From cf0977ad72e97ae7d2aad1f0b6a2fe1a0209bbe7 Mon Sep 17 00:00:00 2001
From: Alan Somers <asomers@FreeBSD.org>
Date: Tue, 12 Jan 2021 15:25:52 -0700
Subject: [PATCH] Parallelize vdev_validate

The runtime of vdev_validate is dominated by the disk accesses in
vdev_label_read_config.  Speed it up by validating all vdevs in
parallel using a taskq.

Sponsored by: Axcient
Reviewed-by: Brian Behlendorf <behlendorf1@llnl.gov>
Signed-off-by: Alan Somers <asomers@gmail.com>
Closes #11470
---
 include/sys/vdev_impl.h |  2 ++
 module/zfs/vdev.c       | 39 +++++++++++++++++++++++++++++++++++++--
 module/zfs/vdev_label.c |  3 ++-
 3 files changed, 41 insertions(+), 3 deletions(-)

diff --git a/include/sys/vdev_impl.h b/include/sys/vdev_impl.h
index c509d390fc76..db4fe1447ebc 100644
--- a/include/sys/vdev_impl.h
+++ b/include/sys/vdev_impl.h
@@ -271,7 +271,9 @@ struct vdev {
 	boolean_t	vdev_nonrot;	/* true if solid state		*/
 	int		vdev_load_error; /* error on last load		*/
 	int		vdev_open_error; /* error on last open		*/
+	int		vdev_validate_error; /* error on last validate	*/
 	kthread_t	*vdev_open_thread; /* thread opening children	*/
+	kthread_t	*vdev_validate_thread; /* thread validating children */
 	uint64_t	vdev_crtxg;	/* txg when top-level was added */
 
 	/*
diff --git a/module/zfs/vdev.c b/module/zfs/vdev.c
index 018e48c3867f..36001e0a6626 100644
--- a/module/zfs/vdev.c
+++ b/module/zfs/vdev.c
@@ -2136,6 +2136,16 @@ vdev_open(vdev_t *vd)
 	return (0);
 }
 
+static void
+vdev_validate_child(void *arg)
+{
+	vdev_t *vd = arg;
+
+	vd->vdev_validate_thread = curthread;
+	vd->vdev_validate_error = vdev_validate(vd);
+	vd->vdev_validate_thread = NULL;
+}
+
 /*
  * Called once the vdevs are all opened, this routine validates the label
  * contents. This needs to be done before vdev_load() so that we don't
@@ -2150,18 +2160,43 @@ int
 vdev_validate(vdev_t *vd)
 {
 	spa_t *spa = vd->vdev_spa;
+	taskq_t *tq = NULL;
 	nvlist_t *label;
 	uint64_t guid = 0, aux_guid = 0, top_guid;
 	uint64_t state;
 	nvlist_t *nvl;
 	uint64_t txg;
+	int children = vd->vdev_children;
 
 	if (vdev_validate_skip)
 		return (0);
 
-	for (uint64_t c = 0; c < vd->vdev_children; c++)
-		if (vdev_validate(vd->vdev_child[c]) != 0)
+	if (children > 0) {
+		tq = taskq_create("vdev_validate", children, minclsyspri,
+		    children, children, TASKQ_PREPOPULATE);
+	}
+
+	for (uint64_t c = 0; c < children; c++) {
+		vdev_t *cvd = vd->vdev_child[c];
+
+		if (tq == NULL || vdev_uses_zvols(cvd)) {
+			vdev_validate_child(cvd);
+		} else {
+			VERIFY(taskq_dispatch(tq, vdev_validate_child, cvd,
+			    TQ_SLEEP) != TASKQID_INVALID);
+		}
+	}
+	if (tq != NULL) {
+		taskq_wait(tq);
+		taskq_destroy(tq);
+	}
+	for (int c = 0; c < children; c++) {
+		int error = vd->vdev_child[c]->vdev_validate_error;
+
+		if (error != 0)
 			return (SET_ERROR(EBADF));
+	}
+
 
 	/*
 	 * If the device has already failed, or was marked offline, don't do
diff --git a/module/zfs/vdev_label.c b/module/zfs/vdev_label.c
index d5966e1bd459..04202a9f8960 100644
--- a/module/zfs/vdev_label.c
+++ b/module/zfs/vdev_label.c
@@ -763,7 +763,8 @@ vdev_label_read_config(vdev_t *vd, uint64_t txg)
 	int flags = ZIO_FLAG_CONFIG_WRITER | ZIO_FLAG_CANFAIL |
 	    ZIO_FLAG_SPECULATIVE;
 
-	ASSERT(spa_config_held(spa, SCL_STATE_ALL, RW_WRITER) == SCL_STATE_ALL);
+	ASSERT(vd->vdev_validate_thread == curthread ||
+	    spa_config_held(spa, SCL_STATE_ALL, RW_WRITER) == SCL_STATE_ALL);
 
 	if (!vdev_readable(vd))
 		return (NULL);