Fix metaslab count assertion in metaslab_group_alloc() for small vdevs

When creating a pool on a small vdev (< 1GB) with a low
zfs_vdev_min_ms_count value (e.g., 2), vdev_metaslab_set_size() could
compute a metaslab shift that results in fewer metaslabs than
zfs_vdev_min_ms_count, triggering an assertion failure in
metaslab_group_alloc():
ASSERT3U(mg->mg_vd->vdev_ms_count, >=, 2) failed (1 >= 2)

The root cause is that highbit64(asize / zfs_vdev_min_ms_count) returns
last set bit, which can yield a metaslab size larger than
asize / zfs_vdev_min_ms_count. For example, with a 500MB vdev and
zfs_vdev_min_ms_count=2:
- asize / 2 = 250MB
- highbit64(250MB) = 28, giving 256MB metaslabs
- 500MB / 256MB = 1 metaslab (violates minimum of 2)

Fix by subtracting 1 from highbit64() result to ensure the metaslab
size is at most asize / zfs_vdev_min_ms_count, guaranteeing at least
zfs_vdev_min_ms_count metaslabs are created.

Test fixes:

trim/autotrim_config.ksh:

Increase post-trim thresholds from 30% to 40% (45% for draid). With the
metaslab fix, smaller vdevs now create more metaslabs as intended,
which increases spacemap metadata overhead. Adjust the post-trim size
threshold to accommodate this expected overhead.

replacement/replace_resilver_sit_out.ksh:

Increase test vdev size from 150MB to 256MB. With the metaslab fix,
150MB vdevs create 18 metaslabs (vs 9 before), changing read
distribution patterns and preventing the sit_out outlier detection
from triggering reliably. Larger vdevs provide more consistent read
patterns for the latency-based sit_out mechanism.

Reviewed-by: Brian Behlendorf <[email protected]>
Reviewed-by: Alexander Motin <[email protected]>
Signed-off-by: Jitendra Patidar <[email protected]>
Closes #19031

module/zfs/vdev.c

@@ -3102,7 +3102,11 @@ vdev_metaslab_set_size(vdev_t *vd)
 	 */
 
 	if (ms_count < zfs_vdev_min_ms_count)
-		ms_shift = highbit64(asize / zfs_vdev_min_ms_count);
+		/*
+		 * Subtract 1 from highbit64() to ensure ms_shift yields
+		 * at least zfs_vdev_min_ms_count metaslabs.
+		 */
+		ms_shift = highbit64(asize / zfs_vdev_min_ms_count) - 1;
 	else if (ms_count > zfs_vdev_default_ms_count)
 		ms_shift = highbit64(asize / zfs_vdev_default_ms_count);
 	else

tests/zfs-tests/tests/functional/replacement/replace_resilver_sit_out.ksh

@@ -130,7 +130,7 @@ function replace_test
 	verify_pool $TESTPOOL1
 }
 
-DEVSIZE="150M"
+DEVSIZE="256M"
 specials_list=""
 i=0
 while [[ $i != 10 ]]; do

tests/zfs-tests/tests/functional/trim/autotrim_config.ksh

@@ -71,7 +71,7 @@ typeset vdev_min_ms_count=$(get_tunable VDEV_MIN_MS_COUNT)
 log_must set_tunable64 VDEV_MIN_MS_COUNT 32
 
 typeset VDEV_MAX_MB=$(( floor(4 * MINVDEVSIZE * 0.75 / 1024 / 1024) ))
-typeset VDEV_MIN_MB=$(( floor(4 * MINVDEVSIZE * 0.30 / 1024 / 1024) ))
+typeset VDEV_MIN_MB=$(( floor(4 * MINVDEVSIZE * 0.40 / 1024 / 1024) ))
 
 for type in "" "mirror" "raidz2" "draid"; do
 
@@ -85,8 +85,9 @@ for type in "" "mirror" "raidz2" "draid"; do
 		VDEVS="$TRIM_VDEV1 $TRIM_VDEV2 $TRIM_VDEV3 $TRIM_VDEV4"
 
 		# The per-vdev utilization is lower due to the capacity
-		# resilverd for the distributed spare.
+		# reserved for the distributed spare.
 		VDEV_MAX_MB=$(( floor(4 * MINVDEVSIZE * 0.50 / 1024 / 1024) ))
+		VDEV_MIN_MB=$(( floor(4 * MINVDEVSIZE * 0.45 / 1024 / 1024) ))
 	fi
 
 	log_must truncate -s $((4 * MINVDEVSIZE)) $VDEVS