cc: Jan Kara <jack@ucw.cz>
Signed-off-by: NJosef Bacik <jbacik@redhat.com>
Signed-off-by: NAndrew Morton <akpm@linux-foundation.org>
Signed-off-by: N"Theodore Ts'o" <tytso@mit.edu>

In JBD2 jbd2_journal_write_commit_record(), clear the buffer_ordered
flag for the bh after barried IO has succeed. This prevents later, if
the same buffer head were submitted to the underlying device, which has
been reconfigured to not support barrier request, the JBD2 commit code
could treat it as a normal IO (without barrier).

This is a port from JBD/ext3 fix from Neil Brown.

More details from Neil:

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. JBD2 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 this in JBD2 for ext4 .

When ext4/JBD2 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.

Cc:  Neil Brown <neilb@suse.de>
Signed-off-by: NDave Kleikamp <shaggy@linux.vnet.ibm.com>
Signed-off-by: NMingming Cao <cmm@us.ibm.com>
Signed-off-by: N"Theodore Ts'o" <tytso@mit.edu>