6754011 SPA 3.0: lock breakup, i/o pipeline refactoring, device failure handling

6667208 zfs/zpool commands on failed pool should not hang
6430480 grabbing config lock as writer during I/O load can take excessively long

1 files changed, 9 insertions, 21 deletions

diff --git a/usr/src/uts/common/fs/zfs/vdev_queue.c b/usr/src/uts/common/fs/zfs/vdev_queue.c
index 0f921e088a..46fca0e3b6 100644
--- a/usr/src/uts/common/fs/zfs/vdev_queue.c
+++ b/usr/src/uts/common/fs/zfs/vdev_queue.c
@@ -19,12 +19,10 @@
  * CDDL HEADER END
  */
 /*
- * Copyright 2007 Sun Microsystems, Inc.  All rights reserved.
+ * Copyright 2008 Sun Microsystems, Inc.  All rights reserved.
  * Use is subject to license terms.
  */
 
-#pragma ident	"%Z%%M%	%I%	%E% SMI"
-
 #include <sys/zfs_context.h>
 #include <sys/spa.h>
 #include <sys/vdev_impl.h>
@@ -191,6 +189,7 @@ vdev_queue_io_to_issue(vdev_queue_t *vq, uint64_t pending_limit)
 	size = fio->io_size;
 
 	while ((dio = AVL_PREV(tree, fio)) != NULL && IS_ADJACENT(dio, fio) &&
+	    !((dio->io_flags | fio->io_flags) & ZIO_FLAG_DONT_AGGREGATE) &&
 	    size + dio->io_size <= zfs_vdev_aggregation_limit) {
 		dio->io_delegate_next = fio;
 		fio = dio;
@@ -198,6 +197,7 @@ vdev_queue_io_to_issue(vdev_queue_t *vq, uint64_t pending_limit)
 	}
 
 	while ((dio = AVL_NEXT(tree, lio)) != NULL && IS_ADJACENT(lio, dio) &&
+	    !((lio->io_flags | dio->io_flags) & ZIO_FLAG_DONT_AGGREGATE) &&
 	    size + dio->io_size <= zfs_vdev_aggregation_limit) {
 		lio->io_delegate_next = dio;
 		lio = dio;
@@ -207,15 +207,12 @@ vdev_queue_io_to_issue(vdev_queue_t *vq, uint64_t pending_limit)
 	if (fio != lio) {
 		char *buf = zio_buf_alloc(size);
 		uint64_t offset = 0;
-		int nagg = 0;
 
 		ASSERT(size <= zfs_vdev_aggregation_limit);
 
-		aio = zio_vdev_child_io(fio, NULL, fio->io_vd,
-		    fio->io_offset, buf, size, fio->io_type,
-		    ZIO_PRIORITY_NOW, ZIO_FLAG_DONT_QUEUE |
-		    ZIO_FLAG_DONT_CACHE | ZIO_FLAG_DONT_PROPAGATE |
-		    ZIO_FLAG_NOBOOKMARK,
+		aio = zio_vdev_delegated_io(fio->io_vd, fio->io_offset,
+		    buf, size, fio->io_type, ZIO_PRIORITY_NOW,
+		    ZIO_FLAG_DONT_CACHE | ZIO_FLAG_DONT_QUEUE,
 		    vdev_queue_agg_io_done, NULL);
 
 		aio->io_delegate_list = fio;
@@ -228,16 +225,10 @@ vdev_queue_io_to_issue(vdev_queue_t *vq, uint64_t pending_limit)
 			offset += dio->io_size;
 			vdev_queue_io_remove(vq, dio);
 			zio_vdev_io_bypass(dio);
-			nagg++;
 		}
 
 		ASSERT(offset == size);
 
-		dprintf("%5s  T=%llu  off=%8llx  agg=%3d  "
-		    "old=%5llx  new=%5llx\n",
-		    zio_type_name[fio->io_type],
-		    fio->io_deadline, fio->io_offset, nagg, fio->io_size, size);
-
 		avl_add(&vq->vq_pending_tree, aio);
 
 		return (aio);
@@ -271,8 +262,7 @@ vdev_queue_io(zio_t *zio)
 
 	mutex_enter(&vq->vq_lock);
 
-	zio->io_deadline = (zio->io_timestamp >> zfs_vdev_time_shift) +
-	    zio->io_priority;
+	zio->io_deadline = (lbolt64 >> zfs_vdev_time_shift) + zio->io_priority;
 
 	vdev_queue_io_add(vq, zio);
 
@@ -295,15 +285,13 @@ void
 vdev_queue_io_done(zio_t *zio)
 {
 	vdev_queue_t *vq = &zio->io_vd->vdev_queue;
-	zio_t *nio;
-	int i;
 
 	mutex_enter(&vq->vq_lock);
 
 	avl_remove(&vq->vq_pending_tree, zio);
 
-	for (i = 0; i < zfs_vdev_ramp_rate; i++) {
-		nio = vdev_queue_io_to_issue(vq, zfs_vdev_max_pending);
+	for (int i = 0; i < zfs_vdev_ramp_rate; i++) {
+		zio_t *nio = vdev_queue_io_to_issue(vq, zfs_vdev_max_pending);
 		if (nio == NULL)
 			break;
 		mutex_exit(&vq->vq_lock);