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

Correct a long-standing performance bug in cluster_rbuild(). Specifically,

Browse Source 
in the case of a file system with a block size that is less than the page
size, cluster_rbuild() looks at too many of the page's valid bits.
Consequently, it may terminate prematurely, resulting in poor performance.

Reported by:	bde
Reviewed by:	tegge
Approved by:	re (kib)
This commit is contained in:
Alan Cox 2009-06-27 21:37:36 +00:00
parent 29bd7d7e9a
commit de0c3e0895
Notes: svn2git 2020-12-20 02:59:44 +00:00 
svn path=/head/; revision=195122
1 changed files with 15 additions and 4 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

19
sys/kern/vfs_cluster.c
View File

@ -310,7 +310,9 @@ cluster_rbuild(vp, filesize, lbn, blkno, size, run, fbp)
struct bufobj *bo;
struct buf *bp, *tbp;
daddr_t bn;
int i, inc, j;
off_t off;
long tinc, tsize;
int i, inc, j, toff;
KASSERT(size == vp->v_mount->mnt_stat.f_iosize,
("cluster_rbuild: size %ld != filesize %jd\n",
@ -402,15 +404,24 @@ cluster_rbuild(vp, filesize, lbn, blkno, size, run, fbp)
* take part in the cluster. If it is partially valid
* then we stop.
*/
off = tbp->b_offset;
tsize = size;
VM_OBJECT_LOCK(tbp->b_bufobj->bo_object);
for (j = 0;j < tbp->b_npages; j++) {
for (j = 0; tsize > 0; j++) {
toff = off & PAGE_MASK;
tinc = tsize;
if (toff + tinc > PAGE_SIZE)
tinc = PAGE_SIZE - toff;
VM_OBJECT_LOCK_ASSERT(tbp->b_pages[j]->object,
MA_OWNED);
if (tbp->b_pages[j]->valid)
if ((tbp->b_pages[j]->valid &
vm_page_bits(toff, tinc)) != 0)
break;
off += tinc;
tsize -= tinc;
}
VM_OBJECT_UNLOCK(tbp->b_bufobj->bo_object);
if (j != tbp->b_npages) {
if (tsize > 0) {
bqrelse(tbp);
break;
}