Add vdev property to toggle adding io to vdev queue

MigeljanImeri · MigeljanImeri · commit a978640bba26 · 2025-05-21T12:32:42.000-06:00
Added vdev property to toggle queueing io to vdev queue when
reading / writing from / to vdev. The intention behind this
property is to improve IOPS performance when using o_direct.

Signed-off-by: MigeljanImeri &lt;ImeriMigel@gmail.com&gt;
diff --git a/include/sys/fs/zfs.h b/include/sys/fs/zfs.h
@@ -385,9 +385,23 @@ typedef enum {
 	VDEV_PROP_TRIM_SUPPORT,
 	VDEV_PROP_TRIM_ERRORS,
 	VDEV_PROP_SLOW_IOS,
+	VDEV_PROP_QUEUE_IO,
 	VDEV_NUM_PROPS
 } vdev_prop_t;
 
+/*
+ * Different queuing behaviors for vdev prop queue_io.
+ * VDEV_QUEUE_AUTO = Don't queue if vdev is nonrot and backed by blkdev,
+ * queue otherwise.
+ * VDEV_QUEUE_CLASSIC = Always queue.
+ * VDEV_QUEUE_NONE = Never Queue.
+ */
+typedef enum {
+	VDEV_QUEUE_AUTO,
+	VDEV_QUEUE_CLASSIC,
+	VDEV_QUEUE_NONE
+} vdev_queue_type_t;
+
 /*
  * Dataset property functions shared between libzfs and kernel.
  */
diff --git a/include/sys/vdev.h b/include/sys/vdev.h
@@ -168,6 +168,7 @@ extern boolean_t vdev_is_spacemap_addressable(vdev_t *vd);
 
 extern void vdev_queue_init(vdev_t *vd);
 extern void vdev_queue_fini(vdev_t *vd);
+extern boolean_t vdev_should_queue_zio(zio_t *zio);
 extern zio_t *vdev_queue_io(zio_t *zio);
 extern void vdev_queue_io_done(zio_t *zio);
 extern void vdev_queue_change_io_priority(zio_t *zio, zio_priority_t priority);
diff --git a/include/sys/vdev_impl.h b/include/sys/vdev_impl.h
@@ -423,6 +423,7 @@ struct vdev {
 	boolean_t	vdev_resilver_deferred;  /* resilver deferred */
 	boolean_t	vdev_kobj_flag; /* kobj event record */
 	boolean_t	vdev_attaching; /* vdev attach ashift handling */
+	boolean_t	vdev_is_blkdev; /* vdev is backed by block device */
 	vdev_queue_t	vdev_queue;	/* I/O deadline schedule queue	*/
 	spa_aux_vdev_t	*vdev_aux;	/* for l2cache and spares vdevs	*/
 	zio_t		*vdev_probe_zio; /* root of current probe	*/
@@ -466,6 +467,7 @@ struct vdev {
 	uint64_t	vdev_io_t;
 	uint64_t	vdev_slow_io_n;
 	uint64_t	vdev_slow_io_t;
+	uint64_t	vdev_queue_io;
 };
 
 #define	VDEV_PAD_SIZE		(8 << 10)
diff --git a/lib/libzfs/libzfs.abi b/lib/libzfs/libzfs.abi
@@ -5899,7 +5899,8 @@
       <enumerator name='VDEV_PROP_TRIM_SUPPORT' value='49'/>
       <enumerator name='VDEV_PROP_TRIM_ERRORS' value='50'/>
       <enumerator name='VDEV_PROP_SLOW_IOS' value='51'/>
-      <enumerator name='VDEV_NUM_PROPS' value='52'/>
+      <enumerator name='VDEV_PROP_QUEUE_IO' value='52'/>
+      <enumerator name='VDEV_NUM_PROPS' value='53'/>
     </enum-decl>
     <typedef-decl name='vdev_prop_t' type-id='1573bec8' id='5aa5c90c'/>
     <class-decl name='zpool_load_policy' size-in-bits='256' is-struct='yes' visibility='default' id='2f65b36f'>
diff --git a/man/man7/vdevprops.7 b/man/man7/vdevprops.7
@@ -157,6 +157,22 @@ If this device should perform new allocations, used to disable a device
 when it is scheduled for later removal.
 See
 .Xr zpool-remove 8 .
+.It Sy queue_io Ns = Ns Sy auto Ns | Ns Sy classic Ns | Ns Sy none
+Controls how I/O requests are added to the vdev queue when reading or
+writing to this vdev.
+.It Sy auto
+Does not add I/O requests to the vdev queue if the vdev is backed by a
+non-rotational block device.
+This can sometimes improve performance for direct IOs.
+I/O requests will be added to the vdev queue if the vdev is backed by a
+rotational block device or file.
+This is the default behavior.
+.It Sy classic
+Always adds I/O requests to the vdev queue.
+.It Sy none
+Never adds I/O requests to the vdev queue.
+This is not recommended for vdevs backed by spinning disks as it could
+result in starvation.
 .El
 .Ss User Properties
 In addition to the standard native properties, ZFS supports arbitrary user
diff --git a/module/os/freebsd/zfs/vdev_geom.c b/module/os/freebsd/zfs/vdev_geom.c
@@ -968,6 +968,9 @@ vdev_geom_open(vdev_t *vd, uint64_t *psize, uint64_t *max_psize,
 	else
 		vd->vdev_nonrot = B_FALSE;
 
+	/* Is backed by a block device. */
+	vd->vdev_is_blkdev = B_TRUE;
+
 	/* Set when device reports it supports TRIM. */
 	error = g_getattr("GEOM::candelete", cp, &has_trim);
 	vd->vdev_has_trim = (error == 0 && has_trim);
diff --git a/module/os/linux/zfs/vdev_disk.c b/module/os/linux/zfs/vdev_disk.c
@@ -447,6 +447,9 @@ vdev_disk_open(vdev_t *v, uint64_t *psize, uint64_t *max_psize,
 	/* Inform the ZIO pipeline that we are non-rotational */
 	v->vdev_nonrot = blk_queue_nonrot(bdev_get_queue(bdev));
 
+	/* Is backed by a block device. */
+	v->vdev_is_blkdev = B_TRUE;
+
 	/* Physical volume size in bytes for the partition */
 	*psize = bdev_capacity(bdev);
 
diff --git a/module/zcommon/zpool_prop.c b/module/zcommon/zpool_prop.c
@@ -326,6 +326,13 @@ vdev_prop_init(void)
 		{ NULL }
 	};
 
+	static const zprop_index_t vdevqueuetype_table[] = {
+		{ "auto",	VDEV_QUEUE_AUTO },
+		{ "classic",	VDEV_QUEUE_CLASSIC },
+		{ "none",	VDEV_QUEUE_NONE },
+		{ NULL }
+	};
+
 	struct zfs_mod_supported_features *sfeatures =
 	    zfs_mod_list_supported(ZFS_SYSFS_VDEV_PROPERTIES);
 
@@ -470,6 +477,9 @@ vdev_prop_init(void)
 	zprop_register_index(VDEV_PROP_TRIM_SUPPORT, "trim_support", 0,
 	    PROP_READONLY, ZFS_TYPE_VDEV, "on | off", "TRIMSUP",
 	    boolean_table, sfeatures);
+	zprop_register_index(VDEV_PROP_QUEUE_IO, "queue_io", VDEV_QUEUE_AUTO,
+	    PROP_DEFAULT, ZFS_TYPE_VDEV, "auto | classic | none", "QUEUE_IO",
+	    vdevqueuetype_table, sfeatures);
 
 	/* default index properties */
 	zprop_register_index(VDEV_PROP_FAILFAST, "failfast", B_TRUE,
diff --git a/module/zfs/vdev.c b/module/zfs/vdev.c
@@ -718,6 +718,8 @@ vdev_alloc_common(spa_t *spa, uint_t id, uint64_t guid, vdev_ops_t *ops)
 	vd->vdev_slow_io_n = vdev_prop_default_numeric(VDEV_PROP_SLOW_IO_N);
 	vd->vdev_slow_io_t = vdev_prop_default_numeric(VDEV_PROP_SLOW_IO_T);
 
+	vd->vdev_queue_io = vdev_prop_default_numeric(VDEV_PROP_QUEUE_IO);
+
 	list_link_init(&vd->vdev_config_dirty_node);
 	list_link_init(&vd->vdev_state_dirty_node);
 	list_link_init(&vd->vdev_initialize_node);
@@ -3860,6 +3862,12 @@ vdev_load(vdev_t *vd)
 		if (error && error != ENOENT)
 			vdev_dbgmsg(vd, "vdev_load: zap_lookup(zap=%llu) "
 			    "failed [error=%d]", (u_longlong_t)zapobj, error);
+
+		error = vdev_prop_get_int(vd, VDEV_PROP_QUEUE_IO,
+		    &vd->vdev_queue_io);
+		if (error && error != ENOENT)
+			vdev_dbgmsg(vd, "vdev_load: zap_lookup(zap=%llu) "
+			    "failed [error=%d]", (u_longlong_t)zapobj, error);
 	}
 
 	/*
@@ -6095,6 +6103,15 @@ vdev_prop_set(vdev_t *vd, nvlist_t *innvl, nvlist_t *outnvl)
 			}
 			vd->vdev_slow_io_t = intval;
 			break;
+		case VDEV_PROP_QUEUE_IO:
+			if (nvpair_value_uint64(elem, &intval) != 0) {
+				error = EINVAL;
+				break;
+			}
+			if (vd->vdev_ops->vdev_op_leaf) {
+				vd->vdev_queue_io = intval;
+			}
+			break;
 		default:
 			/* Most processing is done in vdev_props_set_sync */
 			break;
@@ -6458,6 +6475,7 @@ vdev_prop_get(vdev_t *vd, nvlist_t *innvl, nvlist_t *outnvl)
 			case VDEV_PROP_IO_T:
 			case VDEV_PROP_SLOW_IO_N:
 			case VDEV_PROP_SLOW_IO_T:
+			case VDEV_PROP_QUEUE_IO:
 				err = vdev_prop_get_int(vd, prop, &intval);
 				if (err && err != ENOENT)
 					break;
diff --git a/module/zfs/vdev_file.c b/module/zfs/vdev_file.c
@@ -109,6 +109,9 @@ vdev_file_open(vdev_t *vd, uint64_t *psize, uint64_t *max_psize,
 	 */
 	vd->vdev_nonrot = B_TRUE;
 
+	/* Is not backed by a block device. */
+	vd->vdev_is_blkdev = B_FALSE;
+
 	/*
 	 * Allow TRIM on file based vdevs.  This may not always be supported,
 	 * since it depends on your kernel version and underlying filesystem
diff --git a/module/zfs/vdev_queue.c b/module/zfs/vdev_queue.c
@@ -879,6 +879,35 @@ vdev_queue_io_to_issue(vdev_queue_t *vq)
 	return (zio);
 }
 
+boolean_t
+vdev_should_queue_zio(zio_t *zio)
+{
+	vdev_t *vd = zio->io_vd;
+	boolean_t should_queue = B_TRUE;
+
+	/*
+	 * Add zio with ZIO_FLAG_NODATA to queue as bypass code
+	 * currently does not handle certain cases (gang abd, raidz
+	 * write aggregation).
+	 */
+	if (zio->io_flags & ZIO_FLAG_NODATA)
+		return (B_TRUE);
+
+	switch (vd->vdev_queue_io) {
+		case VDEV_QUEUE_AUTO:
+			if (vd->vdev_nonrot && vd->vdev_is_blkdev)
+				should_queue = B_FALSE;
+			break;
+		case VDEV_QUEUE_CLASSIC:
+			should_queue = B_TRUE;
+			break;
+		case VDEV_QUEUE_NONE:
+			should_queue = B_FALSE;
+			break;
+	}
+	return (should_queue);
+}
+
 zio_t *
 vdev_queue_io(zio_t *zio)
 {
@@ -922,6 +951,11 @@ vdev_queue_io(zio_t *zio)
 	zio->io_flags |= ZIO_FLAG_DONT_QUEUE;
 	zio->io_timestamp = gethrtime();
 
+	if (!vdev_should_queue_zio(zio)) {
+		zio->io_queue_state = ZIO_QS_NONE;
+		return (zio);
+	}
+
 	mutex_enter(&vq->vq_lock);
 	vdev_queue_io_add(vq, zio);
 	nio = vdev_queue_io_to_issue(vq);
@@ -950,10 +984,15 @@ vdev_queue_io_done(zio_t *zio)
 	zio_t *dio, *nio;
 	zio_link_t *zl = NULL;
 
+
 	hrtime_t now = gethrtime();
 	vq->vq_io_complete_ts = now;
 	vq->vq_io_delta_ts = zio->io_delta = now - zio->io_timestamp;
 
+	if (zio->io_queue_state == ZIO_QS_NONE) {
+		return;
+	}
+
 	mutex_enter(&vq->vq_lock);
 	vdev_queue_pending_remove(vq, zio);
 
diff --git a/tests/runfiles/common.run b/tests/runfiles/common.run
@@ -549,7 +549,7 @@ tags = ['functional', 'cli_root', 'zpool_scrub']
 tests = ['zpool_set_001_pos', 'zpool_set_002_neg', 'zpool_set_003_neg',
     'zpool_set_ashift', 'zpool_set_features', 'vdev_set_001_pos',
     'user_property_001_pos', 'user_property_002_neg',
-    'zpool_set_clear_userprop']
+    'zpool_set_clear_userprop','vdev_set_queue_io']
 tags = ['functional', 'cli_root', 'zpool_set']
 
 [tests/functional/cli_root/zpool_split]
diff --git a/tests/zfs-tests/tests/Makefile.am b/tests/zfs-tests/tests/Makefile.am
@@ -1245,6 +1245,7 @@ nobase_dist_datadir_zfs_tests_tests_SCRIPTS += \
 	functional/cli_root/zpool_set/setup.ksh \
 	functional/cli_root/zpool/setup.ksh \
 	functional/cli_root/zpool_set/vdev_set_001_pos.ksh \
+	functional/cli_root/zpool_set/vdev_set_queue_io.ksh \
 	functional/cli_root/zpool_set/zpool_set_common.kshlib \
 	functional/cli_root/zpool_set/zpool_set_001_pos.ksh \
 	functional/cli_root/zpool_set/zpool_set_002_neg.ksh \
diff --git a/tests/zfs-tests/tests/functional/cli_root/zpool_get/vdev_get.cfg b/tests/zfs-tests/tests/functional/cli_root/zpool_get/vdev_get.cfg
@@ -76,4 +76,5 @@ typeset -a properties=(
     trim_support
     trim_errors
     slow_ios
+    queue_io
 )
diff --git a/tests/zfs-tests/tests/functional/cli_root/zpool_set/vdev_set_queue_io.ksh b/tests/zfs-tests/tests/functional/cli_root/zpool_set/vdev_set_queue_io.ksh
@@ -0,0 +1,96 @@
+#!/bin/ksh -p
+# SPDX-License-Identifier: CDDL-1.0
+#
+# CDDL HEADER START
+#
+# The contents of this file are subject to the terms of the
+# Common Development and Distribution License (the "License").
+# You may not use this file except in compliance with the License.
+#
+# You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE
+# or https://opensource.org/licenses/CDDL-1.0.
+# See the License for the specific language governing permissions
+# and limitations under the License.
+#
+# When distributing Covered Code, include this CDDL HEADER in each
+# file and include the License file at usr/src/OPENSOLARIS.LICENSE.
+# If applicable, add the following below this CDDL HEADER, with the
+# fields enclosed by brackets "[]" replaced with your own identifying
+# information: Portions Copyright [yyyy] [name of copyright owner]
+#
+# CDDL HEADER END
+#
+
+#
+# Copyright (c) 2024 by Triad National Security, LLC.
+#
+
+. $STF_SUITE/include/libtest.shlib
+
+#
+# DESCRIPTION:
+# Toggling vdev queue_io property while reading from vdev should not cause panic.
+#
+# STRATEGY:
+# 1. Create a zpool
+# 2. Write a file to the pool.
+# 3. Start reading from file, while also toggling the queue_io property on / off.
+#
+
+verify_runnable "global"
+
+command -v fio > /dev/null || log_unsupported "fio missing"
+log_must save_tunable DIO_ENABLED
+log_must set_tunable32 DIO_ENABLED 1
+
+function toggle_queue_io
+{
+	zpool set queue_io=off $TESTPOOL1 $FILEDEV
+	sleep 0.1
+	zpool set queue_io=on $TESTPOOL1 $FILEDEV
+	sleep 0.1
+}
+
+function cleanup
+{
+	destroy_pool $TESTPOOL1
+	log_must rm -f $FILEDEV
+	log_must restore_tunable DIO_ENABLED
+}
+
+log_assert "Toggling vdev queue_io property while reading from vdev should not cause panic"
+log_onexit cleanup
+
+# 1. Create a pool
+
+FILEDEV="$TEST_BASE_DIR/filedev.$$"
+log_must truncate -s $(($MINVDEVSIZE * 2)) $FILEDEV
+create_pool $TESTPOOL1 $FILEDEV
+
+mntpnt=$(get_prop mountpoint $TESTPOOL1)
+
+# 2. Write a file to the pool, while also toggling the queue_io property on / off.
+
+log_must eval "fio --filename=$mntpnt/foobar --name=write-file \
+		--rw=write --size=$MINVDEVSIZE --bs=128k --numjobs=1 --direct=1 \
+		--ioengine=sync --time_based --runtime=10 &"
+
+ITERATIONS=30
+
+for i in $(seq $ITERATIONS); do
+	log_must toggle_queue_io
+done;
+wait
+
+# 3. Starting reading from file, while also toggling the queue_io property on / off.
+
+log_must eval "fio --filename=$mntpnt/foobar --name=read-file \
+		--rw=read --size=$MINVDEVSIZE --bs=128k --numjobs=1 --direct=1 \
+		--ioengine=sync --time_based --runtime=10 &"
+
+for i in $(seq $ITERATIONS); do
+	log_must toggle_queue_io
+done;
+wait
+
+log_pass "Toggling vdev queue_io property while reading from vdev does not cause panic"
