Before this patch, if GFS2 encountered IO errors while writing to
the journal, it would not report the problem, so they would go
unnoticed, sometimes for many hours. Sometimes this would only be
noticed later, when recovery tried to do journal replay and failed
due to invalid metadata at the blocks that resulted in IO errors.
This patch makes GFS2's log daemon check for IO errors. If it
encounters one, it withdraws from the file system and reports
why in dmesg. A similar action is taken when IO errors occur when
writing to the system statfs file.
These errors are also reported back to any callers of fsync, since
that requires the journal to be flushed. Therefore, any IO errors
that would previously go unnoticed are now noticed and the file
system is withdrawn as early as possible, thus preventing further
file system damage.
Also note that this reintroduces superblock variable sd_log_error,
which Christoph removed with commit f729b66fca.
Signed-off-by: Bob Peterson <rpeterso@redhat.com>
When using cman-3.0.12.1 and gfs2-utils-3.0.12.1, mounting and
unmounting GFS2 file system would cause kernel to hang. The slab
allocator suggests that it is likely a double free memory corruption.
The issue is traced back to v3.9-rc6 where a patch is submitted to
use kzalloc() for storing a bitmap instead of using a local variable.
The intention is to allocate memory during mount and to free memory
during unmount. The original patch misses a code path which has
already freed the memory and caused memory corruption. This patch sets
the memory pointer to NULL after the memory is freed, so that double
free memory corruption will not happen.
gdlm_mount()
'-- set_recover_size() which use kzalloc()
'-- if dlm does not support ops callbacks then
'--- free_recover_size() which use kfree()
gldm_unmount()
'-- free_recover_size() which use kfree()
Previous patch which introduced the double free issue is
commit 57c7310b8e ("GFS2: use kmalloc for lvb bitmap")
Signed-off-by: Thomas Tai <thomas.tai@oracle.com>
Signed-off-by: Bob Peterson <rpeterso@redhat.com>
Reviewed-by: Liam R. Howlett <Liam.Howlett@Oracle.com>
On systems with low memory, it is possible for gfs2 to infinitely
loop in balance_dirty_pages() under heavy IO (creating sparse files).
balance_dirty_pages() attempts to write out the dirty pages via
gfs2_writepages() but none are found because these dirty pages are
being used by the journaling code in the ail. Normally, the journal
has an upper threshold which when hit triggers an automatic flush
of the ail. But this threshold can be higher than the number of
allowable dirty pages and result in the ail never being flushed.
This patch forces an ail flush when gfs2_writepages() fails to write
anything. This is a good indication that the ail might be holding
some dirty pages.
Signed-off-by: Abhi Das <adas@redhat.com>
Signed-off-by: Bob Peterson <rpeterso@redhat.com>
The prepare_to_wait_on_glock and finish_wait_on_glock functions introduced in
commit 56a365be "gfs2: gfs2_glock_get: Wait on freeing glocks" are
better removed, resulting in cleaner code.
Signed-off-by: Andreas Gruenbacher <agruenba@redhat.com>
Signed-off-by: Bob Peterson <rpeterso@redhat.com>
When under memory pressure and an inode's link count has dropped to
zero, defer deleting the inode to the delete workqueue. This avoids
calling into DLM under memory pressure, which can deadlock.
Signed-off-by: Andreas Gruenbacher <agruenba@redhat.com>
Signed-off-by: Bob Peterson <rpeterso@redhat.com>
gfs2_evict_inode is called to free inodes under memory pressure. The
function calls into DLM when an inode's last cluster-wide reference goes
away (remote unlink) and to release the glock and associated DLM lock
before finally destroying the inode. However, if DLM is blocked on
memory to become available, calling into DLM again will deadlock.
Avoid that by decoupling releasing glocks from destroying inodes in that
case: with gfs2_glock_queue_put, glocks will be dequeued asynchronously
in work queue context, when the associated inodes have likely already
been destroyed.
With this change, inodes can end up being unlinked, remote-unlink can be
triggered, and then the inode can be reallocated before all
remote-unlink callbacks are processed. To detect that, revalidate the
link count in gfs2_evict_inode to make sure we're not deleting an
allocated, referenced inode.
Signed-off-by: Andreas Gruenbacher <agruenba@redhat.com>
Signed-off-by: Bob Peterson <rpeterso@redhat.com>
Remove gfs2_set_nlink which prevents the link count of an inode from
becoming non-zero once it has reached zero. The next commit reduces the
amount of waiting on glocks when an inode is evicted from memory. With
that, an inode can become reallocated before all the remote-unlink
callbacks from a previous delete are processed, which causes the link
count to change from zero to non-zero.
Signed-off-by: Andreas Gruenbacher <agruenba@redhat.com>
Signed-off-by: Bob Peterson <rpeterso@redhat.com>
Keep glocks in their hash table until they are freed instead of removing
them when their last reference is dropped. This allows to wait for any
previous instances of a glock to go away in gfs2_glock_get before
creating a new glocks.
Special thanks to Andy Price for finding and fixing a problem which also
required us to delete the rcu_read_unlock from the error case in function
gfs2_glock_get.
Signed-off-by: Andreas Gruenbacher <agruenba@redhat.com>
Signed-off-by: Bob Peterson <rpeterso@redhat.com>
This patch moves the call to gfs2_delete_debugfs_file so that it
comes after the glock hash table has been cleared. This way we
can query the debugfs files if umount hangs.
Signed-off-by: Bob Peterson <rpeterso@redhat.com>
Before this patch, glock_dq would call gfs2_glock_remove_from_lru.
For glocks that are never put on the LRU, such as the transaction
glock, this just takes the spin_lock, determines there's nothing to
be done because the list is empty, then unlocks again. This was
causing unnecessary lock contention on the lru_lock spin_lock.
This patch adds a check for GLOF_LRU in the glops before taking
the spin_lock.
Signed-off-by: Bob Peterson <rpeterso@redhat.com>
This patch removes a call to gfs2_glock_add_to_lru from function
gfs2_clear_rgrpd. The call is just a waste of time because as soon
as it adds it to the lru_list, the call to gfs2_glock_put takes it
back off again.
Signed-off-by: Bob Peterson <rpeterso@redhat.com>
This patch adds some calls to clear gl_object in function
gfs2_delete_inode. Since we are deleting the inode, and the glock
typically outlives the inode in core, we must clear gl_object
so subsequent use of the glock (e.g. for a new inode in its place)
will not have the old pointer sitting there. In error cases we
need to tidy up after ourselves. In non-error cases, we need to
clear gl_object before we set the block free in the bitmap so
residules aren't left for potential inode creators.
Signed-off-by: Bob Peterson <rpeterso@redhat.com>
Reviewed-by: Andreas Gruenbacher <agruenba@redhat.com>
If function gfs2_create_inode fails after the inode has been
created (for example, if the inode_refresh fails for some reason)
the function was setting gl_object but never clearing it again.
The glocks are left pointing to a freed inode. This patch adds
the calls to clear gl_object in the appropriate error paths.
Signed-off-by: Bob Peterson <rpeterso@redhat.com>
Reviewed-by: Andreas Gruenbacher <agruenba@redhat.com>
Before this patch, the inode glock's gl_object was set after a
reference was acquired, but before the block type was verified.
In cases where the block was unlinked, then freed and reused on
another node, a residule delete callback (delete_work) would try
to look up the inode, eventually failing the block check, but
only after it overwrites gl_object with a pointer to the wrong
inode. This patch moves the assignment of gl_object after the
block check so it won't be improperly overwritten.
Likewise, at the end of the function, gfs2_inode_lookup was
clearing gl_object after it unlocked the glock, which meant
another process might free the glock in the meantime. This
patch guards against that case.
Signed-off-by: Bob Peterson <rpeterso@redhat.com>
Reviewed-by: Andreas Gruenbacher <agruenba@redhat.com>
This patch introduces a new helper function in glock.h that
clears gl_object, with an added integrity check. An additional
integrity check has been added to glock_set_object, plus comments.
This is step 1 in a series to ensure gl_object integrity.
Signed-off-by: Bob Peterson <rpeterso@redhat.com>
Reviewed-by: Andreas Gruenbacher <agruenba@redhat.com>
When gfs2 does metadata I/O, only REQ_META is used as a metadata hint of
the bio. But flag REQ_META is just a hint for block trace, not for block
layer code to handle a bio as metadata request.
For some of metadata I/Os of gfs2, A REQ_PRIO flag on the metadata bio
would be very informative to block layer code. For example, if bcache is
used as a I/O cache for gfs2, it will be possible for bcache code to get
the hint and cache the pre-fetched metadata blocks on cache device. This
behavior may be helpful to improve metadata I/O performance if the
following requests hit the cache.
Here are the locations in gfs2 code where a REQ_PRIO flag should be added,
- All places where REQ_READAHEAD is used, gfs2 code uses this flag for
metadata read ahead.
- In gfs2_meta_rq() where the first metadata block is read in.
- In gfs2_write_buf_to_page(), read in quota metadata blocks to have them
up to date.
These metadata blocks are probably to be accessed again in future, adding
a REQ_PRIO flag may have bcache to keep such metadata in fast cache
device. For system without a cache layer, REQ_PRIO can still provide hint
to block layer to handle metadata requests more properly.
Signed-off-by: Coly Li <colyli@suse.de>
Signed-off-by: Bob Peterson <rpeterso@redhat.com>
In inode_go_lock() function, the parameter order of list_add() is error.
According to the define of list_add(), the first parameter is new entry
and the second is the list head, so ip->i_trunc_list should be the
first parameter and the sdp->sd_trunc_list should be second.
Signed-off-by: Wang Xibo<wang.xibo@zte.com.cn>
Signed-off-by: Xiao Likun<xiao.likun@zte.com.cn>
Signed-off-by: Bob Peterson <rpeterso@redhat.com>
When commit 4fd1a57952 moved the call to flush_delayed_work from
gfs2_evict_inode to gfs2_inode_lookup to avoid calling into DLM during
evict, a similar call should have been added to gfs2_create_inode:
that's another code path in which glocks of previous inodes may be
reused.
The flush of the iopen glock work queue added by 4fd1a57952, on the
other hand, is unnecessary and can be removed.
Signed-off-by: Andreas Gruenbacher <agruenba@redhat.com>
Signed-off-by: Bob Peterson <rpeterso@redhat.com>
When new directory 'DIR1' is created in a directory 'DIR0' with SGID bit
set, DIR1 is expected to have SGID bit set (and owning group equal to
the owning group of 'DIR0'). However when 'DIR0' also has some default
ACLs that 'DIR1' inherits, setting these ACLs will result in SGID bit on
'DIR1' to get cleared if user is not member of the owning group.
Fix the problem by moving posix_acl_update_mode() out of
__gfs2_set_acl() into gfs2_set_acl(). That way the function will not be
called when inheriting ACLs which is what we want as it prevents SGID
bit clearing and the mode has been properly set by posix_acl_create()
anyway.
Fixes: 073931017b
Signed-off-by: Jan Kara <jack@suse.cz>
Signed-off-by: Bob Peterson <rpeterso@redhat.com>
Function gfs2_holder_initialized should be used in do_flock as well.
Signed-off-by: Andreas Gruenbacher <agruenba@redhat.com>
Signed-off-by: Bob Peterson <rpeterso@redhat.com>
Before this patch, problems reading in indirect buffers would send
an IO error back to the caller, and release the buffer_head with
brelse() in function gfs2_meta_indirect_buffer, however, it would
still return the address of the buffer_head it released. After the
error was discovered, function gfs2_block_map would call function
release_metapath to free all buffers. That checked:
if (mp->mp_bh[i] == NULL) but since the value was set after the
error, it was non-zero, so brelse was called a second time. This
resulted in the following error:
kernel: WARNING: at fs/buffer.c:1224 __brelse+0x3a/0x40() (Tainted: G W -- ------------ )
kernel: Hardware name: RHEV Hypervisor
kernel: VFS: brelse: Trying to free free buffer
This patch changes gfs2_meta_indirect_buffer so it only sets
the buffer_head pointer in cases where it isn't released.
Signed-off-by: Bob Peterson <rpeterso@redhat.com>
Acked-by: Steven Whitehouse <swhiteho@redhat.com>
-----BEGIN PGP SIGNATURE-----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=3Z93
-----END PGP SIGNATURE-----
Merge tag 'for-linus-v4.13-2' of git://git.kernel.org/pub/scm/linux/kernel/git/jlayton/linux
Pull Writeback error handling updates from Jeff Layton:
"This pile represents the bulk of the writeback error handling fixes
that I have for this cycle. Some of the earlier patches in this pile
may look trivial but they are prerequisites for later patches in the
series.
The aim of this set is to improve how we track and report writeback
errors to userland. Most applications that care about data integrity
will periodically call fsync/fdatasync/msync to ensure that their
writes have made it to the backing store.
For a very long time, we have tracked writeback errors using two flags
in the address_space: AS_EIO and AS_ENOSPC. Those flags are set when a
writeback error occurs (via mapping_set_error) and are cleared as a
side-effect of filemap_check_errors (as you noted yesterday). This
model really sucks for userland.
Only the first task to call fsync (or msync or fdatasync) will see the
error. Any subsequent task calling fsync on a file will get back 0
(unless another writeback error occurs in the interim). If I have
several tasks writing to a file and calling fsync to ensure that their
writes got stored, then I need to have them coordinate with one
another. That's difficult enough, but in a world of containerized
setups that coordination may even not be possible.
But wait...it gets worse!
The calls to filemap_check_errors can be buried pretty far down in the
call stack, and there are internal callers of filemap_write_and_wait
and the like that also end up clearing those errors. Many of those
callers ignore the error return from that function or return it to
userland at nonsensical times (e.g. truncate() or stat()). If I get
back -EIO on a truncate, there is no reason to think that it was
because some previous writeback failed, and a subsequent fsync() will
(incorrectly) return 0.
This pile aims to do three things:
1) ensure that when a writeback error occurs that that error will be
reported to userland on a subsequent fsync/fdatasync/msync call,
regardless of what internal callers are doing
2) report writeback errors on all file descriptions that were open at
the time that the error occurred. This is a user-visible change,
but I think most applications are written to assume this behavior
anyway. Those that aren't are unlikely to be hurt by it.
3) document what filesystems should do when there is a writeback
error. Today, there is very little consistency between them, and a
lot of cargo-cult copying. We need to make it very clear what
filesystems should do in this situation.
To achieve this, the set adds a new data type (errseq_t) and then
builds new writeback error tracking infrastructure around that. Once
all of that is in place, we change the filesystems to use the new
infrastructure for reporting wb errors to userland.
Note that this is just the initial foray into cleaning up this mess.
There is a lot of work remaining here:
1) convert the rest of the filesystems in a similar fashion. Once the
initial set is in, then I think most other fs' will be fairly
simple to convert. Hopefully most of those can in via individual
filesystem trees.
2) convert internal waiters on writeback to use errseq_t for
detecting errors instead of relying on the AS_* flags. I have some
draft patches for this for ext4, but they are not quite ready for
prime time yet.
This was a discussion topic this year at LSF/MM too. If you're
interested in the gory details, LWN has some good articles about this:
https://lwn.net/Articles/718734/https://lwn.net/Articles/724307/"
* tag 'for-linus-v4.13-2' of git://git.kernel.org/pub/scm/linux/kernel/git/jlayton/linux:
btrfs: minimal conversion to errseq_t writeback error reporting on fsync
xfs: minimal conversion to errseq_t writeback error reporting
ext4: use errseq_t based error handling for reporting data writeback errors
fs: convert __generic_file_fsync to use errseq_t based reporting
block: convert to errseq_t based writeback error tracking
dax: set errors in mapping when writeback fails
Documentation: flesh out the section in vfs.txt on storing and reporting writeback errors
mm: set both AS_EIO/AS_ENOSPC and errseq_t in mapping_set_error
fs: new infrastructure for writeback error handling and reporting
lib: add errseq_t type and infrastructure for handling it
mm: don't TestClearPageError in __filemap_fdatawait_range
mm: clear AS_EIO/AS_ENOSPC when writeback initiation fails
jbd2: don't clear and reset errors after waiting on writeback
buffer: set errors in mapping at the time that the error occurs
fs: check for writeback errors after syncing out buffers in generic_file_fsync
buffer: use mapping_set_error instead of setting the flag
mm: fix mapping_set_error call in me_pagecache_dirty
Before commit 88ffbf3e03 "GFS2: Use resizable hash table for glocks",
glocks were freed via call_rcu to allow reading the glock hashtable
locklessly using rcu. This was then changed to free glocks immediately,
which made reading the glock hashtable unsafe. Bring back the original
code for freeing glocks via call_rcu.
Signed-off-by: Andreas Gruenbacher <agruenba@redhat.com>
Signed-off-by: Bob Peterson <rpeterso@redhat.com>
Cc: stable@vger.kernel.org # 4.3+
I noticed on xfs that I could still sometimes get back an error on fsync
on a fd that was opened after the error condition had been cleared.
The problem is that the buffer code sets the write_io_error flag and
then later checks that flag to set the error in the mapping. That flag
perisists for quite a while however. If the file is later opened with
O_TRUNC, the buffers will then be invalidated and the mapping's error
set such that a subsequent fsync will return error. I think this is
incorrect, as there was no writeback between the open and fsync.
Add a new mark_buffer_write_io_error operation that sets the flag and
the error in the mapping at the same time. Replace all calls to
set_buffer_write_io_error with mark_buffer_write_io_error, and remove
the places that check this flag in order to set the error in the
mapping.
This sets the error in the mapping earlier, at the time that it's first
detected.
Signed-off-by: Jeff Layton <jlayton@redhat.com>
Reviewed-by: Jan Kara <jack@suse.cz>
Reviewed-by: Carlos Maiolino <cmaiolino@redhat.com>
1. Andreas Gruenbacher has four patches related to cleaning up the GFS2
inode evict process. This is about half of his patches designed to
fix a long-standing GFS2 hang related to the inode shrinker.
(Shrinker calls gfs2 evict, evict calls DLM, DLM requires memory
and blocks on the shrinker.) These 4 patches have been well tested.
His second set of patches are still being tested, so I plan to hold
them until the next merge window, after we have more weeks of testing.
The first patch eliminates the flush_delayed_work, which can block.
2. Andreas's second patch protects setting of gl_object for rgrps with
a spin_lock to prevent proven races.
3. His third patch introduces a centralized mechanism for queueing glock
work with better reference counting, to prevent more races.
4. His fourth patch retains a reference to inode glocks when an error
occurs while creating an inode. This keeps the subsequent evict from
needing to reacquire the glock, which might call into DLM and block
in low memory conditions.
5. Arvind Yadav has a patch to add const to attribute_group structures.
6. I have a patch to detect directory entry inconsistencies and withdraw
the file system if any are found. Better that than silent corruption.
7. I have a patch to remove a vestigial variable from glock structures,
saving some slab space.
8. I have another patch to remove a vestigial variable from the GFS2
in-core superblock structure.
-----BEGIN PGP SIGNATURE-----
iQEcBAABAgAGBQJZXOIfAAoJENeLYdPf93o7RVcH/jLEK3hmZOd94pDTYg3Damuo
KI3xjyutDgQT83uwg8p5UBPwRYCDnyiOLwOWGBJJvjPEI1S4syrXq/FzOmxmX6cV
nE28ARL/OXCoFEXBMUVHvHL3nK+zEUr8rO6Xz51B1ifVq7GV8iVK+ZgxzRhx0PWP
f+0SVHiQtU0HKyxR5y9p43oygtHZaGbjy4WL0YbmFZM59y5q9A8rBHFACn2JyPBm
/zXN6gF/Orao+BDXLT6OM3vNXZcOQ7FUPWwctguHsAO/bLzWiISyfJxLWJsHvSdW
tzFTN1DByjXvqAhs4HTSuh9JfBDAyxcXkmczXJyATBkCTEJv42Iev+ILmre+wwQ=
=YTwn
-----END PGP SIGNATURE-----
Merge tag 'gfs2-4.13.fixes' of git://git.kernel.org/pub/scm/linux/kernel/git/gfs2/linux-gfs2
Pull GFS2 updates from Bob Peterson:
"We've got eight GFS2 patches for this merge window:
- Andreas Gruenbacher has four patches related to cleaning up the
GFS2 inode evict process. This is about half of his patches
designed to fix a long-standing GFS2 hang related to the inode
shrinker: Shrinker calls gfs2 evict, evict calls DLM, DLM requires
memory and blocks on the shrinker.
These four patches have been well tested. His second set of patches
are still being tested, so I plan to hold them until the next merge
window, after we have more weeks of testing. The first patch
eliminates the flush_delayed_work, which can block.
- Andreas's second patch protects setting of gl_object for rgrps with
a spin_lock to prevent proven races.
- His third patch introduces a centralized mechanism for queueing
glock work with better reference counting, to prevent more races.
-His fourth patch retains a reference to inode glocks when an error
occurs while creating an inode. This keeps the subsequent evict
from needing to reacquire the glock, which might call into DLM and
block in low memory conditions.
- Arvind Yadav has a patch to add const to attribute_group
structures.
- I have a patch to detect directory entry inconsistencies and
withdraw the file system if any are found. Better that than silent
corruption.
- I have a patch to remove a vestigial variable from glock
structures, saving some slab space.
- I have another patch to remove a vestigial variable from the GFS2
in-core superblock structure"
* tag 'gfs2-4.13.fixes' of git://git.kernel.org/pub/scm/linux/kernel/git/gfs2/linux-gfs2:
GFS2: constify attribute_group structures.
gfs2: gfs2_create_inode: Keep glock across iput
gfs2: Clean up glock work enqueuing
gfs2: Protect gl->gl_object by spin lock
gfs2: Get rid of flush_delayed_work in gfs2_evict_inode
GFS2: Eliminate vestigial sd_log_flush_wrapped
GFS2: Remove gl_list from glock structure
GFS2: Withdraw when directory entry inconsistencies are detected
attribute_groups are not supposed to change at runtime. All functions
working with attribute_groups provided by <linux/sysfs.h> work with const
attribute_group. So mark the non-const structs as const.
File size before:
text data bss dec hex filename
5259 1344 8 6611 19d3 fs/gfs2/sys.o
File size After adding 'const':
text data bss dec hex filename
5371 1216 8 6595 19c3 fs/gfs2/sys.o
Signed-off-by: Arvind Yadav <arvind.yadav.cs@gmail.com>
Signed-off-by: Bob Peterson <rpeterso@redhat.com>
On failure, keep the inode glock across the final iput of the new inode
so that gfs2_evict_inode doesn't have to re-acquire the glock. That
way, gfs2_evict_inode won't need to revalidate the block type.
Signed-off-by: Andreas Gruenbacher <agruenba@redhat.com>
Signed-off-by: Bob Peterson <rpeterso@redhat.com>
This patch adds a standardized queueing mechanism for glock work
with spin_lock protection to prevent races.
Signed-off-by: Andreas Gruenbacher <agruenba@redhat.com>
Signed-off-by: Bob Peterson <rpeterso@redhat.com>
Put all remaining accesses to gl->gl_object under the
gl->gl_lockref.lock spinlock to prevent races.
Signed-off-by: Andreas Gruenbacher <agruenba@redhat.com>
Signed-off-by: Bob Peterson <rpeterso@redhat.com>
So far, gfs2_evict_inode clears gl->gl_object and then flushes the glock
work queue to make sure that inode glops which dereference gl->gl_object
have finished running before the inode is destroyed. However, flushing
the work queue may do more work than needed, and in particular, it may
call into DLM, which we want to avoid here. Use a bit lock
(GIF_GLOP_PENDING) to synchronize between the inode glops and
gfs2_evict_inode instead to get rid of the flushing.
In addition, flush the work queues of existing glocks before reusing
them for new inodes to get those glocks into a known state: the glock
state engine currently doesn't handle glock re-appropriation correctly.
(We may be able to fix the glock state engine instead later.)
Based on a patch by Steven Whitehouse <swhiteho@redhat.com>.
Signed-off-by: Andreas Gruenbacher <agruenba@redhat.com>
Signed-off-by: Bob Peterson <rpeterso@redhat.com>
This patch prints an inode consistency error and withdraws the file
system when directory entry counts are mismatched.
Signed-off-by: Bob Peterson <rpeterso@redhat.com>
-----BEGIN PGP SIGNATURE-----
iQEcBAABAgAGBQJZPdbLAAoJEHm+PkMAQRiGx4wH/1nCjfnl6fE8oJ24/1gEAOUh
biFdqJkYZmlLYHVtYfLm4Ueg4adJdg0wx6qM/4RaAzmQVvLfDV34bc1qBf1+P95G
kVF+osWyXrZo5cTwkwapHW/KNu4VJwAx2D1wrlxKDVG5AOrULH1pYOYGOpApEkZU
4N+q5+M0ce0GJpqtUZX+UnI33ygjdDbBxXoFKsr24B7eA0ouGbAJ7dC88WcaETL+
2/7tT01SvDMo0jBSV0WIqlgXwZ5gp3yPGnklC3F4159Yze6VFrzHMKS/UpPF8o8E
W9EbuzwxsKyXUifX2GY348L1f+47glen/1sedbuKnFhP6E9aqUQQJXvEO7ueQl4=
=m2Gx
-----END PGP SIGNATURE-----
Merge tag 'v4.12-rc5' into for-4.13/block
We've already got a few conflicts and upcoming work depends on some of the
changes that have gone into mainline as regression fixes for this series.
Pull in 4.12-rc5 to resolve these conflicts and make it easier on down stream
trees to continue working on 4.13 changes.
Signed-off-by: Jens Axboe <axboe@kernel.dk>
Replace bi_error with a new bi_status to allow for a clear conversion.
Note that device mapper overloaded bi_error with a private value, which
we'll have to keep arround at least for now and thus propagate to a
proper blk_status_t value.
Signed-off-by: Christoph Hellwig <hch@lst.de>
Signed-off-by: Jens Axboe <axboe@fb.com>
For some file systems we still memcpy into it, but in various places this
already allows us to use the proper uuid helpers. More to come..
Signed-off-by: Christoph Hellwig <hch@lst.de>
Reviewed-by: Amir Goldstein <amir73il@gmail.com>
Acked-by: Mimi Zohar <zohar@linux.vnet.ibm.com> (Changes to IMA/EVM)
Reviewed-by: Andy Shevchenko <andriy.shevchenko@linux.intel.com>
Commit b685d3d65a "block: treat REQ_FUA and REQ_PREFLUSH as
synchronous" removed REQ_SYNC flag from WRITE_{FUA|PREFLUSH|...}
definitions. generic_make_request_checks() however strips REQ_FUA and
REQ_PREFLUSH flags from a bio when the storage doesn't report volatile
write cache and thus write effectively becomes asynchronous which can
lead to performance regressions
Fix the problem by making sure all bios which are synchronous are
properly marked with REQ_SYNC.
Fixes: b685d3d65a
CC: Steven Whitehouse <swhiteho@redhat.com>
CC: cluster-devel@redhat.com
CC: stable@vger.kernel.org
Acked-by: Bob Peterson <rpeterso@redhat.com>
Signed-off-by: Jan Kara <jack@suse.cz>
1. Andreas Gruenbacher wrote a patch to replace the deprecated
call to rhashtable_walk_init with rhashtable_walk_enter.
2. Andreas also wrote a patch to eliminate redundant code in
two of our debugfs sequence files.
3. Andreas also cleaned up the rhashtable key ugliness Linus
pointed out during this cycle, following Linus's suggestions.
4. Andreas also wrote a patch to take advantage of his new
function rhashtable_lookup_get_insert_fast. This makes glock
lookup faster and more bullet-proof.
5. Andreas also wrote a patch to revert a patch in the evict
path that caused occasional deadlocks, and is no longer
needed.
6. Andrew Price wrote a patch to re-enable fallocate for the
rindex system file to enable gfs2_grow to grow properly on
secondary file system grow operations.
7. I wrote a patch to initialize an inode number field to make
certain kernel trace points more understandable.
8. I also wrote a patch that makes GFS2 file system "withdraw"
work more like it should by ignoring operations after a
withdraw that would formerly cause a BUG() and kernel panic.
9. I also reworked the entire truncate/delete algorithm,
scrapping the old recursive algorithm in favor of a new
non-recursive algorithm. This was done for performance:
This way, GFS2 no longer needs to lock multiple resource
groups while doing truncates and deletes of files that cross
multiple resource group boundaries, allowing for better
parallelism. It also solves a problem whereby deleting large
files would request a large chunk of kernel memory, which
resulted in a get_page_from_freelist warning.
10. Due to a regression found during testing, I added a new
patch to correct "GFS2: Prevent BUG from occurring when
normal Withdraws occur".
-----BEGIN PGP SIGNATURE-----
iQEcBAABAgAGBQJZDNnaAAoJENeLYdPf93o7B7kIAJzwz7vVDVg2TpWVhMmXIWhf
rZx3Gth5F0h+ZHddW7HzTLg+64XQ5//GyDD3UDtCpkhl5SJH+nt3juHyPJlRwioT
0ua4SjyKLQSoJJVAEgAwu42QjORTXab7NjYn5LEhvRc0Gg/El9WGU+ZgmP2/aAvf
KE2u/IEYNDkoJNS3Oqc7shajAyLYda6wCAASs/1ZGt9u48m/o/I23Zd7wr7EOkzw
rd3gB0x80cJqDAB5IcymGOm111Tg4g34LwsRuyMnWE3H1jOgV+J515FVHEIvZuPq
Wl9X7V8CzktI7nyLKVnZhpuv5JzyMq/vOPiD01tTFx8Oy1JCRezjmATXFjW/zIo=
=MX3c
-----END PGP SIGNATURE-----
Merge tag 'gfs2-4.12.fixes' of git://git.kernel.org/pub/scm/linux/kernel/git/gfs2/linux-gfs2
Pull GFS2 updates from Bob Peterson:
"We've got ten GFS2 patches for this merge window.
- Andreas Gruenbacher wrote a patch to replace the deprecated call to
rhashtable_walk_init with rhashtable_walk_enter.
- Andreas also wrote a patch to eliminate redundant code in two of
our debugfs sequence files.
- Andreas also cleaned up the rhashtable key ugliness Linus pointed
out during this cycle, following Linus's suggestions.
- Andreas also wrote a patch to take advantage of his new function
rhashtable_lookup_get_insert_fast. This makes glock lookup faster
and more bullet-proof.
- Andreas also wrote a patch to revert a patch in the evict path that
caused occasional deadlocks, and is no longer needed.
- Andrew Price wrote a patch to re-enable fallocate for the rindex
system file to enable gfs2_grow to grow properly on secondary file
system grow operations.
- I wrote a patch to initialize an inode number field to make certain
kernel trace points more understandable.
- I also wrote a patch that makes GFS2 file system "withdraw" work
more like it should by ignoring operations after a withdraw that
would formerly cause a BUG() and kernel panic.
- I also reworked the entire truncate/delete algorithm, scrapping the
old recursive algorithm in favor of a new non-recursive algorithm.
This was done for performance: This way, GFS2 no longer needs to
lock multiple resource groups while doing truncates and deletes of
files that cross multiple resource group boundaries, allowing for
better parallelism. It also solves a problem whereby deleting large
files would request a large chunk of kernel memory, which resulted
in a get_page_from_freelist warning.
- Due to a regression found during testing, I added a new patch to
correct 'GFS2: Prevent BUG from occurring when normal Withdraws
occur'."
* tag 'gfs2-4.12.fixes' of git://git.kernel.org/pub/scm/linux/kernel/git/gfs2/linux-gfs2:
GFS2: Allow glocks to be unlocked after withdraw
GFS2: Non-recursive delete
gfs2: Re-enable fallocate for the rindex
Revert "GFS2: Wait for iopen glock dequeues"
gfs2: Switch to rhashtable_lookup_get_insert_fast
GFS2: Temporarily zero i_no_addr when creating a dinode
gfs2: Don't pack struct lm_lockname
gfs2: Deduplicate gfs2_{glocks,glstats}_open
gfs2: Replace rhashtable_walk_init with rhashtable_walk_enter
GFS2: Prevent BUG from occurring when normal Withdraws occur
This bug fixes a regression introduced by patch 0d1c7ae9d8.
The intent of the patch was to stop promoting glocks after a
file system is withdrawn due to a variety of errors, because doing
so results in a BUG(). (You should be able to unmount after a
withdraw rather than having the kernel panic.)
Unfortunately, it also stopped demotions, so glocks could not be
unlocked after withdraw, which means the unmount would hang.
This patch allows function do_xmote to demote locks to an
unlocked state after a withdraw, but not promote them.
Signed-off-by: Bob Peterson <rpeterso@redhat.com>
Now that all bdi structures filesystems use are properly refcounted, we
can remove the SB_I_DYNBDI flag.
Reviewed-by: Christoph Hellwig <hch@lst.de>
Signed-off-by: Jan Kara <jack@suse.cz>
Signed-off-by: Jens Axboe <axboe@fb.com>
Similarly to set_bdev_super() GFS2 just used block device reference to
bdi. Convert it to properly getting bdi reference. The reference will
get automatically dropped on superblock destruction.
CC: Steven Whitehouse <swhiteho@redhat.com>
CC: Bob Peterson <rpeterso@redhat.com>
CC: cluster-devel@redhat.com
Reviewed-by: Christoph Hellwig <hch@lst.de>
Signed-off-by: Jan Kara <jack@suse.cz>
Signed-off-by: Jens Axboe <axboe@fb.com>
Implement truncate/delete as a non-recursive algorithm. The older
algorithm was implemented with recursion to strip off each layer
at a time (going by height, starting with the maximum height.
This version tries to do the same thing but without recursion,
and without needing to allocate new structures or lists in memory.
For example, say you want to truncate a very large file to 1 byte,
and its end-of-file metapath is: 0.505.463.428. The starting
metapath would be 0.0.0.0. Since it's a truncate to non-zero, it
needs to preserve that byte, and all metadata pointing to it.
So it would start at 0.0.0.0, look up all its metadata buffers,
then free all data blocks pointed to at the highest level.
After that buffer is "swept", it moves on to 0.0.0.1, then
0.0.0.2, etc., reading in buffers and sweeping them clean.
When it gets to the end of the 0.0.0 metadata buffer (for 4K
blocks the last valid one is 0.0.0.508), it backs up to the
previous height and starts working on 0.0.1.0, then 0.0.1.1,
and so forth. After it reaches the end and sweeps 0.0.1.508,
it continues with 0.0.2.0, and so on. When that height is
exhausted, and it reaches 0.0.508.508 it backs up another level,
to 0.1.0.0, then 0.1.0.1, through 0.1.0.508. So it has to keep
marching backwards and forwards through the metadata until it's
all swept clean. Once it has all the data blocks freed, it
lowers the strip height, and begins the process all over again,
but with one less height. This time it sweeps 0.0.0 through
0.505.463. When that's clean, it lowers the strip height again
and works to free 0.505. Eventually it strips the lowest height, 0.
For a delete or truncate to 0, all metadata for all heights of
0.0.0.0 would be freed. For a truncate to 1 byte, 0.0.0.0 would
be preserved.
This isn't much different from normal integer incrementing,
where an integer gets incremented from 0000 (0.0.0.0) to 3021
(3.0.2.1). So 0000 gets increments to 0001, 0002, up to 0009,
then on to 0010, 0011 up to 0099, then 0100 and so forth. It's
just that each "digit" goes from 0 to 508 (for a total of 509
pointers) rather than from 0 to 9.
Note that the dinode will only have 483 pointers due to the
dinode structure itself.
Also note: this is just an example. These numbers (509 and 483)
are based on a standard 4K block size. Smaller block sizes will
yield smaller numbers of indirect pointers accordingly.
The truncation process is accomplished with the help of two
major functions and a few helper functions.
Functions do_strip and recursive_scan are obsolete, so removed.
New function sweep_bh_for_rgrps cleans a buffer_head pointed to
by the given metapath and height. By cleaning, I mean it frees
all blocks starting at the offset passed in metapath. It starts
at the first block in the buffer pointed to by the metapath and
identifies its resource group (rgrp). From there it frees all
subsequent block pointers that lie within that rgrp. If it's
already inside a transaction, it stays within it as long as it
can. In other words, it doesn't close a transaction until it knows
it's freed what it can from the resource group. In this way,
multiple buffers may be cleaned in a single transaction, as long
as those blocks in the buffer all lie within the same rgrp.
If it's not in a transaction, it starts one. If the buffer_head
has references to blocks within multiple rgrps, it frees all the
blocks inside the first rgrp it finds, then closes the
transaction. Then it repeats the cycle: identifies the next
unfreed block, uses it to find its rgrp, then starts a new
transaction for that set. It repeats this process repeatedly
until the buffer_head contains no more references to any blocks
past the given metapath.
Function trunc_dealloc has been reworked into a finite state
automaton. It has basically 3 active states:
DEALLOC_MP_FULL, DEALLOC_MP_LOWER, and DEALLOC_FILL_MP:
The DEALLOC_MP_FULL state implies the metapath has a full set
of buffers out to the "shrink height", and therefore, it can
call function sweep_bh_for_rgrps to free the blocks within the
highest height of the metapath. If it's just swept the lowest
level (or an error has occurred) the state machine is ended.
Otherwise it proceeds to the DEALLOC_MP_LOWER state.
The DEALLOC_MP_LOWER state implies we are finished with a given
buffer_head, which may now be released, and therefore we are
then missing some buffer information from the metapath. So we
need to find more buffers to read in. In most cases, this is
just a matter of releasing the buffer_head and moving to the
next pointer from the previous height, so it may be read in and
swept as well. If it can't find another non-null pointer to
process, it checks whether it's reached the end of a height
and needs to lower the strip height, or whether it still needs
move forward through the previous height's metadata. In this
state, all zero-pointers are skipped. From this state, it can
only loop around (once more backing up another height) or,
once a valid metapath is found (one that has non-zero
pointers), proceed to state DEALLOC_FILL_MP.
The DEALLOC_FILL_MP state implies that we have a metapath
but not all its buffers are read in. So we must proceed to read
in buffer_heads until the metapath has a valid buffer for every
height. If the previous state backed us up 3 heights, we may
need to read in a buffer, increment the height, then repeat the
process until buffers have been read in for all required heights.
If it's successful reading a buffer, and it's at the highest
height we need, it proceeds back to the DEALLOC_MP_FULL state.
If it's unable to fill in a buffer, (encounters a hole, etc.)
it tries to find another non-zero block pointer. If they're all
zero, it lowers the height and returns to the DEALLOC_MP_LOWER
state. If it finds a good non-null pointer, it loops around and
reads it in, while keeping the metapath in lock-step with the
pointers it examines.
The state machine runs until the truncation request is
satisfied. Then any transactions are ended, the quota and
statfs data are updated, and the function is complete.
Helper function metaptr1 was introduced to be an easy way to
determine the start of a buffer_head's indirect pointers.
Helper function lookup_mp_height was introduced to find a
metapath index and read in the buffer that corresponds to it.
In this way, function lookup_metapath becomes a simple loop to
call it for every height.
Helper function fillup_metapath is similar to lookup_metapath
except it can do partial lookups. If the state machine
backed up multiple levels (like 2999 wrapping to 3000) it
needs to find out the next starting point and start issuing
metadata reads at that point.
Helper function hptrs is a shortcut to determine how many
pointers should be expected in a buffer. Height 0 is the dinode
which has fewer pointers than the others.
Signed-off-by: Bob Peterson <rpeterso@redhat.com>
Commit 86066914ed "gfs2: Don't support
fallocate on jdata files" removed the ability of gfs2_grow to reserve
space at the end of the rindex, which could prevent a second gfs2_grow
from succeeding if the fs is full. Allow fallocate to work on the rindex
once again.
Signed-off-by: Andrew Price <anprice@redhat.com>
Signed-off-by: Bob Peterson <rpeterso@redhat.com>
Revert commit 86d067a797: it turns out
that waiting for iopen glock dequeues here isn't needed anymore because
the bugs that commit was meant to fix have been fixed otherwise.
In addition, we want to avoid waiting on glocks in gfs2_evict_inode in
shrinker context because the shrinker may be invoked on behalf of DLM,
in which case calling into DLM again would deadlock. This commit makes
the described scenario less likely without completely avoiding it; it's
still a step in the right direction, though.
Signed-off-by: Andreas Gruenbacher <agruenba@redhat.com>
Signed-off-by: Bob Peterson <rpeterso@redhat.com>
Switch from rhashtable_lookup_insert_fast + rhashtable_lookup_fast to
rhashtable_lookup_get_insert_fast, which is cleaner and avoids an extra
rhashtable lookup.
At the same time, turn the retry loop in gfs2_glock_get into an infinite
loop. The lookup or insert will eventually succeed, usually very fast,
but there is no reason to give up trying at a fixed number of
iterations.
Signed-off-by: Andreas Gruenbacher <agruenba@redhat.com>
Signed-off-by: Bob Peterson <rpeterso@redhat.com>
Before this patch i_no_addr was not initialized until after the
return from allocating its block. That meant the i_no_addr was
temporarily uninitialized storage. Ordinarily that's not a concern,
but if inplace_reserve can't find space, it can call try_rgrp_unlink
which references i_no_addr as a block to avoid. That can result in
unpredictable behavior. More importantly, the trace point in
gfs2_alloc_blocks references ip->i_no_addr before it is set, which
is misleading when reading the kernel traces. This patch makes it
look like the new dinode block was assigned in the name of inode 0
rather than a random inode that's completely unrelated.
Signed-off-by: Bob Peterson <rpeterso@redhat.com>
