d/freebsd-dev
1
0
Fork 0
Code Issues Pull Requests Packages Projects Releases Wiki Activity

The UFS dirhash code was attempting to update shared state in the dirhash

Browse Source 
from multiple threads while holding a shared lock during a lookup operation.
This could result in incorrect ENOENT failures which could then be
permanently stored in the name cache.

Specifically, the dirhash code optimizes the case that a single thread is
walking a directory sequentially opening (or stat'ing) each file.  It uses
state in the dirhash structure to determine if a given lookup is using the
optimization.  If the optimization fails, it disables it and restarts the
lookup.  The problem arises when two threads both attempt the optimization
and fail.  The first thread will restart the loop, but the second thread
will incorrectly think that it did not try the optimization and will only
examine a subset of the directory entires in its hash chain.  As a result,
it may fail to find its directory entry and incorrectly fail with ENOENT.

To make this safe for use with shared locks, simplify the state stored in
the dirhash and move some of the state (the part that determines if the
current thread is trying the optimization) into a local variable.  One
result is that we will now try the optimization more often.  We still
update the value under the shared lock, but it is a single atomic store
similar to i_diroff that is stored in UFS directory i-nodes for the
non-dirhash lookup.

Reviewed by:	kib
MFC after:	1 week
This commit is contained in:
John Baldwin 2011-03-07 18:33:29 +00:00
parent cf3943e4d5
commit 8587289fb8
Notes: svn2git 2020-12-20 02:59:44 +00:00 
svn path=/head/; revision=219384
2 changed files with 15 additions and 18 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

1
sys/ufs/ufs/dirhash.h
View File

@ -98,7 +98,6 @@ struct dirhash {
int dh_dirblks; /* number of DIRBLKSIZ blocks in dir */
int dh_firstfree[DH_NFSTATS + 1]; /* first blk with N words free */
int dh_seqopt; /* sequential access optimisation enabled */
doff_t dh_seqoff; /* sequential access optimisation offset */
int dh_score; /* access count for this dirhash */

32
sys/ufs/ufs/ufs_dirhash.c
View File

@ -402,8 +402,7 @@ ufsdirhash_build(struct inode *ip)
dh->dh_firstfree[i] = -1;
dh->dh_firstfree[DH_NFSTATS] = 0;
dh->dh_hused = 0;
dh->dh_seqopt = 0;
dh->dh_seqoff = 0;
dh->dh_seqoff = -1;
dh->dh_score = DH_SCOREINIT;
dh->dh_lastused = time_second;
@ -551,7 +550,7 @@ ufsdirhash_lookup(struct inode *ip, char *name, int namelen, doff_t *offp,
struct direct *dp;
struct vnode *vp;
struct buf *bp;
doff_t blkoff, bmask, offset, prevoff;
doff_t blkoff, bmask, offset, prevoff, seqoff;
int i, slot;
int error;
@ -591,29 +590,30 @@ ufsdirhash_lookup(struct inode *ip, char *name, int namelen, doff_t *offp,
bmask = VFSTOUFS(vp->v_mount)->um_mountp->mnt_stat.f_iosize - 1;
blkoff = -1;
bp = NULL;
seqoff = dh->dh_seqoff;
restart:
slot = ufsdirhash_hash(dh, name, namelen);
if (dh->dh_seqopt) {
if (seqoff != -1) {
/*
* Sequential access optimisation. dh_seqoff contains the
* Sequential access optimisation. seqoff contains the
* offset of the directory entry immediately following
* the last entry that was looked up. Check if this offset
* appears in the hash chain for the name we are looking for.
*/
for (i = slot; (offset = DH_ENTRY(dh, i)) != DIRHASH_EMPTY;
i = WRAPINCR(i, dh->dh_hlen))
if (offset == dh->dh_seqoff)
if (offset == seqoff)
break;
if (offset == dh->dh_seqoff) {
if (offset == seqoff) {
/*
* We found an entry with the expected offset. This
* is probably the entry we want, but if not, the
* code below will turn off seqopt and retry.
* code below will retry.
*/
slot = i;
} else
dh->dh_seqopt = 0;
} else
seqoff = -1;
}
for (; (offset = DH_ENTRY(dh, slot)) != DIRHASH_EMPTY;
@ -655,9 +655,7 @@ ufsdirhash_lookup(struct inode *ip, char *name, int namelen, doff_t *offp,
*prevoffp = prevoff;
}
/* Check for sequential access, and update offset. */
if (dh->dh_seqopt == 0 && dh->dh_seqoff == offset)
dh->dh_seqopt = 1;
/* Update offset. */
dh->dh_seqoff = offset + DIRSIZ(0, dp);
*bpp = bp;
*offp = offset;
@ -666,11 +664,11 @@ ufsdirhash_lookup(struct inode *ip, char *name, int namelen, doff_t *offp,
}
/*
* When the name doesn't match in the seqopt case, go back
* and search normally.
* When the name doesn't match in the sequential
* optimization case, go back and search normally.
*/
if (dh->dh_seqopt) {
dh->dh_seqopt = 0;
if (seqoff != -1) {
seqoff = -1;
goto restart;
}
}