D33921.id101618.diff
No OneTemporary
Actions

Size

13 KB

Referenced Files

None

Subscribers

None

D33921.id101618.diff
View Options

	diff --git a/sys/kern/vfs_bio.c b/sys/kern/vfs_bio.c
	--- a/sys/kern/vfs_bio.c
	+++ b/sys/kern/vfs_bio.c
	@@ -1702,7 +1702,7 @@
	if (freebufs == bd->bd_lofreebuffers)
	bufspace_daemon_wakeup(bd);

	- error = BUF_LOCK(bp, LK_EXCLUSIVE, NULL);
	+ error = BUF_LOCK(bp, LK_EXCLUSIVE \| LK_NOWAIT, NULL);
	KASSERT(error == 0, ("%s: BUF_LOCK on free buf %p: %d.", __func__, bp,
	error));
	(void)error;
	diff --git a/sys/ufs/ffs/ffs_alloc.c b/sys/ufs/ffs/ffs_alloc.c
	--- a/sys/ufs/ffs/ffs_alloc.c
	+++ b/sys/ufs/ffs/ffs_alloc.c
	@@ -81,6 +81,7 @@
	#include <sys/vnode.h>
	#include <sys/mount.h>
	#include <sys/kernel.h>
	+#include <sys/stat.h>
	#include <sys/syscallsubr.h>
	#include <sys/sysctl.h>
	#include <sys/syslog.h>
	@@ -269,7 +270,9 @@
	ump = ITOUMP(ip);
	fs = ump->um_fs;
	bp = NULL;
	- gbflags = (flags & BA_UNMAPPED) != 0 ? GB_UNMAPPED : 0;
	+retry_snap:
	+ gbflags = ((flags & BA_UNMAPPED) != 0 ? GB_UNMAPPED : 0) \|
	+ (IS_SNAPSHOT(ip) ? GB_LOCK_NOWAIT : 0);

	mtx_assert(UFS_MTX(ump), MA_OWNED);
	#ifdef INVARIANTS
	@@ -303,9 +306,14 @@
	* Allocate the extra space in the buffer.
	*/
	error = bread_gb(vp, lbprev, osize, NOCRED, gbflags, &bp);
	- if (error) {
	- return (error);
	+ if (error == EBUSY && IS_SNAPSHOT(ip)) {
	+ error = ffs_snap_relock(vp);
	+ if (error == EBADF)
	+ return (error);
	+ goto retry_snap;
	}
	+ if (error != 0)
	+ return (error);

	if (bp->b_blkno == bp->b_lblkno) {
	if (lbprev >= UFS_NDADDR)
	diff --git a/sys/ufs/ffs/ffs_balloc.c b/sys/ufs/ffs/ffs_balloc.c
	--- a/sys/ufs/ffs/ffs_balloc.c
	+++ b/sys/ufs/ffs/ffs_balloc.c
	@@ -70,6 +70,7 @@
	#include <sys/buf.h>
	#include <sys/lock.h>
	#include <sys/mount.h>
	+#include <sys/stat.h>
	#include <sys/vnode.h>
	#include <sys/vmmeter.h>

	@@ -587,6 +588,42 @@
	return (error);
	}

	+static int
	+ffs_balloc_bread(struct vnode vp, daddr_t lbn, int size, struct ucred cred,
	+ int gbflags, struct buf **bpp)
	+{
	+ int error;
	+
	+ for (;;) {
	+ error = bread_gb(vp, lbn, size, cred, gbflags \|
	+ (IS_SNAPSHOT(VTOI(vp)) ? GB_LOCK_NOWAIT : 0), bpp);
	+ if (error == 0)
	+ return (0);
	+ if (error == EBUSY && IS_SNAPSHOT(VTOI(vp)))
	+ error = ffs_snap_relock(vp);
	+ if (error != 0)
	+ return (error);
	+ }
	+}
	+
	+static int
	+ffs_balloc_getblk(struct vnode *vp, daddr_t lbn, int size, int gbflags,
	+ struct buf **bpp)
	+{
	+ int error;
	+
	+ for (;;) {
	+ error = getblkx(vp, lbn, 0, size, 0, 0, gbflags \|
	+ (IS_SNAPSHOT(VTOI(vp)) ? GB_LOCK_NOWAIT : 0), bpp);
	+ if (error == 0)
	+ return (0);
	+ if (error == EBUSY && IS_SNAPSHOT(VTOI(vp)))
	+ error = ffs_snap_relock(vp);
	+ if (error != 0)
	+ return (error);
	+ }
	+}
	+
	/*
	* Balloc defines the structure of file system storage
	* by allocating the physical blocks on a device given
	@@ -681,9 +718,9 @@
	panic("ffs_balloc_ufs2: BA_METAONLY for ext block");
	nb = dp->di_extb[lbn];
	if (nb != 0 && dp->di_extsize >= smalllblktosize(fs, lbn + 1)) {
	- error = bread_gb(vp, -1 - lbn, fs->fs_bsize, NOCRED,
	- gbflags, &bp);
	- if (error)
	+ error = ffs_balloc_bread(vp, -1 - lbn, fs->fs_bsize,
	+ NOCRED, gbflags, &bp);
	+ if (error != 0)
	goto done;
	bp->b_blkno = fsbtodb(fs, nb);
	bp->b_xflags \|= BX_ALTDATA;
	@@ -697,9 +734,9 @@
	osize = fragroundup(fs, blkoff(fs, dp->di_extsize));
	nsize = fragroundup(fs, size);
	if (nsize <= osize) {
	- error = bread_gb(vp, -1 - lbn, osize, NOCRED,
	- gbflags, &bp);
	- if (error)
	+ error = ffs_balloc_bread(vp, -1 - lbn, osize,
	+ NOCRED, gbflags, &bp);
	+ if (error != 0)
	goto done;
	bp->b_blkno = fsbtodb(fs, nb);
	bp->b_xflags \|= BX_ALTDATA;
	@@ -729,7 +766,12 @@
	nsize, flags, cred, &newb);
	if (error)
	goto done;
	- bp = getblk(vp, -1 - lbn, nsize, 0, 0, gbflags);
	+ error = ffs_balloc_getblk(vp, -1 - lbn, nsize, gbflags,
	+ &bp);
	+ if (error != 0) {
	+ /* XXXKIB free newb */
	+ goto done;
	+ }
	bp->b_blkno = fsbtodb(fs, newb);
	bp->b_xflags \|= BX_ALTDATA;
	if (flags & BA_CLRBUF)
	@@ -786,17 +828,15 @@
	nb = dp->di_db[lbn];
	if (nb != 0 && ip->i_size >= smalllblktosize(fs, lbn + 1)) {
	if ((flags & BA_CLRBUF) != 0) {
	- error = bread_gb(vp, lbn, fs->fs_bsize, NOCRED,
	- gbflags, &bp);
	+ error = ffs_balloc_bread(vp, lbn, fs->fs_bsize,
	+ NOCRED, gbflags, &bp);
	if (error != 0)
	goto done;
	} else {
	- bp = getblk(vp, lbn, fs->fs_bsize, 0, 0,
	- gbflags);
	- if (bp == NULL) {
	- error = EIO;
	+ error = ffs_balloc_getblk(vp, lbn,
	+ fs->fs_bsize, gbflags, &bp);
	+ if (error != 0)
	goto done;
	- }
	vfs_bio_clrbuf(bp);
	}
	bp->b_blkno = fsbtodb(fs, nb);
	@@ -811,9 +851,9 @@
	osize = fragroundup(fs, blkoff(fs, ip->i_size));
	nsize = fragroundup(fs, size);
	if (nsize <= osize) {
	- error = bread_gb(vp, lbn, osize, NOCRED,
	+ error = ffs_balloc_bread(vp, lbn, osize, NOCRED,
	gbflags, &bp);
	- if (error)
	+ if (error != 0)
	goto done;
	bp->b_blkno = fsbtodb(fs, nb);
	} else {
	@@ -838,9 +878,13 @@
	error = ffs_alloc(ip, lbn,
	ffs_blkpref_ufs2(ip, lbn, (int)lbn,
	&dp->di_db[0]), nsize, flags, cred, &newb);
	- if (error)
	+ if (error != 0)
	goto done;
	- bp = getblk(vp, lbn, nsize, 0, 0, gbflags);
	+ error = ffs_balloc_getblk(vp, lbn, nsize, gbflags, &bp);
	+ if (error != 0) {
	+ /* XXXKIB free newb */
	+ goto done;
	+ }
	bp->b_blkno = fsbtodb(fs, newb);
	if (flags & BA_CLRBUF)
	vfs_bio_clrbuf(bp);
	@@ -888,8 +932,12 @@
	MPASS(lbns_remfree < lbns + nitems(lbns));
	*allocblk++ = nb;
	*lbns_remfree++ = indirs[1].in_lbn;
	- bp = getblk(vp, indirs[1].in_lbn, fs->fs_bsize, 0, 0,
	- GB_UNMAPPED);
	+ error = ffs_balloc_getblk(vp, indirs[1].in_lbn, fs->fs_bsize,
	+ GB_UNMAPPED, &bp);
	+ if (error != 0) {
	+ /* XXXKIB free newb */
	+ goto done;
	+ }
	bp->b_blkno = fsbtodb(fs, nb);
	vfs_bio_clrbuf(bp);
	if (DOINGSOFTDEP(vp)) {
	@@ -914,11 +962,10 @@
	*/
	retry:
	for (i = 1;;) {
	- error = bread(vp,
	- indirs[i].in_lbn, (int)fs->fs_bsize, NOCRED, &bp);
	- if (error) {
	+ error = ffs_balloc_bread(vp, indirs[i].in_lbn,
	+ (int)fs->fs_bsize, NOCRED, 0, &bp);
	+ if (error != 0)
	goto fail;
	- }
	bap = (ufs2_daddr_t *)bp->b_data;
	nb = bap[indirs[i].in_off];
	if ((error = UFS_CHECK_BLKNO(mp, ip->i_number, nb,
	@@ -969,8 +1016,8 @@
	MPASS(lbns_remfree < lbns + nitems(lbns));
	*allocblk++ = nb;
	*lbns_remfree++ = indirs[i].in_lbn;
	- nbp = getblk(vp, indirs[i].in_lbn, fs->fs_bsize, 0, 0,
	- GB_UNMAPPED);
	+ error = ffs_balloc_getblk(vp, indirs[i].in_lbn, fs->fs_bsize,
	+ GB_UNMAPPED, &nbp);
	nbp->b_blkno = fsbtodb(fs, nb);
	vfs_bio_clrbuf(nbp);
	if (DOINGSOFTDEP(vp)) {
	@@ -1054,7 +1101,9 @@
	MPASS(lbns_remfree < lbns + nitems(lbns));
	*allocblk++ = nb;
	*lbns_remfree++ = lbn;
	- nbp = getblk(vp, lbn, fs->fs_bsize, 0, 0, gbflags);
	+ error = ffs_balloc_getblk(vp, lbn, fs->fs_bsize, gbflags, &nbp);
	+ if (error != 0)
	+ goto fail;
	nbp->b_blkno = fsbtodb(fs, nb);
	if (flags & BA_CLRBUF)
	vfs_bio_clrbuf(nbp);
	@@ -1091,10 +1140,11 @@
	(vp->v_mount->mnt_flag & MNT_NOCLUSTERR) == 0 &&
	!(vm_page_count_severe() \|\| buf_dirty_count_severe())) {
	error = cluster_read(vp, ip->i_size, lbn,
	- (int)fs->fs_bsize, NOCRED,
	- MAXBSIZE, seqcount, gbflags, &nbp);
	+ (int)fs->fs_bsize, NOCRED, MAXBSIZE, seqcount,
	+ gbflags \| (IS_SNAPSHOT(ip) ? GB_LOCK_NOWAIT : 0),
	+ &nbp);
	} else {
	- error = bread_gb(vp, lbn, (int)fs->fs_bsize,
	+ error = ffs_balloc_bread(vp, lbn, (int)fs->fs_bsize,
	NOCRED, gbflags, &nbp);
	}
	if (error) {
	@@ -1102,7 +1152,9 @@
	goto fail;
	}
	} else {
	- nbp = getblk(vp, lbn, fs->fs_bsize, 0, 0, gbflags);
	+ error = ffs_balloc_getblk(vp, lbn, fs->fs_bsize, gbflags, &nbp);
	+ if (error != 0)
	+ goto fail;
	nbp->b_blkno = fsbtodb(fs, nb);
	}
	curthread_pflags_restore(saved_inbdflush);
	@@ -1131,15 +1183,16 @@
	* XXX Still have to journal the free below
	*/
	(void) ffs_syncvnode(vp, MNT_WAIT, 0);
	+
	for (deallocated = 0, blkp = allociblk, lbns_remfree = lbns;
	blkp < allocblk; blkp++, lbns_remfree++) {
	/*
	* We shall not leave the freed blocks on the vnode
	* buffer object lists.
	*/
	- bp = getblk(vp, *lbns_remfree, fs->fs_bsize, 0, 0,
	- GB_NOCREAT \| GB_UNMAPPED);
	- if (bp != NULL) {
	+ error = ffs_balloc_getblk(vp, *lbns_remfree, fs->fs_bsize,
	+ GB_NOCREAT \| GB_UNMAPPED, &bp);
	+ if (error == 0) {
	KASSERT(bp->b_blkno == fsbtodb(fs, *blkp),
	("mismatch2 l %jd %jd b %ju %ju",
	(intmax_t)bp->b_lblkno, (uintmax_t)*lbns_remfree,
	@@ -1148,6 +1201,10 @@
	bp->b_flags \|= B_INVAL \| B_RELBUF \| B_NOCACHE;
	bp->b_flags &= ~(B_ASYNC \| B_CACHE);
	brelse(bp);
	+ } else if (error != EEXIST) {
	+ /* XXXKIB what to do? */
	+ printf("Can not getblk() to deallocate block, error %d",
	+ error);
	}
	deallocated += fs->fs_bsize;
	}
	@@ -1156,10 +1213,10 @@
	} else if (unwindidx >= 0) {
	int r;

	- r = bread(vp, indirs[unwindidx].in_lbn,
	- (int)fs->fs_bsize, NOCRED, &bp);
	- if (r) {
	- panic("Could not unwind indirect block, error %d", r);
	+ r = ffs_balloc_bread(vp, indirs[unwindidx].in_lbn,
	+ (int)fs->fs_bsize, NOCRED, 0, &bp);
	+ if (r != 0) {
	+ printf("Could not unwind indirect block, error %d", r);
	brelse(bp);
	} else {
	bap = (ufs2_daddr_t *)bp->b_data;
	@@ -1192,9 +1249,9 @@
	#ifdef INVARIANTS
	if (blkp == allociblk)
	lbns_remfree = lbns;
	- bp = getblk(vp, *lbns_remfree, fs->fs_bsize, 0, 0,
	- GB_NOCREAT \| GB_UNMAPPED);
	- if (bp != NULL) {
	+ error = ffs_balloc_getblk(vp, *lbns_remfree, fs->fs_bsize,
	+ GB_NOCREAT \| GB_UNMAPPED, &bp);
	+ if (error == 0) {
	panic("zombie2 %jd %ju %ju",
	(intmax_t)bp->b_lblkno, (uintmax_t)bp->b_blkno,
	(uintmax_t)fsbtodb(fs, *blkp));
	diff --git a/sys/ufs/ffs/ffs_extern.h b/sys/ufs/ffs/ffs_extern.h
	--- a/sys/ufs/ffs/ffs_extern.h
	+++ b/sys/ufs/ffs/ffs_extern.h
	@@ -99,6 +99,7 @@
	int ffs_snapblkfree(struct fs , struct vnode , ufs2_daddr_t, long, ino_t,
	enum vtype, struct workhead *);
	void ffs_snapremove(struct vnode *vp);
	+int ffs_snap_relock(struct vnode *vp);
	int ffs_snapshot(struct mount mp, char snapfile);
	void ffs_snapshot_mount(struct mount *mp);
	void ffs_snapshot_unmount(struct mount *mp);
	diff --git a/sys/ufs/ffs/ffs_inode.c b/sys/ufs/ffs/ffs_inode.c
	--- a/sys/ufs/ffs/ffs_inode.c
	+++ b/sys/ufs/ffs/ffs_inode.c
	@@ -161,25 +161,8 @@
	if (error != EBUSY \|\| (flags & GB_LOCK_NOWAIT) == 0)
	return (error);

	- /*
	- * Wait for our inode block to become available.
	- *
	- * Hold a reference to the vnode to protect against
	- * ffs_snapgone(). Since we hold a reference, it can only
	- * get reclaimed (VIRF_DOOMED flag) in a forcible downgrade
	- * or unmount. For an unmount, the entire filesystem will be
	- * gone, so we cannot attempt to touch anything associated
	- * with it while the vnode is unlocked; all we can do is
	- * pause briefly and try again. If when we relock the vnode
	- * we discover that it has been reclaimed, updating it is no
	- * longer necessary and we can just return an error.
	- */
	- vref(vp);
	- VOP_UNLOCK(vp);
	- pause("ffsupd", 1);
	- vn_lock(vp, LK_EXCLUSIVE \| LK_RETRY);
	- vrele(vp);
	- if (VN_IS_DOOMED(vp))
	+ error = ffs_snap_relock(vp);
	+ if (error == EBADF)
	return (ENOENT);

	/*
	diff --git a/sys/ufs/ffs/ffs_snapshot.c b/sys/ufs/ffs/ffs_snapshot.c
	--- a/sys/ufs/ffs/ffs_snapshot.c
	+++ b/sys/ufs/ffs/ffs_snapshot.c
	@@ -139,6 +139,12 @@
	{
	}

	+int
	+ffs_snap_relock(struct vnode *vp __unused)
	+{
	+ return (0);
	+}
	+
	#else
	FEATURE(ffs_snapshot, "FFS snapshot support");

	@@ -2745,4 +2751,33 @@
	return (sn);
	}

	+/*
	+ * This function should be used after bread_gb(GB_LOCK_NOWAIT)
	+ * returned EBUSY, when caller owns the snapshot lock. It temporarily
	+ * unlocks the snapshot vnode to allow for the block to become
	+ * available.
	+ */
	+int
	+ffs_snap_relock(struct vnode *vp)
	+{
	+ int locked;
	+
	+ MPASS(IS_SNAPSHOT(VTOI(vp)));
	+
	+ /*
	+ * Hold a reference to the vnode to protect against
	+ * ffs_snapgone(), in case caller does not.
	+ */
	+ vref(vp);
	+
	+ locked = VOP_ISLOCKED(vp);
	+ VOP_UNLOCK(vp);
	+ pause("slkrd", 1);
	+ vn_lock(vp, locked \| LK_RETRY);
	+ vrele(vp);
	+ if (!IS_UFS(vp))
	+ return (EBADF);
	+ return (0);
	+}
	+
	#endif
	diff --git a/sys/ufs/ffs/ffs_vnops.c b/sys/ufs/ffs/ffs_vnops.c
	--- a/sys/ufs/ffs/ffs_vnops.c
	+++ b/sys/ufs/ffs/ffs_vnops.c
	@@ -690,7 +690,9 @@
	uio->uio_offset >= fs->fs_maxfilesize)
	return (EOVERFLOW);

	- bflag = GB_UNMAPPED \| (uio->uio_segflg == UIO_NOCOPY ? 0 : GB_NOSPARSE);
	+ bflag = GB_UNMAPPED \|
	+ (uio->uio_segflg == UIO_NOCOPY ? 0 : GB_NOSPARSE) \|
	+ (IS_SNAPSHOT(ip) ? GB_LOCK_NOWAIT : 0);
	for (error = 0, bp = NULL; uio->uio_resid > 0; bp = NULL) {
	if ((bytesinfile = ip->i_size - uio->uio_offset) <= 0)
	break;
	@@ -759,6 +761,12 @@
	*/
	error = bread_gb(vp, lbn, size, NOCRED, bflag, &bp);
	}
	+ if (error == EBUSY && IS_SNAPSHOT(ip)) {
	+ error = ffs_snap_relock(vp);
	+ if (error == EBADF)
	+ break;
	+ continue;
	+ }
	if (error == EJUSTRETURN) {
	error = ffs_read_hole(uio, xfersize, &size);
	if (error == 0)
	@@ -2066,7 +2074,7 @@
	* and respond to dead vnodes by returning ESTALE.
	*/
	VOP_LOCK(vp, vp_locked \| LK_RETRY);
	- if (!VN_IS_DOOMED(vp))
	+ if (IS_UFS(vp))
	return (0);

	/*
	diff --git a/sys/ufs/ufs/inode.h b/sys/ufs/ufs/inode.h
	--- a/sys/ufs/ufs/inode.h
	+++ b/sys/ufs/ufs/inode.h
	@@ -247,6 +247,7 @@
	} while (0)

	#define IS_SNAPSHOT(ip) ((ip)->i_flags & SF_SNAPSHOT)
	+#define IS_UFS(vp) ((vp)->v_data != NULL)

	/*
	* Structure used to pass around logical block paths generated by

File Metadata

Mime Type: text/plain
Expires: Thu, Aug 6, 8:07 AM (1 h, 57 m)
Storage Engine: blob
Storage Format: Raw Data
Storage Handle: 36100797
Default Alt Text: D33921.id101618.diff (13 KB)

D33921.id101618.diffNo OneTemporaryActions

D33921.id101618.diffView Options

File Metadata

Event Timeline

D33921.id101618.diff
No OneTemporary
Actions

D33921.id101618.diff
View Options