linux-stable/fs
George Spelvin 2a18da7a9c fs/namei.c: Improve dcache hash function
Patch 0fed3ac866 improved the hash mixing, but the function is slower
than necessary; there's a 7-instruction dependency chain (10 on x86)
each loop iteration.

Word-at-a-time access is a very tight loop (which is good, because
link_path_walk() is one of the hottest code paths in the entire kernel),
and the hash mixing function must not have a longer latency to avoid
slowing it down.

There do not appear to be any published fast hash functions that:
1) Operate on the input a word at a time, and
2) Don't need to know the length of the input beforehand, and
3) Have a single iterated mixing function, not needing conditional
   branches or unrolling to distinguish different loop iterations.

One of the algorithms which comes closest is Yann Collet's xxHash, but
that's two dependent multiplies per word, which is too much.

The key insights in this design are:

1) Barring expensive ops like multiplies, to diffuse one input bit
   across 64 bits of hash state takes at least log2(64) = 6 sequentially
   dependent instructions.  That is more cycles than we'd like.
2) An operation like "hash ^= hash << 13" requires a second temporary
   register anyway, and on a 2-operand machine like x86, it's three
   instructions.
3) A better use of a second register is to hold a two-word hash state.
   With careful design, no temporaries are needed at all, so it doesn't
   increase register pressure.  And this gets rid of register copying
   on 2-operand machines, so the code is smaller and faster.
4) Using two words of state weakens the requirement for one-round mixing;
   we now have two rounds of mixing before cancellation is possible.
5) A two-word hash state also allows operations on both halves to be
   done in parallel, so on a superscalar processor we get more mixing
   in fewer cycles.

I ended up using a mixing function inspired by the ChaCha and Speck
round functions.  It is 6 simple instructions and 3 cycles per iteration
(assuming multiply by 9 can be done by an "lea" instruction):

		x ^= *input++;
	y ^= x;	x = ROL(x, K1);
	x += y;	y = ROL(y, K2);
	y *= 9;

Not only is this reversible, two consecutive rounds are reversible:
if you are given the initial and final states, but not the intermediate
state, it is possible to compute both input words.  This means that at
least 3 words of input are required to create a collision.

(It also has the property, used by hash_name() to avoid a branch, that
it hashes all-zero to all-zero.)

The rotate constants K1 and K2 were found by experiment.  The search took
a sample of random initial states (I used 1023) and considered the effect
of flipping each of the 64 input bits on each of the 128 output bits two
rounds later.  Each of the 8192 pairs can be considered a biased coin, and
adding up the Shannon entropy of all of them produces a score.

The best-scoring shifts also did well in other tests (flipping bits in y,
trying 3 or 4 rounds of mixing, flipping all 64*63/2 pairs of input bits),
so the choice was made with the additional constraint that the sum of the
shifts is odd and not too close to the word size.

The final state is then folded into a 32-bit hash value by a less carefully
optimized multiply-based scheme.  This also has to be fast, as pathname
components tend to be short (the most common case is one iteration!), but
there's some room for latency, as there is a fair bit of intervening logic
before the hash value is used for anything.

(Performance verified with "bonnie++ -s 0 -n 1536:-2" on tmpfs.  I need
a better benchmark; the numbers seem to show a slight dip in performance
between 4.6.0 and this patch, but they're too noisy to quote.)

Special thanks to Bruce fields for diligent testing which uncovered a
nasty fencepost error in an earlier version of this patch.

[checkpatch.pl formatting complaints noted and respectfully disagreed with.]

Signed-off-by: George Spelvin <linux@sciencehorizons.net>
Tested-by: J. Bruce Fields <bfields@redhat.com>
2016-05-28 15:45:29 -04:00
..
9p mm, fs: get rid of PAGE_CACHE_* and page_cache_{get,release} macros 2016-04-04 10:41:08 -07:00
adfs
affs mm, fs: get rid of PAGE_CACHE_* and page_cache_{get,release} macros 2016-04-04 10:41:08 -07:00
afs mm, fs: get rid of PAGE_CACHE_* and page_cache_{get,release} macros 2016-04-04 10:41:08 -07:00
autofs4 Merge branch 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/viro/vfs 2016-03-19 18:52:29 -07:00
befs
bfs
btrfs Merge branch 'for-linus-4.6' of git://git.kernel.org/pub/scm/linux/kernel/git/mason/linux-btrfs 2016-04-09 10:41:34 -07:00
cachefiles mm, fs: get rid of PAGE_CACHE_* and page_cache_{get,release} macros 2016-04-04 10:41:08 -07:00
ceph libceph: make authorizer destruction independent of ceph_auth_client 2016-04-25 20:54:13 +02:00
cifs mm, fs: remove remaining PAGE_CACHE_* and page_cache_{get,release} usage 2016-04-04 10:41:08 -07:00
coda Merge branch 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/viro/vfs 2016-01-23 12:24:56 -08:00
configfs mm, fs: get rid of PAGE_CACHE_* and page_cache_{get,release} macros 2016-04-04 10:41:08 -07:00
cramfs mm, fs: remove remaining PAGE_CACHE_* and page_cache_{get,release} usage 2016-04-04 10:41:08 -07:00
crypto ext4/fscrypto: avoid RCU lookup in d_revalidate 2016-04-12 20:01:35 -07:00
debugfs debugfs: Make automount point inodes permanently empty 2016-04-12 15:01:53 -07:00
devpts devpts: more pty driver interface cleanups 2016-04-26 15:47:32 -07:00
dlm mm, fs: get rid of PAGE_CACHE_* and page_cache_{get,release} macros 2016-04-04 10:41:08 -07:00
ecryptfs Merge branch 'ovl-fixes' into for-linus 2016-05-11 00:00:29 -04:00
efivarfs mm, fs: get rid of PAGE_CACHE_* and page_cache_{get,release} macros 2016-04-04 10:41:08 -07:00
efs
exofs mm, fs: get rid of PAGE_CACHE_* and page_cache_{get,release} macros 2016-04-04 10:41:08 -07:00
exportfs
ext2 mm, fs: remove remaining PAGE_CACHE_* and page_cache_{get,release} usage 2016-04-04 10:41:08 -07:00
ext4 ext4/fscrypto: avoid RCU lookup in d_revalidate 2016-04-12 20:01:35 -07:00
f2fs fscrypto: don't let data integrity writebacks fail with ENOMEM 2016-04-12 10:25:30 -07:00
fat fat: add config option to set UTF-8 mount option by default 2016-03-22 15:36:02 -07:00
freevxfs mm, fs: get rid of PAGE_CACHE_* and page_cache_{get,release} macros 2016-04-04 10:41:08 -07:00
fscache mm, fs: get rid of PAGE_CACHE_* and page_cache_{get,release} macros 2016-04-04 10:41:08 -07:00
fuse fuse: Fix return value from fuse_get_user_pages() 2016-04-25 13:01:04 +02:00
gfs2 mm, fs: get rid of PAGE_CACHE_* and page_cache_{get,release} macros 2016-04-04 10:41:08 -07:00
hfs mm, fs: get rid of PAGE_CACHE_* and page_cache_{get,release} macros 2016-04-04 10:41:08 -07:00
hfsplus mm, fs: get rid of PAGE_CACHE_* and page_cache_{get,release} macros 2016-04-04 10:41:08 -07:00
hostfs mm, fs: get rid of PAGE_CACHE_* and page_cache_{get,release} macros 2016-04-04 10:41:08 -07:00
hpfs hpfs: don't truncate the file when delete fails 2016-02-27 19:15:51 -05:00
hugetlbfs mm, fs: remove remaining PAGE_CACHE_* and page_cache_{get,release} usage 2016-04-04 10:41:08 -07:00
isofs Merge branch 'ovl-fixes' into for-linus 2016-05-11 00:00:29 -04:00
jbd2 mm, fs: get rid of PAGE_CACHE_* and page_cache_{get,release} macros 2016-04-04 10:41:08 -07:00
jffs2 mm, fs: get rid of PAGE_CACHE_* and page_cache_{get,release} macros 2016-04-04 10:41:08 -07:00
jfs mm, fs: get rid of PAGE_CACHE_* and page_cache_{get,release} macros 2016-04-04 10:41:08 -07:00
kernfs kernfs: kernfs_sop_show_path: don't return 0 after seq_dentry call 2016-05-12 11:03:51 -04:00
lockd
logfs mm, fs: get rid of PAGE_CACHE_* and page_cache_{get,release} macros 2016-04-04 10:41:08 -07:00
minix mm, fs: get rid of PAGE_CACHE_* and page_cache_{get,release} macros 2016-04-04 10:41:08 -07:00
ncpfs mm, fs: get rid of PAGE_CACHE_* and page_cache_{get,release} macros 2016-04-04 10:41:08 -07:00
nfs These changes contains a fix for overlayfs interacting with some 2016-04-07 17:22:20 -07:00
nfs_common
nfsd Various bugfixes, a RDMA update from Chuck Lever, and support for a new 2016-03-24 19:50:32 -07:00
nilfs2 mm, fs: get rid of PAGE_CACHE_* and page_cache_{get,release} macros 2016-04-04 10:41:08 -07:00
nls
notify fsnotify: turn fsnotify reaper thread into a workqueue job 2016-02-18 16:23:24 -08:00
ntfs mm, fs: remove remaining PAGE_CACHE_* and page_cache_{get,release} usage 2016-04-04 10:41:08 -07:00
ocfs2 ocfs2: fix posix_acl_create deadlock 2016-05-12 15:52:50 -07:00
omfs
openpromfs
orangefs Orangefs: cleanups and a strncpy vulnerability fix. 2016-04-09 10:33:58 -07:00
overlayfs ovl: ignore permissions on underlying lookup 2016-05-10 23:58:18 -04:00
proc Revert "proc/base: make prompt shell start from new line after executing "cat /proc/$pid/wchan"" 2016-05-09 17:40:59 -07:00
pstore mm, fs: get rid of PAGE_CACHE_* and page_cache_{get,release} macros 2016-04-04 10:41:08 -07:00
qnx4
qnx6 mm, fs: get rid of PAGE_CACHE_* and page_cache_{get,release} macros 2016-04-04 10:41:08 -07:00
quota quota: Handle Q_GETNEXTQUOTA when quota is disabled 2016-03-29 17:20:10 +02:00
ramfs mm, fs: get rid of PAGE_CACHE_* and page_cache_{get,release} macros 2016-04-04 10:41:08 -07:00
reiserfs mm, fs: remove remaining PAGE_CACHE_* and page_cache_{get,release} usage 2016-04-04 10:41:08 -07:00
romfs
squashfs mm, fs: remove remaining PAGE_CACHE_* and page_cache_{get,release} usage 2016-04-04 10:41:08 -07:00
sysfs
sysv mm, fs: get rid of PAGE_CACHE_* and page_cache_{get,release} macros 2016-04-04 10:41:08 -07:00
tracefs
ubifs mm, fs: remove remaining PAGE_CACHE_* and page_cache_{get,release} usage 2016-04-04 10:41:08 -07:00
udf udf: Fix conversion of 'dstring' fields to UTF8 2016-04-25 15:18:50 +02:00
ufs mm, fs: get rid of PAGE_CACHE_* and page_cache_{get,release} macros 2016-04-04 10:41:08 -07:00
xfs mm, fs: remove remaining PAGE_CACHE_* and page_cache_{get,release} usage 2016-04-04 10:41:08 -07:00
aio.c
anon_inodes.c
attr.c
bad_inode.c
binfmt_aout.c
binfmt_elf.c mm, fs: get rid of PAGE_CACHE_* and page_cache_{get,release} macros 2016-04-04 10:41:08 -07:00
binfmt_elf_fdpic.c mm, fs: get rid of PAGE_CACHE_* and page_cache_{get,release} macros 2016-04-04 10:41:08 -07:00
binfmt_em86.c
binfmt_flat.c
binfmt_misc.c
binfmt_script.c
block_dev.c mm, fs: get rid of PAGE_CACHE_* and page_cache_{get,release} macros 2016-04-04 10:41:08 -07:00
buffer.c mm, fs: get rid of PAGE_CACHE_* and page_cache_{get,release} macros 2016-04-04 10:41:08 -07:00
char_dev.c
compat.c
compat_binfmt_elf.c
compat_ioctl.c Merge 4.5-rc4 into char-misc-next 2016-02-14 14:25:59 -08:00
coredump.c fs/coredump: prevent fsuid=0 dumps into user-controlled directories 2016-03-22 15:36:02 -07:00
dax.c mm, fs: remove remaining PAGE_CACHE_* and page_cache_{get,release} usage 2016-04-04 10:41:08 -07:00
dcache.c fs/namei.c: Add hashlen_string() function 2016-05-28 15:42:50 -04:00
dcookies.c
direct-io.c mm, fs: get rid of PAGE_CACHE_* and page_cache_{get,release} macros 2016-04-04 10:41:08 -07:00
drop_caches.c
eventfd.c eventfd: document lockless access in eventfd_poll 2016-03-22 15:36:02 -07:00
eventpoll.c timer: convert timer_slack_ns from unsigned long to u64 2016-03-17 15:09:34 -07:00
exec.c Merge branch 'mm-pkeys-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/tip 2016-03-20 19:08:56 -07:00
fcntl.c
fhandle.c fs/coredump: prevent fsuid=0 dumps into user-controlled directories 2016-03-22 15:36:02 -07:00
file.c
file_table.c
filesystems.c
fs-writeback.c mm, fs: get rid of PAGE_CACHE_* and page_cache_{get,release} macros 2016-04-04 10:41:08 -07:00
fs_pin.c
fs_struct.c
inode.c writeback: initialize inode members that track writeback history 2016-02-16 14:57:21 -07:00
internal.h
ioctl.c
Kconfig Merge tag 'ofs-pull-tag-1' of git://git.kernel.org/pub/scm/linux/kernel/git/hubcap/linux 2016-03-26 12:59:04 -07:00
Kconfig.binfmt
libfs.c mm, fs: get rid of PAGE_CACHE_* and page_cache_{get,release} macros 2016-04-04 10:41:08 -07:00
locks.c
Makefile Merge tag 'ofs-pull-tag-1' of git://git.kernel.org/pub/scm/linux/kernel/git/hubcap/linux 2016-03-26 12:59:04 -07:00
mbcache.c mbcache: add reusable flag to cache entries 2016-02-22 22:44:04 -05:00
mount.h
mpage.c mm, fs: remove remaining PAGE_CACHE_* and page_cache_{get,release} usage 2016-04-04 10:41:08 -07:00
namei.c fs/namei.c: Improve dcache hash function 2016-05-28 15:45:29 -04:00
namespace.c
no-block.c
nsfs.c
open.c vfs: add vfs_select_inode() helper 2016-05-10 23:55:01 -04:00
pipe.c mm, fs: get rid of PAGE_CACHE_* and page_cache_{get,release} macros 2016-04-04 10:41:08 -07:00
pnode.c propogate_mnt: Handle the first propogated copy being a slave 2016-05-05 09:54:45 -05:00
pnode.h
posix_acl.c
proc_namespace.c vfs: show_vfsstat: do not ignore errors from show_devname method 2016-03-16 13:09:08 -04:00
read_write.c Merge branches 'work.lookups', 'work.misc' and 'work.preadv2' into for-next 2016-03-18 16:07:38 -04:00
readdir.c
select.c timer: convert timer_slack_ns from unsigned long to u64 2016-03-17 15:09:34 -07:00
seq_file.c Make file credentials available to the seqfile interfaces 2016-04-14 12:56:09 -07:00
signalfd.c
splice.c Merge branch 'ovl-fixes' into for-linus 2016-05-11 00:00:29 -04:00
stack.c
stat.c
statfs.c
super.c writeback: flush inode cgroup wb switches instead of pinning super_block 2016-03-03 14:42:50 -07:00
sync.c mm, fs: get rid of PAGE_CACHE_* and page_cache_{get,release} macros 2016-04-04 10:41:08 -07:00
timerfd.c
userfaultfd.c userfaultfd: don't block on the last VM updates at exit time 2016-03-02 09:03:18 -08:00
utimes.c
xattr.c xattr handlers: plug a lock leak in simple_xattr_list 2016-02-20 00:15:51 -05:00