The ffs_balloc_ufs{1,2} functions call bdwrite() while having several

vnode buffers locked at once. In particular, there are indirect buffers
among locked ones. The bdwrite() may start the flushing to keep dirty
buffer list at the bounds. If any buffer on the dirty list requires
translation from logical to physical block number, code may ends up
trying to lock an indirect buffer already locked in ffs_balloc_ufsX.

Prevent the bdflush() activity when several buffers are locked at once
by setting the TDP_INBDFUSH for the problematic code blocks.

Reported and tested by:	pho, Josef Buchsteiner at Juniper
In collaboration with:	kan
MFC after:	1 month

1 files changed, 22 insertions, 2 deletions

diff --git a/sys/ufs/ffs/ffs_balloc.c b/sys/ufs/ffs/ffs_balloc.c
index 4d4fa893182b..0bd895f57410 100644
--- a/sys/ufs/ffs/ffs_balloc.c
+++ b/sys/ufs/ffs/ffs_balloc.c
@@ -104,6 +104,7 @@ ffs_balloc_ufs1(struct vnode *vp, off_t startoffset, int size,
 	ufs1_daddr_t *allocib, *blkp, *allocblk, allociblk[NIADDR + 1];
 	ufs2_daddr_t *lbns_remfree, lbns[NIADDR + 1];
 	int unwindidx = -1;
+	int saved_inbdflush;
 
 	ip = VTOI(vp);
 	dp = ip->i_din1;
@@ -225,6 +226,9 @@ ffs_balloc_ufs1(struct vnode *vp, off_t startoffset, int size,
 	if (num < 1)
 		panic ("ffs_balloc_ufs1: ufs_getlbns returned indirect block");
 #endif
+	saved_inbdflush = ~TDP_INBDFLUSH | (curthread->td_pflags &
+	    TDP_INBDFLUSH);
+	curthread->td_pflags |= TDP_INBDFLUSH;
 	/*
 	 * Fetch the first indirect block allocating if necessary.
 	 */
@@ -237,8 +241,10 @@ ffs_balloc_ufs1(struct vnode *vp, off_t startoffset, int size,
 		UFS_LOCK(ump);
 		pref = ffs_blkpref_ufs1(ip, lbn, 0, (ufs1_daddr_t *)0);
 	        if ((error = ffs_alloc(ip, lbn, pref, (int)fs->fs_bsize,
-		    cred, &newb)) != 0)
+		    cred, &newb)) != 0) {
+			curthread->td_pflags &= saved_inbdflush;
 			return (error);
+		}
 		nb = newb;
 		*allocblk++ = nb;
 		*lbns_remfree++ = indirs[1].in_lbn;
@@ -329,6 +335,7 @@ ffs_balloc_ufs1(struct vnode *vp, off_t startoffset, int size,
 	 * If asked only for the indirect block, then return it.
 	 */
 	if (flags & BA_METAONLY) {
+		curthread->td_pflags &= saved_inbdflush;
 		*bpp = bp;
 		return (0);
 	}
@@ -366,6 +373,7 @@ ffs_balloc_ufs1(struct vnode *vp, off_t startoffset, int size,
 				bp->b_flags |= B_CLUSTEROK;
 			bdwrite(bp);
 		}
+		curthread->td_pflags &= saved_inbdflush;
 		*bpp = nbp;
 		return (0);
 	}
@@ -387,9 +395,11 @@ ffs_balloc_ufs1(struct vnode *vp, off_t startoffset, int size,
 		nbp = getblk(vp, lbn, fs->fs_bsize, 0, 0, 0);
 		nbp->b_blkno = fsbtodb(fs, nb);
 	}
+	curthread->td_pflags &= saved_inbdflush;
 	*bpp = nbp;
 	return (0);
 fail:
+	curthread->td_pflags &= saved_inbdflush;
 	/*
 	 * If we have failed to allocate any blocks, simply return the error.
 	 * This is the usual case and avoids the need to fsync the file.
@@ -489,6 +499,7 @@ ffs_balloc_ufs2(struct vnode *vp, off_t startoffset, int size,
 	ufs2_daddr_t *lbns_remfree, lbns[NIADDR + 1];
 	int deallocated, osize, nsize, num, i, error;
 	int unwindidx = -1;
+	int saved_inbdflush;
 
 	ip = VTOI(vp);
 	dp = ip->i_din2;
@@ -719,6 +730,9 @@ ffs_balloc_ufs2(struct vnode *vp, off_t startoffset, int size,
 	if (num < 1)
 		panic ("ffs_balloc_ufs2: ufs_getlbns returned indirect block");
 #endif
+	saved_inbdflush = ~TDP_INBDFLUSH | (curthread->td_pflags &
+	    TDP_INBDFLUSH);
+	curthread->td_pflags |= TDP_INBDFLUSH;
 	/*
 	 * Fetch the first indirect block allocating if necessary.
 	 */
@@ -731,8 +745,10 @@ ffs_balloc_ufs2(struct vnode *vp, off_t startoffset, int size,
 		UFS_LOCK(ump);
 		pref = ffs_blkpref_ufs2(ip, lbn, 0, (ufs2_daddr_t *)0);
 	        if ((error = ffs_alloc(ip, lbn, pref, (int)fs->fs_bsize,
-		    cred, &newb)) != 0)
+		    cred, &newb)) != 0) {
+			curthread->td_pflags &= saved_inbdflush;
 			return (error);
+		}
 		nb = newb;
 		*allocblk++ = nb;
 		*lbns_remfree++ = indirs[1].in_lbn;
@@ -823,6 +839,7 @@ ffs_balloc_ufs2(struct vnode *vp, off_t startoffset, int size,
 	 * If asked only for the indirect block, then return it.
 	 */
 	if (flags & BA_METAONLY) {
+		curthread->td_pflags &= saved_inbdflush;
 		*bpp = bp;
 		return (0);
 	}
@@ -860,6 +877,7 @@ ffs_balloc_ufs2(struct vnode *vp, off_t startoffset, int size,
 				bp->b_flags |= B_CLUSTEROK;
 			bdwrite(bp);
 		}
+		curthread->td_pflags &= saved_inbdflush;
 		*bpp = nbp;
 		return (0);
 	}
@@ -887,9 +905,11 @@ ffs_balloc_ufs2(struct vnode *vp, off_t startoffset, int size,
 		nbp = getblk(vp, lbn, fs->fs_bsize, 0, 0, 0);
 		nbp->b_blkno = fsbtodb(fs, nb);
 	}
+	curthread->td_pflags &= saved_inbdflush;
 	*bpp = nbp;
 	return (0);
 fail:
+	curthread->td_pflags &= saved_inbdflush;
 	/*
 	 * If we have failed to allocate any blocks, simply return the error.
 	 * This is the usual case and avoids the need to fsync the file.