The IO_NOWDRAIN and B_NOWDRAIN hacks are no longer needed to prevent

deadlocks with vnode backed md(4) devices because md now uses a
kthread to run the bio requests instead of doing it directly from
the bio down path.

7 files changed, 7 insertions, 27 deletions

diff --git a/sys/dev/md/md.c b/sys/dev/md/md.c
index 7071a3e..1577f02 100644
--- a/sys/dev/md/md.c
+++ b/sys/dev/md/md.c
@@ -497,8 +497,7 @@ mdstart_vnode(struct md_s *sc, struct bio *bp)
 	auio.uio_td = curthread;
 	/*
 	 * When reading set IO_DIRECT to try to avoid double-caching
-	 * the data.  When writing IO_DIRECT is not optimal, but we
-	 * must set IO_NOWDRAIN to avoid a wdrain deadlock.
+	 * the data.  When writing IO_DIRECT is not optimal.
 	 */
 	if (bp->bio_cmd == BIO_READ) {
 		vn_lock(sc->vnode, LK_EXCLUSIVE | LK_RETRY, curthread);
@@ -506,7 +505,7 @@ mdstart_vnode(struct md_s *sc, struct bio *bp)
 	} else {
 		(void) vn_start_write(sc->vnode, &mp, V_WAIT);
 		vn_lock(sc->vnode, LK_EXCLUSIVE | LK_RETRY, curthread);
-		error = VOP_WRITE(sc->vnode, &auio, IO_NOWDRAIN, sc->cred);
+		error = VOP_WRITE(sc->vnode, &auio, 0, sc->cred);
 		vn_finished_write(mp);
 	}
 	VOP_UNLOCK(sc->vnode, 0, curthread);
diff --git a/sys/kern/vfs_bio.c b/sys/kern/vfs_bio.c
index 0b4a643..e1c0f9e 100644
--- a/sys/kern/vfs_bio.c
+++ b/sys/kern/vfs_bio.c
@@ -882,13 +882,10 @@ bwrite(struct buf * bp)
 		int rtval = bufwait(bp);
 		brelse(bp);
 		return (rtval);
-	} else if ((oldflags & B_NOWDRAIN) == 0) {
+	} else {
 		/*
 		 * don't allow the async write to saturate the I/O
-		 * system.  Deadlocks can occur only if a device strategy
-		 * routine (like in MD) turns around and issues another
-		 * high-level write, in which case B_NOWDRAIN is expected
-		 * to be set.  Otherwise we will not deadlock here because
+		 * system.  We will not deadlock here because
 		 * we are blocking waiting for I/O that is already in-progress
 		 * to complete.
 		 */
@@ -1461,8 +1458,7 @@ brelse(struct buf * bp)
 	if (bp->b_bufsize || bp->b_kvasize)
 		bufspacewakeup();
 
-	bp->b_flags &= ~(B_ASYNC | B_NOCACHE | B_AGE | B_RELBUF | 
-			B_DIRECT | B_NOWDRAIN);
+	bp->b_flags &= ~(B_ASYNC | B_NOCACHE | B_AGE | B_RELBUF | B_DIRECT);
 	if ((bp->b_flags & B_DELWRI) == 0 && (bp->b_xflags & BX_VNDIRTY))
 		panic("brelse: not dirty");
 	/* unlock */
diff --git a/sys/kern/vfs_cluster.c b/sys/kern/vfs_cluster.c
index 85fb8b3..72e80d6 100644
--- a/sys/kern/vfs_cluster.c
+++ b/sys/kern/vfs_cluster.c
@@ -839,7 +839,7 @@ cluster_wbuild(vp, size, start_lbn, len)
 		bp->b_data = (char *)((vm_offset_t)bp->b_data |
 		    ((vm_offset_t)tbp->b_data & PAGE_MASK));
 		bp->b_flags |= B_CLUSTER |
-				(tbp->b_flags & (B_VMIO | B_NEEDCOMMIT | B_NOWDRAIN));
+				(tbp->b_flags & (B_VMIO | B_NEEDCOMMIT));
 		bp->b_iodone = cluster_callback;
 		pbgetvp(vp, bp);
 		/*
diff --git a/sys/nfsclient/nfs_bio.c b/sys/nfsclient/nfs_bio.c
index 088be65..f36cbcb 100644
--- a/sys/nfsclient/nfs_bio.c
+++ b/sys/nfsclient/nfs_bio.c
@@ -973,12 +973,6 @@ again:
 			}
 			vfs_bio_set_validclean(bp, on, n);
 		}
-		/*
-		 * If IO_NOWDRAIN then set B_NOWDRAIN (nfs-backed MD 
-		 * filesystem)
-		 */
-		if (ioflag & IO_NOWDRAIN)
-			bp->b_flags |= B_NOWDRAIN;
 
 		/*
 		 * If IO_SYNC do bwrite().
diff --git a/sys/sys/buf.h b/sys/sys/buf.h
index 60d0d42..e41110f 100644
--- a/sys/sys/buf.h
+++ b/sys/sys/buf.h
@@ -206,10 +206,6 @@ struct buf {
  *			sticky until the buffer is released and typically
  *			only has an effect when B_RELBUF is also set.
  *
- *	B_NOWDRAIN	This flag should be set when a device (like MD)
- *			does a turn-around VOP_WRITE from its strategy
- *			routine.  This flag prevents bwrite() from blocking
- *			in wdrain, avoiding a deadlock situation.
  */
 
 #define	B_AGE		0x00000001	/* Move to age queue when I/O done. */
@@ -223,7 +219,7 @@ struct buf {
 #define	B_00000100	0x00000100	/* Available flag. */
 #define	B_DONE		0x00000200	/* I/O completed. */
 #define	B_EINTR		0x00000400	/* I/O was interrupted */
-#define	B_NOWDRAIN	0x00000800	/* Avoid wdrain deadlock */
+#define	B_00000800	0x00000800	/* Availabel flag. */
 #define	B_00001000	0x00001000	/* Available flag. */
 #define	B_INVAL		0x00002000	/* Does not contain valid info. */
 #define	B_LOCKED	0x00004000	/* Locked in core (not reusable). */
diff --git a/sys/sys/vnode.h b/sys/sys/vnode.h
index e48d392..230d4d3 100644
--- a/sys/sys/vnode.h
+++ b/sys/sys/vnode.h
@@ -279,7 +279,6 @@ struct vattr {
 #define	IO_INVAL	0x0040		/* invalidate after I/O */
 #define	IO_ASYNC	0x0080		/* bawrite rather then bdwrite */
 #define	IO_DIRECT	0x0100		/* attempt to bypass buffer cache */
-#define	IO_NOWDRAIN	0x0200		/* do not block on wdrain */
 #define	IO_EXT		0x0400		/* operate on external attributes */
 #define	IO_NORMAL	0x0800		/* operate on regular data */
 #define	IO_NOMACCHECK	0x1000		/* MAC checks unnecessary */
diff --git a/sys/ufs/ffs/ffs_vnops.c b/sys/ufs/ffs/ffs_vnops.c
index 9bcd7ff..910c0e4 100644
--- a/sys/ufs/ffs/ffs_vnops.c
+++ b/sys/ufs/ffs/ffs_vnops.c
@@ -702,8 +702,6 @@ ffs_write(ap)
 			vfs_bio_clrbuf(bp);
 		if (ioflag & IO_DIRECT)
 			bp->b_flags |= B_DIRECT;
-		if (ioflag & IO_NOWDRAIN)
-			bp->b_flags |= B_NOWDRAIN;
 
 		if (uio->uio_offset + xfersize > ip->i_size) {
 			ip->i_size = uio->uio_offset + xfersize;
@@ -1163,8 +1161,6 @@ ffs_extwrite(struct vnode *vp, struct uio *uio, int ioflag, struct ucred *ucred)
 			vfs_bio_clrbuf(bp);
 		if (ioflag & IO_DIRECT)
 			bp->b_flags |= B_DIRECT;
-		if (ioflag & IO_NOWDRAIN)
-			bp->b_flags |= B_NOWDRAIN;
 
 		if (uio->uio_offset + xfersize > dp->di_extsize)
 			dp->di_extsize = uio->uio_offset + xfersize;