Simplify the calling convention of these functions since the
xfs_rtalloc_args structure contains the parameters we need.
Signed-off-by: Darrick J. Wong <djwong@kernel.org>
Reviewed-by: Christoph Hellwig <hch@lst.de>
Now that xfs_rtalloc_args holds references to the last-read bitmap and
summary blocks, we don't need to pass the buffer pointer out of
xfs_rtbuf_get.
Callers no longer have to xfs_trans_brelse on their own, though they are
required to call xfs_rtbuf_cache_relse before the xfs_rtalloc_args goes
out of scope.
While we're at it, create some trivial helpers so that we don't have to
remember if "0" means "bitmap" and "1" means "summary".
Signed-off-by: Darrick J. Wong <djwong@kernel.org>
Reviewed-by: Christoph Hellwig <hch@lst.de>
Profiling a workload on a highly fragmented realtime device showed a ton
of CPU cycles being spent in xfs_trans_read_buf() called by
xfs_rtbuf_get(). Further tracing showed that much of that was repeated
calls to xfs_rtbuf_get() for the same block of the realtime bitmap.
These come from xfs_rtallocate_extent_block(): as it walks through
ranges of free bits in the bitmap, each call to xfs_rtcheck_range() and
xfs_rtfind_{forw,back}() gets the same bitmap block. If the bitmap block
is very fragmented, then this is _a lot_ of buffer lookups.
The realtime allocator already passes around a cache of the last used
realtime summary block to avoid repeated reads (the parameters rbpp and
rsb). We can do the same for the realtime bitmap.
This replaces rbpp and rsb with a struct xfs_rtbuf_cache, which caches
the most recently used block for both the realtime bitmap and summary.
xfs_rtbuf_get() now handles the caching instead of the callers, which
requires plumbing xfs_rtbuf_cache to more functions but also makes sure
we don't miss anything.
Signed-off-by: Omar Sandoval <osandov@fb.com>
Reviewed-by: Darrick J. Wong <djwong@kernel.org>
Signed-off-by: Darrick J. Wong <djwong@kernel.org>
Reviewed-by: Christoph Hellwig <hch@lst.de>
Consolidate the arguments passed around the rt allocator into a
struct xfs_rtalloc_arg similar to how the btree allocator arguments
are consolidated in a struct xfs_alloc_arg....
Signed-off-by: Dave Chinner <dchinner@redhat.com>
Reviewed-by: Darrick J. Wong <djwong@kernel.org>
Signed-off-by: Darrick J. Wong <djwong@kernel.org>
Reviewed-by: Christoph Hellwig <hch@lst.de>
Create get and set functions for rtsummary words so that we can redefine
the ondisk format with a specific endianness. Note that this requires
the definition of a distinct type for ondisk summary info words so that
the compiler can perform proper typechecking.
Signed-off-by: Darrick J. Wong <djwong@kernel.org>
Reviewed-by: Christoph Hellwig <hch@lst.de>
Create helper functions that compute the number of blocks or words
necessary to store the rt bitmap.
Signed-off-by: Darrick J. Wong <djwong@kernel.org>
Reviewed-by: Christoph Hellwig <hch@lst.de>
Convert the realtime summary file macros to helper functions so that we
can improve type checking.
Signed-off-by: Darrick J. Wong <djwong@kernel.org>
Reviewed-by: Christoph Hellwig <hch@lst.de>
Replace these macros with typechecked helper functions. Eventually
we're going to add more logic to the helpers and it'll be easier if we
don't have to macro it up.
Signed-off-by: Darrick J. Wong <djwong@kernel.org>
Reviewed-by: Christoph Hellwig <hch@lst.de>
Convert these calls to use the helpers, and clean up all these places
where the same variable can have different units depending on where it
is in the function.
Signed-off-by: Darrick J. Wong <djwong@kernel.org>
Reviewed-by: Christoph Hellwig <hch@lst.de>
Create a helper to compute the misalignment between a file extent
(xfs_extlen_t) and a realtime extent.
Signed-off-by: Darrick J. Wong <djwong@kernel.org>
Reviewed-by: Christoph Hellwig <hch@lst.de>
Create a helper to convert a realtime extent to a realtime block. Later
on we'll change the helper to use bit shifts when possible.
Signed-off-by: Darrick J. Wong <djwong@kernel.org>
Reviewed-by: Christoph Hellwig <hch@lst.de>
Further disambiguate the xfs_rtblock_t uses by creating a new type,
xfs_rtxnum_t, to store the position of an extent within the realtime
section, in units of rtextents.
Signed-off-by: Darrick J. Wong <djwong@kernel.org>
Reviewed-by: Christoph Hellwig <hch@lst.de>
This helper function validates that a range of *blocks* in the
realtime section is completely contained within the realtime section.
It does /not/ validate ranges of *rtextents*. Rename the function to
avoid suggesting that it does, and change the type of the @len parameter
since xfs_rtblock_t is a position unit, not a length unit.
Signed-off-by: Darrick J. Wong <djwong@kernel.org>
Reviewed-by: Christoph Hellwig <hch@lst.de>
XFS uses xfs_rtblock_t for many different uses, which makes it much more
difficult to perform a unit analysis on the codebase. One of these
(ab)uses is when we need to store the length of a free space extent as
stored in the realtime bitmap. Because there can be up to 2^64 realtime
extents in a filesystem, we need a new type that is larger than
xfs_rtxlen_t for callers that are querying the bitmap directly. This
means scrub and growfs.
Create this type as "xfs_rtbxlen_t" and use it to store 64-bit rtx
lengths. 'b' stands for 'bitmap' or 'big'; reader's choice.
Signed-off-by: Darrick J. Wong <djwong@kernel.org>
Reviewed-by: Christoph Hellwig <hch@lst.de>
In most of the filesystem, we use xfs_extlen_t to store the length of a
file (or AG) space mapping in units of fs blocks. Unfortunately, the
realtime allocator also uses it to store the length of a rt space
mapping in units of rt extents. This is confusing, since one rt extent
can consist of many fs blocks.
Separate the two by introducing a new type (xfs_rtxlen_t) to store the
length of a space mapping (in units of realtime extents) that would be
found in a file.
Signed-off-by: Darrick J. Wong <djwong@kernel.org>
Reviewed-by: Christoph Hellwig <hch@lst.de>
Move all the declarations for functionality in xfs_rtbitmap.c into a
separate xfs_rtbitmap.h header file.
Signed-off-by: Darrick J. Wong <djwong@kernel.org>
Reviewed-by: Christoph Hellwig <hch@lst.de>
./fs/xfs/scrub/xfile.c: xfs_format.h is included more than once.
Reported-by: Abaci Robot <abaci@linux.alibaba.com>
Closes: https://bugzilla.openanolis.cn/show_bug.cgi?id=6209
Signed-off-by: Jiapeng Chong <jiapeng.chong@linux.alibaba.com>
Reviewed-by: "Darrick J. Wong" <djwong@kernel.org>
Signed-off-by: Chandan Babu R <chandanbabu@kernel.org>
Harshit Mogalapalli slogged through several reports from our internal
syzbot instance and observed that they all had a common stack trace:
BUG: KASAN: user-memory-access in instrument_atomic_read_write include/linux/instrumented.h:96 [inline]
BUG: KASAN: user-memory-access in atomic_try_cmpxchg_acquire include/linux/atomic/atomic-instrumented.h:1294 [inline]
BUG: KASAN: user-memory-access in queued_spin_lock include/asm-generic/qspinlock.h:111 [inline]
BUG: KASAN: user-memory-access in do_raw_spin_lock include/linux/spinlock.h:187 [inline]
BUG: KASAN: user-memory-access in __raw_spin_lock include/linux/spinlock_api_smp.h:134 [inline]
BUG: KASAN: user-memory-access in _raw_spin_lock+0x76/0xe0 kernel/locking/spinlock.c:154
Write of size 4 at addr 0000001dd87ee280 by task syz-executor365/1543
CPU: 2 PID: 1543 Comm: syz-executor365 Not tainted 6.5.0-syzk #1
Hardware name: Red Hat KVM, BIOS 1.13.0-2.module+el8.3.0+7860+a7792d29 04/01/2014
Call Trace:
<TASK>
__dump_stack lib/dump_stack.c:88 [inline]
dump_stack_lvl+0x83/0xb0 lib/dump_stack.c:106
print_report+0x3f8/0x620 mm/kasan/report.c:478
kasan_report+0xb0/0xe0 mm/kasan/report.c:588
check_region_inline mm/kasan/generic.c:181 [inline]
kasan_check_range+0x139/0x1e0 mm/kasan/generic.c:187
instrument_atomic_read_write include/linux/instrumented.h:96 [inline]
atomic_try_cmpxchg_acquire include/linux/atomic/atomic-instrumented.h:1294 [inline]
queued_spin_lock include/asm-generic/qspinlock.h:111 [inline]
do_raw_spin_lock include/linux/spinlock.h:187 [inline]
__raw_spin_lock include/linux/spinlock_api_smp.h:134 [inline]
_raw_spin_lock+0x76/0xe0 kernel/locking/spinlock.c:154
spin_lock include/linux/spinlock.h:351 [inline]
xchk_stats_merge_one.isra.1+0x39/0x650 fs/xfs/scrub/stats.c:191
xchk_stats_merge+0x5f/0xe0 fs/xfs/scrub/stats.c:225
xfs_scrub_metadata+0x252/0x14e0 fs/xfs/scrub/scrub.c:599
xfs_ioc_scrub_metadata+0xc8/0x160 fs/xfs/xfs_ioctl.c:1646
xfs_file_ioctl+0x3fd/0x1870 fs/xfs/xfs_ioctl.c:1955
vfs_ioctl fs/ioctl.c:51 [inline]
__do_sys_ioctl fs/ioctl.c:871 [inline]
__se_sys_ioctl fs/ioctl.c:857 [inline]
__x64_sys_ioctl+0x199/0x220 fs/ioctl.c:857
do_syscall_x64 arch/x86/entry/common.c:50 [inline]
do_syscall_64+0x3e/0x90 arch/x86/entry/common.c:80
entry_SYSCALL_64_after_hwframe+0x6e/0xd8
RIP: 0033:0x7ff155af753d
Code: 00 c3 66 2e 0f 1f 84 00 00 00 00 00 90 f3 0f 1e fa 48 89 f8 48 89 f7 48 89 d6 48 89 ca 4d 89 c2 4d 89 c8 4c 8b 4c 24 08 0f 05 <48> 3d 01 f0 ff ff 73 01 c3 48 8b 0d 1b 79 2c 00 f7 d8 64 89 01 48
RSP: 002b:00007ffc006e2568 EFLAGS: 00000246 ORIG_RAX: 0000000000000010
RAX: ffffffffffffffda RBX: 0000000000000000 RCX: 00007ff155af753d
RDX: 00000000200000c0 RSI: 00000000c040583c RDI: 0000000000000003
RBP: 00000000ffffffff R08: 00000000004010c0 R09: 00000000004010c0
R10: 00000000004010c0 R11: 0000000000000246 R12: 0000000000400cb0
R13: 00007ffc006e2670 R14: 0000000000000000 R15: 0000000000000000
</TASK>
The root cause here is that xchk_stats_merge_one walks off the end of
the xchk_scrub_stats.cs_stats array because it has been fed a garbage
value in sm->sm_type. That occurs because I put the xchk_stats_merge
in the wrong place -- it should have been after the last xchk_teardown
call on our way out of xfs_scrub_metadata because we only call the
teardown function if we called the setup function, and we don't call the
setup functions if the inputs are obviously garbage.
Thanks to Harshit for triaging the bug reports and bringing this to my
attention.
Fixes: d7a74cad8f ("xfs: track usage statistics of online fsck")
Reported-by: Harshit Mogalapalli <harshit.m.mogalapalli@oracle.com>
Signed-off-by: Darrick J. Wong <djwong@kernel.org>
Reviewed-by: Dave Chinner <dchinner@redhat.com>
* Chandan Babu will be taking over as the XFS release manager. He has
reviewed all the patches that are in this branch, though I'm signing
the branch one last time since I'm still technically maintainer. :P
* Create a maintainer entry profile for XFS in which we lay out the
various roles that I have played for many years. Aside from release
manager, the remaining roles are as yet unfilled.
* Start merging online repair -- we now have in-memory pageable memory
for staging btrees, a bunch of pending fixes, and we've started the
process of refactoring the scrub support code to support more of
repair. In particular, reaping of old blocks from damaged structures.
* Scrub the realtime summary file.
* Fix a bug where scrub's quota iteration only ever returned the root
dquot. Oooops.
* Fix some typos.
Signed-off-by: Darrick J. Wong <djwong@kernel.org>
-----BEGIN PGP SIGNATURE-----
iHUEABYKAB0WIQQ2qTKExjcn+O1o2YRKO3ySh0YRpgUCZOQE2AAKCRBKO3ySh0YR
pvmZAQDe+KceaVx6Dv2f9ihckeS2dILSpDTo1bh9BeXnt005VwD/ceHTaJxEl8lp
u/dixFDkRgp9RYtoTAK2WNiUxYetsAc=
=oZN6
-----END PGP SIGNATURE-----
Merge tag 'xfs-6.6-merge-1' of git://git.kernel.org/pub/scm/fs/xfs/xfs-linux
Pull xfs updates from Chandan Babu:
- Chandan Babu will be taking over as the XFS release manager. He has
reviewed all the patches that are in this branch, though I'm signing
the branch one last time since I'm still technically maintainer. :P
- Create a maintainer entry profile for XFS in which we lay out the
various roles that I have played for many years. Aside from release
manager, the remaining roles are as yet unfilled.
- Start merging online repair -- we now have in-memory pageable memory
for staging btrees, a bunch of pending fixes, and we've started the
process of refactoring the scrub support code to support more of
repair. In particular, reaping of old blocks from damaged structures.
- Scrub the realtime summary file.
- Fix a bug where scrub's quota iteration only ever returned the root
dquot. Oooops.
- Fix some typos.
[ Pull request from Chandan Babu, but signed tag and description from
Darrick Wong, thus the first person singular above is Darrick, not
Chandan ]
* tag 'xfs-6.6-merge-1' of git://git.kernel.org/pub/scm/fs/xfs/xfs-linux: (37 commits)
fs/xfs: Fix typos in comments
xfs: fix dqiterate thinko
xfs: don't check reflink iflag state when checking cow fork
xfs: simplify returns in xchk_bmap
xfs: rewrite xchk_inode_is_allocated to work properly
xfs: hide xfs_inode_is_allocated in scrub common code
xfs: fix agf_fllast when repairing an empty AGFL
xfs: allow userspace to rebuild metadata structures
xfs: clear pagf_agflreset when repairing the AGFL
xfs: allow the user to cancel repairs before we start writing
xfs: don't complain about unfixed metadata when repairs were injected
xfs: implement online scrubbing of rtsummary info
xfs: always rescan allegedly healthy per-ag metadata after repair
xfs: move the realtime summary file scrubber to a separate source file
xfs: wrap ilock/iunlock operations on sc->ip
xfs: get our own reference to inodes that we want to scrub
xfs: track usage statistics of online fsck
xfs: improve xfarray quicksort pivot
xfs: create scaffolding for creating debugfs entries
xfs: cache pages used for xfarray quicksort convergence
...
Any inode on a reflink filesystem can have a cow fork, even if the inode
does not have the reflink iflag set. This happens either because the
inode once had the iflag set but does not now, because we don't free the
incore cow fork until the icache deletes the inode; or because we're
running in alwayscow mode.
Either way, we can collapse both of the xfs_is_reflink_inode calls into
one, and change it to xfs_has_reflink, now that the bmap checker will
return ENOENT if there is no pointer to the incore fork.
Signed-off-by: Darrick J. Wong <djwong@kernel.org>
Reviewed-by: Dave Chinner <dchinner@redhat.com>
Remove the pointless goto and return code in xchk_bmap, since it only
serves to obscure what's going on in the function. Instead, return
whichever error code is appropriate there. For nonexistent forks,
this should have been ENOENT.
Signed-off-by: Darrick J. Wong <djwong@kernel.org>
Reviewed-by: Dave Chinner <dchinner@redhat.com>
Back in the mists of time[1], I proposed this function to assist the
inode btree scrubbers in checking the inode btree contents against the
allocation state of the inode records. The original version performed a
direct lookup in the inode cache and returned the allocation status if
the cached inode hadn't been reused and wasn't in an intermediate state.
Brian thought it would be better to use the usual iget/irele mechanisms,
so that was changed for the final version.
Unfortunately, this hasn't aged well -- the IGET_INCORE flag only has
one user and clutters up the regular iget path, which makes it hard to
reason about how it actually works. Worse yet, the inode inactivation
series silently broke it because iget won't return inodes that are
anywhere in the inactivation machinery, even though the caller is
already required to prevent inode allocation and freeing. Inodes in the
inactivation machinery are still allocated, but the current code's
interactions with the iget code prevent us from being able to say that.
Now that I understand the inode lifecycle better than I did in early
2017, I now realize that as long as the cached inode hasn't been reused
and isn't actively being reclaimed, it's safe to access the i_mode field
(with the AGI, rcu, and i_flags locks held), and we don't need to worry
about the inode being freed out from under us.
Therefore, port the original version to modern code structure, which
fixes the brokennes w.r.t. inactivation. In the next patch we'll remove
IGET_INCORE since it's no longer necessary.
[1] https://lore.kernel.org/linux-xfs/149643868294.23065.8094890990886436794.stgit@birch.djwong.org/
Signed-off-by: Darrick J. Wong <djwong@kernel.org>
Reviewed-by: Dave Chinner <dchinner@redhat.com>
This function is only used by online fsck, so let's move it there.
In the next patch, we'll fix it to work properly and to require that the
caller hold the AGI buffer locked. No major changes aside from
adjusting the signature a bit.
Signed-off-by: Darrick J. Wong <djwong@kernel.org>
Reviewed-by: Dave Chinner <dchinner@redhat.com>
xfs/139 with parent pointers enabled occasionally pops up a corruption
message when online fsck force-rebuild repairs an AGFL:
XFS (sde): Metadata corruption detected at xfs_agf_verify+0x11e/0x220 [xfs], xfs_agf block 0x9e0001
XFS (sde): Unmount and run xfs_repair
XFS (sde): First 128 bytes of corrupted metadata buffer:
00000000: 58 41 47 46 00 00 00 01 00 00 00 4f 00 00 40 00 XAGF.......O..@.
00000010: 00 00 00 01 00 00 00 02 00 00 00 05 00 00 00 01 ................
00000020: 00 00 00 01 00 00 00 01 00 00 00 00 ff ff ff ff ................
00000030: 00 00 00 00 00 00 00 05 00 00 00 05 00 00 00 00 ................
00000040: 91 2e 6f b1 ed 61 4b 4d 8c 9b 6e 87 08 bb f6 36 ..o..aKM..n....6
00000050: 00 00 00 01 00 00 00 01 00 00 00 06 00 00 00 01 ................
00000060: 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 ................
00000070: 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 ................
The root cause of this failure is that prior to the repair, there were
zero blocks in the AGFL. This scenario is set up by the test case, since
it formats with 64MB AGs and tries to ENOSPC the whole filesystem. In
this case of flcount==0, we reset fllast to -1U, which then trips the
write verifier's check that fllast is less than xfs_agfl_size().
Correct this code to set fllast to the last possible slot in the AGFL
when flcount is zero, which mirrors the behavior of xfs_repair phase5
when it has to create a totally empty AGFL.
Fixes: 0e93d3f43e ("xfs: repair the AGFL")
Signed-off-by: Darrick J. Wong <djwong@kernel.org>
Reviewed-by: Dave Chinner <dchinner@redhat.com>
Clear the pagf_agflreset flag when we're repairing the AGFL because we
fix all the same padding problems that xfs_agfl_reset does.
Signed-off-by: Darrick J. Wong <djwong@kernel.org>
Reviewed-by: Dave Chinner <dchinner@redhat.com>
Add a new (superuser-only) flag to the online metadata repair ioctl to
force it to rebuild structures, even if they're not broken. We will use
this to move metadata structures out of the way during a free space
defragmentation operation.
Signed-off-by: Darrick J. Wong <djwong@kernel.org>
Reviewed-by: Dave Chinner <dchinner@redhat.com>
While debugging other parts of online repair, I noticed that if someone
injects FORCE_SCRUB_REPAIR, starts an IFLAG_REPAIR scrub on a piece of
metadata, and the metadata repair fails, we'll log a message about
uncorrected errors in the filesystem.
This isn't strictly true if the scrub function didn't set OFLAG_CORRUPT
and we're only doing the repair because the error injection knob is set.
Repair functions are allowed to abort the entire operation at any point
before committing new metadata, in which case the piece of metadata is
in the same state as it was before. Therefore, the log message should
be gated on the results of the scrub. Refactor the predicate and
rearrange the code flow to make this happen.
Note: If the repair function errors out after it commits the new
metadata, the transaction cancellation will shut down the filesystem,
which is an obvious sign of corrupt metadata.
Signed-off-by: Darrick J. Wong <djwong@kernel.org>
Reviewed-by: Dave Chinner <dchinner@redhat.com>
All online repair functions have the same structure: walk filesystem
metadata structures gathering enough data to rebuild the structure,
stage a new copy, and then commit the new copy.
The gathering steps do not write anything to disk, so they are peppered
with xchk_should_terminate calls to avoid softlockup warnings and to
provide an opportunity to abort the repair (by killing xfs_scrub).
However, it's not clear in the code base when is the last chance to
abort cleanly without having to undo a bunch of structure.
Therefore, add one more call to xchk_should_terminate (along with a
comment) providing the sysadmin with the ability to abort before it's
too late and to make it clear in the source code when it's no longer
convenient or safe to abort a repair. As there are only four repair
functions right now, this patch exists more to establish a precedent for
subsequent additions than to deliver practical functionality.
Signed-off-by: Darrick J. Wong <djwong@kernel.org>
Reviewed-by: Dave Chinner <dchinner@redhat.com>
After an online repair function runs for a per-AG metadata structure,
sc->sick_mask is supposed to reflect the per-AG metadata that the repair
function fixed. Our next move is to re-check the metadata to assess
the completeness of our repair, so we don't want the rebuilt structure
to be excluded from the rescan just because the health system previously
logged a problem with the data structure.
Signed-off-by: Darrick J. Wong <djwong@kernel.org>
Reviewed-by: Dave Chinner <dchinner@redhat.com>
Finish the realtime summary scrubber by adding the functions we need to
compute a fresh copy of the rtsummary info and comparing it to the copy
on disk.
Signed-off-by: Darrick J. Wong <djwong@kernel.org>
Reviewed-by: Dave Chinner <dchinner@redhat.com>
Move the realtime summary file checking code to a separate file in
preparation to actually implement it.
Signed-off-by: Darrick J. Wong <djwong@kernel.org>
Reviewed-by: Dave Chinner <dchinner@redhat.com>
Scrub tracks the resources that it's holding onto in the xfs_scrub
structure. This includes the inode being checked (if applicable) and
the inode lock state of that inode. Replace the open-coded structure
manipulation with a trivial helper to eliminate sources of error.
Signed-off-by: Darrick J. Wong <djwong@kernel.org>
Reviewed-by: Dave Chinner <dchinner@redhat.com>
When we want to scrub a file, get our own reference to the inode
unconditionally. This will make disposal rules simpler in the long run.
Signed-off-by: Darrick J. Wong <djwong@kernel.org>
Reviewed-by: Dave Chinner <dchinner@redhat.com>
Track the usage, outcomes, and run times of the online fsck code, and
report these values via debugfs. The columns in the file are:
* scrubber name
* number of scrub invocations
* clean objects found
* corruptions found
* optimizations found
* cross referencing failures
* inconsistencies found during cross referencing
* incomplete scrubs
* warnings
* number of time scrub had to retry
* cumulative amount of time spent scrubbing (microseconds)
* number of repair inovcations
* successfully repaired objects
* cumuluative amount of time spent repairing (microseconds)
Signed-off-by: Darrick J. Wong <djwong@kernel.org>
Reviewed-by: Dave Chinner <dchinner@redhat.com>
Now that we have the means to do insertion sorts of small in-memory
subsets of an xfarray, use it to improve the quicksort pivot algorithm
by reading 7 records into memory and finding the median of that. This
should prevent bad partitioning when a[lo] and a[hi] end up next to each
other in the final sort, which can happen when sorting for cntbt repair
when the free space is extremely fragmented (e.g. generic/176).
This doesn't speed up the average quicksort run by much, but it will
(hopefully) avoid the quadratic time collapse for which quicksort is
famous.
Signed-off-by: Darrick J. Wong <djwong@kernel.org>
Reviewed-by: Kent Overstreet <kent.overstreet@linux.dev>
Reviewed-by: Dave Chinner <dchinner@redhat.com>
After quicksort picks a pivot item for a particular subsort, it walks
the records in that subset from the outside in, rearranging them so that
every record less than the pivot comes before it, and every record
greater than the pivot comes after it. This scan has a lot of locality,
so we can speed it up quite a bit by grabbing the xfile backing page and
holding onto it as long as we possibly can. Doing so reduces the
runtime by another 5% on the author's computer.
Signed-off-by: Darrick J. Wong <djwong@kernel.org>
Reviewed-by: Kent Overstreet <kent.overstreet@linux.dev>
Reviewed-by: Dave Chinner <dchinner@redhat.com>
If all the records in an xfarray subset live within the same memory
page, we can short-circuit even more quicksort recursion by mapping that
page into the local CPU and using the kernel's heapsort function to sort
the subset. On the author's computer, this reduces the runtime by
another 15% on a 500,000 element array.
Signed-off-by: Darrick J. Wong <djwong@kernel.org>
Reviewed-by: Kent Overstreet <kent.overstreet@linux.dev>
Reviewed-by: Dave Chinner <dchinner@redhat.com>
Certain xfile array operations (such as sorting) can be sped up quite a
bit by allowing xfile users to grab a page to bulk-read the records
contained within it. Create helper methods to facilitate this.
Signed-off-by: Darrick J. Wong <djwong@kernel.org>
Reviewed-by: Kent Overstreet <kent.overstreet@linux.dev>
Reviewed-by: Dave Chinner <dchinner@redhat.com>
In the previous patch, we created a very basic quicksort implementation
for xfile arrays. While the use of an alternate sorting algorithm to
avoid quicksort recursion on very small subsets reduces the runtime
modestly, we could do better than a load and store-heavy insertion sort,
particularly since each load and store requires a page mapping lookup in
the xfile.
For a small increase in kernel memory requirements, we could instead
bulk load the xfarray records into memory, use the kernel's existing
heapsort implementation to sort the records, and bulk store the memory
buffer back into the xfile. On the author's computer, this reduces the
runtime by about 5% on a 500,000 element array.
Signed-off-by: Darrick J. Wong <djwong@kernel.org>
Reviewed-by: Kent Overstreet <kent.overstreet@linux.dev>
Reviewed-by: Dave Chinner <dchinner@redhat.com>
The btree bulk loading code requires that records be provided in the
correct record sort order for the given btree type. In general, repair
code cannot be required to collect records in order, and it is not
feasible to insert new records in the middle of an array to maintain
sort order.
Implement a sorting algorithm so that we can sort the records just prior
to bulk loading. In principle, an xfarray could consume many gigabytes
of memory and its backing pages can be sent out to disk at any time.
This means that we cannot map the entire array into memory at once, so
we must find a way to divide the work into smaller portions (e.g. a
page) that /can/ be mapped into memory.
Quicksort seems like a reasonable fit for this purpose, since it uses a
divide and conquer strategy to keep its average runtime logarithmic.
The solution presented here is a port of the glibc implementation, which
itself is derived from the median-of-three and tail call recursion
strategies outlined by Sedgwick.
Subsequent patches will optimize the implementation further by utilizing
the kernel's heapsort on directly-mapped memory whenever possible, and
improving the quicksort pivot selection algorithm to try to avoid O(n^2)
collapses.
Note: The sorting functionality gets its own patch because the basic big
array mechanisms were plenty for a single code patch.
Signed-off-by: Darrick J. Wong <djwong@kernel.org>
Reviewed-by: Kent Overstreet <kent.overstreet@linux.dev>
Reviewed-by: Dave Chinner <dchinner@redhat.com>
Create a simple 'big array' data structure for storage of fixed-size
metadata records that will be used to reconstruct a btree index. For
repair operations, the most important operations are append, iterate,
and sort.
Earlier implementations of the big array used linked lists and suffered
from severe problems -- pinning all records in kernel memory was not a
good idea and frequently lead to OOM situations; random access was very
inefficient; and record overhead for the lists was unacceptably high at
40-60%.
Therefore, the big memory array relies on the 'xfile' abstraction, which
creates a memfd file and stores the records in page cache pages. Since
the memfd is created in tmpfs, the memory pages can be pushed out to
disk if necessary and we have a built-in usage limit of 50% of physical
memory.
Signed-off-by: Darrick J. Wong <djwong@kernel.org>
Reviewed-by: Kent Overstreet <kent.overstreet@linux.dev>
Reviewed-by: Dave Chinner <dchinner@redhat.com>
The AGFL repair code uses a series of bitmaps to figure out where there
are OWN_AG blocks that are not claimed by the free space and rmap
btrees. These blocks become the new AGFL, and any overflow is reaped.
The bitmaps current track xfs_fsblock_t even though we already know the
AG number.
In the last patch, we introduced a new bitmap "type" for tracking
xfs_agblock_t extents. Port the reaping code and the AGFL repair to use
this new type, which makes it very obvious what we're tracking. This
also eliminates a bunch of unnecessary agblock <-> fsblock conversions.
Signed-off-by: Darrick J. Wong <djwong@kernel.org>
Reviewed-by: Dave Chinner <dchinner@redhat.com>
When we're freeing extents that have been set in a bitmap, break the
bitmap extent into multiple sub-extents organized by fate, and reap the
extents. This enables us to dispose of old resources more efficiently
than doing them block by block.
While we're at it, rename the reaping functions to make it clear that
they're reaping per-AG extents.
Signed-off-by: Darrick J. Wong <djwong@kernel.org>
Reviewed-by: Dave Chinner <dchinner@redhat.com>
After an online repair, we need to invalidate buffers representing the
blocks from the old metadata that we're replacing. It's possible that
parts of a tree that were previously cached in memory are no longer
accessible due to media failure or other corruption on interior nodes,
so repair figures out the old blocks from the reverse mapping data and
scans the buffer cache directly.
In other words, online fsck needs to find all the live (i.e. non-stale)
buffers for a range of fsblocks so that it can invalidate them.
Unfortunately, the current buffer cache code triggers asserts if the
rhashtable lookup finds a non-stale buffer of a different length than
the key we searched for. For regular operation this is desirable, but
for this repair procedure, we don't care since we're going to forcibly
stale the buffer anyway. Add an internal lookup flag to avoid the
assert. Skip buffers that are already XBF_STALE.
Signed-off-by: Darrick J. Wong <djwong@kernel.org>
Reviewed-by: Dave Chinner <dchinner@redhat.com>
Rearrange the logic inside xrep_reap_block to make it more obvious that
crosslinked metadata blocks are handled differently. Add a couple of
tracepoints so that we can tell what's going on at the end of a btree
rebuild operation.
Signed-off-by: Darrick J. Wong <djwong@kernel.org>
Reviewed-by: Dave Chinner <dchinner@redhat.com>
Use deferred frees (EFIs) to reap the blocks of a btree that we just
replaced. This helps us to shrink the window in which those old blocks
could be lost due to a system crash, though we try to flush the EFIs
every few hundred blocks so that we don't also overflow the transaction
reservations during and after we commit the new btree.
Signed-off-by: Darrick J. Wong <djwong@kernel.org>
Reviewed-by: Dave Chinner <dchinner@redhat.com>
Now that we've refactored btree cursors to require the caller to pass in
a perag structure, there are numerous problems in xrep_reap_extents if
it's being called to reap extents for an inode metadata repair. We
don't have any repair functions that can do that, so drop the support
for now.
Signed-off-by: Darrick J. Wong <djwong@kernel.org>
Reviewed-by: Dave Chinner <dchinner@redhat.com>
When we're discarding old btree blocks after a repair, only invalidate
the buffers for the ones that we're freeing -- if the metadata was
crosslinked with another data structure, we don't want to touch it.
Signed-off-by: Darrick J. Wong <djwong@kernel.org>
Reviewed-by: Dave Chinner <dchinner@redhat.com>
Reaping blocks after a repair is a complicated affair involving a lot of
rmap btree lookups and figuring out if we're going to unmap or free old
metadata blocks that might be crosslinked. Eventually, we will need to
be able to reap per-AG metadata blocks, bmbt blocks from inode forks,
garbage CoW staging extents, and (even later) blocks from btrees rooted
in inodes. This results in a lot of reaping code, so we might as well
split that off while it's easy.
Signed-off-by: Darrick J. Wong <djwong@kernel.org>
Reviewed-by: Dave Chinner <dchinner@redhat.com>
These two functions date from the era when I thought that we could
rebuild btrees by creating an alternate root and adding records one by
one. In other words, they predate the btree bulk loader. They're not
necessary now, so remove them.
Signed-off-by: Darrick J. Wong <djwong@kernel.org>
Reviewed-by: Dave Chinner <dchinner@redhat.com>
