android/android_kernel_xiaomi_sm8450
1
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity
fc0561cefc
android_kernel_xiaomi_sm8450/fs
History
Dave Chinner fc0561cefc xfs: optimise away log forces on timestamp updates for fdatasync 
xfs: timestamp updates cause excessive fdatasync log traffic

Sage Weil reported that a ceph test workload was writing to the
log on every fdatasync during an overwrite workload. Event tracing
showed that the only metadata modification being made was the
timestamp updates during the write(2) syscall, but fdatasync(2)
is supposed to ignore them. The key observation was that the
transactions in the log all looked like this:

INODE: #regs: 4   ino: 0x8b  flags: 0x45   dsize: 32

And contained a flags field of 0x45 or 0x85, and had data and
attribute forks following the inode core. This means that the
timestamp updates were triggering dirty relogging of previously
logged parts of the inode that hadn't yet been flushed back to
disk.

There are two parts to this problem. The first is that XFS relogs
dirty regions in subsequent transactions, so it carries around the
fields that have been dirtied since the last time the inode was
written back to disk, not since the last time the inode was forced
into the log.

The second part is that on v5 filesystems, the inode change count
update during inode dirtying also sets the XFS_ILOG_CORE flag, so
on v5 filesystems this makes a timestamp update dirty the entire
inode.

As a result when fdatasync is run, it looks at the dirty fields in
the inode, and sees more than just the timestamp flag, even though
the only metadata change since the last fdatasync was just the
timestamps. Hence we force the log on every subsequent fdatasync
even though it is not needed.

To fix this, add a new field to the inode log item that tracks
changes since the last time fsync/fdatasync forced the log to flush
the changes to the journal. This flag is updated when we dirty the
inode, but we do it before updating the change count so it does not
carry the "core dirty" flag from timestamp updates. The fields are
zeroed when the inode is marked clean (due to writeback/freeing) or
when an fsync/datasync forces the log. Hence if we only dirty the
timestamps on the inode between fsync/fdatasync calls, the fdatasync
will not trigger another log force.

Over 100 runs of the test program:

Ext4 baseline:
	runtime: 1.63s +/- 0.24s
	avg lat: 1.59ms +/- 0.24ms
	iops: ~2000

XFS, vanilla kernel:
        runtime: 2.45s +/- 0.18s
	avg lat: 2.39ms +/- 0.18ms
	log forces: ~400/s
	iops: ~1000

XFS, patched kernel:
        runtime: 1.49s +/- 0.26s
	avg lat: 1.46ms +/- 0.25ms
	log forces: ~30/s
	iops: ~1500

Reported-by: Sage Weil <sage@redhat.com>
Signed-off-by: Dave Chinner <dchinner@redhat.com>
Reviewed-by: Brian Foster <bfoster@redhat.com>
Signed-off-by: Dave Chinner <david@fromorbit.com>
2015-11-03 13:14:59 +11:00
..
9p 9p: fix return code of read() when count is 0 2015-08-23 14:21:36 -05:00
adfs fs/adfs: remove unneeded cast 2015-06-30 19:44:57 -07:00
affs fs/affs: make root lookup from blkdev logical size 2015-09-10 13:29:01 -07:00
afs net: Add a struct net parameter to sock_create_kern 2015-05-11 10:50:17 -04:00
autofs4 make simple_positive() public 2015-06-23 18:02:01 -04:00
befs Merge branch 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/viro/vfs 2015-07-04 19:36:06 -07:00
bfs Merge branch 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/viro/vfs 2015-04-26 17:22:07 -07:00
btrfs Merge branch 'for-linus-4.3' of git://git.kernel.org/pub/scm/linux/kernel/git/mason/linux-btrfs 2015-09-11 12:38:25 -07:00
cachefiles Merge branch 'fscache-fixes' into for-next 2015-06-23 18:01:30 -04:00
ceph Merge branch 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/sage/ceph-client 2015-09-11 12:33:03 -07:00
cifs [CIFS] mount option sec=none not displayed properly in /proc/mounts 2015-09-11 19:37:06 -05:00
coda fs/coda: fix readlink buffer overflow 2015-09-10 13:29:01 -07:00
configfs configfs: fix kernel infoleak through user-controlled format string 2015-07-17 16:39:53 -07:00
cramfs fs/cramfs/inode.c: use linux/uaccess.h 2014-08-08 15:57:25 -07:00
debugfs debugfs: Export bool read/write functions 2015-07-20 18:44:50 +01:00
devpts devpts: if initialization failed, don't crash when opening /dev/ptmx 2015-06-30 19:44:58 -07:00
dlm dlm for 4.3 2015-09-03 12:57:48 -07:00
ecryptfs Invalidate stale eCryptfs dcache entries caused by unlinked lower inodes 2015-09-08 11:26:17 -07:00
efivarfs Merge branch 'x86-urgent-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/tip 2015-05-06 10:57:37 -07:00
efs fs/efs: femove unneeded cast 2015-06-25 17:00:42 -07:00
exofs pagemap.h: move dir_pages() over there 2015-06-23 18:02:00 -04:00
exportfs VFS: (Scripted) Convert S_ISLNK/DIR/REG(dentry->d_inode) to d_is_*(dentry) 2015-02-22 11:38:41 -05:00
ext2 ext2: huge page fault support 2015-09-08 15:35:28 -07:00
ext4 ext4: start transaction before calling into DAX 2015-09-08 15:35:28 -07:00
f2fs Merge tag 'for-f2fs-4.3' of git://git.kernel.org/pub/scm/linux/kernel/git/jaegeuk/f2fs 2015-09-03 13:10:22 -07:00
fat writeback: separate out include/linux/backing-dev-defs.h 2015-06-02 08:33:34 -06:00
freevxfs freevxfs: Grammar s/an negative/a negative/ 2015-08-07 13:59:24 +02:00
fscache FS-Cache: Retain the netfs context in the retrieval op earlier 2015-04-02 14:28:53 +01:00
fuse fs/fuse: fix ioctl type confusion 2015-08-16 12:35:44 -07:00
gfs2 GFS2: merge window 2015-09-11 12:23:51 -07:00
hfs hfs: fix B-tree corruption after insertion at position 0 2015-09-10 13:29:01 -07:00
hfsplus hfs,hfsplus: cache pages correctly between bnode_create and bnode_free 2015-09-10 13:29:01 -07:00
hostfs fs: create and use seq_show_option for escaping 2015-09-04 16:54:41 -07:00
hpfs hpfs: update ctime and mtime on directory modification 2015-09-03 11:55:30 -07:00
hugetlbfs hugetlbfs: add hugetlbfs_fallocate() 2015-09-08 15:35:28 -07:00
isofs VFS: normal filesystems (and lustre): d_inode() annotations 2015-04-15 15:06:57 -04:00
jbd2 jbd2: limit number of reserved credits 2015-08-04 11:21:52 -04:00
jffs2 Merge branch 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/viro/vfs 2015-07-04 19:36:06 -07:00
jfs Merge branch 'for_linus' of git://git.kernel.org/pub/scm/linux/kernel/git/jack/linux-fs 2015-09-03 12:28:30 -07:00
kernfs kernfs: implement kernfs_path_len() 2015-08-18 15:49:15 -07:00
lockd lockd: NLM grace period shouldn't block NFSv4 opens 2015-08-13 10:22:06 -04:00
logfs block: remove bio_get_nr_vecs() 2015-08-13 12:32:04 -06:00
minix Merge branch 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/viro/vfs 2015-07-04 19:36:06 -07:00
ncpfs ncpfs: successful rename() should invalidate caches for parents 2015-06-14 11:31:39 -04:00
nfs NFS client updates for Linux 4.3 2015-09-07 14:02:24 -07:00
nfs_common lockd: NLM grace period shouldn't block NFSv4 opens 2015-08-13 10:22:06 -04:00
nfsd NFS client updates for Linux 4.3 2015-09-07 14:02:24 -07:00
nilfs2 block: remove bio_get_nr_vecs() 2015-08-13 12:32:04 -06:00
nls
notify Merge branch 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/viro/vfs 2015-09-05 20:34:28 -07:00
ntfs ntfs: delete unnecessary checks before calling iput() 2015-09-04 16:54:41 -07:00
ocfs2 revert "ocfs2/dlm: use list_for_each_entry instead of list_for_each" 2015-09-11 15:21:34 -07:00
omfs omfs: fix potential integer overflow in allocator 2015-05-28 18:25:19 -07:00
openpromfs
overlayfs fs: create and use seq_show_option for escaping 2015-09-04 16:54:41 -07:00
proc proc: convert to kstrto*()/kstrto*_from_user() 2015-09-10 13:29:01 -07:00
pstore Merge branch 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/ebiederm/user-namespace 2015-07-03 15:20:57 -07:00
qnx4
qnx6 pagemap.h: move dir_pages() over there 2015-06-23 18:02:00 -04:00
quota Merge branch 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/viro/vfs 2015-09-05 20:34:28 -07:00
ramfs VFS: normal filesystems (and lustre): d_inode() annotations 2015-04-15 15:06:57 -04:00
reiserfs fs: create and use seq_show_option for escaping 2015-09-04 16:54:41 -07:00
romfs make new_sync_{read,write}() static 2015-04-11 22:29:40 -04:00
squashfs fs: cleanup slight list_entry abuse 2015-06-23 18:01:59 -04:00
sysfs vfs: Commit to never having exectuables on proc and sysfs. 2015-07-10 10:39:25 -05:00
sysv pagemap.h: move dir_pages() over there 2015-06-23 18:02:00 -04:00
tracefs Merge branch 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/viro/vfs 2015-07-04 19:36:06 -07:00
ubifs This pull request includes the following UBI/UBIFS changes: 2015-06-25 14:11:34 -07:00
udf udf: Don't modify filesystem for read-only mounts 2015-08-20 14:58:35 +02:00
ufs fix ufs write vs readpage race when writing into a hole 2015-09-09 10:43:12 -07:00
xfs xfs: optimise away log forces on timestamp updates for fdatasync 2015-11-03 13:14:59 +11:00
aio.c mm: move ->mremap() from file_operations to vm_operations_struct 2015-09-04 16:54:41 -07:00
anon_inodes.c
attr.c fs,userns: Change inode_capable to capable_wrt_inode_uidgid 2014-06-10 13:57:22 -07:00
bad_inode.c don't bother with most of the bad_file_ops methods 2015-02-20 04:03:58 -05:00
binfmt_aout.c assorted conversions to %p[dD] 2014-11-19 13:01:20 -05:00
binfmt_elf_fdpic.c handle suicide on late failure exits in execve() in search_binary_handler() 2014-10-09 02:39:00 -04:00
binfmt_elf.c Merge branch 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/viro/vfs 2015-07-04 19:36:06 -07:00
binfmt_em86.c syscalls: implement execveat() system call 2014-12-13 12:42:51 -08:00
binfmt_flat.c fs/binfmt_flat.c: make old_reloc() static 2014-06-04 16:54:21 -07:00
binfmt_misc.c Merge branch 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/viro/vfs 2015-04-26 17:22:07 -07:00
binfmt_script.c syscalls: implement execveat() system call 2014-12-13 12:42:51 -08:00
block_dev.c blockdev: don't set S_DAX for misaligned partitions 2015-09-15 20:08:05 -04:00
buffer.c fs: use helper bio_add_page() instead of open coding on bi_io_vec 2015-08-13 12:32:00 -06:00
char_dev.c fs/char_dev.c: fix incorrect documentation for unregister_chrdev_region 2015-08-05 13:49:35 -07:00
compat_binfmt_elf.c
compat_ioctl.c ioctl_compat: handle FITRIM 2015-07-09 11:42:21 -07:00
compat.c vfs: make first argument of dir_context.actor typed 2014-10-31 17:48:54 -04:00
coredump.c fs: Don't dump core if the corefile would become world-readable. 2015-09-10 13:29:01 -07:00
dax.c dax: fix O_DIRECT I/O to the last block of a blockdev 2015-09-15 20:07:35 -04:00
dcache.c dcache: Reduce the scope of i_lock in d_splice_alias 2015-08-21 02:34:37 -04:00
dcookies.c
direct-io.c block: remove bio_get_nr_vecs() 2015-08-13 12:32:04 -06:00
drop_caches.c inode: convert inode_sb_list_lock to per-sb 2015-08-17 18:39:46 -04:00
eventfd.c eventfd: don't take the spinlock in eventfd_poll 2015-02-17 14:34:52 -08:00
eventpoll.c epoll: optimize setting task running after blocking 2015-02-13 21:21:40 -08:00
exec.c vfs: Commit to never having exectuables on proc and sysfs. 2015-07-10 10:39:25 -05:00
fcntl.c vfs: renumber FMODE_NONOTIFY and add to uniqueness check 2015-01-08 15:10:52 -08:00
fhandle.c vfs: read file_handle only once in handle_to_path 2015-06-02 10:29:07 -07:00
file_table.c fs, file table: reinit files_stat.max_files after deferred memory initialisation 2015-08-07 04:39:40 +03:00
file.c fs/file.c: __fget() and dup2() atomicity rules 2015-07-01 02:31:08 -04:00
filesystems.c
fs_pin.c fs_pin: Allow for the possibility that m_list or s_list go unused. 2015-04-09 11:39:55 -05:00
fs_struct.c
fs-writeback.c fs-writeback: unplug before cond_resched in writeback_sb_inodes 2015-09-19 18:50:19 -07:00
inode.c inode: don't softlockup when evicting inodes 2015-08-18 10:20:09 -07:00
internal.h inode: rename i_wb_list to i_io_list 2015-08-17 23:38:10 -04:00
ioctl.c fsioctl.c: make generic_block_fiemap() signal-tolerant 2015-02-10 14:30:30 -08:00
Kconfig fs: Remove ext3 filesystem driver 2015-07-23 20:59:40 +02:00
Kconfig.binfmt mm: split ET_DYN ASLR from mmap ASLR 2015-04-14 16:49:05 -07:00
libfs.c fs: Set the size of empty dirs to 0. 2015-08-12 15:28:45 -05:00
locks.c fs: fix fs/locks.c kernel-doc warning 2015-08-31 16:27:25 -04:00
Makefile userfaultfd: buildsystem activation 2015-09-04 16:54:41 -07:00
mbcache.c fs/mbcache: replace __builtin_log2() with ilog2() 2014-06-25 22:08:29 -04:00
mount.h fs: use seq_open_private() for proc_mounts 2015-06-30 19:44:56 -07:00
mpage.c block: remove bio_get_nr_vecs() 2015-08-13 12:32:04 -06:00
namei.c namei: fix warning while make xmldocs caused by namei.c 2015-09-10 13:29:01 -07:00
namespace.c Merge branch 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/ebiederm/user-namespace 2015-09-01 16:13:25 -07:00
no-block.c
nsfs.c fs/seq_file: convert int seq_vprint/seq_printf/etc... returns to void 2015-09-11 15:21:34 -07:00
open.c vfs: Commit to never having exectuables on proc and sysfs. 2015-07-10 10:39:25 -05:00
pipe.c VFS: assorted weird filesystems: d_inode() annotations 2015-04-15 15:06:58 -04:00
pnode.c mnt: Don't propagate unmounts to locked mounts 2015-04-02 20:34:20 -05:00
pnode.h mnt: Clarify and correct the disconnect logic in umount_tree 2015-07-22 20:33:27 -05:00
posix_acl.c fs/posix_acl.c: make posix_acl_create() safer and cleaner 2015-06-23 18:01:07 -04:00
proc_namespace.c fs: use seq_open_private() for proc_mounts 2015-06-30 19:44:56 -07:00
read_write.c new_sync_write(): discard ->ki_pos unless the return value is positive 2015-04-11 22:29:46 -04:00
readdir.c vfs: make first argument of dir_context.actor typed 2014-10-31 17:48:54 -04:00
select.c locking/arch: Rename set_mb() to smp_store_mb() 2015-05-19 08:32:00 +02:00
seq_file.c fs/seq_file: convert int seq_vprint/seq_printf/etc... returns to void 2015-09-11 15:21:34 -07:00
signalfd.c signalfd: fix information leak in signalfd_copyinfo 2015-08-07 04:39:40 +03:00
splice.c Merge branch 'akpm' (patches from Andrew) 2015-06-24 20:47:21 -07:00
stack.c fs: fix comment for 'CONFIG_LBADF' 2014-08-26 09:35:56 +02:00
stat.c VFS: assorted d_backing_inode() annotations 2015-04-15 15:06:59 -04:00
statfs.c
super.c Merge branch 'superblock-scaling' of git://git.kernel.org/pub/scm/linux/kernel/git/josef/btrfs-next into for-next 2015-08-21 02:31:20 -04:00
sync.c vfs: add support for a lazytime mount option 2015-02-05 02:45:00 -05:00
timerfd.c fs: Convert show_fdinfo functions to void 2014-11-05 14:13:23 -05:00
userfaultfd.c userfaultfd: add missing mmput() in error path 2015-09-17 21:16:07 -07:00
utimes.c
xattr.c evm: fix potential race when removing xattrs 2015-05-21 13:28:47 -04:00