AuxXxilium/linux_dsm_epyc7002
1
0
Fork 0
mirror of https://github.com/AuxXxilium/linux_dsm_epyc7002.git synced 2025-01-24 10:39:26 +07:00
Code Issues Actions Packages Projects Releases Wiki Activity
99d263d4c5
linux_dsm_epyc7002/fs
History
Linus Torvalds 99d263d4c5 vfs: fix bad hashing of dentries 
Josef Bacik found a performance regression between 3.2 and 3.10 and
narrowed it down to commit bfcfaa77bd ("vfs: use 'unsigned long'
accesses for dcache name comparison and hashing"). He reports:

 "The test case is essentially

      for (i = 0; i < 1000000; i++)
              mkdir("a$i");

  On xfs on a fio card this goes at about 20k dir/sec with 3.2, and 12k
  dir/sec with 3.10.  This is because we spend waaaaay more time in
  __d_lookup on 3.10 than in 3.2.

  The new hashing function for strings is suboptimal for <
  sizeof(unsigned long) string names (and hell even > sizeof(unsigned
  long) string names that I've tested).  I broke out the old hashing
  function and the new one into a userspace helper to get real numbers
  and this is what I'm getting:

      Old hash table had 1000000 entries, 0 dupes, 0 max dupes
      New hash table had 12628 entries, 987372 dupes, 900 max dupes
      We had 11400 buckets with a p50 of 30 dupes, p90 of 240 dupes, p99 of 567 dupes for the new hash

  My test does the hash, and then does the d_hash into a integer pointer
  array the same size as the dentry hash table on my system, and then
  just increments the value at the address we got to see how many
  entries we overlap with.

  As you can see the old hash function ended up with all 1 million
  entries in their own bucket, whereas the new one they are only
  distributed among ~12.5k buckets, which is why we're using so much
  more CPU in __d_lookup".

The reason for this hash regression is two-fold:

 - On 64-bit architectures the down-mixing of the original 64-bit
   word-at-a-time hash into the final 32-bit hash value is very
   simplistic and suboptimal, and just adds the two 32-bit parts
   together.

   In particular, because there is no bit shuffling and the mixing
   boundary is also a byte boundary, similar character patterns in the
   low and high word easily end up just canceling each other out.

 - the old byte-at-a-time hash mixed each byte into the final hash as it
   hashed the path component name, resulting in the low bits of the hash
   generally being a good source of hash data.  That is not true for the
   word-at-a-time case, and the hash data is distributed among all the
   bits.

The fix is the same in both cases: do a better job of mixing the bits up
and using as much of the hash data as possible.  We already have the
"hash_32|64()" functions to do that.

Reported-by: Josef Bacik <jbacik@fb.com>
Cc: Al Viro <viro@zeniv.linux.org.uk>
Cc: Christoph Hellwig <hch@infradead.org>
Cc: Chris Mason <clm@fb.com>
Cc: linux-fsdevel@vger.kernel.org
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2014-09-13 11:30:10 -07:00
..
9p Merge branch 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/viro/vfs 2014-06-12 10:30:18 -07:00
adfs adfs: add __printf verification, fix format/argument mismatches 2014-08-08 15:57:24 -07:00
affs Merge branch 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/viro/vfs 2014-06-12 10:30:18 -07:00
afs AFS: Correctly assemble the client UUID 2014-07-29 10:14:36 -07:00
autofs4 autofs4: comment typo: remove a a doubled word 2014-08-08 15:57:19 -07:00
befs fs/befs/linuxvfs.c: check superblock before dump operation 2014-08-08 15:57:20 -07:00
bfs fs/bfs: use bfs prefix for dump_imap 2014-08-08 15:57:24 -07:00
btrfs Merge branch 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/mason/linux-btrfs 2014-09-12 11:53:30 -07:00
cachefiles fs/cachefiles: replace kerror by pr_err 2014-06-06 16:08:14 -07:00
ceph Merge branch 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/sage/ceph-client 2014-08-13 17:43:29 -06:00
cifs Clarify Kconfig help text for CIFS and SMB2/SMB3 2014-08-25 17:01:05 -05:00
coda fs/coda: use linux/uaccess.h 2014-08-08 15:57:20 -07:00
configfs fs/configfs: use pr_fmt 2014-06-04 16:53:53 -07:00
cramfs fs/cramfs/inode.c: use linux/uaccess.h 2014-08-08 15:57:25 -07:00
debugfs fs: debugfs: remove trailing whitespace 2014-07-09 16:58:21 -07:00
devpts fs/devpts/inode.c: convert printk to pr_foo() 2014-06-06 16:08:14 -07:00
dlm fs/dlm/debug_fs.c: remove unnecessary null test before debugfs_remove 2014-08-08 15:57:27 -07:00
ecryptfs write_iter variants of {__,}generic_file_aio_write() 2014-05-06 17:38:00 -04:00
efivarfs fs/efivarfs/super.c: use static const for dentry_operations 2014-06-04 16:54:14 -07:00
efs fs/efs/namei.c: return is not a function 2014-08-08 15:57:18 -07:00
exofs fs/exofs/ore_raid.c: replace count*size kzalloc by kcalloc 2014-08-08 15:57:24 -07:00
exportfs fs/exportfs/expfs.c: kernel-doc warning fixes 2014-06-04 16:54:14 -07:00
ext2 fs/ext2/super.c: Drop memory allocation cast 2014-07-15 22:40:22 +02:00
ext3 ext3: Count internal journal as bsddf overhead in ext3_statfs 2014-08-19 23:16:51 +02:00
ext4 ext4: avoid trying to kfree an ERR_PTR pointer 2014-09-03 09:37:30 -04:00
f2fs f2fs: reposition unlock_new_inode to prevent accessing invalid inode 2014-09-02 00:22:24 -07:00
fat Merge branch 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/viro/vfs 2014-06-12 10:30:18 -07:00
freevxfs Major changes for 3.14 include support for the newly added ZERO_RANGE 2014-04-04 15:39:39 -07:00
fscache fs/fscache: make ctl_table static 2014-08-06 18:01:12 -07:00
fuse switch iov_iter_get_pages() to passing maximal number of pages 2014-08-07 14:40:11 -04:00
gfs2 Merge branch 'sched/urgent' into sched/core, to merge fixes before applying new changes 2014-07-28 10:03:00 +02:00
hfs write_iter variants of {__,}generic_file_aio_write() 2014-05-06 17:38:00 -04:00
hfsplus Merge branch 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/viro/vfs 2014-06-12 10:30:18 -07:00
hostfs hostfs: support rename flags 2014-08-07 14:40:09 -04:00
hpfs fs/hpfs/dnode.c: fix suspect code indent 2014-08-08 15:57:22 -07:00
hppfs
hugetlbfs fs/hugetlbfs/inode.c: remove null test before kfree 2014-06-04 16:54:11 -07:00
isofs isofs: Fix unbounded recursion when processing relocated directories 2014-08-19 18:29:30 +02:00
jbd fs/jbd/revoke.c: replace shift loop by ilog2 2014-05-21 10:26:13 +02:00
jbd2 jbd2: fix descriptor block size handling errors with journal_csum 2014-08-28 22:22:29 -04:00
jffs2 MTD updates for 3.17-rc1 2014-08-08 18:13:21 -07:00
jfs Merge branch 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/viro/vfs 2014-06-12 10:30:18 -07:00
kernfs Merge 3.16-rc6 into driver-core-next 2014-07-21 10:07:25 -07:00
lockd lockd: fix rpcbind crash on lockd startup failure 2014-09-08 12:03:32 -04:00
logfs fs/logfs/readwrite.c: kernel-doc warning fixes 2014-08-06 18:01:12 -07:00
minix minix zmap block counts calculation fix 2014-08-08 15:57:20 -07:00
ncpfs fs/ncpfs/getopt.c: replace simple_strtoul by kstrtoul 2014-06-04 16:54:21 -07:00
nfs pnfs: fix filelayout_retry_commit when idx > 0 2014-09-10 12:43:45 -07:00
nfs_common fs/nfs_common/nfsacl.c: move EXPORT symbol after functions 2014-07-12 18:43:42 -04:00
nfsd nfsd4: fix rd_dircount enforcement 2014-09-08 12:02:03 -04:00
nilfs2 Merge branch 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/viro/vfs 2014-08-11 11:44:11 -07:00
nls
notify fs/notify: don't show f_handle if exportfs_encode_inode_fh failed 2014-09-10 15:42:12 -07:00
ntfs ntfs: kernel-doc warning fixes 2014-08-06 18:01:12 -07:00
ocfs2 ocfs2: quorum: add a log for node not fenced 2014-08-29 16:28:17 -07:00
omfs fs/omfs/inode.c: replace count*size kzalloc by kcalloc 2014-08-08 15:57:25 -07:00
openpromfs fs: push sync_filesystem() down to the file system's remount_fs() 2014-03-13 10:14:33 -04:00
proc Revert "proc: Point /proc/{mounts,net} at /proc/thread-self/{mounts,net} instead of /proc/self/{mounts,net}" 2014-08-10 21:24:59 -07:00
pstore fs/pstore/ram_core.c: replace count*size kmalloc by kmalloc_array 2014-08-08 15:57:25 -07:00
qnx4 fs: push sync_filesystem() down to the file system's remount_fs() 2014-03-13 10:14:33 -04:00
qnx6 fs/qnx6: update debugging to current functions 2014-08-08 15:57:26 -07:00
quota fs/quota: kernel-doc warning fixes 2014-07-15 22:40:23 +02:00
ramfs fs/ramfs/file-nommu.c: replace count*size kzalloc by kcalloc 2014-08-08 15:57:18 -07:00
reiserfs Merge branch 'for_linus' of git://git.kernel.org/pub/scm/linux/kernel/git/jack/linux-fs 2014-08-13 17:45:40 -06:00
romfs fs/romfs/super.c: add blank line after declarations 2014-08-08 15:57:25 -07:00
squashfs fs/squashfs/super.c: logging cleanup 2014-08-06 18:01:13 -07:00
sysfs kernfs: move the last knowledge of sysfs out from kernfs 2014-06-03 08:11:18 -07:00
sysv write_iter variants of {__,}generic_file_aio_write() 2014-05-06 17:38:00 -04:00
ubifs UBIFS: Add log overlap assertions 2014-07-31 15:52:51 +03:00
udf udf: saner calling conventions for udf_new_inode() 2014-09-04 21:37:41 +02:00
ufs ufs: fix deadlocks introduced by sb mutex merge 2014-09-07 13:26:39 -04:00
xfs xfs: trim eofblocks before collapse range 2014-09-02 12:12:53 +10:00
aio.c aio: block exit_aio() until all context requests are completed 2014-09-04 16:54:47 -04:00
anon_inodes.c vfs: Allocate anon_inode_inode in anon_inode_init() 2014-03-27 09:52:54 -07:00
attr.c fs,userns: Change inode_capable to capable_wrt_inode_uidgid 2014-06-10 13:57:22 -07:00
bad_inode.c bad_inode: add ->rename2() 2014-08-07 14:40:09 -04:00
binfmt_aout.c
binfmt_elf_fdpic.c
binfmt_elf.c Merge branch 'x86/vdso' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/tip into next 2014-06-05 08:05:29 -07:00
binfmt_em86.c
binfmt_flat.c fs/binfmt_flat.c: make old_reloc() static 2014-06-04 16:54:21 -07:00
binfmt_misc.c binfmt_misc: add missing 'break' statement 2014-04-03 16:21:16 -07:00
binfmt_script.c
binfmt_som.c
block_dev.c Merge branch 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/viro/vfs 2014-06-12 10:30:18 -07:00
buffer.c sched: Remove proliferation of wait_on_bit() action functions 2014-07-16 15:10:39 +02:00
char_dev.c
compat_binfmt_elf.c binfmt_elf: add ELF_HWCAP2 to compat auxv entries 2014-03-04 08:05:21 +00:00
compat_ioctl.c Bluetooth: Move HCI socket definitions into its own header file 2014-07-11 13:53:04 +03:00
compat.c locks: rename file-private locks to "open file description locks" 2014-04-22 08:23:58 -04:00
coredump.c coredump: fix the setting of PF_DUMPCORE 2014-07-23 15:10:54 -07:00
dcache.c vfs: fix bad hashing of dentries 2014-09-13 11:30:10 -07:00
dcookies.c
direct-io.c switch iov_iter_get_pages() to passing maximal number of pages 2014-08-07 14:40:11 -04:00
drop_caches.c fs: convert use of typedef ctl_table to struct ctl_table 2014-06-06 16:08:16 -07:00
eventfd.c
eventpoll.c eventpoll: fix uninitialized variable in epoll_ctl 2014-09-10 15:42:12 -07:00
exec.c fork/exec: cleanup mm initialization 2014-08-08 15:57:23 -07:00
fcntl.c shm: add sealing API 2014-08-08 15:57:31 -07:00
fhandle.c
file_table.c Merge branch 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/viro/vfs 2014-06-12 10:30:18 -07:00
file.c fs/file.c: don't open-code kvfree() 2014-05-06 17:31:10 -04:00
filesystems.c sys_sysfs: Add CONFIG_SYSFS_SYSCALL 2014-04-03 16:21:05 -07:00
fs_pin.c make fs/{namespace,super}.c forget about acct.h 2014-08-07 14:40:09 -04:00
fs_struct.c
fs-writeback.c sched: Remove proliferation of wait_on_bit() action functions 2014-07-16 15:10:39 +02:00
inode.c mm: allow drivers to prevent new writable mappings 2014-08-08 15:57:31 -07:00
internal.h make fs/{namespace,super}.c forget about acct.h 2014-08-07 14:40:09 -04:00
ioctl.c
Kconfig kernfs: add CONFIG_KERNFS 2014-02-07 16:08:57 -08:00
Kconfig.binfmt
libfs.c fs/libfs.c: add generic data flush to fsync 2014-06-04 16:53:55 -07:00
locks.c locks: pass correct "before" pointer to locks_unlink_lock in generic_add_lease 2014-08-22 09:58:22 -04:00
Makefile take fs_pin stuff to fs/* 2014-08-07 14:40:08 -04:00
mbcache.c fs/mbcache: replace __builtin_log2() with ilog2() 2014-06-25 22:08:29 -04:00
mount.h death to mnt_pinned 2014-08-07 14:40:09 -04:00
mpage.c fs/block_dev.c: add bdev_read_page() and bdev_write_page() 2014-06-04 16:54:02 -07:00
namei.c vfs: fix bad hashing of dentries 2014-09-13 11:30:10 -07:00
namespace.c fix EBUSY on umount() from MNT_SHRINKABLE 2014-08-30 18:32:05 -04:00
no-block.c
open.c vfs: fix check for fallocate on active swapfile 2014-08-01 02:36:04 -04:00
pipe.c new helper: copy_page_from_iter() 2014-05-06 17:39:42 -04:00
pnode.c get rid of propagate_umount() mistakenly treating slaves as busy. 2014-08-30 18:31:41 -04:00
pnode.h smarter propagate_mnt() 2014-04-01 23:19:08 -04:00
posix_acl.c posix_acl: handle NULL ACL in posix_acl_equiv_mode 2014-05-06 13:58:42 -04:00
proc_namespace.c namespaces: Use task_lock and not rcu to protect nsproxy 2014-07-29 18:08:50 -07:00
read_write.c switch simple generic_file_aio_read() users to ->read_iter() 2014-05-06 17:37:55 -04:00
readdir.c fanotify: create FAN_ACCESS event for readdir 2014-06-04 16:53:52 -07:00
select.c
seq_file.c fs/seq_file: fallback to vmalloc allocation 2014-07-03 09:21:54 -07:00
signalfd.c
splice.c Merge commit '9f12600fe425bc28f0ccba034a77783c09c15af4' into for-linus 2014-06-12 00:28:09 -04:00
stack.c
stat.c
statfs.c
super.c Merge branch 'for_linus' of git://git.kernel.org/pub/scm/linux/kernel/git/jack/linux-fs 2014-08-13 17:45:40 -06:00
sync.c Export sync_filesystem() for modular ->remount_fs() use 2014-09-05 08:16:21 -07:00
timerfd.c timerfd: Use ktime_mono_to_real() 2014-07-23 10:18:02 -07:00
utimes.c
xattr.c simple_xattr: permit 0-size extended attributes 2014-07-23 15:10:55 -07:00