Implement the new readahead aop and convert all callers (block_dev,
exfat, ext2, fat, gfs2, hpfs, isofs, jfs, nilfs2, ocfs2, omfs, qnx6,
reiserfs & udf).
The callers are all trivial except for GFS2 & OCFS2.
Signed-off-by: Matthew Wilcox (Oracle) <willy@infradead.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Reviewed-by: Junxiao Bi <junxiao.bi@oracle.com> # ocfs2
Reviewed-by: Joseph Qi <joseph.qi@linux.alibaba.com> # ocfs2
Reviewed-by: Dave Chinner <dchinner@redhat.com>
Reviewed-by: John Hubbard <jhubbard@nvidia.com>
Reviewed-by: Christoph Hellwig <hch@lst.de>
Reviewed-by: William Kucharski <william.kucharski@oracle.com>
Cc: Chao Yu <yuchao0@huawei.com>
Cc: Cong Wang <xiyou.wangcong@gmail.com>
Cc: Darrick J. Wong <darrick.wong@oracle.com>
Cc: Eric Biggers <ebiggers@google.com>
Cc: Gao Xiang <gaoxiang25@huawei.com>
Cc: Jaegeuk Kim <jaegeuk@kernel.org>
Cc: Michal Hocko <mhocko@suse.com>
Cc: Zi Yan <ziy@nvidia.com>
Cc: Johannes Thumshirn <johannes.thumshirn@wdc.com>
Cc: Miklos Szeredi <mszeredi@redhat.com>
Link: http://lkml.kernel.org/r/20200414150233.24495-17-willy@infradead.org
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Fix several issues in the previous gfs2_find_jhead fix:
* When updating @blocks_submitted, @block refers to the first block block not
submitted yet, not the last block submitted, so fix an off-by-one error.
* We want to ensure that @blocks_submitted is far enough ahead of @blocks_read
to guarantee that there is in-flight I/O. Otherwise, we'll eventually end up
waiting for pages that haven't been submitted, yet.
* It's much easier to compare the number of blocks added with the number of
blocks submitted to limit the maximum bio size.
* Even with bio chaining, we can keep adding blocks until we reach the maximum
bio size, as long as we stop at a page boundary. This simplifies the logic.
Signed-off-by: Andreas Gruenbacher <agruenba@redhat.com>
Reviewed-by: Bob Peterson <rpeterso@redhat.com>
This reverts commit df5db5f9ee.
This patch fixes a regression: patch df5db5f9ee allowed function
run_queue() to bypass its call to do_xmote() if revokes were queued for
the glock. That's wrong because its call to do_xmote() is what is
responsible for calling the go_sync() glops functions to sync both
the ail list and any revokes queued for it. By bypassing the call,
gfs2 could get into a stand-off where the glock could not be demoted
until its revokes are written back, but the revokes would not be
written back because do_xmote() was never called.
It "sort of" works, however, because there are other mechanisms like
the log flush daemon (logd) that can sync the ail items and revokes,
if it deems it necessary. The problem is: without file system pressure,
it might never deem it necessary.
Signed-off-by: Bob Peterson <rpeterso@redhat.com>
Before this patch, if the go_sync operation returned an error during
the do_xmote process (such as unable to sync metadata to the journal)
the code did goto out. That kept the glock locked, so it could not be
given away, which correctly avoids file system corruption. However,
it never set the withdraw bit or requeueing the glock work. So it would
hang forever, unable to ever demote the glock.
This patch changes to goto to a new label, skip_inval, so that errors
from go_sync are treated the same way as errors from go_inval:
The delayed withdraw bit is set and the work is requeued. That way,
the logd should eventually figure out there's a problem and withdraw
properly there.
Signed-off-by: Bob Peterson <rpeterso@redhat.com>
Signed-off-by: Andreas Gruenbacher <agruenba@redhat.com>
This patch rearranges gfs2_add_revoke so that the extra glock
reference is added earlier on in the function to avoid races in which
the glock is freed before the new reference is taken.
Signed-off-by: Andreas Gruenbacher <agruenba@redhat.com>
Signed-off-by: Bob Peterson <rpeterso@redhat.com>
Before this patch, function gfs2_quota_unlock checked if quotas are
turned off, and if so, it branched to label out, which called
gfs2_quota_unhold. With the new system of gfs2_qa_get and put, we
no longer want to call gfs2_quota_unhold or we won't balance our
gets and puts.
Signed-off-by: Bob Peterson <rpeterso@redhat.com>
Signed-off-by: Andreas Gruenbacher <agruenba@redhat.com>
Before this patch, function gfs2_quota_lock checked if it was called
from a privileged user, and if so, it bypassed the quota check:
superuser can operate outside the quotas.
That's the wrong place for the check because the lock/unlock functions
are separate from the lock_check function, and you can do lock and
unlock without actually checking the quotas.
This patch moves the check to gfs2_quota_lock_check.
Signed-off-by: Bob Peterson <rpeterso@redhat.com>
Signed-off-by: Andreas Gruenbacher <agruenba@redhat.com>
This patch removes a check from gfs2_quota_check for whether quotas
are enabled by the superblock. There is a test just prior for the
GIF_QD_LOCKED bit in the inode, and that can only be set by functions
that already check that quotas are enabled in the superblock.
Therefore, the check is redundant.
Signed-off-by: Bob Peterson <rpeterso@redhat.com>
Signed-off-by: Andreas Gruenbacher <agruenba@redhat.com>
Before this patch, gfs2_quota_change() would BUG_ON if the
qa_ref counter was not a positive number. This patch changes it to
be a withdraw instead. That way we can debug things more easily.
Signed-off-by: Bob Peterson <rpeterso@redhat.com>
Signed-off-by: Andreas Gruenbacher <agruenba@redhat.com>
This patch fixes a couple of places in which gfs2_qa_get and gfs2_qa_put are
not balanced: we now keep references around whenever a file is open for writing
(see gfs2_open_common and gfs2_release), so we need to put all references we
grab in function gfs2_create_inode. This was broken in the successful case and
on one error path.
This also means that we don't have a reference to put in gfs2_evict_inode.
In addition, gfs2_qa_put was called for the wrong inode in gfs2_link.
Signed-off-by: Bob Peterson <rpeterso@redhat.com>
Signed-off-by: Andreas Gruenbacher <agruenba@redhat.com>
It turns out that when extending an existing bio, gfs2_find_jhead fails to
check if the block number is consecutive, which leads to incorrect reads for
fragmented journals.
In addition, limit the maximum bio size to an arbitrary value of 2 megabytes:
since commit 07173c3ec2 ("block: enable multipage bvecs"), if we just keep
adding pages until bio_add_page fails, bios will grow much larger than useful,
which pins more memory than necessary with barely any additional performance
gains.
Fixes: f4686c26ec ("gfs2: read journal in large chunks")
Cc: stable@vger.kernel.org # v5.2+
Signed-off-by: Andreas Gruenbacher <agruenba@redhat.com>
Signed-off-by: Bob Peterson <rpeterso@redhat.com>
Make sure we don't walk past the end of the metadata in gfs2_walk_metadata: the
inode holds fewer pointers than indirect blocks.
Slightly clean up gfs2_iomap_get.
Fixes: a27a0c9b6a ("gfs2: gfs2_walk_metadata fix")
Cc: stable@vger.kernel.org # v5.3+
Signed-off-by: Andreas Gruenbacher <agruenba@redhat.com>
Signed-off-by: Bob Peterson <rpeterso@redhat.com>
When the gfs2_logd daemon withdrew, the withdraw sequence called
into make_fs_ro() to make the file system read-only. That caused the
journal descriptors to be freed. However, those journal descriptors
were used by gfs2_logd's call to gfs2_ail_flush_reqd(). This caused
a use-after free and NULL pointer dereference.
This patch changes function gfs2_logd() so that it stops all logd
work until the thread is told to stop. Once a withdraw is done,
it only does an interruptible sleep.
Signed-off-by: Bob Peterson <rpeterso@redhat.com>
Signed-off-by: Andreas Gruenbacher <agruenba@redhat.com>
Before this patch, when the logd daemon was forced to withdraw, it
would try to request its journal be recovered by another cluster node.
However, in single-user cases with lock_nolock, there are no other
nodes to recover the journal. Function signal_our_withdraw() was
recognizing the lock_nolock situation, but not until after it had
evicted its journal inode. Since the journal descriptor that points
to the inode was never removed from the master list, when the unmount
occurred, it did another iput on the evicted inode, which resulted in
a BUG_ON(inode->i_state & I_CLEAR).
This patch moves the check for this situation earlier in function
signal_our_withdraw(), which avoids the extra iput, so the unmount
may happen normally.
Signed-off-by: Bob Peterson <rpeterso@redhat.com>
Signed-off-by: Andreas Gruenbacher <agruenba@redhat.com>
Before this patch, if an error was detected from glock function go_sync
by function do_xmote, it would return. But the function had temporarily
unlocked the gl_lockref spin_lock, and it never re-locked it. When the
caller of do_xmote tried to unlock it again, it was already unlocked,
which resulted in a corrupted spin_lock value.
This patch makes sure the gl_lockref spin_lock is re-locked after it is
unlocked.
Thanks to Wu Bo <wubo40@huawei.com> for reporting this problem.
Signed-off-by: Bob Peterson <rpeterso@redhat.com>
Signed-off-by: Andreas Gruenbacher <agruenba@redhat.com>
After a gfs2 file system withdraw, any attempt to read metadata is
automatically rejected by function gfs2_meta_read() except for reads
of the journal inode. This turns out to be a problem because function
signal_our_withdraw() repeatedly calls check_journal_clean() which reads
the metadata (both its dinode and indirect blocks) to see if the entire
journal is mapped. The dinode read works, but reading the indirect blocks
returns -EIO which gets sent back up and causes a consistency error.
This results in withdraw-from-withdraw, which becomes a deadlock.
This patch changes the test in gfs2_meta_read() to allow all metadata
reads for the journal. Instead of checking the journal block, it now
checks for the journal inode glock which is the same for all blocks in
the journal. This allows check_journal_clean() to properly check the
journal without trying to withdraw recursively.
Signed-off-by: Bob Peterson <rpeterso@redhat.com>
Signed-off-by: Andreas Gruenbacher <agruenba@redhat.com>
are related to corruption that occurs when journals are replayed.
For example:
1. A node fails while writing to the file system.
2. Other nodes use the metadata that was once used by the failed node.
3. When the node returns to the cluster, its journal is replayed,
but the older metadata blocks overwrite the changes from step 2.
- Fixed the recovery sequence to prevent corruption during journal replay.
- Many bug fixes found during recovery testing.
- New improved file system withdraw sequence.
- Fixed how resource group buffers are managed.
- Fixed how metadata revokes are tracked and written.
- Improve processing of IO errors hit by daemons like logd and quotad.
- Improved error checking in metadata writes.
- Fixed how qadata quota data structures are managed.
-----BEGIN PGP SIGNATURE-----
iQEzBAABCAAdFiEE89F0ZrnZapxy/9qS14th09/3ejsFAl6Db/QACgkQ14th09/3
ejvVTgf+IdHXfmpv3ftah8lDDpbsnSKZYRC1NW7skQB+NVG9KtJhtzy1nldaMqMv
s8wQ5aGKrfBfmzg8IZ9Pt3dCItFqC5d8IqcO0M0FtNuyN+27ETUUMnqBf1NwL6wI
iAm/+ncZ/BiZN2P8MgXV3OgRGvaC9ebmz860+nthwyJT+6y8d8Qab7pUfyix5e0d
oTgDhEJqF0DOrGsrlS5rxjTU+RMixtepsAW958D4Eks28OlyduRAj6fAMDoLN2/E
WoDpX6iKeczH0lOZxnIVQOkCztDaa0jDlK2JK7sJRBMpNxj77aUn4cffY+b/A4kk
sR5gjsiHoesdAMEpHIXSdEcYMIstIg==
=VEKB
-----END PGP SIGNATURE-----
Merge tag 'gfs2-for-5.7' of git://git.kernel.org/pub/scm/linux/kernel/git/gfs2/linux-gfs2
Pull gfs2 updates from Bob Peterson:
"We've got a lot of patches (39) for this merge window. Most of these
patches are related to corruption that occurs when journals are
replayed. For example:
1. A node fails while writing to the file system.
2. Other nodes use the metadata that was once used by the failed
node.
3. When the node returns to the cluster, its journal is replayed, but
the older metadata blocks overwrite the changes from step 2.
Summary:
- Fixed the recovery sequence to prevent corruption during journal
replay.
- Many bug fixes found during recovery testing.
- New improved file system withdraw sequence.
- Fixed how resource group buffers are managed.
- Fixed how metadata revokes are tracked and written.
- Improve processing of IO errors hit by daemons like logd and
quotad.
- Improved error checking in metadata writes.
- Fixed how qadata quota data structures are managed"
* tag 'gfs2-for-5.7' of git://git.kernel.org/pub/scm/linux/kernel/git/gfs2/linux-gfs2: (39 commits)
gfs2: Fix oversight in gfs2_ail1_flush
gfs2: change from write to read lock for sd_log_flush_lock in journal replay
gfs2: instrumentation wrt ail1 stuck
gfs2: don't lock sd_log_flush_lock in try_rgrp_unlink
gfs2: Remove unnecessary gfs2_qa_{get,put} pairs
gfs2: Split gfs2_rsqa_delete into gfs2_rs_delete and gfs2_qa_put
gfs2: Change inode qa_data to allow multiple users
gfs2: eliminate gfs2_rsqa_alloc in favor of gfs2_qa_alloc
gfs2: Switch to list_{first,last}_entry
gfs2: Clean up inode initialization and teardown
gfs2: Additional information when gfs2_ail1_flush withdraws
gfs2: leaf_dealloc needs to allocate one more revoke
gfs2: allow journal replay to hold sd_log_flush_lock
gfs2: don't allow releasepage to free bd still used for revokes
gfs2: flesh out delayed withdraw for gfs2_log_flush
gfs2: Do proper error checking for go_sync family of glops functions
gfs2: Don't demote a glock until its revokes are written
gfs2: drain the ail2 list after io errors
gfs2: Withdraw in gfs2_ail1_flush if write_cache_pages fails
gfs2: Do log_flush in gfs2_ail_empty_gl even if ail list is empty
...
Ordinarily, function gfs2_ail1_start_one issues a write request
for one item on the ail1 list, then returns -EBUSY. This makes the
caller, gfs2_ail1_flush, loop around and start another. However,
it was not clearing the -EBUSY return code each time through the loop.
So on rare occasions, like when the wbc runs out of nr_to_write, it
remained set to -EBUSY, which triggered an error and withdraw.
This patch sets the return code to 0 each time through the restart
loop so this won't happen anymore.
Signed-off-by: Bob Peterson <rpeterso@redhat.com>
Function gfs2_recover_func grabs the sd_log_flush_lock rw_semaphore in
write mode. This is unnecessary because we only need to prevent log flush
from using sd_log_bio bio while it does. Therefore, a read lock will be
enough. This is a small step in cleaning up log flush.
Signed-off-by: Bob Peterson <rpeterso@redhat.com>
Before this patch, if the ail1 flush got stuck for some reason, there
were no clues as to why. This patch introduces a check for getting
stuck for more than a minute, and if it happens, it dumps the items
still remaining on the ail1 list.
Signed-off-by: Bob Peterson <rpeterso@redhat.com>
In function try_rgrp_unlink, we added a temporary lock of the
sd_log_flush_lock while searching the bitmaps. This protected us from
problems in which dinodes being freed were still in a state of flux
because the rgrp was in an active transaction. It was a kludge.
Now that we've straightened out the code for inode eviction, deletes,
and all the recovery mess, we no longer need this kludge.
This patch removes it, and should improve performance.
Signed-off-by: Bob Peterson <rpeterso@redhat.com>
We now get the quota data structure when opening a file writable and put it
when closing that writable file descriptor, so there no longer is a need for
gfs2_qa_{get,put} while we're holding a writable file descriptor.
Signed-off-by: Andreas Gruenbacher <agruenba@redhat.com>
Signed-off-by: Bob Peterson <rpeterso@redhat.com>
Keeping reservations and quotas separate helps reviewing the code.
Signed-off-by: Andreas Gruenbacher <agruenba@redhat.com>
Signed-off-by: Bob Peterson <rpeterso@redhat.com>
Before this patch, multiple users called gfs2_qa_alloc which allocated
a qadata structure to the inode, if quotas are turned on. Later, in
file close or evict, the structure was deleted with gfs2_qa_delete.
But there can be several competing processes who need access to the
structure. There were races between file close (release) and the others.
Thus, a release could delete the structure out from under a process
that relied upon its existence. For example, chown.
This patch changes the management of the qadata structures to be
a get/put scheme. Function gfs2_qa_alloc has been changed to gfs2_qa_get
and if the structure is allocated, the count essentially starts out at
1. Function gfs2_qa_delete has been renamed to gfs2_qa_put, and the
last guy to decrement the count to 0 frees the memory.
Signed-off-by: Bob Peterson <rpeterso@redhat.com>
Before this patch, multiple callers called gfs2_rsqa_alloc to force
the existence of a reservations structure and a quota data structure
if needed. However, now the reservations are handled separately, so
the quota data is only the quota data. So we eliminate the one in
favor of just calling gfs2_qa_alloc directly.
Signed-off-by: Bob Peterson <rpeterso@redhat.com>
Replace open-coded versions of list_first_entry and list_last_entry with those
functions.
Signed-off-by: Andreas Gruenbacher <agruenba@redhat.com>
Signed-off-by: Bob Peterson <rpeterso@redhat.com>
When allocating a new inode, mark the iopen glock holder as uninitialized to
make sure gfs2_evict_inode won't fail after an incomplete create or lookup. In
gfs2_evict_inode, allow the inode glock to be NULL and remove the duplicate
iopen glock teardown code. In gfs2_inode_lookup, don't tear down things that
gfs2_evict_inode will already tear down.
Signed-off-by: Andreas Gruenbacher <agruenba@redhat.com>
Signed-off-by: Bob Peterson <rpeterso@redhat.com>
with the way fs/namei.c:do_last() had been done, ->atomic_open()
instances needed to recognize the case when existing file got
found with O_EXCL|O_CREAT, either by falling back to finish_no_open()
or failing themselves. gfs2 one didn't.
Fixes: 6d4ade986f (GFS2: Add atomic_open support)
Cc: stable@kernel.org # v3.11
Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>
Before this patch, if gfs2_ail1_flush gets an error from function
gfs2_ail1_start_one (which comes indirectly from generic_writepages)
the file system is withdrawn, but without any explanation why.
This patch adds an error message if gfs2_ail1_flush gets an error
from gfs2_ail1_start_one.
Signed-off-by: Bob Peterson <rpeterso@redhat.com>
Function leaf_dealloc was not allocating enough journal space for
revokes. Before, it allocated 'l_blocks' revokes, but it needs one
more for the revoke of the dinode that is modified. This patch adds
the needed revoke entry to function leaf_dealloc.
Signed-off-by: Bob Peterson <rpeterso@redhat.com>
Before this patch, journal replays could stomp on log flushes
and each other because both log flushes and journal replays used
the same sd_log_bio. Function gfs2_log_flush prevents other log
flushes from interfering by taking the sd_log_flush_lock rwsem
during the flush. However, it does not protect against journal
replays. This patch allows the journal replay to take the same
sd_log_flush_lock rwsem so use of the sd_log_bio is not stomped.
Signed-off-by: Bob Peterson <rpeterso@redhat.com>
Before this patch, function gfs2_releasepage would free any bd
elements that had been used for the page being released. However,
those bd elements may still be queued to the sd_log_revokes list,
in which case we cannot free them until the revoke has been issued.
This patch adds additional checks for bds that are still being
used for revokes.
Signed-off-by: Bob Peterson <rpeterso@redhat.com>
Function gfs2_log_flush() had a few places where it tried to withdraw
from the file system when errors were encountered. The problem is,
it should delay those withdraws until the log flush lock is no longer
held.
This patch creates a new function just for delayed withdraws for
situations like this. If errors=panic was specified on mount, we
still want to do it the old fashioned way because the panic it does
not help to delay in that situation.
Signed-off-by: Bob Peterson <rpeterso@redhat.com>
Reviewed-by: Andreas Gruenbacher <agruenba@redhat.com>
Before this patch, function do_xmote would try to sync out the glock
dirty data by calling the appropriate glops function XXX_go_sync()
but it did not check for a good return code. If the sync was not
possible due to an io error or whatever, do_xmote would continue on
and call go_inval and release the glock to other cluster nodes.
When those nodes go to replay the journal, they may already be holding
glocks for the journal records that should have been synced, but were
not due to the ignored error.
This patch introduces proper error code checking to the go_sync
family of glops functions.
Signed-off-by: Bob Peterson <rpeterso@redhat.com>
Reviewed-by: Andreas Gruenbacher <agruenba@redhat.com>
Before this patch, run_queue would demote glocks based on whether
there are any more holders. But if the glock has pending revokes that
haven't been written to the media, giving up the glock might end in
file system corruption if the revokes never get written due to
io errors, node crashes and fences, etc. In that case, another node
will replay the metadata blocks associated with the glock, but
because the revoke was never written, it could replay that block
even though the glock had since been granted to another node who
might have made changes.
This patch changes the logic in run_queue so that it never demotes
a glock until its count of pending revokes reaches zero.
Signed-off-by: Bob Peterson <rpeterso@redhat.com>
Reviewed-by: Andreas Gruenbacher <agruenba@redhat.com>
Before this patch, gfs2_logd continually tried to flush its journal
log, after the file system is withdrawn. We don't want to write anything
to the journal, lest we add corruption. Best course of action is to
drain the ail1 into the ail2 list (via gfs2_ail1_empty) then drain the
ail2 list with a new function, ail2_drain.
Signed-off-by: Bob Peterson <rpeterso@redhat.com>
Reviewed-by: Andreas Gruenbacher <agruenba@redhat.com>
Before this patch, function gfs2_ail1_start_one would return any
errors it received from write_cache_pages (except -EBUSY) but it did
not withdraw. Since function gfs2_ail1_flush just checks for the bad
return code and loops, the loop might potentially never end.
This patch adds some logic to allow it to exit the loop and withdraw
properly when errors are received from write_cache_pages.
Signed-off-by: Bob Peterson <rpeterso@redhat.com>
Reviewed-by: Andreas Gruenbacher <agruenba@redhat.com>
Before this patch, if gfs2_ail_empty_gl saw there was nothing on
the ail list, it would return and not flush the log. The problem
is that there could still be a revoke for the rgrp sitting on the
sd_log_le_revoke list that's been recently taken off the ail list.
But that revoke still needs to be written, and the rgrp_go_inval
still needs to call log_flush_wait to ensure the revokes are all
properly written to the journal before we relinquish control of
the glock to another node. If we give the glock to another node
before we have this knowledge, the node might crash and its journal
replayed, in which case the missing revoke would allow the journal
replay to replay the rgrp over top of the rgrp we already gave to
another node, thus overwriting its changes and corrupting the
file system.
This patch makes gfs2_ail_empty_gl still call gfs2_log_flush rather
than returning.
Signed-off-by: Bob Peterson <rpeterso@redhat.com>
Reviewed-by: Andreas Gruenbacher <agruenba@redhat.com>
Before this patch, function do_xmote just assumed all the writes
submitted to the journal were finished and successful, and it
called the go_unlock function to release the dlm lock. But if
they're not, and a revoke failed to make its way to the journal,
a journal replay on another node will cause corruption if we
let the go_inval function continue and tell dlm to release the
glock to another node. This patch adds a couple checks for errors
in do_xmote after the calls to go_sync and go_inval. If an error
is found, we cannot withdraw yet, because the withdraw itself
uses glocks to make the file system read-only. Instead, we flag
the error. Later, asserts should cause another node to replay
the journal before continuing, thus protecting rgrp and dinode
glocks and maintaining the integrity of the metadata. Note that
we only need to do this for journaled glocks. System glocks
should be able to progress even under withdrawn conditions.
Signed-off-by: Bob Peterson <rpeterso@redhat.com>
Reviewed-by: Andreas Gruenbacher <agruenba@redhat.com>
Before this patch, function gfs2_end_log_write would detect any IO
errors writing to the journal and put out an appropriate message,
but it never set a withdrawing condition. Eventually, the log daemon
would see the error and determine it was time to withdraw, but in
the meantime, other processes could continue running as if nothing
bad ever happened. The biggest consequence is that __gfs2_glock_put
would BUG() when it saw that there were still unwritten items.
This patch sets the WITHDRAWING status as soon as an IO error is
detected, and that way, the BUG will be avoided so the file system
can be properly withdrawn and unmounted.
Signed-off-by: Bob Peterson <rpeterso@redhat.com>
Reviewed-by: Andreas Gruenbacher <agruenba@redhat.com>
Before this patch, function gfs2_write_revokes would call
gfs2_ail1_empty, then traverse the sd_ail1_list looking for
transactions that had bds which were no longer queued to a glock.
And if it found some, it would try to issue revokes for them, up to
a predetermined maximum. There were two problems with how it did
this. First was the fact that gfs2_ail1_empty moves transactions
which have nothing remaining on the ail1 list from the sd_ail1_list
to the sd_ail2_list, thus making its traversal of sd_ail1_list
miss them completely, and therefore, never issue revokes for them.
Second was the fact that there were three traversals (or partial
traversals) of the sd_ail1_list, each of which took and then
released the sd_ail_lock lock: First inside gfs2_ail1_empty,
second to determine if there are any revokes to be issued, and
third to actually issue them. All this taking and releasing of the
sd_ail_lock meant other processes could modify the lists and the
conditions in which we're working.
This patch simplies the whole process by adding a new parameter
to function gfs2_ail1_empty, max_revokes. For normal calls, this
is passed in as 0, meaning we don't want to issue any revokes.
For function gfs2_write_revokes, we pass in the maximum number
of revokes we can, thus allowing gfs2_ail1_empty to add the
revokes where needed. This simplies the code, allows for a single
holding of the sd_ail_lock, and allows gfs2_ail1_empty to add
revokes for all the necessary bd items without missing any.
Signed-off-by: Bob Peterson <rpeterso@redhat.com>
Reviewed-by: Andreas Gruenbacher <agruenba@redhat.com>
Before this patch, function check_journal_clean would give messages
related to journal recovery. That's fine for mount time, but when a
node withdraws and forces replay that way, we don't want all those
distracting and misleading messages. This patch adds a new parameter
to make those messages optional.
Signed-off-by: Bob Peterson <rpeterso@redhat.com>
Reviewed-by: Andreas Gruenbacher <agruenba@redhat.com>
Before this patch, the rgrp_go_inval and inode_go_inval functions each
checked if there were any items left on the ail count (by way of a
count), and if so, did a withdraw. But the withdraw code now uses
glocks when changing the file system to read-only status. So we can
not have glock functions withdrawing or a hang will likely result:
The glocks can't be serviced by the work_func if the work_func is
busy doing its own withdraw.
This patch removes the checks from the go_inval functions and adds
a centralized check in do_xmote to warn about the problem and not
withdraw, but flag the error so it's eventually caught when the logd
daemon eventually runs.
Signed-off-by: Bob Peterson <rpeterso@redhat.com>
Reviewed-by: Andreas Gruenbacher <agruenba@redhat.com>
When a node withdraws from a file system, it often leaves its journal
in an incomplete state. This is especially true when the withdraw is
caused by io errors writing to the journal. Before this patch, a
withdraw would try to write a "shutdown" record to the journal, tell
dlm it's done with the file system, and none of the other nodes
know about the problem. Later, when the problem is fixed and the
withdrawn node is rebooted, it would then discover that its own
journal was incomplete, and replay it. However, replaying it at this
point is almost guaranteed to introduce corruption because the other
nodes are likely to have used affected resource groups that appeared
in the journal since the time of the withdraw. Replaying the journal
later will overwrite any changes made, and not through any fault of
dlm, which was instructed during the withdraw to release those
resources.
This patch makes file system withdraws seen by the entire cluster.
Withdrawing nodes dequeue their journal glock to allow recovery.
The remaining nodes check all the journals to see if they are
clean or in need of replay. They try to replay dirty journals, but
only the journals of withdrawn nodes will be "not busy" and
therefore available for replay.
Until the journal replay is complete, no i/o related glocks may be
given out, to ensure that the replay does not cause the
aforementioned corruption: We cannot allow any journal replay to
overwrite blocks associated with a glock once it is held.
The "live" glock which is now used to signal when a withdraw
occurs. When a withdraw occurs, the node signals its withdraw by
dequeueing the "live" glock and trying to enqueue it in EX mode,
thus forcing the other nodes to all see a demote request, by way
of a "1CB" (one callback) try lock. The "live" glock is not
granted in EX; the callback is only just used to indicate a
withdraw has occurred.
Note that all nodes in the cluster must wait for the recovering
node to finish replaying the withdrawing node's journal before
continuing. To this end, it checks that the journals are clean
multiple times in a retry loop.
Also note that the withdraw function may be called from a wide
variety of situations, and therefore, we need to take extra
precautions to make sure pointers are valid before using them in
many circumstances.
We also need to take care when glocks decide to withdraw, since
the withdraw code now uses glocks.
Also, before this patch, if a process encountered an error and
decided to withdraw, if another process was already withdrawing,
the second withdraw would be silently ignored, which set it free
to unlock its glocks. That's correct behavior if the original
withdrawer encounters further errors down the road. But if
secondary waiters don't wait for the journal replay, unlocking
glocks will allow other nodes to use them, despite the fact that
the journal containing those blocks is being replayed. The
replay needs to finish before our glocks are released to other
nodes. IOW, secondary withdraws need to wait for the first
withdraw to finish.
For example, if an rgrp glock is unlocked by a process that didn't
wait for the first withdraw, a journal replay could introduce file
system corruption by replaying a rgrp block that has already been
granted to a different cluster node.
Signed-off-by: Bob Peterson <rpeterso@redhat.com>
We need to allow some glocks to be enqueued, dequeued, promoted, and demoted
when we're withdrawn. For example, to maintain metadata integrity, we should
disallow the use of inode and rgrp glocks when withdrawn. Other glocks, like
iopen or the transaction glocks may be safely used because none of their
metadata goes through the journal. So in general, we should disallow all
glocks with an address space, and allow all the others. One exception is:
we need to allow our active journal to be demoted so others may recover it.
Allowing glocks after withdraw gives us the ability to take appropriate
action (in a following patch) to have our journal properly replayed by
another node rather than just abandoning the current transactions and
pretending nothing bad happened, leaving the other nodes free to modify
the blocks we had in our journal, which may result in file system
corruption.
Signed-off-by: Bob Peterson <rpeterso@redhat.com>
Before this patch function check_journal_clean was in ops_fstype.c.
This patch moves it to util.c so we can make use of it elsewhere
in a future patch.
Signed-off-by: Bob Peterson <rpeterso@redhat.com>
Reviewed-by: Andreas Gruenbacher <agruenba@redhat.com>
When a node fails, user space informs dlm of the node failure,
and dlm instructs gfs2 on the surviving nodes to perform journal
recovery. It does this by calling various callback functions in
lock_dlm.c. To mark its progress, it keeps generation numbers
and recover bits in a dlm "control" lock lvb, which is seen by
all nodes to determine which journals need to be replayed.
The gfs2 on all nodes get the same recovery requests from dlm,
so they all try to do the recovery, but only one will be
granted the exclusive lock on the journal. The others fail
with a "Busy" message on their "try lock."
However, when a node is withdrawn, it cannot safely do any
recovery or replay any journals. To make matters worse,
gfs2 might withdraw as a result of attempting recovery. For
example, this might happen if the device goes offline, or if
an hba fails. But in today's gfs2 code, it doesn't check for
being withdrawn at any step in the recovery process. What's
worse is that these callbacks from dlm have no return code,
so there is no way to indicate failure back to dlm. We can
send a "Recovery failed" uevent eventually, but that tells
user space what happened, not dlm's kernel code.
Before this patch, lock_dlm would perform its recovery steps but
ignore the result, and eventually it would still update its
generation number in the lvb, despite the fact that it may have
withdrawn or encountered an error. The other nodes would then
see the newer generation number in the lvb and conclude that
they don't need to do recovery because the generation number
is newer than the last one they saw. They think a different
node has already recovered the journal.
This patch adds checks to several of the callbacks used by dlm
in its recovery state machine so that the functions are ignored
and skipped if an io error has occurred or if the file system
is withdrawn. That prevents the lvb bits from being updated, and
therefore dlm and user space still see the need for recovery to
take place.
Signed-off-by: Bob Peterson <rpeterso@redhat.com>
Reviewed-by: Andreas Gruenbacher <agruenba@redhat.com>
Before this patch, all io errors received by the quota daemon or the
logd daemon would cause a complaint message to be issued, such as:
gfs2: fsid=dm-13.0: Error 10 writing to journal, jid=0
This patch changes it so that the error message is only issued the
first time the error is encountered.
Also, before this patch function gfs2_end_log_write did not set the
sd_log_error value, so log errors would not cause the file system to
be withdrawn. This patch sets the error code so the file system is
properly withdrawn if an io error is encountered writing to the journal.
WARNING: This change in function breaks check xfstests generic/441
and causes it to fail: io errors writing to the log should cause a
file system to be withdrawn, and no further operations are tolerated.
Signed-off-by: Bob Peterson <rpeterso@redhat.com>
Reviewed-by: Andreas Gruenbacher <agruenba@redhat.com>
Before this patch, gfs2 kept track of journal io errors in two
places sd_log_error and the SDF_AIL1_IO_ERROR flag in sd_flags.
This patch consolidates the two into sd_log_error so that it
reflects the first error encountered writing to the journal.
In future patches, we will take advantage of this by checking
this value rather than having to check both when reacting to
io errors.
In addition, this fixes a tight loop in unmount: If buffers
get on the ail1 list and an io error occurs elsewhere, the
ail1 list would never be cleared because they were always busy.
So unmount would hang, waiting for the ail1 list to empty.
Signed-off-by: Bob Peterson <rpeterso@redhat.com>
Reviewed-by: Andreas Gruenbacher <agruenba@redhat.com>
Before this patch, the rgrp code had a serious problem related to
how it managed buffer_heads for resource groups. The problem caused
file system corruption, especially in cases of journal replay.
When an rgrp glock was demoted to transfer ownership to a
different cluster node, do_xmote() first calls rgrp_go_sync and then
rgrp_go_inval, as expected. When it calls rgrp_go_sync, that called
gfs2_rgrp_brelse() that dropped the buffer_head reference count.
In most cases, the reference count went to zero, which is right.
However, there were other places where the buffers are handled
differently.
After rgrp_go_sync, do_xmote called rgrp_go_inval which called
gfs2_rgrp_brelse a second time, then rgrp_go_inval's call to
truncate_inode_pages_range would get rid of the pages in memory,
but only if the reference count drops to 0.
Unfortunately, gfs2_rgrp_brelse was setting bi->bi_bh = NULL.
So when rgrp_go_sync called gfs2_rgrp_brelse, it lost the pointer
to the buffer_heads in cases where the reference count was still 1.
Therefore, when rgrp_go_inval called gfs2_rgrp_brelse a second time,
it failed the check for "if (bi->bi_bh)" and thus failed to call
brelse a second time. Because of that, the reference count on those
buffers sometimes failed to drop from 1 to 0. And that caused
function truncate_inode_pages_range to keep the pages in page cache
rather than freeing them.
The next time the rgrp glock was acquired, the metadata read of
the rgrp buffers re-used the pages in memory, which were now
wrong because they were likely modified by the other node who
acquired the glock in EX (which is why we demoted the glock).
This re-use of the page cache caused corruption because changes
made by the other nodes were never seen, so the bitmaps were
inaccurate.
For some reason, the problem became most apparent when journal
replay forced the replay of rgrps in memory, which caused newer
rgrp data to be overwritten by the older in-core pages.
A big part of the problem was that the rgrp buffer were released
in multiple places: The go_unlock function would release them when
the glock was released rather than when the glock is demoted,
which is clearly wrong because our intent was to cache them until
the glock is demoted from SH or EX.
This patch attempts to clean up the mess and make one consistent
and centralized mechanism for managing the rgrp buffer_heads by
implementing several changes:
1. It eliminates the call to gfs2_rgrp_brelse() from rgrp_go_sync.
We don't want to release the buffers or zero the pointers when
syncing for the reasons stated above. It only makes sense to
release them when the glock is actually invalidated (go_inval).
And when we do, then we set the bh pointers to NULL.
2. The go_unlock function (which was only used for rgrps) is
eliminated, as we've talked about doing many times before.
The go_unlock function was called too early in the glock dq
process, and should not happen until the glock is invalidated.
3. It also eliminates the call to rgrp_brelse in gfs2_clear_rgrpd.
That will now happen automatically when the rgrp glocks are
demoted, and shouldn't happen any sooner or later than that.
Instead, function gfs2_clear_rgrpd has been modified to demote
the rgrp glocks, and therefore, free those pages, before the
remaining glocks are culled by gfs2_gl_hash_clear. This
prevents the gl_object from hanging around when the glocks are
culled.
Signed-off-by: Bob Peterson <rpeterso@redhat.com>
Reviewed-by: Andreas Gruenbacher <agruenba@redhat.com>
