aboutsummaryrefslogtreecommitdiffstats
diff options
context:
space:
mode:
Diffstat
-rw-r--r--cmd/zpool/zpool_main.c2
-rw-r--r--include/sys/spa.h4
-rw-r--r--include/sys/uberblock_impl.h16
-rw-r--r--include/sys/vdev_impl.h2
-rw-r--r--man/man8/zpool-clear.87
-rw-r--r--module/zfs/mmp.c5
-rw-r--r--module/zfs/spa.c102
-rw-r--r--module/zfs/txg.c9
-rw-r--r--module/zfs/vdev.c22
-rw-r--r--module/zfs/vdev_label.c4
-rw-r--r--module/zfs/zfs_ioctl.c9
-rw-r--r--module/zfs/zio.c6
-rw-r--r--module/zfs/zio_inject.c6
-rw-r--r--tests/runfiles/linux.run2
-rw-r--r--tests/zfs-tests/tests/Makefile.am1
-rwxr-xr-xtests/zfs-tests/tests/functional/mmp/mmp_write_slow_disk.ksh97
16 files changed, 242 insertions, 52 deletions
diff --git a/cmd/zpool/zpool_main.c b/cmd/zpool/zpool_main.c
index 20b1c8506..ed0b8d7a1 100644
--- a/cmd/zpool/zpool_main.c
+++ b/cmd/zpool/zpool_main.c
@@ -8910,7 +8910,7 @@ status_callback(zpool_handle_t *zhp, void *data)
printf_color(ANSI_BOLD, gettext("action: "));
printf_color(ANSI_YELLOW, gettext("Make sure the pool's devices"
" are connected, then reboot your system and\n\timport the "
- "pool.\n"));
+ "pool or run 'zpool clear' to resume the pool.\n"));
break;
case ZPOOL_STATUS_IO_FAILURE_WAIT:
diff --git a/include/sys/spa.h b/include/sys/spa.h
index 3112ba09c..6611141b9 100644
--- a/include/sys/spa.h
+++ b/include/sys/spa.h
@@ -769,7 +769,7 @@ extern int bpobj_enqueue_free_cb(void *arg, const blkptr_t *bp, dmu_tx_t *tx);
#define SPA_ASYNC_CONFIG_UPDATE 0x01
#define SPA_ASYNC_REMOVE 0x02
-#define SPA_ASYNC_PROBE 0x04
+#define SPA_ASYNC_FAULT_VDEV 0x04
#define SPA_ASYNC_RESILVER_DONE 0x08
#define SPA_ASYNC_RESILVER 0x10
#define SPA_ASYNC_AUTOEXPAND 0x20
@@ -1113,6 +1113,8 @@ extern uint32_t spa_get_hostid(spa_t *spa);
extern void spa_activate_allocation_classes(spa_t *, dmu_tx_t *);
extern boolean_t spa_livelist_delete_check(spa_t *spa);
+extern boolean_t spa_mmp_remote_host_activity(spa_t *spa);
+
extern spa_mode_t spa_mode(spa_t *spa);
extern uint64_t zfs_strtonum(const char *str, char **nptr);
diff --git a/include/sys/uberblock_impl.h b/include/sys/uberblock_impl.h
index 03bcfa8f4..13fce9c29 100644
--- a/include/sys/uberblock_impl.h
+++ b/include/sys/uberblock_impl.h
@@ -50,20 +50,20 @@ extern "C" {
#define MMP_SEQ_VALID_BIT 0x02
#define MMP_FAIL_INT_VALID_BIT 0x04
-#define MMP_VALID(ubp) (ubp->ub_magic == UBERBLOCK_MAGIC && \
- ubp->ub_mmp_magic == MMP_MAGIC)
-#define MMP_INTERVAL_VALID(ubp) (MMP_VALID(ubp) && (ubp->ub_mmp_config & \
+#define MMP_VALID(ubp) ((ubp)->ub_magic == UBERBLOCK_MAGIC && \
+ (ubp)->ub_mmp_magic == MMP_MAGIC)
+#define MMP_INTERVAL_VALID(ubp) (MMP_VALID(ubp) && ((ubp)->ub_mmp_config & \
MMP_INTERVAL_VALID_BIT))
-#define MMP_SEQ_VALID(ubp) (MMP_VALID(ubp) && (ubp->ub_mmp_config & \
+#define MMP_SEQ_VALID(ubp) (MMP_VALID(ubp) && ((ubp)->ub_mmp_config & \
MMP_SEQ_VALID_BIT))
-#define MMP_FAIL_INT_VALID(ubp) (MMP_VALID(ubp) && (ubp->ub_mmp_config & \
+#define MMP_FAIL_INT_VALID(ubp) (MMP_VALID(ubp) && ((ubp)->ub_mmp_config & \
MMP_FAIL_INT_VALID_BIT))
-#define MMP_INTERVAL(ubp) ((ubp->ub_mmp_config & 0x00000000FFFFFF00) \
+#define MMP_INTERVAL(ubp) (((ubp)->ub_mmp_config & 0x00000000FFFFFF00) \
>> 8)
-#define MMP_SEQ(ubp) ((ubp->ub_mmp_config & 0x0000FFFF00000000) \
+#define MMP_SEQ(ubp) (((ubp)->ub_mmp_config & 0x0000FFFF00000000) \
>> 32)
-#define MMP_FAIL_INT(ubp) ((ubp->ub_mmp_config & 0xFFFF000000000000) \
+#define MMP_FAIL_INT(ubp) (((ubp)->ub_mmp_config & 0xFFFF000000000000) \
>> 48)
#define MMP_INTERVAL_SET(write) \
diff --git a/include/sys/vdev_impl.h b/include/sys/vdev_impl.h
index 02948894c..8c6ab316f 100644
--- a/include/sys/vdev_impl.h
+++ b/include/sys/vdev_impl.h
@@ -274,7 +274,7 @@ struct vdev {
txg_list_t vdev_dtl_list; /* per-txg dirty DTL lists */
txg_node_t vdev_txg_node; /* per-txg dirty vdev linkage */
boolean_t vdev_remove_wanted; /* async remove wanted? */
- boolean_t vdev_probe_wanted; /* async probe wanted? */
+ boolean_t vdev_fault_wanted; /* async faulted wanted? */
list_node_t vdev_config_dirty_node; /* config dirty list */
list_node_t vdev_state_dirty_node; /* state dirty list */
uint64_t vdev_deflate_ratio; /* deflation ratio (x512) */
diff --git a/man/man8/zpool-clear.8 b/man/man8/zpool-clear.8
index c61ecae48..3e448be87 100644
--- a/man/man8/zpool-clear.8
+++ b/man/man8/zpool-clear.8
@@ -50,9 +50,10 @@ If the pool was suspended it will be brought back online provided the
devices can be accessed.
Pools with
.Sy multihost
-enabled which have been suspended cannot be resumed.
-While the pool was suspended, it may have been imported on
-another host, and resuming I/O could result in pool damage.
+enabled which have been suspended cannot be resumed when there is evidence
+that the pool was imported by another host.
+The same checks performed during an import will be applied before the clear
+proceeds.
.Bl -tag -width Ds
.It Fl -power
Power on the devices's slot in the storage enclosure and wait for the device
diff --git a/module/zfs/mmp.c b/module/zfs/mmp.c
index 25eea0752..8144d8965 100644
--- a/module/zfs/mmp.c
+++ b/module/zfs/mmp.c
@@ -663,12 +663,13 @@ mmp_thread(void *arg)
(gethrtime() - mmp->mmp_last_write) > mmp_fail_ns) {
zfs_dbgmsg("MMP suspending pool '%s': gethrtime %llu "
"mmp_last_write %llu mmp_interval %llu "
- "mmp_fail_intervals %llu mmp_fail_ns %llu",
+ "mmp_fail_intervals %llu mmp_fail_ns %llu txg %llu",
spa_name(spa), (u_longlong_t)gethrtime(),
(u_longlong_t)mmp->mmp_last_write,
(u_longlong_t)mmp_interval,
(u_longlong_t)mmp_fail_intervals,
- (u_longlong_t)mmp_fail_ns);
+ (u_longlong_t)mmp_fail_ns,
+ (u_longlong_t)spa->spa_uberblock.ub_txg);
cmn_err(CE_WARN, "MMP writes to pool '%s' have not "
"succeeded in over %llu ms; suspending pool. "
"Hrtime %llu",
diff --git a/module/zfs/spa.c b/module/zfs/spa.c
index fba784695..251dd8a4d 100644
--- a/module/zfs/spa.c
+++ b/module/zfs/spa.c
@@ -3542,11 +3542,16 @@ spa_activity_check_duration(spa_t *spa, uberblock_t *ub)
}
/*
- * Perform the import activity check. If the user canceled the import or
- * we detected activity then fail.
+ * Remote host activity check.
+ *
+ * error results:
+ * 0 - no activity detected
+ * EREMOTEIO - remote activity detected
+ * EINTR - user canceled the operation
*/
static int
-spa_activity_check(spa_t *spa, uberblock_t *ub, nvlist_t *config)
+spa_activity_check(spa_t *spa, uberblock_t *ub, nvlist_t *config,
+ boolean_t importing)
{
uint64_t txg = ub->ub_txg;
uint64_t timestamp = ub->ub_timestamp;
@@ -3591,19 +3596,23 @@ spa_activity_check(spa_t *spa, uberblock_t *ub, nvlist_t *config)
import_expire = gethrtime() + import_delay;
- spa_import_progress_set_notes(spa, "Checking MMP activity, waiting "
- "%llu ms", (u_longlong_t)NSEC2MSEC(import_delay));
+ if (importing) {
+ spa_import_progress_set_notes(spa, "Checking MMP activity, "
+ "waiting %llu ms", (u_longlong_t)NSEC2MSEC(import_delay));
+ }
- int interations = 0;
+ int iterations = 0;
while ((now = gethrtime()) < import_expire) {
- if (interations++ % 30 == 0) {
+ if (importing && iterations++ % 30 == 0) {
spa_import_progress_set_notes(spa, "Checking MMP "
"activity, %llu ms remaining",
(u_longlong_t)NSEC2MSEC(import_expire - now));
}
- (void) spa_import_progress_set_mmp_check(spa_guid(spa),
- NSEC2SEC(import_expire - gethrtime()));
+ if (importing) {
+ (void) spa_import_progress_set_mmp_check(spa_guid(spa),
+ NSEC2SEC(import_expire - gethrtime()));
+ }
vdev_uberblock_load(rvd, ub, &mmp_label);
@@ -3685,6 +3694,61 @@ out:
return (error);
}
+/*
+ * Called from zfs_ioc_clear for a pool that was suspended
+ * after failing mmp write checks.
+ */
+boolean_t
+spa_mmp_remote_host_activity(spa_t *spa)
+{
+ ASSERT(spa_multihost(spa) && spa_suspended(spa));
+
+ nvlist_t *best_label;
+ uberblock_t best_ub;
+
+ /*
+ * Locate the best uberblock on disk
+ */
+ vdev_uberblock_load(spa->spa_root_vdev, &best_ub, &best_label);
+ if (best_label) {
+ /*
+ * confirm that the best hostid matches our hostid
+ */
+ if (nvlist_exists(best_label, ZPOOL_CONFIG_HOSTID) &&
+ spa_get_hostid(spa) !=
+ fnvlist_lookup_uint64(best_label, ZPOOL_CONFIG_HOSTID)) {
+ nvlist_free(best_label);
+ return (B_TRUE);
+ }
+ nvlist_free(best_label);
+ } else {
+ return (B_TRUE);
+ }
+
+ if (!MMP_VALID(&best_ub) ||
+ !MMP_FAIL_INT_VALID(&best_ub) ||
+ MMP_FAIL_INT(&best_ub) == 0) {
+ return (B_TRUE);
+ }
+
+ if (best_ub.ub_txg != spa->spa_uberblock.ub_txg ||
+ best_ub.ub_timestamp != spa->spa_uberblock.ub_timestamp) {
+ zfs_dbgmsg("txg mismatch detected during pool clear "
+ "txg %llu ub_txg %llu timestamp %llu ub_timestamp %llu",
+ (u_longlong_t)spa->spa_uberblock.ub_txg,
+ (u_longlong_t)best_ub.ub_txg,
+ (u_longlong_t)spa->spa_uberblock.ub_timestamp,
+ (u_longlong_t)best_ub.ub_timestamp);
+ return (B_TRUE);
+ }
+
+ /*
+ * Perform an activity check looking for any remote writer
+ */
+ return (spa_activity_check(spa, &spa->spa_uberblock, spa->spa_config,
+ B_FALSE) != 0);
+}
+
static int
spa_verify_host(spa_t *spa, nvlist_t *mos_config)
{
@@ -4005,7 +4069,8 @@ spa_ld_select_uberblock(spa_t *spa, spa_import_type_t type)
return (spa_vdev_err(rvd, VDEV_AUX_ACTIVE, EREMOTEIO));
}
- int error = spa_activity_check(spa, ub, spa->spa_config);
+ int error =
+ spa_activity_check(spa, ub, spa->spa_config, B_TRUE);
if (error) {
nvlist_free(label);
return (error);
@@ -8589,15 +8654,16 @@ spa_async_remove(spa_t *spa, vdev_t *vd)
}
static void
-spa_async_probe(spa_t *spa, vdev_t *vd)
+spa_async_fault_vdev(spa_t *spa, vdev_t *vd)
{
- if (vd->vdev_probe_wanted) {
- vd->vdev_probe_wanted = B_FALSE;
- vdev_reopen(vd); /* vdev_open() does the actual probe */
+ if (vd->vdev_fault_wanted) {
+ vd->vdev_fault_wanted = B_FALSE;
+ vdev_set_state(vd, B_TRUE, VDEV_STATE_FAULTED,
+ VDEV_AUX_ERR_EXCEEDED);
}
for (int c = 0; c < vd->vdev_children; c++)
- spa_async_probe(spa, vd->vdev_child[c]);
+ spa_async_fault_vdev(spa, vd->vdev_child[c]);
}
static void
@@ -8685,11 +8751,11 @@ spa_async_thread(void *arg)
}
/*
- * See if any devices need to be probed.
+ * See if any devices need to be marked faulted.
*/
- if (tasks & SPA_ASYNC_PROBE) {
+ if (tasks & SPA_ASYNC_FAULT_VDEV) {
spa_vdev_state_enter(spa, SCL_NONE);
- spa_async_probe(spa, spa->spa_root_vdev);
+ spa_async_fault_vdev(spa, spa->spa_root_vdev);
(void) spa_vdev_state_exit(spa, NULL, 0);
}
diff --git a/module/zfs/txg.c b/module/zfs/txg.c
index a67c04344..5ce6be69b 100644
--- a/module/zfs/txg.c
+++ b/module/zfs/txg.c
@@ -551,6 +551,15 @@ txg_sync_thread(void *arg)
}
/*
+ * When we're suspended, nothing should be changing and for
+ * MMP we don't want to bump anything that would make it
+ * harder to detect if another host is changing it when
+ * resuming after a MMP suspend.
+ */
+ if (spa_suspended(spa))
+ continue;
+
+ /*
* Wait until the quiesce thread hands off a txg to us,
* prompting it to do so if necessary.
*/
diff --git a/module/zfs/vdev.c b/module/zfs/vdev.c
index 6d8eb50a1..981da4e98 100644
--- a/module/zfs/vdev.c
+++ b/module/zfs/vdev.c
@@ -1661,6 +1661,7 @@ vdev_metaslab_fini(vdev_t *vd)
typedef struct vdev_probe_stats {
boolean_t vps_readable;
boolean_t vps_writeable;
+ boolean_t vps_zio_done_probe;
int vps_flags;
} vdev_probe_stats_t;
@@ -1704,6 +1705,17 @@ vdev_probe_done(zio_t *zio)
(void) zfs_ereport_post(FM_EREPORT_ZFS_PROBE_FAILURE,
spa, vd, NULL, NULL, 0);
zio->io_error = SET_ERROR(ENXIO);
+
+ /*
+ * If this probe was initiated from zio pipeline, then
+ * change the state in a spa_async_request. Probes that
+ * were initiated from a vdev_open can change the state
+ * as part of the open call.
+ */
+ if (vps->vps_zio_done_probe) {
+ vd->vdev_fault_wanted = B_TRUE;
+ spa_async_request(spa, SPA_ASYNC_FAULT_VDEV);
+ }
}
mutex_enter(&vd->vdev_probe_lock);
@@ -1754,6 +1766,7 @@ vdev_probe(vdev_t *vd, zio_t *zio)
vps->vps_flags = ZIO_FLAG_CANFAIL | ZIO_FLAG_PROBE |
ZIO_FLAG_DONT_AGGREGATE | ZIO_FLAG_TRYHARD;
+ vps->vps_zio_done_probe = (zio != NULL);
if (spa_config_held(spa, SCL_ZIO, RW_WRITER)) {
/*
@@ -1780,15 +1793,6 @@ vdev_probe(vdev_t *vd, zio_t *zio)
vd->vdev_probe_zio = pio = zio_null(NULL, spa, vd,
vdev_probe_done, vps,
vps->vps_flags | ZIO_FLAG_DONT_PROPAGATE);
-
- /*
- * We can't change the vdev state in this context, so we
- * kick off an async task to do it on our behalf.
- */
- if (zio != NULL) {
- vd->vdev_probe_wanted = B_TRUE;
- spa_async_request(spa, SPA_ASYNC_PROBE);
- }
}
if (zio != NULL)
diff --git a/module/zfs/vdev_label.c b/module/zfs/vdev_label.c
index 737d8b33e..5c0e750c4 100644
--- a/module/zfs/vdev_label.c
+++ b/module/zfs/vdev_label.c
@@ -1982,6 +1982,7 @@ retry:
/*
* If this isn't a resync due to I/O errors,
* and nothing changed in this transaction group,
+ * and multihost protection isn't enabled,
* and the vdev configuration hasn't changed,
* then there's nothing to do.
*/
@@ -1989,7 +1990,8 @@ retry:
boolean_t changed = uberblock_update(ub, spa->spa_root_vdev,
txg, spa->spa_mmp.mmp_delay);
- if (!changed && list_is_empty(&spa->spa_config_dirty_list))
+ if (!changed && list_is_empty(&spa->spa_config_dirty_list) &&
+ !spa_multihost(spa))
return (0);
}
diff --git a/module/zfs/zfs_ioctl.c b/module/zfs/zfs_ioctl.c
index ac8329185..b07837113 100644
--- a/module/zfs/zfs_ioctl.c
+++ b/module/zfs/zfs_ioctl.c
@@ -5800,10 +5800,13 @@ zfs_ioc_clear(zfs_cmd_t *zc)
/*
* If multihost is enabled, resuming I/O is unsafe as another
- * host may have imported the pool.
+ * host may have imported the pool. Check for remote activity.
*/
- if (spa_multihost(spa) && spa_suspended(spa))
- return (SET_ERROR(EINVAL));
+ if (spa_multihost(spa) && spa_suspended(spa) &&
+ spa_mmp_remote_host_activity(spa)) {
+ spa_close(spa, FTAG);
+ return (SET_ERROR(EREMOTEIO));
+ }
spa_vdev_state_enter(spa, SCL_NONE);
diff --git a/module/zfs/zio.c b/module/zfs/zio.c
index d0b401623..046e6d64c 100644
--- a/module/zfs/zio.c
+++ b/module/zfs/zio.c
@@ -2503,8 +2503,10 @@ zio_suspend(spa_t *spa, zio_t *zio, zio_suspend_reason_t reason)
"failure and the failure mode property for this pool "
"is set to panic.", spa_name(spa));
- cmn_err(CE_WARN, "Pool '%s' has encountered an uncorrectable I/O "
- "failure and has been suspended.\n", spa_name(spa));
+ if (reason != ZIO_SUSPEND_MMP) {
+ cmn_err(CE_WARN, "Pool '%s' has encountered an uncorrectable "
+ "I/O failure and has been suspended.\n", spa_name(spa));
+ }
(void) zfs_ereport_post(FM_EREPORT_ZFS_IO_FAILURE, spa, NULL,
NULL, NULL, 0);
diff --git a/module/zfs/zio_inject.c b/module/zfs/zio_inject.c
index 609182f4a..66ad72fb8 100644
--- a/module/zfs/zio_inject.c
+++ b/module/zfs/zio_inject.c
@@ -605,9 +605,11 @@ zio_handle_io_delay(zio_t *zio)
if (vd->vdev_guid != handler->zi_record.zi_guid)
continue;
+ /* also match on I/O type (e.g., -T read) */
if (handler->zi_record.zi_iotype != ZIO_TYPES &&
- handler->zi_record.zi_iotype != zio->io_type)
- continue;
+ handler->zi_record.zi_iotype != zio->io_type) {
+ continue;
+ }
/*
* Defensive; should never happen as the array allocation
diff --git a/tests/runfiles/linux.run b/tests/runfiles/linux.run
index a0b74ef4a..92ce09ec6 100644
--- a/tests/runfiles/linux.run
+++ b/tests/runfiles/linux.run
@@ -146,7 +146,7 @@ tags = ['functional', 'mmap']
tests = ['mmp_on_thread', 'mmp_on_uberblocks', 'mmp_on_off', 'mmp_interval',
'mmp_active_import', 'mmp_inactive_import', 'mmp_exported_import',
'mmp_write_uberblocks', 'mmp_reset_interval', 'multihost_history',
- 'mmp_on_zdb', 'mmp_write_distribution', 'mmp_hostid']
+ 'mmp_on_zdb', 'mmp_write_distribution', 'mmp_hostid', 'mmp_write_slow_disk']
tags = ['functional', 'mmp']
[tests/functional/mount:Linux]
diff --git a/tests/zfs-tests/tests/Makefile.am b/tests/zfs-tests/tests/Makefile.am
index 8befed077..cc66d762f 100644
--- a/tests/zfs-tests/tests/Makefile.am
+++ b/tests/zfs-tests/tests/Makefile.am
@@ -1588,6 +1588,7 @@ nobase_dist_datadir_zfs_tests_tests_SCRIPTS += \
functional/mmp/mmp_on_zdb.ksh \
functional/mmp/mmp_reset_interval.ksh \
functional/mmp/mmp_write_distribution.ksh \
+ functional/mmp/mmp_write_slow_disk.ksh \
functional/mmp/mmp_write_uberblocks.ksh \
functional/mmp/multihost_history.ksh \
functional/mmp/setup.ksh \
diff --git a/tests/zfs-tests/tests/functional/mmp/mmp_write_slow_disk.ksh b/tests/zfs-tests/tests/functional/mmp/mmp_write_slow_disk.ksh
new file mode 100755
index 000000000..8b118684a
--- /dev/null
+++ b/tests/zfs-tests/tests/functional/mmp/mmp_write_slow_disk.ksh
@@ -0,0 +1,97 @@
+#!/bin/ksh -p
+#
+# CDDL HEADER START
+#
+# This file and its contents are supplied under the terms of the
+# Common Development and Distribution License ("CDDL"), version 1.0.
+# You may only use this file in accordance with the terms of version
+# 1.0 of the CDDL.
+#
+# A full copy of the text of the CDDL should have accompanied this
+# source. A copy of the CDDL is also available via the Internet at
+# http://www.illumos.org/license/CDDL.
+#
+# CDDL HEADER END
+#
+
+#
+# Copyright (c) 2024, Klara Inc
+#
+
+# DESCRIPTION:
+# Verify that long VDEV probes do not cause MMP checks to suspend pool
+# Note: without PR-15839 fix, this test will suspend the pool.
+#
+# A device that is returning unexpected errors will trigger a vdev_probe.
+# When the device additionally has slow response times, the probe can hold
+# the spa config lock as a writer for a long period of time such that the
+# mmp uberblock updates stall when trying to acquire the spa config lock.
+#
+# STRATEGY:
+# 1. Create a pool with multiple leaf vdevs
+# 2. Enable multihost and multihost_history
+# 3. Delay for MMP writes to occur
+# 4. Verify that a long VDEV probe didn't cause MMP check to suspend pool
+#
+
+. $STF_SUITE/include/libtest.shlib
+. $STF_SUITE/tests/functional/mmp/mmp.cfg
+. $STF_SUITE/tests/functional/mmp/mmp.kshlib
+
+verify_runnable "both"
+
+function cleanup
+{
+ log_must zinject -c all
+
+ if [[ $(zpool list -H -o health $MMP_POOL) == "SUSPENDED" ]]; then
+ log_must zpool clear $MMP_POOL
+ zpool get state $MMP_POOL $MMP_DIR/file.3
+ zpool events | grep ".fs.zfs." | grep -v "history_event"
+ fi
+
+ poolexists $MMP_POOL && destroy_pool $MMP_POOL
+ log_must rm -r $MMP_DIR
+ log_must mmp_clear_hostid
+}
+
+log_assert "A long VDEV probe doesn't cause a MMP check suspend"
+log_onexit cleanup
+
+MMP_HISTORY_URL=/proc/spl/kstat/zfs/$MMP_POOL/multihost
+
+# Create a multiple drive pool
+log_must zpool events -c
+log_must mkdir -p $MMP_DIR
+log_must truncate -s 128M $MMP_DIR/file.{0,1,2,3,4,5}
+log_must zpool create -f $MMP_POOL \
+ mirror $MMP_DIR/file.{0,1,2} \
+ mirror $MMP_DIR/file.{3,4,5}
+
+# Enable MMP
+log_must mmp_set_hostid $HOSTID1
+log_must zpool set multihost=on $MMP_POOL
+clear_mmp_history
+
+# Inject vdev write error along with a delay
+log_must zinject -f 33 -e io -L pad2 -T write -d $MMP_DIR/file.3 $MMP_POOL
+log_must zinject -f 50 -e io -L uber -T write -d $MMP_DIR/file.3 $MMP_POOL
+log_must zinject -D 2000:4 -T write -d $MMP_DIR/file.3 $MMP_POOL
+
+log_must dd if=/dev/urandom of=/$MMP_POOL/data bs=1M count=5
+sleep 10
+sync_pool $MMP_POOL
+
+# Confirm mmp writes to the non-slow disks have taken place
+for x in {0,1,2,4}; do
+ write_count=$(grep -c file.${x} $MMP_HISTORY_URL)
+ [[ $write_count -gt 0 ]] || log_fail "expecting mmp writes"
+done
+
+# Expect that the pool was not suspended
+log_must check_state $MMP_POOL "" "ONLINE"
+health=$(zpool list -H -o health $MMP_POOL)
+log_note "$MMP_POOL health is $health"
+[[ "$health" == "SUSPENDED" ]] && log_fail "$MMP_POOL $health unexpected"
+
+log_pass "A long VDEV probe doesn't cause a MMP check suspend"
generated by cgit v1.2.3 (git 2.39.1) at 2024-12-26 06:50:59 +0000