Tag 2.3.0-rc3

Signed-off-by: Brian Behlendorf <behlendorf1@llnl.gov>

.github/workflows/codeql.yml

@@ -11,7 +11,7 @@ concurrency:
 jobs:
   analyze:
     name: Analyze
-    runs-on: ubuntu-latest
+    runs-on: ubuntu-22.04
     permissions:
       actions: read
       contents: read

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

@@ -18,19 +18,21 @@ ssh-keygen -t ed25519 -f ~/.ssh/id_ed25519 -q -N ""
 
 # we expect RAM shortage
 cat << EOF | sudo tee /etc/ksmtuned.conf > /dev/null
+# /etc/ksmtuned.conf - Configuration file for ksmtuned
 # https://docs.redhat.com/en/documentation/red_hat_enterprise_linux/7/html/virtualization_tuning_and_optimization_guide/chap-ksm
 KSM_MONITOR_INTERVAL=60
 
 # Millisecond sleep between ksm scans for 16Gb server.
 # Smaller servers sleep more, bigger sleep less.
-KSM_SLEEP_MSEC=10
+KSM_SLEEP_MSEC=30
-KSM_NPAGES_BOOST=300
-KSM_NPAGES_DECAY=-50
-KSM_NPAGES_MIN=64
-KSM_NPAGES_MAX=2048
 
-KSM_THRES_COEF=25
+KSM_NPAGES_BOOST=0
-KSM_THRES_CONST=2048
+KSM_NPAGES_DECAY=0
+KSM_NPAGES_MIN=1000
+KSM_NPAGES_MAX=25000
+
+KSM_THRES_COEF=80
+KSM_THRES_CONST=8192
 
 LOGFILE=/var/log/ksmtuned.log
 DEBUG=1

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

@@ -14,7 +14,7 @@ OSv=$OS
 
 # compressed with .zst extension
 REPO="https://github.com/mcmilk/openzfs-freebsd-images"
-FREEBSD="$REPO/releases/download/v2024-09-16"
+FREEBSD="$REPO/releases/download/v2024-10-05"
 URLzs=""
 
 # Ubuntu mirrors
@@ -52,43 +52,55 @@ case "$OS" in
     OSNAME="Debian 12"
     URL="https://cloud.debian.org/images/cloud/bookworm/latest/debian-12-generic-amd64.qcow2"
     ;;
-  fedora39)
-    OSNAME="Fedora 39"
-    OSv="fedora39"
-    URL="https://download.fedoraproject.org/pub/fedora/linux/releases/39/Cloud/x86_64/images/Fedora-Cloud-Base-39-1.5.x86_64.qcow2"
-    ;;
   fedora40)
     OSNAME="Fedora 40"
-    OSv="fedora39"
+    OSv="fedora-unknown"
     URL="https://download.fedoraproject.org/pub/fedora/linux/releases/40/Cloud/x86_64/images/Fedora-Cloud-Base-Generic.x86_64-40-1.14.qcow2"
     ;;
-  freebsd13r)
+  fedora41)
+    OSNAME="Fedora 41"
+    OSv="fedora-unknown"
+    URL="https://download.fedoraproject.org/pub/fedora/linux/releases/41/Cloud/x86_64/images/Fedora-Cloud-Base-Generic-41-1.4.x86_64.qcow2"
+    ;;
+  freebsd13-3r)
+    OSNAME="FreeBSD 13.3-RELEASE"
+    OSv="freebsd13.0"
+    URLzs="$FREEBSD/amd64-freebsd-13.3-RELEASE.qcow2.zst"
+    BASH="/usr/local/bin/bash"
+    NIC="rtl8139"
+    ;;
+  freebsd13-4r)
     OSNAME="FreeBSD 13.4-RELEASE"
     OSv="freebsd13.0"
     URLzs="$FREEBSD/amd64-freebsd-13.4-RELEASE.qcow2.zst"
     BASH="/usr/local/bin/bash"
     NIC="rtl8139"
     ;;
-  freebsd13)
+  freebsd14-0r)
-    OSNAME="FreeBSD 13.4-STABLE"
+    OSNAME="FreeBSD 14.0-RELEASE"
-    OSv="freebsd13.0"
+    OSv="freebsd14.0"
-    URLzs="$FREEBSD/amd64-freebsd-13.4-STABLE.qcow2.zst"
+    URLzs="$FREEBSD/amd64-freebsd-14.0-RELEASE.qcow2.zst"
     BASH="/usr/local/bin/bash"
-    NIC="rtl8139"
     ;;
-  freebsd14r)
+  freebsd14-1r)
     OSNAME="FreeBSD 14.1-RELEASE"
     OSv="freebsd14.0"
     URLzs="$FREEBSD/amd64-freebsd-14.1-RELEASE.qcow2.zst"
     BASH="/usr/local/bin/bash"
     ;;
-  freebsd14)
+  freebsd13-4s)
+    OSNAME="FreeBSD 13.4-STABLE"
+    OSv="freebsd13.0"
+    URLzs="$FREEBSD/amd64-freebsd-13.4-STABLE.qcow2.zst"
+    BASH="/usr/local/bin/bash"
+    ;;
+  freebsd14-1s)
     OSNAME="FreeBSD 14.1-STABLE"
     OSv="freebsd14.0"
     URLzs="$FREEBSD/amd64-freebsd-14.1-STABLE.qcow2.zst"
     BASH="/usr/local/bin/bash"
     ;;
-  freebsd15)
+  freebsd15-0c)
     OSNAME="FreeBSD 15.0-CURRENT"
     OSv="freebsd14.0"
     URLzs="$FREEBSD/amd64-freebsd-15.0-CURRENT.qcow2.zst"

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

@@ -13,10 +13,10 @@ function archlinux() {
   echo "##[endgroup]"
 
   echo "##[group]Install Development Tools"
-  sudo pacman -Sy --noconfirm base-devel bc cpio dhclient dkms fakeroot \
+  sudo pacman -Sy --noconfirm base-devel bc cpio cryptsetup dhclient dkms \
-    fio gdb inetutils jq less linux linux-headers lsscsi nfs-utils parted \
+    fakeroot fio gdb inetutils jq less linux linux-headers lsscsi nfs-utils \
-    pax perf python-packaging python-setuptools qemu-guest-agent ksh samba \
+    parted pax perf python-packaging python-setuptools qemu-guest-agent ksh \
-    sysstat rng-tools rsync wget xxhash
+    samba sysstat rng-tools rsync wget xxhash
   echo "##[endgroup]"
 }
 
@@ -30,11 +30,11 @@ function debian() {
 
   echo "##[group]Install Development Tools"
   sudo apt-get install -y \
-    acl alien attr autoconf bc cpio curl dbench dh-python dkms fakeroot \
+    acl alien attr autoconf bc cpio cryptsetup curl dbench dh-python dkms \
-    fio gdb gdebi git ksh lcov isc-dhcp-client jq libacl1-dev libaio-dev \
+    fakeroot fio gdb gdebi git ksh lcov isc-dhcp-client jq libacl1-dev \
-    libattr1-dev libblkid-dev libcurl4-openssl-dev libdevmapper-dev libelf-dev \
+    libaio-dev libattr1-dev libblkid-dev libcurl4-openssl-dev libdevmapper-dev \
-    libffi-dev libmount-dev libpam0g-dev libselinux-dev libssl-dev libtool \
+    libelf-dev libffi-dev libmount-dev libpam0g-dev libselinux-dev libssl-dev \
-    libtool-bin libudev-dev libunwind-dev linux-headers-$(uname -r) \
+    libtool libtool-bin libudev-dev libunwind-dev linux-headers-$(uname -r) \
     lsscsi nfs-kernel-server pamtester parted python3 python3-all-dev \
     python3-cffi python3-dev python3-distlib python3-packaging \
     python3-setuptools python3-sphinx qemu-guest-agent rng-tools rpm2cpio \
@@ -66,16 +66,23 @@ function rhel() {
   echo "##[endgroup]"
 
   echo "##[group]Install Development Tools"
-  sudo dnf group install -y "Development Tools"
+
+  # Alma wants "Development Tools", Fedora 41 wants "development-tools"
+  if ! sudo dnf group install -y "Development Tools" ; then
+    echo "Trying 'development-tools' instead of 'Development Tools'"
+    sudo dnf group install -y development-tools
+  fi
+
   sudo dnf install -y \
-    acl attr bc bzip2 curl dbench dkms elfutils-libelf-devel fio gdb git \
+    acl attr bc bzip2 cryptsetup curl dbench dkms elfutils-libelf-devel fio \
-    jq kernel-rpm-macros ksh libacl-devel libaio-devel libargon2-devel \
+    gdb git jq kernel-rpm-macros ksh libacl-devel libaio-devel \
-    libattr-devel libblkid-devel libcurl-devel libffi-devel ncompress \
+    libargon2-devel libattr-devel libblkid-devel libcurl-devel libffi-devel \
-    libselinux-devel libtirpc-devel libtool libudev-devel libuuid-devel \
+    ncompress libselinux-devel libtirpc-devel libtool libudev-devel \
-    lsscsi mdadm nfs-utils openssl-devel pam-devel pamtester parted perf \
+    libuuid-devel lsscsi mdadm nfs-utils openssl-devel pam-devel pamtester \
-    python3 python3-cffi python3-devel python3-packaging kernel-devel \
+    parted perf python3 python3-cffi python3-devel python3-packaging \
-    python3-setuptools qemu-guest-agent rng-tools rpcgen rpm-build rsync \
+    kernel-devel python3-setuptools qemu-guest-agent rng-tools rpcgen \
-    samba sysstat systemd watchdog wget xfsprogs-devel xxhash zlib-devel
+    rpm-build rsync samba sysstat systemd watchdog wget xfsprogs-devel xxhash \
+    zlib-devel
   echo "##[endgroup]"
 }
 
@@ -111,6 +118,7 @@ case "$1" in
     archlinux
     ;;
   debian*)
+    echo 'debconf debconf/frontend select Noninteractive' | sudo debconf-set-selections
     debian
     echo "##[group]Install Debian specific"
     sudo apt-get install -yq linux-perf dh-sequence-dkms

.github/workflows/scripts/qemu-4-build.sh

@@ -83,7 +83,7 @@ function rpm_build_and_install() {
   echo "##[endgroup]"
 
   echo "##[group]Install"
-  run sudo dnf -y --skip-broken localinstall $(ls *.rpm | grep -v src.rpm)
+  run sudo dnf -y --nobest install $(ls *.rpm | grep -v src.rpm)
   echo "##[endgroup]"
 
 }

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

@@ -14,17 +14,21 @@ PID=$(pidof /usr/bin/qemu-system-x86_64)
 tail --pid=$PID -f /dev/null
 sudo virsh undefine openzfs
 
-# definitions of per operating system
+# default values per test vm:
+VMs=2
+CPU=2
+
+# cpu pinning
+CPUSET=("0,1" "2,3")
+
 case "$OS" in
   freebsd*)
-    VMs=2
+    # FreeBSD can't be optimized via ksmtuned
-    CPU=3
     RAM=6
     ;;
   *)
-    VMs=2
+    # Linux can be optimized via ksmtuned
-    CPU=3
+    RAM=8
-    RAM=7
     ;;
 esac
 
@@ -73,6 +77,7 @@ EOF
     --cpu host-passthrough \
     --virt-type=kvm --hvm \
     --vcpus=$CPU,sockets=1 \
+    --cpuset=${CPUSET[$((i-1))]} \
     --memory $((1024*RAM)) \
     --memballoon model=virtio \
     --graphics none \

.github/workflows/scripts/qemu-9-summary-page.sh

@@ -11,12 +11,10 @@ function output() {
 }
 
 function outfile() {
-  test -s "$1" || return
   cat "$1" >> "out-$logfile.md"
 }
 
 function outfile_plain() {
-  test -s "$1" || return
   output "<pre>"
   cat "$1" >> "out-$logfile.md"
   output "</pre>"
@@ -45,6 +43,8 @@ if [ ! -f out-1.md ]; then
     tar xf "$tarfile"
     test -s env.txt || continue
     source env.txt
+    # when uname.txt is there, the other files are also ok
+    test -s uname.txt || continue
     output "\n## Functional Tests: $OSNAME\n"
     outfile_plain uname.txt
     outfile_plain summary.txt

.github/workflows/zfs-qemu.yml

@@ -22,8 +22,8 @@ jobs:
       - name: Generate OS config and CI type
         id: os
         run: |
-          FULL_OS='["almalinux8", "almalinux9", "centos-stream9", "debian11", "debian12", "fedora39", "fedora40", "freebsd13", "freebsd13r", "freebsd14", "freebsd14r", "ubuntu20", "ubuntu22", "ubuntu24"]'
+          FULL_OS='["almalinux8", "almalinux9", "centos-stream9", "debian11", "debian12", "fedora40", "fedora41", "freebsd13-4r", "freebsd14-0r", "freebsd14-1s", "ubuntu20", "ubuntu22", "ubuntu24"]'
-          QUICK_OS='["almalinux8", "almalinux9", "debian12", "fedora40", "freebsd13", "freebsd14", "ubuntu24"]'
+          QUICK_OS='["almalinux8", "almalinux9", "debian12", "fedora41", "freebsd13-3r", "freebsd14-1r", "ubuntu24"]'
           # determine CI type when running on PR
           ci_type="full"
           if ${{ github.event_name == 'pull_request' }}; then
@@ -46,10 +46,12 @@ jobs:
     strategy:
       fail-fast: false
       matrix:
-        # all:
+        # rhl:     almalinux8, almalinux9, centos-stream9, fedora40, fedora41
-        # os:  [almalinux8, almalinux9, archlinux, centos-stream9, fedora39, fedora40, debian11, debian12, freebsd13, freebsd13r, freebsd14, freebsd14r, freebsd15, ubuntu20, ubuntu22, ubuntu24]
+        # debian:  debian11, debian12, ubuntu20, ubuntu22, ubuntu24
-        # openzfs:
+        # misc:    archlinux, tumbleweed
-        # os: [almalinux8, almalinux9, centos-stream9, debian11, debian12, fedora39, fedora40, freebsd13, freebsd13r, freebsd14, freebsd14r, ubuntu20, ubuntu22, ubuntu24]
+        # FreeBSD Release: freebsd13-3r, freebsd13-4r, freebsd14-0r, freebsd14-1r
+        # FreeBSD Stable:  freebsd13-4s, freebsd14-1s
+        # FreeBSD Current: freebsd15-0c
         os: ${{ fromJson(needs.test-config.outputs.test_os) }}
     runs-on: ubuntu-24.04
     steps:

META

@@ -2,7 +2,7 @@ Meta:          1
 Name:          zfs
 Branch:        1.0
 Version:       2.3.0
-Release:       rc1
+Release:       rc3
 Release-Tags:  relext
 License:       CDDL
 Author:        OpenZFS

cmd/zdb/zdb.c

@@ -1131,7 +1131,7 @@ dump_zap(objset_t *os, uint64_t object, void *data, size_t size)
 		    !!(zap_getflags(zc.zc_zap) & ZAP_FLAG_UINT64_KEY);
 
 		if (key64)
-			(void) printf("\t\t0x%010lx = ",
+			(void) printf("\t\t0x%010" PRIu64 "x = ",
 			    *(uint64_t *)attrp->za_name);
 		else
 			(void) printf("\t\t%s = ", attrp->za_name);
@@ -2152,14 +2152,21 @@ dump_brt(spa_t *spa)
 	if (dump_opt['T'] < 3)
 		return;
 
+	/* -TTT shows a per-vdev histograms; -TTTT shows all entries */
+	boolean_t do_histo = dump_opt['T'] == 3;
+
 	char dva[64];
-	printf("\n%-16s %-10s\n", "DVA", "REFCNT");
+
+	if (!do_histo)
+		printf("\n%-16s %-10s\n", "DVA", "REFCNT");
 
 	for (uint64_t vdevid = 0; vdevid < brt->brt_nvdevs; vdevid++) {
 		brt_vdev_t *brtvd = &brt->brt_vdevs[vdevid];
 		if (brtvd == NULL || !brtvd->bv_initiated)
 			continue;
 
+		uint64_t counts[64] = {};
+
 		zap_cursor_t zc;
 		zap_attribute_t *za = zap_attribute_alloc();
 		for (zap_cursor_init(&zc, brt->brt_mos, brtvd->bv_mos_entries);
@@ -2172,14 +2179,26 @@ dump_brt(spa_t *spa)
 			    za->za_integer_length, za->za_num_integers,
 			    &refcnt));
 
-			uint64_t offset = *(const uint64_t *)za->za_name;
+			if (do_histo)
+				counts[highbit64(refcnt)]++;
+			else {
+				uint64_t offset =
+				    *(const uint64_t *)za->za_name;
 
-			snprintf(dva, sizeof (dva), "%" PRIu64 ":%llx", vdevid,
+				snprintf(dva, sizeof (dva), "%" PRIu64 ":%llx",
-			    (u_longlong_t)offset);
+				    vdevid, (u_longlong_t)offset);
-			printf("%-16s %-10llu\n", dva, (u_longlong_t)refcnt);
+				printf("%-16s %-10llu\n", dva,
+				    (u_longlong_t)refcnt);
+			}
 		}
 		zap_cursor_fini(&zc);
 		zap_attribute_free(za);
+
+		if (do_histo) {
+			printf("\nBRT: vdev %" PRIu64
+			    ": DVAs with 2^n refcnts:\n", vdevid);
+			dump_histogram(counts, 64, 0);
+		}
 	}
 }
 
@@ -4266,6 +4285,10 @@ dump_uberblock(uberblock_t *ub, const char *header, const char *footer)
 	(void) printf("\ttimestamp = %llu UTC = %s",
 	    (u_longlong_t)ub->ub_timestamp, ctime(&timestamp));
 
+	char blkbuf[BP_SPRINTF_LEN];
+	snprintf_blkptr(blkbuf, sizeof (blkbuf), &ub->ub_rootbp);
+	(void) printf("\tbp = %s\n", blkbuf);
+
 	(void) printf("\tmmp_magic = %016llx\n",
 	    (u_longlong_t)ub->ub_mmp_magic);
 	if (MMP_VALID(ub)) {

cmd/zed/zed_disk_event.c

@@ -139,7 +139,8 @@ dev_event_nvlist(struct udev_device *dev)
 		 * is /dev/sda.
 		 */
 		struct udev_device *parent_dev = udev_device_get_parent(dev);
-		if ((value = udev_device_get_sysattr_value(parent_dev, "size"))
+		if (parent_dev != NULL &&
+		    (value = udev_device_get_sysattr_value(parent_dev, "size"))
 		    != NULL) {
 			uint64_t numval = DEV_BSIZE;
 

cmd/zfs/zfs_main.c

@@ -2162,6 +2162,7 @@ zfs_do_get(int argc, char **argv)
 	cb.cb_type = ZFS_TYPE_DATASET;
 
 	struct option long_options[] = {
+		{"json", no_argument, NULL, 'j'},
 		{"json-int", no_argument, NULL, ZFS_OPTION_JSON_NUMS_AS_INT},
 		{0, 0, 0, 0}
 	};
@@ -3760,8 +3761,13 @@ collect_dataset(zfs_handle_t *zhp, list_cbdata_t *cb)
 		if (cb->cb_json) {
 			if (pl->pl_prop == ZFS_PROP_NAME)
 				continue;
+			const char *prop_name;
+			if (pl->pl_prop != ZPROP_USERPROP)
+				prop_name = zfs_prop_to_name(pl->pl_prop);
+			else
+				prop_name = pl->pl_user_prop;
 			if (zprop_nvlist_one_property(
-			    zfs_prop_to_name(pl->pl_prop), propstr,
+			    prop_name, propstr,
 			    sourcetype, source, NULL, props,
 			    cb->cb_json_as_int) != 0)
 				nomem();
@@ -3852,6 +3858,7 @@ zfs_do_list(int argc, char **argv)
 	nvlist_t *data = NULL;
 
 	struct option long_options[] = {
+		{"json", no_argument, NULL, 'j'},
 		{"json-int", no_argument, NULL, ZFS_OPTION_JSON_NUMS_AS_INT},
 		{0, 0, 0, 0}
 	};
@@ -7436,9 +7443,15 @@ share_mount(int op, int argc, char **argv)
 	uint_t nthr;
 	jsobj = data = item = NULL;
 
+	struct option long_options[] = {
+		{"json", no_argument, NULL, 'j'},
+		{0, 0, 0, 0}
+	};
+
 	/* check options */
-	while ((c = getopt(argc, argv, op == OP_MOUNT ? ":ajRlvo:Of" : "al"))
+	while ((c = getopt_long(argc, argv,
-	    != -1) {
+	    op == OP_MOUNT ? ":ajRlvo:Of" : "al",
+	    op == OP_MOUNT ? long_options : NULL, NULL)) != -1) {
 		switch (c) {
 		case 'a':
 			do_all = 1;
@@ -8374,8 +8387,14 @@ zfs_do_channel_program(int argc, char **argv)
 	boolean_t sync_flag = B_TRUE, json_output = B_FALSE;
 	zpool_handle_t *zhp;
 
+	struct option long_options[] = {
+		{"json", no_argument, NULL, 'j'},
+		{0, 0, 0, 0}
+	};
+
 	/* check options */
-	while ((c = getopt(argc, argv, "nt:m:j")) != -1) {
+	while ((c = getopt_long(argc, argv, "nt:m:j", long_options,
+	    NULL)) != -1) {
 		switch (c) {
 		case 't':
 		case 'm': {
@@ -9083,7 +9102,13 @@ zfs_do_version(int argc, char **argv)
 	int c;
 	nvlist_t *jsobj = NULL, *zfs_ver = NULL;
 	boolean_t json = B_FALSE;
-	while ((c = getopt(argc, argv, "j")) != -1) {
+
+	struct option long_options[] = {
+		{"json", no_argument, NULL, 'j'},
+		{0, 0, 0, 0}
+	};
+
+	while ((c = getopt_long(argc, argv, "j", long_options, NULL)) != -1) {
 		switch (c) {
 		case 'j':
 			json = B_TRUE;
@@ -9187,7 +9212,7 @@ main(int argc, char **argv)
 	 * Special case '-V|--version'
 	 */
 	if ((strcmp(cmdname, "-V") == 0) || (strcmp(cmdname, "--version") == 0))
-		return (zfs_do_version(argc, argv));
+		return (zfs_version_print() != 0);
 
 	/*
 	 * Special case 'help'

cmd/zpool/zpool_main.c

@@ -7340,6 +7340,7 @@ zpool_do_list(int argc, char **argv)
 	current_prop_type = ZFS_TYPE_POOL;
 
 	struct option long_options[] = {
+		{"json", no_argument, NULL, 'j'},
 		{"json-int", no_argument, NULL, ZPOOL_OPTION_JSON_NUMS_AS_INT},
 		{"json-pool-key-guid", no_argument, NULL,
 		    ZPOOL_OPTION_POOL_KEY_GUID},
@@ -7965,8 +7966,11 @@ zpool_do_online(int argc, char **argv)
 
 	poolname = argv[0];
 
-	if ((zhp = zpool_open(g_zfs, poolname)) == NULL)
+	if ((zhp = zpool_open(g_zfs, poolname)) == NULL) {
+		(void) fprintf(stderr, gettext("failed to open pool "
+		    "\"%s\""), poolname);
 		return (1);
+	}
 
 	for (i = 1; i < argc; i++) {
 		vdev_state_t oldstate;
@@ -7987,12 +7991,15 @@ zpool_do_online(int argc, char **argv)
 		    &l2cache, NULL);
 		if (tgt == NULL) {
 			ret = 1;
+			(void) fprintf(stderr, gettext("couldn't find device "
+			"\"%s\" in pool \"%s\"\n"), argv[i], poolname);
 			continue;
 		}
 		uint_t vsc;
 		oldstate = ((vdev_stat_t *)fnvlist_lookup_uint64_array(tgt,
 		    ZPOOL_CONFIG_VDEV_STATS, &vsc))->vs_state;
-		if (zpool_vdev_online(zhp, argv[i], flags, &newstate) == 0) {
+		if ((rc = zpool_vdev_online(zhp, argv[i], flags,
+		    &newstate)) == 0) {
 			if (newstate != VDEV_STATE_HEALTHY) {
 				(void) printf(gettext("warning: device '%s' "
 				    "onlined, but remains in faulted state\n"),
@@ -8018,6 +8025,9 @@ zpool_do_online(int argc, char **argv)
 				}
 			}
 		} else {
+			(void) fprintf(stderr, gettext("Failed to online "
+			    "\"%s\" in pool \"%s\": %d\n"),
+			    argv[i], poolname, rc);
 			ret = 1;
 		}
 	}
@@ -8102,8 +8112,11 @@ zpool_do_offline(int argc, char **argv)
 
 	poolname = argv[0];
 
-	if ((zhp = zpool_open(g_zfs, poolname)) == NULL)
+	if ((zhp = zpool_open(g_zfs, poolname)) == NULL) {
+		(void) fprintf(stderr, gettext("failed to open pool "
+		    "\"%s\""), poolname);
 		return (1);
+	}
 
 	for (i = 1; i < argc; i++) {
 		uint64_t guid = zpool_vdev_path_to_guid(zhp, argv[i]);
@@ -9224,6 +9237,12 @@ vdev_stats_nvlist(zpool_handle_t *zhp, status_cbdata_t *cb, nvlist_t *nv,
 		}
 	}
 
+	if (cb->cb_print_dio_verify) {
+		nice_num_str_nvlist(vds, "dio_verify_errors",
+		    vs->vs_dio_verify_errors, cb->cb_literal,
+		    cb->cb_json_as_int, ZFS_NICENUM_1024);
+	}
+
 	if (nvlist_lookup_uint64(nv, ZPOOL_CONFIG_NOT_PRESENT,
 	    &notpresent) == 0) {
 		nice_num_str_nvlist(vds, ZPOOL_CONFIG_NOT_PRESENT,
@@ -10975,6 +10994,7 @@ zpool_do_status(int argc, char **argv)
 
 	struct option long_options[] = {
 		{"power", no_argument, NULL, ZPOOL_OPTION_POWER},
+		{"json", no_argument, NULL, 'j'},
 		{"json-int", no_argument, NULL, ZPOOL_OPTION_JSON_NUMS_AS_INT},
 		{"json-flat-vdevs", no_argument, NULL,
 		    ZPOOL_OPTION_JSON_FLAT_VDEVS},
@@ -12583,6 +12603,7 @@ zpool_do_get(int argc, char **argv)
 	current_prop_type = cb.cb_type;
 
 	struct option long_options[] = {
+		{"json", no_argument, NULL, 'j'},
 		{"json-int", no_argument, NULL, ZPOOL_OPTION_JSON_NUMS_AS_INT},
 		{"json-pool-key-guid", no_argument, NULL,
 		    ZPOOL_OPTION_POOL_KEY_GUID},
@@ -13497,7 +13518,12 @@ zpool_do_version(int argc, char **argv)
 	int c;
 	nvlist_t *jsobj = NULL, *zfs_ver = NULL;
 	boolean_t json = B_FALSE;
-	while ((c = getopt(argc, argv, "j")) != -1) {
+
+	struct option long_options[] = {
+		{"json", no_argument, NULL, 'j'},
+	};
+
+	while ((c = getopt_long(argc, argv, "j", long_options, NULL)) != -1) {
 		switch (c) {
 		case 'j':
 			json = B_TRUE;
@@ -13613,7 +13639,7 @@ main(int argc, char **argv)
 	 * Special case '-V|--version'
 	 */
 	if ((strcmp(cmdname, "-V") == 0) || (strcmp(cmdname, "--version") == 0))
-		return (zpool_do_version(argc, argv));
+		return (zfs_version_print() != 0);
 
 	/*
 	 * Special case 'help'

cmd/ztest.c

@@ -6717,6 +6717,17 @@ out:
  *
  * Only after a full scrub has been completed is it safe to start injecting
  * data corruption.  See the comment in zfs_fault_inject().
+ *
+ * EBUSY may be returned for the following six cases.  It's the callers
+ * responsibility to handle them accordingly.
+ *
+ * Current state                Requested
+ * 1. Normal Scrub Running      Normal Scrub or Error Scrub
+ * 2. Normal Scrub Paused       Error Scrub
+ * 3. Normal Scrub Paused       Pause Normal Scrub
+ * 4. Error Scrub Running       Normal Scrub or Error Scrub
+ * 5. Error Scrub Paused        Pause Error Scrub
+ * 6. Resilvering               Anything else
  */
 static int
 ztest_scrub_impl(spa_t *spa)
@@ -8082,8 +8093,14 @@ ztest_raidz_expand_check(spa_t *spa)
 		(void) printf("verifying an interrupted raidz "
 		    "expansion using a pool scrub ...\n");
 	}
+
 	/* Will fail here if there is non-recoverable corruption detected */
-	VERIFY0(ztest_scrub_impl(spa));
+	int error = ztest_scrub_impl(spa);
+	if (error == EBUSY)
+		error = 0;
+
+	VERIFY0(error);
+
 	if (ztest_opts.zo_verbose >= 1) {
 		(void) printf("raidz expansion scrub check complete\n");
 	}

config/deb.am

@@ -58,9 +58,9 @@ deb-utils: deb-local rpm-utils-initramfs
 	pkg1=$${name}-$${version}.$${arch}.rpm; \
 	pkg2=libnvpair3-$${version}.$${arch}.rpm; \
 	pkg3=libuutil3-$${version}.$${arch}.rpm; \
-	pkg4=libzfs5-$${version}.$${arch}.rpm; \
+	pkg4=libzfs6-$${version}.$${arch}.rpm; \
-	pkg5=libzpool5-$${version}.$${arch}.rpm; \
+	pkg5=libzpool6-$${version}.$${arch}.rpm; \
-	pkg6=libzfs5-devel-$${version}.$${arch}.rpm; \
+	pkg6=libzfs6-devel-$${version}.$${arch}.rpm; \
 	pkg7=$${name}-test-$${version}.$${arch}.rpm; \
 	pkg8=$${name}-dracut-$${version}.noarch.rpm; \
 	pkg9=$${name}-initramfs-$${version}.$${arch}.rpm; \
@@ -72,7 +72,7 @@ deb-utils: deb-local rpm-utils-initramfs
 	path_prepend=`mktemp -d /tmp/intercept.XXXXXX`; \
 	echo "#!$(SHELL)" > $${path_prepend}/dh_shlibdeps; \
 	echo "`which dh_shlibdeps` -- \
-	 -xlibuutil3linux -xlibnvpair3linux -xlibzfs5linux -xlibzpool5linux" \
+	 -xlibuutil3linux -xlibnvpair3linux -xlibzfs6linux -xlibzpool6linux" \
 	 >> $${path_prepend}/dh_shlibdeps; \
 ## These -x arguments are passed to dpkg-shlibdeps, which exclude the
 ## Debianized packages from the auto-generated dependencies of the new debs,
@@ -93,13 +93,17 @@ debian:
 	cp -r contrib/debian debian; chmod +x debian/rules;
 
 native-deb-utils: native-deb-local debian
+	while [ -f debian/deb-build.lock ]; do sleep 1; done; \
+	echo "native-deb-utils" > debian/deb-build.lock; \
 	cp contrib/debian/control debian/control; \
-	$(DPKGBUILD) -b -rfakeroot -us -uc;
+	$(DPKGBUILD) -b -rfakeroot -us -uc; \
+	$(RM) -f debian/deb-build.lock
 
 native-deb-kmod: native-deb-local debian
+	while [ -f debian/deb-build.lock ]; do sleep 1; done; \
+	echo "native-deb-kmod" > debian/deb-build.lock; \
 	sh scripts/make_gitrev.sh; \
-	fakeroot debian/rules override_dh_binary-modules;
+	fakeroot debian/rules override_dh_binary-modules; \
+	$(RM) -f debian/deb-build.lock
 
 native-deb: native-deb-utils native-deb-kmod
-
-.NOTPARALLEL: native-deb native-deb-utils native-deb-kmod

config/kernel-kthread.m4

@@ -17,14 +17,21 @@ AC_DEFUN([ZFS_AC_KERNEL_KTHREAD_COMPLETE_AND_EXIT], [
 
 AC_DEFUN([ZFS_AC_KERNEL_KTHREAD_DEQUEUE_SIGNAL], [
 	dnl #
-	dnl # 5.17 API: enum pid_type * as new 4th dequeue_signal() argument,
+	dnl # prehistory:
-	dnl # 5768d8906bc23d512b1a736c1e198aa833a6daa4 ("signal: Requeue signals in the appropriate queue")
+	dnl #     int dequeue_signal(struct task_struct *task, sigset_t *mask,
+	dnl #         siginfo_t *info)
 	dnl #
-	dnl # int dequeue_signal(struct task_struct *task, sigset_t *mask, kernel_siginfo_t *info);
+	dnl # 4.20: kernel_siginfo_t introduced, replaces siginfo_t
-	dnl # int dequeue_signal(struct task_struct *task, sigset_t *mask, kernel_siginfo_t *info, enum pid_type *type);
+	dnl #     int dequeue_signal(struct task_struct *task, sigset_t *mask,
+	dnl           kernel_siginfo_t *info)
 	dnl #
-	dnl # 6.12 API: first arg struct_task* removed
+	dnl # 5.17: enum pid_type introduced as 4th arg
-	dnl # int dequeue_signal(sigset_t *mask, kernel_siginfo_t *info, enum pid_type *type);
+	dnl #     int dequeue_signal(struct task_struct *task, sigset_t *mask,
+	dnl #         kernel_siginfo_t *info, enum pid_type *type)
+	dnl #
+	dnl # 6.12: first arg struct_task* removed
+	dnl #     int dequeue_signal(sigset_t *mask, kernel_siginfo_t *info,
+	dnl #         enum pid_type *type)
 	dnl #
 	AC_MSG_CHECKING([whether dequeue_signal() takes 4 arguments])
 	ZFS_LINUX_TEST_RESULT([kthread_dequeue_signal_4arg], [
@@ -33,11 +40,11 @@ AC_DEFUN([ZFS_AC_KERNEL_KTHREAD_DEQUEUE_SIGNAL], [
 		    [dequeue_signal() takes 4 arguments])
 	], [
 		AC_MSG_RESULT(no)
-		AC_MSG_CHECKING([whether dequeue_signal() a task argument])
+		AC_MSG_CHECKING([whether 3-arg dequeue_signal() takes a type argument])
-		ZFS_LINUX_TEST_RESULT([kthread_dequeue_signal_3arg_task], [
+		ZFS_LINUX_TEST_RESULT([kthread_dequeue_signal_3arg_type], [
 			AC_MSG_RESULT(yes)
-			AC_DEFINE(HAVE_DEQUEUE_SIGNAL_3ARG_TASK, 1,
+			AC_DEFINE(HAVE_DEQUEUE_SIGNAL_3ARG_TYPE, 1,
-			    [dequeue_signal() takes a task argument])
+			    [3-arg dequeue_signal() takes a type argument])
 		], [
 			AC_MSG_RESULT(no)
 		])
@@ -56,17 +63,6 @@ AC_DEFUN([ZFS_AC_KERNEL_SRC_KTHREAD_COMPLETE_AND_EXIT], [
 ])
 
 AC_DEFUN([ZFS_AC_KERNEL_SRC_KTHREAD_DEQUEUE_SIGNAL], [
-	ZFS_LINUX_TEST_SRC([kthread_dequeue_signal_3arg_task], [
-		#include <linux/sched/signal.h>
-	], [
-		struct task_struct *task = NULL;
-		sigset_t *mask = NULL;
-		kernel_siginfo_t *info = NULL;
-		int error __attribute__ ((unused));
-
-		error = dequeue_signal(task, mask, info);
-	])
-
 	ZFS_LINUX_TEST_SRC([kthread_dequeue_signal_4arg], [
 		#include <linux/sched/signal.h>
 	], [
@@ -78,6 +74,17 @@ AC_DEFUN([ZFS_AC_KERNEL_SRC_KTHREAD_DEQUEUE_SIGNAL], [
 
 		error = dequeue_signal(task, mask, info, type);
 	])
+
+	ZFS_LINUX_TEST_SRC([kthread_dequeue_signal_3arg_type], [
+		#include <linux/sched/signal.h>
+	], [
+		sigset_t *mask = NULL;
+		kernel_siginfo_t *info = NULL;
+		enum pid_type *type = NULL;
+		int error __attribute__ ((unused));
+
+		error = dequeue_signal(mask, info, type);
+	])
 ])
 
 AC_DEFUN([ZFS_AC_KERNEL_KTHREAD], [

config/kernel-vfs-invalidate_folio.m4

@@ -1,33 +0,0 @@
-dnl #
-dnl # Linux 5.18 uses invalidate_folio in lieu of invalidate_page
-dnl #
-AC_DEFUN([ZFS_AC_KERNEL_SRC_VFS_INVALIDATE_FOLIO], [
-	ZFS_LINUX_TEST_SRC([vfs_has_invalidate_folio], [
-		#include <linux/fs.h>
-
-		static void
-		test_invalidate_folio(struct folio *folio, size_t offset,
-		                      size_t len) {
-			(void) folio; (void) offset; (void) len;
-			return;
-		}
-
-		static const struct address_space_operations
-		    aops __attribute__ ((unused)) = {
-			.invalidate_folio	= test_invalidate_folio,
-		};
-	],[])
-])
-
-AC_DEFUN([ZFS_AC_KERNEL_VFS_INVALIDATE_FOLIO], [
-	dnl #
-	dnl # Linux 5.18 uses invalidate_folio in lieu of invalidate_page
-	dnl #
-	AC_MSG_CHECKING([whether invalidate_folio exists])
-	ZFS_LINUX_TEST_RESULT([vfs_has_invalidate_folio], [
-		AC_MSG_RESULT([yes])
-		AC_DEFINE(HAVE_VFS_INVALIDATE_FOLIO, 1, [invalidate_folio exists])
-	],[
-		AC_MSG_RESULT([no])
-	])
-])

config/kernel-vfs-migrate_folio.m4

@@ -0,0 +1,27 @@
+dnl #
+dnl # Linux 6.0 uses migrate_folio in lieu of migrate_page
+dnl #
+AC_DEFUN([ZFS_AC_KERNEL_SRC_VFS_MIGRATE_FOLIO], [
+	ZFS_LINUX_TEST_SRC([vfs_has_migrate_folio], [
+		#include <linux/fs.h>
+		#include <linux/migrate.h>
+
+		static const struct address_space_operations
+		    aops __attribute__ ((unused)) = {
+			.migrate_folio	= migrate_folio,
+		};
+	],[])
+])
+
+AC_DEFUN([ZFS_AC_KERNEL_VFS_MIGRATE_FOLIO], [
+	dnl #
+	dnl # Linux 6.0 uses migrate_folio in lieu of migrate_page
+	dnl #
+	AC_MSG_CHECKING([whether migrate_folio exists])
+	ZFS_LINUX_TEST_RESULT([vfs_has_migrate_folio], [
+		AC_MSG_RESULT([yes])
+		AC_DEFINE(HAVE_VFS_MIGRATE_FOLIO, 1, [migrate_folio exists])
+	],[
+		AC_MSG_RESULT([no])
+	])
+])

config/kernel-vfs-release_folio.m4

@@ -1,32 +0,0 @@
-dnl #
-dnl # Linux 5.19 uses release_folio in lieu of releasepage
-dnl #
-AC_DEFUN([ZFS_AC_KERNEL_SRC_VFS_RELEASE_FOLIO], [
-	ZFS_LINUX_TEST_SRC([vfs_has_release_folio], [
-		#include <linux/fs.h>
-
-		static bool
-		test_release_folio(struct folio *folio, gfp_t gfp) {
-			(void) folio; (void) gfp;
-			return (0);
-		}
-
-		static const struct address_space_operations
-		    aops __attribute__ ((unused)) = {
-			.release_folio	= test_release_folio,
-		};
-	],[])
-])
-
-AC_DEFUN([ZFS_AC_KERNEL_VFS_RELEASE_FOLIO], [
-	dnl #
-	dnl # Linux 5.19 uses release_folio in lieu of releasepage
-	dnl #
-	AC_MSG_CHECKING([whether release_folio exists])
-	ZFS_LINUX_TEST_RESULT([vfs_has_release_folio], [
-		AC_MSG_RESULT([yes])
-		AC_DEFINE(HAVE_VFS_RELEASE_FOLIO, 1, [release_folio exists])
-	],[
-		AC_MSG_RESULT([no])
-	])
-])

config/kernel.m4

@@ -77,8 +77,7 @@ AC_DEFUN([ZFS_AC_KERNEL_TEST_SRC], [
 	ZFS_AC_KERNEL_SRC_SGET
 	ZFS_AC_KERNEL_SRC_VFS_FILEMAP_DIRTY_FOLIO
 	ZFS_AC_KERNEL_SRC_VFS_READ_FOLIO
-	ZFS_AC_KERNEL_SRC_VFS_RELEASE_FOLIO
+	ZFS_AC_KERNEL_SRC_VFS_MIGRATE_FOLIO
-	ZFS_AC_KERNEL_SRC_VFS_INVALIDATE_FOLIO
 	ZFS_AC_KERNEL_SRC_VFS_FSYNC_2ARGS
 	ZFS_AC_KERNEL_SRC_VFS_DIRECT_IO
 	ZFS_AC_KERNEL_SRC_VFS_READPAGES
@@ -189,8 +188,7 @@ AC_DEFUN([ZFS_AC_KERNEL_TEST_RESULT], [
 	ZFS_AC_KERNEL_SGET
 	ZFS_AC_KERNEL_VFS_FILEMAP_DIRTY_FOLIO
 	ZFS_AC_KERNEL_VFS_READ_FOLIO
-	ZFS_AC_KERNEL_VFS_RELEASE_FOLIO
+	ZFS_AC_KERNEL_VFS_MIGRATE_FOLIO
-	ZFS_AC_KERNEL_VFS_INVALIDATE_FOLIO
 	ZFS_AC_KERNEL_VFS_FSYNC_2ARGS
 	ZFS_AC_KERNEL_VFS_DIRECT_IO
 	ZFS_AC_KERNEL_VFS_READPAGES

config/user.m4

@@ -33,7 +33,7 @@ AC_DEFUN([ZFS_AC_CONFIG_USER], [
 	ZFS_AC_CONFIG_USER_MAKEDEV_IN_MKDEV
 	ZFS_AC_CONFIG_USER_ZFSEXEC
 
-	AC_CHECK_FUNCS([execvpe issetugid mlockall strlcat strlcpy gettid])
+	AC_CHECK_FUNCS([execvpe issetugid mlockall strerror_l strlcat strlcpy gettid])
 
 	AC_SUBST(RM)
 ])

contrib/debian/Makefile.am

@@ -12,14 +12,14 @@ dist_noinst_DATA += %D%/openzfs-libpam-zfs.postinst
 dist_noinst_DATA += %D%/openzfs-libpam-zfs.prerm
 dist_noinst_DATA += %D%/openzfs-libuutil3.docs
 dist_noinst_DATA += %D%/openzfs-libuutil3.install.in
-dist_noinst_DATA += %D%/openzfs-libzfs4.docs
+dist_noinst_DATA += %D%/openzfs-libzfs6.docs
-dist_noinst_DATA += %D%/openzfs-libzfs4.install.in
+dist_noinst_DATA += %D%/openzfs-libzfs6.install.in
 dist_noinst_DATA += %D%/openzfs-libzfsbootenv1.docs
 dist_noinst_DATA += %D%/openzfs-libzfsbootenv1.install.in
 dist_noinst_DATA += %D%/openzfs-libzfs-dev.docs
 dist_noinst_DATA += %D%/openzfs-libzfs-dev.install.in
-dist_noinst_DATA += %D%/openzfs-libzpool5.docs
+dist_noinst_DATA += %D%/openzfs-libzpool6.docs
-dist_noinst_DATA += %D%/openzfs-libzpool5.install.in
+dist_noinst_DATA += %D%/openzfs-libzpool6.install.in
 dist_noinst_DATA += %D%/openzfs-python3-pyzfs.install
 dist_noinst_DATA += %D%/openzfs-zfs-dkms.config
 dist_noinst_DATA += %D%/openzfs-zfs-dkms.dkms

contrib/debian/clean

@@ -6,6 +6,6 @@ contrib/pyzfs/libzfs_core/bindings/__pycache__/
 contrib/pyzfs/pyzfs.egg-info/
 debian/openzfs-libnvpair3.install
 debian/openzfs-libuutil3.install
-debian/openzfs-libzfs4.install
+debian/openzfs-libzfs6.install
 debian/openzfs-libzfs-dev.install
-debian/openzfs-libzpool5.install
+debian/openzfs-libzpool6.install

contrib/debian/control

@@ -78,9 +78,9 @@ Architecture: linux-any
 Depends: libssl-dev | libssl1.0-dev,
          openzfs-libnvpair3 (= ${binary:Version}),
          openzfs-libuutil3 (= ${binary:Version}),
-         openzfs-libzfs4 (= ${binary:Version}),
+         openzfs-libzfs6 (= ${binary:Version}),
          openzfs-libzfsbootenv1 (= ${binary:Version}),
-         openzfs-libzpool5 (= ${binary:Version}),
+         openzfs-libzpool6 (= ${binary:Version}),
          ${misc:Depends}
 Replaces: libzfslinux-dev
 Conflicts: libzfslinux-dev
@@ -90,18 +90,18 @@ Description: OpenZFS filesystem development files for Linux
  libraries of OpenZFS filesystem.
  .
  This package includes the development files of libnvpair3, libuutil3,
- libzpool5 and libzfs4.
+ libzpool6 and libzfs6.
 
-Package: openzfs-libzfs4
+Package: openzfs-libzfs6
 Section: contrib/libs
 Architecture: linux-any
 Depends: ${misc:Depends}, ${shlibs:Depends}
 # The libcurl4 is loaded through dlopen("libcurl.so.4").
 # https://bugs.debian.org/cgi-bin/bugreport.cgi?bug=988521
 Recommends: libcurl4
-Breaks: libzfs2, libzfs4
+Breaks: libzfs2, libzfs4, libzfs4linux, libzfs6linux
-Replaces: libzfs2, libzfs4, libzfs4linux
+Replaces: libzfs2, libzfs4, libzfs4linux, libzfs6linux
-Conflicts: libzfs4linux
+Conflicts: libzfs6linux
 Description: OpenZFS filesystem library for Linux - general support
  OpenZFS is a storage platform that encompasses the functionality of
  traditional filesystems and volume managers. It supports data checksums,
@@ -123,13 +123,13 @@ Description: OpenZFS filesystem library for Linux - label info support
  .
  The zfsbootenv library provides support for modifying ZFS label information.
 
-Package: openzfs-libzpool5
+Package: openzfs-libzpool6
 Section: contrib/libs
 Architecture: linux-any
 Depends: ${misc:Depends}, ${shlibs:Depends}
-Breaks: libzpool2, libzpool5
+Breaks: libzpool2, libzpool5, libzpool5linux, libzpool6linux
-Replaces: libzpool2, libzpool5, libzpool5linux
+Replaces: libzpool2, libzpool5, libzpool5linux, libzpool6linux
-Conflicts: libzpool5linux
+Conflicts: libzpool6linux
 Description: OpenZFS pool library for Linux
  OpenZFS is a storage platform that encompasses the functionality of
  traditional filesystems and volume managers. It supports data checksums,
@@ -246,8 +246,8 @@ Architecture: linux-any
 Pre-Depends: ${misc:Pre-Depends}
 Depends: openzfs-libnvpair3 (= ${binary:Version}),
          openzfs-libuutil3 (= ${binary:Version}),
-         openzfs-libzfs4 (= ${binary:Version}),
+         openzfs-libzfs6 (= ${binary:Version}),
-         openzfs-libzpool5 (= ${binary:Version}),
+         openzfs-libzpool6 (= ${binary:Version}),
          python3,
          ${misc:Depends},
          ${shlibs:Depends}

contrib/debian/openzfs-zfsutils.install

@@ -98,6 +98,7 @@ usr/share/man/man8/zpool-attach.8
 usr/share/man/man8/zpool-checkpoint.8
 usr/share/man/man8/zpool-clear.8
 usr/share/man/man8/zpool-create.8
+usr/share/man/man8/zpool-ddtprune.8
 usr/share/man/man8/zpool-destroy.8
 usr/share/man/man8/zpool-detach.8
 usr/share/man/man8/zpool-ddtprune.8
@@ -113,6 +114,7 @@ usr/share/man/man8/zpool-list.8
 usr/share/man/man8/zpool-offline.8
 usr/share/man/man8/zpool-online.8
 usr/share/man/man8/zpool-prefetch.8
+usr/share/man/man8/zpool-prefetch.8
 usr/share/man/man8/zpool-reguid.8
 usr/share/man/man8/zpool-remove.8
 usr/share/man/man8/zpool-reopen.8

contrib/initramfs/scripts/zfs

@@ -344,7 +344,7 @@ mount_fs()
 
 	# Need the _original_ datasets mountpoint!
 	mountpoint=$(get_fs_value "$fs" mountpoint)
-	ZFS_CMD="mount -o zfsutil -t zfs"
+	ZFS_CMD="mount.zfs -o zfsutil"
 	if [ "$mountpoint" = "legacy" ] || [ "$mountpoint" = "none" ]; then
 		# Can't use the mountpoint property. Might be one of our
 		# clones. Check the 'org.zol:mountpoint' property set in
@@ -359,9 +359,8 @@ mount_fs()
 				# isn't the root fs.
 				return 0
 			fi
-			# Don't use mount.zfs -o zfsutils for legacy mountpoint
 			if [ "$mountpoint" = "legacy" ]; then
-				ZFS_CMD="mount -t zfs"
+				ZFS_CMD="mount.zfs"
 			fi
 			# Last hail-mary: Hope 'rootmnt' is set!
 			mountpoint=""

include/libzutil.h

@@ -276,7 +276,11 @@ _LIBZUTIL_H void update_vdev_config_dev_sysfs_path(nvlist_t *nv,
  * Thread-safe strerror() for use in ZFS libraries
  */
 static inline char *zfs_strerror(int errnum) {
+#ifdef HAVE_STRERROR_L
 	return (strerror_l(errnum, uselocale(0)));
+#else
+	return (strerror(errnum));
+#endif
 }
 
 #ifdef	__cplusplus

include/os/freebsd/linux/compiler.h

@@ -1,10 +1,5 @@
 /*
- * Copyright (c) 2010 Isilon Systems, Inc.
+ * Copyright (c) 2024 Warner Losh.
- * Copyright (c) 2010 iXsystems, Inc.
- * Copyright (c) 2010 Panasas, Inc.
- * Copyright (c) 2013-2016 Mellanox Technologies, Ltd.
- * Copyright (c) 2015 François Tigeot
- * All rights reserved.
  *
  * Redistribution and use in source and binary forms, with or without
  * modification, are permitted provided that the following conditions
@@ -26,76 +21,14 @@
  * THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
  * (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF
  * THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
- *
- * $FreeBSD$
  */
-#ifndef	_LINUX_COMPILER_H_
-#define	_LINUX_COMPILER_H_
 
-#include <sys/cdefs.h>
+/*
+ * FreeBSD's LinuxKPI compiler.h as far back as FreeBSD 12 has what we need,
+ * except zfs_fallthrough.
+ */
+#pragma once
+
+#include <compat/linuxkpi/common/include/linux/compiler.h>
 
-#define	__user
-#define	__kernel
-#define	__safe
-#define	__force
-#define	__nocast
-#define	__iomem
-#define	__chk_user_ptr(x)		((void)0)
-#define	__chk_io_ptr(x)			((void)0)
-#define	__builtin_warning(x, y...)	(1)
-#define	__acquires(x)
-#define	__releases(x)
-#define	__acquire(x)			do { } while (0)
-#define	__release(x)			do { } while (0)
-#define	__cond_lock(x, c)		(c)
-#define	__bitwise
-#define	__devinitdata
-#define	__deprecated
-#define	__init
-#define	__initconst
-#define	__devinit
-#define	__devexit
-#define	__exit
-#define	__rcu
-#define	__percpu
-#define	__weak __weak_symbol
-#define	__malloc
-#define	___stringify(...)		#__VA_ARGS__
-#define	__stringify(...)		___stringify(__VA_ARGS__)
-#define	__attribute_const__		__attribute__((__const__))
-#undef __always_inline
-#define	__always_inline			inline
-#define	noinline			__noinline
-#define	____cacheline_aligned		__aligned(CACHE_LINE_SIZE)
 #define	zfs_fallthrough			__attribute__((__fallthrough__))
-
-#if !defined(_KERNEL) && !defined(_STANDALONE)
-#define	likely(x)			__builtin_expect(!!(x), 1)
-#define	unlikely(x)			__builtin_expect(!!(x), 0)
-#endif
-#define	typeof(x)			__typeof(x)
-
-#define	uninitialized_var(x)		x = x
-#define	__maybe_unused			__unused
-#define	__always_unused			__unused
-#define	__must_check			__result_use_check
-
-#define	__printf(a, b)			__printflike(a, b)
-
-#define	barrier()			__asm__ __volatile__("": : :"memory")
-#define	___PASTE(a, b) a##b
-#define	__PASTE(a, b) ___PASTE(a, b)
-
-#define	ACCESS_ONCE(x)			(*(volatile __typeof(x) *)&(x))
-
-#define	WRITE_ONCE(x, v) do {		\
-	barrier();			\
-	ACCESS_ONCE(x) = (v);		\
-	barrier();			\
-} while (0)
-
-#define	lockless_dereference(p) READ_ONCE(p)
-
-#define	_AT(T, X)	((T)(X))
-
-#endif	/* _LINUX_COMPILER_H_ */

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

@@ -70,15 +70,6 @@ hlist_del(struct hlist_node *n)
 		n->next->pprev = n->pprev;
 }
 	/* BEGIN CSTYLED */
-#define	READ_ONCE(x) ({			\
-	__typeof(x) __var = ({		\
-		barrier();		\
-		ACCESS_ONCE(x);		\
-	});				\
-	barrier();			\
-	__var;				\
-})
-
 #define	HLIST_HEAD_INIT { }
 #define	HLIST_HEAD(name) struct hlist_head name = HLIST_HEAD_INIT
 #define	INIT_HLIST_HEAD(head) (head)->first = NULL

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

@@ -95,10 +95,6 @@ spl_assert(const char *buf, const char *file, const char *func, int line)
 #ifndef expect
 #define	expect(expr, value) (__builtin_expect((expr), (value)))
 #endif
-#ifndef __linux__
-#define	likely(expr)   expect((expr) != 0, 1)
-#define	unlikely(expr) expect((expr) != 0, 0)
-#endif
 
 #define	PANIC(fmt, a...)						\
 	spl_panic(__FILE__, __FUNCTION__, __LINE__, fmt, ## a)

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

@@ -50,7 +50,7 @@
 #define	kfpu_fini()		do {} while (0)
 #endif
 
-#define	simd_stat_init()	0
+#define	simd_stat_init()	do {} while (0)
-#define	simd_stat_fini()	0
+#define	simd_stat_fini()	do {} while (0)
 
 #endif

include/os/freebsd/zfs/sys/abd_os.h

@@ -26,8 +26,10 @@
 #ifndef _ABD_OS_H
 #define	_ABD_OS_H
 
+#ifdef _KERNEL
 #include <sys/vm.h>
 #include <vm/vm_page.h>
+#endif
 
 #ifdef __cplusplus
 extern "C" {
@@ -47,8 +49,10 @@ struct abd_linear {
 #endif
 };
 
+#ifdef _KERNEL
 __attribute__((malloc))
 struct abd *abd_alloc_from_pages(vm_page_t *, unsigned long, uint64_t);
+#endif
 
 #ifdef __cplusplus
 }

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

@@ -38,8 +38,7 @@
 #include <sys/rwlock.h>
 #include <sys/wait.h>
 #include <sys/wmsum.h>
-
+#include <sys/kstat.h>
-typedef struct kstat_s kstat_t;
 
 #define	TASKQ_NAMELEN		31
 

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

@@ -42,7 +42,7 @@
 #define	TS_ZOMB				EXIT_ZOMBIE
 #define	TS_STOPPED			TASK_STOPPED
 
-typedef void (*thread_func_t)(void *);
+typedef void (*thread_func_t)(void *) __attribute__((noreturn));
 
 #define	thread_create_named(name, stk, stksize, func, arg, len,	\
     pp, state, pri)	\

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

@@ -30,6 +30,8 @@
 extern "C" {
 #endif
 
+struct abd;
+
 struct abd_scatter {
 	uint_t		abd_offset;
 	uint_t		abd_nents;
@@ -41,10 +43,8 @@ struct abd_linear {
 	struct scatterlist *abd_sgl; /* for LINEAR_PAGE */
 };
 
-typedef struct abd abd_t;
-
 typedef int abd_iter_page_func_t(struct page *, size_t, size_t, void *);
-int abd_iterate_page_func(abd_t *, size_t, size_t, abd_iter_page_func_t *,
+int abd_iterate_page_func(struct abd *, size_t, size_t, abd_iter_page_func_t *,
     void *);
 
 /*
@@ -52,11 +52,11 @@ int abd_iterate_page_func(abd_t *, size_t, size_t, abd_iter_page_func_t *,
  * Note: these are only needed to support vdev_classic. See comment in
  * vdev_disk.c.
  */
-unsigned int abd_bio_map_off(struct bio *, abd_t *, unsigned int, size_t);
+unsigned int abd_bio_map_off(struct bio *, struct abd *, unsigned int, size_t);
-unsigned long abd_nr_pages_off(abd_t *, unsigned int, size_t);
+unsigned long abd_nr_pages_off(struct abd *, unsigned int, size_t);
 
 __attribute__((malloc))
-abd_t *abd_alloc_from_pages(struct page **, unsigned long, uint64_t);
+struct abd *abd_alloc_from_pages(struct page **, unsigned long, uint64_t);
 
 #ifdef __cplusplus
 }

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

@@ -69,6 +69,7 @@ typedef struct vfs {
 	boolean_t	vfs_do_relatime;
 	boolean_t	vfs_nbmand;
 	boolean_t	vfs_do_nbmand;
+	kmutex_t	vfs_mntpt_lock;
 } vfs_t;
 
 typedef struct zfs_mnt {

include/sys/dbuf.h

@@ -171,7 +171,6 @@ typedef struct dbuf_dirty_record {
 			 * gets COW'd in a subsequent transaction group.
 			 */
 			arc_buf_t *dr_data;
-			blkptr_t dr_overridden_by;
 			override_states_t dr_override_state;
 			uint8_t dr_copies;
 			boolean_t dr_nopwrite;
@@ -179,14 +178,21 @@ typedef struct dbuf_dirty_record {
 			boolean_t dr_diowrite;
 			boolean_t dr_has_raw_params;
 
-			/*
+			/* Override and raw params are mutually exclusive. */
-			 * If dr_has_raw_params is set, the following crypt
+			union {
-			 * params will be set on the BP that's written.
+				blkptr_t dr_overridden_by;
-			 */
+				struct {
-			boolean_t dr_byteorder;
+					/*
-			uint8_t	dr_salt[ZIO_DATA_SALT_LEN];
+					 * If dr_has_raw_params is set, the
-			uint8_t	dr_iv[ZIO_DATA_IV_LEN];
+					 * following crypt params will be set
-			uint8_t	dr_mac[ZIO_DATA_MAC_LEN];
+					 * on the BP that's written.
+					 */
+					boolean_t dr_byteorder;
+					uint8_t	dr_salt[ZIO_DATA_SALT_LEN];
+					uint8_t	dr_iv[ZIO_DATA_IV_LEN];
+					uint8_t	dr_mac[ZIO_DATA_MAC_LEN];
+				};
+			};
 		} dl;
 		struct dirty_lightweight_leaf {
 			/*
@@ -264,6 +270,27 @@ typedef struct dmu_buf_impl {
 	 */
 	uint8_t db_level;
 
+	/* This block was freed while a read or write was active. */
+	uint8_t db_freed_in_flight;
+
+	/*
+	 * Evict user data as soon as the dirty and reference counts are equal.
+	 */
+	uint8_t db_user_immediate_evict;
+
+	/*
+	 * dnode_evict_dbufs() or dnode_evict_bonus() tried to evict this dbuf,
+	 * but couldn't due to outstanding references.  Evict once the refcount
+	 * drops to 0.
+	 */
+	uint8_t db_pending_evict;
+
+	/* Number of TXGs in which this buffer is dirty. */
+	uint8_t db_dirtycnt;
+
+	/* The buffer was partially read.  More reads may follow. */
+	uint8_t db_partial_read;
+
 	/*
 	 * Protects db_buf's contents if they contain an indirect block or data
 	 * block of the meta-dnode. We use this lock to protect the structure of
@@ -288,6 +315,9 @@ typedef struct dmu_buf_impl {
 	 */
 	dbuf_states_t db_state;
 
+	/* In which dbuf cache this dbuf is, if any. */
+	dbuf_cached_state_t db_caching_status;
+
 	/*
 	 * Refcount accessed by dmu_buf_{hold,rele}.
 	 * If nonzero, the buffer can't be destroyed.
@@ -304,39 +334,10 @@ typedef struct dmu_buf_impl {
 	/* Link in dbuf_cache or dbuf_metadata_cache */
 	multilist_node_t db_cache_link;
 
-	/* Tells us which dbuf cache this dbuf is in, if any */
-	dbuf_cached_state_t db_caching_status;
-
 	uint64_t db_hash;
 
-	/* Data which is unique to data (leaf) blocks: */
-
 	/* User callback information. */
 	dmu_buf_user_t *db_user;
-
-	/*
-	 * Evict user data as soon as the dirty and reference
-	 * counts are equal.
-	 */
-	uint8_t db_user_immediate_evict;
-
-	/*
-	 * This block was freed while a read or write was
-	 * active.
-	 */
-	uint8_t db_freed_in_flight;
-
-	/*
-	 * dnode_evict_dbufs() or dnode_evict_bonus() tried to
-	 * evict this dbuf, but couldn't due to outstanding
-	 * references.  Evict once the refcount drops to 0.
-	 */
-	uint8_t db_pending_evict;
-
-	uint8_t db_dirtycnt;
-
-	/* The buffer was partially read.  More reads may follow. */
-	uint8_t db_partial_read;
 } dmu_buf_impl_t;
 
 #define	DBUF_HASH_MUTEX(h, idx) \
@@ -351,6 +352,8 @@ typedef struct dbuf_hash_table {
 
 typedef void (*dbuf_prefetch_fn)(void *, uint64_t, uint64_t, boolean_t);
 
+extern kmem_cache_t *dbuf_dirty_kmem_cache;
+
 uint64_t dbuf_whichblock(const struct dnode *di, const int64_t level,
     const uint64_t offset);
 

include/sys/fm/fs/zfs.h

@@ -42,7 +42,8 @@ extern "C" {
 #define	FM_EREPORT_ZFS_DATA			"data"
 #define	FM_EREPORT_ZFS_DELAY			"delay"
 #define	FM_EREPORT_ZFS_DEADMAN			"deadman"
-#define	FM_EREPORT_ZFS_DIO_VERIFY		"dio_verify"
+#define	FM_EREPORT_ZFS_DIO_VERIFY_WR		"dio_verify_wr"
+#define	FM_EREPORT_ZFS_DIO_VERIFY_RD		"dio_verify_rd"
 #define	FM_EREPORT_ZFS_POOL			"zpool"
 #define	FM_EREPORT_ZFS_DEVICE_UNKNOWN		"vdev.unknown"
 #define	FM_EREPORT_ZFS_DEVICE_OPEN_FAILED	"vdev.open_failed"

include/sys/vdev_raidz.h

@@ -57,7 +57,7 @@ void vdev_raidz_reconstruct(struct raidz_map *, const int *, int);
 void vdev_raidz_child_done(zio_t *);
 void vdev_raidz_io_done(zio_t *);
 void vdev_raidz_checksum_error(zio_t *, struct raidz_col *, abd_t *);
-struct raidz_row *vdev_raidz_row_alloc(int);
+struct raidz_row *vdev_raidz_row_alloc(int, zio_t *);
 void vdev_raidz_reflow_copy_scratch(spa_t *);
 void raidz_dtl_reassessed(vdev_t *);
 

include/sys/zio.h

@@ -208,25 +208,25 @@ typedef uint64_t zio_flag_t;
 #define	ZIO_FLAG_PROBE		(1ULL << 16)
 #define	ZIO_FLAG_TRYHARD	(1ULL << 17)
 #define	ZIO_FLAG_OPTIONAL	(1ULL << 18)
-
+#define	ZIO_FLAG_DIO_READ	(1ULL << 19)
 #define	ZIO_FLAG_VDEV_INHERIT	(ZIO_FLAG_DONT_QUEUE - 1)
 
 	/*
 	 * Flags not inherited by any children.
 	 */
-#define	ZIO_FLAG_DONT_QUEUE	(1ULL << 19)	/* must be first for INHERIT */
+#define	ZIO_FLAG_DONT_QUEUE	(1ULL << 20)	/* must be first for INHERIT */
-#define	ZIO_FLAG_DONT_PROPAGATE	(1ULL << 20)
+#define	ZIO_FLAG_DONT_PROPAGATE	(1ULL << 21)
-#define	ZIO_FLAG_IO_BYPASS	(1ULL << 21)
+#define	ZIO_FLAG_IO_BYPASS	(1ULL << 22)
-#define	ZIO_FLAG_IO_REWRITE	(1ULL << 22)
+#define	ZIO_FLAG_IO_REWRITE	(1ULL << 23)
-#define	ZIO_FLAG_RAW_COMPRESS	(1ULL << 23)
+#define	ZIO_FLAG_RAW_COMPRESS	(1ULL << 24)
-#define	ZIO_FLAG_RAW_ENCRYPT	(1ULL << 24)
+#define	ZIO_FLAG_RAW_ENCRYPT	(1ULL << 25)
-#define	ZIO_FLAG_GANG_CHILD	(1ULL << 25)
+#define	ZIO_FLAG_GANG_CHILD	(1ULL << 26)
-#define	ZIO_FLAG_DDT_CHILD	(1ULL << 26)
+#define	ZIO_FLAG_DDT_CHILD	(1ULL << 27)
-#define	ZIO_FLAG_GODFATHER	(1ULL << 27)
+#define	ZIO_FLAG_GODFATHER	(1ULL << 28)
-#define	ZIO_FLAG_NOPWRITE	(1ULL << 28)
+#define	ZIO_FLAG_NOPWRITE	(1ULL << 29)
-#define	ZIO_FLAG_REEXECUTED	(1ULL << 29)
+#define	ZIO_FLAG_REEXECUTED	(1ULL << 30)
-#define	ZIO_FLAG_DELEGATED	(1ULL << 30)
+#define	ZIO_FLAG_DELEGATED	(1ULL << 31)
-#define	ZIO_FLAG_DIO_CHKSUM_ERR	(1ULL << 31)
+#define	ZIO_FLAG_DIO_CHKSUM_ERR	(1ULL << 32)
 
 #define	ZIO_ALLOCATOR_NONE	(-1)
 #define	ZIO_HAS_ALLOCATOR(zio)	((zio)->io_allocator != ZIO_ALLOCATOR_NONE)
@@ -647,6 +647,7 @@ extern void zio_vdev_io_redone(zio_t *zio);
 extern void zio_change_priority(zio_t *pio, zio_priority_t priority);
 
 extern void zio_checksum_verified(zio_t *zio);
+extern void zio_dio_chksum_verify_error_report(zio_t *zio);
 extern int zio_worst_error(int e1, int e2);
 
 extern enum zio_checksum zio_checksum_select(enum zio_checksum child,

lib/libspl/backtrace.c

@@ -25,19 +25,32 @@
 
 #include <sys/backtrace.h>
 #include <sys/types.h>
+#include <sys/debug.h>
 #include <unistd.h>
 
 /*
- * libspl_backtrace() must be safe to call from inside a signal hander. This
+ * Output helpers. libspl_backtrace() must not block, must be thread-safe and
- * mostly means it must not allocate, and so we can't use things like printf.
+ * must be safe to call from a signal handler. At least, that means not having
+ * printf, so we end up having to call write() directly on the fd. That's
+ * awkward, as we always have to pass through a length, and some systems will
+ * complain if we don't consume the return. So we have some macros to make
+ * things a little more palatable.
  */
+#define	spl_bt_write_n(fd, s, n) \
+	do { ssize_t r __maybe_unused = write(fd, s, n); } while (0)
+#define	spl_bt_write(fd, s)		spl_bt_write_n(fd, s, sizeof (s))
 
 #if defined(HAVE_LIBUNWIND)
 #define	UNW_LOCAL_ONLY
 #include <libunwind.h>
 
+/*
+ * Convert `v` to ASCII hex characters. The bottom `n` nybbles (4-bits ie one
+ * hex digit) will be written, up to `buflen`. The buffer will not be
+ * null-terminated. Returns the number of digits written.
+ */
 static size_t
-libspl_u64_to_hex_str(uint64_t v, size_t digits, char *buf, size_t buflen)
+spl_bt_u64_to_hex_str(uint64_t v, size_t n, char *buf, size_t buflen)
 {
 	static const char hexdigits[] = {
 	    '0', '1', '2', '3', '4', '5', '6', '7',
@@ -45,10 +58,10 @@ libspl_u64_to_hex_str(uint64_t v, size_t digits, char *buf, size_t buflen)
 	};
 
 	size_t pos = 0;
-	boolean_t want = (digits == 0);
+	boolean_t want = (n == 0);
 	for (int i = 15; i >= 0; i--) {
 		const uint64_t d = v >> (i * 4) & 0xf;
-		if (!want && (d != 0 || digits > i))
+		if (!want && (d != 0 || n > i))
 			want = B_TRUE;
 		if (want) {
 			buf[pos++] = hexdigits[d];
@@ -62,40 +75,181 @@ libspl_u64_to_hex_str(uint64_t v, size_t digits, char *buf, size_t buflen)
 void
 libspl_backtrace(int fd)
 {
-	ssize_t ret __attribute__((unused));
 	unw_context_t uc;
 	unw_cursor_t cp;
-	unw_word_t loc;
+	unw_word_t v;
 	char buf[128];
 	size_t n;
+	int err;
 
-	ret = write(fd, "Call trace:\n", 12);
+	/* Snapshot the current frame and state. */
 	unw_getcontext(&uc);
+
+	/*
+	 * TODO: walk back to the frame that tripped the assertion / the place
+	 *       where the signal was recieved.
+	 */
+
+	/*
+	 * Register dump. We're going to loop over all the registers in the
+	 * top frame, and show them, with names, in a nice three-column
+	 * layout, which keeps us within 80 columns.
+	 */
+	spl_bt_write(fd, "Registers:\n");
+
+	/* Initialise a frame cursor, starting at the current frame */
 	unw_init_local(&cp, &uc);
-	while (unw_step(&cp) > 0) {
+
-		unw_get_reg(&cp, UNW_REG_IP, &loc);
+	/*
-		ret = write(fd, "  [0x", 5);
+	 * libunwind's list of possible registers for this architecture is an
-		n = libspl_u64_to_hex_str(loc, 10, buf, sizeof (buf));
+	 * enum, unw_regnum_t. UNW_TDEP_LAST_REG is the highest-numbered
-		ret = write(fd, buf, n);
+	 * register in that list, however, not all register numbers in this
-		ret = write(fd, "] ", 2);
+	 * range are defined by the architecture, and not all defined registers
-		unw_get_proc_name(&cp, buf, sizeof (buf), &loc);
+	 * will be present on every implementation of that architecture.
-		for (n = 0; n < sizeof (buf) && buf[n] != '\0'; n++) {}
+	 * Moreover, libunwind provides nice names for most, but not all
-		ret = write(fd, buf, n);
+	 * registers, but these are hardcoded; a name being available does not
-		ret = write(fd, "+0x", 3);
+	 * mean that register is available.
-		n = libspl_u64_to_hex_str(loc, 2, buf, sizeof (buf));
+	 *
-		ret = write(fd, buf, n);
+	 * So, we have to pull this all together here. We try to get the value
-#ifdef HAVE_LIBUNWIND_ELF
+	 * of every possible register. If we get a value for it, then the
-		ret = write(fd, " (in ", 5);
+	 * register must exist, and so we get its name. If libunwind has no
-		unw_get_elf_filename(&cp, buf, sizeof (buf), &loc);
+	 * name for it, we synthesize something. These cases should be rare,
-		for (n = 0; n < sizeof (buf) && buf[n] != '\0'; n++) {}
+	 * and they're usually for uninteresting or niche registers, so it
-		ret = write(fd, buf, n);
+	 * shouldn't really matter. We can see the value, and that's the main
-		ret = write(fd, " +0x", 4);
+	 * thing.
-		n = libspl_u64_to_hex_str(loc, 2, buf, sizeof (buf));
+	 */
-		ret = write(fd, buf, n);
+	uint_t cols = 0;
-		ret = write(fd, ")", 1);
+	for (uint_t regnum = 0; regnum <= UNW_TDEP_LAST_REG; regnum++) {
-#endif
+		/*
-		ret = write(fd, "\n", 1);
+		 * Get the value. Any error probably means the register
+		 * doesn't exist, and we skip it.
+		 */
+		if (unw_get_reg(&cp, regnum, &v) < 0)
+			continue;
+
+		/*
+		 * Register name. If libunwind doesn't have a name for it,
+		 * it will return "???". As a shortcut, we just treat '?'
+		 * is an alternate end-of-string character.
+		 */
+		const char *name = unw_regname(regnum);
+		for (n = 0; name[n] != '\0' && name[n] != '?'; n++) {}
+		if (n == 0) {
+			/*
+			 * No valid name, so make one of the form "?xx", where
+			 * "xx" is the two-char hex of libunwind's register
+			 * number.
+			 */
+			buf[0] = '?';
+			n = spl_bt_u64_to_hex_str(regnum, 2,
+			    &buf[1], sizeof (buf)-1) + 1;
+			name = buf;
+		}
+
+		/*
+		 * Two spaces of padding before each column, plus extra
+		 * spaces to align register names shorter than three chars.
+		 */
+		spl_bt_write_n(fd, "      ", 5-MIN(n, 3));
+
+		/* Register name and column punctuation */
+		spl_bt_write_n(fd, name, n);
+		spl_bt_write(fd, ": 0x");
+
+		/*
+		 * Convert register value (from unw_get_reg()) to hex. We're
+		 * assuming that all registers are 64-bits wide, which is
+		 * probably fine for any general-purpose registers on any
+		 * machine currently in use. A more generic way would be to
+		 * look at the width of unw_word_t, but that would also
+		 * complicate the column code a bit. This is fine.
+		 */
+		n = spl_bt_u64_to_hex_str(v, 16, buf, sizeof (buf));
+		spl_bt_write_n(fd, buf, n);
+
+		/* Every third column, emit a newline */
+		if (!(++cols % 3))
+			spl_bt_write(fd, "\n");
 	}
+
+	/* If we finished before the third column, emit a newline. */
+	if (cols % 3)
+		spl_bt_write(fd, "\n");
+
+	/* Now the main event, the backtrace. */
+	spl_bt_write(fd, "Call trace:\n");
+
+	/* Reset the cursor to the top again. */
+	unw_init_local(&cp, &uc);
+
+	do {
+		/*
+		 * Getting the IP should never fail; libunwind handles it
+		 * specially, because its used a lot internally. Still, no
+		 * point being silly about it, as the last thing we want is
+		 * our crash handler to crash. So if it ever does fail, we'll
+		 * show an error line, but keep going to the next frame.
+		 */
+		if (unw_get_reg(&cp, UNW_REG_IP, &v) < 0) {
+			spl_bt_write(fd, "  [couldn't get IP register; "
+			    "corrupt frame?]");
+			continue;
+		}
+
+		/* IP & punctuation */
+		n = spl_bt_u64_to_hex_str(v, 16, buf, sizeof (buf));
+		spl_bt_write(fd, "  [0x");
+		spl_bt_write_n(fd, buf, n);
+		spl_bt_write(fd, "] ");
+
+		/*
+		 * Function ("procedure") name for the current frame. `v`
+		 * receives the offset from the named function to the IP, which
+		 * we show as a "+offset" suffix.
+		 *
+		 * If libunwind can't determine the name, we just show "???"
+		 * instead. We've already displayed the IP above; that will
+		 * have to do.
+		 *
+		 * unw_get_proc_name() will return ENOMEM if the buffer is too
+		 * small, instead truncating the name. So we treat that as a
+		 * success and use whatever is in the buffer.
+		 */
+		err = unw_get_proc_name(&cp, buf, sizeof (buf), &v);
+		if (err == 0 || err == -UNW_ENOMEM) {
+			for (n = 0; n < sizeof (buf) && buf[n] != '\0'; n++) {}
+			spl_bt_write_n(fd, buf, n);
+
+			/* Offset from proc name */
+			spl_bt_write(fd, "+0x");
+			n = spl_bt_u64_to_hex_str(v, 2, buf, sizeof (buf));
+			spl_bt_write_n(fd, buf, n);
+		} else
+			spl_bt_write(fd, "???");
+
+#ifdef HAVE_LIBUNWIND_ELF
+		/*
+		 * Newer libunwind has unw_get_elf_filename(), which gets
+		 * the name of the ELF object that the frame was executing in.
+		 * Like `unw_get_proc_name()`, `v` recieves the offset within
+		 * the file, and UNW_ENOMEM indicates that a truncate filename
+		 * was left in the buffer.
+		 */
+		err = unw_get_elf_filename(&cp, buf, sizeof (buf), &v);
+		if (err == 0 || err == -UNW_ENOMEM) {
+			for (n = 0; n < sizeof (buf) && buf[n] != '\0'; n++) {}
+			spl_bt_write(fd, " (in ");
+			spl_bt_write_n(fd, buf, n);
+
+			/* Offset within file */
+			spl_bt_write(fd, " +0x");
+			n = spl_bt_u64_to_hex_str(v, 2, buf, sizeof (buf));
+			spl_bt_write_n(fd, buf, n);
+			spl_bt_write(fd, ")");
+		}
+#endif
+		spl_bt_write(fd, "\n");
+	} while (unw_step(&cp) > 0);
 }
 #elif defined(HAVE_BACKTRACE)
 #include <execinfo.h>
@@ -103,15 +257,12 @@ libspl_backtrace(int fd)
 void
 libspl_backtrace(int fd)
 {
-	ssize_t ret __attribute__((unused));
 	void *btptrs[64];
 	size_t nptrs = backtrace(btptrs, 64);
-	ret = write(fd, "Call trace:\n", 12);
+	spl_bt_write(fd, "Call trace:\n");
 	backtrace_symbols_fd(btptrs, nptrs, fd);
 }
 #else
-#include <sys/debug.h>
-
 void
 libspl_backtrace(int fd __maybe_unused)
 {

lib/libspl/include/sys/uio.h

@@ -92,20 +92,20 @@ zfs_dio_page_aligned(void *buf)
 static inline boolean_t
 zfs_dio_offset_aligned(uint64_t offset, uint64_t blksz)
 {
-	return (IS_P2ALIGNED(offset, blksz));
+	return ((IS_P2ALIGNED(offset, blksz)) ? B_TRUE : B_FALSE);
 }
 
 static inline boolean_t
 zfs_dio_size_aligned(uint64_t size, uint64_t blksz)
 {
-	return ((size % blksz) == 0);
+	return (((size % blksz) == 0) ? B_TRUE : B_FALSE);
 }
 
 static inline boolean_t
 zfs_dio_aligned(uint64_t offset, uint64_t size, uint64_t blksz)
 {
-	return (zfs_dio_offset_aligned(offset, blksz) &&
+	return ((zfs_dio_offset_aligned(offset, blksz) &&
-	    zfs_dio_size_aligned(size, blksz));
+	    zfs_dio_size_aligned(size, blksz)) ? B_TRUE : B_FALSE);
 }
 
 static inline void

lib/libzfs/Makefile.am

@@ -70,7 +70,7 @@ if BUILD_FREEBSD
 libzfs_la_LIBADD += -lutil -lgeom
 endif
 
-libzfs_la_LDFLAGS += -version-info 5:0:1
+libzfs_la_LDFLAGS += -version-info 6:0:0
 
 pkgconfig_DATA += %D%/libzfs.pc
 

lib/libzfs/libzfs.abi

@@ -1,4 +1,4 @@
-<abi-corpus version='2.0' architecture='elf-amd-x86_64' soname='libzfs.so.4'>
+<abi-corpus version='2.0' architecture='elf-amd-x86_64' soname='libzfs.so.6'>
   <elf-needed>
     <dependency name='libzfs_core.so.3'/>
     <dependency name='libnvpair.so.3'/>

lib/libzfs/libzfs_pool.c

@@ -2796,7 +2796,7 @@ zpool_scan(zpool_handle_t *zhp, pool_scan_func_t func, pool_scrub_cmd_t cmd)
 	}
 
 	/*
-	 * With EBUSY, five cases are possible:
+	 * With EBUSY, six cases are possible:
 	 *
 	 * Current state		Requested
 	 * 1. Normal Scrub Running	Normal Scrub or Error Scrub

lib/libzpool/Makefile.am

@@ -212,7 +212,7 @@ if BUILD_FREEBSD
 libzpool_la_LIBADD += -lgeom
 endif
 
-libzpool_la_LDFLAGS += -version-info 5:0:0
+libzpool_la_LDFLAGS += -version-info 6:0:0
 
 if TARGET_CPU_POWERPC
 module/zfs/libzpool_la-vdev_raidz_math_powerpc_altivec.$(OBJEXT) : CFLAGS += -maltivec

lib/libzpool/zfs_debug.c

@@ -35,9 +35,25 @@ typedef struct zfs_dbgmsg {
 
 static list_t zfs_dbgmsgs;
 static kmutex_t zfs_dbgmsgs_lock;
+static uint_t zfs_dbgmsg_size = 0;
+static uint_t zfs_dbgmsg_maxsize = 4<<20; /* 4MB */
 
 int zfs_dbgmsg_enable = B_TRUE;
 
+static void
+zfs_dbgmsg_purge(uint_t max_size)
+{
+	while (zfs_dbgmsg_size > max_size) {
+		zfs_dbgmsg_t *zdm = list_remove_head(&zfs_dbgmsgs);
+		if (zdm == NULL)
+			return;
+
+		uint_t size = zdm->zdm_size;
+		kmem_free(zdm, size);
+		zfs_dbgmsg_size -= size;
+	}
+}
+
 void
 zfs_dbgmsg_init(void)
 {
@@ -74,6 +90,8 @@ __zfs_dbgmsg(char *buf)
 
 	mutex_enter(&zfs_dbgmsgs_lock);
 	list_insert_tail(&zfs_dbgmsgs, zdm);
+	zfs_dbgmsg_size += size;
+	zfs_dbgmsg_purge(zfs_dbgmsg_maxsize);
 	mutex_exit(&zfs_dbgmsgs_lock);
 }
 

man/man4/zfs.4

@@ -18,7 +18,7 @@
 .\"
 .\" Copyright (c) 2024, Klara, Inc.
 .\"
-.Dd October 2, 2024
+.Dd November 1, 2024
 .Dt ZFS 4
 .Os
 .
@@ -436,7 +436,7 @@ write.
 It can also help to identify if reported checksum errors are tied to Direct I/O
 writes.
 Each verify error causes a
-.Sy dio_verify
+.Sy dio_verify_wr
 zevent.
 Direct Write I/O checkum verify errors can be seen with
 .Nm zpool Cm status Fl d .
@@ -1333,9 +1333,10 @@ results in vector instructions
 from the respective CPU instruction set being used.
 .
 .It Sy zfs_bclone_enabled Ns = Ns Sy 1 Ns | Ns 0 Pq int
-Enable the experimental block cloning feature.
+Enables access to the block cloning feature.
 If this setting is 0, then even if feature@block_cloning is enabled,
-attempts to clone blocks will act as though the feature is disabled.
+using functions and system calls that attempt to clone blocks will act as
+though the feature is disabled.
 .
 .It Sy zfs_bclone_wait_dirty Ns = Ns Sy 0 Ns | Ns 1 Pq int
 When set to 1 the FICLONE and FICLONERANGE ioctls wait for dirty data to be

man/man5/vdev_id.conf.5

@@ -92,6 +92,11 @@ before a generic mapping for the same slot.
 In this way a custom mapping may be applied to a particular channel
 and a default mapping applied to the others.
 .
+.It Sy zpad_slot Ar digits
+Pad slot numbers with zeros to make them
+.Ar digits
+long, which can help to make disk names a consistent length and easier to sort.
+.
 .It Sy multipath Sy yes Ns | Ns Sy no
 Specifies whether
 .Xr vdev_id 8
@@ -122,7 +127,7 @@ device is connected to.
 The default is
 .Sy 4 .
 .
-.It Sy slot Sy bay Ns | Ns Sy phy Ns | Ns Sy port Ns | Ns Sy id Ns | Ns Sy lun Ns | Ns Sy ses
+.It Sy slot Sy bay Ns | Ns Sy phy Ns | Ns Sy port Ns | Ns Sy id Ns | Ns Sy lun Ns | Ns Sy bay_lun Ns | Ns Sy ses
 Specifies from which element of a SAS identifier the slot number is
 taken.
 The default is
@@ -138,6 +143,9 @@ use the SAS port as the slot number.
 use the scsi id as the slot number.
 .It Sy lun
 use the scsi lun as the slot number.
+.It Sy bay_lun
+read the slot number from the bay identifier and append the lun number.
+Useful for multi-lun multi-actuator hard drives.
 .It Sy ses
 use the SCSI Enclosure Services (SES) enclosure device slot number,
 as reported by

man/man8/zdb.8

@@ -14,7 +14,7 @@
 .\" Copyright (c) 2017 Lawrence Livermore National Security, LLC.
 .\" Copyright (c) 2017 Intel Corporation.
 .\"
-.Dd November 18, 2023
+.Dd October 27, 2024
 .Dt ZDB 8
 .Os
 .
@@ -408,6 +408,8 @@ blocks cloned, the space saving as a result of cloning, and the saving ratio.
 .It Fl TT
 Display the per-vdev BRT statistics, including total references.
 .It Fl TTT
+Display histograms of per-vdev BRT refcounts.
+.It Fl TTTT
 Dump the contents of the block reference tables.
 .It Fl u , -uberblock
 Display the current uberblock.

man/man8/zfs-list.8

@@ -71,7 +71,7 @@ The following fields are displayed:
 Used for scripting mode.
 Do not print headers and separate fields by a single tab instead of arbitrary
 white space.
-.It Fl j Op Ar --json-int
+.It Fl j , -json Op Ar --json-int
 Print the output in JSON format.
 Specify
 .Sy --json-int

man/man8/zfs-mount.8

@@ -59,7 +59,7 @@
 .Xc
 Displays all ZFS file systems currently mounted.
 .Bl -tag -width "-j"
-.It Fl j
+.It Fl j , -json
 Displays all mounted file systems in JSON format.
 .El
 .It Xo

man/man8/zfs-program.8

@@ -50,7 +50,7 @@ and any attempts to access or modify other pools will cause an error.
 .
 .Sh OPTIONS
 .Bl -tag -width "-t"
-.It Fl j
+.It Fl j , -json
 Display channel program output in JSON format.
 When this flag is specified and standard output is empty -
 channel program encountered an error.

man/man8/zfs-set.8

@@ -130,7 +130,7 @@ The value
 can be used to display all properties that apply to the given dataset's type
 .Pq Sy filesystem , volume , snapshot , No or Sy bookmark .
 .Bl -tag -width "-s source"
-.It Fl j Op Ar --json-int
+.It Fl j , -json Op Ar --json-int
 Display the output in JSON format.
 Specify
 .Sy --json-int

man/man8/zpool-events.8

@@ -98,7 +98,10 @@ This can be an indicator of problems with the underlying storage device.
 The number of delay events is ratelimited by the
 .Sy zfs_slow_io_events_per_second
 module parameter.
-.It Sy dio_verify
+.It Sy dio_verify_rd
+Issued when there was a checksum verify error after a Direct I/O read has been
+issued.
+.It Sy dio_verify_wr
 Issued when there was a checksum verify error after a Direct I/O write has been
 issued.
 This event can only take place if the module parameter

man/man8/zpool-get.8

@@ -98,7 +98,7 @@ See the
 .Xr zpoolprops 7
 manual page for more information on the available pool properties.
 .Bl -tag -compact -offset Ds -width "-o field"
-.It Fl j Op Ar --json-int, --json-pool-key-guid
+.It Fl j , -json Op Ar --json-int, --json-pool-key-guid
 Display the list of properties in JSON format.
 Specify
 .Sy --json-int
@@ -157,7 +157,7 @@ See the
 .Xr vdevprops 7
 manual page for more information on the available pool properties.
 .Bl -tag -compact -offset Ds -width "-o field"
-.It Fl j Op Ar --json-int
+.It Fl j , -json Op Ar --json-int
 Display the list of properties in JSON format.
 Specify
 .Sy --json-int

man/man8/zpool-list.8

@@ -59,7 +59,7 @@ is specified, the command exits after
 .Ar count
 reports are printed.
 .Bl -tag -width Ds
-.It Fl j Op Ar --json-int, --json-pool-key-guid
+.It Fl j , -json Op Ar --json-int, --json-pool-key-guid
 Display the list of pools in JSON format.
 Specify
 .Sy --json-int

man/man8/zpool-status.8

@@ -70,7 +70,7 @@ See the
 option of
 .Nm zpool Cm iostat
 for complete details.
-.It Fl j Op Ar --json-int, --json-flat-vdevs, --json-pool-key-guid
+.It Fl j , -json Op Ar --json-int, --json-flat-vdevs, --json-pool-key-guid
 Display the status for ZFS pools in JSON format.
 Specify
 .Sy --json-int
@@ -82,14 +82,18 @@ Specify
 .Sy --json-pool-key-guid
 to set pool GUID as key for pool objects instead of pool names.
 .It Fl d
-Display the number of Direct I/O write checksum verify errors that have occured
+Display the number of Direct I/O read/write checksum verify errors that have
-on a top-level VDEV.
+occured on a top-level VDEV.
 See
 .Sx zfs_vdev_direct_write_verify
 in
 .Xr zfs 4
 for details about the conditions that can cause Direct I/O write checksum
 verify failures to occur.
+Direct I/O reads checksum verify errors can also occur if the contents of the
+buffer are being manipulated after the I/O has been issued and is in flight.
+In the case of Direct I/O read checksum verify errors, the I/O will be reissued
+through the ARC.
 .It Fl D
 Display a histogram of deduplication statistics, showing the allocated
 .Pq physically present on disk

module/os/freebsd/zfs/abd_os.c

@@ -620,9 +620,16 @@ abd_borrow_buf_copy(abd_t *abd, size_t n)
 
 /*
  * Return a borrowed raw buffer to an ABD. If the ABD is scattered, this will
- * no change the contents of the ABD and will ASSERT that you didn't modify
+ * not change the contents of the ABD. If you want any changes you made to
- * the buffer since it was borrowed. If you want any changes you made to buf to
+ * buf to be copied back to abd, use abd_return_buf_copy() instead. If the
- * be copied back to abd, use abd_return_buf_copy() instead.
+ * ABD is not constructed from user pages from Direct I/O then an ASSERT
+ * checks to make sure the contents of the buffer have not changed since it was
+ * borrowed. We can not ASSERT the contents of the buffer have not changed if
+ * it is composed of user pages. While Direct I/O write pages are placed under
+ * write protection and can not be changed, this is not the case for Direct I/O
+ * reads. The pages of a Direct I/O read could be manipulated at any time.
+ * Checksum verifications in the ZIO pipeline check for this issue and handle
+ * it by returning an error on checksum verification failure.
  */
 void
 abd_return_buf(abd_t *abd, void *buf, size_t n)
@@ -632,8 +639,34 @@ abd_return_buf(abd_t *abd, void *buf, size_t n)
 #ifdef ZFS_DEBUG
 	(void) zfs_refcount_remove_many(&abd->abd_children, n, buf);
 #endif
-	if (abd_is_linear(abd)) {
+	if (abd_is_from_pages(abd)) {
+		if (!abd_is_linear_page(abd))
+			zio_buf_free(buf, n);
+	} else if (abd_is_linear(abd)) {
 		ASSERT3P(buf, ==, abd_to_buf(abd));
+	} else if (abd_is_gang(abd)) {
+#ifdef ZFS_DEBUG
+		/*
+		 * We have to be careful with gang ABD's that we do not ASSERT
+		 * for any ABD's that contain user pages from Direct I/O. See
+		 * the comment above about Direct I/O read buffers possibly
+		 * being manipulated. In order to handle this, we jsut iterate
+		 * through the gang ABD and only verify ABD's that are not from
+		 * user pages.
+		 */
+		void *cmp_buf = buf;
+
+		for (abd_t *cabd = list_head(&ABD_GANG(abd).abd_gang_chain);
+		    cabd != NULL;
+		    cabd = list_next(&ABD_GANG(abd).abd_gang_chain, cabd)) {
+			if (!abd_is_from_pages(cabd)) {
+				ASSERT0(abd_cmp_buf(cabd, cmp_buf,
+				    cabd->abd_size));
+			}
+			cmp_buf = (char *)cmp_buf + cabd->abd_size;
+		}
+#endif
+		zio_buf_free(buf, n);
 	} else {
 		ASSERT0(abd_cmp_buf(abd, buf, n));
 		zio_buf_free(buf, n);

module/os/freebsd/zfs/zio_crypt.c

@@ -1686,11 +1686,10 @@ zio_do_crypt_data(boolean_t encrypt, zio_crypt_key_t *key,
 	freebsd_crypt_session_t *tmpl = NULL;
 	uint8_t *authbuf = NULL;
 
-
+	memset(&puio_s, 0, sizeof (puio_s));
+	memset(&cuio_s, 0, sizeof (cuio_s));
 	zfs_uio_init(&puio, &puio_s);
 	zfs_uio_init(&cuio, &cuio_s);
-	memset(GET_UIO_STRUCT(&puio), 0, sizeof (struct uio));
-	memset(GET_UIO_STRUCT(&cuio), 0, sizeof (struct uio));
 
 #ifdef FCRYPTO_DEBUG
 	printf("%s(%s, %p, %p, %d, %p, %p, %u, %s, %p, %p, %p)\n",

module/os/linux/spl/spl-thread.c

@@ -171,11 +171,11 @@ issig(void)
 #if defined(HAVE_DEQUEUE_SIGNAL_4ARG)
 	enum pid_type __type;
 	if (dequeue_signal(current, &set, &__info, &__type) != 0) {
-#elif defined(HAVE_DEQUEUE_SIGNAL_3ARG_TASK)
+#elif defined(HAVE_DEQUEUE_SIGNAL_3ARG_TYPE)
-	if (dequeue_signal(current, &set, &__info) != 0) {
-#else
 	enum pid_type __type;
 	if (dequeue_signal(&set, &__info, &__type) != 0) {
+#else
+	if (dequeue_signal(current, &set, &__info) != 0) {
 #endif
 		spin_unlock_irq(&current->sighand->siglock);
 		kernel_signal_stop();

module/os/linux/zfs/abd_os.c

@@ -701,6 +701,8 @@ abd_free_linear_page(abd_t *abd)
 	/* When backed by user page unmap it */
 	if (abd_is_from_pages(abd))
 		zfs_kunmap(sg_page(sg));
+	else
+		abd_update_scatter_stats(abd, ABDSTAT_DECR);
 
 	abd->abd_flags &= ~ABD_FLAG_LINEAR;
 	abd->abd_flags &= ~ABD_FLAG_LINEAR_PAGE;
@@ -1008,7 +1010,9 @@ abd_borrow_buf_copy(abd_t *abd, size_t n)
  * borrowed. We can not ASSERT that the contents of the buffer have not changed
  * if it is composed of user pages because the pages can not be placed under
  * write protection and the user could have possibly changed the contents in
- * the pages at any time.
+ * the pages at any time. This is also an issue for Direct I/O reads. Checksum
+ * verifications in the ZIO pipeline check for this issue and handle it by
+ * returning an error on checksum verification failure.
  */
 void
 abd_return_buf(abd_t *abd, void *buf, size_t n)

module/os/linux/zfs/vdev_disk.c

@@ -801,24 +801,13 @@ vbio_completion(struct bio *bio)
 	bio_put(bio);
 
 	/*
-	 * If we copied the ABD before issuing it, clean up and return the copy
+	 * We're likely in an interrupt context so we can't do ABD/memory work
-	 * to the ADB, with changes if appropriate.
+	 * here; instead we stash vbio on the zio and take care of it in the
+	 * done callback.
 	 */
-	if (vbio->vbio_abd != NULL) {
+	ASSERT3P(zio->io_bio, ==, NULL);
-		void *buf = abd_to_buf(vbio->vbio_abd);
+	zio->io_bio = vbio;
-		abd_free(vbio->vbio_abd);
-		vbio->vbio_abd = NULL;
 
-		if (zio->io_type == ZIO_TYPE_READ)
-			abd_return_buf_copy(zio->io_abd, buf, zio->io_size);
-		else
-			abd_return_buf(zio->io_abd, buf, zio->io_size);
-	}
-
-	/* Final cleanup */
-	kmem_free(vbio, sizeof (vbio_t));
-
-	/* All done, submit for processing */
 	zio_delay_interrupt(zio);
 }
 
@@ -834,38 +823,61 @@ vbio_completion(struct bio *bio)
  * NOTE: if you change this function, change the copy in
  * tests/zfs-tests/tests/functional/vdev_disk/page_alignment.c, and add test
  * data there to validate the change you're making.
- *
  */
 typedef struct {
-	uint_t  bmask;
+	size_t	blocksize;
-	uint_t  npages;
+	int	seen_first;
-	uint_t  end;
+	int	seen_last;
-} vdev_disk_check_pages_t;
+} vdev_disk_check_alignment_t;
 
 static int
-vdev_disk_check_pages_cb(struct page *page, size_t off, size_t len, void *priv)
+vdev_disk_check_alignment_cb(struct page *page, size_t off, size_t len,
+    void *priv)
 {
 	(void) page;
-	vdev_disk_check_pages_t *s = priv;
+	vdev_disk_check_alignment_t *s = priv;
 
 	/*
-	 * If we didn't finish on a block size boundary last time, then there
+	 * The cardinal rule: a single on-disk block must never cross an
-	 * would be a gap if we tried to use this ABD as-is, so abort.
+	 * physical (order-0) page boundary, as the kernel expects to be able
+	 * to split at both LBS and page boundaries.
+	 *
+	 * This implies various alignment rules for the blocks in this
+	 * (possibly compound) page, which we can check for.
 	 */
-	if (s->end != 0)
+
+	/*
+	 * If the previous page did not end on a page boundary, then we
+	 * can't proceed without creating a hole.
+	 */
+	if (s->seen_last)
+		return (1);
+
+	/* This page must contain only whole LBS-sized blocks. */
+	if (!IS_P2ALIGNED(len, s->blocksize))
 		return (1);
 
 	/*
-	 * Note if we're taking less than a full block, so we can check it
+	 * If this is not the first page in the ABD, then the data must start
-	 * above on the next call.
+	 * on a page-aligned boundary (so the kernel can split on page
+	 * boundaries without having to deal with a hole). If it is, then
+	 * it can start on LBS-alignment.
 	 */
-	s->end = (off+len) & s->bmask;
+	if (s->seen_first) {
+		if (!IS_P2ALIGNED(off, PAGESIZE))
+			return (1);
+	} else {
+		if (!IS_P2ALIGNED(off, s->blocksize))
+			return (1);
+		s->seen_first = 1;
+	}
 
-	/* All blocks after the first must start on a block size boundary. */
+	/*
-	if (s->npages != 0 && (off & s->bmask) != 0)
+	 * If this data does not end on a page-aligned boundary, then this
-		return (1);
+	 * must be the last page in the ABD, for the same reason.
+	 */
+	s->seen_last = !IS_P2ALIGNED(off+len, PAGESIZE);
 
-	s->npages++;
 	return (0);
 }
 
@@ -874,15 +886,14 @@ vdev_disk_check_pages_cb(struct page *page, size_t off, size_t len, void *priv)
  * the number of pages, or 0 if it can't be submitted like this.
  */
 static boolean_t
-vdev_disk_check_pages(abd_t *abd, uint64_t size, struct block_device *bdev)
+vdev_disk_check_alignment(abd_t *abd, uint64_t size, struct block_device *bdev)
 {
-	vdev_disk_check_pages_t s = {
+	vdev_disk_check_alignment_t s = {
-	    .bmask = bdev_logical_block_size(bdev)-1,
+	    .blocksize = bdev_logical_block_size(bdev),
-	    .npages = 0,
-	    .end = 0,
 	};
 
-	if (abd_iterate_page_func(abd, 0, size, vdev_disk_check_pages_cb, &s))
+	if (abd_iterate_page_func(abd, 0, size,
+	    vdev_disk_check_alignment_cb, &s))
 		return (B_FALSE);
 
 	return (B_TRUE);
@@ -916,37 +927,32 @@ vdev_disk_io_rw(zio_t *zio)
 
 	/*
 	 * Check alignment of the incoming ABD. If any part of it would require
-	 * submitting a page that is not aligned to the logical block size,
+	 * submitting a page that is not aligned to both the logical block size
-	 * then we take a copy into a linear buffer and submit that instead.
+	 * and the page size, then we take a copy into a new memory region with
-	 * This should be impossible on a 512b LBS, and fairly rare on 4K,
+	 * correct alignment.  This should be impossible on a 512b LBS. On
-	 * usually requiring abnormally-small data blocks (eg gang blocks)
+	 * larger blocks, this can happen at least when a small number of
-	 * mixed into the same ABD as larger ones (eg aggregated).
+	 * blocks (usually 1) are allocated from a shared slab, or when
+	 * abnormally-small data regions (eg gang headers) are mixed into the
+	 * same ABD as larger allocations (eg aggregations).
 	 */
 	abd_t *abd = zio->io_abd;
-	if (!vdev_disk_check_pages(abd, zio->io_size, bdev)) {
+	if (!vdev_disk_check_alignment(abd, zio->io_size, bdev)) {
-		void *buf;
+		/* Allocate a new memory region with guaranteed alignment */
-		if (zio->io_type == ZIO_TYPE_READ)
+		abd = abd_alloc_for_io(zio->io_size,
-			buf = abd_borrow_buf(zio->io_abd, zio->io_size);
+		    zio->io_abd->abd_flags & ABD_FLAG_META);
-		else
+
-			buf = abd_borrow_buf_copy(zio->io_abd, zio->io_size);
+		/* If we're writing copy our data into it */
+		if (zio->io_type == ZIO_TYPE_WRITE)
+			abd_copy(abd, zio->io_abd, zio->io_size);
 
 		/*
-		 * Wrap the copy in an abd_t, so we can use the same iterators
+		 * False here would mean the new allocation has an invalid
-		 * to count and fill the vbio later.
+		 * alignment too, which would mean that abd_alloc() is not
-		 */
+		 * guaranteeing this, or our logic in
-		abd = abd_get_from_buf(buf, zio->io_size);
+		 * vdev_disk_check_alignment() is wrong. In either case,
-
-		/*
-		 * False here would mean the borrowed copy has an invalid
-		 * alignment too, which would mean we've somehow been passed a
-		 * linear ABD with an interior page that has a non-zero offset
-		 * or a size not a multiple of PAGE_SIZE. This is not possible.
-		 * It would mean either zio_buf_alloc() or its underlying
-		 * allocators have done something extremely strange, or our
-		 * math in vdev_disk_check_pages() is wrong. In either case,
 		 * something in seriously wrong and its not safe to continue.
 		 */
-		VERIFY(vdev_disk_check_pages(abd, zio->io_size, bdev));
+		VERIFY(vdev_disk_check_alignment(abd, zio->io_size, bdev));
 	}
 
 	/* Allocate vbio, with a pointer to the borrowed ABD if necessary */
@@ -1437,6 +1443,28 @@ vdev_disk_io_start(zio_t *zio)
 static void
 vdev_disk_io_done(zio_t *zio)
 {
+	/* If this was a read or write, we need to clean up the vbio */
+	if (zio->io_bio != NULL) {
+		vbio_t *vbio = zio->io_bio;
+		zio->io_bio = NULL;
+
+		/*
+		 * If we copied the ABD before issuing it, clean up and return
+		 * the copy to the ADB, with changes if appropriate.
+		 */
+		if (vbio->vbio_abd != NULL) {
+			if (zio->io_type == ZIO_TYPE_READ)
+				abd_copy(zio->io_abd, vbio->vbio_abd,
+				    zio->io_size);
+
+			abd_free(vbio->vbio_abd);
+			vbio->vbio_abd = NULL;
+		}
+
+		/* Final cleanup */
+		kmem_free(vbio, sizeof (vbio_t));
+	}
+
 	/*
 	 * If the device returned EIO, we revalidate the media.  If it is
 	 * determined the media has changed this triggers the asynchronous

module/os/linux/zfs/vdev_file.c

@@ -33,11 +33,13 @@
 #include <sys/fs/zfs.h>
 #include <sys/fm/fs/zfs.h>
 #include <sys/abd.h>
-#include <sys/fcntl.h>
 #include <sys/vnode.h>
 #include <sys/zfs_file.h>
 #ifdef _KERNEL
 #include <linux/falloc.h>
+#include <sys/fcntl.h>
+#else
+#include <fcntl.h>
 #endif
 /*
  * Virtual device vector for files.

module/os/linux/zfs/zfs_ctldir.c

@@ -767,9 +767,6 @@ zfsctl_snapshot_path_objset(zfsvfs_t *zfsvfs, uint64_t objsetid,
 	uint64_t id, pos = 0;
 	int error = 0;
 
-	if (zfsvfs->z_vfs->vfs_mntpoint == NULL)
-		return (SET_ERROR(ENOENT));
-
 	cookie = spl_fstrans_mark();
 	snapname = kmem_alloc(ZFS_MAX_DATASET_NAME_LEN, KM_SLEEP);
 
@@ -786,8 +783,14 @@ zfsctl_snapshot_path_objset(zfsvfs_t *zfsvfs, uint64_t objsetid,
 			break;
 	}
 
-	snprintf(full_path, path_len, "%s/.zfs/snapshot/%s",
+	mutex_enter(&zfsvfs->z_vfs->vfs_mntpt_lock);
-	    zfsvfs->z_vfs->vfs_mntpoint, snapname);
+	if (zfsvfs->z_vfs->vfs_mntpoint != NULL) {
+		snprintf(full_path, path_len, "%s/.zfs/snapshot/%s",
+		    zfsvfs->z_vfs->vfs_mntpoint, snapname);
+	} else
+		error = SET_ERROR(ENOENT);
+	mutex_exit(&zfsvfs->z_vfs->vfs_mntpt_lock);
+
 out:
 	kmem_free(snapname, ZFS_MAX_DATASET_NAME_LEN);
 	spl_fstrans_unmark(cookie);
@@ -1049,6 +1052,66 @@ exportfs_flush(void)
 	(void) call_usermodehelper(argv[0], argv, envp, UMH_WAIT_PROC);
 }
 
+/*
+ * Returns the path in char format for given struct path. Uses
+ * d_path exported by kernel to convert struct path to char
+ * format. Returns the correct path for mountpoints and chroot
+ * environments.
+ *
+ * If chroot environment has directories that are mounted with
+ * --bind or --rbind flag, d_path returns the complete path inside
+ * chroot environment but does not return the absolute path, i.e.
+ * the path to chroot environment is missing.
+ */
+static int
+get_root_path(struct path *path, char *buff, int len)
+{
+	char *path_buffer, *path_ptr;
+	int error = 0;
+
+	path_get(path);
+	path_buffer = kmem_zalloc(len, KM_SLEEP);
+	path_ptr = d_path(path, path_buffer, len);
+	if (IS_ERR(path_ptr))
+		error = SET_ERROR(-PTR_ERR(path_ptr));
+	else
+		strcpy(buff, path_ptr);
+
+	kmem_free(path_buffer, len);
+	path_put(path);
+	return (error);
+}
+
+/*
+ * Returns if the current process root is chrooted or not. Linux
+ * kernel exposes the task_struct for current process and init.
+ * Since init process root points to actual root filesystem when
+ * Linux runtime is reached, we can compare the current process
+ * root with init process root to determine if root of the current
+ * process is different from init, which can reliably determine if
+ * current process is in chroot context or not.
+ */
+static int
+is_current_chrooted(void)
+{
+	struct task_struct *curr = current, *global = &init_task;
+	struct path cr_root, gl_root;
+
+	task_lock(curr);
+	get_fs_root(curr->fs, &cr_root);
+	task_unlock(curr);
+
+	task_lock(global);
+	get_fs_root(global->fs, &gl_root);
+	task_unlock(global);
+
+	int chrooted = !path_equal(&cr_root, &gl_root);
+	path_put(&gl_root);
+	path_put(&cr_root);
+
+	return (chrooted);
+}
+
 /*
  * Attempt to unmount a snapshot by making a call to user space.
  * There is no assurance that this can or will succeed, is just a
@@ -1123,14 +1186,50 @@ zfsctl_snapshot_mount(struct path *path, int flags)
 	if (error)
 		goto error;
 
+	if (is_current_chrooted() == 0) {
+		/*
+		 * Current process is not in chroot context
+		 */
+
+		char *m = kmem_zalloc(MAXPATHLEN, KM_SLEEP);
+		struct path mnt_path;
+		mnt_path.mnt = path->mnt;
+		mnt_path.dentry = path->mnt->mnt_root;
+
+		/*
+		 * Get path to current mountpoint
+		 */
+		error = get_root_path(&mnt_path, m, MAXPATHLEN);
+		if (error != 0) {
+			kmem_free(m, MAXPATHLEN);
+			goto error;
+		}
+		mutex_enter(&zfsvfs->z_vfs->vfs_mntpt_lock);
+		if (zfsvfs->z_vfs->vfs_mntpoint != NULL) {
+			/*
+			 * If current mnountpoint and vfs_mntpoint are not same,
+			 * store current mountpoint in vfs_mntpoint.
+			 */
+			if (strcmp(zfsvfs->z_vfs->vfs_mntpoint, m) != 0) {
+				kmem_strfree(zfsvfs->z_vfs->vfs_mntpoint);
+				zfsvfs->z_vfs->vfs_mntpoint = kmem_strdup(m);
+			}
+		} else
+			zfsvfs->z_vfs->vfs_mntpoint = kmem_strdup(m);
+		mutex_exit(&zfsvfs->z_vfs->vfs_mntpt_lock);
+		kmem_free(m, MAXPATHLEN);
+	}
+
 	/*
 	 * Construct a mount point path from sb of the ctldir inode and dirent
 	 * name, instead of from d_path(), so that chroot'd process doesn't fail
 	 * on mount.zfs(8).
 	 */
+	mutex_enter(&zfsvfs->z_vfs->vfs_mntpt_lock);
 	snprintf(full_path, MAXPATHLEN, "%s/.zfs/snapshot/%s",
 	    zfsvfs->z_vfs->vfs_mntpoint ? zfsvfs->z_vfs->vfs_mntpoint : "",
 	    dname(dentry));
+	mutex_exit(&zfsvfs->z_vfs->vfs_mntpt_lock);
 
 	snprintf(options, 7, "%s",
 	    zfs_snapshot_no_setuid ? "nosuid" : "suid");

module/os/linux/zfs/zfs_vfsops.c

@@ -115,7 +115,7 @@ zfsvfs_vfs_free(vfs_t *vfsp)
 	if (vfsp != NULL) {
 		if (vfsp->vfs_mntpoint != NULL)
 			kmem_strfree(vfsp->vfs_mntpoint);
-
+		mutex_destroy(&vfsp->vfs_mntpt_lock);
 		kmem_free(vfsp, sizeof (vfs_t));
 	}
 }
@@ -197,10 +197,11 @@ zfsvfs_parse_option(char *option, int token, substring_t *args, vfs_t *vfsp)
 		vfsp->vfs_do_nbmand = B_TRUE;
 		break;
 	case TOKEN_MNTPOINT:
+		if (vfsp->vfs_mntpoint != NULL)
+			kmem_strfree(vfsp->vfs_mntpoint);
 		vfsp->vfs_mntpoint = match_strdup(&args[0]);
 		if (vfsp->vfs_mntpoint == NULL)
 			return (SET_ERROR(ENOMEM));
-
 		break;
 	default:
 		break;
@@ -219,6 +220,7 @@ zfsvfs_parse_options(char *mntopts, vfs_t **vfsp)
 	int error;
 
 	tmp_vfsp = kmem_zalloc(sizeof (vfs_t), KM_SLEEP);
+	mutex_init(&tmp_vfsp->vfs_mntpt_lock, NULL, MUTEX_DEFAULT, NULL);
 
 	if (mntopts != NULL) {
 		substring_t args[MAX_OPT_ARGS];

module/os/linux/zfs/zfs_vnops_os.c

@@ -260,15 +260,6 @@ update_pages(znode_t *zp, int64_t start, int len, objset_t *os)
 			} else {
 				ClearPageError(pp);
 				SetPageUptodate(pp);
-				if (!PagePrivate(pp)) {
-					/*
-					 * Set private bit so page migration
-					 * will wait for us to finish writeback
-					 * before calling migrate_folio().
-					 */
-					SetPagePrivate(pp);
-					get_page(pp);
-				}
 
 				if (mapping_writably_mapped(mp))
 					flush_dcache_page(pp);
@@ -4090,14 +4081,6 @@ zfs_fillpage(struct inode *ip, struct page *pp)
 	} else {
 		ClearPageError(pp);
 		SetPageUptodate(pp);
-		if (!PagePrivate(pp)) {
-			/*
-			 * Set private bit so page migration will wait for us to
-			 * finish writeback before calling migrate_folio().
-			 */
-			SetPagePrivate(pp);
-			get_page(pp);
-		}
 	}
 
 	return (error);

module/os/linux/zfs/zfs_znode_os.c

@@ -1577,14 +1577,6 @@ zfs_zero_partial_page(znode_t *zp, uint64_t start, uint64_t len)
 		mark_page_accessed(pp);
 		SetPageUptodate(pp);
 		ClearPageError(pp);
-		if (!PagePrivate(pp)) {
-			/*
-			 * Set private bit so page migration will wait for us to
-			 * finish writeback before calling migrate_folio().
-			 */
-			SetPagePrivate(pp);
-			get_page(pp);
-		}
 		unlock_page(pp);
 		put_page(pp);
 	}

module/os/linux/zfs/zpl_file.c

@@ -28,6 +28,7 @@
 #include <linux/compat.h>
 #endif
 #include <linux/fs.h>
+#include <linux/migrate.h>
 #include <sys/file.h>
 #include <sys/dmu_objset.h>
 #include <sys/zfs_znode.h>
@@ -607,42 +608,6 @@ zpl_writepage(struct page *pp, struct writeback_control *wbc)
 	return (zpl_putpage(pp, wbc, &for_sync));
 }
 
-static int
-zpl_releasepage(struct page *pp, gfp_t gfp)
-{
-	if (PagePrivate(pp)) {
-		ClearPagePrivate(pp);
-		put_page(pp);
-	}
-	return (1);
-}
-
-#ifdef HAVE_VFS_RELEASE_FOLIO
-static bool
-zpl_release_folio(struct folio *folio, gfp_t gfp)
-{
-	return (zpl_releasepage(&folio->page, gfp));
-}
-#endif
-
-#ifdef HAVE_VFS_INVALIDATE_FOLIO
-static void
-zpl_invalidate_folio(struct folio *folio, size_t offset, size_t len)
-{
-	if ((offset == 0) && (len == PAGE_SIZE)) {
-		zpl_releasepage(&folio->page, 0);
-	}
-}
-#else
-static void
-zpl_invalidatepage(struct page *pp, unsigned int offset, unsigned int len)
-{
-	if ((offset == 0) && (len == PAGE_SIZE)) {
-		zpl_releasepage(pp, 0);
-	}
-}
-#endif
-
 /*
  * The flag combination which matches the behavior of zfs_space() is
  * FALLOC_FL_KEEP_SIZE | FALLOC_FL_PUNCH_HOLE.  The FALLOC_FL_PUNCH_HOLE
@@ -1126,15 +1091,10 @@ const struct address_space_operations zpl_address_space_operations = {
 #ifdef HAVE_VFS_FILEMAP_DIRTY_FOLIO
 	.dirty_folio	= filemap_dirty_folio,
 #endif
-#ifdef HAVE_VFS_RELEASE_FOLIO
+#ifdef HAVE_VFS_MIGRATE_FOLIO
-	.release_folio	= zpl_release_folio,
+	.migrate_folio	= migrate_folio,
 #else
-	.releasepage	= zpl_releasepage,
+	.migratepage	= migrate_page,
-#endif
-#ifdef HAVE_VFS_INVALIDATE_FOLIO
-	.invalidate_folio = zpl_invalidate_folio,
-#else
-	.invalidatepage = zpl_invalidatepage,
 #endif
 };
 

module/zcommon/zfs_valstr.c

@@ -206,6 +206,7 @@ _VALSTR_BITFIELD_IMPL(zio_flag,
 	{ '.', "PR", "PROBE" },
 	{ '.', "TH", "TRYHARD" },
 	{ '.', "OP", "OPTIONAL" },
+	{ '.', "RD", "DIO_READ" },
 	{ '.', "DQ", "DONT_QUEUE" },
 	{ '.', "DP", "DONT_PROPAGATE" },
 	{ '.', "BY", "IO_BYPASS" },

module/zfs/dbuf.c

@@ -182,6 +182,7 @@ static void dbuf_sync_leaf_verify_bonus_dnode(dbuf_dirty_record_t *dr);
  * Global data structures and functions for the dbuf cache.
  */
 static kmem_cache_t *dbuf_kmem_cache;
+kmem_cache_t *dbuf_dirty_kmem_cache;
 static taskq_t *dbu_evict_taskq;
 
 static kthread_t *dbuf_cache_evict_thread;
@@ -966,6 +967,8 @@ dbuf_init(void)
 	dbuf_kmem_cache = kmem_cache_create("dmu_buf_impl_t",
 	    sizeof (dmu_buf_impl_t),
 	    0, dbuf_cons, dbuf_dest, NULL, NULL, NULL, 0);
+	dbuf_dirty_kmem_cache = kmem_cache_create("dbuf_dirty_record_t",
+	    sizeof (dbuf_dirty_record_t), 0, NULL, NULL, NULL, NULL, NULL, 0);
 
 	for (int i = 0; i < hmsize; i++)
 		mutex_init(&h->hash_mutexes[i], NULL, MUTEX_NOLOCKDEP, NULL);
@@ -1041,6 +1044,7 @@ dbuf_fini(void)
 	    sizeof (kmutex_t));
 
 	kmem_cache_destroy(dbuf_kmem_cache);
+	kmem_cache_destroy(dbuf_dirty_kmem_cache);
 	taskq_destroy(dbu_evict_taskq);
 
 	mutex_enter(&dbuf_evict_lock);
@@ -2343,7 +2347,8 @@ dbuf_dirty(dmu_buf_impl_t *db, dmu_tx_t *tx)
 	 * to make a copy of it so that the changes we make in this
 	 * transaction group won't leak out when we sync the older txg.
 	 */
-	dr = kmem_zalloc(sizeof (dbuf_dirty_record_t), KM_SLEEP);
+	dr = kmem_cache_alloc(dbuf_dirty_kmem_cache, KM_SLEEP);
+	memset(dr, 0, sizeof (*dr));
 	list_link_init(&dr->dr_dirty_node);
 	list_link_init(&dr->dr_dbuf_node);
 	dr->dr_dnode = dn;
@@ -2526,7 +2531,7 @@ dbuf_undirty_bonus(dbuf_dirty_record_t *dr)
 		mutex_destroy(&dr->dt.di.dr_mtx);
 		list_destroy(&dr->dt.di.dr_children);
 	}
-	kmem_free(dr, sizeof (dbuf_dirty_record_t));
+	kmem_cache_free(dbuf_dirty_kmem_cache, dr);
 	ASSERT3U(db->db_dirtycnt, >, 0);
 	db->db_dirtycnt -= 1;
 }
@@ -2616,7 +2621,7 @@ dbuf_undirty(dmu_buf_impl_t *db, dmu_tx_t *tx)
 		}
 	}
 
-	kmem_free(dr, sizeof (dbuf_dirty_record_t));
+	kmem_cache_free(dbuf_dirty_kmem_cache, dr);
 
 	ASSERT(db->db_dirtycnt > 0);
 	db->db_dirtycnt -= 1;
@@ -2941,7 +2946,7 @@ dmu_buf_set_crypt_params(dmu_buf_t *db_fake, boolean_t byteorder,
 	 * (see dbuf_sync_dnode_leaf_crypt()).
 	 */
 	ASSERT3U(db->db.db_object, ==, DMU_META_DNODE_OBJECT);
-	ASSERT3U(db->db_level, ==, 0);
+	ASSERT0(db->db_level);
 	ASSERT(db->db_objset->os_raw_receive);
 
 	dmu_buf_will_dirty_impl(db_fake,
@@ -2950,6 +2955,7 @@ dmu_buf_set_crypt_params(dmu_buf_t *db_fake, boolean_t byteorder,
 	dr = dbuf_find_dirty_eq(db, tx->tx_txg);
 
 	ASSERT3P(dr, !=, NULL);
+	ASSERT3U(dr->dt.dl.dr_override_state, ==, DR_NOT_OVERRIDDEN);
 
 	dr->dt.dl.dr_has_raw_params = B_TRUE;
 	dr->dt.dl.dr_byteorder = byteorder;
@@ -2964,10 +2970,14 @@ dbuf_override_impl(dmu_buf_impl_t *db, const blkptr_t *bp, dmu_tx_t *tx)
 	struct dirty_leaf *dl;
 	dbuf_dirty_record_t *dr;
 
+	ASSERT3U(db->db.db_object, !=, DMU_META_DNODE_OBJECT);
+	ASSERT0(db->db_level);
+
 	dr = list_head(&db->db_dirty_records);
 	ASSERT3P(dr, !=, NULL);
 	ASSERT3U(dr->dr_txg, ==, tx->tx_txg);
 	dl = &dr->dt.dl;
+	ASSERT0(dl->dr_has_raw_params);
 	dl->dr_overridden_by = *bp;
 	dl->dr_override_state = DR_OVERRIDDEN;
 	BP_SET_LOGICAL_BIRTH(&dl->dr_overridden_by, dr->dr_txg);
@@ -3040,6 +3050,7 @@ dmu_buf_write_embedded(dmu_buf_t *dbuf, void *data,
 	ASSERT3P(dr, !=, NULL);
 	ASSERT3U(dr->dr_txg, ==, tx->tx_txg);
 	dl = &dr->dt.dl;
+	ASSERT0(dl->dr_has_raw_params);
 	encode_embedded_bp_compressed(&dl->dr_overridden_by,
 	    data, comp, uncompressed_size, compressed_size);
 	BPE_SET_ETYPE(&dl->dr_overridden_by, etype);
@@ -5083,7 +5094,7 @@ dbuf_write_done(zio_t *zio, arc_buf_t *buf, void *vdb)
 	dsl_pool_undirty_space(dmu_objset_pool(os), dr->dr_accounted,
 	    zio->io_txg);
 
-	kmem_free(dr, sizeof (dbuf_dirty_record_t));
+	kmem_cache_free(dbuf_dirty_kmem_cache, dr);
 }
 
 static void

module/zfs/dmu.c

@@ -1895,6 +1895,7 @@ dmu_sync_done(zio_t *zio, arc_buf_t *buf, void *varg)
 	mutex_enter(&db->db_mtx);
 	ASSERT(dr->dt.dl.dr_override_state == DR_IN_DMU_SYNC);
 	if (zio->io_error == 0) {
+		ASSERT0(dr->dt.dl.dr_has_raw_params);
 		dr->dt.dl.dr_nopwrite = !!(zio->io_flags & ZIO_FLAG_NOPWRITE);
 		if (dr->dt.dl.dr_nopwrite) {
 			blkptr_t *bp = zio->io_bp;
@@ -2190,6 +2191,7 @@ dmu_sync(zio_t *pio, uint64_t txg, dmu_sync_cb_t *done, zgd_t *zgd)
 		return (SET_ERROR(EALREADY));
 	}
 
+	ASSERT0(dr->dt.dl.dr_has_raw_params);
 	ASSERT(dr->dt.dl.dr_override_state == DR_NOT_OVERRIDDEN);
 	dr->dt.dl.dr_override_state = DR_IN_DMU_SYNC;
 	mutex_exit(&db->db_mtx);
@@ -2657,6 +2659,7 @@ dmu_brt_clone(objset_t *os, uint64_t object, uint64_t offset, uint64_t length,
 		db = (dmu_buf_impl_t *)dbuf;
 		bp = &bps[i];
 
+		ASSERT3U(db->db.db_object, !=, DMU_META_DNODE_OBJECT);
 		ASSERT0(db->db_level);
 		ASSERT(db->db_blkid != DMU_BONUS_BLKID);
 		ASSERT(db->db_blkid != DMU_SPILL_BLKID);
@@ -2672,11 +2675,6 @@ dmu_brt_clone(objset_t *os, uint64_t object, uint64_t offset, uint64_t length,
 		db = (dmu_buf_impl_t *)dbuf;
 		bp = &bps[i];
 
-		ASSERT0(db->db_level);
-		ASSERT(db->db_blkid != DMU_BONUS_BLKID);
-		ASSERT(db->db_blkid != DMU_SPILL_BLKID);
-		ASSERT(BP_IS_HOLE(bp) || dbuf->db_size == BP_GET_LSIZE(bp));
-
 		dmu_buf_will_clone_or_dio(dbuf, tx);
 
 		mutex_enter(&db->db_mtx);
@@ -2685,6 +2683,7 @@ dmu_brt_clone(objset_t *os, uint64_t object, uint64_t offset, uint64_t length,
 		VERIFY(dr != NULL);
 		ASSERT3U(dr->dr_txg, ==, tx->tx_txg);
 		dl = &dr->dt.dl;
+		ASSERT0(dl->dr_has_raw_params);
 		dl->dr_overridden_by = *bp;
 		if (!BP_IS_HOLE(bp) || BP_GET_LOGICAL_BIRTH(bp) != 0) {
 			if (!BP_IS_EMBEDDED(bp)) {

module/zfs/dmu_direct.c

@@ -180,6 +180,7 @@ dmu_write_direct(zio_t *pio, dmu_buf_impl_t *db, abd_t *data, dmu_tx_t *tx)
 	if (list_next(&db->db_dirty_records, dr_head) != NULL)
 		zp.zp_nopwrite = B_FALSE;
 
+	ASSERT0(dr_head->dt.dl.dr_has_raw_params);
 	ASSERT3S(dr_head->dt.dl.dr_override_state, ==, DR_NOT_OVERRIDDEN);
 	dr_head->dt.dl.dr_override_state = DR_IN_DMU_SYNC;
 
@@ -330,7 +331,7 @@ dmu_read_abd(dnode_t *dn, uint64_t offset, uint64_t size,
 		 */
 		zio_t *cio = zio_read(rio, spa, bp, mbuf, db->db.db_size,
 		    dmu_read_abd_done, NULL, ZIO_PRIORITY_SYNC_READ,
-		    ZIO_FLAG_CANFAIL, &zb);
+		    ZIO_FLAG_CANFAIL | ZIO_FLAG_DIO_READ, &zb);
 		mutex_exit(&db->db_mtx);
 
 		zfs_racct_read(spa, db->db.db_size, 1, flags);

module/zfs/dmu_send.c

@@ -180,6 +180,8 @@ struct send_range {
 			 */
 			dnode_phys_t		*dnp;
 			blkptr_t		bp;
+			/* Piggyback unmodified spill block */
+			struct send_range	*spill_range;
 		} object;
 		struct srr {
 			uint32_t		datablksz;
@@ -231,6 +233,8 @@ range_free(struct send_range *range)
 		size_t size = sizeof (dnode_phys_t) *
 		    (range->sru.object.dnp->dn_extra_slots + 1);
 		kmem_free(range->sru.object.dnp, size);
+		if (range->sru.object.spill_range)
+			range_free(range->sru.object.spill_range);
 	} else if (range->type == DATA) {
 		mutex_enter(&range->sru.data.lock);
 		while (range->sru.data.io_outstanding)
@@ -617,7 +621,7 @@ dump_spill(dmu_send_cookie_t *dscp, const blkptr_t *bp, uint64_t object,
 	drrs->drr_length = blksz;
 	drrs->drr_toguid = dscp->dsc_toguid;
 
-	/* See comment in dump_dnode() for full details */
+	/* See comment in piggyback_unmodified_spill() for full details */
 	if (zfs_send_unmodified_spill_blocks &&
 	    (BP_GET_LOGICAL_BIRTH(bp) <= dscp->dsc_fromtxg)) {
 		drrs->drr_flags |= DRR_SPILL_UNMODIFIED;
@@ -793,35 +797,6 @@ dump_dnode(dmu_send_cookie_t *dscp, const blkptr_t *bp, uint64_t object,
 	    (dnp->dn_datablkszsec << SPA_MINBLOCKSHIFT), DMU_OBJECT_END) != 0)
 		return (SET_ERROR(EINTR));
 
-	/*
-	 * Send DRR_SPILL records for unmodified spill blocks.	This is useful
-	 * because changing certain attributes of the object (e.g. blocksize)
-	 * can cause old versions of ZFS to incorrectly remove a spill block.
-	 * Including these records in the stream forces an up to date version
-	 * to always be written ensuring they're never lost.  Current versions
-	 * of the code which understand the DRR_FLAG_SPILL_BLOCK feature can
-	 * ignore these unmodified spill blocks.
-	 */
-	if (zfs_send_unmodified_spill_blocks &&
-	    (dnp->dn_flags & DNODE_FLAG_SPILL_BLKPTR) &&
-	    (BP_GET_LOGICAL_BIRTH(DN_SPILL_BLKPTR(dnp)) <= dscp->dsc_fromtxg)) {
-		struct send_range record;
-		blkptr_t *bp = DN_SPILL_BLKPTR(dnp);
-
-		memset(&record, 0, sizeof (struct send_range));
-		record.type = DATA;
-		record.object = object;
-		record.eos_marker = B_FALSE;
-		record.start_blkid = DMU_SPILL_BLKID;
-		record.end_blkid = record.start_blkid + 1;
-		record.sru.data.bp = *bp;
-		record.sru.data.obj_type = dnp->dn_type;
-		record.sru.data.datablksz = BP_GET_LSIZE(bp);
-
-		if (do_dump(dscp, &record) != 0)
-			return (SET_ERROR(EINTR));
-	}
-
 	if (dscp->dsc_err != 0)
 		return (SET_ERROR(EINTR));
 
@@ -911,6 +886,9 @@ do_dump(dmu_send_cookie_t *dscp, struct send_range *range)
 	case OBJECT:
 		err = dump_dnode(dscp, &range->sru.object.bp, range->object,
 		    range->sru.object.dnp);
+		/* Dump piggybacked unmodified spill block */
+		if (!err && range->sru.object.spill_range)
+			err = do_dump(dscp, range->sru.object.spill_range);
 		return (err);
 	case OBJECT_RANGE: {
 		ASSERT3U(range->start_blkid + 1, ==, range->end_blkid);
@@ -939,34 +917,7 @@ do_dump(dmu_send_cookie_t *dscp, struct send_range *range)
 
 		ASSERT3U(srdp->datablksz, ==, BP_GET_LSIZE(bp));
 		ASSERT3U(range->start_blkid + 1, ==, range->end_blkid);
-		if (BP_GET_TYPE(bp) == DMU_OT_SA) {
-			arc_flags_t aflags = ARC_FLAG_WAIT;
-			zio_flag_t zioflags = ZIO_FLAG_CANFAIL;
 
-			if (dscp->dsc_featureflags & DMU_BACKUP_FEATURE_RAW) {
-				ASSERT(BP_IS_PROTECTED(bp));
-				zioflags |= ZIO_FLAG_RAW;
-			}
-
-			zbookmark_phys_t zb;
-			ASSERT3U(range->start_blkid, ==, DMU_SPILL_BLKID);
-			zb.zb_objset = dmu_objset_id(dscp->dsc_os);
-			zb.zb_object = range->object;
-			zb.zb_level = 0;
-			zb.zb_blkid = range->start_blkid;
-
-			arc_buf_t *abuf = NULL;
-			if (!dscp->dsc_dso->dso_dryrun && arc_read(NULL, spa,
-			    bp, arc_getbuf_func, &abuf, ZIO_PRIORITY_ASYNC_READ,
-			    zioflags, &aflags, &zb) != 0)
-				return (SET_ERROR(EIO));
-
-			err = dump_spill(dscp, bp, zb.zb_object,
-			    (abuf == NULL ? NULL : abuf->b_data));
-			if (abuf != NULL)
-				arc_buf_destroy(abuf, &abuf);
-			return (err);
-		}
 		if (send_do_embed(bp, dscp->dsc_featureflags)) {
 			err = dump_write_embedded(dscp, range->object,
 			    range->start_blkid * srdp->datablksz,
@@ -975,8 +926,9 @@ do_dump(dmu_send_cookie_t *dscp, struct send_range *range)
 		}
 		ASSERT(range->object > dscp->dsc_resume_object ||
 		    (range->object == dscp->dsc_resume_object &&
+		    (range->start_blkid == DMU_SPILL_BLKID ||
 		    range->start_blkid * srdp->datablksz >=
-		    dscp->dsc_resume_offset));
+		    dscp->dsc_resume_offset)));
 		/* it's a level-0 block of a regular object */
 
 		mutex_enter(&srdp->lock);
@@ -1006,8 +958,6 @@ do_dump(dmu_send_cookie_t *dscp, struct send_range *range)
 		ASSERT(dscp->dsc_dso->dso_dryrun ||
 		    srdp->abuf != NULL || srdp->abd != NULL);
 
-		uint64_t offset = range->start_blkid * srdp->datablksz;
-
 		char *data = NULL;
 		if (srdp->abd != NULL) {
 			data = abd_to_buf(srdp->abd);
@@ -1016,6 +966,14 @@ do_dump(dmu_send_cookie_t *dscp, struct send_range *range)
 			data = srdp->abuf->b_data;
 		}
 
+		if (BP_GET_TYPE(bp) == DMU_OT_SA) {
+			ASSERT3U(range->start_blkid, ==, DMU_SPILL_BLKID);
+			err = dump_spill(dscp, bp, range->object, data);
+			return (err);
+		}
+
+		uint64_t offset = range->start_blkid * srdp->datablksz;
+
 		/*
 		 * If we have large blocks stored on disk but the send flags
 		 * don't allow us to send large blocks, we split the data from
@@ -1098,6 +1056,8 @@ range_alloc(enum type type, uint64_t object, uint64_t start_blkid,
 		range->sru.data.io_outstanding = 0;
 		range->sru.data.io_err = 0;
 		range->sru.data.io_compressed = B_FALSE;
+	} else if (type == OBJECT) {
+		range->sru.object.spill_range = NULL;
 	}
 	return (range);
 }
@@ -1742,6 +1702,45 @@ enqueue_range(struct send_reader_thread_arg *srta, bqueue_t *q, dnode_t *dn,
 	bqueue_enqueue(q, range, datablksz);
 }
 
+/*
+ * Send DRR_SPILL records for unmodified spill blocks.	This is useful
+ * because changing certain attributes of the object (e.g. blocksize)
+ * can cause old versions of ZFS to incorrectly remove a spill block.
+ * Including these records in the stream forces an up to date version
+ * to always be written ensuring they're never lost.  Current versions
+ * of the code which understand the DRR_FLAG_SPILL_BLOCK feature can
+ * ignore these unmodified spill blocks.
+ *
+ * We piggyback the spill_range to dnode range instead of enqueueing it
+ * so send_range_after won't complain.
+ */
+static uint64_t
+piggyback_unmodified_spill(struct send_reader_thread_arg *srta,
+    struct send_range *range)
+{
+	ASSERT3U(range->type, ==, OBJECT);
+
+	dnode_phys_t *dnp = range->sru.object.dnp;
+	uint64_t fromtxg = srta->smta->to_arg->fromtxg;
+
+	if (!zfs_send_unmodified_spill_blocks ||
+	    !(dnp->dn_flags & DNODE_FLAG_SPILL_BLKPTR) ||
+	    !(BP_GET_LOGICAL_BIRTH(DN_SPILL_BLKPTR(dnp)) <= fromtxg))
+		return (0);
+
+	blkptr_t *bp = DN_SPILL_BLKPTR(dnp);
+	struct send_range *spill_range = range_alloc(DATA, range->object,
+	    DMU_SPILL_BLKID, DMU_SPILL_BLKID+1, B_FALSE);
+	spill_range->sru.data.bp = *bp;
+	spill_range->sru.data.obj_type = dnp->dn_type;
+	spill_range->sru.data.datablksz = BP_GET_LSIZE(bp);
+
+	issue_data_read(srta, spill_range);
+	range->sru.object.spill_range = spill_range;
+
+	return (BP_GET_LSIZE(bp));
+}
+
 /*
  * This thread is responsible for two things: First, it retrieves the correct
  * blkptr in the to ds if we need to send the data because of something from
@@ -1773,17 +1772,20 @@ send_reader_thread(void *arg)
 	uint64_t last_obj_exists = B_TRUE;
 	while (!range->eos_marker && !srta->cancel && smta->error == 0 &&
 	    err == 0) {
+		uint64_t spill = 0;
 		switch (range->type) {
 		case DATA:
 			issue_data_read(srta, range);
 			bqueue_enqueue(outq, range, range->sru.data.datablksz);
 			range = get_next_range_nofree(inq, range);
 			break;
-		case HOLE:
 		case OBJECT:
+			spill = piggyback_unmodified_spill(srta, range);
+			zfs_fallthrough;
+		case HOLE:
 		case OBJECT_RANGE:
 		case REDACT: // Redacted blocks must exist
-			bqueue_enqueue(outq, range, sizeof (*range));
+			bqueue_enqueue(outq, range, sizeof (*range) + spill);
 			range = get_next_range_nofree(inq, range);
 			break;
 		case PREVIOUSLY_REDACTED: {

module/zfs/dmu_tx.c

@@ -1377,6 +1377,13 @@ dmu_tx_pool(dmu_tx_t *tx)
 	return (tx->tx_pool);
 }
 
+/*
+ * Register a callback to be executed at the end of a TXG.
+ *
+ * Note: This currently exists for outside consumers, specifically the ZFS OSD
+ * for Lustre. Please do not remove before checking that project. For examples
+ * on how to use this see `ztest_commit_callback`.
+ */
 void
 dmu_tx_callback_register(dmu_tx_t *tx, dmu_tx_callback_func_t *func, void *data)
 {

module/zfs/dnode_sync.c

@@ -566,7 +566,7 @@ dnode_undirty_dbufs(list_t *list)
 			mutex_destroy(&dr->dt.di.dr_mtx);
 			list_destroy(&dr->dt.di.dr_children);
 		}
-		kmem_free(dr, sizeof (dbuf_dirty_record_t));
+		kmem_cache_free(dbuf_dirty_kmem_cache, dr);
 		dbuf_rele_and_unlock(db, (void *)(uintptr_t)txg, B_FALSE);
 	}
 }

module/zfs/dsl_dataset.c

@@ -2987,6 +2987,7 @@ dsl_dataset_rename_snapshot_sync_impl(dsl_pool_t *dp,
 	dsl_dataset_t *ds;
 	uint64_t val;
 	dmu_tx_t *tx = ddrsa->ddrsa_tx;
+	char *oldname, *newname;
 	int error;
 
 	error = dsl_dataset_snap_lookup(hds, ddrsa->ddrsa_oldsnapname, &val);
@@ -3011,8 +3012,14 @@ dsl_dataset_rename_snapshot_sync_impl(dsl_pool_t *dp,
 	VERIFY0(zap_add(dp->dp_meta_objset,
 	    dsl_dataset_phys(hds)->ds_snapnames_zapobj,
 	    ds->ds_snapname, 8, 1, &ds->ds_object, tx));
-	zvol_rename_minors(dp->dp_spa, ddrsa->ddrsa_oldsnapname,
+
-	    ddrsa->ddrsa_newsnapname, B_TRUE);
+	oldname = kmem_asprintf("%s@%s", ddrsa->ddrsa_fsname,
+	    ddrsa->ddrsa_oldsnapname);
+	newname = kmem_asprintf("%s@%s", ddrsa->ddrsa_fsname,
+	    ddrsa->ddrsa_newsnapname);
+	zvol_rename_minors(dp->dp_spa, oldname, newname, B_TRUE);
+	kmem_strfree(oldname);
+	kmem_strfree(newname);
 
 	dsl_dataset_rele(ds, FTAG);
 	return (0);

module/zfs/vdev.c

@@ -2205,10 +2205,11 @@ vdev_open(vdev_t *vd)
 		vd->vdev_max_asize = max_asize;
 
 		/*
-		 * If the vdev_ashift was not overridden at creation time,
+		 * If the vdev_ashift was not overridden at creation time
+		 * (0) or the override value is impossible for the device,
 		 * then set it the logical ashift and optimize the ashift.
 		 */
-		if (vd->vdev_ashift == 0) {
+		if (vd->vdev_ashift < vd->vdev_logical_ashift) {
 			vd->vdev_ashift = vd->vdev_logical_ashift;
 
 			if (vd->vdev_logical_ashift > ASHIFT_MAX) {

module/zfs/vdev_draid.c

@@ -1026,7 +1026,7 @@ vdev_draid_map_alloc_row(zio_t *zio, raidz_row_t **rrp, uint64_t io_offset,
 
 	ASSERT3U(vdc->vdc_nparity, >, 0);
 
-	raidz_row_t *rr = vdev_raidz_row_alloc(groupwidth);
+	raidz_row_t *rr = vdev_raidz_row_alloc(groupwidth, zio);
 	rr->rr_bigcols = bc;
 	rr->rr_firstdatacol = vdc->vdc_nparity;
 #ifdef ZFS_DEBUG

module/zfs/vdev_indirect.c

@@ -34,6 +34,7 @@
 #include <sys/zap.h>
 #include <sys/abd.h>
 #include <sys/zthr.h>
+#include <sys/fm/fs/zfs.h>
 
 /*
  * An indirect vdev corresponds to a vdev that has been removed.  Since
@@ -1832,6 +1833,19 @@ vdev_indirect_io_done(zio_t *zio)
 
 	zio_bad_cksum_t zbc;
 	int ret = zio_checksum_error(zio, &zbc);
+	/*
+	 * Any Direct I/O read that has a checksum error must be treated as
+	 * suspicious as the contents of the buffer could be getting
+	 * manipulated while the I/O is taking place. The checksum verify error
+	 * will be reported to the top-level VDEV.
+	 */
+	if (zio->io_flags & ZIO_FLAG_DIO_READ && ret == ECKSUM) {
+		zio->io_error = ret;
+		zio->io_flags |= ZIO_FLAG_DIO_CHKSUM_ERR;
+		zio_dio_chksum_verify_error_report(zio);
+		ret = 0;
+	}
+
 	if (ret == 0) {
 		zio_checksum_verified(zio);
 		return;

module/zfs/vdev_mirror.c

@@ -764,6 +764,27 @@ vdev_mirror_io_done(zio_t *zio)
 
 	ASSERT(zio->io_type == ZIO_TYPE_READ);
 
+	/*
+	 * Any Direct I/O read that has a checksum error must be treated as
+	 * suspicious as the contents of the buffer could be getting
+	 * manipulated while the I/O is taking place. The checksum verify error
+	 * will be reported to the top-level Mirror VDEV.
+	 *
+	 * There will be no attampt at reading any additional data copies. If
+	 * the buffer is still being manipulated while attempting to read from
+	 * another child, there exists a possibly that the checksum could be
+	 * verified as valid. However, the buffer contents could again get
+	 * manipulated after verifying the checksum. This would lead to bad data
+	 * being written out during self healing.
+	 */
+	if ((zio->io_flags & ZIO_FLAG_DIO_READ) &&
+	    (zio->io_flags & ZIO_FLAG_DIO_CHKSUM_ERR)) {
+		zio_dio_chksum_verify_error_report(zio);
+		zio->io_error = vdev_mirror_worst_error(mm);
+		ASSERT3U(zio->io_error, ==, ECKSUM);
+		return;
+	}
+
 	/*
 	 * If we don't have a good copy yet, keep trying other children.
 	 */

module/zfs/vdev_raidz.c

@@ -433,7 +433,7 @@ const zio_vsd_ops_t vdev_raidz_vsd_ops = {
 };
 
 raidz_row_t *
-vdev_raidz_row_alloc(int cols)
+vdev_raidz_row_alloc(int cols, zio_t *zio)
 {
 	raidz_row_t *rr =
 	    kmem_zalloc(offsetof(raidz_row_t, rr_col[cols]), KM_SLEEP);
@@ -445,7 +445,17 @@ vdev_raidz_row_alloc(int cols)
 		raidz_col_t *rc = &rr->rr_col[c];
 		rc->rc_shadow_devidx = INT_MAX;
 		rc->rc_shadow_offset = UINT64_MAX;
-		rc->rc_allow_repair = 1;
+		/*
+		 * We can not allow self healing to take place for Direct I/O
+		 * reads. There is nothing that stops the buffer contents from
+		 * being manipulated while the I/O is in flight. It is possible
+		 * that the checksum could be verified on the buffer and then
+		 * the contents of that buffer are manipulated afterwards. This
+		 * could lead to bad data being written out during self
+		 * healing.
+		 */
+		if (!(zio->io_flags & ZIO_FLAG_DIO_READ))
+			rc->rc_allow_repair = 1;
 	}
 	return (rr);
 }
@@ -619,7 +629,7 @@ vdev_raidz_map_alloc(zio_t *zio, uint64_t ashift, uint64_t dcols,
 	}
 
 	ASSERT3U(acols, <=, scols);
-	rr = vdev_raidz_row_alloc(scols);
+	rr = vdev_raidz_row_alloc(scols, zio);
 	rm->rm_row[0] = rr;
 	rr->rr_cols = acols;
 	rr->rr_bigcols = bc;
@@ -765,7 +775,7 @@ vdev_raidz_map_alloc_expanded(zio_t *zio,
 
 	for (uint64_t row = 0; row < rows; row++) {
 		boolean_t row_use_scratch = B_FALSE;
-		raidz_row_t *rr = vdev_raidz_row_alloc(cols);
+		raidz_row_t *rr = vdev_raidz_row_alloc(cols, zio);
 		rm->rm_row[row] = rr;
 
 		/* The starting RAIDZ (parent) vdev sector of the row. */
@@ -2633,6 +2643,20 @@ raidz_checksum_verify(zio_t *zio)
 	raidz_map_t *rm = zio->io_vsd;
 
 	int ret = zio_checksum_error(zio, &zbc);
+	/*
+	 * Any Direct I/O read that has a checksum error must be treated as
+	 * suspicious as the contents of the buffer could be getting
+	 * manipulated while the I/O is taking place. The checksum verify error
+	 * will be reported to the top-level RAIDZ VDEV.
+	 */
+	if (zio->io_flags & ZIO_FLAG_DIO_READ && ret == ECKSUM) {
+		zio->io_error = ret;
+		zio->io_flags |= ZIO_FLAG_DIO_CHKSUM_ERR;
+		zio_dio_chksum_verify_error_report(zio);
+		zio_checksum_verified(zio);
+		return (0);
+	}
+
 	if (ret != 0 && zbc.zbc_injected != 0)
 		rm->rm_ecksuminjected = 1;
 
@@ -2776,6 +2800,11 @@ vdev_raidz_io_done_verified(zio_t *zio, raidz_row_t *rr)
 			    (rc->rc_error == 0 || rc->rc_size == 0)) {
 				continue;
 			}
+			/*
+			 * We do not allow self healing for Direct I/O reads.
+			 * See comment in vdev_raid_row_alloc().
+			 */
+			ASSERT0(zio->io_flags & ZIO_FLAG_DIO_READ);
 
 			zfs_dbgmsg("zio=%px repairing c=%u devidx=%u "
 			    "offset=%llx",
@@ -2979,6 +3008,8 @@ raidz_reconstruct(zio_t *zio, int *ltgts, int ntgts, int nparity)
 
 	/* Check for success */
 	if (raidz_checksum_verify(zio) == 0) {
+		if (zio->io_flags & ZIO_FLAG_DIO_CHKSUM_ERR)
+			return (0);
 
 		/* Reconstruction succeeded - report errors */
 		for (int i = 0; i < rm->rm_nrows; i++) {
@@ -3379,7 +3410,6 @@ vdev_raidz_io_done_unrecoverable(zio_t *zio)
 			zio_bad_cksum_t zbc;
 			zbc.zbc_has_cksum = 0;
 			zbc.zbc_injected = rm->rm_ecksuminjected;
-
 			mutex_enter(&cvd->vdev_stat_lock);
 			cvd->vdev_stat.vs_checksum_errors++;
 			mutex_exit(&cvd->vdev_stat_lock);
@@ -3444,6 +3474,9 @@ vdev_raidz_io_done(zio_t *zio)
 		}
 
 		if (raidz_checksum_verify(zio) == 0) {
+			if (zio->io_flags & ZIO_FLAG_DIO_CHKSUM_ERR)
+				goto done;
+
 			for (int i = 0; i < rm->rm_nrows; i++) {
 				raidz_row_t *rr = rm->rm_row[i];
 				vdev_raidz_io_done_verified(zio, rr);
@@ -3538,6 +3571,7 @@ vdev_raidz_io_done(zio_t *zio)
 			}
 		}
 	}
+done:
 	if (rm->rm_lr != NULL) {
 		zfs_rangelock_exit(rm->rm_lr);
 		rm->rm_lr = NULL;

module/zfs/zfs_vnops.c

@@ -58,9 +58,9 @@
 #include <sys/zfs_znode.h>
 
 /*
- * Enable the experimental block cloning feature.  If this setting is 0, then
+ * Enables access to the block cloning feature. If this setting is 0, then even
- * even if feature@block_cloning is enabled, attempts to clone blocks will act
+ * if feature@block_cloning is enabled, using functions and system calls that
- * as though the feature is disabled.
+ * attempt to clone blocks will act as though the feature is disabled.
  */
 int zfs_bclone_enabled = 1;
 
@@ -303,6 +303,7 @@ zfs_read(struct znode *zp, zfs_uio_t *uio, int ioflag, cred_t *cr)
 	(void) cr;
 	int error = 0;
 	boolean_t frsync = B_FALSE;
+	boolean_t dio_checksum_failure = B_FALSE;
 
 	zfsvfs_t *zfsvfs = ZTOZSB(zp);
 	if ((error = zfs_enter_verify_zp(zfsvfs, zp, FTAG)) != 0)
@@ -424,8 +425,26 @@ zfs_read(struct znode *zp, zfs_uio_t *uio, int ioflag, cred_t *cr)
 
 		if (error) {
 			/* convert checksum errors into IO errors */
-			if (error == ECKSUM)
+			if (error == ECKSUM) {
-				error = SET_ERROR(EIO);
+				/*
+				 * If a Direct I/O read returned a checksum
+				 * verify error, then it must be treated as
+				 * suspicious. The contents of the buffer could
+				 * have beeen manipulated while the I/O was in
+				 * flight. In this case, the remainder of I/O
+				 * request will just be reissued through the
+				 * ARC.
+				 */
+				if (uio->uio_extflg & UIO_DIRECT) {
+					dio_checksum_failure = B_TRUE;
+					uio->uio_extflg &= ~UIO_DIRECT;
+					n += dio_remaining_resid;
+					dio_remaining_resid = 0;
+					continue;
+				} else {
+					error = SET_ERROR(EIO);
+				}
+			}
 
 #if defined(__linux__)
 			/*
@@ -472,6 +491,9 @@ zfs_read(struct znode *zp, zfs_uio_t *uio, int ioflag, cred_t *cr)
 out:
 	zfs_rangelock_exit(lr);
 
+	if (dio_checksum_failure == B_TRUE)
+		uio->uio_extflg |= UIO_DIRECT;
+
 	/*
 	 * Cleanup for Direct I/O if requested.
 	 */

module/zfs/zio.c

@@ -804,11 +804,11 @@ zio_notify_parent(zio_t *pio, zio_t *zio, enum zio_wait_type wait,
 	pio->io_reexecute |= zio->io_reexecute;
 	ASSERT3U(*countp, >, 0);
 
-	if (zio->io_flags & ZIO_FLAG_DIO_CHKSUM_ERR) {
+	/*
-		ASSERT3U(*errorp, ==, EIO);
+	 * Propogate the Direct I/O checksum verify failure to the parent.
-		ASSERT3U(pio->io_child_type, ==, ZIO_CHILD_LOGICAL);
+	 */
+	if (zio->io_flags & ZIO_FLAG_DIO_CHKSUM_ERR)
 		pio->io_flags |= ZIO_FLAG_DIO_CHKSUM_ERR;
-	}
 
 	(*countp)--;
 
@@ -1573,6 +1573,14 @@ zio_vdev_child_io(zio_t *pio, blkptr_t *bp, vdev_t *vd, uint64_t offset,
 		 */
 		pipeline |= ZIO_STAGE_CHECKSUM_VERIFY;
 		pio->io_pipeline &= ~ZIO_STAGE_CHECKSUM_VERIFY;
+		/*
+		 * We never allow the mirror VDEV to attempt reading from any
+		 * additional data copies after the first Direct I/O checksum
+		 * verify failure. This is to avoid bad data being written out
+		 * through the mirror during self healing. See comment in
+		 * vdev_mirror_io_done() for more details.
+		 */
+		ASSERT0(pio->io_flags & ZIO_FLAG_DIO_CHKSUM_ERR);
 	} else if (type == ZIO_TYPE_WRITE &&
 	    pio->io_prop.zp_direct_write == B_TRUE) {
 		/*
@@ -4555,18 +4563,18 @@ zio_vdev_io_assess(zio_t *zio)
 	}
 
 	/*
-	 * If a Direct I/O write checksum verify error has occurred then this
+	 * If a Direct I/O operation has a checksum verify error then this I/O
-	 * I/O should not attempt to be issued again. Instead the EIO will
+	 * should not attempt to be issued again.
-	 * be returned.
 	 */
 	if (zio->io_flags & ZIO_FLAG_DIO_CHKSUM_ERR) {
-		ASSERT3U(zio->io_child_type, ==, ZIO_CHILD_LOGICAL);
+		if (zio->io_type == ZIO_TYPE_WRITE) {
-		ASSERT3U(zio->io_error, ==, EIO);
+			ASSERT3U(zio->io_child_type, ==, ZIO_CHILD_LOGICAL);
+			ASSERT3U(zio->io_error, ==, EIO);
+		}
 		zio->io_pipeline = ZIO_INTERLOCK_PIPELINE;
 		return (zio);
 	}
 
-
 	if (zio_injection_enabled && zio->io_error == 0)
 		zio->io_error = zio_handle_fault_injection(zio, EIO);
 
@@ -4864,16 +4872,40 @@ zio_checksum_verify(zio_t *zio)
 		ASSERT3U(zio->io_prop.zp_checksum, ==, ZIO_CHECKSUM_LABEL);
 	}
 
+	ASSERT0(zio->io_flags & ZIO_FLAG_DIO_CHKSUM_ERR);
+	IMPLY(zio->io_flags & ZIO_FLAG_DIO_READ,
+	    !(zio->io_flags & ZIO_FLAG_SPECULATIVE));
+
 	if ((error = zio_checksum_error(zio, &info)) != 0) {
 		zio->io_error = error;
 		if (error == ECKSUM &&
 		    !(zio->io_flags & ZIO_FLAG_SPECULATIVE)) {
-			mutex_enter(&zio->io_vd->vdev_stat_lock);
+			if (zio->io_flags & ZIO_FLAG_DIO_READ) {
-			zio->io_vd->vdev_stat.vs_checksum_errors++;
+				zio->io_flags |= ZIO_FLAG_DIO_CHKSUM_ERR;
-			mutex_exit(&zio->io_vd->vdev_stat_lock);
+				zio_t *pio = zio_unique_parent(zio);
-			(void) zfs_ereport_start_checksum(zio->io_spa,
+				/*
-			    zio->io_vd, &zio->io_bookmark, zio,
+				 * Any Direct I/O read that has a checksum
-			    zio->io_offset, zio->io_size, &info);
+				 * error must be treated as suspicous as the
+				 * contents of the buffer could be getting
+				 * manipulated while the I/O is taking place.
+				 *
+				 * The checksum verify error will only be
+				 * reported here for disk and file VDEV's and
+				 * will be reported on those that the failure
+				 * occurred on. Other types of VDEV's report the
+				 * verify failure in their own code paths.
+				 */
+				if (pio->io_child_type == ZIO_CHILD_LOGICAL) {
+					zio_dio_chksum_verify_error_report(zio);
+				}
+			} else {
+				mutex_enter(&zio->io_vd->vdev_stat_lock);
+				zio->io_vd->vdev_stat.vs_checksum_errors++;
+				mutex_exit(&zio->io_vd->vdev_stat_lock);
+				(void) zfs_ereport_start_checksum(zio->io_spa,
+				    zio->io_vd, &zio->io_bookmark, zio,
+				    zio->io_offset, zio->io_size, &info);
+			}
 		}
 	}
 
@@ -4899,22 +4931,8 @@ zio_dio_checksum_verify(zio_t *zio)
 	if ((error = zio_checksum_error(zio, NULL)) != 0) {
 		zio->io_error = error;
 		if (error == ECKSUM) {
-			mutex_enter(&zio->io_vd->vdev_stat_lock);
-			zio->io_vd->vdev_stat.vs_dio_verify_errors++;
-			mutex_exit(&zio->io_vd->vdev_stat_lock);
-			zio->io_error = SET_ERROR(EIO);
 			zio->io_flags |= ZIO_FLAG_DIO_CHKSUM_ERR;
-
+			zio_dio_chksum_verify_error_report(zio);
-			/*
-			 * The EIO error must be propagated up to the logical
-			 * parent ZIO in zio_notify_parent() so it can be
-			 * returned to dmu_write_abd().
-			 */
-			zio->io_flags &= ~ZIO_FLAG_DONT_PROPAGATE;
-
-			(void) zfs_ereport_post(FM_EREPORT_ZFS_DIO_VERIFY,
-			    zio->io_spa, zio->io_vd, &zio->io_bookmark,
-			    zio, 0);
 		}
 	}
 
@@ -4932,6 +4950,39 @@ zio_checksum_verified(zio_t *zio)
 	zio->io_pipeline &= ~ZIO_STAGE_CHECKSUM_VERIFY;
 }
 
+/*
+ * Report Direct I/O checksum verify error and create ZED event.
+ */
+void
+zio_dio_chksum_verify_error_report(zio_t *zio)
+{
+	ASSERT(zio->io_flags & ZIO_FLAG_DIO_CHKSUM_ERR);
+
+	if (zio->io_child_type == ZIO_CHILD_LOGICAL)
+		return;
+
+	mutex_enter(&zio->io_vd->vdev_stat_lock);
+	zio->io_vd->vdev_stat.vs_dio_verify_errors++;
+	mutex_exit(&zio->io_vd->vdev_stat_lock);
+	if (zio->io_type == ZIO_TYPE_WRITE) {
+		/*
+		 * Convert checksum error for writes into EIO.
+		 */
+		zio->io_error = SET_ERROR(EIO);
+		/*
+		 * Report dio_verify_wr ZED event.
+		 */
+		(void) zfs_ereport_post(FM_EREPORT_ZFS_DIO_VERIFY_WR,
+		    zio->io_spa,  zio->io_vd, &zio->io_bookmark, zio, 0);
+	} else {
+		/*
+		 * Report dio_verify_rd ZED event.
+		 */
+		(void) zfs_ereport_post(FM_EREPORT_ZFS_DIO_VERIFY_RD,
+		    zio->io_spa, zio->io_vd, &zio->io_bookmark, zio, 0);
+	}
+}
+
 /*
  * ==========================================================================
  * Error rank.  Error are ranked in the order 0, ENXIO, ECKSUM, EIO, other.
@@ -5343,10 +5394,9 @@ zio_done(zio_t *zio)
 
 	if (zio->io_reexecute) {
 		/*
-		 * A Direct I/O write that has a checksum verify error should
+		 * A Direct I/O operation that has a checksum verify error
-		 * not attempt to reexecute. Instead, EAGAIN should just be
+		 * should not attempt to reexecute. Instead, the error should
-		 * propagated back up so the write can be attempt to be issued
+		 * just be propagated back.
-		 * through the ARC.
 		 */
 		ASSERT(!(zio->io_flags & ZIO_FLAG_DIO_CHKSUM_ERR));
 

rpm/generic/zfs.spec.in

@@ -99,10 +99,10 @@ License:        @ZFS_META_LICENSE@
 URL:            https://github.com/openzfs/zfs
 Source0:        %{name}-%{version}.tar.gz
 BuildRoot:      %{_tmppath}/%{name}-%{version}-%{release}-root-%(%{__id_u} -n)
-Requires:       libzpool5%{?_isa} = %{version}-%{release}
+Requires:       libzpool6%{?_isa} = %{version}-%{release}
 Requires:       libnvpair3%{?_isa} = %{version}-%{release}
 Requires:       libuutil3%{?_isa} = %{version}-%{release}
-Requires:       libzfs5%{?_isa} = %{version}-%{release}
+Requires:       libzfs6%{?_isa} = %{version}-%{release}
 Requires:       %{name}-kmod = %{version}
 Provides:       %{name}-kmod-common = %{version}-%{release}
 Obsoletes:      spl <= %{version}
@@ -150,21 +150,22 @@ Requires:  sysstat
 %description
 This package contains the core ZFS command line utilities.
 
-%package -n libzpool5
+%package -n libzpool6
 Summary:        Native ZFS pool library for Linux
 Group:          System Environment/Kernel
 Obsoletes:      libzpool2 <= %{version}
 Obsoletes:      libzpool4 <= %{version}
+Obsoletes:      libzpool5 <= %{version}
 
-%description -n libzpool5
+%description -n libzpool6
 This package contains the zpool library, which provides support
 for managing zpools
 
 %if %{defined ldconfig_scriptlets}
-%ldconfig_scriptlets -n libzpool5
+%ldconfig_scriptlets -n libzpool6
 %else
-%post -n libzpool5 -p /sbin/ldconfig
+%post -n libzpool6 -p /sbin/ldconfig
-%postun -n libzpool5 -p /sbin/ldconfig
+%postun -n libzpool6 -p /sbin/ldconfig
 %endif
 
 %package -n libnvpair3
@@ -211,37 +212,39 @@ This library provides a variety of compatibility functions for OpenZFS:
 # The library version is encoded in the package name.  When updating the
 # version information it is important to add an obsoletes line below for
 # the previous version of the package.
-%package -n libzfs5
+%package -n libzfs6
 Summary:        Native ZFS filesystem library for Linux
 Group:          System Environment/Kernel
 Obsoletes:      libzfs2 <= %{version}
 Obsoletes:      libzfs4 <= %{version}
+Obsoletes:      libzfs5 <= %{version}
 
-%description -n libzfs5
+%description -n libzfs6
 This package provides support for managing ZFS filesystems
 
 %if %{defined ldconfig_scriptlets}
-%ldconfig_scriptlets -n libzfs5
+%ldconfig_scriptlets -n libzfs6
 %else
-%post -n libzfs5 -p /sbin/ldconfig
+%post -n libzfs6 -p /sbin/ldconfig
-%postun -n libzfs5 -p /sbin/ldconfig
+%postun -n libzfs6 -p /sbin/ldconfig
 %endif
 
-%package -n libzfs5-devel
+%package -n libzfs6-devel
 Summary:        Development headers
 Group:          System Environment/Kernel
-Requires:       libzfs5%{?_isa} = %{version}-%{release}
+Requires:       libzfs6%{?_isa} = %{version}-%{release}
-Requires:       libzpool5%{?_isa} = %{version}-%{release}
+Requires:       libzpool6%{?_isa} = %{version}-%{release}
 Requires:       libnvpair3%{?_isa} = %{version}-%{release}
 Requires:       libuutil3%{?_isa} = %{version}-%{release}
-Provides:       libzpool5-devel = %{version}-%{release}
+Provides:       libzpool6-devel = %{version}-%{release}
 Provides:       libnvpair3-devel = %{version}-%{release}
 Provides:       libuutil3-devel = %{version}-%{release}
 Obsoletes:      zfs-devel <= %{version}
 Obsoletes:      libzfs2-devel <= %{version}
 Obsoletes:      libzfs4-devel <= %{version}
+Obsoletes:      libzfs5-devel <= %{version}
 
-%description -n libzfs5-devel
+%description -n libzfs6-devel
 This package contains the header files needed for building additional
 applications against the ZFS libraries.
 
@@ -290,7 +293,7 @@ Summary:        Python %{python_version} wrapper for libzfs_core
 Group:          Development/Languages/Python
 License:        Apache-2.0
 BuildArch:      noarch
-Requires:       libzfs5 = %{version}-%{release}
+Requires:       libzfs6 = %{version}-%{release}
 Requires:       libnvpair3 = %{version}-%{release}
 Requires:       libffi
 Requires:       python%{__python_pkg_version}
@@ -534,7 +537,7 @@ systemctl --system daemon-reload >/dev/null || true
 %config(noreplace) %{_bashcompletiondir}/zfs
 %config(noreplace) %{_bashcompletiondir}/zpool
 
-%files -n libzpool5
+%files -n libzpool6
 %{_libdir}/libzpool.so.*
 
 %files -n libnvpair3
@@ -543,10 +546,10 @@ systemctl --system daemon-reload >/dev/null || true
 %files -n libuutil3
 %{_libdir}/libuutil.so.*
 
-%files -n libzfs5
+%files -n libzfs6
 %{_libdir}/libzfs*.so.*
 
-%files -n libzfs5-devel
+%files -n libzfs6-devel
 %{_pkgconfigdir}/libzfs.pc
 %{_pkgconfigdir}/libzfsbootenv.pc
 %{_pkgconfigdir}/libzfs_core.pc

tests/runfiles/common.run

@@ -697,8 +697,8 @@ tags = ['functional', 'delegate']
 tests = ['dio_aligned_block', 'dio_async_always', 'dio_async_fio_ioengines',
     'dio_compression', 'dio_dedup', 'dio_encryption', 'dio_grow_block',
     'dio_max_recordsize', 'dio_mixed', 'dio_mmap', 'dio_overwrites',
-    'dio_property', 'dio_random', 'dio_recordsize', 'dio_unaligned_block',
+    'dio_property', 'dio_random', 'dio_read_verify', 'dio_recordsize',
-    'dio_unaligned_filesize']
+    'dio_unaligned_block', 'dio_unaligned_filesize']
 tags = ['functional', 'direct']
 
 [tests/functional/exec]

tests/runfiles/linux.run

@@ -147,6 +147,12 @@ tags = ['functional', 'largest_pool']
 tests = ['longname_001_pos', 'longname_002_pos', 'longname_003_pos']
 tags = ['functional', 'longname']
 
+[tests/functional/luks:Linux]
+pre =
+post =
+tests = ['luks_sanity']
+tags = ['functional', 'luks']
+
 [tests/functional/mmap:Linux]
 tests = ['mmap_libaio_001_pos', 'mmap_sync_001_pos']
 tags = ['functional', 'mmap']

tests/test-runner/bin/zts-report.py.in

@@ -213,6 +213,7 @@ maybe = {
     'cli_root/zfs_unshare/zfs_unshare_006_pos': ['SKIP', na_reason],
     'cli_root/zpool_add/zpool_add_004_pos': ['FAIL', known_reason],
     'cli_root/zpool_destroy/zpool_destroy_001_pos': ['SKIP', 6145],
+    'cli_root/zpool_import/import_devices_missing': ['FAIL', 16669],
     'cli_root/zpool_import/zpool_import_missing_003_pos': ['SKIP', 6839],
     'cli_root/zpool_initialize/zpool_initialize_import_export':
         ['FAIL', 11948],
@@ -275,7 +276,8 @@ if sys.platform.startswith('freebsd'):
         'pool_checkpoint/checkpoint_big_rewind': ['FAIL', 12622],
         'pool_checkpoint/checkpoint_indirect': ['FAIL', 12623],
         'resilver/resilver_restart_001': ['FAIL', known_reason],
-        'snapshot/snapshot_002_pos': ['FAIL', '14831'],
+        'snapshot/snapshot_002_pos': ['FAIL', 14831],
+        'zvol/zvol_misc/zvol_misc_volmode': ['FAIL', 16668],
         'bclone/bclone_crossfs_corner_cases': ['SKIP', cfr_cross_reason],
         'bclone/bclone_crossfs_corner_cases_limited':
             ['SKIP', cfr_cross_reason],

tests/zfs-tests/cmd/getversion.c

@@ -19,9 +19,9 @@
  */
 
 #include <sys/ioctl.h>
-#include <sys/fcntl.h>
 #include <linux/fs.h>
 #include <err.h>
+#include <fcntl.h>
 #include <stdio.h>
 #include <stdlib.h>
 #include <unistd.h>

tests/zfs-tests/cmd/manipulate_user_buffer.c

@@ -20,7 +20,7 @@
  */
 
 /*
- * Copyright (c) 2022 by Triad National Security, LLC.
+ * Copyright (c) 2024 by Triad National Security, LLC.
  */
 
 #include <sys/types.h>
@@ -39,51 +39,59 @@
 #define	MIN(a, b)	((a) < (b)) ? (a) : (b)
 #endif
 
-static char *outputfile = NULL;
+static char *filename = NULL;
 static int blocksize = 131072; /* 128K */
-static int wr_err_expected = 0;
+static int err_expected = 0;
+static int read_op = 0;
+static int write_op = 0;
 static int numblocks = 100;
 static char *execname = NULL;
 static int print_usage = 0;
 static int randompattern = 0;
-static int ofd;
+static int fd;
 char *buf = NULL;
 
 typedef struct {
-	int entire_file_written;
+	int entire_file_completed;
 } pthread_args_t;
 
 static void
 usage(void)
 {
 	(void) fprintf(stderr,
-	    "usage %s -o outputfile [-b blocksize] [-e wr_error_expected]\n"
+	    "usage %s -f filename [-b blocksize] [-e wr_error_expected]\n"
-	    "         [-n numblocks] [-p randpattern] [-h help]\n"
+	    "         [-n numblocks] [-p randompattern] -r read_op \n"
+	    "         -w write_op [-h help]\n"
 	    "\n"
 	    "Testing whether checksum verify works correctly for O_DIRECT.\n"
 	    "when manipulating the contents of a userspace buffer.\n"
 	    "\n"
-	    "    outputfile:      File to write to.\n"
+	    "    filename:       File to read or write to.\n"
-	    "    blocksize:       Size of each block to write (must be at \n"
+	    "    blocksize:      Size of each block to write (must be at \n"
-	    "                     least >= 512).\n"
+	    "                    least >= 512).\n"
-	    "    wr_err_expected: Whether pwrite() is expected to return EIO\n"
+	    "    err_expected:   Whether write() is expected to return EIO\n"
-	    "                     while manipulating the contents of the\n"
+	    "                    while manipulating the contents of the\n"
-	    "                     buffer.\n"
+	    "                    buffer.\n"
-	    "    numblocks:       Total number of blocksized blocks to\n"
+	    "    numblocks:      Total number of blocksized blocks to\n"
-	    "                     write.\n"
+	    "                    write.\n"
-	    "    randpattern:     Fill data buffer with random data. Default\n"
+	    "    read_op:        Perform reads to the filename file while\n"
-	    "                     behavior is to fill the buffer with the \n"
+	    "                    while manipulating the buffer contents\n"
-	    "                     known data pattern (0xdeadbeef).\n"
+	    "    write_op:       Perform writes to the filename file while\n"
+	    "                    manipulating the buffer contents\n"
+	    "    randompattern:  Fill data buffer with random data for \n"
+	    "                    writes. Default behavior is to fill the \n"
+	    "                    buffer with known data pattern (0xdeadbeef)\n"
 	    "    help:           Print usage information and exit.\n"
 	    "\n"
 	    "    Required parameters:\n"
-	    "    outputfile\n"
+	    "    filename\n"
+	    "    read_op or write_op\n"
 	    "\n"
 	    "    Default Values:\n"
 	    "    blocksize       -> 131072\n"
 	    "    wr_err_expexted -> false\n"
 	    "    numblocks       -> 100\n"
-	    "    randpattern     -> false\n",
+	    "    randompattern   -> false\n",
 	    execname);
 	(void) exit(1);
 }
@@ -97,16 +105,21 @@ parse_options(int argc, char *argv[])
 	extern int optind, optopt;
 	execname = argv[0];
 
-	while ((c = getopt(argc, argv, "b:ehn:o:p")) != -1) {
+	while ((c = getopt(argc, argv, "b:ef:hn:rw")) != -1) {
 		switch (c) {
 			case 'b':
 				blocksize = atoi(optarg);
 				break;
 
 			case 'e':
-				wr_err_expected = 1;
+				err_expected = 1;
 				break;
 
+			case 'f':
+				filename = optarg;
+				break;
+
+
 			case 'h':
 				print_usage = 1;
 				break;
@@ -115,12 +128,12 @@ parse_options(int argc, char *argv[])
 				numblocks = atoi(optarg);
 				break;
 
-			case 'o':
+			case 'r':
-				outputfile = optarg;
+				read_op = 1;
 				break;
 
-			case 'p':
+			case 'w':
-				randompattern = 1;
+				write_op = 1;
 				break;
 
 			case ':':
@@ -141,7 +154,8 @@ parse_options(int argc, char *argv[])
 	if (errflag || print_usage == 1)
 		(void) usage();
 
-	if (blocksize < 512 || outputfile == NULL || numblocks <= 0) {
+	if (blocksize < 512 || filename == NULL || numblocks <= 0 ||
+	    (read_op == 0 && write_op == 0)) {
 		(void) fprintf(stderr,
 		    "Required paramater(s) missing or invalid.\n");
 		(void) usage();
@@ -160,10 +174,10 @@ write_thread(void *arg)
 	ssize_t wrote = 0;
 	pthread_args_t *args = (pthread_args_t *)arg;
 
-	while (!args->entire_file_written) {
+	while (!args->entire_file_completed) {
-		wrote = pwrite(ofd, buf, blocksize, offset);
+		wrote = pwrite(fd, buf, blocksize, offset);
 		if (wrote != blocksize) {
-			if (wr_err_expected)
+			if (err_expected)
 				assert(errno == EIO);
 			else
 				exit(2);
@@ -173,7 +187,35 @@ write_thread(void *arg)
 		left -= blocksize;
 
 		if (left == 0)
-			args->entire_file_written = 1;
+			args->entire_file_completed = 1;
+	}
+
+	pthread_exit(NULL);
+}
+
+/*
+ * Read blocksize * numblocks to the file using O_DIRECT.
+ */
+static void *
+read_thread(void *arg)
+{
+	size_t offset = 0;
+	int total_data = blocksize * numblocks;
+	int left = total_data;
+	ssize_t read = 0;
+	pthread_args_t *args = (pthread_args_t *)arg;
+
+	while (!args->entire_file_completed) {
+		read = pread(fd, buf, blocksize, offset);
+		if (read != blocksize) {
+			exit(2);
+		}
+
+		offset = ((offset + blocksize) % total_data);
+		left -= blocksize;
+
+		if (left == 0)
+			args->entire_file_completed = 1;
 	}
 
 	pthread_exit(NULL);
@@ -189,7 +231,7 @@ manipulate_buf_thread(void *arg)
 	char rand_char;
 	pthread_args_t *args = (pthread_args_t *)arg;
 
-	while (!args->entire_file_written) {
+	while (!args->entire_file_completed) {
 		rand_offset = (rand() % blocksize);
 		rand_char = (rand() % (126 - 33) + 33);
 		buf[rand_offset] = rand_char;
@@ -202,9 +244,9 @@ int
 main(int argc, char *argv[])
 {
 	const char *datapattern = "0xdeadbeef";
-	int ofd_flags = O_WRONLY | O_CREAT | O_DIRECT;
+	int fd_flags = O_DIRECT;
 	mode_t mode = S_IRUSR | S_IWUSR;
-	pthread_t write_thr;
+	pthread_t io_thr;
 	pthread_t manipul_thr;
 	int left = blocksize;
 	int offset = 0;
@@ -213,9 +255,15 @@ main(int argc, char *argv[])
 
 	parse_options(argc, argv);
 
-	ofd = open(outputfile, ofd_flags, mode);
+	if (write_op) {
-	if (ofd == -1) {
+		fd_flags |= (O_WRONLY | O_CREAT);
-		(void) fprintf(stderr, "%s, %s\n", execname, outputfile);
+	} else {
+		fd_flags |= O_RDONLY;
+	}
+
+	fd = open(filename, fd_flags, mode);
+	if (fd == -1) {
+		(void) fprintf(stderr, "%s, %s\n", execname, filename);
 		perror("open");
 		exit(2);
 	}
@@ -228,24 +276,22 @@ main(int argc, char *argv[])
 		exit(2);
 	}
 
-	if (!randompattern) {
+	if (write_op) {
-		/* Putting known data pattern in buffer */
+		if (!randompattern) {
-		while (left) {
+			/* Putting known data pattern in buffer */
-			size_t amt = MIN(strlen(datapattern), left);
+			while (left) {
-			memcpy(&buf[offset], datapattern, amt);
+				size_t amt = MIN(strlen(datapattern), left);
-			offset += amt;
+				memcpy(&buf[offset], datapattern, amt);
-			left -= amt;
+				offset += amt;
+				left -= amt;
+			}
+		} else {
+			/* Putting random data in buffer */
+			for (int i = 0; i < blocksize; i++)
+				buf[i] = rand();
 		}
-	} else {
-		/* Putting random data in buffer */
-		for (int i = 0; i < blocksize; i++)
-			buf[i] = rand();
 	}
 
-	/*
-	 * Writing using O_DIRECT while manipulating the buffer contents until
-	 * the entire file is written.
-	 */
 	if ((rc = pthread_create(&manipul_thr, NULL, manipulate_buf_thread,
 	    &args))) {
 		fprintf(stderr, "error: pthreads_create, manipul_thr, "
@@ -253,18 +299,34 @@ main(int argc, char *argv[])
 		exit(2);
 	}
 
-	if ((rc = pthread_create(&write_thr, NULL, write_thread, &args))) {
+	if (write_op) {
-		fprintf(stderr, "error: pthreads_create, write_thr, "
+		/*
-		    "rc: %d\n", rc);
+		 * Writing using O_DIRECT while manipulating the buffer contents
-		exit(2);
+		 * until the entire file is written.
+		 */
+		if ((rc = pthread_create(&io_thr, NULL, write_thread, &args))) {
+			fprintf(stderr, "error: pthreads_create, io_thr, "
+			    "rc: %d\n", rc);
+			exit(2);
+		}
+	} else {
+		/*
+		 * Reading using O_DIRECT while manipulating the buffer contents
+		 * until the entire file is read.
+		 */
+		if ((rc = pthread_create(&io_thr, NULL, read_thread, &args))) {
+			fprintf(stderr, "error: pthreads_create, io_thr, "
+			    "rc: %d\n", rc);
+			exit(2);
+		}
 	}
 
-	pthread_join(write_thr, NULL);
+	pthread_join(io_thr, NULL);
 	pthread_join(manipul_thr, NULL);
 
-	assert(args.entire_file_written == 1);
+	assert(args.entire_file_completed == 1);
 
-	(void) close(ofd);
+	(void) close(fd);
 
 	free(buf);
 

tests/zfs-tests/include/commands.cfg

@@ -129,6 +129,7 @@ export SYSTEM_FILES_LINUX='attr
     blkdiscard
     blockdev
     chattr
+    cryptsetup
     exportfs
     fallocate
     flock

tests/zfs-tests/tests/Makefile.am

@@ -80,7 +80,8 @@ if BUILD_LINUX
 nobase_dist_datadir_zfs_tests_tests_SCRIPTS += \
 	functional/simd/simd_supported.ksh \
 	functional/tmpfile/cleanup.ksh \
-	functional/tmpfile/setup.ksh
+	functional/tmpfile/setup.ksh \
+	functional/luks/luks_sanity.ksh
 endif
 
 nobase_dist_datadir_zfs_tests_tests_DATA += \
@@ -1477,6 +1478,7 @@ nobase_dist_datadir_zfs_tests_tests_SCRIPTS += \
 	functional/direct/dio_overwrites.ksh \
 	functional/direct/dio_property.ksh \
 	functional/direct/dio_random.ksh \
+	functional/direct/dio_read_verify.ksh \
 	functional/direct/dio_recordsize.ksh \
 	functional/direct/dio_unaligned_block.ksh \
 	functional/direct/dio_unaligned_filesize.ksh \

tests/zfs-tests/tests/functional/cli_root/json/json_sanity.ksh

@@ -30,28 +30,39 @@
 # STRATEGY:
 # 1. Run different zfs/zpool -j commands and check for valid JSON
 
+#
+# -j and --json mean the same thing. Each command will be run twice, replacing
+# JSONFLAG with the flag under test.
 list=(
-    "zpool status -j -g --json-int --json-flat-vdevs --json-pool-key-guid"
+    "zpool status JSONFLAG -g --json-int --json-flat-vdevs --json-pool-key-guid"
-    "zpool status -p -j -g --json-int --json-flat-vdevs --json-pool-key-guid"
+    "zpool status -p JSONFLAG -g --json-int --json-flat-vdevs --json-pool-key-guid"
-    "zpool status -j -c upath"
+    "zpool status JSONFLAG -c upath"
-    "zpool status -j"
+    "zpool status JSONFLAG"
-    "zpool status -j testpool1"
+    "zpool status JSONFLAG testpool1"
-    "zpool list -j"
+    "zpool list JSONFLAG"
-    "zpool list -j -g"
+    "zpool list JSONFLAG -g"
-    "zpool list -j -o fragmentation"
+    "zpool list JSONFLAG -o fragmentation"
-    "zpool get -j size"
+    "zpool get JSONFLAG size"
-    "zpool get -j all"
+    "zpool get JSONFLAG all"
-    "zpool version -j"
+    "zpool version JSONFLAG"
-    "zfs list -j"
+    "zfs list JSONFLAG"
-    "zfs list -j testpool1"
+    "zfs list JSONFLAG testpool1"
-    "zfs get -j all"
+    "zfs get JSONFLAG all"
-    "zfs get -j available"
+    "zfs get JSONFLAG available"
-    "zfs mount -j"
+    "zfs mount JSONFLAG"
-    "zfs version -j"
+    "zfs version JSONFLAG"
 )
 
-for cmd in "${list[@]}" ; do
+function run_json_tests
-    log_must eval "$cmd | jq > /dev/null" 
+{
-done
+	typeset flag=$1
+	for cmd in "${list[@]}" ; do
+	    cmd=${cmd//JSONFLAG/$flag}
+	    log_must eval "$cmd | jq > /dev/null"
+	done
+}
+
+log_must run_json_tests -j
+log_must run_json_tests --json
 
 log_pass "zpool and zfs commands outputted valid JSON"

tests/zfs-tests/tests/functional/cli_root/zpool_add/zpool_add_dryrun_output.ksh

@@ -148,9 +148,9 @@ done
 
 # Foreach test create pool, add -n devices and check output.
 for (( i=0; i < ${#tests[@]}; i+=1 )); do
-	typeset tree="${tests[$i].tree}"
+	tree="${tests[$i].tree}"
-	typeset add="${tests[$i].add}"
+	add="${tests[$i].add}"
-	typeset want="${tests[$i].want}"
+	want="${tests[$i].want}"
 
 	log_must eval zpool create "$TESTPOOL" $tree
 	log_must poolexists "$TESTPOOL"

tests/zfs-tests/tests/functional/cli_root/zpool_create/zpool_create_dryrun_output.ksh

@@ -124,8 +124,8 @@ done
 
 # Foreach test create pool, add -n devices and check output.
 for (( i=0; i < ${#tests[@]}; i+=1 )); do
-	typeset tree="${tests[$i].tree}"
+	tree="${tests[$i].tree}"
-	typeset want="${tests[$i].want}"
+	want="${tests[$i].want}"
 
 	typeset out="$(log_must eval "zpool create -n '$TESTPOOL' $tree" | \
 	    sed /^SUCCESS/d)"

tests/zfs-tests/tests/functional/cli_root/zpool_import/zpool_import_parallel_pos.ksh

@@ -113,7 +113,7 @@ wait
 parallel_time=$SECONDS
 log_note "asyncronously imported 4 pools in $parallel_time seconds"
 
-log_must test $parallel_time -lt $(($sequential_time / 3))
+log_must test $parallel_time -lt $(($sequential_time / 2))
 
 #
 # export pools with import delay injectors
@@ -132,6 +132,6 @@ log_must zpool import -a -d $DEVICE_DIR -f
 parallel_time=$SECONDS
 log_note "asyncronously imported 4 pools in $parallel_time seconds"
 
-log_must test $parallel_time -lt $(($sequential_time / 3))
+log_must test $parallel_time -lt $(($sequential_time / 2))
 
 log_pass "Pool imports occur in parallel"