Author: kib
Date: Thu Aug 22 18:26:45 2013
New Revision: 254668
URL: http://svnweb.freebsd.org/changeset/base/254668
Log:
Both cluster_rbuild() and cluster_wbuild() sometimes set the pages
shared busy without first draining the hard busy state. Previously it
went unnoticed since VPO_BUSY and m->busy fields were distinct, and
vm_page_io_start() did not verified that the passed page has VPO_BUSY
flag cleared, but such page state is wrong. New implementation is
more strict and catched this case.
Drain the busy state as needed, before calling vm_page_sbusy().
Tested by: pho, jkim
Sponsored by: The FreeBSD Foundation
Modified:
head/sys/kern/vfs_bio.c
head/sys/kern/vfs_cluster.c
head/sys/sys/buf.h
Modified: head/sys/kern/vfs_bio.c
==============================================================================
--- head/sys/kern/vfs_bio.c Thu Aug 22 18:12:24 2013 (r254667)
+++ head/sys/kern/vfs_bio.c Thu Aug 22 18:26:45 2013 (r254668)
@@ -108,7 +108,6 @@ static void vm_hold_load_pages(struct bu
static void vfs_page_set_valid(struct buf *bp, vm_ooffset_t off, vm_page_t m);
static void vfs_page_set_validclean(struct buf *bp, vm_ooffset_t off,
vm_page_t m);
-static void vfs_drain_busy_pages(struct buf *bp);
static void vfs_clean_pages_dirty_buf(struct buf *bp);
static void vfs_setdirty_locked_object(struct buf *bp);
static void vfs_vmio_release(struct buf *bp);
@@ -3983,7 +3982,7 @@ vfs_page_set_validclean(struct buf *bp,
* Ensure that all buffer pages are not exclusive busied. If any page is
* exclusive busy, drain it.
*/
-static void
+void
vfs_drain_busy_pages(struct buf *bp)
{
vm_page_t m;
Modified: head/sys/kern/vfs_cluster.c
==============================================================================
--- head/sys/kern/vfs_cluster.c Thu Aug 22 18:12:24 2013 (r254667)
+++ head/sys/kern/vfs_cluster.c Thu Aug 22 18:26:45 2013 (r254668)
@@ -315,7 +315,7 @@ cluster_rbuild(struct vnode *vp, u_quad_
daddr_t bn;
off_t off;
long tinc, tsize;
- int i, inc, j, toff;
+ int i, inc, j, k, toff;
KASSERT(size == vp->v_mount->mnt_stat.f_iosize,
("cluster_rbuild: size %ld != f_iosize %jd\n",
@@ -378,7 +378,15 @@ cluster_rbuild(struct vnode *vp, u_quad_
inc = btodb(size);
bo = &vp->v_bufobj;
for (bn = blkno, i = 0; i < run; ++i, bn += inc) {
- if (i != 0) {
+ if (i == 0) {
+ VM_OBJECT_WLOCK(tbp->b_bufobj->bo_object);
+ vfs_drain_busy_pages(tbp);
+ vm_object_pip_add(tbp->b_bufobj->bo_object,
+ tbp->b_npages);
+ for (k = 0; k < tbp->b_npages; k++)
+ vm_page_sbusy(tbp->b_pages[k]);
+ VM_OBJECT_WUNLOCK(tbp->b_bufobj->bo_object);
+ } else {
if ((bp->b_npages * PAGE_SIZE) +
round_page(size) > vp->v_mount->mnt_iosize_max) {
break;
@@ -424,14 +432,23 @@ cluster_rbuild(struct vnode *vp, u_quad_
if ((tbp->b_pages[j]->valid &
vm_page_bits(toff, tinc)) != 0)
break;
+ if (vm_page_xbusied(tbp->b_pages[j]))
+ break;
+ vm_object_pip_add(tbp->b_bufobj->bo_object, 1);
+ vm_page_sbusy(tbp->b_pages[j]);
off += tinc;
tsize -= tinc;
}
- VM_OBJECT_WUNLOCK(tbp->b_bufobj->bo_object);
if (tsize > 0) {
+clean_sbusy:
+ vm_object_pip_add(tbp->b_bufobj->bo_object, -j);
+ for (k = 0; k < j; k++)
+ vm_page_sunbusy(tbp->b_pages[k]);
+ VM_OBJECT_WUNLOCK(tbp->b_bufobj->bo_object);
bqrelse(tbp);
break;
}
+ VM_OBJECT_WUNLOCK(tbp->b_bufobj->bo_object);
/*
* Set a read-ahead mark as appropriate
@@ -451,8 +468,8 @@ cluster_rbuild(struct vnode *vp, u_quad_
if (tbp->b_blkno == tbp->b_lblkno) {
tbp->b_blkno = bn;
} else if (tbp->b_blkno != bn) {
- brelse(tbp);
- break;
+ VM_OBJECT_WLOCK(tbp->b_bufobj->bo_object);
+ goto clean_sbusy;
}
}
/*
@@ -466,10 +483,8 @@ cluster_rbuild(struct vnode *vp, u_quad_
for (j = 0; j < tbp->b_npages; j += 1) {
vm_page_t m;
m = tbp->b_pages[j];
- vm_page_sbusy(m);
- vm_object_pip_add(m->object, 1);
if ((bp->b_npages == 0) ||
- (bp->b_pages[bp->b_npages-1] != m)) {
+ (bp->b_pages[bp->b_npages-1] != m)) {
bp->b_pages[bp->b_npages] = m;
bp->b_npages++;
}
@@ -944,7 +959,9 @@ cluster_wbuild(struct vnode *vp, long si
vm_page_t m;
VM_OBJECT_WLOCK(tbp->b_bufobj->bo_object);
- if (i != 0) { /* if not first buffer */
+ if (i == 0) {
+ vfs_drain_busy_pages(tbp);
+ } else { /* if not first buffer */
for (j = 0; j < tbp->b_npages; j += 1) {
m = tbp->b_pages[j];
if (vm_page_xbusied(m)) {
Modified: head/sys/sys/buf.h
==============================================================================
--- head/sys/sys/buf.h Thu Aug 22 18:12:24 2013 (r254667)
+++ head/sys/sys/buf.h Thu Aug 22 18:26:45 2013 (r254668)
@@ -501,6 +501,7 @@ void bufstrategy(struct bufobj *, struct
void brelse(struct buf *);
void bqrelse(struct buf *);
int vfs_bio_awrite(struct buf *);
+void vfs_drain_busy_pages(struct buf *bp);
struct buf * getpbuf(int *);
struct buf *incore(struct bufobj *, daddr_t);
struct buf *gbincore(struct bufobj *, daddr_t);
_______________________________________________
svn-src-head@freebsd.org mailing list
http://lists.freebsd.org/mailman/listinfo/svn-src-head
To unsubscribe, send any mail to "svn-src-head-unsubscr...@freebsd.org"
