Skip to content

K
Kernel

openeuler / Kernel
1 年多 前同步成功

通知 8
Star 0
Fork 0
K
Kernel
切换分支/标签
  1. 08 4月, 2021 10 次提交
  3. 26 3月, 2021 2 次提交
    • D
      xfs: initialise attr fork on inode create · e6a688c3
      由 Dave Chinner 提交于 
      When we allocate a new inode, we often need to add an attribute to
the inode as part of the create. This can happen as a result of
needing to add default ACLs or security labels before the inode is
made visible to userspace.

This is highly inefficient right now. We do the create transaction
to allocate the inode, then we do an "add attr fork" transaction to
modify the just created empty inode to set the inode fork offset to
allow attributes to be stored, then we go and do the attribute
creation.

This means 3 transactions instead of 1 to allocate an inode, and
this greatly increases the load on the CIL commit code, resulting in
excessive contention on the CIL spin locks and performance
degradation:

 18.99%  [kernel]                [k] __pv_queued_spin_lock_slowpath
  3.57%  [kernel]                [k] do_raw_spin_lock
  2.51%  [kernel]                [k] __raw_callee_save___pv_queued_spin_unlock
  2.48%  [kernel]                [k] memcpy
  2.34%  [kernel]                [k] xfs_log_commit_cil

The typical profile resulting from running fsmark on a selinux enabled
filesytem is adds this overhead to the create path:

  - 15.30% xfs_init_security
     - 15.23% security_inode_init_security
	- 13.05% xfs_initxattrs
	   - 12.94% xfs_attr_set
	      - 6.75% xfs_bmap_add_attrfork
		 - 5.51% xfs_trans_commit
		    - 5.48% __xfs_trans_commit
		       - 5.35% xfs_log_commit_cil
			  - 3.86% _raw_spin_lock
			     - do_raw_spin_lock
				  __pv_queued_spin_lock_slowpath
		 - 0.70% xfs_trans_alloc
		      0.52% xfs_trans_reserve
	      - 5.41% xfs_attr_set_args
		 - 5.39% xfs_attr_set_shortform.constprop.0
		    - 4.46% xfs_trans_commit
		       - 4.46% __xfs_trans_commit
			  - 4.33% xfs_log_commit_cil
			     - 2.74% _raw_spin_lock
				- do_raw_spin_lock
				     __pv_queued_spin_lock_slowpath
			       0.60% xfs_inode_item_format
		      0.90% xfs_attr_try_sf_addname
	- 1.99% selinux_inode_init_security
	   - 1.02% security_sid_to_context_force
	      - 1.00% security_sid_to_context_core
		 - 0.92% sidtab_entry_to_string
		    - 0.90% sidtab_sid2str_get
			 0.59% sidtab_sid2str_put.part.0
	   - 0.82% selinux_determine_inode_label
	      - 0.77% security_transition_sid
		   0.70% security_compute_sid.part.0

And fsmark creation rate performance drops by ~25%. The key point to
note here is that half the additional overhead comes from adding the
attribute fork to the newly created inode. That's crazy, considering
we can do this same thing at inode create time with a couple of
lines of code and no extra overhead.

So, if we know we are going to add an attribute immediately after
creating the inode, let's just initialise the attribute fork inside
the create transaction and chop that whole chunk of code out of
the create fast path. This completely removes the performance
drop caused by enabling SELinux, and the profile looks like:

     - 8.99% xfs_init_security
         - 9.00% security_inode_init_security
            - 6.43% xfs_initxattrs
               - 6.37% xfs_attr_set
                  - 5.45% xfs_attr_set_args
                     - 5.42% xfs_attr_set_shortform.constprop.0
                        - 4.51% xfs_trans_commit
                           - 4.54% __xfs_trans_commit
                              - 4.59% xfs_log_commit_cil
                                 - 2.67% _raw_spin_lock
                                    - 3.28% do_raw_spin_lock
                                         3.08% __pv_queued_spin_lock_slowpath
                                   0.66% xfs_inode_item_format
                        - 0.90% xfs_attr_try_sf_addname
                  - 0.60% xfs_trans_alloc
            - 2.35% selinux_inode_init_security
               - 1.25% security_sid_to_context_force
                  - 1.21% security_sid_to_context_core
                     - 1.19% sidtab_entry_to_string
                        - 1.20% sidtab_sid2str_get
                           - 0.86% sidtab_sid2str_put.part.0
                              - 0.62% _raw_spin_lock_irqsave
                                 - 0.77% do_raw_spin_lock
                                      __pv_queued_spin_lock_slowpath
               - 0.84% selinux_determine_inode_label
                  - 0.83% security_transition_sid
                       0.86% security_compute_sid.part.0

Which indicates the XFS overhead of creating the selinux xattr has
been halved. This doesn't fix the CIL lock contention problem, just
means it's not a limiting factor for this workload. Lock contention
in the security subsystems is going to be an issue soon, though...
Signed-off-by: NDave Chinner <dchinner@redhat.com>
Reviewed-by: NChristoph Hellwig <hch@lst.de>
[djwong: fix compilation error when CONFIG_SECURITY=n]
Reviewed-by: NDarrick J. Wong <djwong@kernel.org>
Signed-off-by: NDarrick J. Wong <djwong@kernel.org>
Reviewed-by: NGao Xiang <hsiangkao@redhat.com>
      e6a688c3
    • D
      xfs: prevent metadata files from being inactivated · 383e32b0
      由 Darrick J. Wong 提交于 
      Files containing metadata (quota records, rt bitmap and summary info)
are fully managed by the filesystem, which means that all resource
cleanup must be explicit, not automatic.  This means that they should
never be subjected automatic to post-eof truncation, nor should they be
freed automatically even if the link count drops to zero.

In other words, xfs_inactive() should leave these files alone.  Add the
necessary predicate functions to make this happen.  This adds a second
layer of prevention for the kinds of fs corruption that was fixed by
commit f4c32e87.  If we ever decide to support removing metadata
files, we should make all those metadata updates explicit.

Rearrange the order of #includes to fix compiler errors, since
xfs_mount.h is supposed to be included before xfs_inode.h

Followup-to: f4c32e87 ("xfs: fix realtime bitmap/summary file truncation when growing rt volume")
Signed-off-by: NDarrick J. Wong <djwong@kernel.org>
Reviewed-by: NChristoph Hellwig <hch@lst.de>
      383e32b0
  5. 24 1月, 2021 1 次提交
  7. 13 12月, 2020 1 次提交
  9. 16 9月, 2020 1 次提交
  11. 07 9月, 2020 1 次提交
    • D
      xfs: xfs_iflock is no longer a completion · 718ecc50
      由 Dave Chinner 提交于 
      With the recent rework of the inode cluster flushing, we no longer
ever wait on the the inode flush "lock". It was never a lock in the
first place, just a completion to allow callers to wait for inode IO
to complete. We now never wait for flush completion as all inode
flushing is non-blocking. Hence we can get rid of all the iflock
infrastructure and instead just set and check a state flag.

Rename the XFS_IFLOCK flag to XFS_IFLUSHING, convert all the
xfs_iflock_nowait() test-and-set operations on that flag, and
replace all the xfs_ifunlock() calls to clear operations.
Signed-off-by: NDave Chinner <dchinner@redhat.com>
Reviewed-by: NDarrick J. Wong <darrick.wong@oracle.com>
Signed-off-by: NDarrick J. Wong <darrick.wong@oracle.com>
      718ecc50
  13. 07 7月, 2020 3 次提交
    • D
      xfs: rework xfs_iflush_cluster() dirty inode iteration · 5717ea4d
      由 Dave Chinner 提交于 
      Now that we have all the dirty inodes attached to the cluster
buffer, we don't actually have to do radix tree lookups to find
them. Sure, the radix tree is efficient, but walking a linked list
of just the dirty inodes attached to the buffer is much better.

We are also no longer dependent on having a locked inode passed into
the function to determine where to start the lookup. This means we
can drop it from the function call and treat all inodes the same.

We also make xfs_iflush_cluster skip inodes marked with
XFS_IRECLAIM. This we avoid races with inodes that reclaim is
actively referencing or are being re-initialised by inode lookup. If
they are actually dirty, they'll get written by a future cluster
flush....

We also add a shutdown check after obtaining the flush lock so that
we catch inodes that are dirty in memory and may have inconsistent
state due to the shutdown in progress. We abort these inodes
directly and so they remove themselves directly from the buffer list
and the AIL rather than having to wait for the buffer to be failed
and callbacks run to be processed correctly.
Signed-off-by: NDave Chinner <dchinner@redhat.com>
Reviewed-by: NDarrick J. Wong <darrick.wong@oracle.com>
Signed-off-by: NDarrick J. Wong <darrick.wong@oracle.com>
      5717ea4d
    • D
      xfs: xfs_iflush() is no longer necessary · 90c60e16
      由 Dave Chinner 提交于 
      Now we have a cached buffer on inode log items, we don't need
to do buffer lookups when flushing inodes anymore - all we need
to do is lock the buffer and we are ready to go.

This largely gets rid of the need for xfs_iflush(), which is
essentially just a mechanism to look up the buffer and flush the
inode to it. Instead, we can just call xfs_iflush_cluster() with a
few modifications to ensure it also flushes the inode we already
hold locked.

This allows the AIL inode item pushing to be almost entirely
non-blocking in XFS - we won't block unless memory allocation
for the cluster inode lookup blocks or the block device queues are
full.

Writeback during inode reclaim becomes a little more complex because
we now have to lock the buffer ourselves, but otherwise this change
is largely a functional no-op that removes a whole lot of code.
Signed-off-by: NDave Chinner <dchinner@redhat.com>
Reviewed-by: NDarrick J. Wong <darrick.wong@oracle.com>
Reviewed-by: NBrian Foster <bfoster@redhat.com>
Signed-off-by: NDarrick J. Wong <darrick.wong@oracle.com>
      90c60e16
    • D
      xfs: move helpers that lock and unlock two inodes against userspace IO · e2aaee9c
      由 Darrick J. Wong 提交于 
      Move the double-inode locking helpers to xfs_inode.c since they're not
specific to reflink.
Signed-off-by: NDarrick J. Wong <darrick.wong@oracle.com>
Reviewed-by: NBrian Foster <bfoster@redhat.com>
      e2aaee9c
  15. 30 5月, 2020 1 次提交
  17. 20 5月, 2020 3 次提交
  19. 13 5月, 2020 1 次提交
  21. 05 5月, 2020 1 次提交
  23. 06 4月, 2020 1 次提交
  25. 14 11月, 2019 1 次提交
  27. 11 11月, 2019 1 次提交
  29. 28 10月, 2019 1 次提交
  31. 23 4月, 2019 1 次提交
    • D
      xfs: widen inode delalloc block counter to 64-bits · 394aafdc
      由 Darrick J. Wong 提交于 
      Widen the incore inode's i_delayed_blks counter to be a 64-bit integer.
This is necessary to fix an integer overflow problem that can be
reproduced easily now that we use the counter to track blocks that are
assigned to the inode in memory but not on disk.  This includes actual
delalloc reservations as well as real extents in the COW fork that
are waiting to be remapped into the data fork.

These 'delayed mapping' blocks can easily exceed 2^32 blocks if one
creates a very large sparse file of size approximately 2^33 bytes with
one byte written every 2^23 bytes, sets a very large COW extent size
hint of 2^23 blocks, reflinks the first file into a second file, and
then writes a single byte every 2^23 blocks in the original file.

When this happens, we'll try to create approximately 1024 2^23 extent
reservations in the COW fork, which will overflow the counter and cause
problems.

Note that on x64 we end up filling a 4-byte gap in the structure so this
doesn't increase the incore size.
Signed-off-by: NDarrick J. Wong <darrick.wong@oracle.com>
Reviewed-by: NAllison Collins <allison.henderson@oracle.com>
Reviewed-by: NDave Chinner <dchinner@redhat.com>
Reviewed-by: NChristoph Hellwig <hch@lst.de>
      394aafdc
  33. 17 4月, 2019 1 次提交
    • D
      xfs: implement per-inode writeback completion queues · cb357bf3
      由 Darrick J. Wong 提交于 
      When scheduling writeback of dirty file data in the page cache, XFS uses
IO completion workqueue items to ensure that filesystem metadata only
updates after the write completes successfully.  This is essential for
converting unwritten extents to real extents at the right time and
performing COW remappings.

Unfortunately, XFS queues each IO completion work item to an unbounded
workqueue, which means that the kernel can spawn dozens of threads to
try to handle the items quickly.  These threads need to take the ILOCK
to update file metadata, which results in heavy ILOCK contention if a
large number of the work items target a single file, which is
inefficient.

Worse yet, the writeback completion threads get stuck waiting for the
ILOCK while holding transaction reservations, which can use up all
available log reservation space.  When that happens, metadata updates to
other parts of the filesystem grind to a halt, even if the filesystem
could otherwise have handled it.

Even worse, if one of the things grinding to a halt happens to be a
thread in the middle of a defer-ops finish holding the same ILOCK and
trying to obtain more log reservation having exhausted the permanent
reservation, we now have an ABBA deadlock - writeback completion has a
transaction reserved and wants the ILOCK, and someone else has the ILOCK
and wants a transaction reservation.

Therefore, we create a per-inode writeback io completion queue + work
item.  When writeback finishes, it can add the ioend to the per-inode
queue and let the single worker item process that queue.  This
dramatically cuts down on the number of kworkers and ILOCK contention in
the system, and seems to have eliminated an occasional deadlock I was
seeing while running generic/476.

Testing with a program that simulates a heavy random-write workload to a
single file demonstrates that the number of kworkers drops from
approximately 120 threads per file to 1, without dramatically changing
write bandwidth or pagecache access latency.

Note that we leave the xfs-conv workqueue's max_active alone because we
still want to be able to run ioend processing for as many inodes as the
system can handle.
Signed-off-by: NDarrick J. Wong <darrick.wong@oracle.com>
Reviewed-by: NBrian Foster <bfoster@redhat.com>
      cb357bf3
  35. 15 4月, 2019 1 次提交
  37. 12 2月, 2019 1 次提交
    • D
      xfs: cache unlinked pointers in an rhashtable · 9b247179
      由 Darrick J. Wong 提交于 
      Use a rhashtable to cache the unlinked list incore.  This should speed
up unlinked processing considerably when there are a lot of inodes on
the unlinked list because iunlink_remove no longer has to traverse an
entire bucket list to find which inode points to the one being removed.

The incore list structure records "X.next_unlinked = Y" relations, with
the rhashtable using Y to index the records.  This makes finding the
inode X that points to a inode Y very quick.  If our cache fails to find
anything we can always fall back on the old method.

FWIW this drastically reduces the amount of time it takes to remove
inodes from the unlinked list.  I wrote a program to open a lot of
O_TMPFILE files and then close them in the same order, which takes
a very long time if we have to traverse the unlinked lists.  With the
ptach, I see:

+ /d/t/tmpfile/tmpfile
Opened 193531 files in 6.33s.
Closed 193531 files in 5.86s

real    0m12.192s
user    0m0.064s
sys     0m11.619s
+ cd /
+ umount /mnt

real    0m0.050s
user    0m0.004s
sys     0m0.030s

And without the patch:

+ /d/t/tmpfile/tmpfile
Opened 193588 files in 6.35s.
Closed 193588 files in 751.61s

real    12m38.853s
user    0m0.084s
sys     12m34.470s
+ cd /
+ umount /mnt

real    0m0.086s
user    0m0.000s
sys     0m0.060s
Signed-off-by: NDarrick J. Wong <darrick.wong@oracle.com>
Reviewed-by: NBrian Foster <bfoster@redhat.com>
      9b247179
  39. 03 8月, 2018 1 次提交
  41. 30 7月, 2018 2 次提交
  43. 27 7月, 2018 2 次提交
  45. 12 7月, 2018 1 次提交
  47. 07 6月, 2018 1 次提交
    • D
      xfs: convert to SPDX license tags · 0b61f8a4
      由 Dave Chinner 提交于 
      Remove the verbose license text from XFS files and replace them
with SPDX tags. This does not change the license of any of the code,
merely refers to the common, up-to-date license files in LICENSES/

This change was mostly scripted. fs/xfs/Makefile and
fs/xfs/libxfs/xfs_fs.h were modified by hand, the rest were detected
and modified by the following command:

for f in `git grep -l "GNU General" fs/xfs/` ; do
	echo $f
	cat $f | awk -f hdr.awk > $f.new
	mv -f $f.new $f
done

And the hdr.awk script that did the modification (including
detecting the difference between GPL-2.0 and GPL-2.0+ licenses)
is as follows:

$ cat hdr.awk
BEGIN {
	hdr = 1.0
	tag = "GPL-2.0"
	str = ""
}

/^ \* This program is free software/ {
	hdr = 2.0;
	next
}

/any later version./ {
	tag = "GPL-2.0+"
	next
}

/^ \*\// {
	if (hdr > 0.0) {
		print "// SPDX-License-Identifier: " tag
		print str
		print $0
		str=""
		hdr = 0.0
		next
	}
	print $0
	next
}

/^ \* / {
	if (hdr > 1.0)
		next
	if (hdr > 0.0) {
		if (str != "")
			str = str "\n"
		str = str $0
		next
	}
	print $0
	next
}

/^ \*/ {
	if (hdr > 0.0)
		next
	print $0
	next
}

// {
	if (hdr > 0.0) {
		if (str != "")
			str = str "\n"
		str = str $0
		next
	}
	print $0
}

END { }
$
Signed-off-by: NDave Chinner <dchinner@redhat.com>
Reviewed-by: NDarrick J. Wong <darrick.wong@oracle.com>
Signed-off-by: NDarrick J. Wong <darrick.wong@oracle.com>
      0b61f8a4