zinject: Introduce ready delay fault injection

This adds a pause to the ZIO pipeline in the ready stage for
matching I/O (data, dnode, or raw bookmark).

Reviewed-by: Brian Behlendorf <behlendorf1@llnl.gov>
Reviewed-by: Rob Norris <robn@despairlabs.com>
Reviewed-by: Tony Hutter <hutter2@llnl.gov>
Reviewed-by: Akash B <akash-b@hpe.com>
Signed-off-by: Robert Evans <evansr@google.com>
Closes #17787

.github/ISSUE_TEMPLATE/feature_request.md

@@ -14,7 +14,7 @@ Please check our issue tracker before opening a new feature request.
 Filling out the following template will help other contributors better understand your proposed feature.
 -->
 
-### Describe the feature would like to see added to OpenZFS
+### Describe the feature you would like to see added to OpenZFS
 
 <!--
 Provide a clear and concise description of the feature.

.github/PULL_REQUEST_TEMPLATE.md

@@ -2,11 +2,6 @@
 
 <!--- Provide a general summary of your changes in the Title above -->
 
-<!---
-Documentation on ZFS Buildbot options can be found at
-https://openzfs.github.io/openzfs-docs/Developer%20Resources/Buildbot%20Options.html
--->
-
 ### Motivation and Context
 <!--- Why is this change required? What problem does it solve? -->
 <!--- If it fixes an open issue, please link to the issue here. -->

.github/workflows/scripts/generate-ci-type.py

@@ -65,7 +65,7 @@ if __name__ == '__main__':
 
     # check last (HEAD) commit message
     last_commit_message_raw = subprocess.run([
-        'git', 'show', '-s', '--format=%B', 'HEAD'
+        'git', 'show', '-s', '--format=%B', head
     ], check=True, stdout=subprocess.PIPE, stderr=subprocess.PIPE)
 
     for line in last_commit_message_raw.stdout.decode().splitlines():

.github/workflows/scripts/qemu-1-setup.sh

@@ -6,6 +6,13 @@
 
 set -eu
 
+# We've been seeing this script take over 15min to run.  This may or
+# may not be normal.  Just to get a little more insight, print out
+# a message to stdout with the top running process, and do this every
+# 30 seconds.  We can delete this watchdog later once we get a better
+# handle on what the timeout value should be.
+(while [ 1 ] ; do sleep 30 && echo "[watchdog: $(ps -eo cmd --sort=-pcpu  | head -n 2 | tail -n 1)}')]"; done) &
+
 # install needed packages
 export DEBIAN_FRONTEND="noninteractive"
 sudo apt-get -y update
@@ -65,3 +72,6 @@ sudo zpool create -f -o ashift=12 zpool $SSD1 $SSD2 -O relatime=off \
 for i in /sys/block/s*/queue/scheduler; do
   echo "none" | sudo tee $i
 done
+
+# Kill off our watchdog
+kill $(jobs -p)

.github/workflows/scripts/qemu-2-start.sh

@@ -121,7 +121,7 @@ case "$OS" in
     KSRC="$FREEBSD_SNAP/../amd64/$FreeBSD/src.txz"
     ;;
   freebsd15-0c)
-    FreeBSD="15.0-PRERELEASE"
+    FreeBSD="15.0-ALPHA3"
     OSNAME="FreeBSD $FreeBSD"
     OSv="freebsd14.0"
     URLxz="$FREEBSD_SNAP/$FreeBSD/amd64/Latest/FreeBSD-$FreeBSD-amd64-BASIC-CI-ufs.raw.xz"

.github/workflows/scripts/qemu-3-deps-vm.sh

@@ -20,7 +20,7 @@ function archlinux() {
   sudo pacman -Sy --noconfirm base-devel bc cpio cryptsetup dhclient dkms \
     fakeroot fio gdb inetutils jq less linux linux-headers lsscsi nfs-utils \
     parted pax perf python-packaging python-setuptools qemu-guest-agent ksh \
-    samba sysstat rng-tools rsync wget xxhash
+    samba strace sysstat rng-tools rsync wget xxhash
   echo "##[endgroup]"
 }
 
@@ -43,7 +43,8 @@ function debian() {
     lsscsi nfs-kernel-server pamtester parted python3 python3-all-dev \
     python3-cffi python3-dev python3-distlib python3-packaging libtirpc-dev \
     python3-setuptools python3-sphinx qemu-guest-agent rng-tools rpm2cpio \
-    rsync samba sysstat uuid-dev watchdog wget xfslibs-dev  xxhash zlib1g-dev
+    rsync samba strace sysstat uuid-dev watchdog wget xfslibs-dev xxhash \
+    zlib1g-dev
   echo "##[endgroup]"
 }
 
@@ -87,8 +88,8 @@ function rhel() {
     libuuid-devel lsscsi mdadm nfs-utils openssl-devel pam-devel pamtester \
     parted perf python3 python3-cffi python3-devel python3-packaging \
     kernel-devel python3-setuptools qemu-guest-agent rng-tools rpcgen \
-    rpm-build rsync samba sysstat systemd watchdog wget xfsprogs-devel xxhash \
+    rpm-build rsync samba strace sysstat systemd watchdog wget xfsprogs-devel \
-    zlib-devel
+    xxhash zlib-devel
   echo "##[endgroup]"
 }
 
@@ -104,7 +105,7 @@ function install_fedora_experimental_kernel {
   our_version="$1"
   sudo dnf -y copr enable @kernel-vanilla/stable
   sudo dnf -y copr enable @kernel-vanilla/mainline
-  all="$(sudo dnf list --showduplicates kernel-*)"
+  all="$(sudo dnf list --showduplicates kernel-* python3-perf* perf* bpftool*)"
   echo "Available versions:"
   echo "$all"
 

.github/workflows/scripts/qemu-5-setup.sh

@@ -108,19 +108,30 @@ echo '*/5 * * * *  /root/cronjob.sh' > crontab.txt
 sudo crontab crontab.txt
 rm crontab.txt
 
-# check if the machines are okay
-echo "Waiting for vm's to come up...  (${VMs}x CPU=$CPU RAM=$RAM)"
-for ((i=1; i<=VMs; i++)); do
-  .github/workflows/scripts/qemu-wait-for-vm.sh vm$i
-done
-echo "All $VMs VMs are up now."
-
 # Save the VM's serial output (ttyS0) to /var/tmp/console.txt
 # - ttyS0 on the VM corresponds to a local /dev/pty/N entry
 # - use 'virsh ttyconsole' to lookup the /dev/pty/N entry
 for ((i=1; i<=VMs; i++)); do
   mkdir -p $RESPATH/vm$i
   read "pty" <<< $(sudo virsh ttyconsole vm$i)
+
+  # Create the file so we can tail it, even if there's no output.
+  touch $RESPATH/vm$i/console.txt
+
   sudo nohup bash -c "cat $pty > $RESPATH/vm$i/console.txt" &
+
+  # Write all VM boot lines to the console to aid in debugging failed boots.
+  # The boot lines from all the VMs will be munged together, so prepend each
+  # line with the vm hostname (like 'vm1:').
+  (while IFS=$'\n' read -r line; do echo "vm$i: $line" ; done < <(sudo tail -f $RESPATH/vm$i/console.txt)) &
+
 done
 echo "Console logging for ${VMs}x $OS started."
+
+
+# check if the machines are okay
+echo "Waiting for vm's to come up...  (${VMs}x CPU=$CPU RAM=$RAM)"
+for ((i=1; i<=VMs; i++)); do
+  .github/workflows/scripts/qemu-wait-for-vm.sh vm$i
+done
+echo "All $VMs VMs are up now."

.github/workflows/scripts/qemu-6-tests.sh

@@ -111,7 +111,7 @@ fi
 sudo dmesg -c > dmesg-prerun.txt
 mount > mount.txt
 df -h > df-prerun.txt
-$TDIR/zfs-tests.sh -vK -s 3GB -T $TAGS
+$TDIR/zfs-tests.sh -vKO -s 3GB -T $TAGS
 RV=$?
 df -h > df-postrun.txt
 echo $RV > tests-exitcode.txt

.github/workflows/zfs-qemu.yml

@@ -44,7 +44,7 @@ jobs:
             os_selection="$FULL_OS"
           fi
 
-          if [ ${{ github.event.inputs.fedora_kernel_ver }} != "" ] ; then
+          if ${{ github.event.inputs.fedora_kernel_ver != '' }}; then
               # They specified a custom kernel version for Fedora.  Use only
               # Fedora runners.
               os_json=$(echo ${os_selection} | jq -c '[.[] | select(startswith("fedora"))]')
@@ -53,9 +53,8 @@ jobs:
               os_json=$(echo ${os_selection} | jq -c)
           fi
 
-          echo $os_json
+          echo "os=$os_json" | tee -a $GITHUB_OUTPUT
-          echo "os=$os_json" >> $GITHUB_OUTPUT
+          echo "ci_type=$ci_type" | tee -a $GITHUB_OUTPUT
-          echo "ci_type=$ci_type" >> $GITHUB_OUTPUT
 
   qemu-vm:
     name: qemu-x86
@@ -78,8 +77,12 @@ jobs:
         ref: ${{ github.event.pull_request.head.sha }}
 
     - name: Setup QEMU
-      timeout-minutes: 10
+      timeout-minutes: 20
-      run: .github/workflows/scripts/qemu-1-setup.sh
+      run: |
+        # Add a timestamp to each line to debug timeouts
+        while IFS=$'\n' read -r line; do
+            echo "$(date +'%H:%M:%S') $line"
+        done < <(.github/workflows/scripts/qemu-1-setup.sh)
 
     - name: Start build machine
       timeout-minutes: 10

META

@@ -1,10 +1,10 @@
 Meta:          1
 Name:          zfs
 Branch:        1.0
-Version:       2.4.0
+Version:       2.4.99
-Release:       rc1
+Release:       1
 Release-Tags:  relext
 License:       CDDL
 Author:        OpenZFS
-Linux-Maximum: 6.16
+Linux-Maximum: 6.17
 Linux-Minimum: 4.18

Makefile.am

@@ -1,6 +1,7 @@
 CLEANFILES =
 dist_noinst_DATA =
 INSTALL_DATA_HOOKS =
+INSTALL_EXEC_HOOKS =
 ALL_LOCAL =
 CLEAN_LOCAL =
 CHECKS = shellcheck checkbashisms
@@ -71,6 +72,9 @@ all: gitrev
 PHONY += install-data-hook $(INSTALL_DATA_HOOKS)
 install-data-hook: $(INSTALL_DATA_HOOKS)
 
+PHONY += install-exec-hook $(INSTALL_EXEC_HOOKS)
+install-exec-hook: $(INSTALL_EXEC_HOOKS)
+
 PHONY += maintainer-clean-local
 maintainer-clean-local:
 	-$(RM) $(GITREV)

cmd/Makefile.am

@@ -98,17 +98,16 @@ endif
 
 
 if USING_PYTHON
-bin_SCRIPTS      += arc_summary     arcstat        dbufstat        zilstat
+bin_SCRIPTS      += zarcsummary     zarcstat        dbufstat        zilstat
-CLEANFILES       += arc_summary     arcstat        dbufstat        zilstat
+CLEANFILES       += zarcsummary     zarcstat        dbufstat        zilstat
-dist_noinst_DATA += %D%/arc_summary %D%/arcstat.in %D%/dbufstat.in %D%/zilstat.in
+dist_noinst_DATA += %D%/zarcsummary %D%/zarcstat.in %D%/dbufstat.in %D%/zilstat.in
 
-$(call SUBST,arcstat,%D%/)
+$(call SUBST,zarcstat,%D%/)
 $(call SUBST,dbufstat,%D%/)
 $(call SUBST,zilstat,%D%/)
-arc_summary: %D%/arc_summary
+zarcsummary: %D%/zarcsummary
 	$(AM_V_at)cp $< $@
 endif
 
-
 PHONY += cmd
 cmd: $(bin_SCRIPTS) $(bin_PROGRAMS) $(sbin_SCRIPTS) $(sbin_PROGRAMS) $(dist_bin_SCRIPTS) $(zfsexec_PROGRAMS) $(mounthelper_PROGRAMS)

cmd/zarcstat.in

@@ -2,7 +2,7 @@
 # SPDX-License-Identifier: CDDL-1.0
 #
 # Print out ZFS ARC Statistics exported via kstat(1)
-# For a definition of fields, or usage, use arcstat -v
+# For a definition of fields, or usage, use zarcstat -v
 #
 # This script was originally a fork of the original arcstat.pl (0.1)
 # by Neelakanth Nadgir, originally published on his Sun blog on
@@ -56,6 +56,7 @@ import time
 import getopt
 import re
 import copy
+import os
 
 from signal import signal, SIGINT, SIGWINCH, SIG_DFL
 
@@ -171,7 +172,7 @@ cols = {
     "zactive":    [7, 1000, "zfetch prefetches active per second"],
 }
 
-# ARC structural breakdown from arc_summary
+# ARC structural breakdown from zarcsummary
 structfields = {
     "cmp":      ["compressed", "Compressed"],
     "ovh":      ["overhead", "Overhead"],
@@ -187,7 +188,7 @@ structstats = {                             # size stats
     "sz":       ["_size", "size"],
 }
 
-# ARC types breakdown from arc_summary
+# ARC types breakdown from zarcsummary
 typefields = {
     "data":     ["data", "ARC data"],
     "meta":     ["metadata", "ARC metadata"],
@@ -198,7 +199,7 @@ typestats = {                               # size stats
     "sz":       ["_size", "size"],
 }
 
-# ARC states breakdown from arc_summary
+# ARC states breakdown from zarcsummary
 statefields = {
     "ano":      ["anon", "Anonymous"],
     "mfu":      ["mfu", "MFU"],
@@ -261,7 +262,7 @@ hdr_intr = 20          # Print header every 20 lines of output
 opfile = None
 sep = "  "              # Default separator is 2 spaces
 l2exist = False
-cmd = ("Usage: arcstat [-havxp] [-f fields] [-o file] [-s string] [interval "
+cmd = ("Usage: zarcstat [-havxp] [-f fields] [-o file] [-s string] [interval "
        "[count]]\n")
 cur = {}
 d = {}
@@ -348,10 +349,10 @@ def usage():
                      "character or string\n")
     sys.stderr.write("\t -p : Disable auto-scaling of numerical fields\n")
     sys.stderr.write("\nExamples:\n")
-    sys.stderr.write("\tarcstat -o /tmp/a.log 2 10\n")
+    sys.stderr.write("\tzarcstat -o /tmp/a.log 2 10\n")
-    sys.stderr.write("\tarcstat -s \",\" -o /tmp/a.log 2 10\n")
+    sys.stderr.write("\tzarcstat -s \",\" -o /tmp/a.log 2 10\n")
-    sys.stderr.write("\tarcstat -v\n")
+    sys.stderr.write("\tzarcstat -v\n")
-    sys.stderr.write("\tarcstat -f time,hit%,dh%,ph%,mh% 1\n")
+    sys.stderr.write("\tzarcstat -f time,hit%,dh%,ph%,mh% 1\n")
     sys.stderr.write("\n")
 
     sys.exit(1)
@@ -366,7 +367,7 @@ def snap_stats():
 
     cur = kstat
 
-    # fill in additional values from arc_summary
+    # fill in additional values from zarcsummary
     cur["caches_size"] = caches_size = cur["anon_data"]+cur["anon_metadata"]+\
         cur["mfu_data"]+cur["mfu_metadata"]+cur["mru_data"]+cur["mru_metadata"]+\
         cur["uncached_data"]+cur["uncached_metadata"]
@@ -766,6 +767,7 @@ def calculate():
 
 
 def main():
+
     global sint
     global count
     global hdr_intr

cmd/zarcsummary

@@ -34,7 +34,7 @@ Provides basic information on the ARC, its efficiency, the L2ARC (if present),
 the Data Management Unit (DMU), Virtual Devices (VDEVs), and tunables. See
 the in-source documentation and code at
 https://github.com/openzfs/zfs/blob/master/module/zfs/arc.c for details.
-The original introduction to arc_summary can be found at
+The original introduction to zarcsummary can be found at
 http://cuddletech.com/?p=454
 """
 
@@ -161,7 +161,7 @@ elif sys.platform.startswith('linux'):
         return get_params(TUNABLES_PATH)
 
     def get_version_impl(request):
-        # The original arc_summary called /sbin/modinfo/{spl,zfs} to get
+        # The original zarcsummary called /sbin/modinfo/{spl,zfs} to get
         # the version information. We switch to /sys/module/{spl,zfs}/version
         # to make sure we get what is really loaded in the kernel
         try:
@@ -439,7 +439,7 @@ def print_header():
     """
 
     # datetime is now recommended over time but we keep the exact formatting
-    # from the older version of arc_summary in case there are scripts
+    # from the older version of zarcsummary in case there are scripts
     # that expect it in this way
     daydate = time.strftime(DATE_FORMAT)
     spc_date = LINE_LENGTH-len(daydate)

cmd/zdb/zdb.c

@@ -107,7 +107,9 @@ extern uint_t zfs_reconstruct_indirect_combinations_max;
 extern uint_t zfs_btree_verify_intensity;
 
 static const char cmdname[] = "zdb";
-uint8_t dump_opt[256];
+uint8_t dump_opt[512];
+
+#define	ALLOCATED_OPT	256
 
 typedef void object_viewer_t(objset_t *, uint64_t, void *data, size_t size);
 
@@ -383,7 +385,7 @@ verify_livelist_allocs(metaslab_verify_t *mv, uint64_t txg,
 	sublivelist_verify_block_t svb = {{{0}}};
 	DVA_SET_VDEV(&svb.svb_dva, mv->mv_vdid);
 	DVA_SET_OFFSET(&svb.svb_dva, offset);
-	DVA_SET_ASIZE(&svb.svb_dva, size);
+	DVA_SET_ASIZE(&svb.svb_dva, 0);
 	zfs_btree_index_t where;
 	uint64_t end_offset = offset + size;
 
@@ -1666,6 +1668,16 @@ dump_metaslab_stats(metaslab_t *msp)
 	dump_histogram(rt->rt_histogram, ZFS_RANGE_TREE_HISTOGRAM_SIZE, 0);
 }
 
+static void
+dump_allocated(void *arg, uint64_t start, uint64_t size)
+{
+	uint64_t *off = arg;
+	if (*off != start)
+		(void) printf("ALLOC: %"PRIu64" %"PRIu64"\n", *off,
+		    start - *off);
+	*off = start + size;
+}
+
 static void
 dump_metaslab(metaslab_t *msp)
 {
@@ -1682,13 +1694,24 @@ dump_metaslab(metaslab_t *msp)
 	    (u_longlong_t)msp->ms_id, (u_longlong_t)msp->ms_start,
 	    (u_longlong_t)space_map_object(sm), freebuf);
 
-	if (dump_opt['m'] > 2 && !dump_opt['L']) {
+	if (dump_opt[ALLOCATED_OPT] ||
+	    (dump_opt['m'] > 2 && !dump_opt['L'])) {
 		mutex_enter(&msp->ms_lock);
 		VERIFY0(metaslab_load(msp));
+	}
+
+	if (dump_opt['m'] > 2 && !dump_opt['L']) {
 		zfs_range_tree_stat_verify(msp->ms_allocatable);
 		dump_metaslab_stats(msp);
-		metaslab_unload(msp);
+	}
-		mutex_exit(&msp->ms_lock);
+
+	if (dump_opt[ALLOCATED_OPT]) {
+		uint64_t off = msp->ms_start;
+		zfs_range_tree_walk(msp->ms_allocatable, dump_allocated,
+		    &off);
+		if (off != msp->ms_start + msp->ms_size)
+			(void) printf("ALLOC: %"PRIu64" %"PRIu64"\n", off,
+			    msp->ms_size - off);
 	}
 
 	if (dump_opt['m'] > 1 && sm != NULL &&
@@ -1703,6 +1726,12 @@ dump_metaslab(metaslab_t *msp)
 		    SPACE_MAP_HISTOGRAM_SIZE, sm->sm_shift);
 	}
 
+	if (dump_opt[ALLOCATED_OPT] ||
+	    (dump_opt['m'] > 2 && !dump_opt['L'])) {
+		metaslab_unload(msp);
+		mutex_exit(&msp->ms_lock);
+	}
+
 	if (vd->vdev_ops == &vdev_draid_ops)
 		ASSERT3U(msp->ms_size, <=, 1ULL << vd->vdev_ms_shift);
 	else
@@ -1739,8 +1768,9 @@ print_vdev_metaslab_header(vdev_t *vd)
 		}
 	}
 
-	(void) printf("\tvdev %10llu   %s",
+	(void) printf("\tvdev %10llu\t%s  metaslab shift %4llu",
-	    (u_longlong_t)vd->vdev_id, bias_str);
+	    (u_longlong_t)vd->vdev_id, bias_str,
+	    (u_longlong_t)vd->vdev_ms_shift);
 
 	if (ms_flush_data_obj != 0) {
 		(void) printf("   ms_unflushed_phys object %llu",
@@ -2634,8 +2664,8 @@ print_indirect(spa_t *spa, blkptr_t *bp, const zbookmark_phys_t *zb,
 		}
 		if (BP_GET_LEVEL(bp) != zb->zb_level) {
 			(void) printf(" (ERROR: Block pointer level "
-			    "(%llu) does not match bookmark level (%ld))",
+			    "(%llu) does not match bookmark level (%lld))",
-			    BP_GET_LEVEL(bp), zb->zb_level);
+			    BP_GET_LEVEL(bp), (longlong_t)zb->zb_level);
 			corruption_found = B_TRUE;
 		}
 	}
@@ -2688,8 +2718,8 @@ visit_indirect(spa_t *spa, const dnode_phys_t *dnp,
 			if (fill != BP_GET_FILL(bp)) {
 				(void) printf("%16llx: Block pointer "
 				    "fill (%llu) does not match calculated "
-				    "value (%lu)\n", offset, BP_GET_FILL(bp),
+				    "value (%llu)\n", offset, BP_GET_FILL(bp),
-				    fill);
+				    (u_longlong_t)fill);
 				corruption_found = B_TRUE;
 			}
 		}
@@ -3271,6 +3301,7 @@ zdb_derive_key(dsl_dir_t *dd, uint8_t *key_out)
 	uint64_t keyformat, salt, iters;
 	int i;
 	unsigned char c;
+	FILE *f;
 
 	VERIFY0(zap_lookup(dd->dd_pool->dp_meta_objset, dd->dd_crypto_obj,
 	    zfs_prop_to_name(ZFS_PROP_KEYFORMAT), sizeof (uint64_t),
@@ -3303,6 +3334,25 @@ zdb_derive_key(dsl_dir_t *dd, uint8_t *key_out)
 
 		break;
 
+	case ZFS_KEYFORMAT_RAW:
+		if ((f = fopen(key_material, "r")) == NULL)
+			return (B_FALSE);
+
+		if (fread(key_out, 1, WRAPPING_KEY_LEN, f) !=
+		    WRAPPING_KEY_LEN) {
+			(void) fclose(f);
+			return (B_FALSE);
+		}
+
+		/* Check the key length */
+		if (fgetc(f) != EOF) {
+			(void) fclose(f);
+			return (B_FALSE);
+		}
+
+		(void) fclose(f);
+		break;
+
 	default:
 		fatal("no support for key format %u\n",
 		    (unsigned int) keyformat);
@@ -9375,6 +9425,8 @@ main(int argc, char **argv)
 		{"all-reconstruction",	no_argument,		NULL, 'Y'},
 		{"livelist",		no_argument,		NULL, 'y'},
 		{"zstd-headers",	no_argument,		NULL, 'Z'},
+		{"allocated-map",	no_argument,		NULL,
+		    ALLOCATED_OPT},
 		{0, 0, 0, 0}
 	};
 
@@ -9405,6 +9457,7 @@ main(int argc, char **argv)
 		case 'u':
 		case 'y':
 		case 'Z':
+		case ALLOCATED_OPT:
 			dump_opt[c]++;
 			dump_all = 0;
 			break;

cmd/zdb/zdb.h

@@ -29,6 +29,6 @@
 #define	_ZDB_H
 
 void dump_intent_log(zilog_t *);
-extern uint8_t dump_opt[256];
+extern uint8_t dump_opt[512];
 
 #endif	/* _ZDB_H */

cmd/zdb/zdb_il.c

@@ -48,8 +48,6 @@
 
 #include "zdb.h"
 
-extern uint8_t dump_opt[256];
-
 static char tab_prefix[4] = "\t\t\t";
 
 static void

cmd/zed/zed.d/Makefile.am

@@ -9,18 +9,18 @@ dist_zedexec_SCRIPTS = \
 	%D%/all-debug.sh \
 	%D%/all-syslog.sh \
 	%D%/data-notify.sh \
-	%D%/deadman-slot_off.sh \
+	%D%/deadman-sync-slot_off.sh \
 	%D%/generic-notify.sh \
-	%D%/pool_import-led.sh \
+	%D%/pool_import-sync-led.sh \
 	%D%/resilver_finish-notify.sh \
 	%D%/resilver_finish-start-scrub.sh \
 	%D%/scrub_finish-notify.sh \
-	%D%/statechange-led.sh \
+	%D%/statechange-sync-led.sh \
 	%D%/statechange-notify.sh \
-	%D%/statechange-slot_off.sh \
+	%D%/statechange-sync-slot_off.sh \
 	%D%/trim_finish-notify.sh \
-	%D%/vdev_attach-led.sh \
+	%D%/vdev_attach-sync-led.sh \
-	%D%/vdev_clear-led.sh
+	%D%/vdev_clear-sync-led.sh
 
 nodist_zedexec_SCRIPTS = \
 	%D%/history_event-zfs-list-cacher.sh
@@ -30,17 +30,17 @@ SUBSTFILES += $(nodist_zedexec_SCRIPTS)
 zedconfdefaults = \
 	all-syslog.sh \
 	data-notify.sh \
-	deadman-slot_off.sh \
+	deadman-sync-slot_off.sh \
 	history_event-zfs-list-cacher.sh \
-	pool_import-led.sh \
+	pool_import-sync-led.sh \
 	resilver_finish-notify.sh \
 	resilver_finish-start-scrub.sh \
 	scrub_finish-notify.sh \
-	statechange-led.sh \
+	statechange-sync-led.sh \
 	statechange-notify.sh \
-	statechange-slot_off.sh \
+	statechange-sync-slot_off.sh \
-	vdev_attach-led.sh \
+	vdev_attach-sync-led.sh \
-	vdev_clear-led.sh
+	vdev_clear-sync-led.sh
 
 dist_noinst_DATA += %D%/README
 

cmd/zed/zed.d/pool_import-led.sh

@@ -1 +0,0 @@
-statechange-led.sh

cmd/zed/zed.d/pool_import-sync-led.sh

@@ -0,0 +1 @@
+statechange-sync-led.sh

cmd/zed/zed.d/vdev_attach-led.sh

@@ -1 +0,0 @@
-statechange-led.sh

cmd/zed/zed.d/vdev_attach-sync-led.sh

@@ -0,0 +1 @@
+statechange-sync-led.sh

cmd/zed/zed.d/vdev_clear-led.sh

@@ -1 +0,0 @@
-statechange-led.sh

cmd/zed/zed.d/vdev_clear-sync-led.sh

@@ -0,0 +1 @@
+statechange-sync-led.sh

cmd/zed/zed_exec.c

@@ -196,37 +196,29 @@ _nop(int sig)
 	(void) sig;
 }
 
-static void *
+static void
-_reap_children(void *arg)
+wait_for_children(boolean_t do_pause, boolean_t wait)
 {
-	(void) arg;
-	struct launched_process_node node, *pnode;
 	pid_t pid;
-	int status;
 	struct rusage usage;
-	struct sigaction sa = {};
+	int status;
-
+	struct launched_process_node node, *pnode;
-	(void) sigfillset(&sa.sa_mask);
-	(void) sigdelset(&sa.sa_mask, SIGCHLD);
-	(void) pthread_sigmask(SIG_SETMASK, &sa.sa_mask, NULL);
-
-	(void) sigemptyset(&sa.sa_mask);
-	sa.sa_handler = _nop;
-	sa.sa_flags = SA_NOCLDSTOP;
-	(void) sigaction(SIGCHLD, &sa, NULL);
 
 	for (_reap_children_stop = B_FALSE; !_reap_children_stop; ) {
 		(void) pthread_mutex_lock(&_launched_processes_lock);
-		pid = wait4(0, &status, WNOHANG, &usage);
+		pid = wait4(0, &status, wait ? 0 : WNOHANG, &usage);
-
 		if (pid == 0 || pid == (pid_t)-1) {
 			(void) pthread_mutex_unlock(&_launched_processes_lock);
-			if (pid == 0 || errno == ECHILD)
+			if ((pid == 0) || (errno == ECHILD)) {
-				pause();
+				if (do_pause)
-			else if (errno != EINTR)
+					pause();
+			} else if (errno != EINTR)
 				zed_log_msg(LOG_WARNING,
 				    "Failed to wait for children: %s",
 				    strerror(errno));
+			if (!do_pause)
+				return;
+
 		} else {
 			memset(&node, 0, sizeof (node));
 			node.pid = pid;
@@ -278,6 +270,25 @@ _reap_children(void *arg)
 		}
 	}
 
+}
+
+static void *
+_reap_children(void *arg)
+{
+	(void) arg;
+	struct sigaction sa = {};
+
+	(void) sigfillset(&sa.sa_mask);
+	(void) sigdelset(&sa.sa_mask, SIGCHLD);
+	(void) pthread_sigmask(SIG_SETMASK, &sa.sa_mask, NULL);
+
+	(void) sigemptyset(&sa.sa_mask);
+	sa.sa_handler = _nop;
+	sa.sa_flags = SA_NOCLDSTOP;
+	(void) sigaction(SIGCHLD, &sa, NULL);
+
+	wait_for_children(B_TRUE, B_FALSE);
+
 	return (NULL);
 }
 
@@ -306,6 +317,45 @@ zed_exec_fini(void)
 	_reap_children_tid = (pthread_t)-1;
 }
 
+/*
+ * Check if the zedlet name indicates if it is a synchronous zedlet
+ *
+ * Synchronous zedlets have a "-sync-" immediately following the event name in
+ * their zedlet filename, like:
+ *
+ * EVENT_NAME-sync-ZEDLETNAME.sh
+ *
+ * For example, if you wanted a synchronous statechange script:
+ *
+ * statechange-sync-myzedlet.sh
+ *
+ * Synchronous zedlets are guaranteed to be the only zedlet running.  No other
+ * zedlets may run in parallel with a synchronous zedlet.  A synchronous
+ * zedlet will wait for all previously spawned zedlets to finish before running.
+ * Users should be careful to only use synchronous zedlets when needed, since
+ * they decrease parallelism.
+ */
+static boolean_t
+zedlet_is_sync(const char *zedlet, const char *event)
+{
+	const char *sync_str = "-sync-";
+	size_t sync_str_len;
+	size_t zedlet_len;
+	size_t event_len;
+
+	sync_str_len = strlen(sync_str);
+	zedlet_len = strlen(zedlet);
+	event_len = strlen(event);
+
+	if (event_len + sync_str_len >= zedlet_len)
+		return (B_FALSE);
+
+	if (strncmp(&zedlet[event_len], sync_str, sync_str_len) == 0)
+		return (B_TRUE);
+
+	return (B_FALSE);
+}
+
 /*
  * Process the event [eid] by synchronously invoking all zedlets with a
  * matching class prefix.
@@ -368,9 +418,28 @@ zed_exec_process(uint64_t eid, const char *class, const char *subclass,
 	    z = zed_strings_next(zcp->zedlets)) {
 		for (csp = class_strings; *csp; csp++) {
 			n = strlen(*csp);
-			if ((strncmp(z, *csp, n) == 0) && !isalpha(z[n]))
+			if ((strncmp(z, *csp, n) == 0) && !isalpha(z[n])) {
+				boolean_t is_sync = zedlet_is_sync(z, *csp);
+
+				if (is_sync) {
+					/*
+					 * Wait for previous zedlets to
+					 * finish
+					 */
+					wait_for_children(B_FALSE, B_TRUE);
+				}
+
 				_zed_exec_fork_child(eid, zcp->zedlet_dir,
 				    z, e, zcp->zevent_fd, zcp->do_foreground);
+
+				if (is_sync) {
+					/*
+					 * Wait for sync zedlet we just launched
+					 * to finish.
+					 */
+					wait_for_children(B_FALSE, B_TRUE);
+				}
+			}
 		}
 	}
 	free(e);

cmd/zfs/zfs_main.c

@@ -930,19 +930,15 @@ usage:
 }
 
 /*
- * Return a default volblocksize for the pool which always uses more than
+ * Calculate the minimum allocation size based on the top-level vdevs.
- * half of the data sectors.  This primarily applies to dRAID which always
- * writes full stripe widths.
  */
 static uint64_t
-default_volblocksize(zpool_handle_t *zhp, nvlist_t *props)
+calculate_volblocksize(nvlist_t *config)
 {
-	uint64_t volblocksize, asize = SPA_MINBLOCKSIZE;
+	uint64_t asize = SPA_MINBLOCKSIZE;
 	nvlist_t *tree, **vdevs;
 	uint_t nvdevs;
 
-	nvlist_t *config = zpool_get_config(zhp, NULL);
-
 	if (nvlist_lookup_nvlist(config, ZPOOL_CONFIG_VDEV_TREE, &tree) != 0 ||
 	    nvlist_lookup_nvlist_array(tree, ZPOOL_CONFIG_CHILDREN,
 	    &vdevs, &nvdevs) != 0) {
@@ -973,6 +969,24 @@ default_volblocksize(zpool_handle_t *zhp, nvlist_t *props)
 		}
 	}
 
+	return (asize);
+}
+
+/*
+ * Return a default volblocksize for the pool which always uses more than
+ * half of the data sectors.  This primarily applies to dRAID which always
+ * writes full stripe widths.
+ */
+static uint64_t
+default_volblocksize(zpool_handle_t *zhp, nvlist_t *props)
+{
+	uint64_t volblocksize, asize = SPA_MINBLOCKSIZE;
+
+	nvlist_t *config = zpool_get_config(zhp, NULL);
+
+	if (nvlist_lookup_uint64(config, ZPOOL_CONFIG_MAX_ALLOC, &asize) != 0)
+		asize = calculate_volblocksize(config);
+
 	/*
 	 * Calculate the target volblocksize such that more than half
 	 * of the asize is used. The following table is for 4k sectors.
@@ -5303,6 +5317,7 @@ zfs_do_receive(int argc, char **argv)
 #define	ZFS_DELEG_PERM_MOUNT		"mount"
 #define	ZFS_DELEG_PERM_SHARE		"share"
 #define	ZFS_DELEG_PERM_SEND		"send"
+#define	ZFS_DELEG_PERM_SEND_RAW		"send:raw"
 #define	ZFS_DELEG_PERM_RECEIVE		"receive"
 #define	ZFS_DELEG_PERM_RECEIVE_APPEND	"receive:append"
 #define	ZFS_DELEG_PERM_ALLOW		"allow"
@@ -5345,6 +5360,7 @@ static zfs_deleg_perm_tab_t zfs_deleg_perm_tbl[] = {
 	{ ZFS_DELEG_PERM_RENAME, ZFS_DELEG_NOTE_RENAME },
 	{ ZFS_DELEG_PERM_ROLLBACK, ZFS_DELEG_NOTE_ROLLBACK },
 	{ ZFS_DELEG_PERM_SEND, ZFS_DELEG_NOTE_SEND },
+	{ ZFS_DELEG_PERM_SEND_RAW, ZFS_DELEG_NOTE_SEND_RAW },
 	{ ZFS_DELEG_PERM_SHARE, ZFS_DELEG_NOTE_SHARE },
 	{ ZFS_DELEG_PERM_SNAPSHOT, ZFS_DELEG_NOTE_SNAPSHOT },
 	{ ZFS_DELEG_PERM_BOOKMARK, ZFS_DELEG_NOTE_BOOKMARK },
@@ -5929,6 +5945,10 @@ deleg_perm_comment(zfs_deleg_note_t note)
 	case ZFS_DELEG_NOTE_SEND:
 		str = gettext("");
 		break;
+	case ZFS_DELEG_NOTE_SEND_RAW:
+		str = gettext("Allow sending ONLY encrypted (raw) replication"
+		    "\n\t\t\t\tstreams");
+		break;
 	case ZFS_DELEG_NOTE_SHARE:
 		str = gettext("Allows sharing file systems over NFS or SMB"
 		    "\n\t\t\t\tprotocols");
@@ -6858,17 +6878,17 @@ print_holds(boolean_t scripted, int nwidth, int tagwidth, nvlist_t *nvl,
 
 			if (scripted) {
 				if (parsable) {
-					(void) printf("%s\t%s\t%ld\n", zname,
+					(void) printf("%s\t%s\t%lld\n", zname,
-					    tagname, time);
+					    tagname, (long long)time);
 				} else {
 					(void) printf("%s\t%s\t%s\n", zname,
 					    tagname, tsbuf);
 				}
 			} else {
 				if (parsable) {
-					(void) printf("%-*s  %-*s  %ld\n",
+					(void) printf("%-*s  %-*s  %lld\n",
 					    nwidth, zname, tagwidth,
-					    tagname, time);
+					    tagname, (long long)time);
 				} else {
 					(void) printf("%-*s  %-*s  %s\n",
 					    nwidth, zname, tagwidth,

cmd/zhack.c

@@ -54,6 +54,7 @@
 #include <sys/dmu_tx.h>
 #include <zfeature_common.h>
 #include <libzutil.h>
+#include <sys/metaslab_impl.h>
 
 static importargs_t g_importargs;
 static char *g_pool;
@@ -69,7 +70,8 @@ static __attribute__((noreturn)) void
 usage(void)
 {
 	(void) fprintf(stderr,
-	    "Usage: zhack [-c cachefile] [-d dir] <subcommand> <args> ...\n"
+	    "Usage: zhack [-o tunable] [-c cachefile] [-d dir] <subcommand> "
+	    "<args> ...\n"
 	    "where <subcommand> <args> is one of the following:\n"
 	    "\n");
 
@@ -93,7 +95,10 @@ usage(void)
 	    "        -c repair corrupted label checksums\n"
 	    "        -u restore the label on a detached device\n"
 	    "\n"
-	    "    <device> : path to vdev\n");
+	    "    <device> : path to vdev\n"
+	    "\n"
+	    "    metaslab leak <pool>\n"
+	    "        apply allocation map from zdb to specified pool\n");
 	exit(1);
 }
 
@@ -363,10 +368,12 @@ feature_incr_sync(void *arg, dmu_tx_t *tx)
 	zfeature_info_t *feature = arg;
 	uint64_t refcount;
 
+	mutex_enter(&spa->spa_feat_stats_lock);
 	VERIFY0(feature_get_refcount_from_disk(spa, feature, &refcount));
 	feature_sync(spa, feature, refcount + 1, tx);
 	spa_history_log_internal(spa, "zhack feature incr", tx,
 	    "name=%s", feature->fi_guid);
+	mutex_exit(&spa->spa_feat_stats_lock);
 }
 
 static void
@@ -376,10 +383,12 @@ feature_decr_sync(void *arg, dmu_tx_t *tx)
 	zfeature_info_t *feature = arg;
 	uint64_t refcount;
 
+	mutex_enter(&spa->spa_feat_stats_lock);
 	VERIFY0(feature_get_refcount_from_disk(spa, feature, &refcount));
 	feature_sync(spa, feature, refcount - 1, tx);
 	spa_history_log_internal(spa, "zhack feature decr", tx,
 	    "name=%s", feature->fi_guid);
+	mutex_exit(&spa->spa_feat_stats_lock);
 }
 
 static void
@@ -496,6 +505,186 @@ zhack_do_feature(int argc, char **argv)
 	return (0);
 }
 
+static boolean_t
+strstarts(const char *a, const char *b)
+{
+	return (strncmp(a, b, strlen(b)) == 0);
+}
+
+static void
+metaslab_force_alloc(metaslab_t *msp, uint64_t start, uint64_t size,
+    dmu_tx_t *tx)
+{
+	ASSERT(msp->ms_disabled);
+	ASSERT(MUTEX_HELD(&msp->ms_lock));
+	uint64_t txg = dmu_tx_get_txg(tx);
+
+	uint64_t off = start;
+	while (off < start + size) {
+		uint64_t ostart, osize;
+		boolean_t found = zfs_range_tree_find_in(msp->ms_allocatable,
+		    off, start + size - off, &ostart, &osize);
+		if (!found)
+			break;
+		zfs_range_tree_remove(msp->ms_allocatable, ostart, osize);
+
+		if (zfs_range_tree_is_empty(msp->ms_allocating[txg & TXG_MASK]))
+			vdev_dirty(msp->ms_group->mg_vd, VDD_METASLAB, msp,
+			    txg);
+
+		zfs_range_tree_add(msp->ms_allocating[txg & TXG_MASK], ostart,
+		    osize);
+		msp->ms_allocating_total += osize;
+		off = ostart + osize;
+	}
+}
+
+static void
+zhack_do_metaslab_leak(int argc, char **argv)
+{
+	int c;
+	char *target;
+	spa_t *spa;
+
+	optind = 1;
+	boolean_t force = B_FALSE;
+	while ((c = getopt(argc, argv, "f")) != -1) {
+		switch (c) {
+		case 'f':
+			force = B_TRUE;
+			break;
+		default:
+			usage();
+			break;
+		}
+	}
+
+	argc -= optind;
+	argv += optind;
+
+	if (argc < 1) {
+		(void) fprintf(stderr, "error: missing pool name\n");
+		usage();
+	}
+	target = argv[0];
+
+	zhack_spa_open(target, B_FALSE, FTAG, &spa);
+	spa_config_enter(spa, SCL_VDEV | SCL_ALLOC, FTAG, RW_READER);
+
+	char *line = NULL;
+	size_t cap = 0;
+
+	vdev_t *vd = NULL;
+	metaslab_t *prev = NULL;
+	dmu_tx_t *tx = NULL;
+	while (getline(&line, &cap, stdin) > 0) {
+		if (strstarts(line, "\tvdev ")) {
+			uint64_t vdev_id, ms_shift;
+			if (sscanf(line,
+			    "\tvdev %10"PRIu64"\t%*s  metaslab shift %4"PRIu64,
+			    &vdev_id, &ms_shift) == 1) {
+				VERIFY3U(sscanf(line, "\tvdev %"PRIu64
+				    "\t  metaslab shift %4"PRIu64,
+				    &vdev_id, &ms_shift), ==, 2);
+			}
+			vd = vdev_lookup_top(spa, vdev_id);
+			if (vd == NULL) {
+				fprintf(stderr, "error: no such vdev with "
+				    "id %"PRIu64"\n", vdev_id);
+				break;
+			}
+			if (tx) {
+				dmu_tx_commit(tx);
+				mutex_exit(&prev->ms_lock);
+				metaslab_enable(prev, B_FALSE, B_FALSE);
+				tx = NULL;
+				prev = NULL;
+			}
+			if (vd->vdev_ms_shift != ms_shift) {
+				fprintf(stderr, "error: ms_shift mismatch: %"
+				    PRIu64" != %"PRIu64"\n", vd->vdev_ms_shift,
+				    ms_shift);
+				break;
+			}
+		} else if (strstarts(line, "\tmetaslabs ")) {
+			uint64_t ms_count;
+			VERIFY3U(sscanf(line, "\tmetaslabs %"PRIu64, &ms_count),
+			    ==, 1);
+			ASSERT(vd);
+			if (!force && vd->vdev_ms_count != ms_count) {
+				fprintf(stderr, "error: ms_count mismatch: %"
+				    PRIu64" != %"PRIu64"\n", vd->vdev_ms_count,
+				    ms_count);
+				break;
+			}
+		} else if (strstarts(line, "ALLOC:")) {
+			uint64_t start, size;
+			VERIFY3U(sscanf(line, "ALLOC: %"PRIu64" %"PRIu64"\n",
+			    &start, &size), ==, 2);
+
+			ASSERT(vd);
+			metaslab_t *cur =
+			    vd->vdev_ms[start >> vd->vdev_ms_shift];
+			if (prev != cur) {
+				if (prev) {
+					dmu_tx_commit(tx);
+					mutex_exit(&prev->ms_lock);
+					metaslab_enable(prev, B_FALSE, B_FALSE);
+				}
+				ASSERT(cur);
+				metaslab_disable(cur);
+				mutex_enter(&cur->ms_lock);
+				metaslab_load(cur);
+				prev = cur;
+				tx = dmu_tx_create_dd(
+				    spa_get_dsl(vd->vdev_spa)->dp_root_dir);
+				dmu_tx_assign(tx, DMU_TX_WAIT);
+			}
+
+			metaslab_force_alloc(cur, start, size, tx);
+		} else {
+			continue;
+		}
+	}
+	if (tx) {
+		dmu_tx_commit(tx);
+		mutex_exit(&prev->ms_lock);
+		metaslab_enable(prev, B_FALSE, B_FALSE);
+		tx = NULL;
+		prev = NULL;
+	}
+	if (line)
+		free(line);
+
+	spa_config_exit(spa, SCL_VDEV | SCL_ALLOC, FTAG);
+	spa_close(spa, FTAG);
+}
+
+static int
+zhack_do_metaslab(int argc, char **argv)
+{
+	char *subcommand;
+
+	argc--;
+	argv++;
+	if (argc == 0) {
+		(void) fprintf(stderr,
+		    "error: no metaslab operation specified\n");
+		usage();
+	}
+
+	subcommand = argv[0];
+	if (strcmp(subcommand, "leak") == 0) {
+		zhack_do_metaslab_leak(argc, argv);
+	} else {
+		(void) fprintf(stderr, "error: unknown subcommand: %s\n",
+		    subcommand);
+		usage();
+	}
+
+	return (0);
+}
+
 #define	ASHIFT_UBERBLOCK_SHIFT(ashift)	\
 	MIN(MAX(ashift, UBERBLOCK_SHIFT), \
 	MAX_UBERBLOCK_SHIFT)
@@ -525,6 +714,23 @@ zhack_repair_read_label(const int fd, vdev_label_t *vl,
 	return (0);
 }
 
+static int
+zhack_repair_get_byteswap(const zio_eck_t *vdev_eck, const int l, int *byteswap)
+{
+	if (vdev_eck->zec_magic == ZEC_MAGIC) {
+		*byteswap = B_FALSE;
+	} else if (vdev_eck->zec_magic == BSWAP_64((uint64_t)ZEC_MAGIC)) {
+		*byteswap = B_TRUE;
+	} else {
+		(void) fprintf(stderr, "error: label %d: "
+		    "Expected the nvlist checksum magic number but instead got "
+		    "0x%" PRIx64 "\n",
+		    l, vdev_eck->zec_magic);
+		return (1);
+	}
+	return (0);
+}
+
 static void
 zhack_repair_calc_cksum(const int byteswap, void *data, const uint64_t offset,
     const uint64_t abdsize, zio_eck_t *eck, zio_cksum_t *cksum)
@@ -551,33 +757,10 @@ zhack_repair_calc_cksum(const int byteswap, void *data, const uint64_t offset,
 }
 
 static int
-zhack_repair_check_label(uberblock_t *ub, const int l, const char **cfg_keys,
+zhack_repair_get_ashift(nvlist_t *cfg, const int l, uint64_t *ashift)
-    const size_t cfg_keys_len, nvlist_t *cfg, nvlist_t *vdev_tree_cfg,
-    uint64_t *ashift)
 {
 	int err;
-
+	nvlist_t *vdev_tree_cfg;
-	if (ub->ub_txg != 0) {
-		(void) fprintf(stderr,
-		    "error: label %d: UB TXG of 0 expected, but got %"
-		    PRIu64 "\n",
-		    l, ub->ub_txg);
-		(void) fprintf(stderr, "It would appear the device was not "
-		    "properly removed.\n");
-		return (1);
-	}
-
-	for (int i = 0; i < cfg_keys_len; i++) {
-		uint64_t val;
-		err = nvlist_lookup_uint64(cfg, cfg_keys[i], &val);
-		if (err) {
-			(void) fprintf(stderr,
-			    "error: label %d, %d: "
-			    "cannot find nvlist key %s\n",
-			    l, i, cfg_keys[i]);
-			return (err);
-		}
-	}
 
 	err = nvlist_lookup_nvlist(cfg,
 	    ZPOOL_CONFIG_VDEV_TREE, &vdev_tree_cfg);
@@ -601,7 +784,7 @@ zhack_repair_check_label(uberblock_t *ub, const int l, const char **cfg_keys,
 		(void) fprintf(stderr,
 		    "error: label %d: nvlist key %s is zero\n",
 		    l, ZPOOL_CONFIG_ASHIFT);
-		return (err);
+		return (1);
 	}
 
 	return (0);
@@ -616,30 +799,35 @@ zhack_repair_undetach(uberblock_t *ub, nvlist_t *cfg, const int l)
 	 */
 	if (BP_GET_LOGICAL_BIRTH(&ub->ub_rootbp) != 0) {
 		const uint64_t txg = BP_GET_LOGICAL_BIRTH(&ub->ub_rootbp);
+		int err;
+
 		ub->ub_txg = txg;
 
-		if (nvlist_remove_all(cfg, ZPOOL_CONFIG_CREATE_TXG) != 0) {
+		err = nvlist_remove_all(cfg, ZPOOL_CONFIG_CREATE_TXG);
+		if (err) {
 			(void) fprintf(stderr,
 			    "error: label %d: "
 			    "Failed to remove pool creation TXG\n",
 			    l);
-			return (1);
+			return (err);
 		}
 
-		if (nvlist_remove_all(cfg, ZPOOL_CONFIG_POOL_TXG) != 0) {
+		err = nvlist_remove_all(cfg, ZPOOL_CONFIG_POOL_TXG);
+		if (err) {
 			(void) fprintf(stderr,
 			    "error: label %d: Failed to remove pool TXG to "
 			    "be replaced.\n",
 			    l);
-			return (1);
+			return (err);
 		}
 
-		if (nvlist_add_uint64(cfg, ZPOOL_CONFIG_POOL_TXG, txg) != 0) {
+		err = nvlist_add_uint64(cfg, ZPOOL_CONFIG_POOL_TXG, txg);
+		if (err) {
 			(void) fprintf(stderr,
 			    "error: label %d: "
 			    "Failed to add pool TXG of %" PRIu64 "\n",
 			    l, txg);
-			return (1);
+			return (err);
 		}
 	}
 
@@ -733,6 +921,7 @@ zhack_repair_test_cksum(const int byteswap, void *vdev_data,
 	    BSWAP_64(ZEC_MAGIC) : ZEC_MAGIC;
 	const uint64_t actual_magic = vdev_eck->zec_magic;
 	int err = 0;
+
 	if (actual_magic != expected_magic) {
 		(void) fprintf(stderr, "error: label %d: "
 		    "Expected "
@@ -754,6 +943,36 @@ zhack_repair_test_cksum(const int byteswap, void *vdev_data,
 	return (err);
 }
 
+static int
+zhack_repair_unpack_cfg(vdev_label_t *vl, const int l, nvlist_t **cfg)
+{
+	const char *cfg_keys[] = { ZPOOL_CONFIG_VERSION,
+	    ZPOOL_CONFIG_POOL_STATE, ZPOOL_CONFIG_GUID };
+	int err;
+
+	err = nvlist_unpack(vl->vl_vdev_phys.vp_nvlist,
+	    VDEV_PHYS_SIZE - sizeof (zio_eck_t), cfg, 0);
+	if (err) {
+		(void) fprintf(stderr,
+		    "error: cannot unpack nvlist label %d\n", l);
+		return (err);
+	}
+
+	for (int i = 0; i < ARRAY_SIZE(cfg_keys); i++) {
+		uint64_t val;
+		err = nvlist_lookup_uint64(*cfg, cfg_keys[i], &val);
+		if (err) {
+			(void) fprintf(stderr,
+			    "error: label %d, %d: "
+			    "cannot find nvlist key %s\n",
+			    l, i, cfg_keys[i]);
+			return (err);
+		}
+	}
+
+	return (0);
+}
+
 static void
 zhack_repair_one_label(const zhack_repair_op_t op, const int fd,
     vdev_label_t *vl, const uint64_t label_offset, const int l,
@@ -767,10 +986,7 @@ zhack_repair_one_label(const zhack_repair_op_t op, const int fd,
 	    (zio_eck_t *)((char *)(vdev_data) + VDEV_PHYS_SIZE) - 1;
 	const uint64_t vdev_phys_offset =
 	    label_offset + offsetof(vdev_label_t, vl_vdev_phys);
-	const char *cfg_keys[] = { ZPOOL_CONFIG_VERSION,
-	    ZPOOL_CONFIG_POOL_STATE, ZPOOL_CONFIG_GUID };
 	nvlist_t *cfg;
-	nvlist_t *vdev_tree_cfg = NULL;
 	uint64_t ashift;
 	int byteswap;
 
@@ -778,18 +994,9 @@ zhack_repair_one_label(const zhack_repair_op_t op, const int fd,
 	if (err)
 		return;
 
-	if (vdev_eck->zec_magic == 0) {
+	err = zhack_repair_get_byteswap(vdev_eck, l, &byteswap);
-		(void) fprintf(stderr, "error: label %d: "
+	if (err)
-		    "Expected the nvlist checksum magic number to not be zero"
-		    "\n",
-		    l);
-		(void) fprintf(stderr, "There should already be a checksum "
-		    "for the label.\n");
 		return;
-	}
-
-	byteswap =
-	    (vdev_eck->zec_magic == BSWAP_64((uint64_t)ZEC_MAGIC));
 
 	if (byteswap) {
 		byteswap_uint64_array(&vdev_eck->zec_cksum,
@@ -805,16 +1012,7 @@ zhack_repair_one_label(const zhack_repair_op_t op, const int fd,
 		return;
 	}
 
-	err = nvlist_unpack(vl->vl_vdev_phys.vp_nvlist,
+	err = zhack_repair_unpack_cfg(vl, l, &cfg);
-	    VDEV_PHYS_SIZE - sizeof (zio_eck_t), &cfg, 0);
-	if (err) {
-		(void) fprintf(stderr,
-		    "error: cannot unpack nvlist label %d\n", l);
-		return;
-	}
-
-	err = zhack_repair_check_label(ub,
-	    l, cfg_keys, ARRAY_SIZE(cfg_keys), cfg, vdev_tree_cfg, &ashift);
 	if (err)
 		return;
 
@@ -822,6 +1020,19 @@ zhack_repair_one_label(const zhack_repair_op_t op, const int fd,
 		char *buf;
 		size_t buflen;
 
+		if (ub->ub_txg != 0) {
+			(void) fprintf(stderr,
+			    "error: label %d: UB TXG of 0 expected, but got %"
+			    PRIu64 "\n", l, ub->ub_txg);
+			(void) fprintf(stderr, "It would appear the device was "
+			    "not properly detached.\n");
+			return;
+		}
+
+		err = zhack_repair_get_ashift(cfg, l, &ashift);
+		if (err)
+			return;
+
 		err = zhack_repair_undetach(ub, cfg, l);
 		if (err)
 			return;
@@ -981,7 +1192,7 @@ main(int argc, char **argv)
 	dprintf_setup(&argc, argv);
 	zfs_prop_init();
 
-	while ((c = getopt(argc, argv, "+c:d:")) != -1) {
+	while ((c = getopt(argc, argv, "+c:d:o:")) != -1) {
 		switch (c) {
 		case 'c':
 			g_importargs.cachefile = optarg;
@@ -990,6 +1201,10 @@ main(int argc, char **argv)
 			assert(g_importargs.paths < MAX_NUM_PATHS);
 			g_importargs.path[g_importargs.paths++] = optarg;
 			break;
+		case 'o':
+			if (handle_tunable_option(optarg, B_FALSE) != 0)
+				exit(1);
+			break;
 		default:
 			usage();
 			break;
@@ -1011,6 +1226,8 @@ main(int argc, char **argv)
 		rv = zhack_do_feature(argc, argv);
 	} else if (strcmp(subcommand, "label") == 0) {
 		return (zhack_do_label(argc, argv));
+	} else if (strcmp(subcommand, "metaslab") == 0) {
+		rv = zhack_do_metaslab(argc, argv);
 	} else {
 		(void) fprintf(stderr, "error: unknown subcommand: %s\n",
 		    subcommand);

cmd/zinject/zinject.c

@@ -107,6 +107,8 @@
  * 	zinject
  * 	zinject <-a | -u pool>
  * 	zinject -c <id|all>
+ * 	zinject -E <delay> [-a] [-m] [-f freq] [-l level] [-r range]
+ *	    [-T iotype] [-t type object | -b bookmark pool]
  * 	zinject [-q] <-t type> [-f freq] [-u] [-a] [-m] [-e errno] [-l level]
  *	    [-r range] <object>
  * 	zinject [-f freq] [-a] [-m] [-u] -b objset:object:level:start:end pool
@@ -132,14 +134,18 @@
  * The '-f' flag controls the frequency of errors injected, expressed as a
  * real number percentage between 0.0001 and 100.  The default is 100.
  *
- * The this form is responsible for actually injecting the handler into the
+ * The <object> form is responsible for actually injecting the handler into the
  * framework.  It takes the arguments described above, translates them to the
  * internal tuple using libzpool, and then issues an ioctl() to register the
  * handler.
  *
- * The final form can target a specific bookmark, regardless of whether a
+ * The '-b' option can target a specific bookmark, regardless of whether a
  * human-readable interface has been designed.  It allows developers to specify
  * a particular block by number.
+ *
+ * The '-E' option injects pipeline ready stage delays for the given object or
+ * bookmark. The delay is specified in milliseconds, and it supports I/O type
+ * and range filters.
  */
 
 #include <errno.h>
@@ -346,6 +352,13 @@ usage(void)
 	    "\t\tsuch that the operation takes a minimum of supplied seconds\n"
 	    "\t\tto complete.\n"
 	    "\n"
+	    "\tzinject -E <delay> [-a] [-m] [-f freq] [-l level] [-r range]\n"
+	    "\t\t[-T iotype] [-t type object | -b bookmark pool]\n"
+	    "\n"
+	    "\t\tInject pipeline ready stage delays for the given object path\n"
+	    "\t\t(data or dnode) or raw bookmark. The delay is specified in\n"
+	    "\t\tmilliseconds.\n"
+	    "\n"
 	    "\tzinject -I [-s <seconds> | -g <txgs>] pool\n"
 	    "\t\tCause the pool to stop writing blocks yet not\n"
 	    "\t\treport errors for a duration.  Simulates buggy hardware\n"
@@ -724,12 +737,15 @@ register_handler(const char *pool, int flags, zinject_record_t *record,
 	if (quiet) {
 		(void) printf("%llu\n", (u_longlong_t)zc.zc_guid);
 	} else {
+		boolean_t show_object = B_FALSE;
+		boolean_t show_iotype = B_FALSE;
 		(void) printf("Added handler %llu with the following "
 		    "properties:\n", (u_longlong_t)zc.zc_guid);
 		(void) printf("  pool: %s\n", pool);
 		if (record->zi_guid) {
 			(void) printf("  vdev: %llx\n",
 			    (u_longlong_t)record->zi_guid);
+			show_iotype = B_TRUE;
 		} else if (record->zi_func[0] != '\0') {
 			(void) printf("  panic function: %s\n",
 			    record->zi_func);
@@ -742,7 +758,18 @@ register_handler(const char *pool, int flags, zinject_record_t *record,
 		} else if (record->zi_timer > 0) {
 			(void) printf(" timer: %lld ms\n",
 			    (u_longlong_t)NSEC2MSEC(record->zi_timer));
+			if (record->zi_cmd == ZINJECT_DELAY_READY) {
+				show_object = B_TRUE;
+				show_iotype = B_TRUE;
+			}
 		} else {
+			show_object = B_TRUE;
+		}
+		if (show_iotype) {
+			(void) printf("iotype: %s\n",
+			    iotype_to_str(record->zi_iotype));
+		}
+		if (show_object) {
 			(void) printf("objset: %llu\n",
 			    (u_longlong_t)record->zi_objset);
 			(void) printf("object: %llu\n",
@@ -910,6 +937,7 @@ main(int argc, char **argv)
 	int ret;
 	int flags = 0;
 	uint32_t dvas = 0;
+	hrtime_t ready_delay = -1;
 
 	if ((g_zfs = libzfs_init()) == NULL) {
 		(void) fprintf(stderr, "%s\n", libzfs_error_init(errno));
@@ -940,7 +968,7 @@ main(int argc, char **argv)
 	}
 
 	while ((c = getopt(argc, argv,
-	    ":aA:b:C:d:D:f:Fg:qhIc:t:T:l:mr:s:e:uL:p:P:")) != -1) {
+	    ":aA:b:C:d:D:E:f:Fg:qhIc:t:T:l:mr:s:e:uL:p:P:")) != -1) {
 		switch (c) {
 		case 'a':
 			flags |= ZINJECT_FLUSH_ARC;
@@ -1113,6 +1141,18 @@ main(int argc, char **argv)
 		case 'u':
 			flags |= ZINJECT_UNLOAD_SPA;
 			break;
+		case 'E':
+			ready_delay = MSEC2NSEC(strtol(optarg, &end, 10));
+			if (ready_delay <= 0 || *end != '\0') {
+				(void) fprintf(stderr, "invalid delay '%s': "
+				    "must be a positive duration\n", optarg);
+				usage();
+				libzfs_fini(g_zfs);
+				return (1);
+			}
+			record.zi_cmd = ZINJECT_DELAY_READY;
+			record.zi_timer = ready_delay;
+			break;
 		case 'L':
 			if ((label = name_to_type(optarg)) == TYPE_INVAL &&
 			    !LABEL_TYPE(type)) {
@@ -1150,7 +1190,7 @@ main(int argc, char **argv)
 		 */
 		if (raw != NULL || range != NULL || type != TYPE_INVAL ||
 		    level != 0 || record.zi_cmd != ZINJECT_UNINITIALIZED ||
-		    record.zi_freq > 0 || dvas != 0) {
+		    record.zi_freq > 0 || dvas != 0 || ready_delay >= 0) {
 			(void) fprintf(stderr, "cancel (-c) incompatible with "
 			    "any other options\n");
 			usage();
@@ -1186,7 +1226,7 @@ main(int argc, char **argv)
 		 */
 		if (raw != NULL || range != NULL || type != TYPE_INVAL ||
 		    level != 0 || record.zi_cmd != ZINJECT_UNINITIALIZED ||
-		    dvas != 0) {
+		    dvas != 0 || ready_delay >= 0) {
 			(void) fprintf(stderr, "device (-d) incompatible with "
 			    "data error injection\n");
 			usage();
@@ -1276,13 +1316,23 @@ main(int argc, char **argv)
 			return (1);
 		}
 
-		record.zi_cmd = ZINJECT_DATA_FAULT;
+		if (record.zi_cmd == ZINJECT_UNINITIALIZED) {
+			record.zi_cmd = ZINJECT_DATA_FAULT;
+			if (!error)
+				error = EIO;
+		} else if (error != 0) {
+			(void) fprintf(stderr, "error type -e incompatible "
+			    "with delay injection\n");
+			libzfs_fini(g_zfs);
+			return (1);
+		} else {
+			record.zi_iotype = io_type;
+		}
+
 		if (translate_raw(raw, &record) != 0) {
 			libzfs_fini(g_zfs);
 			return (1);
 		}
-		if (!error)
-			error = EIO;
 	} else if (record.zi_cmd == ZINJECT_PANIC) {
 		if (raw != NULL || range != NULL || type != TYPE_INVAL ||
 		    level != 0 || device != NULL || record.zi_freq > 0 ||
@@ -1410,6 +1460,13 @@ main(int argc, char **argv)
 			record.zi_dvas = dvas;
 		}
 
+		if (record.zi_cmd != ZINJECT_UNINITIALIZED && error != 0) {
+			(void) fprintf(stderr, "error type -e incompatible "
+			    "with delay injection\n");
+			libzfs_fini(g_zfs);
+			return (1);
+		}
+
 		if (error == EACCES) {
 			if (type != TYPE_DATA) {
 				(void) fprintf(stderr, "decryption errors "
@@ -1425,8 +1482,12 @@ main(int argc, char **argv)
 			 * not found.
 			 */
 			error = ECKSUM;
-		} else {
+		} else if (record.zi_cmd == ZINJECT_UNINITIALIZED) {
 			record.zi_cmd = ZINJECT_DATA_FAULT;
+			if (!error)
+				error = EIO;
+		} else {
+			record.zi_iotype = io_type;
 		}
 
 		if (translate_record(type, argv[0], range, level, &record, pool,
@@ -1434,8 +1495,6 @@ main(int argc, char **argv)
 			libzfs_fini(g_zfs);
 			return (1);
 		}
-		if (!error)
-			error = EIO;
 	}
 
 	/*

cmd/zpool/Makefile.am

@@ -148,6 +148,7 @@ dist_zpoolcompat_DATA = \
 	%D%/compatibility.d/openzfs-2.1-linux \
 	%D%/compatibility.d/openzfs-2.2 \
 	%D%/compatibility.d/openzfs-2.3 \
+	%D%/compatibility.d/openzfs-2.4 \
 	%D%/compatibility.d/openzfsonosx-1.7.0 \
 	%D%/compatibility.d/openzfsonosx-1.8.1 \
 	%D%/compatibility.d/openzfsonosx-1.9.3 \
@@ -187,7 +188,9 @@ zpoolcompatlinks = \
 	"openzfs-2.2		openzfs-2.2-linux" \
 	"openzfs-2.2		openzfs-2.2-freebsd" \
 	"openzfs-2.3		openzfs-2.3-linux" \
-	"openzfs-2.3		openzfs-2.3-freebsd"
+	"openzfs-2.3		openzfs-2.3-freebsd" \
+	"openzfs-2.4		openzfs-2.4-linux" \
+	"openzfs-2.4		openzfs-2.4-freebsd"
 
 zpoolconfdir = $(sysconfdir)/zfs/zpool.d
 INSTALL_DATA_HOOKS += zpool-install-data-hook

cmd/zpool/compatibility.d/openzfs-2.4

@@ -0,0 +1,48 @@
+# Features supported by OpenZFS 2.4 on Linux and FreeBSD
+allocation_classes
+async_destroy
+blake3
+block_cloning
+block_cloning_endian
+bookmark_v2
+bookmark_written
+bookmarks
+device_rebuild
+device_removal
+draid
+dynamic_gang_header
+edonr
+embedded_data
+empty_bpobj
+enabled_txg
+encryption
+extensible_dataset
+fast_dedup
+filesystem_limits
+head_errlog
+hole_birth
+large_blocks
+large_dnode
+large_microzap
+livelist
+log_spacemap
+longname
+lz4_compress
+multi_vdev_crash_dump
+obsolete_counts
+physical_rewrite
+project_quota
+raidz_expansion
+redacted_datasets
+redaction_bookmarks
+redaction_list_spill
+resilver_defer
+sha512
+skein
+spacemap_histogram
+spacemap_v2
+userobj_accounting
+vdev_zaps_v2
+zilsaxattr
+zpool_checkpoint
+zstd_compress

cmd/zpool/zpool_iter.c

@@ -26,6 +26,7 @@
 
 /*
  * Copyright 2016 Igor Kozhukhov <ikozhukhov@gmail.com>.
+ * Copyright (c) 2025, Klara, Inc.
  */
 
 #include <libintl.h>
@@ -52,7 +53,7 @@
 typedef struct zpool_node {
 	zpool_handle_t	*zn_handle;
 	uu_avl_node_t	zn_avlnode;
-	int		zn_mark;
+	hrtime_t	zn_last_refresh;
 } zpool_node_t;
 
 struct zpool_list {
@@ -62,6 +63,7 @@ struct zpool_list {
 	uu_avl_pool_t	*zl_pool;
 	zprop_list_t	**zl_proplist;
 	zfs_type_t	zl_type;
+	hrtime_t	zl_last_refresh;
 };
 
 static int
@@ -81,32 +83,47 @@ zpool_compare(const void *larg, const void *rarg, void *unused)
  * of known pools.
  */
 static int
-add_pool(zpool_handle_t *zhp, void *data)
+add_pool(zpool_handle_t *zhp, zpool_list_t *zlp)
 {
-	zpool_list_t *zlp = data;
+	zpool_node_t *node, *new = safe_malloc(sizeof (zpool_node_t));
-	zpool_node_t *node = safe_malloc(sizeof (zpool_node_t));
 	uu_avl_index_t idx;
 
-	node->zn_handle = zhp;
+	new->zn_handle = zhp;
-	uu_avl_node_init(node, &node->zn_avlnode, zlp->zl_pool);
+	uu_avl_node_init(new, &new->zn_avlnode, zlp->zl_pool);
-	if (uu_avl_find(zlp->zl_avl, node, NULL, &idx) == NULL) {
+
+	node = uu_avl_find(zlp->zl_avl, new, NULL, &idx);
+	if (node == NULL) {
 		if (zlp->zl_proplist &&
 		    zpool_expand_proplist(zhp, zlp->zl_proplist,
 		    zlp->zl_type, zlp->zl_literal) != 0) {
 			zpool_close(zhp);
-			free(node);
+			free(new);
 			return (-1);
 		}
-		uu_avl_insert(zlp->zl_avl, node, idx);
+		new->zn_last_refresh = zlp->zl_last_refresh;
+		uu_avl_insert(zlp->zl_avl, new, idx);
 	} else {
+		node->zn_last_refresh = zlp->zl_last_refresh;
 		zpool_close(zhp);
-		free(node);
+		free(new);
 		return (-1);
 	}
 
 	return (0);
 }
 
+/*
+ * add_pool(), but always returns 0. This allows zpool_iter() to continue
+ * even if a pool exists in the tree, or we fail to get the properties for
+ * a new one.
+ */
+static int
+add_pool_cb(zpool_handle_t *zhp, void *data)
+{
+	(void) add_pool(zhp, data);
+	return (0);
+}
+
 /*
  * Create a list of pools based on the given arguments.  If we're given no
  * arguments, then iterate over all pools in the system and add them to the AVL
@@ -135,9 +152,10 @@ pool_list_get(int argc, char **argv, zprop_list_t **proplist, zfs_type_t type,
 	zlp->zl_type = type;
 
 	zlp->zl_literal = literal;
+	zlp->zl_last_refresh = gethrtime();
 
 	if (argc == 0) {
-		(void) zpool_iter(g_zfs, add_pool, zlp);
+		(void) zpool_iter(g_zfs, add_pool_cb, zlp);
 		zlp->zl_findall = B_TRUE;
 	} else {
 		int i;
@@ -159,15 +177,69 @@ pool_list_get(int argc, char **argv, zprop_list_t **proplist, zfs_type_t type,
 }
 
 /*
- * Search for any new pools, adding them to the list.  We only add pools when no
+ * Refresh the state of all pools on the list. Additionally, if no options were
- * options were given on the command line.  Otherwise, we keep the list fixed as
+ * given on the command line, add any new pools and remove any that are no
- * those that were explicitly specified.
+ * longer available.
  */
-void
+int
-pool_list_update(zpool_list_t *zlp)
+pool_list_refresh(zpool_list_t *zlp)
 {
-	if (zlp->zl_findall)
+	zlp->zl_last_refresh = gethrtime();
-		(void) zpool_iter(g_zfs, add_pool, zlp);
+
+	if (!zlp->zl_findall) {
+		/*
+		 * This list is a fixed list of pools, so we must not add
+		 * or remove any. Just walk over them and refresh their
+		 * state.
+		 */
+		int navail = 0;
+		for (zpool_node_t *node = uu_avl_first(zlp->zl_avl);
+		    node != NULL; node = uu_avl_next(zlp->zl_avl, node)) {
+			boolean_t missing;
+			zpool_refresh_stats(node->zn_handle, &missing);
+			navail += !missing;
+			node->zn_last_refresh = zlp->zl_last_refresh;
+		}
+		return (navail);
+	}
+
+	/*
+	 * Search for any new pools and add them to the list. zpool_iter()
+	 * will call zpool_refresh_stats() as part of its work, so this has
+	 * the side effect of updating all active handles.
+	 */
+	(void) zpool_iter(g_zfs, add_pool_cb, zlp);
+
+	/*
+	 * Walk the list for any that weren't refreshed, and update and remove
+	 * them. It's not enough to just skip available ones, as zpool_iter()
+	 * won't update them, so they'll still appear active in our list.
+	 */
+	zpool_node_t *node, *next;
+	for (node = uu_avl_first(zlp->zl_avl); node != NULL; node = next) {
+		next = uu_avl_next(zlp->zl_avl, node);
+
+		/*
+		 * Skip any that were refreshed and are online; they're already
+		 * handled.
+		 */
+		if (node->zn_last_refresh == zlp->zl_last_refresh &&
+		    zpool_get_state(node->zn_handle) != POOL_STATE_UNAVAIL)
+			continue;
+
+		/* Do the refresh ourselves, just in case. */
+		boolean_t missing;
+		zpool_refresh_stats(node->zn_handle, &missing);
+		if (missing) {
+			uu_avl_remove(zlp->zl_avl, node);
+			zpool_close(node->zn_handle);
+			free(node);
+		} else {
+			node->zn_last_refresh = zlp->zl_last_refresh;
+		}
+	}
+
+	return (uu_avl_numnodes(zlp->zl_avl));
 }
 
 /*
@@ -190,23 +262,6 @@ pool_list_iter(zpool_list_t *zlp, int unavail, zpool_iter_f func,
 	return (ret);
 }
 
-/*
- * Remove the given pool from the list.  When running iostat, we want to remove
- * those pools that no longer exist.
- */
-void
-pool_list_remove(zpool_list_t *zlp, zpool_handle_t *zhp)
-{
-	zpool_node_t search, *node;
-
-	search.zn_handle = zhp;
-	if ((node = uu_avl_find(zlp->zl_avl, &search, NULL, NULL)) != NULL) {
-		uu_avl_remove(zlp->zl_avl, node);
-		zpool_close(node->zn_handle);
-		free(node);
-	}
-}
-
 /*
  * Free all the handles associated with this list.
  */

cmd/zpool/zpool_main.c

@@ -33,7 +33,7 @@
  * Copyright (c) 2017, Intel Corporation.
  * Copyright (c) 2019, loli10K <ezomori.nozomu@gmail.com>
  * Copyright (c) 2021, Colm Buckley <colm@tuatha.org>
- * Copyright (c) 2021, 2023, Klara Inc.
+ * Copyright (c) 2021, 2023, 2025, Klara, Inc.
  * Copyright (c) 2021, 2025 Hewlett Packard Enterprise Development LP.
  */
 
@@ -456,7 +456,7 @@ get_usage(zpool_help_t idx)
 		    "<pool> <vdev> ...\n"));
 	case HELP_ATTACH:
 		return (gettext("\tattach [-fsw] [-o property=value] "
-		    "<pool> <device> <new-device>\n"));
+		    "<pool> <vdev> <new-device>\n"));
 	case HELP_CLEAR:
 		return (gettext("\tclear [[--power]|[-nF]] <pool> [device]\n"));
 	case HELP_CREATE:
@@ -5761,24 +5761,6 @@ children:
 	return (ret);
 }
 
-static int
-refresh_iostat(zpool_handle_t *zhp, void *data)
-{
-	iostat_cbdata_t *cb = data;
-	boolean_t missing;
-
-	/*
-	 * If the pool has disappeared, remove it from the list and continue.
-	 */
-	if (zpool_refresh_stats(zhp, &missing) != 0)
-		return (-1);
-
-	if (missing)
-		pool_list_remove(cb->cb_list, zhp);
-
-	return (0);
-}
-
 /*
  * Callback to print out the iostats for the given pool.
  */
@@ -6359,15 +6341,14 @@ get_namewidth_iostat(zpool_handle_t *zhp, void *data)
  * This command can be tricky because we want to be able to deal with pool
  * creation/destruction as well as vdev configuration changes.  The bulk of this
  * processing is handled by the pool_list_* routines in zpool_iter.c.  We rely
- * on pool_list_update() to detect the addition of new pools.  Configuration
+ * on pool_list_refresh() to detect the addition and removal of pools.
- * changes are all handled within libzfs.
+ * Configuration changes are all handled within libzfs.
  */
 int
 zpool_do_iostat(int argc, char **argv)
 {
 	int c;
 	int ret;
-	int npools;
 	float interval = 0;
 	unsigned long count = 0;
 	zpool_list_t *list;
@@ -6618,10 +6599,24 @@ zpool_do_iostat(int argc, char **argv)
 		return (1);
 	}
 
+	int last_npools = 0;
 	for (;;) {
-		if ((npools = pool_list_count(list)) == 0)
+		/*
+		 * Refresh all pools in list, adding or removing pools as
+		 * necessary.
+		 */
+		int npools = pool_list_refresh(list);
+		if (npools == 0) {
 			(void) fprintf(stderr, gettext("no pools available\n"));
-		else {
+		} else {
+			/*
+			 * If the list of pools has changed since last time
+			 * around, reset the iteration count to force the
+			 * header to be redisplayed.
+			 */
+			if (last_npools != npools)
+				cb.cb_iteration = 0;
+
 			/*
 			 * If this is the first iteration and -y was supplied
 			 * we skip any printing.
@@ -6629,15 +6624,6 @@ zpool_do_iostat(int argc, char **argv)
 			boolean_t skip = (omit_since_boot &&
 			    cb.cb_iteration == 0);
 
-			/*
-			 * Refresh all statistics.  This is done as an
-			 * explicit step before calculating the maximum name
-			 * width, so that any * configuration changes are
-			 * properly accounted for.
-			 */
-			(void) pool_list_iter(list, B_FALSE, refresh_iostat,
-			    &cb);
-
 			/*
 			 * Iterate over all pools to determine the maximum width
 			 * for the pool / device name column across all pools.
@@ -6728,6 +6714,8 @@ zpool_do_iostat(int argc, char **argv)
 
 		(void) fflush(stdout);
 		(void) fsleep(interval);
+
+		last_npools = npools;
 	}
 
 	pool_list_free(list);
@@ -7644,7 +7632,7 @@ zpool_do_replace(int argc, char **argv)
 }
 
 /*
- * zpool attach [-fsw] [-o property=value] <pool> <device>|<vdev> <new_device>
+ * zpool attach [-fsw] [-o property=value] <pool> <vdev> <new_device>
  *
  *	-f	Force attach, even if <new_device> appears to be in use.
  *	-s	Use sequential instead of healing reconstruction for resilver.
@@ -7652,9 +7640,9 @@ zpool_do_replace(int argc, char **argv)
  *	-w	Wait for resilvering (mirror) or expansion (raidz) to complete
  *		before returning.
  *
- * Attach <new_device> to a <device> or <vdev>, where the vdev can be of type
+ * Attach <new_device> to a <vdev>, where the vdev can be of type
- * mirror or raidz. If <device> is not part of a mirror, then <device> will
+ * device, mirror or raidz. If <vdev> is not part of a mirror, then <vdev> will
- * be transformed into a mirror of <device> and <new_device>. When a mirror
+ * be transformed into a mirror of <vdev> and <new_device>. When a mirror
  * is involved, <new_device> will begin life with a DTL of [0, now], and will
  * immediately begin to resilver itself. For the raidz case, a expansion will
  * commence and reflow the raidz data across all the disks including the

cmd/zpool/zpool_util.h

@@ -76,11 +76,10 @@ typedef struct zpool_list zpool_list_t;
 
 zpool_list_t *pool_list_get(int, char **, zprop_list_t **, zfs_type_t,
     boolean_t, int *);
-void pool_list_update(zpool_list_t *);
+int pool_list_refresh(zpool_list_t *);
 int pool_list_iter(zpool_list_t *, int unavail, zpool_iter_f, void *);
 void pool_list_free(zpool_list_t *);
 int pool_list_count(zpool_list_t *);
-void pool_list_remove(zpool_list_t *, zpool_handle_t *);
 
 extern libzfs_handle_t *g_zfs;
 

cmd/zpool/zpool_vdev.c

@@ -609,22 +609,28 @@ get_replication(nvlist_t *nvroot, boolean_t fatal)
 				    ZPOOL_CONFIG_PATH, &path) == 0);
 
 				/*
-				 * If we have a raidz/mirror that combines disks
+				 * Skip active spares they should never cause
-				 * with files, report it as an error.
+				 * the pool to be evaluated as inconsistent.
 				 */
-				if (!dontreport && type != NULL &&
+				if (is_spare(NULL, path))
+					continue;
+
+				/*
+				 * If we have a raidz/mirror that combines disks
+				 * with files, only report it as an error when
+				 * fatal is set to ensure all the replication
+				 * checks aren't skipped in check_replication().
+				 */
+				if (fatal && !dontreport && type != NULL &&
 				    strcmp(type, childtype) != 0) {
 					if (ret != NULL)
 						free(ret);
 					ret = NULL;
-					if (fatal)
+					vdev_error(gettext(
-						vdev_error(gettext(
+					    "mismatched replication "
-						    "mismatched replication "
+					    "level: %s contains both "
-						    "level: %s contains both "
+					    "files and devices\n"),
-						    "files and devices\n"),
+					    rep.zprl_type);
-						    rep.zprl_type);
-					else
-						return (NULL);
 					dontreport = B_TRUE;
 				}
 

cmd/zstream/Makefile.am

@@ -18,6 +18,7 @@ zstream_LDADD = \
 	libzpool.la \
 	libnvpair.la
 
-PHONY += install-exec-hook
+cmd-zstream-install-exec-hook:
-install-exec-hook:
 	cd $(DESTDIR)$(sbindir) && $(LN_S) -f zstream zstreamdump
+
+INSTALL_EXEC_HOOKS += cmd-zstream-install-exec-hook

config/Shellcheck.am

@@ -16,10 +16,14 @@ SHELLCHECK_OPTS         = $(call JUST_SHELLCHECK_OPTS,$(1)) $(call JUST_CHECKBAS
 
 PHONY += shellcheck
 
+shellcheck_verbose = $(shellcheck_verbose_@AM_V@)
+shellcheck_verbose_ = $(shellcheck_verbose_@AM_DEFAULT_V@)
+shellcheck_verbose_0 = @echo SHELLCHECK $(_STGT);
+
 _STGT = $(subst ^,/,$(subst shellcheck-here-,,$@))
 shellcheck-here-%:
 if HAVE_SHELLCHECK
-	shellcheck --format=gcc --enable=all --exclude=SC1090,SC1091,SC2039,SC2250,SC2312,SC2317,SC3043 $$([ -n "$(SHELLCHECK_SHELL)" ] && echo "--shell=$(SHELLCHECK_SHELL)") "$$([ -e "$(_STGT)" ] || echo "$(srcdir)/")$(_STGT)"
+	$(shellcheck_verbose)shellcheck --format=gcc --enable=all --exclude=SC1090,SC1091,SC2039,SC2250,SC2312,SC2317,SC3043 $$([ -n "$(SHELLCHECK_SHELL)" ] && echo "--shell=$(SHELLCHECK_SHELL)") "$$([ -e "$(_STGT)" ] || echo "$(srcdir)/")$(_STGT)"
 else
 	@echo "skipping shellcheck of" $(_STGT) "because shellcheck is not installed"
 endif
@@ -29,11 +33,15 @@ shellcheck: $(SHELLCHECKSCRIPTS) $(call JUST_SHELLCHECK_OPTS,$(SHELLCHECKSCRIPTS
 
 PHONY += checkbashisms
 
+checkbashisms_verbose = $(checkbashisms_verbose_@AM_V@)
+checkbashisms_verbose_ = $(checkbashisms_verbose_@AM_DEFAULT_V@)
+checkbashisms_verbose_0 = @echo CHECKBASHISMS $(_BTGT);
+
 # command -v *is* specified by POSIX and every shell in existence supports it
 _BTGT = $(subst ^,/,$(subst checkbashisms-here-,,$@))
 checkbashisms-here-%:
 if HAVE_CHECKBASHISMS
-	! { [ -n "$(SHELLCHECK_SHELL)" ] && echo '#!/bin/$(SHELLCHECK_SHELL)'; cat "$$([ -e "$(_BTGT)" ] || echo "$(srcdir)/")$(_BTGT)"; } | \
+	$(checkbashisms_verbose)! { [ -n "$(SHELLCHECK_SHELL)" ] && echo '#!/bin/$(SHELLCHECK_SHELL)'; cat "$$([ -e "$(_BTGT)" ] || echo "$(srcdir)/")$(_BTGT)"; } | \
 	checkbashisms -npx 2>&1 | grep -vFe "'command' with option other than -p" -e 'command -v' -e 'any possible bashisms' $(CHECKBASHISMS_IGNORE) >&2
 else
 	@echo "skipping checkbashisms of" $(_BTGT) "because checkbashisms is not installed"

config/always-arch.m4

@@ -34,8 +34,26 @@ AC_DEFUN([ZFS_AC_CONFIG_ALWAYS_ARCH], [
 	esac
 
 	AM_CONDITIONAL([TARGET_CPU_AARCH64], test $TARGET_CPU = aarch64)
+	AM_CONDITIONAL([TARGET_CPU_I386],    test $TARGET_CPU = i386)
 	AM_CONDITIONAL([TARGET_CPU_X86_64],  test $TARGET_CPU = x86_64)
 	AM_CONDITIONAL([TARGET_CPU_POWERPC], test $TARGET_CPU = powerpc)
 	AM_CONDITIONAL([TARGET_CPU_SPARC64], test $TARGET_CPU = sparc64)
 	AM_CONDITIONAL([TARGET_CPU_ARM],     test $TARGET_CPU = arm)
 ])
+dnl #
+dnl # Check for conflicting environment variables
+dnl #
+dnl # If ARCH env variable is set up, then kernel Makefile in the /usr/src/kernel
+dnl # can misbehave during the zfs ./configure test of the module compilation.
+AC_DEFUN([ZFS_AC_CONFIG_CHECK_ARCH_VAR], [
+	AC_MSG_CHECKING([for conflicting environment variables])
+	if test -n "$ARCH"; then
+		AC_MSG_RESULT([warning])
+		AC_MSG_WARN(m4_normalize([ARCH environment variable is set to "$ARCH".
+    This can cause build kernel modules support check failure.
+    Please unset it.]))
+	else
+		AC_MSG_RESULT([done])
+	fi
+])
+

config/always-compiler-options.m4

@@ -155,6 +155,34 @@ AC_DEFUN([ZFS_AC_CONFIG_ALWAYS_CC_NO_FORMAT_ZERO_LENGTH], [
 	AC_SUBST([NO_FORMAT_ZERO_LENGTH])
 ])
 
+dnl #
+dnl # Check if kernel cc supports -Wno-format-zero-length option.
+dnl #
+AC_DEFUN([ZFS_AC_CONFIG_ALWAYS_KERNEL_CC_NO_FORMAT_ZERO_LENGTH], [
+	saved_cc="$CC"
+	AS_IF(
+		[ test -n "$KERNEL_CC" ], [ CC="$KERNEL_CC" ],
+		[ test -n "$KERNEL_LLVM" ], [ CC="clang" ],
+		[ CC="gcc" ]
+	)
+	AC_MSG_CHECKING([whether $CC supports -Wno-format-zero-length])
+
+	saved_flags="$CFLAGS"
+	CFLAGS="$CFLAGS -Werror -Wno-format-zero-length"
+
+	AC_COMPILE_IFELSE([AC_LANG_PROGRAM([], [])], [
+		KERNEL_NO_FORMAT_ZERO_LENGTH=-Wno-format-zero-length
+		AC_MSG_RESULT([yes])
+	], [
+		KERNEL_NO_FORMAT_ZERO_LENGTH=
+		AC_MSG_RESULT([no])
+	])
+
+	CC="$saved_cc"
+	CFLAGS="$saved_flags"
+	AC_SUBST([KERNEL_NO_FORMAT_ZERO_LENGTH])
+])
+
 dnl #
 dnl # Check if cc supports -Wno-clobbered option.
 dnl #
@@ -181,6 +209,27 @@ AC_DEFUN([ZFS_AC_CONFIG_ALWAYS_CC_NO_CLOBBERED], [
 	AC_SUBST([NO_CLOBBERED])
 ])
 
+dnl #
+dnl # Check if cc supports -Wno-atomic-alignment option.
+dnl #
+AC_DEFUN([ZFS_AC_CONFIG_ALWAYS_CC_NO_ATOMIC_ALIGNMENT], [
+	AC_MSG_CHECKING([whether $CC supports -Wno-atomic-alignment])
+
+	saved_flags="$CFLAGS"
+	CFLAGS="$CFLAGS -Werror -Wno-atomic-alignment"
+
+	AC_COMPILE_IFELSE([AC_LANG_PROGRAM([], [])], [
+		NO_ATOMIC_ALIGNMENT=-Wno-atomic-alignment
+		AC_MSG_RESULT([yes])
+	], [
+		NO_ATOMIC_ALIGNMENT=
+		AC_MSG_RESULT([no])
+	])
+
+	CFLAGS="$saved_flags"
+	AC_SUBST([NO_ATOMIC_ALIGNMENT])
+])
+
 dnl #
 dnl # Check if cc supports -Wimplicit-fallthrough option.
 dnl #
@@ -231,19 +280,16 @@ dnl #
 dnl # Check if kernel cc supports -Winfinite-recursion option.
 dnl #
 AC_DEFUN([ZFS_AC_CONFIG_ALWAYS_KERNEL_CC_INFINITE_RECURSION], [
-	AC_MSG_CHECKING([whether $KERNEL_CC supports -Winfinite-recursion])
-
 	saved_cc="$CC"
-	saved_flags="$CFLAGS"
+	AS_IF(
-	CC="gcc"
+		[ test -n "$KERNEL_CC" ], [ CC="$KERNEL_CC" ],
-	CFLAGS="$CFLAGS -Werror -Winfinite-recursion"
+		[ test -n "$KERNEL_LLVM" ], [ CC="clang" ],
+		[ CC="gcc" ]
+	)
+	AC_MSG_CHECKING([whether $CC supports -Winfinite-recursion])
 
-	AS_IF([ test -n "$KERNEL_CC" ], [
+	saved_flags="$CFLAGS"
-		CC="$KERNEL_CC"
+	CFLAGS="$CFLAGS -Werror -Winfinite-recursion"
-	])
-	AS_IF([ test -n "$KERNEL_LLVM" ], [
-		CC="clang"
-	])
 
 	AC_COMPILE_IFELSE([AC_LANG_PROGRAM([], [])], [
 		KERNEL_INFINITE_RECURSION=-Winfinite-recursion
@@ -329,19 +375,16 @@ dnl #
 dnl # Check if kernel cc supports -fno-ipa-sra option.
 dnl #
 AC_DEFUN([ZFS_AC_CONFIG_ALWAYS_KERNEL_CC_NO_IPA_SRA], [
-	AC_MSG_CHECKING([whether $KERNEL_CC supports -fno-ipa-sra])
-
 	saved_cc="$CC"
-	saved_flags="$CFLAGS"
+	AS_IF(
-	CC="gcc"
+		[ test -n "$KERNEL_CC" ], [ CC="$KERNEL_CC" ],
-	CFLAGS="$CFLAGS -Werror -fno-ipa-sra"
+		[ test -n "$KERNEL_LLVM" ], [ CC="clang" ],
+		[ CC="gcc" ]
+	)
+	AC_MSG_CHECKING([whether $CC supports -fno-ipa-sra])
 
-	AS_IF([ test -n "$KERNEL_CC" ], [
+	saved_flags="$CFLAGS"
-		CC="$KERNEL_CC"
+	CFLAGS="$CFLAGS -Werror -fno-ipa-sra"
-	])
-	AS_IF([ test -n "$KERNEL_LLVM" ], [
-		CC="clang"
-	])
 
 	AC_COMPILE_IFELSE([AC_LANG_PROGRAM([], [])], [
 		KERNEL_NO_IPA_SRA=-fno-ipa-sra

config/kernel-blkdev.m4

@@ -29,9 +29,8 @@ AC_DEFUN([ZFS_AC_KERNEL_SRC_BLKDEV_GET_BY_PATH_4ARG], [
 		const char *path = "path";
 		fmode_t mode = 0;
 		void *holder = NULL;
-		struct blk_holder_ops h;
 
-		bdev = blkdev_get_by_path(path, mode, holder, &h);
+		bdev = blkdev_get_by_path(path, mode, holder, NULL);
 	])
 ])
 
@@ -48,9 +47,8 @@ AC_DEFUN([ZFS_AC_KERNEL_SRC_BLKDEV_BDEV_OPEN_BY_PATH], [
 		const char *path = "path";
 		fmode_t mode = 0;
 		void *holder = NULL;
-		struct blk_holder_ops h;
 
-		bdh = bdev_open_by_path(path, mode, holder, &h);
+		bdh = bdev_open_by_path(path, mode, holder, NULL);
 	])
 ])
 
@@ -68,9 +66,8 @@ AC_DEFUN([ZFS_AC_KERNEL_SRC_BDEV_FILE_OPEN_BY_PATH], [
 		const char *path = "path";
 		fmode_t mode = 0;
 		void *holder = NULL;
-		struct blk_holder_ops h;
 
-		file = bdev_file_open_by_path(path, mode, holder, &h);
+		file = bdev_file_open_by_path(path, mode, holder, NULL);
 	])
 ])
 

config/kernel-dentry-operations.m4

@@ -24,6 +24,9 @@ dnl #
 dnl # 2.6.38 API change
 dnl # Added d_set_d_op() helper function.
 dnl #
+dnl # 6.17 API change
+dnl # d_set_d_op() removed. No direct replacement.
+dnl #
 AC_DEFUN([ZFS_AC_KERNEL_SRC_D_SET_D_OP], [
 	ZFS_LINUX_TEST_SRC([d_set_d_op], [
 		#include <linux/dcache.h>
@@ -34,22 +37,46 @@ AC_DEFUN([ZFS_AC_KERNEL_SRC_D_SET_D_OP], [
 
 AC_DEFUN([ZFS_AC_KERNEL_D_SET_D_OP], [
 	AC_MSG_CHECKING([whether d_set_d_op() is available])
-	ZFS_LINUX_TEST_RESULT_SYMBOL([d_set_d_op],
+	ZFS_LINUX_TEST_RESULT([d_set_d_op], [
-	    [d_set_d_op], [fs/dcache.c], [
 		AC_MSG_RESULT(yes)
+		AC_DEFINE(HAVE_D_SET_D_OP, 1,
+		    [Define if d_set_d_op() is available])
 	], [
-		ZFS_LINUX_TEST_ERROR([d_set_d_op])
+		AC_MSG_RESULT(no)
+	])
+])
+
+dnl #
+dnl # 6.17 API change
+dnl # sb->s_d_op removed; set_default_d_op(sb, dop) added
+dnl #
+AC_DEFUN([ZFS_AC_KERNEL_SRC_SET_DEFAULT_D_OP], [
+	ZFS_LINUX_TEST_SRC([set_default_d_op], [
+		#include <linux/dcache.h>
+	], [
+		set_default_d_op(NULL, NULL);
+	])
+])
+
+AC_DEFUN([ZFS_AC_KERNEL_SET_DEFAULT_D_OP], [
+	AC_MSG_CHECKING([whether set_default_d_op() is available])
+	ZFS_LINUX_TEST_RESULT([set_default_d_op], [
+		AC_MSG_RESULT(yes)
+		AC_DEFINE(HAVE_SET_DEFAULT_D_OP, 1,
+		    [Define if set_default_d_op() is available])
+	], [
+		AC_MSG_RESULT(no)
 	])
 ])
 
 AC_DEFUN([ZFS_AC_KERNEL_SRC_DENTRY], [
         ZFS_AC_KERNEL_SRC_D_OBTAIN_ALIAS
         ZFS_AC_KERNEL_SRC_D_SET_D_OP
-        ZFS_AC_KERNEL_SRC_S_D_OP
+        ZFS_AC_KERNEL_SRC_SET_DEFAULT_D_OP
 ])
 
 AC_DEFUN([ZFS_AC_KERNEL_DENTRY], [
         ZFS_AC_KERNEL_D_OBTAIN_ALIAS
         ZFS_AC_KERNEL_D_SET_D_OP
-        ZFS_AC_KERNEL_S_D_OP
+        ZFS_AC_KERNEL_SET_DEFAULT_D_OP
 ])

config/kernel.m4

@@ -70,6 +70,7 @@ AC_DEFUN([ZFS_AC_KERNEL_TEST_SRC], [
 	ZFS_AC_KERNEL_SRC_COMMIT_METADATA
 	ZFS_AC_KERNEL_SRC_SETATTR_PREPARE
 	ZFS_AC_KERNEL_SRC_INSERT_INODE_LOCKED
+	ZFS_AC_KERNEL_SRC_DENTRY
 	ZFS_AC_KERNEL_SRC_TRUNCATE_SETSIZE
 	ZFS_AC_KERNEL_SRC_SECURITY_INODE
 	ZFS_AC_KERNEL_SRC_FST_MOUNT
@@ -188,6 +189,7 @@ AC_DEFUN([ZFS_AC_KERNEL_TEST_RESULT], [
 	ZFS_AC_KERNEL_COMMIT_METADATA
 	ZFS_AC_KERNEL_SETATTR_PREPARE
 	ZFS_AC_KERNEL_INSERT_INODE_LOCKED
+	ZFS_AC_KERNEL_DENTRY
 	ZFS_AC_KERNEL_TRUNCATE_SETSIZE
 	ZFS_AC_KERNEL_SECURITY_INODE
 	ZFS_AC_KERNEL_FST_MOUNT

config/user-statx.m4

@@ -2,7 +2,7 @@ dnl #
 dnl # Check for statx() function and STATX_MNT_ID availability
 dnl #
 AC_DEFUN([ZFS_AC_CONFIG_USER_STATX], [
-	AC_CHECK_HEADERS([linux/stat.h],
+	AC_CHECK_HEADERS([sys/stat.h],
 		[have_stat_headers=yes],
 		[have_stat_headers=no])
 
@@ -14,7 +14,7 @@ AC_DEFUN([ZFS_AC_CONFIG_USER_STATX], [
 			AC_MSG_CHECKING([for STATX_MNT_ID])
 			AC_COMPILE_IFELSE([
 				AC_LANG_PROGRAM([[
-					#include <linux/stat.h>
+					#include <sys/stat.h>
 				]], [[
 					struct statx stx;
 					int mask = STATX_MNT_ID;
@@ -29,6 +29,6 @@ AC_DEFUN([ZFS_AC_CONFIG_USER_STATX], [
 			])
 		])
 	], [
-		AC_MSG_WARN([linux/stat.h not found; skipping statx support])
+		AC_MSG_WARN([sys/stat.h not found; skipping statx support])
 	])
 ])  dnl end AC_DEFUN

config/zfs-build.m4

@@ -252,10 +252,12 @@ AC_DEFUN([ZFS_AC_CONFIG_ALWAYS], [
 	ZFS_AC_CONFIG_ALWAYS_CC_NO_CLOBBERED
 	ZFS_AC_CONFIG_ALWAYS_CC_INFINITE_RECURSION
 	ZFS_AC_CONFIG_ALWAYS_KERNEL_CC_INFINITE_RECURSION
+	ZFS_AC_CONFIG_ALWAYS_CC_NO_ATOMIC_ALIGNMENT
 	ZFS_AC_CONFIG_ALWAYS_CC_IMPLICIT_FALLTHROUGH
 	ZFS_AC_CONFIG_ALWAYS_CC_FRAME_LARGER_THAN
 	ZFS_AC_CONFIG_ALWAYS_CC_NO_FORMAT_TRUNCATION
 	ZFS_AC_CONFIG_ALWAYS_CC_NO_FORMAT_ZERO_LENGTH
+	ZFS_AC_CONFIG_ALWAYS_KERNEL_CC_NO_FORMAT_ZERO_LENGTH
 	ZFS_AC_CONFIG_ALWAYS_CC_FORMAT_OVERFLOW
 	ZFS_AC_CONFIG_ALWAYS_CC_NO_OMIT_FRAME_POINTER
 	ZFS_AC_CONFIG_ALWAYS_CC_NO_IPA_SRA
@@ -265,6 +267,7 @@ AC_DEFUN([ZFS_AC_CONFIG_ALWAYS], [
 	ZFS_AC_CONFIG_ALWAYS_TOOLCHAIN_SIMD
 	ZFS_AC_CONFIG_ALWAYS_SYSTEM
 	ZFS_AC_CONFIG_ALWAYS_ARCH
+	ZFS_AC_CONFIG_CHECK_ARCH_VAR
 	ZFS_AC_CONFIG_ALWAYS_PYTHON
 	ZFS_AC_CONFIG_ALWAYS_PYZFS
 	ZFS_AC_CONFIG_ALWAYS_SED

contrib/debian/copyright

@@ -4,7 +4,7 @@ The detailed contributor information can be found in [2][3].
 
 Files: contrib/debian/*
 Copyright:
- 2013-2016, Aron Xu <aron@debian.org>
+ 2013-2025, Aron Xu <aron@debian.org>
  2016, Petter Reinholdtsen <pere@hungry.com>
  2013, Carlos Alberto Lopez Perez <clopez@igalia.com>
  2013, Turbo Fredriksson <turbo@bayour.com>
@@ -12,6 +12,8 @@ Copyright:
  2011-2013, Darik Horn <dajhorn@vanadac.com>
  2018-2019, Mo Zhou <cdluminate@gmail.com>
  2018-2020, Mo Zhou <lumin@debian.org>
+ 2023-2024, Shengqi Chen <harry-chen@outlook.com>
+ 2024-2025, Shengqi Chen <harry@debian.org>
 License: GPL-2+
 
 [1] https://tracker.debian.org/pkg/zfs-linux

contrib/debian/not-installed

@@ -1,4 +1,4 @@
-usr/bin/arc_summary.py
+usr/bin/zarcsummary.py
 usr/share/zfs/zfs-helpers.sh
 etc/default/zfs
 etc/init.d
@@ -9,4 +9,4 @@ etc/zfs/vdev_id.conf.sas_direct.example
 etc/zfs/vdev_id.conf.sas_switch.example
 etc/zfs/vdev_id.conf.scsi.example
 etc/zfs/zfs-functions
-lib/systemd/system/zfs-import.service
+usr/lib/systemd/system/zfs-import.service

contrib/debian/openzfs-libnvpair3.install.in

@@ -1 +1 @@
-lib/@DEB_HOST_MULTIARCH@/libnvpair.so.*
+usr/lib/@DEB_HOST_MULTIARCH@/libnvpair.so.*

contrib/debian/openzfs-libpam-zfs.install

@@ -1,2 +1,2 @@
-lib/*/security/pam_zfs_key.so
+usr/lib/*/security/pam_zfs_key.so
 usr/share/pam-configs/zfs_key

contrib/debian/openzfs-libpam-zfs.postinst

@@ -1,7 +1,7 @@
 #!/bin/sh
 set -e
 
-if ! $(ldd "/lib/$(dpkg-architecture -qDEB_HOST_MULTIARCH)/security/pam_zfs_key.so" | grep -q "libasan") ; then
+if ! $(ldd "/usr/lib/$(dpkg-architecture -qDEB_HOST_MULTIARCH)/security/pam_zfs_key.so" | grep -q "libasan") ; then
 	pam-auth-update --package
 fi
 

contrib/debian/openzfs-libuutil3.install.in

@@ -1 +1 @@
-lib/@DEB_HOST_MULTIARCH@/libuutil.so.*
+usr/lib/@DEB_HOST_MULTIARCH@/libuutil.so.*

contrib/debian/openzfs-libzfs-dev.install.in

@@ -1,3 +1,5 @@
-lib/@DEB_HOST_MULTIARCH@/*.a usr/lib/@DEB_HOST_MULTIARCH@
+usr/lib/@DEB_HOST_MULTIARCH@/*.a
+usr/lib/@DEB_HOST_MULTIARCH@/*.so
+usr/lib/@DEB_HOST_MULTIARCH@/pkgconfig
 usr/include
-usr/lib/@DEB_HOST_MULTIARCH@
+

contrib/debian/openzfs-libzfs6.install.in

@@ -1,2 +1,2 @@
-lib/@DEB_HOST_MULTIARCH@/libzfs.so.*
+usr/lib/@DEB_HOST_MULTIARCH@/libzfs.so.*
-lib/@DEB_HOST_MULTIARCH@/libzfs_core.so.*
+usr/lib/@DEB_HOST_MULTIARCH@/libzfs_core.so.*

contrib/debian/openzfs-libzfsbootenv1.install.in

@@ -1 +1 @@
-lib/@DEB_HOST_MULTIARCH@/libzfsbootenv.so.*
+usr/lib/@DEB_HOST_MULTIARCH@/libzfsbootenv.so.*

contrib/debian/openzfs-libzpool6.install.in

@@ -1 +1 @@
-lib/@DEB_HOST_MULTIARCH@/libzpool.so.*
+usr/lib/@DEB_HOST_MULTIARCH@/libzpool.so.*

contrib/debian/openzfs-zfs-test.install

@@ -1,4 +1,4 @@
-sbin/ztest
+usr/sbin/ztest
 usr/bin/raidz_test
 usr/share/man/man1/raidz_test.1
 usr/share/man/man1/test-runner.1

contrib/debian/openzfs-zfs-zed.install

@@ -1,5 +1,5 @@
 etc/zfs/zed.d/*
-lib/systemd/system/zfs-zed.service
+usr/lib/systemd/system/zfs-zed.service
 usr/lib/zfs-linux/zed.d/*
 usr/sbin/zed
 usr/share/man/man8/zed.8

contrib/debian/openzfs-zfsutils.install

@@ -1,48 +1,48 @@
 etc/default/zfs
 etc/zfs/zfs-functions
 etc/zfs/zpool.d/
-lib/systemd/system-generators/
+usr/lib/systemd/system-generators/
-lib/systemd/system-preset/
+usr/lib/systemd/system-preset/
-lib/systemd/system/zfs-import-cache.service
+usr/lib/systemd/system/zfs-import-cache.service
-lib/systemd/system/zfs-import-scan.service
+usr/lib/systemd/system/zfs-import-scan.service
-lib/systemd/system/zfs-import.target
+usr/lib/systemd/system/zfs-import.target
-lib/systemd/system/zfs-load-key.service
+usr/lib/systemd/system/zfs-load-key.service
-lib/systemd/system/zfs-mount.service
+usr/lib/systemd/system/zfs-mount.service
-lib/systemd/system/zfs-mount@.service
+usr/lib/systemd/system/zfs-mount@.service
-lib/systemd/system/zfs-scrub-monthly@.timer
+usr/lib/systemd/system/zfs-scrub-monthly@.timer
-lib/systemd/system/zfs-scrub-weekly@.timer
+usr/lib/systemd/system/zfs-scrub-weekly@.timer
-lib/systemd/system/zfs-scrub@.service
+usr/lib/systemd/system/zfs-scrub@.service
-lib/systemd/system/zfs-trim-monthly@.timer
+usr/lib/systemd/system/zfs-trim-monthly@.timer
-lib/systemd/system/zfs-trim-weekly@.timer
+usr/lib/systemd/system/zfs-trim-weekly@.timer
-lib/systemd/system/zfs-trim@.service
+usr/lib/systemd/system/zfs-trim@.service
-lib/systemd/system/zfs-share.service
+usr/lib/systemd/system/zfs-share.service
-lib/systemd/system/zfs-volume-wait.service
+usr/lib/systemd/system/zfs-volume-wait.service
-lib/systemd/system/zfs-volumes.target
+usr/lib/systemd/system/zfs-volumes.target
-lib/systemd/system/zfs.target
+usr/lib/systemd/system/zfs.target
-lib/udev/
+usr/lib/udev/
-sbin/fsck.zfs
+usr/sbin/fsck.zfs
-sbin/mount.zfs
+usr/sbin/mount.zfs
-sbin/zdb
+usr/sbin/zdb
-sbin/zfs
+usr/sbin/zfs
-sbin/zfs_ids_to_path
+usr/sbin/zfs_ids_to_path
-sbin/zgenhostid
+usr/sbin/zgenhostid
-sbin/zhack
+usr/sbin/zhack
-sbin/zinject
+usr/sbin/zinject
-sbin/zpool
+usr/sbin/zpool
-sbin/zstream
+usr/sbin/zstream
-sbin/zstreamdump
+usr/sbin/zstreamdump
 usr/bin/zvol_wait
-usr/lib/modules-load.d/ lib/
+usr/lib/modules-load.d/
 usr/lib/zfs-linux/zpool.d/
 usr/lib/zfs-linux/zpool_influxdb
 usr/lib/zfs-linux/zfs_prepare_disk
-usr/sbin/arc_summary
+usr/bin/zarcsummary
-usr/sbin/arcstat
+usr/bin/zarcstat
-usr/sbin/dbufstat
+usr/bin/dbufstat usr/sbin
-usr/sbin/zilstat
+usr/bin/zilstat
 usr/share/zfs/compatibility.d/
 usr/share/bash-completion/completions
-usr/share/man/man1/arcstat.1
+usr/share/man/man1/zarcstat.1
 usr/share/man/man1/zhack.1
 usr/share/man/man1/zvol_wait.1
 usr/share/man/man5/

contrib/debian/openzfs-zfsutils.links

@@ -0,0 +1,3 @@
+usr/sbin/zfs      usr/bin/zfs
+usr/sbin/zpool    usr/bin/zpool
+usr/lib/zfs-linux/zpool_influxdb  usr/bin/zpool_influxdb

contrib/debian/rules.in

@@ -37,18 +37,19 @@ override_dh_auto_configure:
 	@# Build the userland, but don't build the kernel modules.
 	dh_auto_configure -- @CFGOPTS@ \
 	  --bindir=/usr/bin \
-	  --sbindir=/sbin \
+	  --sbindir=/usr/sbin \
-	  --libdir=/lib/"$(DEB_HOST_MULTIARCH)" \
+	  --with-mounthelperdir=/usr/sbin \
-	  --with-udevdir=/lib/udev \
+	  --libdir=/usr/lib/"$(DEB_HOST_MULTIARCH)" \
+	  --with-udevdir=/usr/lib/udev \
 	  --with-zfsexecdir=/usr/lib/zfs-linux \
 	  --enable-systemd \
 	  --enable-pyzfs \
 	  --with-python=python3 \
-	  --with-pammoduledir='/lib/$(DEB_HOST_MULTIARCH)/security' \
+	  --with-pammoduledir='/usr/lib/$(DEB_HOST_MULTIARCH)/security' \
 	  --with-pkgconfigdir='/usr/lib/$(DEB_HOST_MULTIARCH)/pkgconfig' \
-	  --with-systemdunitdir=/lib/systemd/system \
+	  --with-systemdunitdir=/usr/lib/systemd/system \
-	  --with-systemdpresetdir=/lib/systemd/system-preset \
+	  --with-systemdpresetdir=/usr/lib/systemd/system-preset \
-	  --with-systemdgeneratordir=/lib/systemd/system-generators \
+	  --with-systemdgeneratordir=/usr/lib/systemd/system-generators \
 	  --with-config=user
 
 	for i in $(wildcard $(CURDIR)/debian/*.install.in) ; do \
@@ -77,19 +78,6 @@ override_dh_auto_install:
 	@# Install the utilities.
 	$(MAKE) install DESTDIR='$(CURDIR)/debian/tmp'
 
-	# Move from bin_dir to /usr/sbin
-	# Remove suffix (.py) as per policy 10.4 - Scripts
-	# https://www.debian.org/doc/debian-policy/ch-files.html#s-scripts
-	mkdir -p '$(CURDIR)/debian/tmp/usr/sbin/'
-	mv '$(CURDIR)/debian/tmp/usr/bin/arc_summary' '$(CURDIR)/debian/tmp/usr/sbin/arc_summary'
-	mv '$(CURDIR)/debian/tmp/usr/bin/arcstat' '$(CURDIR)/debian/tmp/usr/sbin/arcstat'
-	mv '$(CURDIR)/debian/tmp/usr/bin/dbufstat' '$(CURDIR)/debian/tmp/usr/sbin/dbufstat'
-	mv '$(CURDIR)/debian/tmp/usr/bin/zilstat' '$(CURDIR)/debian/tmp/usr/sbin/zilstat'
-
-	@# Zed has dependencies outside of the system root.
-	mv '$(CURDIR)/debian/tmp/sbin/zed' '$(CURDIR)/debian/tmp/usr/sbin/zed'
-	sed -i 's|ExecStart=/sbin/|ExecStart=/usr/sbin/|g' '$(CURDIR)/debian/tmp/lib/systemd/system/zfs-zed.service'
-
 	@# Install the DKMS source.
 	@# We only want the files needed to build the modules
 	install -D -t '$(CURDIR)/debian/tmp/usr/src/$(NAME)-$(DEB_VERSION_UPSTREAM)/scripts' \
@@ -131,11 +119,6 @@ override_dh_auto_install:
 	cd '$(CURDIR)/debian/tmp/usr/src/$(NAME)-$(DEB_VERSION_UPSTREAM)'; ./autogen.sh
 	rm -fr '$(CURDIR)/debian/tmp/usr/src/$(NAME)-$(DEB_VERSION_UPSTREAM)/autom4te.cache'
 
-	for i in `ls $(CURDIR)/debian/tmp/lib/$(DEB_HOST_MULTIARCH)/*.so`; do \
-		ln -s '/lib/$(DEB_HOST_MULTIARCH)/'`readlink $${i}` '$(CURDIR)/debian/tmp/usr/lib/$(DEB_HOST_MULTIARCH)/'`basename $${i}`; \
-		rm $${i}; \
-	done
-
 	chmod a-x '$(CURDIR)/debian/tmp/etc/zfs/zfs-functions'
 	chmod a-x '$(CURDIR)/debian/tmp/etc/default/zfs'
 
@@ -159,7 +142,7 @@ override_dh_auto_clean:
 	@if test -e META.orig; then mv META.orig META; fi
 
 override_dh_install:
-	find debian/tmp/lib -name '*.la' -delete
+	find debian/tmp/usr/lib -name '*.la' -delete
 	dh_install
 
 override_dh_missing:
@@ -173,8 +156,8 @@ override_dh_installinit:
 	dh_installinit -R --name zfs-zed
 
 override_dh_installsystemd:
-	mkdir -p debian/openzfs-zfsutils/lib/systemd/system
+	mkdir -p debian/openzfs-zfsutils/usr/lib/systemd/system
-	ln -sr /dev/null debian/openzfs-zfsutils/lib/systemd/system/zfs-import.service
+	ln -sr /dev/null debian/openzfs-zfsutils/usr/lib/systemd/system/zfs-import.service
 	dh_installsystemd --no-stop-on-upgrade -X zfs-zed.service
 	dh_installsystemd --name zfs-zed
 

contrib/debian/tree/zfs-initramfs/usr/share/initramfs-tools/hooks/zdev

@@ -5,7 +5,7 @@
 
 PREREQ="udev"
 PREREQ_UDEV_RULES="60-zvol.rules 69-vdev.rules"
-COPY_EXEC_LIST="/lib/udev/zvol_id /lib/udev/vdev_id"
+COPY_EXEC_LIST="/usr/lib/udev/zvol_id /usr/lib/udev/vdev_id"
 
 # Generic result code.
 RC=0

contrib/dracut/90zfs/module-setup.sh.in

@@ -16,7 +16,8 @@ depends() {
 }
 
 installkernel() {
-	instmods -c zfs
+	hostonly='' instmods -c zfs
+	instmods mpt3sas virtio_blk
 }
 
 install() {

contrib/initramfs/hooks/zfsunlock.in

@@ -8,3 +8,12 @@ fi
 . /usr/share/initramfs-tools/hook-functions
 
 copy_exec /usr/share/initramfs-tools/zfsunlock /usr/bin/zfsunlock
+
+if [ -f /etc/initramfs-tools/etc/motd ]; then
+	copy_file text /etc/initramfs-tools/etc/motd /etc/motd
+else
+	tmpf=$(mktemp)
+	echo "If you use zfs encrypted root filesystems, you can use \`zfsunlock\` to manually unlock it" > "$tmpf"
+	copy_file text "$tmpf" /etc/motd
+	rm -f "$tmpf"
+fi

contrib/intel_qat/readme.md

@@ -8,7 +8,7 @@ This contrib contains community compatibility patches to get Intel QAT working o
 These patches are based on the following Intel QAT version:
 [1.7.l.4.10.0-00014](https://01.org/sites/default/files/downloads/qat1.7.l.4.10.0-00014.tar.gz)
 
-When using QAT with above kernels versions, the following patches needs to be applied using:
+When using QAT with the above kernel versions, the following patches need to be applied using:
 patch -p1 < _$PATCH_
 _Where $PATCH refers to the path of the patch in question_
 

contrib/pam_zfs_key/pam_zfs_key.c

@@ -391,7 +391,11 @@ static int
 zfs_key_config_load(pam_handle_t *pamh, zfs_key_config_t *config,
     int argc, const char **argv)
 {
+#if	defined(__FreeBSD__)
+	config->homes_prefix = strdup("zroot/home");
+#else
 	config->homes_prefix = strdup("rpool/home");
+#endif
 	if (config->homes_prefix == NULL) {
 		pam_syslog(pamh, LOG_ERR, "strdup failure");
 		return (PAM_SERVICE_ERR);

contrib/pyzfs/libzfs_core/exceptions.py

@@ -604,5 +604,4 @@ class RaidzExpansionRunning(ZFSError):
     errno = ZFS_ERR_RAIDZ_EXPAND_IN_PROGRESS
     message = "A raidz device is currently expanding"
 
-
 # vim: softtabstop=4 tabstop=4 expandtab shiftwidth=4

contrib/pyzfs/libzfs_core/test/test_libzfs_core.py

@@ -4223,7 +4223,7 @@ class _TempPool(object):
             self.getRoot().reset()
             return
 
-        # On the Buildbot builders this may fail with "pool is busy"
+        # On the CI builders this may fail with "pool is busy"
         # Retry 5 times before raising an error
         retry = 0
         while True:

etc/init.d/README.md

@@ -1,5 +1,5 @@
 DESCRIPTION
-  These script were written with the primary intention of being portable and
+  These scripts were written with the primary intention of being portable and
   usable on as many systems as possible.
 
   This is, in practice, usually not possible. But the intention is there.

include/os/freebsd/spl/sys/time.h

@@ -62,6 +62,17 @@ typedef longlong_t	hrtime_t;
 #define	SEC_TO_TICK(sec)	((sec) * hz)
 #define	NSEC_TO_TICK(nsec)	((nsec) / (NANOSEC / hz))
 
+static __inline hrtime_t
+getlrtime(void)
+{
+	struct timespec ts;
+	hrtime_t nsec;
+
+	getnanouptime(&ts);
+	nsec = ((hrtime_t)ts.tv_sec * NANOSEC) + ts.tv_nsec;
+	return (nsec);
+}
+
 static __inline hrtime_t
 gethrtime(void)
 {

include/os/linux/kernel/linux/blkdev_compat.h

@@ -542,24 +542,6 @@ blk_generic_alloc_queue(make_request_fn make_request, int node_id)
 }
 #endif /* !HAVE_SUBMIT_BIO_IN_BLOCK_DEVICE_OPERATIONS */
 
-/*
- * All the io_*() helper functions below can operate on a bio, or a rq, but
- * not both.  The older submit_bio() codepath will pass a bio, and the
- * newer blk-mq codepath will pass a rq.
- */
-static inline int
-io_data_dir(struct bio *bio, struct request *rq)
-{
-	if (rq != NULL) {
-		if (op_is_write(req_op(rq))) {
-			return (WRITE);
-		} else {
-			return (READ);
-		}
-	}
-	return (bio_data_dir(bio));
-}
-
 static inline int
 io_is_flush(struct bio *bio, struct request *rq)
 {

include/os/linux/kernel/linux/dcache_compat.h

@@ -60,32 +60,6 @@
 	} while (0)
 #endif
 
-/*
- * 2.6.30 API change,
- * The const keyword was added to the 'struct dentry_operations' in
- * the dentry structure.  To handle this we define an appropriate
- * dentry_operations_t typedef which can be used.
- */
-typedef const struct dentry_operations	dentry_operations_t;
-
-/*
- * 2.6.38 API addition,
- * Added d_clear_d_op() helper function which clears some flags and the
- * registered dentry->d_op table.  This is required because d_set_d_op()
- * issues a warning when the dentry operations table is already set.
- * For the .zfs control directory to work properly we must be able to
- * override the default operations table and register custom .d_automount
- * and .d_revalidate callbacks.
- */
-static inline void
-d_clear_d_op(struct dentry *dentry)
-{
-	dentry->d_op = NULL;
-	dentry->d_flags &= ~(
-	    DCACHE_OP_HASH | DCACHE_OP_COMPARE |
-	    DCACHE_OP_REVALIDATE | DCACHE_OP_DELETE);
-}
-
 /*
  * Walk and invalidate all dentry aliases of an inode
  * unless it's a mountpoint

include/os/linux/spl/sys/rwlock.h

@@ -130,7 +130,7 @@ RW_READ_HELD(krwlock_t *rwp)
 /*
  * The Linux rwsem implementation does not require a matching destroy.
  */
-#define	rw_destroy(rwp)		((void) 0)
+#define	rw_destroy(rwp)		ASSERT(!(RW_LOCK_HELD(rwp)))
 
 /*
  * Upgrading a rwsem from a reader to a writer is not supported by the

include/os/linux/spl/sys/stat.h

@@ -25,6 +25,6 @@
 #ifndef _SPL_STAT_H
 #define	_SPL_STAT_H
 
-#include <linux/stat.h>
+#include <sys/stat.h>
 
 #endif /* SPL_STAT_H */

include/os/linux/spl/sys/time.h

@@ -79,6 +79,14 @@ gethrestime_sec(void)
 	return (ts.tv_sec);
 }
 
+static inline hrtime_t
+getlrtime(void)
+{
+	inode_timespec_t ts;
+	ktime_get_coarse_ts64(&ts);
+	return (((hrtime_t)ts.tv_sec * NSEC_PER_SEC) + ts.tv_nsec);
+}
+
 static inline hrtime_t
 gethrtime(void)
 {

include/os/linux/zfs/sys/zpl.h

@@ -55,6 +55,7 @@ extern const struct file_operations zpl_dir_file_operations;
 extern void zpl_prune_sb(uint64_t nr_to_scan, void *arg);
 
 extern const struct super_operations zpl_super_operations;
+extern const struct dentry_operations zpl_dentry_operations;
 extern const struct export_operations zpl_export_operations;
 extern struct file_system_type zpl_fs_type;
 

include/sys/dmu.h

@@ -414,9 +414,9 @@ typedef struct dmu_buf {
 #define	DMU_POOL_ZPOOL_CHECKPOINT	"com.delphix:zpool_checkpoint"
 #define	DMU_POOL_LOG_SPACEMAP_ZAP	"com.delphix:log_spacemap_zap"
 #define	DMU_POOL_DELETED_CLONES		"com.delphix:deleted_clones"
-#define	DMU_POOL_TXG_LOG_TIME_MINUTES	"com.klaraystems:txg_log_time:minutes"
+#define	DMU_POOL_TXG_LOG_TIME_MINUTES	"com.klarasystems:txg_log_time:minutes"
-#define	DMU_POOL_TXG_LOG_TIME_DAYS	"com.klaraystems:txg_log_time:days"
+#define	DMU_POOL_TXG_LOG_TIME_DAYS	"com.klarasystems:txg_log_time:days"
-#define	DMU_POOL_TXG_LOG_TIME_MONTHS	"com.klaraystems:txg_log_time:months"
+#define	DMU_POOL_TXG_LOG_TIME_MONTHS	"com.klarasystems:txg_log_time:months"
 
 /*
  * Allocate an object from this objset.  The range of object numbers

include/sys/dsl_deleg.h

@@ -46,6 +46,7 @@ extern "C" {
 #define	ZFS_DELEG_PERM_MOUNT		"mount"
 #define	ZFS_DELEG_PERM_SHARE		"share"
 #define	ZFS_DELEG_PERM_SEND		"send"
+#define	ZFS_DELEG_PERM_SEND_RAW		"send:raw"
 #define	ZFS_DELEG_PERM_RECEIVE		"receive"
 #define	ZFS_DELEG_PERM_RECEIVE_APPEND	"receive:append"
 #define	ZFS_DELEG_PERM_ALLOW		"allow"

include/sys/fm/fs/zfs.h

@@ -58,6 +58,7 @@ extern "C" {
 #define	FM_EREPORT_ZFS_PROBE_FAILURE		"probe_failure"
 #define	FM_EREPORT_ZFS_LOG_REPLAY		"log_replay"
 #define	FM_EREPORT_ZFS_CONFIG_CACHE_WRITE	"config_cache_write"
+#define	FM_EREPORT_ZFS_SITOUT			"sitout"
 
 #define	FM_EREPORT_PAYLOAD_ZFS_POOL		"pool"
 #define	FM_EREPORT_PAYLOAD_ZFS_POOL_FAILMODE	"pool_failmode"

include/sys/fs/zfs.h

@@ -385,6 +385,8 @@ typedef enum {
 	VDEV_PROP_TRIM_SUPPORT,
 	VDEV_PROP_TRIM_ERRORS,
 	VDEV_PROP_SLOW_IOS,
+	VDEV_PROP_SIT_OUT,
+	VDEV_PROP_AUTOSIT,
 	VDEV_NUM_PROPS
 } vdev_prop_t;
 
@@ -746,6 +748,8 @@ typedef struct zpool_load_policy {
 #define	ZPOOL_CONFIG_METASLAB_SHIFT	"metaslab_shift"
 #define	ZPOOL_CONFIG_ASHIFT		"ashift"
 #define	ZPOOL_CONFIG_ASIZE		"asize"
+#define	ZPOOL_CONFIG_MIN_ALLOC		"min_alloc"
+#define	ZPOOL_CONFIG_MAX_ALLOC		"max_alloc"
 #define	ZPOOL_CONFIG_DTL		"DTL"
 #define	ZPOOL_CONFIG_SCAN_STATS		"scan_stats"	/* not stored on disk */
 #define	ZPOOL_CONFIG_REMOVAL_STATS	"removal_stats"	/* not stored on disk */
@@ -1673,6 +1677,7 @@ typedef enum {
 	ZFS_ERR_RAIDZ_EXPAND_IN_PROGRESS,
 	ZFS_ERR_ASHIFT_MISMATCH,
 	ZFS_ERR_STREAM_LARGE_MICROZAP,
+	ZFS_ERR_TOO_MANY_SITOUTS,
 } zfs_errno_t;
 
 /*

include/sys/range_tree.h

@@ -238,8 +238,7 @@ zfs_rs_set_end_raw(zfs_range_seg_t *rs, zfs_range_tree_t *rt, uint64_t end)
 }
 
 static inline void
-zfs_zfs_rs_set_fill_raw(zfs_range_seg_t *rs, zfs_range_tree_t *rt,
+zfs_rs_set_fill_raw(zfs_range_seg_t *rs, zfs_range_tree_t *rt, uint64_t fill)
-    uint64_t fill)
 {
 	ASSERT3U(rt->rt_type, <=, ZFS_RANGE_SEG_NUM_TYPES);
 	switch (rt->rt_type) {
@@ -277,7 +276,7 @@ static inline void
 zfs_rs_set_fill(zfs_range_seg_t *rs, zfs_range_tree_t *rt, uint64_t fill)
 {
 	ASSERT(IS_P2ALIGNED(fill, 1ULL << rt->rt_shift));
-	zfs_zfs_rs_set_fill_raw(rs, rt, fill >> rt->rt_shift);
+	zfs_rs_set_fill_raw(rs, rt, fill >> rt->rt_shift);
 }
 
 typedef void zfs_range_tree_func_t(void *arg, uint64_t start, uint64_t size);

include/sys/spa.h

@@ -1084,6 +1084,7 @@ extern pool_state_t spa_state(spa_t *spa);
 extern spa_load_state_t spa_load_state(spa_t *spa);
 extern uint64_t spa_freeze_txg(spa_t *spa);
 extern uint64_t spa_get_worst_case_asize(spa_t *spa, uint64_t lsize);
+extern void spa_get_min_alloc_range(spa_t *spa, uint64_t *min, uint64_t *max);
 extern uint64_t spa_get_dspace(spa_t *spa);
 extern uint64_t spa_get_checkpoint_space(spa_t *spa);
 extern uint64_t spa_get_slop_space(spa_t *spa);

include/sys/spa_impl.h

@@ -265,6 +265,7 @@ struct spa {
 	uint64_t	spa_min_ashift;		/* of vdevs in normal class */
 	uint64_t	spa_max_ashift;		/* of vdevs in normal class */
 	uint64_t	spa_min_alloc;		/* of vdevs in normal class */
+	uint64_t	spa_max_alloc;		/* of vdevs in normal class */
 	uint64_t	spa_gcd_alloc;		/* of vdevs in normal class */
 	uint64_t	spa_config_guid;	/* config pool guid */
 	uint64_t	spa_load_guid;		/* spa_load initialized guid */

include/sys/vdev_impl.h

@@ -279,10 +279,12 @@ struct vdev {
 	uint64_t	vdev_noalloc;	/* device is passivated?	*/
 	uint64_t	vdev_removing;	/* device is being removed?	*/
 	uint64_t	vdev_failfast;	/* device failfast setting	*/
+	boolean_t	vdev_autosit;	/* automatic sitout management	*/
 	boolean_t	vdev_rz_expanding; /* raidz is being expanded?	*/
 	boolean_t	vdev_ishole;	/* is a hole in the namespace	*/
 	uint64_t	vdev_top_zap;
 	vdev_alloc_bias_t vdev_alloc_bias; /* metaslab allocation bias	*/
+	uint64_t	vdev_last_latency_check;
 
 	/* pool checkpoint related */
 	space_map_t	*vdev_checkpoint_sm;	/* contains reserved blocks */
@@ -431,6 +433,10 @@ struct vdev {
 	hrtime_t	vdev_mmp_pending; /* 0 if write finished	*/
 	uint64_t	vdev_mmp_kstat_id;	/* to find kstat entry */
 	uint64_t	vdev_expansion_time;	/* vdev's last expansion time */
+	/* used to calculate average read latency */
+	uint64_t	*vdev_prev_histo;
+	int64_t		vdev_outlier_count;	/* read outlier amongst peers */
+	hrtime_t	vdev_read_sit_out_expire; /* end of sit out period    */
 	list_node_t	vdev_leaf_node;		/* leaf vdev list */
 
 	/*

include/sys/vdev_raidz.h

@@ -61,6 +61,9 @@ void vdev_raidz_checksum_error(zio_t *, struct raidz_col *, abd_t *);
 struct raidz_row *vdev_raidz_row_alloc(int, zio_t *);
 void vdev_raidz_reflow_copy_scratch(spa_t *);
 void raidz_dtl_reassessed(vdev_t *);
+boolean_t vdev_sit_out_reads(vdev_t *, zio_flag_t);
+void vdev_raidz_sit_child(vdev_t *, uint64_t);
+void vdev_raidz_unsit_child(vdev_t *);
 
 extern const zio_vsd_ops_t vdev_raidz_vsd_ops;
 

include/sys/vdev_raidz_impl.h

@@ -119,6 +119,7 @@ typedef struct raidz_col {
 	uint8_t rc_need_orig_restore:1;	/* need to restore from orig_data? */
 	uint8_t rc_force_repair:1;	/* Write good data to this column */
 	uint8_t rc_allow_repair:1;	/* Allow repair I/O to this column */
+	uint8_t rc_latency_outlier:1;	/* Latency outlier for this device */
 	int rc_shadow_devidx;		/* for double write during expansion */
 	int rc_shadow_error;		/* for double write during expansion */
 	uint64_t rc_shadow_offset;	/* for double write during expansion */
@@ -133,6 +134,7 @@ typedef struct raidz_row {
 	int rr_firstdatacol;		/* First data column/parity count */
 	abd_t *rr_abd_empty;		/* dRAID empty sector buffer */
 	int rr_nempty;			/* empty sectors included in parity */
+	int rr_outlier_cnt;		/* Count of latency outlier devices */
 #ifdef ZFS_DEBUG
 	uint64_t rr_offset;		/* Logical offset for *_io_verify() */
 	uint64_t rr_size;		/* Physical size for *_io_verify() */

include/sys/zfs_file.h

@@ -46,7 +46,7 @@ void zfs_file_close(zfs_file_t *fp);
 
 int zfs_file_write(zfs_file_t *fp, const void *buf, size_t len, ssize_t *resid);
 int zfs_file_pwrite(zfs_file_t *fp, const void *buf, size_t len, loff_t off,
-    ssize_t *resid);
+    uint8_t ashift, ssize_t *resid);
 int zfs_file_read(zfs_file_t *fp, void *buf, size_t len, ssize_t *resid);
 int zfs_file_pread(zfs_file_t *fp, void *buf, size_t len, loff_t off,
     ssize_t *resid);

include/sys/zfs_ioctl.h

@@ -455,6 +455,7 @@ typedef enum zinject_type {
 	ZINJECT_DECRYPT_FAULT,
 	ZINJECT_DELAY_IMPORT,
 	ZINJECT_DELAY_EXPORT,
+	ZINJECT_DELAY_READY,
 } zinject_type_t;
 
 typedef enum zinject_iotype {

include/sys/zio.h

@@ -82,7 +82,8 @@ gbh_nblkptrs(uint64_t size) {
 static inline zio_eck_t *
 gbh_eck(zio_gbh_phys_t *gbh, uint64_t size) {
 	ASSERT(IS_P2ALIGNED(size, sizeof (blkptr_t)));
-	return ((zio_eck_t *)((uintptr_t)gbh + size - sizeof (zio_eck_t)));
+	return ((zio_eck_t *)((uintptr_t)gbh + (size_t)size -
+	    sizeof (zio_eck_t)));
 }
 
 static inline blkptr_t *
@@ -717,6 +718,7 @@ extern void zio_handle_ignored_writes(zio_t *zio);
 extern hrtime_t zio_handle_io_delay(zio_t *zio);
 extern void zio_handle_import_delay(spa_t *spa, hrtime_t elapsed);
 extern void zio_handle_export_delay(spa_t *spa, hrtime_t elapsed);
+extern hrtime_t zio_handle_ready_delay(zio_t *zio);
 
 /*
  * Checksum ereport functions

include/sys/zvol.h

@@ -53,7 +53,7 @@ extern int zvol_set_volsize(const char *, uint64_t);
 extern int zvol_set_volthreading(const char *, boolean_t);
 extern int zvol_set_common(const char *, zfs_prop_t, zprop_source_t, uint64_t);
 extern int zvol_set_ro(const char *, boolean_t);
-extern zvol_state_handle_t *zvol_suspend(const char *);
+extern int zvol_suspend(const char *, zvol_state_handle_t **);
 extern int zvol_resume(zvol_state_handle_t *);
 extern void *zvol_tag(zvol_state_handle_t *);
 

include/zfs_deleg.h

@@ -55,6 +55,7 @@ typedef enum {
 	ZFS_DELEG_NOTE_PROMOTE,
 	ZFS_DELEG_NOTE_RENAME,
 	ZFS_DELEG_NOTE_SEND,
+	ZFS_DELEG_NOTE_SEND_RAW,
 	ZFS_DELEG_NOTE_RECEIVE,
 	ZFS_DELEG_NOTE_ALLOW,
 	ZFS_DELEG_NOTE_USERPROP,

lib/libspl/Makefile.am

@@ -2,6 +2,9 @@ include $(srcdir)/%D%/include/Makefile.am
 
 libspl_assert_la_CFLAGS = $(AM_CFLAGS) $(LIBRARY_CFLAGS) $(LIBUNWIND_CFLAGS)
 libspl_la_CFLAGS        = $(libspl_assert_la_CFLAGS)
+if TARGET_CPU_I386
+libspl_la_CFLAGS        += $(NO_ATOMIC_ALIGNMENT)
+endif
 
 noinst_LTLIBRARIES += libspl_assert.la libspl.la
 CPPCHECKTARGETS    += libspl_assert.la libspl.la

lib/libspl/include/os/linux/sys/stat.h

@@ -33,7 +33,7 @@
 
 #ifdef HAVE_STATX
 #include <fcntl.h>
-#include <linux/stat.h>
+#include <sys/stat.h>
 #endif
 
 /*

lib/libspl/include/sys/time.h

@@ -97,6 +97,15 @@ gethrestime_sec(void)
 	return (tv.tv_sec);
 }
 
+static inline hrtime_t
+getlrtime(void)
+{
+	struct timeval tv;
+	(void) gettimeofday(&tv, NULL);
+	return ((((uint64_t)tv.tv_sec) * NANOSEC) +
+	    ((uint64_t)tv.tv_usec * NSEC_PER_USEC));
+}
+
 static inline hrtime_t
 gethrtime(void)
 {

lib/libzfs/libzfs.abi

@@ -6117,7 +6117,9 @@
       <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_SIT_OUT' value='52'/>
+      <enumerator name='VDEV_PROP_AUTOSIT' value='53'/>
+      <enumerator name='VDEV_NUM_PROPS' value='54'/>
     </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'>

lib/libzfs/libzfs_pool.c

@@ -5549,6 +5549,8 @@ zpool_get_vdev_prop_value(nvlist_t *nvprop, vdev_prop_t prop, char *prop_name,
 			/* Only use if provided by the RAIDZ VDEV above */
 			if (prop == VDEV_PROP_RAIDZ_EXPANDING)
 				return (ENOENT);
+			if (prop == VDEV_PROP_SIT_OUT)
+				return (ENOENT);
 		}
 		if (vdev_prop_index_to_string(prop, intval,
 		    (const char **)&strval) != 0)
@@ -5718,8 +5720,16 @@ zpool_set_vdev_prop(zpool_handle_t *zhp, const char *vdevname,
 	nvlist_free(nvl);
 	nvlist_free(outnvl);
 
-	if (ret)
+	if (ret) {
-		(void) zpool_standard_error(zhp->zpool_hdl, errno, errbuf);
+		if (errno == ENOTSUP) {
+			zfs_error_aux(zhp->zpool_hdl, dgettext(TEXT_DOMAIN,
+			    "property not supported for this vdev"));
+			(void) zfs_error(zhp->zpool_hdl, EZFS_PROPTYPE, errbuf);
+		} else {
+			(void) zpool_standard_error(zhp->zpool_hdl, errno,
+			    errbuf);
+		}
+	}
 
 	return (ret);
 }

lib/libzfs/libzfs_util.c

@@ -776,6 +776,11 @@ zpool_standard_error_fmt(libzfs_handle_t *hdl, int error, const char *fmt, ...)
 	case ZFS_ERR_ASHIFT_MISMATCH:
 		zfs_verror(hdl, EZFS_ASHIFT_MISMATCH, fmt, ap);
 		break;
+	case ZFS_ERR_TOO_MANY_SITOUTS:
+		zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, "too many disks "
+		    "already sitting out"));
+		zfs_verror(hdl, EZFS_BUSY, fmt, ap);
+		break;
 	default:
 		zfs_error_aux(hdl, "%s", zfs_strerror(error));
 		zfs_verror(hdl, EZFS_UNKNOWN, fmt, ap);

lib/libzpool/kernel.c

@@ -1238,7 +1238,7 @@ zfs_file_write(zfs_file_t *fp, const void *buf, size_t count, ssize_t *resid)
  */
 int
 zfs_file_pwrite(zfs_file_t *fp, const void *buf,
-    size_t count, loff_t pos, ssize_t *resid)
+    size_t count, loff_t pos, uint8_t ashift, ssize_t *resid)
 {
 	ssize_t rc, split, done;
 	int sectors;
@@ -1248,8 +1248,8 @@ zfs_file_pwrite(zfs_file_t *fp, const void *buf,
 	 * system calls so that the process can be killed in between.
 	 * This is used by ztest to simulate realistic failure modes.
 	 */
-	sectors = count >> SPA_MINBLOCKSHIFT;
+	sectors = count >> ashift;
-	split = (sectors > 0 ? rand() % sectors : 0) << SPA_MINBLOCKSHIFT;
+	split = (sectors > 0 ? rand() % sectors : 0) << ashift;
 	rc = pwrite64(fp->f_fd, buf, split, pos);
 	if (rc != -1) {
 		done = rc;

man/Makefile.am

@@ -2,7 +2,7 @@ dist_noinst_man_MANS = \
 	%D%/man1/cstyle.1
 
 dist_man_MANS = \
-	%D%/man1/arcstat.1 \
+	%D%/man1/zarcstat.1 \
 	%D%/man1/raidz_test.1 \
 	%D%/man1/test-runner.1 \
 	%D%/man1/zhack.1 \
@@ -124,10 +124,21 @@ dist_noinst_DATA += $(dist_noinst_man_MANS) $(dist_man_MANS)
 
 SUBSTFILES += $(nodist_man_MANS)
 
-CHECKS += mancheck
+MANFILES = $(dist_noinst_man_MANS) $(dist_man_MANS) $(nodist_man_MANS)
-mancheck:
-	$(top_srcdir)/scripts/mancheck.sh $(srcdir)/%D%
 
+PHONY += mancheck
+
+mancheck_verbose = $(mancheck_verbose_@AM_V@)
+mancheck_verbose_ = $(mancheck_verbose_@AM_DEFAULT_V@)
+mancheck_verbose_0 = @echo MANCHECK $(_MTGT);
+
+_MTGT = $(subst ^,/,$(subst mancheck-,,$@))
+mancheck-%:
+	$(mancheck_verbose)scripts/mancheck.sh $(_MTGT)
+
+mancheck: $(foreach manfile, $(MANFILES), $(addprefix mancheck-,$(subst /,^,$(manfile))))
+
+CHECKS += mancheck
 
 if BUILD_LINUX
 # The manual pager in most Linux distros defaults to "BSD" when .Os is blank,

man/man1/cstyle.1

@@ -21,7 +21,7 @@
 .\"
 .\" CDDL HEADER END
 .\"
-.Dd May 26, 2021
+.Dd April 4, 2022
 .Dt CSTYLE 1
 .Os
 .