Merge hickup on my part, due to a clash between the writeback
changes and the EOPNOTSUPP removal in _submit_bh().
Signed-off-by: Jens Axboe <axboe@fb.com>
As concurrent write sharing of an inode is expected to be very rare
and memcg only tracks page ownership on first-use basis severely
confining the usefulness of such sharing, cgroup writeback tracks
ownership per-inode. While the support for concurrent write sharing
of an inode is deemed unnecessary, an inode being written to by
different cgroups at different points in time is a lot more common,
and, more importantly, charging only by first-use can too readily lead
to grossly incorrect behaviors (single foreign page can lead to
gigabytes of writeback to be incorrectly attributed).
To resolve this issue, cgroup writeback detects the majority dirtier
of an inode and will transfer the ownership to it. To avoid
unnnecessary oscillation, the detection mechanism keeps track of
history and gives out the switch verdict only if the foreign usage
pattern is stable over a certain amount of time and/or writeback
attempts.
The detection mechanism has fairly low space and computation overhead.
It adds 8 bytes to struct inode (one int and two u16's) and minimal
amount of calculation per IO. The detection mechanism converges to
the correct answer usually in several seconds of IO time when there's
a clear majority dirtier. Even when there isn't, it can reach an
acceptable answer fairly quickly under most circumstances.
Please see wb_detach_inode() for more details.
This patch only implements detection. Following patches will
implement actual switching.
v2: wbc_account_io() now checks whether the wbc is associated with a
wb before dereferencing it. This can happen when pageout() is
writing pages directly without going through the usual writeback
path. As pageout() path is single-threaded, we don't want it to
be blocked behind a slow cgroup and ultimately want it to delegate
actual writing to the usual writeback path.
Signed-off-by: Tejun Heo <tj@kernel.org>
Cc: Jens Axboe <axboe@kernel.dk>
Cc: Jan Kara <jack@suse.cz>
Cc: Wu Fengguang <fengguang.wu@intel.com>
Cc: Greg Thelen <gthelen@google.com>
Signed-off-by: Jens Axboe <axboe@fb.com>
Currently, for cgroup writeback, the IO submission paths directly
associate the bio's with the blkcg from inode_to_wb_blkcg_css();
however, it'd be necessary to keep more writeback context to implement
foreign inode writeback detection. wbc (writeback_control) is the
natural fit for the extra context - it persists throughout the
writeback of each inode and is passed all the way down to IO
submission paths.
This patch adds wbc_attach_and_unlock_inode(), wbc_detach_inode(), and
wbc_attach_fdatawrite_inode() which are used to associate wbc with the
inode being written back. IO submission paths now use wbc_init_bio()
instead of directly associating bio's with blkcg themselves. This
leaves inode_to_wb_blkcg_css() w/o any user. The function is removed.
wbc currently only tracks the associated wb (bdi_writeback). Future
patches will add more for foreign inode detection. The association is
established under i_lock which will be depended upon when migrating
foreign inodes to other wb's.
As currently, once established, inode to wb association never changes,
going through wbc when initializing bio's doesn't cause any behavior
changes.
v2: submit_blk_blkcg() now checks whether the wbc is associated with a
wb before dereferencing it. This can happen when pageout() is
writing pages directly without going through the usual writeback
path. As pageout() path is single-threaded, we don't want it to
be blocked behind a slow cgroup and ultimately want it to delegate
actual writing to the usual writeback path.
Signed-off-by: Tejun Heo <tj@kernel.org>
Cc: Jens Axboe <axboe@kernel.dk>
Cc: Jan Kara <jack@suse.cz>
Cc: Wu Fengguang <fengguang.wu@intel.com>
Cc: Greg Thelen <gthelen@google.com>
Signed-off-by: Jens Axboe <axboe@fb.com>
[__]block_write_full_page() is used to implement ->writepage in
various filesystems. All writeback logic is now updated to handle
cgroup writeback and the block cgroup to issue IOs for is encoded in
writeback_control and can be retrieved from the inode; however,
[__]block_write_full_page() currently ignores the blkcg indicated by
inode and issues all bio's without explicit blkcg association.
This patch adds submit_bh_blkcg() which associates the bio with the
specified blkio cgroup before issuing and uses it in
__block_write_full_page() so that the issued bio's are associated with
inode_to_wb_blkcg_css(inode).
v2: Updated for per-inode wb association.
Signed-off-by: Tejun Heo <tj@kernel.org>
Cc: Jens Axboe <axboe@kernel.dk>
Cc: Jan Kara <jack@suse.cz>
Cc: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Jens Axboe <axboe@fb.com>
When modifying PG_Dirty on cached file pages, update the new
MEM_CGROUP_STAT_DIRTY counter. This is done in the same places where
global NR_FILE_DIRTY is managed. The new memcg stat is visible in the
per memcg memory.stat cgroupfs file. The most recent past attempt at
this was http://thread.gmane.org/gmane.linux.kernel.cgroups/8632
The new accounting supports future efforts to add per cgroup dirty
page throttling and writeback. It also helps an administrator break
down a container's memory usage and provides evidence to understand
memcg oom kills (the new dirty count is included in memcg oom kill
messages).
The ability to move page accounting between memcg
(memory.move_charge_at_immigrate) makes this accounting more
complicated than the global counter. The existing
mem_cgroup_{begin,end}_page_stat() lock is used to serialize move
accounting with stat updates.
Typical update operation:
memcg = mem_cgroup_begin_page_stat(page)
if (TestSetPageDirty()) {
[...]
mem_cgroup_update_page_stat(memcg)
}
mem_cgroup_end_page_stat(memcg)
Summary of mem_cgroup_end_page_stat() overhead:
- Without CONFIG_MEMCG it's a no-op
- With CONFIG_MEMCG and no inter memcg task movement, it's just
rcu_read_lock()
- With CONFIG_MEMCG and inter memcg task movement, it's
rcu_read_lock() + spin_lock_irqsave()
A memcg parameter is added to several routines because their callers
now grab mem_cgroup_begin_page_stat() which returns the memcg later
needed by for mem_cgroup_update_page_stat().
Because mem_cgroup_begin_page_stat() may disable interrupts, some
adjustments are needed:
- move __mark_inode_dirty() from __set_page_dirty() to its caller.
__mark_inode_dirty() locking does not want interrupts disabled.
- use spin_lock_irqsave(tree_lock) rather than spin_lock_irq() in
__delete_from_page_cache(), replace_page_cache_page(),
invalidate_complete_page2(), and __remove_mapping().
text data bss dec hex filename
8925147 1774832 1785856 12485835 be84cb vmlinux-!CONFIG_MEMCG-before
8925339 1774832 1785856 12486027 be858b vmlinux-!CONFIG_MEMCG-after
+192 text bytes
8965977 1784992 1785856 12536825 bf4bf9 vmlinux-CONFIG_MEMCG-before
8966750 1784992 1785856 12537598 bf4efe vmlinux-CONFIG_MEMCG-after
+773 text bytes
Performance tests run on v4.0-rc1-36-g4f671fe2f952. Lower is better for
all metrics, they're all wall clock or cycle counts. The read and write
fault benchmarks just measure fault time, they do not include I/O time.
* CONFIG_MEMCG not set:
baseline patched
kbuild 1m25.030000(+-0.088% 3 samples) 1m25.426667(+-0.120% 3 samples)
dd write 100 MiB 0.859211561 +-15.10% 0.874162885 +-15.03%
dd write 200 MiB 1.670653105 +-17.87% 1.669384764 +-11.99%
dd write 1000 MiB 8.434691190 +-14.15% 8.474733215 +-14.77%
read fault cycles 254.0(+-0.000% 10 samples) 253.0(+-0.000% 10 samples)
write fault cycles 2021.2(+-3.070% 10 samples) 1984.5(+-1.036% 10 samples)
* CONFIG_MEMCG=y root_memcg:
baseline patched
kbuild 1m25.716667(+-0.105% 3 samples) 1m25.686667(+-0.153% 3 samples)
dd write 100 MiB 0.855650830 +-14.90% 0.887557919 +-14.90%
dd write 200 MiB 1.688322953 +-12.72% 1.667682724 +-13.33%
dd write 1000 MiB 8.418601605 +-14.30% 8.673532299 +-15.00%
read fault cycles 266.0(+-0.000% 10 samples) 266.0(+-0.000% 10 samples)
write fault cycles 2051.7(+-1.349% 10 samples) 2049.6(+-1.686% 10 samples)
* CONFIG_MEMCG=y non-root_memcg:
baseline patched
kbuild 1m26.120000(+-0.273% 3 samples) 1m25.763333(+-0.127% 3 samples)
dd write 100 MiB 0.861723964 +-15.25% 0.818129350 +-14.82%
dd write 200 MiB 1.669887569 +-13.30% 1.698645885 +-13.27%
dd write 1000 MiB 8.383191730 +-14.65% 8.351742280 +-14.52%
read fault cycles 265.7(+-0.172% 10 samples) 267.0(+-0.000% 10 samples)
write fault cycles 2070.6(+-1.512% 10 samples) 2084.4(+-2.148% 10 samples)
As expected anon page faults are not affected by this patch.
tj: Updated to apply on top of the recent cancel_dirty_page() changes.
Signed-off-by: Sha Zhengju <handai.szj@gmail.com>
Signed-off-by: Greg Thelen <gthelen@google.com>
Signed-off-by: Tejun Heo <tj@kernel.org>
Signed-off-by: Jens Axboe <axboe@fb.com>
cancel_dirty_page() had some issues and b9ea25152e ("page_writeback:
clean up mess around cancel_dirty_page()") replaced it with
account_page_cleaned() which makes the caller responsible for clearing
the dirty bit; unfortunately, the planned changes for cgroup writeback
support requires synchronization between dirty bit manipulation and
stat updates. While we can open-code such synchronization in each
account_page_cleaned() callsite, that's gonna be unnecessarily awkward
and verbose.
This patch revives cancel_dirty_page() but in a more restricted form.
All it does is TestClearPageDirty() followed by account_page_cleaned()
invocation if the page was dirty. This helper covers all
account_page_cleaned() usages except for __delete_from_page_cache()
which is a special case anyway and left alone. As this leaves no
module user for account_page_cleaned(), EXPORT_SYMBOL() is dropped
from it.
This patch just revives cancel_dirty_page() as a trivial wrapper to
replace equivalent usages and doesn't introduce any functional
changes.
Signed-off-by: Tejun Heo <tj@kernel.org>
Cc: Konstantin Khlebnikov <khlebnikov@yandex-team.ru>
Signed-off-by: Jens Axboe <axboe@fb.com>
Since the big barrier rewrite/removal in 2007 we never fail FLUSH or
FUA requests, which means we can remove the magic BIO_EOPNOTSUPP flag
to help propagating those to the buffer_head layer.
Signed-off-by: Christoph Hellwig <hch@lst.de>
Reviewed-by: Jeff Moyer <jmoyer@redhat.com>
Acked-by: Tejun Heo <tj@kernel.org>
Signed-off-by: Jens Axboe <axboe@fb.com>
This patch replaces cancel_dirty_page() with a helper function
account_page_cleaned() which only updates counters. It's called from
truncate_complete_page() and from try_to_free_buffers() (hack for ext3).
Page is locked in both cases, page-lock protects against concurrent
dirtiers: see commit 2d6d7f9828 ("mm: protect set_page_dirty() from
ongoing truncation").
Delete_from_page_cache() shouldn't be called for dirty pages, they must
be handled by caller (either written or truncated). This patch treats
final dirty accounting fixup at the end of __delete_from_page_cache() as
a debug check and adds WARN_ON_ONCE() around it. If something removes
dirty pages without proper handling that might be a bug and unwritten
data might be lost.
Hugetlbfs has no dirty pages accounting, ClearPageDirty() is enough
here.
cancel_dirty_page() in nfs_wb_page_cancel() is redundant. This is
helper for nfs_invalidate_page() and it's called only in case complete
invalidation.
The mess was started in v2.6.20 after commits 46d2277c79 ("Clean up
and make try_to_free_buffers() not race with dirty pages") and
3e67c0987d ("truncate: clear page dirtiness before running
try_to_free_buffers()") first was reverted right in v2.6.20 in commit
ecdfc9787f ("Resurrect 'try_to_free_buffers()' VM hackery"), second in
v2.6.25 commit a2b345642f ("Fix dirty page accounting leak with ext3
data=journal").
Custom fixes were introduced between these points. NFS in v2.6.23, commit
1b3b4a1a2d ("NFS: Fix a write request leak in nfs_invalidate_page()").
Kludge in __delete_from_page_cache() in v2.6.24, commit 3a6927906f ("Do
dirty page accounting when removing a page from the page cache"). Since
v2.6.25 all of them are redundant.
[akpm@linux-foundation.org: coding-style fixes]
Signed-off-by: Konstantin Khlebnikov <khlebnikov@yandex-team.ru>
Cc: Tejun Heo <tj@kernel.org>
Cc: Jan Kara <jack@suse.cz>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
When quiet_error applies rate limiting to buffer_io_error calls, what the
they apply to is unclear because the name is so generic, particularly
if the messages are interleaved with others:
[ 1936.063572] quiet_error: 664293 callbacks suppressed
[ 1936.065297] Buffer I/O error on dev sdr, logical block 257429952, lost async page write
[ 1936.067814] Buffer I/O error on dev sdr, logical block 257429953, lost async page write
Also, the function uses printk_ratelimit(), although printk.h includes a
comment advising "Please don't use... Instead use printk_ratelimited()."
Change buffer_io_error to check the BH_Quiet bit itself, drop the
printk_ratelimit call, and print using printk_ratelimited.
This makes the messages look like:
[ 387.208839] buffer_io_error: 676394 callbacks suppressed
[ 387.210693] Buffer I/O error on dev sdr, logical block 211291776, lost async page write
[ 387.213432] Buffer I/O error on dev sdr, logical block 211291777, lost async page write
Signed-off-by: Robert Elliott <elliott@hp.com>
Reviewed-by: Webb Scales <webbnh@hp.com>
Signed-off-by: Jens Axboe <axboe@fb.com>
buffer.c uses two printk calls to print these messages:
[67353.422338] Buffer I/O error on device sdr, logical block 212868488
[67353.422338] lost page write due to I/O error on sdr
In a busy system, they may be interleaved with other prints,
losing the context for the second message. Merge them into
one line with one printk call so the prints are atomic.
Also, differentiate between async page writes, sync page writes, and
async page reads.
Also, shorten "device" to "dev" to match the block layer prints:
[67353.467906] blk_update_request: critical target error, dev sdr, sector
1707107328
Also, use %llu rather than %Lu.
Resulting prints look like:
[ 1356.437006] blk_update_request: critical target error, dev sdr, sector 1719693992
[ 1361.383522] quiet_error: 659876 callbacks suppressed
[ 1361.385816] Buffer I/O error on dev sdr, logical block 256902912, lost async page write
[ 1361.385819] Buffer I/O error on dev sdr, logical block 256903644, lost async page write
Signed-off-by: Robert Elliott <elliott@hp.com>
Reviewed-by: Webb Scales <webbnh@hp.com>
Signed-off-by: Jens Axboe <axboe@fb.com>
Pull ext4 updates from Ted Ts'o:
"A large number of cleanups and bug fixes, with some (minor) journal
optimizations"
[ This got sent to me before -rc1, but was stuck in my spam folder. - Linus ]
* tag 'ext4_for_linus' of git://git.kernel.org/pub/scm/linux/kernel/git/tytso/ext4: (67 commits)
ext4: check s_chksum_driver when looking for bg csum presence
ext4: move error report out of atomic context in ext4_init_block_bitmap()
ext4: Replace open coded mdata csum feature to helper function
ext4: delete useless comments about ext4_move_extents
ext4: fix reservation overflow in ext4_da_write_begin
ext4: add ext4_iget_normal() which is to be used for dir tree lookups
ext4: don't orphan or truncate the boot loader inode
ext4: grab missed write_count for EXT4_IOC_SWAP_BOOT
ext4: optimize block allocation on grow indepth
ext4: get rid of code duplication
ext4: fix over-defensive complaint after journal abort
ext4: fix return value of ext4_do_update_inode
ext4: fix mmap data corruption when blocksize < pagesize
vfs: fix data corruption when blocksize < pagesize for mmaped data
ext4: fold ext4_nojournal_sops into ext4_sops
ext4: support freezing ext2 (nojournal) file systems
ext4: fold ext4_sync_fs_nojournal() into ext4_sync_fs()
ext4: don't check quota format when there are no quota files
jbd2: simplify calling convention around __jbd2_journal_clean_checkpoint_list
jbd2: avoid pointless scanning of checkpoint lists
...
It's very common for the buffer heads in the lru to have different block
numbers. By comparing the blocknr before the bdev and size we can
reduce the cost of searching in the very common case where all the
entries have the same bdev and size.
In quick hot cache cycle counting tests on a single fs workstation this
cut the cost of a miss by about 20%.
A diff of the disassembly shows the reordering of the bdev and blocknr
comparisons. This is in such a tiny loop that skipping one comparison
is a meaningful portion of the total work being done:
1628: 83 c1 01 add $0x1,%ecx
162b: 83 f9 08 cmp $0x8,%ecx
162e: 74 60 je 1690 <__find_get_block+0xa0>
1630: 89 c8 mov %ecx,%eax
1632: 65 4c 8b 04 c5 00 00 mov %gs:0x0(,%rax,8),%r8
1639: 00 00
163b: 4d 85 c0 test %r8,%r8
163e: 4c 89 c3 mov %r8,%rbx
1641: 74 e5 je 1628 <__find_get_block+0x38>
- 1643: 4d 3b 68 30 cmp 0x30(%r8),%r13
+ 1643: 4d 3b 68 18 cmp 0x18(%r8),%r13
1647: 75 df jne 1628 <__find_get_block+0x38>
- 1649: 4d 3b 60 18 cmp 0x18(%r8),%r12
+ 1649: 4d 3b 60 30 cmp 0x30(%r8),%r12
164d: 75 d9 jne 1628 <__find_get_block+0x38>
164f: 49 39 50 20 cmp %rdx,0x20(%r8)
1653: 75 d3 jne 1628 <__find_get_block+0x38>
Signed-off-by: Zach Brown <zab@zabbo.net>
Cc: Al Viro <viro@zeniv.linux.org.uk>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Pull vfs updates from Al Viro:
"The big thing in this pile is Eric's unmount-on-rmdir series; we
finally have everything we need for that. The final piece of prereqs
is delayed mntput() - now filesystem shutdown always happens on
shallow stack.
Other than that, we have several new primitives for iov_iter (Matt
Wilcox, culled from his XIP-related series) pushing the conversion to
->read_iter()/ ->write_iter() a bit more, a bunch of fs/dcache.c
cleanups and fixes (including the external name refcounting, which
gives consistent behaviour of d_move() wrt procfs symlinks for long
and short names alike) and assorted cleanups and fixes all over the
place.
This is just the first pile; there's a lot of stuff from various
people that ought to go in this window. Starting with
unionmount/overlayfs mess... ;-/"
* 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/viro/vfs: (60 commits)
fs/file_table.c: Update alloc_file() comment
vfs: Deduplicate code shared by xattr system calls operating on paths
reiserfs: remove pointless forward declaration of struct nameidata
don't need that forward declaration of struct nameidata in dcache.h anymore
take dname_external() into fs/dcache.c
let path_init() failures treated the same way as subsequent link_path_walk()
fix misuses of f_count() in ppp and netlink
ncpfs: use list_for_each_entry() for d_subdirs walk
vfs: move getname() from callers to do_mount()
gfs2_atomic_open(): skip lookups on hashed dentry
[infiniband] remove pointless assignments
gadgetfs: saner API for gadgetfs_create_file()
f_fs: saner API for ffs_sb_create_file()
jfs: don't hash direct inode
[s390] remove pointless assignment of ->f_op in vmlogrdr ->open()
ecryptfs: ->f_op is never NULL
android: ->f_op is never NULL
nouveau: __iomem misannotations
missing annotation in fs/file.c
fs: namespace: suppress 'may be used uninitialized' warnings
...
Increase the buffer-head per-CPU LRU size to allow efficient filesystem
operations that access many blocks for each transaction. For example,
creating a file in a large ext4 directory with quota enabled will access
multiple buffer heads and will overflow the LRU at the default 8-block LRU
size:
* parent directory inode table block (ctime, nlinks for subdirs)
* new inode bitmap
* inode table block
* 2 quota blocks
* directory leaf block (not reused, but pollutes one cache entry)
* 2 levels htree blocks (only one is reused, other pollutes cache)
* 2 levels indirect/index blocks (only one is reused)
The buffer-head per-CPU LRU size is raised to 16, as it shows in metadata
performance benchmarks up to 10% gain for create, 4% for lookup and 7% for
destroy.
Signed-off-by: Liang Zhen <liang.zhen@intel.com>
Signed-off-by: Andreas Dilger <andreas.dilger@intel.com>
Signed-off-by: Sebastien Buisson <sebastien.buisson@bull.net>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Add guard_bio_eod() check for mpage code in order to allow us to do IO
even on the odd last sectors of a device, even if the block size is some
multiple of the physical sector size.
Using mpage_readpages() for block device requires this guard check.
Signed-off-by: Akinobu Mita <akinobu.mita@gmail.com>
Cc: Jens Axboe <axboe@kernel.dk>
Cc: Alexander Viro <viro@zeniv.linux.org.uk>
Cc: Jeff Moyer <jmoyer@redhat.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
This patchset implements readpages() operation for block device by using
mpage_readpages() which can create multipage BIOs instead of BIOs for each
page and reduce system CPU time consumption.
This patch (of 3):
guard_bh_eod() is used in submit_bh() to allow us to do IO even on the odd
last sectors of a device, even if the block size is some multiple of the
physical sector size. This makes guard_bh_eod() more generic and renames
it guard_bio_eod() so that we can use it without struct buffer_head
argument.
The reason for this change is that using mpage_readpages() for block
device requires to add this guard check in mpage code.
Signed-off-by: Akinobu Mita <akinobu.mita@gmail.com>
Cc: Jens Axboe <axboe@kernel.dk>
Cc: Alexander Viro <viro@zeniv.linux.org.uk>
Cc: Jeff Moyer <jmoyer@redhat.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
This patch makes it possible to kill a process looping in
cont_expand_zero. A process may spend a lot of time in this function, so
it is desirable to be able to kill it.
It happened to me that I wanted to copy a piece data from the disk to a
file. By mistake, I used the "seek" parameter to dd instead of "skip". Due
to the "seek" parameter, dd attempted to extend the file and became stuck
doing so - the only possibility was to reset the machine or wait many
hours until the filesystem runs out of space and cont_expand_zero fails.
We need this patch to be able to terminate the process.
Signed-off-by: Mikulas Patocka <mpatocka@redhat.com>
Cc: stable@vger.kernel.org
Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>
->page_mkwrite() is used by filesystems to allocate blocks under a page
which is becoming writeably mmapped in some process' address space. This
allows a filesystem to return a page fault if there is not enough space
available, user exceeds quota or similar problem happens, rather than
silently discarding data later when writepage is called.
However VFS fails to call ->page_mkwrite() in all the cases where
filesystems need it when blocksize < pagesize. For example when
blocksize = 1024, pagesize = 4096 the following is problematic:
ftruncate(fd, 0);
pwrite(fd, buf, 1024, 0);
map = mmap(NULL, 1024, PROT_WRITE, MAP_SHARED, fd, 0);
map[0] = 'a'; ----> page_mkwrite() for index 0 is called
ftruncate(fd, 10000); /* or even pwrite(fd, buf, 1, 10000) */
mremap(map, 1024, 10000, 0);
map[4095] = 'a'; ----> no page_mkwrite() called
At the moment ->page_mkwrite() is called, filesystem can allocate only
one block for the page because i_size == 1024. Otherwise it would create
blocks beyond i_size which is generally undesirable. But later at
->writepage() time, we also need to store data at offset 4095 but we
don't have block allocated for it.
This patch introduces a helper function filesystems can use to have
->page_mkwrite() called at all the necessary moments.
Signed-off-by: Jan Kara <jack@suse.cz>
Signed-off-by: Theodore Ts'o <tytso@mit.edu>
Cc: stable@vger.kernel.org
On 32-bit architectures, the legacy buffer_head functions are not always
handling the sector number with the proper 64-bit types, and will thus
fail on 4TB+ disks.
Any code that uses __getblk() (and thus bread(), breadahead(),
sb_bread(), sb_breadahead(), sb_getblk()), and calls it using a 64-bit
block on a 32-bit arch (where "long" is 32-bit) causes an inifinite loop
in __getblk_slow() with an infinite stream of errors logged to dmesg
like this:
__find_get_block_slow() failed. block=6740375944, b_blocknr=2445408648
b_state=0x00000020, b_size=512
device sda1 blocksize: 512
Note how in hex block is 0x191C1F988 and b_blocknr is 0x91C1F988 i.e. the
top 32-bits are missing (in this case the 0x1 at the top).
This is because grow_dev_page() is broken and has a 32-bit overflow due
to shifting the page index value (a pgoff_t - which is just 32 bits on
32-bit architectures) left-shifted as the block number. But the top
bits to get lost as the pgoff_t is not type cast to sector_t / 64-bit
before the shift.
This patch fixes this issue by type casting "index" to sector_t before
doing the left shift.
Note this is not a theoretical bug but has been seen in the field on a
4TiB hard drive with logical sector size 512 bytes.
This patch has been verified to fix the infinite loop problem on 3.17-rc5
kernel using a 4TB disk image mounted using "-o loop". Without this patch
doing a "find /nt" where /nt is an NTFS volume causes the inifinite loop
100% reproducibly whilst with the patch it works fine as expected.
Signed-off-by: Anton Altaparmakov <aia21@cantab.net>
Cc: stable@vger.kernel.org
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
A buffer cache is allocated from movable area because it is referred
for a while and released soon. But some filesystems are taking buffer
cache for a long time and it can disturb page migration.
New APIs are introduced to allocate buffer cache with user specific
flag. *_gfp APIs are for user want to set page allocation flag for
page cache allocation. And *_unmovable APIs are for the user wants to
allocate page cache from non-movable area.
Signed-off-by: Gioh Kim <gioh.kim@lge.com>
Signed-off-by: Theodore Ts'o <tytso@mit.edu>
Reviewed-by: Jan Kara <jack@suse.cz>
The current "wait_on_bit" interface requires an 'action'
function to be provided which does the actual waiting.
There are over 20 such functions, many of them identical.
Most cases can be satisfied by one of just two functions, one
which uses io_schedule() and one which just uses schedule().
So:
Rename wait_on_bit and wait_on_bit_lock to
wait_on_bit_action and wait_on_bit_lock_action
to make it explicit that they need an action function.
Introduce new wait_on_bit{,_lock} and wait_on_bit{,_lock}_io
which are *not* given an action function but implicitly use
a standard one.
The decision to error-out if a signal is pending is now made
based on the 'mode' argument rather than being encoded in the action
function.
All instances of the old wait_on_bit and wait_on_bit_lock which
can use the new version have been changed accordingly and their
action functions have been discarded.
wait_on_bit{_lock} does not return any specific error code in the
event of a signal so the caller must check for non-zero and
interpolate their own error code as appropriate.
The wait_on_bit() call in __fscache_wait_on_invalidate() was
ambiguous as it specified TASK_UNINTERRUPTIBLE but used
fscache_wait_bit_interruptible as an action function.
David Howells confirms this should be uniformly
"uninterruptible"
The main remaining user of wait_on_bit{,_lock}_action is NFS
which needs to use a freezer-aware schedule() call.
A comment in fs/gfs2/glock.c notes that having multiple 'action'
functions is useful as they display differently in the 'wchan'
field of 'ps'. (and /proc/$PID/wchan).
As the new bit_wait{,_io} functions are tagged "__sched", they
will not show up at all, but something higher in the stack. So
the distinction will still be visible, only with different
function names (gds2_glock_wait versus gfs2_glock_dq_wait in the
gfs2/glock.c case).
Since first version of this patch (against 3.15) two new action
functions appeared, on in NFS and one in CIFS. CIFS also now
uses an action function that makes the same freezer aware
schedule call as NFS.
Signed-off-by: NeilBrown <neilb@suse.de>
Acked-by: David Howells <dhowells@redhat.com> (fscache, keys)
Acked-by: Steven Whitehouse <swhiteho@redhat.com> (gfs2)
Acked-by: Peter Zijlstra <peterz@infradead.org>
Cc: Oleg Nesterov <oleg@redhat.com>
Cc: Steve French <sfrench@samba.org>
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Link: http://lkml.kernel.org/r/20140707051603.28027.72349.stgit@notabene.brown
Signed-off-by: Ingo Molnar <mingo@kernel.org>
aops->write_begin may allocate a new page and make it visible only to have
mark_page_accessed called almost immediately after. Once the page is
visible the atomic operations are necessary which is noticable overhead
when writing to an in-memory filesystem like tmpfs but should also be
noticable with fast storage. The objective of the patch is to initialse
the accessed information with non-atomic operations before the page is
visible.
The bulk of filesystems directly or indirectly use
grab_cache_page_write_begin or find_or_create_page for the initial
allocation of a page cache page. This patch adds an init_page_accessed()
helper which behaves like the first call to mark_page_accessed() but may
called before the page is visible and can be done non-atomically.
The primary APIs of concern in this care are the following and are used
by most filesystems.
find_get_page
find_lock_page
find_or_create_page
grab_cache_page_nowait
grab_cache_page_write_begin
All of them are very similar in detail to the patch creates a core helper
pagecache_get_page() which takes a flags parameter that affects its
behavior such as whether the page should be marked accessed or not. Then
old API is preserved but is basically a thin wrapper around this core
function.
Each of the filesystems are then updated to avoid calling
mark_page_accessed when it is known that the VM interfaces have already
done the job. There is a slight snag in that the timing of the
mark_page_accessed() has now changed so in rare cases it's possible a page
gets to the end of the LRU as PageReferenced where as previously it might
have been repromoted. This is expected to be rare but it's worth the
filesystem people thinking about it in case they see a problem with the
timing change. It is also the case that some filesystems may be marking
pages accessed that previously did not but it makes sense that filesystems
have consistent behaviour in this regard.
The test case used to evaulate this is a simple dd of a large file done
multiple times with the file deleted on each iterations. The size of the
file is 1/10th physical memory to avoid dirty page balancing. In the
async case it will be possible that the workload completes without even
hitting the disk and will have variable results but highlight the impact
of mark_page_accessed for async IO. The sync results are expected to be
more stable. The exception is tmpfs where the normal case is for the "IO"
to not hit the disk.
The test machine was single socket and UMA to avoid any scheduling or NUMA
artifacts. Throughput and wall times are presented for sync IO, only wall
times are shown for async as the granularity reported by dd and the
variability is unsuitable for comparison. As async results were variable
do to writback timings, I'm only reporting the maximum figures. The sync
results were stable enough to make the mean and stddev uninteresting.
The performance results are reported based on a run with no profiling.
Profile data is based on a separate run with oprofile running.
async dd
3.15.0-rc3 3.15.0-rc3
vanilla accessed-v2
ext3 Max elapsed 13.9900 ( 0.00%) 11.5900 ( 17.16%)
tmpfs Max elapsed 0.5100 ( 0.00%) 0.4900 ( 3.92%)
btrfs Max elapsed 12.8100 ( 0.00%) 12.7800 ( 0.23%)
ext4 Max elapsed 18.6000 ( 0.00%) 13.3400 ( 28.28%)
xfs Max elapsed 12.5600 ( 0.00%) 2.0900 ( 83.36%)
The XFS figure is a bit strange as it managed to avoid a worst case by
sheer luck but the average figures looked reasonable.
samples percentage
ext3 86107 0.9783 vmlinux-3.15.0-rc4-vanilla mark_page_accessed
ext3 23833 0.2710 vmlinux-3.15.0-rc4-accessed-v3r25 mark_page_accessed
ext3 5036 0.0573 vmlinux-3.15.0-rc4-accessed-v3r25 init_page_accessed
ext4 64566 0.8961 vmlinux-3.15.0-rc4-vanilla mark_page_accessed
ext4 5322 0.0713 vmlinux-3.15.0-rc4-accessed-v3r25 mark_page_accessed
ext4 2869 0.0384 vmlinux-3.15.0-rc4-accessed-v3r25 init_page_accessed
xfs 62126 1.7675 vmlinux-3.15.0-rc4-vanilla mark_page_accessed
xfs 1904 0.0554 vmlinux-3.15.0-rc4-accessed-v3r25 init_page_accessed
xfs 103 0.0030 vmlinux-3.15.0-rc4-accessed-v3r25 mark_page_accessed
btrfs 10655 0.1338 vmlinux-3.15.0-rc4-vanilla mark_page_accessed
btrfs 2020 0.0273 vmlinux-3.15.0-rc4-accessed-v3r25 init_page_accessed
btrfs 587 0.0079 vmlinux-3.15.0-rc4-accessed-v3r25 mark_page_accessed
tmpfs 59562 3.2628 vmlinux-3.15.0-rc4-vanilla mark_page_accessed
tmpfs 1210 0.0696 vmlinux-3.15.0-rc4-accessed-v3r25 init_page_accessed
tmpfs 94 0.0054 vmlinux-3.15.0-rc4-accessed-v3r25 mark_page_accessed
[akpm@linux-foundation.org: don't run init_page_accessed() against an uninitialised pointer]
Signed-off-by: Mel Gorman <mgorman@suse.de>
Cc: Johannes Weiner <hannes@cmpxchg.org>
Cc: Vlastimil Babka <vbabka@suse.cz>
Cc: Jan Kara <jack@suse.cz>
Cc: Michal Hocko <mhocko@suse.cz>
Cc: Hugh Dickins <hughd@google.com>
Cc: Dave Hansen <dave.hansen@intel.com>
Cc: Theodore Ts'o <tytso@mit.edu>
Cc: "Paul E. McKenney" <paulmck@linux.vnet.ibm.com>
Cc: Oleg Nesterov <oleg@redhat.com>
Cc: Rik van Riel <riel@redhat.com>
Cc: Peter Zijlstra <peterz@infradead.org>
Tested-by: Prabhakar Lad <prabhakar.csengg@gmail.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Discarding buffers uses a bunch of atomic operations when discarding
buffers because ...... I can't think of a reason. Use a cmpxchg loop to
clear all the necessary flags. In most (all?) cases this will be a single
atomic operations.
[akpm@linux-foundation.org: move BUFFER_FLAGS_DISCARD into the .c file]
Signed-off-by: Mel Gorman <mgorman@suse.de>
Cc: Johannes Weiner <hannes@cmpxchg.org>
Cc: Vlastimil Babka <vbabka@suse.cz>
Cc: Jan Kara <jack@suse.cz>
Cc: Michal Hocko <mhocko@suse.cz>
Cc: Hugh Dickins <hughd@google.com>
Cc: Dave Hansen <dave.hansen@intel.com>
Cc: Theodore Ts'o <tytso@mit.edu>
Cc: "Paul E. McKenney" <paulmck@linux.vnet.ibm.com>
Cc: Oleg Nesterov <oleg@redhat.com>
Cc: Rik van Riel <riel@redhat.com>
Cc: Peter Zijlstra <peterz@infradead.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
The last in-tree caller of block_write_full_page_endio() was removed in
January 2013. It's time to remove the EXPORT_SYMBOL, which leaves
block_write_full_page() as the only caller of
block_write_full_page_endio(), so inline block_write_full_page_endio()
into block_write_full_page().
Signed-off-by: Matthew Wilcox <matthew.r.wilcox@intel.com>
Cc: Hugh Dickins <hughd@google.com>
Cc: Dave Chinner <david@fromorbit.com>
Cc: Dheeraj Reddy <dheeraj.reddy@intel.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Jens Axboe
Tejun Heo
Greg Thelen
Julia Lawall
Christoph Hellwig
Konstantin Khlebnikov
Robert Elliott
Linus Torvalds
Zach Brown
Sebastien Buisson
Akinobu Mita
Mikulas Patocka
Jan Kara
Anton Altaparmakov
Gioh Kim
NeilBrown
Mel Gorman
Matthew Wilcox