Tag zfs-2.3.6

META file and changelog updated.

Signed-off-by: Tony Hutter <hutter2@llnl.gov>

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

@@ -6,12 +6,19 @@
 
 set -eu
 
-# We've been seeing this script take over 15min to run.  This may or
+# The default 'azure.archive.ubuntu.com' mirrors can be really slow.
-# may not be normal.  Just to get a little more insight, print out
+# Prioritize the official Ubuntu mirrors.
-# a message to stdout with the top running process, and do this every
+#
-# 30 seconds.  We can delete this watchdog later once we get a better
+# The normal apt-mirrors.txt will look like:
-# handle on what the timeout value should be.
+#
-(while [ 1 ] ; do sleep 30 && echo "[watchdog: $(ps -eo cmd --sort=-pcpu  | head -n 2 | tail -n 1)}')]"; done) &
+# http://azure.archive.ubuntu.com/ubuntu/       priority:1
+# https://archive.ubuntu.com/ubuntu/    priority:2
+# https://security.ubuntu.com/ubuntu/   priority:3
+#
+# Just delete the 'azure.archive.ubuntu.com' line.
+sudo sed -i '/azure.archive.ubuntu.com/d' /etc/apt/apt-mirrors.txt
+echo "Using mirrors:"
+cat /etc/apt/apt-mirrors.txt
 
 # install needed packages
 export DEBIAN_FRONTEND="noninteractive"
@@ -27,35 +34,89 @@ ssh-keygen -t ed25519 -f ~/.ssh/id_ed25519 -q -N ""
 sudo systemctl stop docker.socket
 sudo systemctl stop multipathd.socket
 
-# remove default swapfile and /mnt
 sudo swapoff -a
-sudo umount -l /mnt
+
-DISK="/dev/disk/cloud/azure_resource-part1"
+# Special case:
-sudo sed -e "s|^$DISK.*||g" -i /etc/fstab
+#
-sudo wipefs -aq $DISK
+# For reasons unknown, the runner can boot-up with two different block device
-sudo systemctl daemon-reload
+# configurations.  On one config you get two 75GB block devices, and on the
+# other you get a single 150GB block device. Here's what both look like:
+#
+# --- Two 75GB block devices ---
+# NAME    MAJ:MIN RM  SIZE RO TYPE MOUNTPOINTS
+# sda       8:0    0  150G  0 disk
+# ├─sda1    8:1    0  149G  0 part /
+# ├─sda14   8:14   0    4M  0 part
+# ├─sda15   8:15   0  106M  0 part /boot/efi
+# └─sda16 259:0    0  913M  0 part /boot
+#
+# lrwxrwxrwx 1 root root  9 Jan 29 18:07 azure_root -> ../../sda
+# lrwxrwxrwx 1 root root 10 Jan 29 18:07 azure_root-part1 -> ../../sda1
+# lrwxrwxrwx 1 root root 11 Jan 29 18:07 azure_root-part14 -> ../../sda14
+# lrwxrwxrwx 1 root root 11 Jan 29 18:07 azure_root-part15 -> ../../sda15
+# lrwxrwxrwx 1 root root 11 Jan 29 18:07 azure_root-part16 -> ../../sda16
+#
+# --- One 150GB block device ---
+# NAME    MAJ:MIN RM  SIZE RO TYPE MOUNTPOINTS
+# sda       8:0    0   75G  0 disk
+# ├─sda1    8:1    0   74G  0 part /
+# ├─sda14   8:14   0    4M  0 part
+# ├─sda15   8:15   0  106M  0 part /boot/efi
+# └─sda16 259:0    0  913M  0 part /boot
+# sdb       8:16   0   75G  0 disk
+# └─sdb1    8:17   0   75G  0 part
+#
+# lrwxrwxrwx 1 root root  9 Jan 29 18:07 azure_resource -> ../../sdb
+# lrwxrwxrwx 1 root root 10 Jan 29 18:07 azure_resource-part1 -> ../../sdb1
+# lrwxrwxrwx 1 root root  9 Jan 29 18:07 azure_root -> ../../sda
+# lrwxrwxrwx 1 root root 10 Jan 29 18:07 azure_root-part1 -> ../../sda1
+# lrwxrwxrwx 1 root root 11 Jan 29 18:07 azure_root-part14 -> ../../sda14
+# lrwxrwxrwx 1 root root 11 Jan 29 18:07 azure_root-part15 -> ../../sda15
+#
+# If we have the azure_resource-part1 partition, umount it, partition it, and
+# use it as our ZFS disk and swap partition.  If not, just create a file VDEV
+# and swap file and use that instead.
+
+# remove default swapfile and /mnt
+if [ -e /dev/disk/cloud/azure_resource-part1 ] ; then
+  sudo umount -l /mnt
+  DISK="/dev/disk/cloud/azure_resource-part1"
+  sudo sed -e "s|^$DISK.*||g" -i /etc/fstab
+  sudo wipefs -aq $DISK
+  sudo systemctl daemon-reload
+fi
 
 sudo modprobe loop
 sudo modprobe zfs
 
-# partition the disk as needed
+if [ -e /dev/disk/cloud/azure_resource-part1 ] ; then
-DISK="/dev/disk/cloud/azure_resource"
+  echo "We have two 75GB block devices"
-sudo sgdisk --zap-all $DISK
+  # partition the disk as needed
-sudo sgdisk -p \
+  DISK="/dev/disk/cloud/azure_resource"
- -n 1:0:+16G -c 1:"swap" \
+  sudo sgdisk --zap-all $DISK
- -n 2:0:0    -c 2:"tests" \
+  sudo sgdisk -p \
-$DISK
+   -n 1:0:+16G -c 1:"swap" \
-sync
+   -n 2:0:0    -c 2:"tests" \
-sleep 1
+   $DISK
+  sync
+  sleep 1
+
+  sudo fallocate -l 12G /test.ssd2
+  DISKS="$DISK-part2 /test.ssd2"
+
+  SWAP=$DISK-part1
+else
+  echo "We have a single 150GB block device"
+  sudo fallocate -l 72G /test.ssd2
+  SWAP=/swapfile.ssd
+  sudo fallocate -l 16G $SWAP
+  sudo chmod 600 $SWAP
+  DISKS="/test.ssd2"
+fi
 
 # swap with same size as RAM (16GiB)
-sudo mkswap $DISK-part1
+sudo mkswap $SWAP
-sudo swapon $DISK-part1
+sudo swapon $SWAP
-
-# JBOD 2xdisk for OpenZFS storage (test vm's)
-SSD1="$DISK-part2"
-sudo fallocate -l 12G /test.ssd2
-SSD2=$(sudo losetup -b 4096 -f /test.ssd2 --show)
 
 # adjust zfs module parameter and create pool
 exec 1>/dev/null
@@ -64,7 +125,7 @@ ARC_MAX=$((1024*1024*512))
 echo $ARC_MIN | sudo tee /sys/module/zfs/parameters/zfs_arc_min
 echo $ARC_MAX | sudo tee /sys/module/zfs/parameters/zfs_arc_max
 echo 1 | sudo tee /sys/module/zfs/parameters/zvol_use_blk_mq
-sudo zpool create -f -o ashift=12 zpool $SSD1 $SSD2 -O relatime=off \
+sudo zpool create -f -o ashift=12 zpool $DISKS -O relatime=off \
   -O atime=off -O xattr=sa -O compression=lz4 -O sync=disabled \
   -O redundant_metadata=none -O mountpoint=/mnt/tests
 
@@ -72,6 +133,3 @@ sudo zpool create -f -o ashift=12 zpool $SSD1 $SSD2 -O relatime=off \
 for i in /sys/block/s*/queue/scheduler; do
   echo "none" | sudo tee $i
 done
-
-# Kill off our watchdog
-kill $(jobs -p)

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

@@ -43,6 +43,12 @@ case "$OS" in
     OSv="almalinux9"
     URL="https://repo.almalinux.org/almalinux/10/cloud/x86_64/images/AlmaLinux-10-GenericCloud-latest.x86_64.qcow2"
     ;;
+  alpine3-23)
+    OSNAME="Alpine Linux 3.23.2"
+    # Alpine Linux v3.22 and v3.23 are unknown to osinfo as of 2025-12-26.
+    OSv="alpinelinux3.21"
+    URL="https://dl-cdn.alpinelinux.org/alpine/v3.23/releases/cloud/generic_alpine-3.23.2-x86_64-bios-cloudinit-r0.qcow2"
+    ;;
   archlinux)
     OSNAME="Archlinux"
     URL="https://geo.mirror.pkgbuild.com/images/latest/Arch-Linux-x86_64-cloudimg.qcow2"
@@ -223,13 +229,21 @@ if [ ${OS:0:7} != "freebsd" ]; then
 hostname: $OS
 
 users:
-- name: root
+  - name: root
-  shell: $BASH
+    shell: /bin/bash
-- name: zfs
+    sudo: ['ALL=(ALL) NOPASSWD:ALL']
-  sudo: ALL=(ALL) NOPASSWD:ALL
+  - name: zfs
-  shell: $BASH
+    shell: /bin/bash
-  ssh_authorized_keys:
+    sudo: ['ALL=(ALL) NOPASSWD:ALL']
-    - $PUBKEY
+    ssh_authorized_keys:
+      - $PUBKEY
+    # Workaround for Alpine Linux.
+    lock_passwd: false
+    passwd: '*'
+
+packages:
+  - sudo
+  - bash
 
 growpart:
   mode: auto
@@ -312,3 +326,23 @@ else
   scp ~/src.txz "root@vm0:/tmp/src.txz"
   ssh root@vm0 'tar -C / -zxf /tmp/src.txz'
 fi
+
+#
+# Config for Alpine Linux similar to FreeBSD.
+#
+if [ ${OS:0:6} == "alpine" ]; then
+  while pidof /usr/bin/qemu-system-x86_64 >/dev/null; do
+    ssh 2>/dev/null zfs@vm0 "uname -a" && break
+  done
+  # Enable community and testing repositories.
+  ssh zfs@vm0 "sudo rm -rf /etc/apk/repositories"
+  ssh zfs@vm0 "sudo setup-apkrepos -c1"
+  ssh zfs@vm0 "echo '@testing http://dl-cdn.alpinelinux.org/alpine/edge/testing' | sudo tee -a /etc/apk/repositories"
+  # Upgrade to edge or latest-stable.
+  #ssh zfs@vm0 "sudo sed -i 's#/v[0-9]\+\.[0-9]\+/#/edge/#g' /etc/apk/repositories"
+  #ssh zfs@vm0 "sudo sed -i 's#/v[0-9]\+\.[0-9]\+/#/latest-stable/#g' /etc/apk/repositories"
+  # Update and upgrade after repository setup.
+  ssh zfs@vm0 "sudo apk update"
+  ssh zfs@vm0 "sudo apk add --upgrade apk-tools"
+  ssh zfs@vm0 "sudo apk upgrade --available"
+fi

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

@@ -10,6 +10,32 @@
 
 set -eu
 
+function alpine() {
+  echo "##[group]Install Development Tools"
+  sudo apk add \
+    acl alpine-sdk attr autoconf automake bash build-base clang21 coreutils \
+    cpio cryptsetup curl curl-dev dhcpcd eudev eudev-dev eudev-libs findutils \
+    fio gawk gdb gettext-dev git grep jq libaio libaio-dev libcurl \
+    libtirpc-dev libtool libunwind libunwind-dev linux-headers linux-tools \
+    linux-virt linux-virt-dev lsscsi m4 make nfs-utils openssl-dev parted \
+    pax procps py3-cffi py3-distlib py3-packaging py3-setuptools python3 \
+    python3-dev qemu-guest-agent rng-tools rsync samba samba-server sed \
+    strace sysstat util-linux util-linux-dev wget words xfsprogs xxhash \
+    zlib-dev pamtester@testing
+  echo "##[endgroup]"
+
+  echo "##[group]Switch to eudev"
+  sudo setup-devd udev
+  echo "##[endgroup]"
+
+  echo "##[group]Install ksh93 from Source"
+  git clone --depth 1 https://github.com/ksh93/ksh.git /tmp/ksh
+  cd /tmp/ksh
+  ./bin/package make
+  sudo ./bin/package install /
+  echo "##[endgroup]"
+}
+
 function archlinux() {
   echo "##[group]Running pacman -Syu"
   sudo btrfs filesystem resize max /
@@ -27,6 +53,10 @@ function archlinux() {
 function debian() {
   export DEBIAN_FRONTEND="noninteractive"
 
+  echo "##[group]Wait for cloud-init to finish"
+  cloud-init status --wait
+  echo "##[endgroup]"
+
   echo "##[group]Running apt-get update+upgrade"
   sudo sed -i '/[[:alpha:]]-backports/d' /etc/apt/sources.list
   sudo apt-get update -y
@@ -90,6 +120,11 @@ function rhel() {
     kernel-devel python3-setuptools qemu-guest-agent rng-tools rpcgen \
     rpm-build rsync samba strace sysstat systemd watchdog wget xfsprogs-devel \
     xxhash zlib-devel
+
+  # These are needed for building Lustre.  We only install these on EL VMs since
+  # we don't plan to test build Lustre on other platforms.
+  sudo dnf install -y libnl3-devel libyaml-devel libmount-devel
+
   echo "##[endgroup]"
 }
 
@@ -140,6 +175,9 @@ case "$1" in
     sudo dnf install -y kernel-abi-stablelists
     echo "##[endgroup]"
     ;;
+  alpine*)
+    alpine
+    ;;
   archlinux)
     archlinux
     ;;
@@ -188,6 +226,16 @@ test -z "${ONLY_DEPS:-}" || exit 0
 # Start services
 echo "##[group]Enable services"
 case "$1" in
+  alpine*)
+    sudo -E rc-update add qemu-guest-agent
+    sudo -E rc-update add nfs
+    sudo -E rc-update add samba
+    sudo -E rc-update add dhcpcd
+    # Remove services related to cloud-init.
+    sudo -E rc-update del cloud-init default
+    sudo -E rc-update del cloud-final default
+    sudo -E rc-update del cloud-config default
+    ;;
   freebsd*)
     # add virtio things
     echo 'virtio_load="YES"' | sudo -E tee -a /boot/loader.conf
@@ -243,7 +291,7 @@ case "$1" in
 esac
 
 case "$1" in
-  archlinux|freebsd*)
+  alpine*|archlinux|freebsd*)
     true
     ;;
   *)

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

@@ -58,13 +58,21 @@ for ((i=1; i<=VMs; i++)); do
 fqdn: vm$i
 
 users:
-- name: root
+  - name: root
-  shell: $BASH
+    shell: /bin/bash
-- name: zfs
+    sudo: ['ALL=(ALL) NOPASSWD:ALL']
-  sudo: ALL=(ALL) NOPASSWD:ALL
+  - name: zfs
-  shell: $BASH
+    shell: /bin/bash
-  ssh_authorized_keys:
+    sudo: ['ALL=(ALL) NOPASSWD:ALL']
-    - $PUBKEY
+    ssh_authorized_keys:
+      - $PUBKEY
+    # Workaround for Alpine Linux.
+    lock_passwd: false
+    passwd: '*'
+
+packages:
+  - sudo
+  - bash
 
 growpart:
   mode: auto

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

@@ -0,0 +1,51 @@
+#!/usr/bin/env bash
+
+######################################################################
+# 6) Test if Lustre can still build against ZFS
+######################################################################
+set -e
+
+# Build from the latest Lustre tag rather than the master branch.  We do this
+# under the assumption that master is going to have a lot of churn thus will be
+# more prone to breaking the build than a point release.  We don't want ZFS
+# PR's reporting bad test results simply because upstream Lustre accidentally
+# broke their build.
+#
+# Skip any RC tags, or any tags where the last version digit is 50 or more.
+# Versions with 50 or more are development versions of Lustre.
+repo=https://github.com/lustre/lustre-release.git
+tag="$(git ls-remote --refs --exit-code --sort=version:refname --tags $repo | \
+	awk -F '_' '/-RC/{next}; /refs\/tags\/v/{if ($NF < 50){print}}' | \
+	tail -n 1 | sed 's/.*\///')"
+
+echo "Cloning Lustre tag $tag"
+git clone --depth 1 --branch "$tag" "$repo"
+
+cd lustre-release
+
+# Include Lustre patches to build against master/zfs-2.4.x.  Once these
+# patches are merged we can remove these lines.
+patches=('https://review.whamcloud.com/changes/fs%2Flustre-release~62101/revisions/2/patch?download'
+	'https://review.whamcloud.com/changes/fs%2Flustre-release~63267/revisions/9/patch?download')
+
+for p in "${patches[@]}" ; do
+	curl $p | base64 -d > patch
+	patch -p1 < patch || true
+done
+
+echo "Configure Lustre"
+./autogen.sh
+# EL 9 needs '--disable-gss-keyring'
+./configure --with-zfs --disable-gss-keyring
+echo "Building Lustre RPMs"
+make rpms
+ls *.rpm
+
+# There's only a handful of Lustre RPMs we actually need to install
+lustrerpms="$(ls *.rpm | grep -E 'kmod-lustre-osd-zfs-[0-9]|kmod-lustre-[0-9]|lustre-osd-zfs-mount-[0-9]')"
+echo "Installing: $lustrerpms"
+sudo dnf -y install $lustrerpms
+sudo modprobe -v lustre
+
+# Should see some Lustre lines in dmesg
+sudo dmesg | grep -Ei 'lnet|lustre'

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

@@ -4,7 +4,10 @@
 # 6) load openzfs module and run the tests
 #
 # called on runner:  qemu-6-tests.sh
-# called on qemu-vm: qemu-6-tests.sh $OS $2/$3
+# called on qemu-vm: qemu-6-tests.sh $OS $2 $3 [--lustre|--builtin] [quick|default]
+#
+# --lustre: Test build lustre in addition to the normal tests
+# --builtin: Test build ZFS as a kernel built-in in addition to the normal tests
 ######################################################################
 
 set -eu
@@ -38,6 +41,54 @@ function prefix() {
   fi
 }
 
+function do_lustre_build() {
+  local rc=0
+  $HOME/zfs/.github/workflows/scripts/qemu-6-lustre-tests-vm.sh &> /var/tmp/lustre.txt || rc=$?
+  echo "$rc" > /var/tmp/lustre-exitcode.txt
+  if [ "$rc" != "0" ] ; then
+      echo "$rc" > /var/tmp/tests-exitcode.txt
+  fi
+}
+export -f do_lustre_build
+
+# Test build ZFS into the kernel directly
+function do_builtin_build() {
+  local rc=0
+  # Get currently full kernel version (like '6.18.8')
+  fullver=$(uname -r | grep -Eo '^[0-9]+\.[0-9]+\.[0-9]+')
+
+  # Get just the major ('6')
+  major=$(echo $fullver | grep -Eo '^[0-9]+')
+  (
+  set -e
+
+  wget https://cdn.kernel.org/pub/linux/kernel/v${major}.x/linux-$fullver.tar.xz
+  tar -xf $HOME/linux-$fullver.tar.xz
+  cd $HOME/linux-$fullver
+  make tinyconfig
+  ./scripts/config --enable EFI_PARTITON
+  ./scripts/config --enable BLOCK
+  # BTRFS_FS is easiest config option to enable CONFIG_ZLIB_INFLATE|DEFLATE
+  ./scripts/config --enable BTRFS_FS
+  yes "" | make oldconfig
+  make prepare
+
+  cd $HOME/zfs
+  ./configure --with-linux=$HOME/linux-$fullver --enable-linux-builtin --enable-debug
+  ./copy-builtin $HOME/linux-$fullver
+
+  cd $HOME/linux-$fullver
+  ./scripts/config --enable ZFS
+  yes "" | make oldconfig
+  make -j `nproc`
+  ) &> /var/tmp/builtin.txt || rc=$?
+  echo "$rc" > /var/tmp/builtin-exitcode.txt
+  if [ "$rc" != "0" ] ; then
+      echo "$rc" > /var/tmp/tests-exitcode.txt
+  fi
+}
+export -f do_builtin_build
+
 # called directly on the runner
 if [ -z ${1:-} ]; then
   cd "/var/tmp"
@@ -49,8 +100,24 @@ if [ -z ${1:-} ]; then
 
   for ((i=1; i<=VMs; i++)); do
     IP="192.168.122.1$i"
+
+    # We do an additional test build of Lustre against ZFS if we're vm2
+    # on almalinux*.  At the time of writing, the vm2 tests were
+    # completing roughly 15min before the vm1 tests, so it makes sense
+    # to have vm2 do the build.
+    #
+    # In addition, we do an additional test build of ZFS as a Linux
+    # kernel built-in on Fedora.  Again, we do it on vm2 to exploit vm2's
+    # early finish time.
+    extra=""
+    if [[ "$OS" == almalinux* ]] && [[ "$i" == "2" ]] ; then
+        extra="--lustre"
+    elif [[ "$OS" == fedora* ]] && [[ "$i" == "2" ]] ; then
+        extra="--builtin"
+    fi
+
     daemonize -c /var/tmp -p vm${i}.pid -o vm${i}log.txt -- \
-      $SSH zfs@$IP $TESTS $OS $i $VMs $CI_TYPE
+      $SSH zfs@$IP $TESTS $OS $i $VMs $extra $CI_TYPE
     # handly line by line and add info prefix
     stdbuf -oL tail -fq vm${i}log.txt \
       | while read -r line; do prefix "$i" "$line"; done &
@@ -70,9 +137,35 @@ if [ -z ${1:-} ]; then
   exit 0
 fi
 
-# this part runs inside qemu vm
+
+#############################################
+# Everything from here on runs inside qemu vm
+#############################################
+
+# Process cmd line args
+OS="$1"
+shift
+NUM="$1"
+shift
+DEN="$1"
+shift
+
+BUILD_LUSTRE=0
+BUILD_BUILTIN=0
+if [ "$1" == "--lustre" ] ; then
+  BUILD_LUSTRE=1
+  shift
+elif [ "$1" == "--builtin" ] ; then
+  BUILD_BUILTIN=1
+  shift
+fi
+
+if [ "$1" == "quick" ] ; then
+  export RUNFILES="sanity.run"
+fi
+
 export PATH="$PATH:/bin:/sbin:/usr/bin:/usr/sbin:/usr/local/sbin:/usr/local/bin"
-case "$1" in
+case "$OS" in
   freebsd*)
     TDIR="/usr/local/share/zfs"
     sudo kldstat -n zfs 2>/dev/null && sudo kldunload zfs
@@ -95,23 +188,42 @@ case "$1" in
     ;;
 esac
 
-# enable io_uring on el9/el10
+# Distribution-specific settings.
-case "$1" in
+case "$OS" in
   almalinux9|almalinux10|centos-stream*)
+    # Enable io_uring on Enterprise Linux 9 and 10.
     sudo sysctl kernel.io_uring_disabled=0 > /dev/null
     ;;
+  alpine*)
+    # Ensure `/etc/zfs/zpool.cache` exists.
+    sudo mkdir -p /etc/zfs
+    sudo touch /etc/zfs/zpool.cache
+    sudo chmod 644 /etc/zfs/zpool.cache
+    ;;
 esac
 
+# Lustre calls a number of exported ZFS module symbols.  To make sure we don't
+# change the symbols and break Lustre, do a quick Lustre build of the latest
+# released Lustre against ZFS.
+#
+# Note that we do the Lustre test build in parallel with ZTS.  ZTS isn't very
+# CPU intensive, so we can use idle CPU cycles "guilt free" for the build.
+# The Lustre build on its own takes ~15min.
+if [ "$BUILD_LUSTRE" == "1" ] ; then
+  do_lustre_build &
+elif [ "$BUILD_BUILTIN" == "1" ] ; then
+  # Try building ZFS directly into the Linux kernel (not as a module)
+  do_builtin_build &
+fi
+
 # run functional testings and save exitcode
 cd /var/tmp
-TAGS=$2/$3
+TAGS=$NUM/$DEN
-if [ "$4" == "quick" ]; then
-  export RUNFILES="sanity.run"
-fi
 sudo dmesg -c > dmesg-prerun.txt
 mount > mount.txt
 df -h > df-prerun.txt
 $TDIR/zfs-tests.sh -vKO -s 3GB -T $TAGS
+
 RV=$?
 df -h > df-postrun.txt
 echo $RV > tests-exitcode.txt

.github/workflows/scripts/qemu-8-summary.sh

@@ -31,6 +31,12 @@ EOF
   rm -f tmp$$
 }
 
+function showfile_tail() {
+  echo "##[group]$2 (final lines)"
+  tail -n 80 $1
+  echo "##[endgroup]"
+}
+
 # overview
 cat /tmp/summary.txt
 echo ""
@@ -46,6 +52,32 @@ fi
 echo -e "\nFull logs for download:\n    $1\n"
 
 for ((i=1; i<=VMs; i++)); do
+
+  # Print Lustre build test results (the build is only done on vm2)
+  if [ -f vm$i/lustre-exitcode.txt ] ; then
+    rv=$(< vm$i/lustre-exitcode.txt)
+    if [ $rv = 0 ]; then
+      vm="vm$i"
+    else
+      vm="vm$i"
+      touch /tmp/have_failed_tests
+    fi
+    file="vm$i/lustre.txt"
+    test -s "$file" && showfile_tail "$file" "$vm: Lustre build"
+  fi
+
+  if [ -f vm$i/builtin-exitcode.txt ] ; then
+    rv=$(< vm$i/builtin-exitcode.txt)
+    if [ $rv = 0 ]; then
+      vm="vm$i"
+    else
+      vm="vm$i"
+      touch /tmp/have_failed_tests
+    fi
+    file="vm$i/builtin.txt"
+    test -s "$file" && showfile_tail "$file" "$vm: Linux built-in build"
+  fi
+
   rv=$(cat vm$i/tests-exitcode.txt)
 
   if [ $rv = 0 ]; then

.github/workflows/scripts/qemu-test-repo-vm.sh

@@ -4,7 +4,11 @@
 #
 # USAGE:
 #
-# 	./qemu-test-repo-vm [URL]
+# 	./qemu-test-repo-vm [--install] [URL]
+#
+# --lookup:	When testing a repo, only lookup the latest package versions,
+#		don't try to install them.  Installing all of them takes over
+#		an hour, so this is much quicker.
 #
 # URL:		URL to use instead of http://download.zfsonlinux.org
 #		If blank, use the default repo from zfs-release RPM.
@@ -15,6 +19,13 @@ source /etc/os-release
 OS="$ID"
 VERSION="$VERSION_ID"
 
+
+LOOKUP=""
+if [ -n "$1" ] && [ "$1" == "--lookup" ] ; then
+	LOOKUP=1
+	shift
+fi
+
 ALTHOST=""
 if [ -n "$1" ] ; then
 	ALTHOST="$1"
@@ -42,7 +53,19 @@ function test_install {
 		sudo sed -i "s;baseurl=http://download.zfsonlinux.org;baseurl=$host;g" /etc/yum.repos.d/zfs.repo
 	fi
 
-	sudo dnf -y install $args zfs zfs-test
+	baseurl=$(grep -A 5 "\[$repo\]" /etc/yum.repos.d/zfs.repo  | awk -F'=' '/baseurl=/{print $2; exit}')
+
+	# Just do a version lookup - don't try to install any RPMs
+	if [ "$LOOKUP" == "1" ] ; then
+		 package="$(dnf list $args zfs | tail -n 1 | awk '{print $2}')"
+		 echo "$repo ${package} $baseurl" >> $SUMMARY
+		 return
+	fi
+
+	if ! sudo dnf -y install $args zfs zfs-test ; then
+		echo "$repo ${package}...[FAILED] $baseurl" >> $SUMMARY
+		return
+	fi
 
 	# Load modules and create a simple pool as a sanity test.
 	sudo /usr/share/zfs/zfs.sh -r
@@ -51,7 +74,6 @@ function test_install {
 	sudo zpool status
 
 	# Print out repo name, rpm installed (kmod or dkms), and repo URL
-	baseurl=$(grep -A 5 "\[$repo\]" /etc/yum.repos.d/zfs.repo  | awk -F'=' '/baseurl=/{print $2; exit}')
 	package=$(sudo rpm -qa | grep zfs | grep -E 'kmod|dkms')
 
 	echo "$repo $package $baseurl" >> $SUMMARY
@@ -70,16 +92,19 @@ almalinux*)
 	name=$(curl -Ls $url | grep 'dnf install' | grep -Eo 'zfs-release-[0-9]+-[0-9]+')
 	sudo dnf -y install https://zfsonlinux.org/epel/$name$(rpm --eval "%{dist}").noarch.rpm 2>&1
 	sudo rpm -qi zfs-release
-	test_install zfs $ALTHOST
+	for i in zfs zfs-kmod zfs-testing zfs-testing-kmod zfs-latest \
-	test_install zfs-kmod $ALTHOST
+		zfs-latest-kmod zfs-legacy zfs-legacy-kmod zfs-2.2 \
-	test_install zfs-testing $ALTHOST
+		zfs-2.2-kmod zfs-2.3 zfs-2.3-kmod zfs-2.4 zfs-2.4-kmod; do
-	test_install zfs-testing-kmod $ALTHOST
+		test_install $i $ALTHOST
+	done
 	;;
 fedora*)
 	url='https://raw.githubusercontent.com/openzfs/openzfs-docs/refs/heads/master/docs/Getting%20Started/Fedora/index.rst'
 	name=$(curl -Ls $url | grep 'dnf install' | grep -Eo 'zfs-release-[0-9]+-[0-9]+')
 	sudo dnf -y install -y https://zfsonlinux.org/fedora/$name$(rpm --eval "%{dist}").noarch.rpm
-	test_install zfs $ALTHOST
+	for i in zfs zfs-latest zfs-legacy zfs-2.2 zfs-2.3 zfs-2.4 ; do
+		test_install $i $ALTHOST
+	done
 	;;
 esac
 echo "##[endgroup]"

.github/workflows/smatch.yml

@@ -0,0 +1,52 @@
+name: smatch
+
+on:
+  push:
+  pull_request:
+
+concurrency:
+  group: ${{ github.workflow }}-${{ github.head_ref || github.run_id }}
+  cancel-in-progress: true
+
+jobs:
+  smatch:
+    runs-on: ubuntu-24.04
+    steps:
+    - name: Checkout smatch
+      uses: actions/checkout@v4
+      with:
+        repository: error27/smatch
+        ref: master
+        path: smatch
+    - name: Install smatch dependencies
+      run: |
+        sudo apt-get install -y llvm gcc make sqlite3 libsqlite3-dev libdbd-sqlite3-perl libssl-dev libtry-tiny-perl
+    - name: Make smatch
+      run: |
+        cd $GITHUB_WORKSPACE/smatch
+        make -j$(nproc)
+    - name: Checkout OpenZFS
+      uses: actions/checkout@v4
+      with:
+        ref: ${{ github.event.pull_request.head.sha }}
+        path: zfs
+    - name: Install OpenZFS dependencies
+      run: |
+        cd $GITHUB_WORKSPACE/zfs
+        sudo apt-get purge -y snapd google-chrome-stable firefox
+        ONLY_DEPS=1 .github/workflows/scripts/qemu-3-deps-vm.sh ubuntu24
+    - name: Autogen.sh OpenZFS
+      run: |
+        cd $GITHUB_WORKSPACE/zfs
+        ./autogen.sh
+    - name: Configure OpenZFS
+      run: |
+        cd $GITHUB_WORKSPACE/zfs
+        ./configure --enable-debug
+    - name: Make OpenZFS
+      run: |
+        cd $GITHUB_WORKSPACE/zfs
+        make -j$(nproc) CHECK="$GITHUB_WORKSPACE/smatch/smatch" CC=$GITHUB_WORKSPACE/smatch/cgcc | tee $GITHUB_WORKSPACE/smatch.log
+    - name: Smatch results log
+      run: |
+        grep -E 'error:|warn:|warning:' $GITHUB_WORKSPACE/smatch.log

.github/workflows/zfs-qemu-packages.yml

@@ -42,6 +42,12 @@ on:
         required: false
         default: ""
         description: "(optional) repo URL (blank: use http://download.zfsonlinux.org)"
+      lookup:
+        type: boolean
+        required: false
+        default: false
+        description: "(optional) do version lookup only on repo test"
+
 concurrency:
   group: ${{ github.workflow }}-${{ github.head_ref || github.run_id }}
   cancel-in-progress: true
@@ -60,20 +66,16 @@ jobs:
         ref: ${{ github.event.pull_request.head.sha }}
 
     - name: Setup QEMU
-      timeout-minutes: 10
       run: .github/workflows/scripts/qemu-1-setup.sh
 
     - name: Start build machine
-      timeout-minutes: 10
       run: .github/workflows/scripts/qemu-2-start.sh ${{ matrix.os }}
 
     - name: Install dependencies
-      timeout-minutes: 20
       run: |
         .github/workflows/scripts/qemu-3-deps.sh ${{ matrix.os }}
 
     - name: Build modules or Test repo
-      timeout-minutes: 30
       run: |
         set -e
         if [ "${{ github.event.inputs.test_type }}" == "Test repo" ] ; then
@@ -81,7 +83,12 @@ jobs:
                 .github/workflows/scripts/qemu-prepare-for-build.sh
 
                 mkdir -p /tmp/repo
-                ssh zfs@vm0 '$HOME/zfs/.github/workflows/scripts/qemu-test-repo-vm.sh' ${{ github.event.inputs.repo_url }}
+                EXTRA=""
+                if [ "${{ github.event.inputs.lookup }}" == 'true' ] ; then
+                        EXTRA="--lookup"
+                fi
+
+                ssh zfs@vm0 '$HOME/zfs/.github/workflows/scripts/qemu-test-repo-vm.sh' $EXTRA ${{ github.event.inputs.repo_url }}
         else
                 EXTRA=""
                 if [ -n "${{ github.event.inputs.patch_level }}" ] ; then
@@ -94,7 +101,6 @@ jobs:
 
     - name: Prepare artifacts
       if: always()
-      timeout-minutes: 10
       run: |
         rsync -a zfs@vm0:/tmp/repo /tmp || true
         .github/workflows/scripts/replace-dupes-with-symlinks.sh /tmp/repo

.github/workflows/zfs-qemu.yml

@@ -10,6 +10,11 @@ on:
         required: false
         default: ""
         description: "(optional) Experimental kernel version to install on Fedora (like '6.14' or '6.13.3-0.rc3')"
+      specific_os:
+        type: string
+        required: false
+        default: ""
+        description: "(optional) Only run on this specific OS (like 'fedora42' or 'alpine3-23')"
 
 concurrency:
   group: ${{ github.workflow }}-${{ github.head_ref || github.run_id }}
@@ -58,6 +63,9 @@ jobs:
               # They specified a custom kernel version for Fedora.
               # Use only Fedora runners.
               os_json=$(echo ${os_selection} | jq -c '[.[] | select(startswith("fedora"))]')
+          elif ${{ github.event.inputs.specific_os != '' }}; then
+              # Use only the specified runner.
+              os_json=$(jq -cn --arg os "${{ github.event.inputs.specific_os }}" '[ $os ]')
           else
               # Normal case
               os_json=$(echo ${os_selection} | jq -c)
@@ -87,19 +95,15 @@ jobs:
         ref: ${{ github.event.pull_request.head.sha }}
 
     - name: Setup QEMU
-      timeout-minutes: 20
+      timeout-minutes: 60
-      run: |
+      run: .github/workflows/scripts/qemu-1-setup.sh
-        # Add a timestamp to each line to debug timeouts
-        while IFS=$'\n' read -r line; do
-            echo "$(date +'%H:%M:%S') $line"
-        done < <(.github/workflows/scripts/qemu-1-setup.sh)
 
     - name: Start build machine
       timeout-minutes: 10
       run: .github/workflows/scripts/qemu-2-start.sh ${{ matrix.os }}
 
     - name: Install dependencies
-      timeout-minutes: 20
+      timeout-minutes: 60
       run: .github/workflows/scripts/qemu-3-deps.sh ${{ matrix.os }} ${{ github.event.inputs.fedora_kernel_ver }}
 
     - name: Build modules

META

@@ -1,10 +1,10 @@
 Meta:          1
 Name:          zfs
 Branch:        1.0
-Version:       2.3.5
+Version:       2.3.6
 Release:       1
 Release-Tags:  relext
 License:       CDDL
 Author:        OpenZFS
-Linux-Maximum: 6.17
+Linux-Maximum: 6.19
 Linux-Minimum: 4.18

cmd/raidz_test/raidz_test.c

@@ -264,9 +264,21 @@ cmp_data(raidz_test_opts_t *opts, raidz_map_t *rm)
 
 static int
 init_rand(void *data, size_t size, void *private)
+{
+	size_t *offsetp = (size_t *)private;
+	size_t offset = *offsetp;
+
+	VERIFY3U(offset + size, <=, SPA_MAXBLOCKSIZE);
+	memcpy(data, (char *)rand_data + offset, size);
+	*offsetp = offset + size;
+	return (0);
+}
+
+static int
+corrupt_rand_fill(void *data, size_t size, void *private)
 {
 	(void) private;
-	memcpy(data, rand_data, size);
+	memset(data, 0xAA, size);
 	return (0);
 }
 
@@ -278,7 +290,7 @@ corrupt_colums(raidz_map_t *rm, const int *tgts, const int cnt)
 		for (int i = 0; i < cnt; i++) {
 			raidz_col_t *col = &rr->rr_col[tgts[i]];
 			abd_iterate_func(col->rc_abd, 0, col->rc_size,
-			    init_rand, NULL);
+			    corrupt_rand_fill, NULL);
 		}
 	}
 }
@@ -286,7 +298,8 @@ corrupt_colums(raidz_map_t *rm, const int *tgts, const int cnt)
 void
 init_zio_abd(zio_t *zio)
 {
-	abd_iterate_func(zio->io_abd, 0, zio->io_size, init_rand, NULL);
+	size_t offset = 0;
+	abd_iterate_func(zio->io_abd, 0, zio->io_size, init_rand, &offset);
 }
 
 static void
@@ -373,7 +386,7 @@ init_raidz_map(raidz_test_opts_t *opts, zio_t **zio, const int parity)
 
 	*zio = umem_zalloc(sizeof (zio_t), UMEM_NOFAIL);
 
-	(*zio)->io_offset = 0;
+	(*zio)->io_offset = opts->rto_offset;
 	(*zio)->io_size = alloc_dsize;
 	(*zio)->io_abd = raidz_alloc(alloc_dsize);
 	init_zio_abd(*zio);
@@ -834,6 +847,8 @@ main(int argc, char **argv)
 		err = run_test(NULL);
 	}
 
+	mprotect(rand_data, SPA_MAXBLOCKSIZE, PROT_READ | PROT_WRITE);
+
 	umem_free(rand_data, SPA_MAXBLOCKSIZE);
 	kernel_fini();
 

cmd/raidz_test/raidz_test.h

@@ -72,7 +72,7 @@ typedef struct raidz_test_opts {
 
 static const raidz_test_opts_t rto_opts_defaults = {
 	.rto_ashift = 9,
-	.rto_offset = 1ULL << 0,
+	.rto_offset = 0,
 	.rto_dcols = 8,
 	.rto_dsize = 1<<19,
 	.rto_v = D_ALL,

cmd/zdb/zdb.c

@@ -107,7 +107,9 @@ extern uint_t zfs_reconstruct_indirect_combinations_max;
 extern uint_t zfs_btree_verify_intensity;
 
 static const char cmdname[] = "zdb";
-uint8_t dump_opt[256];
+uint8_t dump_opt[512];
+
+#define	ALLOCATED_OPT	256
 
 typedef void object_viewer_t(objset_t *, uint64_t, void *data, size_t size);
 
@@ -1651,6 +1653,16 @@ dump_metaslab_stats(metaslab_t *msp)
 	dump_histogram(rt->rt_histogram, ZFS_RANGE_TREE_HISTOGRAM_SIZE, 0);
 }
 
+static void
+dump_allocated(void *arg, uint64_t start, uint64_t size)
+{
+	uint64_t *off = arg;
+	if (*off != start)
+		(void) printf("ALLOC: %"PRIu64" %"PRIu64"\n", *off,
+		    start - *off);
+	*off = start + size;
+}
+
 static void
 dump_metaslab(metaslab_t *msp)
 {
@@ -1667,13 +1679,24 @@ dump_metaslab(metaslab_t *msp)
 	    (u_longlong_t)msp->ms_id, (u_longlong_t)msp->ms_start,
 	    (u_longlong_t)space_map_object(sm), freebuf);
 
-	if (dump_opt['m'] > 2 && !dump_opt['L']) {
+	if (dump_opt[ALLOCATED_OPT] ||
+	    (dump_opt['m'] > 2 && !dump_opt['L'])) {
 		mutex_enter(&msp->ms_lock);
 		VERIFY0(metaslab_load(msp));
+	}
+
+	if (dump_opt['m'] > 2 && !dump_opt['L']) {
 		zfs_range_tree_stat_verify(msp->ms_allocatable);
 		dump_metaslab_stats(msp);
-		metaslab_unload(msp);
+	}
-		mutex_exit(&msp->ms_lock);
+
+	if (dump_opt[ALLOCATED_OPT]) {
+		uint64_t off = msp->ms_start;
+		zfs_range_tree_walk(msp->ms_allocatable, dump_allocated,
+		    &off);
+		if (off != msp->ms_start + msp->ms_size)
+			(void) printf("ALLOC: %"PRIu64" %"PRIu64"\n", off,
+			    msp->ms_size - off);
 	}
 
 	if (dump_opt['m'] > 1 && sm != NULL &&
@@ -1688,6 +1711,12 @@ dump_metaslab(metaslab_t *msp)
 		    SPACE_MAP_HISTOGRAM_SIZE, sm->sm_shift);
 	}
 
+	if (dump_opt[ALLOCATED_OPT] ||
+	    (dump_opt['m'] > 2 && !dump_opt['L'])) {
+		metaslab_unload(msp);
+		mutex_exit(&msp->ms_lock);
+	}
+
 	if (vd->vdev_ops == &vdev_draid_ops)
 		ASSERT3U(msp->ms_size, <=, 1ULL << vd->vdev_ms_shift);
 	else
@@ -1724,8 +1753,9 @@ print_vdev_metaslab_header(vdev_t *vd)
 		}
 	}
 
-	(void) printf("\tvdev %10llu   %s",
+	(void) printf("\tvdev %10llu\t%s  metaslab shift %4llu",
-	    (u_longlong_t)vd->vdev_id, bias_str);
+	    (u_longlong_t)vd->vdev_id, bias_str,
+	    (u_longlong_t)vd->vdev_ms_shift);
 
 	if (ms_flush_data_obj != 0) {
 		(void) printf("   ms_unflushed_phys object %llu",
@@ -9318,6 +9348,8 @@ main(int argc, char **argv)
 		{"all-reconstruction",	no_argument,		NULL, 'Y'},
 		{"livelist",		no_argument,		NULL, 'y'},
 		{"zstd-headers",	no_argument,		NULL, 'Z'},
+		{"allocated-map",	no_argument,		NULL,
+		    ALLOCATED_OPT},
 		{0, 0, 0, 0}
 	};
 
@@ -9348,6 +9380,7 @@ main(int argc, char **argv)
 		case 'u':
 		case 'y':
 		case 'Z':
+		case ALLOCATED_OPT:
 			dump_opt[c]++;
 			dump_all = 0;
 			break;

cmd/zdb/zdb.h

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

cmd/zdb/zdb_il.c

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

cmd/zhack.c

@@ -52,12 +52,15 @@
 #include <sys/zio_compress.h>
 #include <sys/zfeature.h>
 #include <sys/dmu_tx.h>
+#include <sys/backtrace.h>
 #include <zfeature_common.h>
 #include <libzutil.h>
+#include <sys/metaslab_impl.h>
 
 static importargs_t g_importargs;
 static char *g_pool;
 static boolean_t g_readonly;
+static boolean_t g_dump_dbgmsg;
 
 typedef enum {
 	ZHACK_REPAIR_OP_UNKNOWN  = 0,
@@ -69,11 +72,23 @@ static __attribute__((noreturn)) void
 usage(void)
 {
 	(void) fprintf(stderr,
-	    "Usage: zhack [-c cachefile] [-d dir] <subcommand> <args> ...\n"
+	    "Usage: zhack [-o tunable] [-c cachefile] [-d dir] [-G] "
-	    "where <subcommand> <args> is one of the following:\n"
+	    "<subcommand> <args> ...\n"
+	    "       where <subcommand> <args> is one of the following:\n"
 	    "\n");
 
 	(void) fprintf(stderr,
+	    "    global options:\n"
+	    "    -c <cachefile>   reads config from the given cachefile\n"
+	    "    -d <dir>         directory with vdevs for import\n"
+	    "    -o var=value...  set global variable to an unsigned "
+	    "32-bit integer\n"
+	    "    -G               dump zfs_dbgmsg buffer before exiting\n"
+	    "\n"
+	    "    action idle <pool> [-f] [-t seconds]\n"
+	    "        import the pool for a set time then export it\n"
+	    "        -t <seconds> sets the time the pool is imported\n"
+	    "\n"
 	    "    feature stat <pool>\n"
 	    "        print information about enabled features\n"
 	    "    feature enable [-r] [-d desc] <pool> <feature>\n"
@@ -93,10 +108,46 @@ usage(void)
 	    "        -c repair corrupted label checksums\n"
 	    "        -u restore the label on a detached device\n"
 	    "\n"
-	    "    <device> : path to vdev\n");
+	    "    <device> : path to vdev\n"
+	    "\n"
+	    "    metaslab leak <pool>\n"
+	    "        apply allocation map from zdb to specified pool\n");
 	exit(1);
 }
 
+static void
+dump_debug_buffer(void)
+{
+	ssize_t ret __attribute__((unused));
+
+	if (!g_dump_dbgmsg)
+		return;
+
+	/*
+	 * We use write() instead of printf() so that this function
+	 * is safe to call from a signal handler.
+	 */
+	ret = write(STDERR_FILENO, "\n", 1);
+	zfs_dbgmsg_print(STDERR_FILENO, "zhack");
+}
+
+static void sig_handler(int signo)
+{
+	struct sigaction action;
+
+	libspl_backtrace(STDERR_FILENO);
+	dump_debug_buffer();
+
+	/*
+	 * Restore default action and re-raise signal so SIGSEGV and
+	 * SIGABRT can trigger a core dump.
+	 */
+	action.sa_handler = SIG_DFL;
+	sigemptyset(&action.sa_mask);
+	action.sa_flags = 0;
+	(void) sigaction(signo, &action, NULL);
+	raise(signo);
+}
 
 static __attribute__((format(printf, 3, 4))) __attribute__((noreturn)) void
 fatal(spa_t *spa, const void *tag, const char *fmt, ...)
@@ -114,6 +165,8 @@ fatal(spa_t *spa, const void *tag, const char *fmt, ...)
 	va_end(ap);
 	(void) fputc('\n', stderr);
 
+	dump_debug_buffer();
+
 	exit(1);
 }
 
@@ -169,7 +222,7 @@ zhack_import(char *target, boolean_t readonly)
 
 	zfeature_checks_disable = B_TRUE;
 	error = spa_import(target, config, props,
-	    (readonly ?  ZFS_IMPORT_SKIP_MMP : ZFS_IMPORT_NORMAL));
+	    (readonly ? ZFS_IMPORT_SKIP_MMP : ZFS_IMPORT_NORMAL));
 	fnvlist_free(config);
 	zfeature_checks_disable = B_FALSE;
 	if (error == EEXIST)
@@ -500,6 +553,259 @@ zhack_do_feature(int argc, char **argv)
 	return (0);
 }
 
+static void
+zhack_do_action_idle(int argc, char **argv)
+{
+	spa_t *spa;
+	char *target, *tmp;
+	int idle_time = 0;
+	int c;
+
+	optind = 1;
+	while ((c = getopt(argc, argv, "+t:")) != -1) {
+		switch (c) {
+		case 't':
+			idle_time = strtol(optarg, &tmp, 0);
+			if (*tmp) {
+				(void) fprintf(stderr, "error: time must "
+				    "be an integer in seconds: %s\n", tmp);
+				usage();
+			}
+			if (idle_time < 0) {
+				(void) fprintf(stderr, "error: time must "
+				    "not be negative: %d\n", idle_time);
+				usage();
+			}
+			break;
+		default:
+			usage();
+			break;
+		}
+	}
+	argc -= optind;
+	argv += optind;
+
+	if (argc < 1) {
+		(void) fprintf(stderr, "error: missing pool name\n");
+		usage();
+	}
+	target = argv[0];
+
+	zhack_spa_open(target, B_FALSE, FTAG, &spa);
+
+	fprintf(stdout, "Imported pool %s, idle for %d seconds\n",
+	    target, idle_time);
+	sleep(idle_time);
+
+	spa_close(spa, FTAG);
+}
+
+static int
+zhack_do_action(int argc, char **argv)
+{
+	char *subcommand;
+
+	argc--;
+	argv++;
+	if (argc == 0) {
+		(void) fprintf(stderr,
+		    "error: no import operation specified\n");
+		usage();
+	}
+
+	subcommand = argv[0];
+	if (strcmp(subcommand, "idle") == 0) {
+		zhack_do_action_idle(argc, argv);
+	} else {
+		(void) fprintf(stderr, "error: unknown subcommand: %s\n",
+		    subcommand);
+		usage();
+	}
+
+	return (0);
+}
+
+
+static boolean_t
+strstarts(const char *a, const char *b)
+{
+	return (strncmp(a, b, strlen(b)) == 0);
+}
+
+static void
+metaslab_force_alloc(metaslab_t *msp, uint64_t start, uint64_t size,
+    dmu_tx_t *tx)
+{
+	ASSERT(msp->ms_disabled);
+	ASSERT(MUTEX_HELD(&msp->ms_lock));
+	uint64_t txg = dmu_tx_get_txg(tx);
+
+	uint64_t off = start;
+	while (off < start + size) {
+		uint64_t ostart, osize;
+		boolean_t found = zfs_range_tree_find_in(msp->ms_allocatable,
+		    off, start + size - off, &ostart, &osize);
+		if (!found)
+			break;
+		zfs_range_tree_remove(msp->ms_allocatable, ostart, osize);
+
+		if (zfs_range_tree_is_empty(msp->ms_allocating[txg & TXG_MASK]))
+			vdev_dirty(msp->ms_group->mg_vd, VDD_METASLAB, msp,
+			    txg);
+
+		zfs_range_tree_add(msp->ms_allocating[txg & TXG_MASK], ostart,
+		    osize);
+		msp->ms_allocating_total += osize;
+		off = ostart + osize;
+	}
+}
+
+static void
+zhack_do_metaslab_leak(int argc, char **argv)
+{
+	int c;
+	char *target;
+	spa_t *spa;
+
+	optind = 1;
+	boolean_t force = B_FALSE;
+	while ((c = getopt(argc, argv, "f")) != -1) {
+		switch (c) {
+		case 'f':
+			force = B_TRUE;
+			break;
+		default:
+			usage();
+			break;
+		}
+	}
+
+	argc -= optind;
+	argv += optind;
+
+	if (argc < 1) {
+		(void) fprintf(stderr, "error: missing pool name\n");
+		usage();
+	}
+	target = argv[0];
+
+	zhack_spa_open(target, B_FALSE, FTAG, &spa);
+	spa_config_enter(spa, SCL_VDEV | SCL_ALLOC, FTAG, RW_READER);
+
+	char *line = NULL;
+	size_t cap = 0;
+
+	vdev_t *vd = NULL;
+	metaslab_t *prev = NULL;
+	dmu_tx_t *tx = NULL;
+	while (getline(&line, &cap, stdin) > 0) {
+		if (strstarts(line, "\tvdev ")) {
+			uint64_t vdev_id, ms_shift;
+			if (sscanf(line,
+			    "\tvdev %10"PRIu64"\t%*s  metaslab shift %4"PRIu64,
+			    &vdev_id, &ms_shift) == 1) {
+				VERIFY3U(sscanf(line, "\tvdev %"PRIu64
+				    "\t  metaslab shift %4"PRIu64,
+				    &vdev_id, &ms_shift), ==, 2);
+			}
+			vd = vdev_lookup_top(spa, vdev_id);
+			if (vd == NULL) {
+				fprintf(stderr, "error: no such vdev with "
+				    "id %"PRIu64"\n", vdev_id);
+				break;
+			}
+			if (tx) {
+				dmu_tx_commit(tx);
+				mutex_exit(&prev->ms_lock);
+				metaslab_enable(prev, B_FALSE, B_FALSE);
+				tx = NULL;
+				prev = NULL;
+			}
+			if (vd->vdev_ms_shift != ms_shift) {
+				fprintf(stderr, "error: ms_shift mismatch: %"
+				    PRIu64" != %"PRIu64"\n", vd->vdev_ms_shift,
+				    ms_shift);
+				break;
+			}
+		} else if (strstarts(line, "\tmetaslabs ")) {
+			uint64_t ms_count;
+			VERIFY3U(sscanf(line, "\tmetaslabs %"PRIu64, &ms_count),
+			    ==, 1);
+			ASSERT(vd);
+			if (!force && vd->vdev_ms_count != ms_count) {
+				fprintf(stderr, "error: ms_count mismatch: %"
+				    PRIu64" != %"PRIu64"\n", vd->vdev_ms_count,
+				    ms_count);
+				break;
+			}
+		} else if (strstarts(line, "ALLOC:")) {
+			uint64_t start, size;
+			VERIFY3U(sscanf(line, "ALLOC: %"PRIu64" %"PRIu64"\n",
+			    &start, &size), ==, 2);
+
+			ASSERT(vd);
+			metaslab_t *cur =
+			    vd->vdev_ms[start >> vd->vdev_ms_shift];
+			if (prev != cur) {
+				if (prev) {
+					dmu_tx_commit(tx);
+					mutex_exit(&prev->ms_lock);
+					metaslab_enable(prev, B_FALSE, B_FALSE);
+				}
+				ASSERT(cur);
+				metaslab_disable(cur);
+				mutex_enter(&cur->ms_lock);
+				metaslab_load(cur);
+				prev = cur;
+				tx = dmu_tx_create_dd(
+				    spa_get_dsl(vd->vdev_spa)->dp_root_dir);
+				dmu_tx_assign(tx, DMU_TX_WAIT);
+			}
+
+			metaslab_force_alloc(cur, start, size, tx);
+		} else {
+			continue;
+		}
+	}
+	if (tx) {
+		dmu_tx_commit(tx);
+		mutex_exit(&prev->ms_lock);
+		metaslab_enable(prev, B_FALSE, B_FALSE);
+		tx = NULL;
+		prev = NULL;
+	}
+	if (line)
+		free(line);
+
+	spa_config_exit(spa, SCL_VDEV | SCL_ALLOC, FTAG);
+	spa_close(spa, FTAG);
+}
+
+static int
+zhack_do_metaslab(int argc, char **argv)
+{
+	char *subcommand;
+
+	argc--;
+	argv++;
+	if (argc == 0) {
+		(void) fprintf(stderr,
+		    "error: no metaslab operation specified\n");
+		usage();
+	}
+
+	subcommand = argv[0];
+	if (strcmp(subcommand, "leak") == 0) {
+		zhack_do_metaslab_leak(argc, argv);
+	} else {
+		(void) fprintf(stderr, "error: unknown subcommand: %s\n",
+		    subcommand);
+		usage();
+	}
+
+	return (0);
+}
+
 #define	ASHIFT_UBERBLOCK_SHIFT(ashift)	\
 	MIN(MAX(ashift, UBERBLOCK_SHIFT), \
 	MAX_UBERBLOCK_SHIFT)
@@ -975,17 +1281,35 @@ zhack_do_label(int argc, char **argv)
 int
 main(int argc, char **argv)
 {
+	struct sigaction action;
 	char *path[MAX_NUM_PATHS];
 	const char *subcommand;
 	int rv = 0;
 	int c;
 
+	/*
+	 * Set up signal handlers, so if we crash due to bad on-disk data we
+	 * can get more info. Unlike ztest, we don't bail out if we can't set
+	 * up signal handlers, because zhack is very useful without them.
+	 */
+	action.sa_handler = sig_handler;
+	sigemptyset(&action.sa_mask);
+	action.sa_flags = 0;
+	if (sigaction(SIGSEGV, &action, NULL) < 0) {
+		(void) fprintf(stderr, "zhack: cannot catch SIGSEGV: %s\n",
+		    strerror(errno));
+	}
+	if (sigaction(SIGABRT, &action, NULL) < 0) {
+		(void) fprintf(stderr, "zhack: cannot catch SIGABRT: %s\n",
+		    strerror(errno));
+	}
+
 	g_importargs.path = path;
 
 	dprintf_setup(&argc, argv);
 	zfs_prop_init();
 
-	while ((c = getopt(argc, argv, "+c:d:")) != -1) {
+	while ((c = getopt(argc, argv, "+c:d:Go:")) != -1) {
 		switch (c) {
 		case 'c':
 			g_importargs.cachefile = optarg;
@@ -994,6 +1318,13 @@ main(int argc, char **argv)
 			assert(g_importargs.paths < MAX_NUM_PATHS);
 			g_importargs.path[g_importargs.paths++] = optarg;
 			break;
+		case 'G':
+			g_dump_dbgmsg = B_TRUE;
+			break;
+		case 'o':
+			if (set_global_var(optarg) != 0)
+				exit(1);
+			break;
 		default:
 			usage();
 			break;
@@ -1011,10 +1342,14 @@ main(int argc, char **argv)
 
 	subcommand = argv[0];
 
-	if (strcmp(subcommand, "feature") == 0) {
+	if (strcmp(subcommand, "action") == 0) {
+		rv = zhack_do_action(argc, argv);
+	} else if (strcmp(subcommand, "feature") == 0) {
 		rv = zhack_do_feature(argc, argv);
 	} else if (strcmp(subcommand, "label") == 0) {
 		return (zhack_do_label(argc, argv));
+	} else if (strcmp(subcommand, "metaslab") == 0) {
+		rv = zhack_do_metaslab(argc, argv);
 	} else {
 		(void) fprintf(stderr, "error: unknown subcommand: %s\n",
 		    subcommand);
@@ -1026,6 +1361,9 @@ main(int argc, char **argv)
 		    "changes may not be committed to disk\n");
 	}
 
+	if (g_dump_dbgmsg)
+		dump_debug_buffer();
+
 	kernel_fini();
 
 	return (rv);

cmd/zpool/zpool_main.c

@@ -3883,6 +3883,9 @@ do_import(nvlist_t *config, const char *newname, const char *mntopts,
 			    hostid, ctime(&timestamp));
 		}
 
+		if (getenv("ZFS_LOAD_INFO_DEBUG"))
+			dump_nvlist(nvinfo, 4);
+
 		return (1);
 	}
 

cmd/zpool/zpool_vdev.c

@@ -270,14 +270,13 @@ is_spare(nvlist_t *config, const char *path)
  *	draid*		Virtual dRAID spare
  */
 static nvlist_t *
-make_leaf_vdev(nvlist_t *props, const char *arg, boolean_t is_primary)
+make_leaf_vdev(const char *arg, boolean_t is_primary, uint64_t ashift)
 {
 	char path[MAXPATHLEN];
 	struct stat64 statbuf;
 	nvlist_t *vdev = NULL;
 	const char *type = NULL;
 	boolean_t wholedisk = B_FALSE;
-	uint64_t ashift = 0;
 	int err;
 
 	/*
@@ -381,31 +380,6 @@ make_leaf_vdev(nvlist_t *props, const char *arg, boolean_t is_primary)
 		verify(nvlist_add_uint64(vdev, ZPOOL_CONFIG_WHOLE_DISK,
 		    (uint64_t)wholedisk) == 0);
 
-	/*
-	 * Override defaults if custom properties are provided.
-	 */
-	if (props != NULL) {
-		const char *value = NULL;
-
-		if (nvlist_lookup_string(props,
-		    zpool_prop_to_name(ZPOOL_PROP_ASHIFT), &value) == 0) {
-			if (zfs_nicestrtonum(NULL, value, &ashift) != 0) {
-				(void) fprintf(stderr,
-				    gettext("ashift must be a number.\n"));
-				return (NULL);
-			}
-			if (ashift != 0 &&
-			    (ashift < ASHIFT_MIN || ashift > ASHIFT_MAX)) {
-				(void) fprintf(stderr,
-				    gettext("invalid 'ashift=%" PRIu64 "' "
-				    "property: only values between %" PRId32 " "
-				    "and %" PRId32 " are allowed.\n"),
-				    ashift, ASHIFT_MIN, ASHIFT_MAX);
-				return (NULL);
-			}
-		}
-	}
-
 	/*
 	 * If the device is known to incorrectly report its physical sector
 	 * size explicitly provide the known correct value.
@@ -1502,6 +1476,29 @@ construct_spec(nvlist_t *props, int argc, char **argv)
 	const char *type, *fulltype;
 	boolean_t is_log, is_special, is_dedup, is_spare;
 	boolean_t seen_logs;
+	uint64_t ashift = 0;
+
+	if (props != NULL) {
+		const char *value = NULL;
+
+		if (nvlist_lookup_string(props,
+		    zpool_prop_to_name(ZPOOL_PROP_ASHIFT), &value) == 0) {
+			if (zfs_nicestrtonum(NULL, value, &ashift) != 0) {
+				(void) fprintf(stderr,
+				    gettext("ashift must be a number.\n"));
+				return (NULL);
+			}
+			if (ashift != 0 &&
+			    (ashift < ASHIFT_MIN || ashift > ASHIFT_MAX)) {
+				(void) fprintf(stderr,
+				    gettext("invalid 'ashift=%" PRIu64 "' "
+				    "property: only values between %" PRId32 " "
+				    "and %" PRId32 " are allowed.\n"),
+				    ashift, ASHIFT_MIN, ASHIFT_MAX);
+				return (NULL);
+			}
+		}
+	}
 
 	top = NULL;
 	toplevels = 0;
@@ -1608,9 +1605,9 @@ construct_spec(nvlist_t *props, int argc, char **argv)
 				    children * sizeof (nvlist_t *));
 				if (child == NULL)
 					zpool_no_memory();
-				if ((nv = make_leaf_vdev(props, argv[c],
+				if ((nv = make_leaf_vdev(argv[c],
 				    !(is_log || is_special || is_dedup ||
-				    is_spare))) == NULL) {
+				    is_spare), ashift)) == NULL) {
 					for (c = 0; c < children - 1; c++)
 						nvlist_free(child[c]);
 					free(child);
@@ -1674,6 +1671,10 @@ construct_spec(nvlist_t *props, int argc, char **argv)
 					    ZPOOL_CONFIG_ALLOCATION_BIAS,
 					    VDEV_ALLOC_BIAS_DEDUP) == 0);
 				}
+				if (ashift > 0) {
+					fnvlist_add_uint64(nv,
+					    ZPOOL_CONFIG_ASHIFT, ashift);
+				}
 				if (strcmp(type, VDEV_TYPE_RAIDZ) == 0) {
 					verify(nvlist_add_uint64(nv,
 					    ZPOOL_CONFIG_NPARITY,
@@ -1701,8 +1702,9 @@ construct_spec(nvlist_t *props, int argc, char **argv)
 			 * We have a device.  Pass off to make_leaf_vdev() to
 			 * construct the appropriate nvlist describing the vdev.
 			 */
-			if ((nv = make_leaf_vdev(props, argv[0], !(is_log ||
+			if ((nv = make_leaf_vdev(argv[0], !(is_log ||
-			    is_special || is_dedup || is_spare))) == NULL)
+			    is_special || is_dedup || is_spare),
+			    ashift)) == NULL)
 				goto spec_out;
 
 			verify(nvlist_add_uint64(nv,

config/kernel-blkdev.m4

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

config/kernel-block-device-operations.m4

@@ -119,15 +119,49 @@ AC_DEFUN([ZFS_AC_KERNEL_BLOCK_DEVICE_OPERATIONS_REVALIDATE_DISK], [
 	])
 ])
 
+dnl #
+dnl # 6.18 API change
+dnl # block_device_operation->getgeo takes struct gendisk* as first arg
+dnl #
+AC_DEFUN([ZFS_AC_KERNEL_SRC_BLOCK_DEVICE_OPERATIONS_GETGEO_GENDISK], [
+	ZFS_LINUX_TEST_SRC([block_device_operations_getgeo_gendisk], [
+		#include <linux/blkdev.h>
+
+		static int blk_getgeo(struct gendisk *disk, struct hd_geometry *geo)
+		{
+			(void) disk, (void) geo;
+			return (0);
+		}
+
+		static const struct block_device_operations
+		    bops __attribute__ ((unused)) = {
+			.getgeo	= blk_getgeo,
+		};
+	], [], [])
+])
+
+AC_DEFUN([ZFS_AC_KERNEL_BLOCK_DEVICE_OPERATIONS_GETGEO_GENDISK], [
+	AC_MSG_CHECKING([whether bops->getgeo() takes gendisk as first arg])
+	ZFS_LINUX_TEST_RESULT([block_device_operations_getgeo_gendisk], [
+		AC_MSG_RESULT(yes)
+		AC_DEFINE([HAVE_BLOCK_DEVICE_OPERATIONS_GETGEO_GENDISK], [1],
+			[Define if getgeo() in block_device_operations takes struct gendisk * as its first arg])
+	],[
+		AC_MSG_RESULT(no)
+	])
+])
+
 AC_DEFUN([ZFS_AC_KERNEL_SRC_BLOCK_DEVICE_OPERATIONS], [
 	ZFS_AC_KERNEL_SRC_BLOCK_DEVICE_OPERATIONS_CHECK_EVENTS
 	ZFS_AC_KERNEL_SRC_BLOCK_DEVICE_OPERATIONS_RELEASE_VOID
 	ZFS_AC_KERNEL_SRC_BLOCK_DEVICE_OPERATIONS_RELEASE_1ARG
 	ZFS_AC_KERNEL_SRC_BLOCK_DEVICE_OPERATIONS_REVALIDATE_DISK
+	ZFS_AC_KERNEL_SRC_BLOCK_DEVICE_OPERATIONS_GETGEO_GENDISK
 ])
 
 AC_DEFUN([ZFS_AC_KERNEL_BLOCK_DEVICE_OPERATIONS], [
 	ZFS_AC_KERNEL_BLOCK_DEVICE_OPERATIONS_CHECK_EVENTS
 	ZFS_AC_KERNEL_BLOCK_DEVICE_OPERATIONS_RELEASE_VOID
 	ZFS_AC_KERNEL_BLOCK_DEVICE_OPERATIONS_REVALIDATE_DISK
+	ZFS_AC_KERNEL_BLOCK_DEVICE_OPERATIONS_GETGEO_GENDISK
 ])

config/kernel-dentry-operations.m4

@@ -46,14 +46,37 @@ AC_DEFUN([ZFS_AC_KERNEL_D_SET_D_OP], [
 	])
 ])
 
+dnl #
+dnl # 6.17 API change
+dnl # sb->s_d_op removed; set_default_d_op(sb, dop) added
+dnl #
+AC_DEFUN([ZFS_AC_KERNEL_SRC_SET_DEFAULT_D_OP], [
+	ZFS_LINUX_TEST_SRC([set_default_d_op], [
+		#include <linux/dcache.h>
+	], [
+		set_default_d_op(NULL, NULL);
+	])
+])
+
+AC_DEFUN([ZFS_AC_KERNEL_SET_DEFAULT_D_OP], [
+	AC_MSG_CHECKING([whether set_default_d_op() is available])
+	ZFS_LINUX_TEST_RESULT([set_default_d_op], [
+		AC_MSG_RESULT(yes)
+		AC_DEFINE(HAVE_SET_DEFAULT_D_OP, 1,
+		    [Define if set_default_d_op() is available])
+	], [
+		AC_MSG_RESULT(no)
+	])
+])
+
 AC_DEFUN([ZFS_AC_KERNEL_SRC_DENTRY], [
         ZFS_AC_KERNEL_SRC_D_OBTAIN_ALIAS
         ZFS_AC_KERNEL_SRC_D_SET_D_OP
-        ZFS_AC_KERNEL_SRC_S_D_OP
+        ZFS_AC_KERNEL_SRC_SET_DEFAULT_D_OP
 ])
 
 AC_DEFUN([ZFS_AC_KERNEL_DENTRY], [
         ZFS_AC_KERNEL_D_OBTAIN_ALIAS
         ZFS_AC_KERNEL_D_SET_D_OP
-        ZFS_AC_KERNEL_S_D_OP
+        ZFS_AC_KERNEL_SET_DEFAULT_D_OP
 ])

config/kernel-drop-inode.m4

@@ -0,0 +1,24 @@
+dnl #
+dnl # 6.18 API change
+dnl # - generic_drop_inode() renamed to inode_generic_drop()
+dnl # - generic_delete_inode() renamed to inode_just_drop()
+dnl #
+AC_DEFUN([ZFS_AC_KERNEL_SRC_INODE_GENERIC_DROP], [
+	ZFS_LINUX_TEST_SRC([inode_generic_drop], [
+		#include <linux/fs.h>
+	],[
+		struct inode *ip = NULL;
+		inode_generic_drop(ip);
+	])
+])
+
+AC_DEFUN([ZFS_AC_KERNEL_INODE_GENERIC_DROP], [
+	AC_MSG_CHECKING([whether inode_generic_drop() exists])
+	ZFS_LINUX_TEST_RESULT([inode_generic_drop], [
+		AC_MSG_RESULT(yes)
+		AC_DEFINE(HAVE_INODE_GENERIC_DROP, 1,
+			[inode_generic_drop() exists])
+	],[
+		AC_MSG_RESULT(no)
+	])
+])

config/kernel-inode-state.m4

@@ -0,0 +1,23 @@
+dnl #
+dnl # 6.19 API change. inode->i_state no longer accessible directly; helper
+dnl # functions exist.
+dnl #
+AC_DEFUN([ZFS_AC_KERNEL_SRC_INODE_STATE_READ_ONCE], [
+	ZFS_LINUX_TEST_SRC([inode_state_read_once], [
+		#include <linux/fs.h>
+	], [
+		struct inode i = {};
+		inode_state_read_once(&i);
+	],[])
+])
+
+AC_DEFUN([ZFS_AC_KERNEL_INODE_STATE_READ_ONCE], [
+	AC_MSG_CHECKING([whether inode_state_read_once() exists])
+	ZFS_LINUX_TEST_RESULT([inode_state_read_once], [
+		AC_MSG_RESULT(yes)
+		AC_DEFINE(HAVE_INODE_STATE_READ_ONCE, 1,
+		    [inode_state_read_once() exists])
+	],[
+		AC_MSG_RESULT(no)
+	])
+])

config/kernel-kmap-atomic-args.m4

@@ -7,7 +7,7 @@ AC_DEFUN([ZFS_AC_KERNEL_SRC_KMAP_ATOMIC_ARGS], [
 	ZFS_LINUX_TEST_SRC([kmap_atomic], [
 		#include <linux/pagemap.h>
 	],[
-		struct page page;
+		struct page page = {};
 		kmap_atomic(&page);
 	])
 ])

config/kernel-mm-page-flags.m4

@@ -16,9 +16,36 @@ AC_DEFUN([ZFS_AC_KERNEL_MM_PAGE_FLAG_ERROR], [
 	])
 ])
 
+dnl #
+dnl # Linux 6.18+ uses a struct typedef (memdesc_flags_t) instead of an
+dnl # 'unsigned long' for the 'flags' field in 'struct page'.
+dnl #
+AC_DEFUN([ZFS_AC_KERNEL_SRC_MM_PAGE_FLAGS_STRUCT], [
+	ZFS_LINUX_TEST_SRC([mm_page_flags_struct], [
+		#include <linux/mm.h>
+
+		static const struct page p __attribute__ ((unused)) = {
+			.flags = { .f = 0 }
+		};
+	],[])
+])
+
+AC_DEFUN([ZFS_AC_KERNEL_MM_PAGE_FLAGS_STRUCT], [
+	AC_MSG_CHECKING([whether 'flags' in 'struct page' is a struct])
+	ZFS_LINUX_TEST_RESULT([mm_page_flags_struct], [
+		AC_MSG_RESULT([yes])
+		AC_DEFINE(HAVE_MM_PAGE_FLAGS_STRUCT, 1,
+			['flags' in 'struct page' is a struct])
+	],[
+		AC_MSG_RESULT([no])
+	])
+])
+
 AC_DEFUN([ZFS_AC_KERNEL_SRC_MM_PAGE_FLAGS], [
 	ZFS_AC_KERNEL_SRC_MM_PAGE_FLAG_ERROR
+	ZFS_AC_KERNEL_SRC_MM_PAGE_FLAGS_STRUCT
 ])
 AC_DEFUN([ZFS_AC_KERNEL_MM_PAGE_FLAGS], [
 	ZFS_AC_KERNEL_MM_PAGE_FLAG_ERROR
+	ZFS_AC_KERNEL_MM_PAGE_FLAGS_STRUCT
 ])

config/kernel-namespace.m4

@@ -0,0 +1,31 @@
+dnl #
+dnl # 6.18 API change
+dnl # ns->ops->type was moved to ns->ns.ns_type (struct ns_common)
+dnl #
+AC_DEFUN([ZFS_AC_KERNEL_SRC_NS_COMMON_TYPE], [
+	ZFS_LINUX_TEST_SRC([ns_common_type], [
+		#include <linux/user_namespace.h>
+	],[
+		struct user_namespace ns;
+		ns.ns.ns_type = 0;
+	])
+])
+
+AC_DEFUN([ZFS_AC_KERNEL_NS_COMMON_TYPE], [
+	AC_MSG_CHECKING([whether ns_type is accessible through ns_common])
+	ZFS_LINUX_TEST_RESULT([ns_common_type], [
+		AC_MSG_RESULT(yes)
+		AC_DEFINE([HAVE_NS_COMMON_TYPE], 1,
+			[Define if ns_type is accessible through ns_common])
+	],[
+		AC_MSG_RESULT(no)
+	])
+])
+
+AC_DEFUN([ZFS_AC_KERNEL_SRC_NAMESPACE], [
+	ZFS_AC_KERNEL_SRC_NS_COMMON_TYPE
+])
+
+AC_DEFUN([ZFS_AC_KERNEL_NAMESPACE], [
+	ZFS_AC_KERNEL_NS_COMMON_TYPE
+])

config/kernel-userns-capabilities.m4

@@ -1,79 +0,0 @@
-dnl #
-dnl # 2.6.38 API change
-dnl # ns_capable() was introduced
-dnl #
-AC_DEFUN([ZFS_AC_KERNEL_SRC_NS_CAPABLE], [
-	ZFS_LINUX_TEST_SRC([ns_capable], [
-		#include <linux/capability.h>
-	],[
-		ns_capable((struct user_namespace *)NULL, CAP_SYS_ADMIN);
-	])
-])
-
-AC_DEFUN([ZFS_AC_KERNEL_NS_CAPABLE], [
-	AC_MSG_CHECKING([whether ns_capable exists])
-	ZFS_LINUX_TEST_RESULT([ns_capable], [
-		AC_MSG_RESULT(yes)
-	],[
-		ZFS_LINUX_TEST_ERROR([ns_capable()])
-	])
-])
-
-dnl #
-dnl # 2.6.39 API change
-dnl # struct user_namespace was added to struct cred_t as cred->user_ns member
-dnl #
-AC_DEFUN([ZFS_AC_KERNEL_SRC_CRED_USER_NS], [
-	ZFS_LINUX_TEST_SRC([cred_user_ns], [
-		#include <linux/cred.h>
-	],[
-		struct cred cr;
-		cr.user_ns = (struct user_namespace *)NULL;
-	])
-])
-
-AC_DEFUN([ZFS_AC_KERNEL_CRED_USER_NS], [
-	AC_MSG_CHECKING([whether cred_t->user_ns exists])
-	ZFS_LINUX_TEST_RESULT([cred_user_ns], [
-		AC_MSG_RESULT(yes)
-	],[
-		ZFS_LINUX_TEST_ERROR([cred_t->user_ns()])
-	])
-])
-
-dnl #
-dnl # 3.4 API change
-dnl # kuid_has_mapping() and kgid_has_mapping() were added to distinguish
-dnl # between internal kernel uids/gids and user namespace uids/gids.
-dnl #
-AC_DEFUN([ZFS_AC_KERNEL_SRC_KUID_HAS_MAPPING], [
-	ZFS_LINUX_TEST_SRC([kuid_has_mapping], [
-		#include <linux/uidgid.h>
-	],[
-		kuid_has_mapping((struct user_namespace *)NULL, KUIDT_INIT(0));
-		kgid_has_mapping((struct user_namespace *)NULL, KGIDT_INIT(0));
-	])
-])
-
-AC_DEFUN([ZFS_AC_KERNEL_KUID_HAS_MAPPING], [
-	AC_MSG_CHECKING([whether kuid_has_mapping/kgid_has_mapping exist])
-	ZFS_LINUX_TEST_RESULT([kuid_has_mapping], [
-		AC_MSG_RESULT(yes)
-	],[
-		ZFS_LINUX_TEST_ERROR([kuid_has_mapping()])
-	])
-])
-
-AC_DEFUN([ZFS_AC_KERNEL_SRC_USERNS_CAPABILITIES], [
-	ZFS_AC_KERNEL_SRC_NS_CAPABLE
-	ZFS_AC_KERNEL_SRC_HAS_CAPABILITY
-	ZFS_AC_KERNEL_SRC_CRED_USER_NS
-	ZFS_AC_KERNEL_SRC_KUID_HAS_MAPPING
-])
-
-AC_DEFUN([ZFS_AC_KERNEL_USERNS_CAPABILITIES], [
-	ZFS_AC_KERNEL_NS_CAPABLE
-	ZFS_AC_KERNEL_HAS_CAPABILITY
-	ZFS_AC_KERNEL_CRED_USER_NS
-	ZFS_AC_KERNEL_KUID_HAS_MAPPING
-])

config/kernel-writeback.m4

@@ -0,0 +1,58 @@
+AC_DEFUN([ZFS_AC_KERNEL_SRC_WRITEPAGE_T], [
+	dnl #
+	dnl # 6.3 API change
+	dnl # The writepage_t function type now has its first argument as
+	dnl # struct folio* instead of struct page*
+	dnl #
+	ZFS_LINUX_TEST_SRC([writepage_t_folio], [
+		#include <linux/writeback.h>
+		static int putpage(struct folio *folio,
+		    struct writeback_control *wbc, void *data)
+		{ return 0; }
+		writepage_t func = putpage;
+	],[])
+])
+
+AC_DEFUN([ZFS_AC_KERNEL_WRITEPAGE_T], [
+	AC_MSG_CHECKING([whether int (*writepage_t)() takes struct folio*])
+	ZFS_LINUX_TEST_RESULT([writepage_t_folio], [
+		AC_MSG_RESULT(yes)
+		AC_DEFINE(HAVE_WRITEPAGE_T_FOLIO, 1,
+		   [int (*writepage_t)() takes struct folio*])
+	],[
+		AC_MSG_RESULT(no)
+	])
+])
+
+AC_DEFUN([ZFS_AC_KERNEL_SRC_WRITE_CACHE_PAGES], [
+	dnl #
+	dnl # 6.18 API change
+	dnl # write_cache_pages() has been removed.
+	dnl #
+	ZFS_LINUX_TEST_SRC([write_cache_pages], [
+		#include <linux/writeback.h>
+	], [
+		(void) write_cache_pages(NULL, NULL, NULL, NULL);
+	])
+])
+
+AC_DEFUN([ZFS_AC_KERNEL_WRITE_CACHE_PAGES], [
+	AC_MSG_CHECKING([whether write_cache_pages() is available])
+	ZFS_LINUX_TEST_RESULT([write_cache_pages], [
+		AC_MSG_RESULT(yes)
+		AC_DEFINE(HAVE_WRITE_CACHE_PAGES, 1,
+		    [write_cache_pages() is available])
+	],[
+		AC_MSG_RESULT(no)
+	])
+])
+
+AC_DEFUN([ZFS_AC_KERNEL_SRC_WRITEBACK], [
+	ZFS_AC_KERNEL_SRC_WRITEPAGE_T
+	ZFS_AC_KERNEL_SRC_WRITE_CACHE_PAGES
+])
+
+AC_DEFUN([ZFS_AC_KERNEL_WRITEBACK], [
+	ZFS_AC_KERNEL_WRITEPAGE_T
+	ZFS_AC_KERNEL_WRITE_CACHE_PAGES
+])

config/kernel-writepage_t.m4

@@ -1,26 +0,0 @@
-AC_DEFUN([ZFS_AC_KERNEL_SRC_WRITEPAGE_T], [
-	dnl #
-	dnl # 6.3 API change
-	dnl # The writepage_t function type now has its first argument as
-	dnl # struct folio* instead of struct page*
-	dnl #
-	ZFS_LINUX_TEST_SRC([writepage_t_folio], [
-		#include <linux/writeback.h>
-		static int putpage(struct folio *folio,
-		    struct writeback_control *wbc, void *data)
-		{ return 0; }
-		writepage_t func = putpage;
-	],[])
-])
-
-AC_DEFUN([ZFS_AC_KERNEL_WRITEPAGE_T], [
-	AC_MSG_CHECKING([whether int (*writepage_t)() takes struct folio*])
-	ZFS_LINUX_TEST_RESULT([writepage_t_folio], [
-		AC_MSG_RESULT(yes)
-		AC_DEFINE(HAVE_WRITEPAGE_T_FOLIO, 1,
-		   [int (*writepage_t)() takes struct folio*])
-	],[
-		AC_MSG_RESULT(no)
-	])
-])
-

config/kernel.m4

@@ -59,6 +59,7 @@ AC_DEFUN([ZFS_AC_KERNEL_TEST_SRC], [
 	ZFS_AC_KERNEL_SRC_ACL
 	ZFS_AC_KERNEL_SRC_INODE_SETATTR
 	ZFS_AC_KERNEL_SRC_INODE_GETATTR
+	ZFS_AC_KERNEL_SRC_INODE_STATE_READ_ONCE
 	ZFS_AC_KERNEL_SRC_SHOW_OPTIONS
 	ZFS_AC_KERNEL_SRC_SHRINKER
 	ZFS_AC_KERNEL_SRC_MKDIR
@@ -70,6 +71,7 @@ AC_DEFUN([ZFS_AC_KERNEL_TEST_SRC], [
 	ZFS_AC_KERNEL_SRC_COMMIT_METADATA
 	ZFS_AC_KERNEL_SRC_SETATTR_PREPARE
 	ZFS_AC_KERNEL_SRC_INSERT_INODE_LOCKED
+	ZFS_AC_KERNEL_SRC_DENTRY
 	ZFS_AC_KERNEL_SRC_TRUNCATE_SETSIZE
 	ZFS_AC_KERNEL_SRC_SECURITY_INODE
 	ZFS_AC_KERNEL_SRC_FST_MOUNT
@@ -120,7 +122,7 @@ AC_DEFUN([ZFS_AC_KERNEL_TEST_SRC], [
 	ZFS_AC_KERNEL_SRC_IDMAP_MNT_API
 	ZFS_AC_KERNEL_SRC_IDMAP_NO_USERNS
 	ZFS_AC_KERNEL_SRC_IATTR_VFSID
-	ZFS_AC_KERNEL_SRC_WRITEPAGE_T
+	ZFS_AC_KERNEL_SRC_WRITEBACK
 	ZFS_AC_KERNEL_SRC_RECLAIMED
 	ZFS_AC_KERNEL_SRC_REGISTER_SYSCTL_TABLE
 	ZFS_AC_KERNEL_SRC_REGISTER_SYSCTL_SZ
@@ -135,6 +137,8 @@ AC_DEFUN([ZFS_AC_KERNEL_TEST_SRC], [
 	ZFS_AC_KERNEL_SRC_TIMER
 	ZFS_AC_KERNEL_SRC_SUPER_BLOCK_S_WB_ERR
 	ZFS_AC_KERNEL_SRC_SOPS_FREE_INODE
+	ZFS_AC_KERNEL_SRC_NAMESPACE
+	ZFS_AC_KERNEL_SRC_INODE_GENERIC_DROP
 	case "$host_cpu" in
 		powerpc*)
 			ZFS_AC_KERNEL_SRC_CPU_HAS_FEATURE
@@ -177,6 +181,7 @@ AC_DEFUN([ZFS_AC_KERNEL_TEST_RESULT], [
 	ZFS_AC_KERNEL_ACL
 	ZFS_AC_KERNEL_INODE_SETATTR
 	ZFS_AC_KERNEL_INODE_GETATTR
+	ZFS_AC_KERNEL_INODE_STATE_READ_ONCE
 	ZFS_AC_KERNEL_SHOW_OPTIONS
 	ZFS_AC_KERNEL_SHRINKER
 	ZFS_AC_KERNEL_MKDIR
@@ -188,6 +193,7 @@ AC_DEFUN([ZFS_AC_KERNEL_TEST_RESULT], [
 	ZFS_AC_KERNEL_COMMIT_METADATA
 	ZFS_AC_KERNEL_SETATTR_PREPARE
 	ZFS_AC_KERNEL_INSERT_INODE_LOCKED
+	ZFS_AC_KERNEL_DENTRY
 	ZFS_AC_KERNEL_TRUNCATE_SETSIZE
 	ZFS_AC_KERNEL_SECURITY_INODE
 	ZFS_AC_KERNEL_FST_MOUNT
@@ -238,7 +244,7 @@ AC_DEFUN([ZFS_AC_KERNEL_TEST_RESULT], [
 	ZFS_AC_KERNEL_IDMAP_MNT_API
 	ZFS_AC_KERNEL_IDMAP_NO_USERNS
 	ZFS_AC_KERNEL_IATTR_VFSID
-	ZFS_AC_KERNEL_WRITEPAGE_T
+	ZFS_AC_KERNEL_WRITEBACK
 	ZFS_AC_KERNEL_RECLAIMED
 	ZFS_AC_KERNEL_REGISTER_SYSCTL_TABLE
 	ZFS_AC_KERNEL_REGISTER_SYSCTL_SZ
@@ -254,6 +260,8 @@ AC_DEFUN([ZFS_AC_KERNEL_TEST_RESULT], [
 	ZFS_AC_KERNEL_TIMER
 	ZFS_AC_KERNEL_SUPER_BLOCK_S_WB_ERR
 	ZFS_AC_KERNEL_SOPS_FREE_INODE
+	ZFS_AC_KERNEL_NAMESPACE
+	ZFS_AC_KERNEL_INODE_GENERIC_DROP
 	case "$host_cpu" in
 		powerpc*)
 			ZFS_AC_KERNEL_CPU_HAS_FEATURE

config/user-statx.m4

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

contrib/pyzfs/libzfs_core/exceptions.py

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

copy-builtin

@@ -8,7 +8,9 @@ usage()
 	exit 1
 }
 
-[ "$#" -eq 1 ] || usage
+if ! [ -d "$1" ] ; then
+	usage
+fi
 KERNEL_DIR="$1"
 
 if ! [ -e 'zfs_config.h' ]
@@ -31,6 +33,7 @@ cat > "$KERNEL_DIR/fs/zfs/Kconfig" <<EOF
 config ZFS
 	tristate "ZFS filesystem support"
 	depends on EFI_PARTITION
+	depends on BLOCK
 	select ZLIB_INFLATE
 	select ZLIB_DEFLATE
 	help

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

@@ -104,6 +104,9 @@
 #define	spa_taskq_write_param_set_args(var) \
     CTLTYPE_STRING, NULL, 0, spa_taskq_write_param, "A"
 
+#define	spa_taskq_free_param_set_args(var) \
+    CTLTYPE_STRING, NULL, 0, spa_taskq_free_param, "A"
+
 #define	fletcher_4_param_set_args(var) \
     CTLTYPE_STRING, NULL, 0, fletcher_4_param, "A"
 

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

@@ -290,80 +290,11 @@ extern unsigned char bcd_to_byte[256];
 #define	offsetof(type, field)	__offsetof(type, field)
 #endif
 
-/*
+#define	highbit(x)		flsl(x)
- * Find highest one bit set.
+#define	lowbit(x)		ffsl(x)
- *      Returns bit number + 1 of highest bit that is set, otherwise returns 0.
- * High order bit is 31 (or 63 in _LP64 kernel).
- */
-static __inline int
-highbit(ulong_t i)
-{
-#if defined(HAVE_INLINE_FLSL)
-	return (flsl(i));
-#else
-	int h = 1;
 
-	if (i == 0)
+#define	highbit64(x)		flsll(x)
-		return (0);
+#define	lowbit64(x)		ffsll(x)
-#ifdef _LP64
-	if (i & 0xffffffff00000000ul) {
-		h += 32; i >>= 32;
-	}
-#endif
-	if (i & 0xffff0000) {
-		h += 16; i >>= 16;
-	}
-	if (i & 0xff00) {
-		h += 8; i >>= 8;
-	}
-	if (i & 0xf0) {
-		h += 4; i >>= 4;
-	}
-	if (i & 0xc) {
-		h += 2; i >>= 2;
-	}
-	if (i & 0x2) {
-		h += 1;
-	}
-	return (h);
-#endif
-}
-
-/*
- * Find highest one bit set.
- *	Returns bit number + 1 of highest bit that is set, otherwise returns 0.
- */
-static __inline int
-highbit64(uint64_t i)
-{
-#if defined(HAVE_INLINE_FLSLL)
-	return (flsll(i));
-#else
-	int h = 1;
-
-	if (i == 0)
-		return (0);
-	if (i & 0xffffffff00000000ULL) {
-		h += 32; i >>= 32;
-	}
-	if (i & 0xffff0000) {
-		h += 16; i >>= 16;
-	}
-	if (i & 0xff00) {
-		h += 8; i >>= 8;
-	}
-	if (i & 0xf0) {
-		h += 4; i >>= 4;
-	}
-	if (i & 0xc) {
-		h += 2; i >>= 2;
-	}
-	if (i & 0x2) {
-		h += 1;
-	}
-	return (h);
-#endif
-}
 
 #ifdef	__cplusplus
 }

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

@@ -34,6 +34,17 @@
 
 #define	d_alias			d_u.d_alias
 
+#ifdef HAVE_MM_PAGE_FLAGS_STRUCT
+/*
+ * Starting from Linux 6.18, the 'flags' field in 'struct page' is defined
+ * to a struct ('memdesc_flags_t' typedef) instead of an unsigned long for
+ * improved typesafety.
+ */
+#define	page_flags flags.f
+#else
+#define	page_flags flags
+#endif
+
 /*
  * Starting from Linux 5.13, flush_dcache_page() becomes an inline function
  * and under some configurations, may indirectly referencing GPL-only
@@ -44,8 +55,8 @@
 #include <linux/simd_powerpc.h>
 #define	flush_dcache_page(page)	do {					\
 		if (!cpu_has_feature(CPU_FTR_COHERENT_ICACHE) &&	\
-		    test_bit(PG_dcache_clean, &(page)->flags))		\
+		    test_bit(PG_dcache_clean, &(page)->page_flags))	\
-			clear_bit(PG_dcache_clean, &(page)->flags);	\
+			clear_bit(PG_dcache_clean, &(page)->page_flags);\
 	} while (0)
 #endif
 /*
@@ -55,8 +66,8 @@
  */
 #if defined __riscv && defined HAVE_FLUSH_DCACHE_PAGE_GPL_ONLY
 #define	flush_dcache_page(page)	do {					\
-		if (test_bit(PG_dcache_clean, &(page)->flags))		\
+		if (test_bit(PG_dcache_clean, &(page)->page_flags))	\
-			clear_bit(PG_dcache_clean, &(page)->flags);	\
+			clear_bit(PG_dcache_clean, &(page)->page_flags);\
 	} while (0)
 #endif
 

include/os/linux/kernel/linux/vfs_compat.h

@@ -23,6 +23,7 @@
 /*
  * Copyright (C) 2011 Lawrence Livermore National Security, LLC.
  * Copyright (C) 2015 Jörg Thalheim.
+ * Copyright (c) 2025, Rob Norris <robn@despairlabs.com>
  */
 
 #ifndef _ZFS_VFS_H
@@ -262,4 +263,18 @@ zpl_is_32bit_api(void)
 #define	zpl_generic_fillattr(user_ns, ip, sp)	generic_fillattr(ip, sp)
 #endif
 
+#ifdef HAVE_INODE_GENERIC_DROP
+/* 6.18 API change. These were renamed, alias the old names to the new. */
+#define	generic_delete_inode(ip)	inode_just_drop(ip)
+#define	generic_drop_inode(ip)		inode_generic_drop(ip)
+#endif
+
+#ifndef HAVE_INODE_STATE_READ_ONCE
+/*
+ * 6.19 API change. We should no longer access i_state directly. If the new
+ * helper function doesn't exist, define our own.
+ */
+#define	inode_state_read_once(ip)	READ_ONCE(ip->i_state)
+#endif
+
 #endif /* _ZFS_VFS_H */

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

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

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

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

include/sys/fs/zfs.h

@@ -863,6 +863,10 @@ typedef struct zpool_load_policy {
 #define	ZPOOL_CONFIG_MMP_SEQ		"mmp_seq"	/* not stored on disk */
 #define	ZPOOL_CONFIG_MMP_HOSTNAME	"mmp_hostname"	/* not stored on disk */
 #define	ZPOOL_CONFIG_MMP_HOSTID		"mmp_hostid"	/* not stored on disk */
+#define	ZPOOL_CONFIG_MMP_RESULT		"mmp_result"	/* not stored on disk */
+#define	ZPOOL_CONFIG_MMP_TRYIMPORT_NS	"mmp_tryimport_ns"	/* not stored */
+#define	ZPOOL_CONFIG_MMP_IMPORT_NS	"mmp_import_ns"	/* not stored on disk */
+#define	ZPOOL_CONFIG_MMP_CLAIM_NS	"mmp_claim_ns"	/* not stored on disk */
 #define	ZPOOL_CONFIG_ALLOCATION_BIAS	"alloc_bias"	/* not stored on disk */
 #define	ZPOOL_CONFIG_EXPANSION_TIME	"expansion_time"	/* not stored */
 #define	ZPOOL_CONFIG_REBUILD_STATS	"org.openzfs:rebuild_stats"

include/sys/mmp.h

@@ -33,6 +33,7 @@ extern "C" {
 #define	MMP_DEFAULT_IMPORT_INTERVALS	20
 #define	MMP_DEFAULT_FAIL_INTERVALS	10
 #define	MMP_MIN_FAIL_INTERVALS		2	/* min if != 0 */
+#define	MMP_IMPORT_VERIFY_ITERS		10
 #define	MMP_IMPORT_SAFETY_FACTOR	200	/* pct */
 #define	MMP_INTERVAL_OK(interval)	MAX(interval, MMP_MIN_INTERVAL)
 #define	MMP_FAIL_INTVS_OK(fails)	(fails == 0 ? 0 : MAX(fails, \
@@ -53,6 +54,9 @@ typedef struct mmp_thread {
 	vdev_t		*mmp_last_leaf;	/* last mmp write sent here */
 	uint64_t	mmp_leaf_last_gen;	/* last mmp write sent here */
 	uint32_t	mmp_seq;	/* intra-second update counter */
+	uint64_t	mmp_tryimport_ns; /* tryimport activity check time */
+	uint64_t	mmp_import_ns;	/* import activity check time */
+	uint64_t	mmp_claim_ns;	/* claim activity check time */
 } mmp_thread_t;
 
 
@@ -62,6 +66,7 @@ extern void mmp_thread_start(struct spa *spa);
 extern void mmp_thread_stop(struct spa *spa);
 extern void mmp_update_uberblock(struct spa *spa, struct uberblock *ub);
 extern void mmp_signal_all_threads(void);
+extern int mmp_claim_uberblock(spa_t *spa, vdev_t *vd, uberblock_t *ub);
 
 /* Global tuning */
 extern int param_set_multihost_interval(ZFS_MODULE_PARAM_ARGS);

include/sys/spa.h

@@ -1044,6 +1044,7 @@ extern void spa_set_rootblkptr(spa_t *spa, const blkptr_t *bp);
 extern void spa_altroot(spa_t *, char *, size_t);
 extern uint32_t spa_sync_pass(spa_t *spa);
 extern char *spa_name(spa_t *spa);
+extern char *spa_load_name(spa_t *spa);
 extern uint64_t spa_guid(spa_t *spa);
 extern uint64_t spa_load_guid(spa_t *spa);
 extern uint64_t spa_last_synced_txg(spa_t *spa);

include/sys/spa_impl.h

@@ -224,6 +224,7 @@ struct spa {
 	 * Fields protected by spa_namespace_lock.
 	 */
 	char		spa_name[ZFS_MAX_DATASET_NAME_LEN];	/* pool name */
+	char		*spa_load_name;		/* unmodified pool name */
 	char		*spa_comment;		/* comment */
 	avl_node_t	spa_avl;		/* node in spa_namespace_avl */
 	nvlist_t	*spa_config;		/* last synced config */
@@ -303,6 +304,7 @@ struct spa {
 	void		*spa_cksum_tmpls[ZIO_CHECKSUM_FUNCTIONS];
 	uberblock_t	spa_ubsync;		/* last synced uberblock */
 	uberblock_t	spa_uberblock;		/* current uberblock */
+	boolean_t	spa_activity_check; 	/* activity check required */
 	boolean_t	spa_extreme_rewind;	/* rewind past deferred frees */
 	kmutex_t	spa_scrub_lock;		/* resilver/scrub lock */
 	uint64_t	spa_scrub_inflight;	/* in-flight scrub bytes */

include/sys/uberblock_impl.h

@@ -51,6 +51,12 @@ extern "C" {
 #define	MMP_SEQ_VALID_BIT	0x02
 #define	MMP_FAIL_INT_VALID_BIT	0x04
 
+#define	MMP_INTERVAL_MASK	0x00000000FFFFFF00
+#define	MMP_SEQ_MASK		0x0000FFFF00000000
+#define	MMP_FAIL_INT_MASK	0xFFFF000000000000
+
+#define	MMP_SEQ_MAX		UINT16_MAX
+
 #define	MMP_VALID(ubp)		((ubp)->ub_magic == UBERBLOCK_MAGIC && \
 				    (ubp)->ub_mmp_magic == MMP_MAGIC)
 #define	MMP_INTERVAL_VALID(ubp)	(MMP_VALID(ubp) && ((ubp)->ub_mmp_config & \
@@ -60,21 +66,25 @@ extern "C" {
 #define	MMP_FAIL_INT_VALID(ubp)	(MMP_VALID(ubp) && ((ubp)->ub_mmp_config & \
 				    MMP_FAIL_INT_VALID_BIT))
 
-#define	MMP_INTERVAL(ubp)	(((ubp)->ub_mmp_config & 0x00000000FFFFFF00) \
+#define	MMP_INTERVAL(ubp)	(((ubp)->ub_mmp_config & MMP_INTERVAL_MASK) \
 				    >> 8)
-#define	MMP_SEQ(ubp)		(((ubp)->ub_mmp_config & 0x0000FFFF00000000) \
+#define	MMP_SEQ(ubp)		(((ubp)->ub_mmp_config & MMP_SEQ_MASK) \
 				    >> 32)
-#define	MMP_FAIL_INT(ubp)	(((ubp)->ub_mmp_config & 0xFFFF000000000000) \
+#define	MMP_FAIL_INT(ubp)	(((ubp)->ub_mmp_config & MMP_FAIL_INT_MASK) \
 				    >> 48)
 
 #define	MMP_INTERVAL_SET(write) \
-	    (((uint64_t)(write & 0xFFFFFF) << 8) | MMP_INTERVAL_VALID_BIT)
+	    (((uint64_t)((write) & 0xFFFFFF) << 8) | MMP_INTERVAL_VALID_BIT)
 
 #define	MMP_SEQ_SET(seq) \
-	    (((uint64_t)(seq & 0xFFFF) << 32) | MMP_SEQ_VALID_BIT)
+	    (((uint64_t)((seq) & 0xFFFF) << 32) | MMP_SEQ_VALID_BIT)
 
 #define	MMP_FAIL_INT_SET(fail) \
-	    (((uint64_t)(fail & 0xFFFF) << 48) | MMP_FAIL_INT_VALID_BIT)
+	    (((uint64_t)((fail) & 0xFFFF) << 48) | MMP_FAIL_INT_VALID_BIT)
+
+
+#define	MMP_SEQ_CLEAR(ubp) \
+	    ((ubp)->ub_mmp_config &= ~(MMP_SEQ_MASK | MMP_SEQ_VALID_BIT))
 
 /*
  * RAIDZ expansion reflow information.

include/sys/vdev.h

@@ -212,6 +212,8 @@ extern void vdev_label_write(zio_t *zio, vdev_t *vd, int l, abd_t *buf, uint64_t
 extern int vdev_label_read_bootenv(vdev_t *, nvlist_t *);
 extern int vdev_label_write_bootenv(vdev_t *, nvlist_t *);
 extern int vdev_uberblock_sync_list(vdev_t **, int, struct uberblock *, int);
+extern int vdev_uberblock_compare(const struct uberblock *,
+    const struct uberblock *);
 extern int vdev_check_boot_reserve(spa_t *, vdev_t *);
 
 typedef enum {

include/sys/zfs_file.h

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

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

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

lib/libzfs/libzfs_pool.c

@@ -2238,6 +2238,11 @@ zpool_import_props(libzfs_handle_t *hdl, nvlist_t *config, const char *newname,
 
 	zpool_get_load_policy(config, &policy);
 
+	if (getenv("ZFS_LOAD_INFO_DEBUG") && nv != NULL &&
+	    nvlist_lookup_nvlist(nv, ZPOOL_CONFIG_LOAD_INFO, &nvinfo) == 0) {
+		dump_nvlist(nvinfo, 4);
+	}
+
 	if (error) {
 		char desc[1024];
 		char aux[256];

lib/libzfs/libzfs_status.c

@@ -98,57 +98,57 @@ static const char *const zfs_msgid_table[] = {
 #define	NMSGID	(sizeof (zfs_msgid_table) / sizeof (zfs_msgid_table[0]))
 
 static int
-vdev_missing(vdev_stat_t *vs, uint_t vsc)
+vdev_missing(vdev_stat_t *vs, uint_t vsc, void *arg)
 {
-	(void) vsc;
+	(void) vsc, (void) arg;
 	return (vs->vs_state == VDEV_STATE_CANT_OPEN &&
 	    vs->vs_aux == VDEV_AUX_OPEN_FAILED);
 }
 
 static int
-vdev_faulted(vdev_stat_t *vs, uint_t vsc)
+vdev_faulted(vdev_stat_t *vs, uint_t vsc, void *arg)
 {
-	(void) vsc;
+	(void) vsc, (void) arg;
 	return (vs->vs_state == VDEV_STATE_FAULTED);
 }
 
 static int
-vdev_errors(vdev_stat_t *vs, uint_t vsc)
+vdev_errors(vdev_stat_t *vs, uint_t vsc, void *arg)
 {
-	(void) vsc;
+	(void) vsc, (void) arg;
 	return (vs->vs_state == VDEV_STATE_DEGRADED ||
 	    vs->vs_read_errors != 0 || vs->vs_write_errors != 0 ||
 	    vs->vs_checksum_errors != 0);
 }
 
 static int
-vdev_broken(vdev_stat_t *vs, uint_t vsc)
+vdev_broken(vdev_stat_t *vs, uint_t vsc, void *arg)
 {
-	(void) vsc;
+	(void) vsc, (void) arg;
 	return (vs->vs_state == VDEV_STATE_CANT_OPEN);
 }
 
 static int
-vdev_offlined(vdev_stat_t *vs, uint_t vsc)
+vdev_offlined(vdev_stat_t *vs, uint_t vsc, void *arg)
 {
-	(void) vsc;
+	(void) vsc, (void) arg;
 	return (vs->vs_state == VDEV_STATE_OFFLINE);
 }
 
 static int
-vdev_removed(vdev_stat_t *vs, uint_t vsc)
+vdev_removed(vdev_stat_t *vs, uint_t vsc, void *arg)
 {
-	(void) vsc;
+	(void) vsc, (void) arg;
 	return (vs->vs_state == VDEV_STATE_REMOVED);
 }
 
 static int
-vdev_non_native_ashift(vdev_stat_t *vs, uint_t vsc)
+vdev_non_native_ashift(vdev_stat_t *vs, uint_t vsc, void *arg)
 {
-	if (getenv("ZPOOL_STATUS_NON_NATIVE_ASHIFT_IGNORE") != NULL)
+	uint64_t ashift = *(uint64_t *)arg;
-		return (0);
 
 	return (VDEV_STAT_VALID(vs_physical_ashift, vsc) &&
+	    (ashift == 0 || vs->vs_configured_ashift < ashift) &&
 	    vs->vs_configured_ashift < vs->vs_physical_ashift);
 }
 
@@ -156,8 +156,8 @@ vdev_non_native_ashift(vdev_stat_t *vs, uint_t vsc)
  * Detect if any leaf devices that have seen errors or could not be opened.
  */
 static boolean_t
-find_vdev_problem(nvlist_t *vdev, int (*func)(vdev_stat_t *, uint_t),
+find_vdev_problem(nvlist_t *vdev, int (*func)(vdev_stat_t *, uint_t, void *),
-    boolean_t ignore_replacing)
+    void *arg, boolean_t ignore_replacing)
 {
 	nvlist_t **child;
 	uint_t c, children;
@@ -177,14 +177,16 @@ find_vdev_problem(nvlist_t *vdev, int (*func)(vdev_stat_t *, uint_t),
 
 	if (nvlist_lookup_nvlist_array(vdev, ZPOOL_CONFIG_CHILDREN, &child,
 	    &children) == 0) {
-		for (c = 0; c < children; c++)
+		for (c = 0; c < children; c++) {
-			if (find_vdev_problem(child[c], func, ignore_replacing))
+			if (find_vdev_problem(child[c], func, arg,
+			    ignore_replacing))
 				return (B_TRUE);
+		}
 	} else {
 		uint_t vsc;
 		vdev_stat_t *vs = (vdev_stat_t *)fnvlist_lookup_uint64_array(
 		    vdev, ZPOOL_CONFIG_VDEV_STATS, &vsc);
-		if (func(vs, vsc) != 0)
+		if (func(vs, vsc, arg) != 0)
 			return (B_TRUE);
 	}
 
@@ -193,9 +195,11 @@ find_vdev_problem(nvlist_t *vdev, int (*func)(vdev_stat_t *, uint_t),
 	 */
 	if (nvlist_lookup_nvlist_array(vdev, ZPOOL_CONFIG_L2CACHE, &child,
 	    &children) == 0) {
-		for (c = 0; c < children; c++)
+		for (c = 0; c < children; c++) {
-			if (find_vdev_problem(child[c], func, ignore_replacing))
+			if (find_vdev_problem(child[c], func, arg,
+			    ignore_replacing))
 				return (B_TRUE);
+		}
 	}
 
 	return (B_FALSE);
@@ -220,7 +224,7 @@ find_vdev_problem(nvlist_t *vdev, int (*func)(vdev_stat_t *, uint_t),
  */
 static zpool_status_t
 check_status(nvlist_t *config, boolean_t isimport,
-    zpool_errata_t *erratap, const char *compat)
+    zpool_errata_t *erratap, const char *compat, uint64_t ashift)
 {
 	pool_scan_stat_t *ps = NULL;
 	uint_t vsc, psc;
@@ -371,15 +375,15 @@ check_status(nvlist_t *config, boolean_t isimport,
 	 * Bad devices in non-replicated config.
 	 */
 	if (vs->vs_state == VDEV_STATE_CANT_OPEN &&
-	    find_vdev_problem(nvroot, vdev_faulted, B_TRUE))
+	    find_vdev_problem(nvroot, vdev_faulted, NULL, B_TRUE))
 		return (ZPOOL_STATUS_FAULTED_DEV_NR);
 
 	if (vs->vs_state == VDEV_STATE_CANT_OPEN &&
-	    find_vdev_problem(nvroot, vdev_missing, B_TRUE))
+	    find_vdev_problem(nvroot, vdev_missing, NULL, B_TRUE))
 		return (ZPOOL_STATUS_MISSING_DEV_NR);
 
 	if (vs->vs_state == VDEV_STATE_CANT_OPEN &&
-	    find_vdev_problem(nvroot, vdev_broken, B_TRUE))
+	    find_vdev_problem(nvroot, vdev_broken, NULL, B_TRUE))
 		return (ZPOOL_STATUS_CORRUPT_LABEL_NR);
 
 	/*
@@ -402,35 +406,37 @@ check_status(nvlist_t *config, boolean_t isimport,
 	/*
 	 * Missing devices in a replicated config.
 	 */
-	if (find_vdev_problem(nvroot, vdev_faulted, B_TRUE))
+	if (find_vdev_problem(nvroot, vdev_faulted, NULL, B_TRUE))
 		return (ZPOOL_STATUS_FAULTED_DEV_R);
-	if (find_vdev_problem(nvroot, vdev_missing, B_TRUE))
+	if (find_vdev_problem(nvroot, vdev_missing, NULL, B_TRUE))
 		return (ZPOOL_STATUS_MISSING_DEV_R);
-	if (find_vdev_problem(nvroot, vdev_broken, B_TRUE))
+	if (find_vdev_problem(nvroot, vdev_broken, NULL, B_TRUE))
 		return (ZPOOL_STATUS_CORRUPT_LABEL_R);
 
 	/*
 	 * Devices with errors
 	 */
-	if (!isimport && find_vdev_problem(nvroot, vdev_errors, B_TRUE))
+	if (!isimport && find_vdev_problem(nvroot, vdev_errors, NULL, B_TRUE))
 		return (ZPOOL_STATUS_FAILING_DEV);
 
 	/*
 	 * Offlined devices
 	 */
-	if (find_vdev_problem(nvroot, vdev_offlined, B_TRUE))
+	if (find_vdev_problem(nvroot, vdev_offlined, NULL, B_TRUE))
 		return (ZPOOL_STATUS_OFFLINE_DEV);
 
 	/*
 	 * Removed device
 	 */
-	if (find_vdev_problem(nvroot, vdev_removed, B_TRUE))
+	if (find_vdev_problem(nvroot, vdev_removed, NULL, B_TRUE))
 		return (ZPOOL_STATUS_REMOVED_DEV);
 
 	/*
 	 * Suboptimal, but usable, ashift configuration.
 	 */
-	if (find_vdev_problem(nvroot, vdev_non_native_ashift, B_FALSE))
+	if (!isimport &&
+	    getenv("ZPOOL_STATUS_NON_NATIVE_ASHIFT_IGNORE") == NULL &&
+	    find_vdev_problem(nvroot, vdev_non_native_ashift, &ashift, B_FALSE))
 		return (ZPOOL_STATUS_NON_NATIVE_ASHIFT);
 
 	/*
@@ -509,8 +515,10 @@ zpool_get_status(zpool_handle_t *zhp, const char **msgid,
 	    ZFS_MAXPROPLEN, NULL, B_FALSE) != 0)
 		compatibility[0] = '\0';
 
+	uint64_t ashift = zpool_get_prop_int(zhp, ZPOOL_PROP_ASHIFT, NULL);
+
 	zpool_status_t ret = check_status(zhp->zpool_config, B_FALSE, errata,
-	    compatibility);
+	    compatibility, ashift);
 
 	if (msgid != NULL) {
 		if (ret >= NMSGID)
@@ -525,7 +533,7 @@ zpool_status_t
 zpool_import_status(nvlist_t *config, const char **msgid,
     zpool_errata_t *errata)
 {
-	zpool_status_t ret = check_status(config, B_TRUE, errata, NULL);
+	zpool_status_t ret = check_status(config, B_TRUE, errata, NULL, 0);
 
 	if (ret >= NMSGID)
 		*msgid = NULL;

lib/libzpool/kernel.c

@@ -1175,7 +1175,7 @@ zfs_file_write(zfs_file_t *fp, const void *buf, size_t count, ssize_t *resid)
  */
 int
 zfs_file_pwrite(zfs_file_t *fp, const void *buf,
-    size_t count, loff_t pos, ssize_t *resid)
+    size_t count, loff_t pos, uint8_t ashift, ssize_t *resid)
 {
 	ssize_t rc, split, done;
 	int sectors;
@@ -1185,8 +1185,8 @@ zfs_file_pwrite(zfs_file_t *fp, const void *buf,
 	 * system calls so that the process can be killed in between.
 	 * This is used by ztest to simulate realistic failure modes.
 	 */
-	sectors = count >> SPA_MINBLOCKSHIFT;
+	sectors = count >> ashift;
-	split = (sectors > 0 ? rand() % sectors : 0) << SPA_MINBLOCKSHIFT;
+	split = (sectors > 0 ? rand() % sectors : 0) << ashift;
 	rc = pwrite64(fp->f_fd, buf, split, pos);
 	if (rc != -1) {
 		done = rc;

man/man1/zhack.1

@@ -122,6 +122,24 @@ Example:
 .Nm zhack Cm label repair Fl cu Ar device
   Fix checksums and undetach a device
 .
+.It Xo
+.Nm zhack
+.Cm metaslab leak
+.Op Fl f
+.Ar pool
+.Xc
+Apply a fragmentation profile generated by
+.Sy zdb
+to the specified
+.Ar pool Ns
+\&.
+.Pp
+The
+.Fl f
+flag forces the profile to apply even if the vdevs in the
+.Ar pool
+don't have the same number of metaslabs as the fragmentation profile.
+.
 .El
 .
 .Sh GLOBAL OPTIONS
@@ -143,6 +161,8 @@ Search for
 members in
 .Ar dir .
 Can be specified more than once.
+.It Fl o Ar var Ns = Ns Ar value
+Set the given tunable to the provided value.
 .El
 .
 .Sh EXAMPLES

man/man4/zfs.4

@@ -17,7 +17,7 @@
 .\" own identifying information:
 .\" Portions Copyright [yyyy] [name of copyright owner]
 .\"
-.Dd May 24, 2025
+.Dd September 15, 2025
 .Dt ZFS 4
 .Os
 .
@@ -2551,6 +2551,49 @@ the xattr so as to not accumulate duplicates.
 .It Sy zio_requeue_io_start_cut_in_line Ns = Ns Sy 0 Ns | Ns 1 Pq int
 Prioritize requeued I/O.
 .
+.It Sy zfs_delete_inode Ns = Ns Sy 0 Ns | Ns 1 Pq int
+Sets whether the kernel should free an inode structure when the last reference
+is released, or cache it in memory.
+Intended for testing/debugging.
+.Pp
+A live inode structure "pins" versious internal OpenZFS structures in memory,
+which can result in large amounts of "unusable" memory on systems with lots of
+infrequently-accessed files, until the kernel's memory pressure mechanism
+asks OpenZFS to release them.
+.Pp
+The default value of
+.Sy 0
+always caches inodes that appear to still exist on disk.
+Setting it to
+.Sy 1
+will immediately release unused inodes and their associated memory back to the
+dbuf cache or the ARC for reuse, but may reduce performance if inodes are
+frequently evicted and reloaded.
+.Pp
+This parameter is only available on Linux.
+.
+.It Sy zfs_delete_dentry Ns = Ns Sy 0 Ns | Ns 1 Pq int
+Sets whether the kernel should free a dentry structure when it is no longer
+required, or hold it in the dentry cache.
+Intended for testing/debugging.
+.
+Since a dentry structure holds an inode reference, a cached dentry can "pin"
+an inode in memory indefinitely, along with associated OpenZFS structures (See
+.Sy zfs_delete_inode ) .
+.Pp
+The default value of
+.Sy 0
+instructs the kernel to cache entries and their associated inodes when they
+are no longer directly referenced.
+They will be reclaimed as part of the kernel's normal cache management
+processes.
+Setting it to
+.Sy 1
+will instruct the kernel to release directory entries and their inodes as soon
+as they are no longer referenced by the filesystem.
+.Pp
+This parameter is only available on Linux.
+.
 .It Sy zio_taskq_batch_pct Ns = Ns Sy 80 Ns % Pq uint
 Percentage of online CPUs which will run a worker thread for I/O.
 These workers are responsible for I/O work such as compression, encryption,
@@ -2585,12 +2628,50 @@ Set value only applies to pools imported/created after that.
 Set the queue and thread configuration for the IO read queues.
 This is an advanced debugging parameter.
 Don't change this unless you understand what it does.
+Each of the four values corresponds to the issue, issue high-priority,
+interrupt, and interrupt high-priority queues.
+Valid values are
+.Sy fixed,N,M
+(M queues with N threads each),
+.Sy scale[,MIN]
+(scale with CPUs, minimum MIN total threads),
+.Sy sync ,
+and
+.Sy null .
 Set values only apply to pools imported/created after that.
 .
 .It Sy zio_taskq_write Ns = Ns Sy sync null scale null Pq charp
 Set the queue and thread configuration for the IO write queues.
 This is an advanced debugging parameter.
 Don't change this unless you understand what it does.
+Each of the four values corresponds to the issue, issue high-priority,
+interrupt, and interrupt high-priority queues.
+Valid values are
+.Sy fixed,N,M
+(M queues with N threads each),
+.Sy scale[,MIN]
+(scale with CPUs, minimum MIN total threads),
+.Sy sync ,
+and
+.Sy null .
+Set values only apply to pools imported/created after that.
+.
+.It Sy zio_taskq_free Ns = Ns Sy scale,32 null null null Pq charp
+Set the queue and thread configuration for the IO free queues.
+This is an advanced debugging parameter.
+Don't change this unless you understand what it does.
+Each of the four values corresponds to the issue, issue high-priority,
+interrupt, and interrupt high-priority queues.
+Valid values are
+.Sy fixed,N,M
+(M queues with N threads each),
+.Sy scale[,MIN]
+(scale with CPUs, minimum MIN total threads),
+.Sy sync ,
+and
+.Sy null .
+The default uses a minimum of 32 threads to improve parallelism for
+DDT and BRT metadata operations during frees.
 Set values only apply to pools imported/created after that.
 .
 .It Sy zvol_inhibit_dev Ns = Ns Sy 0 Ns | Ns 1 Pq uint

man/man8/zdb.8

@@ -69,6 +69,13 @@
 .Op Fl U Ar cache
 .Ar poolname Op Ar vdev Oo Ar metaslab Oc Ns …
 .Nm
+.Fl -allocated-map
+.Op Fl mAFLPXY
+.Op Fl e Oo Fl V Oc Oo Fl p Ar path Oc Ns …
+.Op Fl t Ar txg
+.Op Fl U Ar cache
+.Ar poolname Op Ar vdev Oo Ar metaslab Oc Ns …
+.Nm
 .Fl O
 .Op Fl K Ar key
 .Ar dataset path
@@ -128,6 +135,11 @@ that zdb may interpret inconsistent pool data and behave erratically.
 .Sh OPTIONS
 Display options:
 .Bl -tag -width Ds
+.It Fl Sy -allocated-map
+Prints out a list of all the allocated regions in the pool.
+Primarily intended for use with the
+.Nm zhack metaslab leak
+subcommand.
 .It Fl b , -block-stats
 Display statistics regarding the number, size
 .Pq logical, physical and allocated

module/Kbuild.in

@@ -2,7 +2,7 @@
 # first.  This ensures its module initialization function is run before
 # any of the other module initialization functions which depend on it.
 
-ZFS_MODULE_CFLAGS += -std=gnu99 -Wno-declaration-after-statement
+ZFS_MODULE_CFLAGS += -std=gnu11 -Wno-declaration-after-statement
 ZFS_MODULE_CFLAGS += -Wmissing-prototypes
 ZFS_MODULE_CFLAGS += @KERNEL_DEBUG_CFLAGS@  @NO_FORMAT_ZERO_LENGTH@
 

module/Makefile.bsd

@@ -62,11 +62,6 @@ CFLAGS+= -DZFS_DEBUG -g
 CFLAGS += -DNDEBUG
 .endif
 
-.if defined(WITH_VFS_DEBUG) && ${WITH_VFS_DEBUG} == "true"
-# kernel must also be built with this option for this to work
-CFLAGS+= -DDEBUG_VFS_LOCKS
-.endif
-
 .if defined(WITH_GCOV) && ${WITH_GCOV} == "true"
 CFLAGS+=	 -fprofile-arcs -ftest-coverage
 .endif

module/icp/algs/sha2/sha2_generic.c

@@ -77,7 +77,8 @@ static const uint32_t SHA256_K[64] = {
 	h = g, g = f, f = e, e = d + T1; \
 	d = c, c = b, b = a, a = T1 + T2;
 
-static void sha256_generic(uint32_t state[8], const void *data, size_t num_blks)
+static void
+icp_sha256_generic(uint32_t state[8], const void *data, size_t num_blks)
 {
 	uint64_t blk;
 
@@ -173,7 +174,8 @@ static const uint64_t SHA512_K[80] = {
 	0x5fcb6fab3ad6faec, 0x6c44198c4a475817
 };
 
-static void sha512_generic(uint64_t state[8], const void *data, size_t num_blks)
+static void
+icp_sha512_generic(uint64_t state[8], const void *data, size_t num_blks)
 {
 	uint64_t blk;
 
@@ -226,7 +228,8 @@ static void sha512_generic(uint64_t state[8], const void *data, size_t num_blks)
 	}
 }
 
-static void sha256_update(sha256_ctx *ctx, const uint8_t *data, size_t len)
+static void
+icp_sha256_update(sha256_ctx *ctx, const uint8_t *data, size_t len)
 {
 	uint64_t pos = ctx->count[0];
 	uint64_t total = ctx->count[1];
@@ -258,7 +261,8 @@ static void sha256_update(sha256_ctx *ctx, const uint8_t *data, size_t len)
 	ctx->count[1] = total;
 }
 
-static void sha512_update(sha512_ctx *ctx, const uint8_t *data, size_t len)
+static void
+icp_sha512_update(sha512_ctx *ctx, const uint8_t *data, size_t len)
 {
 	uint64_t pos = ctx->count[0];
 	uint64_t total = ctx->count[1];
@@ -290,7 +294,8 @@ static void sha512_update(sha512_ctx *ctx, const uint8_t *data, size_t len)
 	ctx->count[1] = total;
 }
 
-static void sha256_final(sha256_ctx *ctx, uint8_t *result, int bits)
+static void
+icp_sha256_final(sha256_ctx *ctx, uint8_t *result, int bits)
 {
 	uint64_t mlen, pos = ctx->count[0];
 	uint8_t *m = ctx->wbuf;
@@ -334,7 +339,8 @@ static void sha256_final(sha256_ctx *ctx, uint8_t *result, int bits)
 	memset(ctx, 0, sizeof (*ctx));
 }
 
-static void sha512_final(sha512_ctx *ctx, uint8_t *result, int bits)
+static void
+icp_sha512_final(sha512_ctx *ctx, uint8_t *result, int bits)
 {
 	uint64_t mlen, pos = ctx->count[0];
 	uint8_t *m = ctx->wbuf, *r;
@@ -461,14 +467,14 @@ SHA2Update(SHA2_CTX *ctx, const void *data, size_t len)
 
 	switch (ctx->algotype) {
 		case SHA256:
-			sha256_update(&ctx->sha256, data, len);
+			icp_sha256_update(&ctx->sha256, data, len);
 			break;
 		case SHA512:
 		case SHA512_HMAC_MECH_INFO_TYPE:
-			sha512_update(&ctx->sha512, data, len);
+			icp_sha512_update(&ctx->sha512, data, len);
 			break;
 		case SHA512_256:
-			sha512_update(&ctx->sha512, data, len);
+			icp_sha512_update(&ctx->sha512, data, len);
 			break;
 	}
 }
@@ -479,32 +485,33 @@ SHA2Final(void *digest, SHA2_CTX *ctx)
 {
 	switch (ctx->algotype) {
 		case SHA256:
-			sha256_final(&ctx->sha256, digest, 256);
+			icp_sha256_final(&ctx->sha256, digest, 256);
 			break;
 		case SHA512:
 		case SHA512_HMAC_MECH_INFO_TYPE:
-			sha512_final(&ctx->sha512, digest, 512);
+			icp_sha512_final(&ctx->sha512, digest, 512);
 			break;
 		case SHA512_256:
-			sha512_final(&ctx->sha512, digest, 256);
+			icp_sha512_final(&ctx->sha512, digest, 256);
 			break;
 	}
 }
 
 /* the generic implementation is always okay */
-static boolean_t sha2_is_supported(void)
+static boolean_t
+icp_sha2_is_supported(void)
 {
 	return (B_TRUE);
 }
 
 const sha256_ops_t sha256_generic_impl = {
 	.name = "generic",
-	.transform = sha256_generic,
+	.transform = icp_sha256_generic,
-	.is_supported = sha2_is_supported
+	.is_supported = icp_sha2_is_supported
 };
 
 const sha512_ops_t sha512_generic_impl = {
 	.name = "generic",
-	.transform = sha512_generic,
+	.transform = icp_sha512_generic,
-	.is_supported = sha2_is_supported
+	.is_supported = icp_sha2_is_supported
 };

module/nvpair/nvpair.c

@@ -3246,7 +3246,8 @@ nvs_xdr_nvl_fini(nvstream_t *nvs)
  * xdrproc_t-compatible callbacks for xdr_array()
  */
 
-#if defined(_KERNEL) && defined(__linux__) /* Linux kernel */
+#if (defined(__FreeBSD_version) && __FreeBSD_version >= 1600010) || \
+    defined(_KERNEL) && defined(__linux__) /* Linux kernel */
 
 #define	NVS_BUILD_XDRPROC_T(type)		\
 static bool_t					\
@@ -3255,7 +3256,7 @@ nvs_xdr_nvp_##type(XDR *xdrs, void *ptr)	\
 	return (xdr_##type(xdrs, ptr));		\
 }
 
-#elif !defined(_KERNEL) && defined(XDR_CONTROL) /* tirpc */
+#elif !defined(_KERNEL) && defined(XDR_CONTROL) /* tirpc, FreeBSD < 16 */
 
 #define	NVS_BUILD_XDRPROC_T(type)		\
 static bool_t					\
@@ -3271,7 +3272,7 @@ nvs_xdr_nvp_##type(XDR *xdrs, ...)		\
 	return (xdr_##type(xdrs, ptr));		\
 }
 
-#else /* FreeBSD, sunrpc */
+#else /* FreeBSD kernel < 16, sunrpc */
 
 #define	NVS_BUILD_XDRPROC_T(type)		\
 static bool_t					\

module/os/freebsd/zfs/zfs_file_os.c

@@ -164,8 +164,9 @@ zfs_file_write(zfs_file_t *fp, const void *buf, size_t count, ssize_t *resid)
 
 int
 zfs_file_pwrite(zfs_file_t *fp, const void *buf, size_t count, loff_t off,
-    ssize_t *resid)
+    uint8_t ashift, ssize_t *resid)
 {
+	(void) ashift;
 	return (zfs_file_write_impl(fp, buf, count, &off, resid));
 }
 

module/os/freebsd/zfs/zfs_vnops_os.c

@@ -100,14 +100,6 @@
 
 VFS_SMR_DECLARE;
 
-#ifdef DEBUG_VFS_LOCKS
-#define	VNCHECKREF(vp)				  \
-	VNASSERT((vp)->v_holdcnt > 0 && (vp)->v_usecount > 0, vp,	\
-	    ("%s: wrong ref counts", __func__));
-#else
-#define	VNCHECKREF(vp)
-#endif
-
 #if __FreeBSD_version >= 1400045
 typedef uint64_t cookie_t;
 #else
@@ -965,9 +957,6 @@ zfs_create(znode_t *dzp, const char *name, vattr_t *vap, int excl, int mode,
 	zfs_acl_ids_t   acl_ids;
 	boolean_t	fuid_dirtied;
 	uint64_t	txtype;
-#ifdef DEBUG_VFS_LOCKS
-	vnode_t	*dvp = ZTOV(dzp);
-#endif
 
 	if (is_nametoolong(zfsvfs, name))
 		return (SET_ERROR(ENAMETOOLONG));
@@ -1097,7 +1086,8 @@ zfs_create(znode_t *dzp, const char *name, vattr_t *vap, int excl, int mode,
 	getnewvnode_drop_reserve();
 
 out:
-	VNCHECKREF(dvp);
+	VNASSERT(ZTOV(dzp)->v_holdcnt > 0 && ZTOV(dzp)->v_usecount > 0,
+	    ZTOV(dzp), ("%s: wrong ref counts", __func__));
 	if (error == 0) {
 		*zpp = zp;
 	}

module/os/freebsd/zfs/zio_crypt.c

@@ -371,9 +371,6 @@ error:
 	return (ret);
 }
 
-void *failed_decrypt_buf;
-int failed_decrypt_size;
-
 /*
  * This function handles all encryption and decryption in zfs. When
  * encrypting it expects puio to reference the plaintext and cuio to
@@ -1663,9 +1660,6 @@ error:
 	return (ret);
 }
 
-void *failed_decrypt_buf;
-int faile_decrypt_size;
-
 /*
  * Primary encryption / decryption entrypoint for zio data.
  */
@@ -1758,13 +1752,6 @@ zio_do_crypt_data(boolean_t encrypt, zio_crypt_key_t *key,
 	return (0);
 
 error:
-	if (!encrypt) {
-		if (failed_decrypt_buf != NULL)
-			kmem_free(failed_decrypt_buf, failed_decrypt_size);
-		failed_decrypt_buf = kmem_alloc(datalen, KM_SLEEP);
-		failed_decrypt_size = datalen;
-		memcpy(failed_decrypt_buf, cipherbuf, datalen);
-	}
 	if (locked)
 		rw_exit(&key->zk_salt_lock);
 	if (authbuf != NULL)

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

@@ -25,6 +25,10 @@
  * SUCH DAMAGE.
  */
 
+/*
+ * Copyright (c) 2025, Rob Norris <robn@despairlabs.com>
+ */
+
 #include <sys/types.h>
 #include <sys/sysmacros.h>
 #include <sys/kmem.h>
@@ -56,6 +60,19 @@ typedef struct zone_dataset {
 } zone_dataset_t;
 
 #ifdef CONFIG_USER_NS
+
+/*
+ * Linux 6.18 moved the generic namespace type away from ns->ops->type onto
+ * ns_common itself.
+ */
+#ifdef HAVE_NS_COMMON_TYPE
+#define	ns_is_newuser(ns)	\
+	((ns)->ns_type == CLONE_NEWUSER)
+#else
+#define	ns_is_newuser(ns)	\
+	((ns)->ops != NULL && (ns)->ops->type == CLONE_NEWUSER)
+#endif
+
 /*
  * Returns:
  * - 0 on success
@@ -84,7 +101,7 @@ user_ns_get(int fd, struct user_namespace **userns)
 		goto done;
 	}
 	ns = get_proc_ns(file_inode(nsfile));
-	if (ns->ops->type != CLONE_NEWUSER) {
+	if (!ns_is_newuser(ns)) {
 		error = ENOTTY;
 		goto done;
 	}

module/os/linux/zfs/abd_os.c

@@ -23,6 +23,7 @@
  * Copyright (c) 2014 by Chunwei Chen. All rights reserved.
  * Copyright (c) 2019 by Delphix. All rights reserved.
  * Copyright (c) 2023, 2024, Klara Inc.
+ * Copyright (c) 2025, Rob Norris <robn@despairlabs.com>
  */
 
 /*
@@ -891,6 +892,14 @@ abd_iter_advance(struct abd_iter *aiter, size_t amount)
 	}
 }
 
+#ifndef nth_page
+/*
+ * Since 6.18 nth_page() no longer exists, and is no longer required to iterate
+ * within a single SG entry, so we replace it with a simple addition.
+ */
+#define	nth_page(p, n)	((p)+(n))
+#endif
+
 /*
  * Map the current chunk into aiter. This can be safely called when the aiter
  * has already exhausted, in which case this does nothing.
@@ -918,7 +927,14 @@ abd_iter_map(struct abd_iter *aiter)
 		aiter->iter_mapsize = MIN(aiter->iter_sg->length - offset,
 		    aiter->iter_abd->abd_size - aiter->iter_pos);
 
-		paddr = zfs_kmap_local(sg_page(aiter->iter_sg));
+		struct page *page = sg_page(aiter->iter_sg);
+		if (PageHighMem(page)) {
+			page = nth_page(page, offset / PAGE_SIZE);
+			offset &= PAGE_SIZE - 1;
+			aiter->iter_mapsize = MIN(aiter->iter_mapsize,
+			    PAGE_SIZE - offset);
+		}
+		paddr = zfs_kmap_local(page);
 	}
 
 	aiter->iter_mapaddr = (char *)paddr + offset;
@@ -936,8 +952,14 @@ abd_iter_unmap(struct abd_iter *aiter)
 		return;
 
 	if (!abd_is_linear(aiter->iter_abd)) {
+		size_t offset = aiter->iter_offset;
+
+		struct page *page = sg_page(aiter->iter_sg);
+		if (PageHighMem(page))
+			offset &= PAGE_SIZE - 1;
+
 		/* LINTED E_FUNC_SET_NOT_USED */
-		zfs_kunmap_local(aiter->iter_mapaddr - aiter->iter_offset);
+		zfs_kunmap_local(aiter->iter_mapaddr - offset);
 	}
 
 	ASSERT3P(aiter->iter_mapaddr, !=, NULL);

module/os/linux/zfs/zfs_file_os.c

@@ -115,8 +115,9 @@ zfs_file_write(zfs_file_t *fp, const void *buf, size_t count, ssize_t *resid)
  */
 int
 zfs_file_pwrite(zfs_file_t *fp, const void *buf, size_t count, loff_t off,
-    ssize_t *resid)
+    uint8_t ashift, ssize_t *resid)
 {
+	(void) ashift;
 	ssize_t rc;
 
 	rc  = kernel_write(fp, buf, count, &off);

module/os/linux/zfs/zfs_uio.c

@@ -100,15 +100,17 @@ zfs_uiomove_bvec_impl(void *p, size_t n, zfs_uio_rw_t rw, zfs_uio_t *uio)
 
 	while (n && uio->uio_resid) {
 		void *paddr;
-		cnt = MIN(bv->bv_len - skip, n);
+		size_t offset = bv->bv_offset + skip;
+		cnt = MIN(PAGE_SIZE - (offset & ~PAGE_MASK),
+		    MIN(bv->bv_len - skip, n));
 
-		paddr = zfs_kmap_local(bv->bv_page);
+		paddr = zfs_kmap_local(bv->bv_page + (offset >> PAGE_SHIFT));
 		if (rw == UIO_READ) {
 			/* Copy from buffer 'p' to the bvec data */
-			memcpy(paddr + bv->bv_offset + skip, p, cnt);
+			memcpy(paddr + (offset & ~PAGE_MASK), p, cnt);
 		} else {
 			/* Copy from bvec data to buffer 'p' */
-			memcpy(p, paddr + bv->bv_offset + skip, cnt);
+			memcpy(p, paddr + (offset & ~PAGE_MASK), cnt);
 		}
 		zfs_kunmap_local(paddr);
 

module/os/linux/zfs/zfs_vfsops.c

@@ -1521,6 +1521,12 @@ zfs_domount(struct super_block *sb, zfs_mnt_t *zm, int silent)
 	sb->s_xattr = zpl_xattr_handlers;
 	sb->s_export_op = &zpl_export_operations;
 
+#ifdef HAVE_SET_DEFAULT_D_OP
+	set_default_d_op(sb, &zpl_dentry_operations);
+#else
+	sb->s_d_op = &zpl_dentry_operations;
+#endif
+
 	/* Set features for file system. */
 	zfs_set_fuid_feature(zfsvfs);
 

module/os/linux/zfs/zfs_vnops_os.c

@@ -3516,7 +3516,8 @@ zfs_link(znode_t *tdzp, znode_t *szp, char *name, cred_t *cr,
 	boolean_t	is_tmpfile = 0;
 	uint64_t	txg;
 
-	is_tmpfile = (sip->i_nlink == 0 && (sip->i_state & I_LINKABLE));
+	is_tmpfile = (sip->i_nlink == 0 &&
+	    (inode_state_read_once(sip) & I_LINKABLE));
 
 	ASSERT(S_ISDIR(ZTOI(tdzp)->i_mode));
 

module/os/linux/zfs/zpl_file.c

@@ -22,6 +22,7 @@
 /*
  * Copyright (c) 2011, Lawrence Livermore National Security, LLC.
  * Copyright (c) 2015 by Chunwei Chen. All rights reserved.
+ * Copyright (c) 2025, Rob Norris <robn@despairlabs.com>
  */
 
 
@@ -444,6 +445,7 @@ zpl_putpage(struct page *pp, struct writeback_control *wbc, void *data)
 	return (ret);
 }
 
+#ifdef HAVE_WRITE_CACHE_PAGES
 #ifdef HAVE_WRITEPAGE_T_FOLIO
 static int
 zpl_putfolio(struct folio *pp, struct writeback_control *wbc, void *data)
@@ -465,6 +467,78 @@ zpl_write_cache_pages(struct address_space *mapping,
 #endif
 	return (result);
 }
+#else
+static inline int
+zpl_write_cache_pages(struct address_space *mapping,
+    struct writeback_control *wbc, void *data)
+{
+	pgoff_t start = wbc->range_start >> PAGE_SHIFT;
+	pgoff_t end = wbc->range_end >> PAGE_SHIFT;
+
+	struct folio_batch fbatch;
+	folio_batch_init(&fbatch);
+
+	/*
+	 * This atomically (-ish) tags all DIRTY pages in the range with
+	 * TOWRITE, allowing users to continue dirtying or undirtying pages
+	 * while we get on with writeback, without us treading on each other.
+	 */
+	tag_pages_for_writeback(mapping, start, end);
+
+	int err = 0;
+	unsigned int npages;
+
+	/*
+	 * Grab references to the TOWRITE pages just flagged. This may not get
+	 * all of them, so we do it in a loop until there are none left.
+	 */
+	while ((npages = filemap_get_folios_tag(mapping, &start, end,
+	    PAGECACHE_TAG_TOWRITE, &fbatch)) != 0) {
+
+		/* Loop over each page and write it out. */
+		struct folio *folio;
+		while ((folio = folio_batch_next(&fbatch)) != NULL) {
+			folio_lock(folio);
+
+			/*
+			 * If the folio has been remapped, or is no longer
+			 * dirty, then there's nothing to do.
+			 */
+			if (folio->mapping != mapping ||
+			    !folio_test_dirty(folio)) {
+				folio_unlock(folio);
+				continue;
+			}
+
+			/*
+			 * If writeback is already in progress, wait for it to
+			 * finish. We continue after this even if the page
+			 * ends up clean; zfs_putpage() will skip it if no
+			 * further work is required.
+			 */
+			while (folio_test_writeback(folio))
+				folio_wait_bit(folio, PG_writeback);
+
+			/*
+			 * Write it out and collect any error. zfs_putpage()
+			 * will clear the TOWRITE and DIRTY flags, and return
+			 * with the page unlocked.
+			 */
+			int ferr = zpl_putpage(&folio->page, wbc, data);
+			if (err == 0 && ferr != 0)
+				err = ferr;
+
+			/* Housekeeping for the caller. */
+			wbc->nr_to_write -= folio_nr_pages(folio);
+		}
+
+		/* Release any remaining references on the batch. */
+		folio_batch_release(&fbatch);
+	}
+
+	return (err);
+}
+#endif
 
 static int
 zpl_writepages(struct address_space *mapping, struct writeback_control *wbc)

module/os/linux/zfs/zpl_super.c

@@ -22,6 +22,8 @@
 /*
  * Copyright (c) 2011, Lawrence Livermore National Security, LLC.
  * Copyright (c) 2023, Datto Inc. All rights reserved.
+ * Copyright (c) 2025, Klara, Inc.
+ * Copyright (c) 2025, Rob Norris <robn@despairlabs.com>
  */
 
 
@@ -32,7 +34,22 @@
 #include <sys/zpl.h>
 #include <linux/iversion.h>
 #include <linux/version.h>
+#include <linux/vfs_compat.h>
 
+/*
+ * What to do when the last reference to an inode is released. If 0, the kernel
+ * will cache it on the superblock. If 1, the inode will be freed immediately.
+ * See zpl_drop_inode().
+ */
+int zfs_delete_inode = 0;
+
+/*
+ * What to do when the last reference to a dentry is released. If 0, the kernel
+ * will cache it until the entry (file) is destroyed. If 1, the dentry will be
+ * marked for cleanup, at which time its inode reference will be released. See
+ * zpl_dentry_delete().
+ */
+int zfs_delete_dentry = 0;
 
 static struct inode *
 zpl_inode_alloc(struct super_block *sb)
@@ -77,11 +94,36 @@ zpl_dirty_inode(struct inode *ip, int flags)
 }
 
 /*
- * When ->drop_inode() is called its return value indicates if the
+ * ->drop_inode() is called when the last reference to an inode is released.
- * inode should be evicted from the inode cache.  If the inode is
+ * Its return value indicates if the inode should be destroyed immediately, or
- * unhashed and has no links the default policy is to evict it
+ * cached on the superblock structure.
- * immediately.
  *
+ * By default (zfs_delete_inode=0), we call generic_drop_inode(), which returns
+ * "destroy immediately" if the inode is unhashed and has no links (roughly: no
+ * longer exists on disk). On datasets with millions of rarely-accessed files,
+ * this can cause a large amount of memory to be "pinned" by cached inodes,
+ * which in turn pin their associated dnodes and dbufs, until the kernel starts
+ * reporting memory pressure and requests OpenZFS release some memory (see
+ * zfs_prune()).
+ *
+ * When set to 1, we call generic_delete_inode(), which always returns "destroy
+ * immediately", resulting in inodes being destroyed immediately, releasing
+ * their associated dnodes and dbufs to the dbuf cached and the ARC to be
+ * evicted as normal.
+ *
+ * Note that the "last reference" doesn't always mean the last _userspace_
+ * reference; the dentry cache also holds a reference, so "busy" inodes will
+ * still be kept alive that way (subject to dcache tuning).
+ */
+static int
+zpl_drop_inode(struct inode *ip)
+{
+	if (zfs_delete_inode)
+		return (generic_delete_inode(ip));
+	return (generic_drop_inode(ip));
+}
+
+/*
  * The ->evict_inode() callback must minimally truncate the inode pages,
  * and call clear_inode().  For 2.6.35 and later kernels this will
  * simply update the inode state, with the sync occurring before the
@@ -470,6 +512,7 @@ const struct super_operations zpl_super_operations = {
 	.destroy_inode		= zpl_inode_destroy,
 	.dirty_inode		= zpl_dirty_inode,
 	.write_inode		= NULL,
+	.drop_inode		= zpl_drop_inode,
 	.evict_inode		= zpl_evict_inode,
 	.put_super		= zpl_put_super,
 	.sync_fs		= zpl_sync_fs,
@@ -480,6 +523,35 @@ const struct super_operations zpl_super_operations = {
 	.show_stats		= NULL,
 };
 
+/*
+ * ->d_delete() is called when the last reference to a dentry is released. Its
+ *  return value indicates if the dentry should be destroyed immediately, or
+ *  retained in the dentry cache.
+ *
+ * By default (zfs_delete_dentry=0) the kernel will always cache unused
+ * entries.  Each dentry holds an inode reference, so cached dentries can hold
+ * the final inode reference indefinitely, leading to the inode and its related
+ * data being pinned (see zpl_drop_inode()).
+ *
+ * When set to 1, we signal that the dentry should be destroyed immediately and
+ * never cached. This reduces memory usage, at the cost of higher overheads to
+ * lookup a file, as the inode and its underlying data (dnode/dbuf) need to be
+ * reloaded and reinflated.
+ *
+ * Note that userspace does not have direct control over dentry references and
+ * reclaim; rather, this is part of the kernel's caching and reclaim subsystems
+ * (eg vm.vfs_cache_pressure).
+ */
+static int
+zpl_dentry_delete(const struct dentry *dentry)
+{
+	return (zfs_delete_dentry ? 1 : 0);
+}
+
+const struct dentry_operations zpl_dentry_operations = {
+	.d_delete = zpl_dentry_delete,
+};
+
 struct file_system_type zpl_fs_type = {
 	.owner			= THIS_MODULE,
 	.name			= ZFS_DRIVER,
@@ -491,3 +563,10 @@ struct file_system_type zpl_fs_type = {
 	.mount			= zpl_mount,
 	.kill_sb		= zpl_kill_sb,
 };
+
+ZFS_MODULE_PARAM(zfs, zfs_, delete_inode, INT, ZMOD_RW,
+	"Delete inodes as soon as the last reference is released.");
+
+ZFS_MODULE_PARAM(zfs, zfs_, delete_dentry, INT, ZMOD_RW,
+	"Delete dentries from dentry cache as soon as the last reference is "
+	"released.");

module/os/linux/zfs/zvol_os.c

@@ -21,7 +21,7 @@
  */
 /*
  * Copyright (c) 2012, 2020 by Delphix. All rights reserved.
- * Copyright (c) 2024, Rob Norris <robn@despairlabs.com>
+ * Copyright (c) 2024, 2025, Rob Norris <robn@despairlabs.com>
  * Copyright (c) 2024, Klara, Inc.
  */
 
@@ -1066,12 +1066,13 @@ zvol_os_clear_private(zvol_state_t *zv)
  * tiny devices.  For devices over 1 Mib a standard head and sector count
  * is used to keep the cylinders count reasonable.
  */
-static int
+static inline int
-zvol_getgeo(struct block_device *bdev, struct hd_geometry *geo)
+zvol_getgeo_impl(struct gendisk *disk, struct hd_geometry *geo)
 {
-	zvol_state_t *zv = bdev->bd_disk->private_data;
+	zvol_state_t *zv = disk->private_data;
 	sector_t sectors;
 
+	ASSERT3P(zv, !=, NULL);
 	ASSERT3U(zv->zv_open_count, >, 0);
 
 	sectors = get_capacity(zv->zv_zso->zvo_disk);
@@ -1090,6 +1091,20 @@ zvol_getgeo(struct block_device *bdev, struct hd_geometry *geo)
 	return (0);
 }
 
+#ifdef HAVE_BLOCK_DEVICE_OPERATIONS_GETGEO_GENDISK
+static int
+zvol_getgeo(struct gendisk *disk, struct hd_geometry *geo)
+{
+	return (zvol_getgeo_impl(disk, geo));
+}
+#else
+static int
+zvol_getgeo(struct block_device *bdev, struct hd_geometry *geo)
+{
+	return (zvol_getgeo_impl(bdev->bd_disk, geo));
+}
+#endif
+
 /*
  * Why have two separate block_device_operations structs?
  *
@@ -1531,7 +1546,7 @@ zvol_os_free(zvol_state_t *zv)
 	if (zv->zv_zso->use_blk_mq)
 		blk_mq_free_tag_set(&zv->zv_zso->tag_set);
 
-	ida_simple_remove(&zvol_ida,
+	ida_free(&zvol_ida,
 	    MINOR(zv->zv_zso->zvo_dev) >> ZVOL_MINOR_BITS);
 
 	cv_destroy(&zv->zv_removing_cv);
@@ -1665,7 +1680,7 @@ zvol_os_create_minor(const char *name)
 	if (zvol_inhibit_dev)
 		return (0);
 
-	idx = ida_simple_get(&zvol_ida, 0, 0, kmem_flags_convert(KM_SLEEP));
+	idx = ida_alloc(&zvol_ida, kmem_flags_convert(KM_SLEEP));
 	if (idx < 0)
 		return (SET_ERROR(-idx));
 	minor = idx << ZVOL_MINOR_BITS;
@@ -1673,7 +1688,7 @@ zvol_os_create_minor(const char *name)
 		/* too many partitions can cause an overflow */
 		zfs_dbgmsg("zvol: create minor overflow: %s, minor %u/%u",
 		    name, minor, MINOR(minor));
-		ida_simple_remove(&zvol_ida, idx);
+		ida_free(&zvol_ida, idx);
 		return (SET_ERROR(EINVAL));
 	}
 
@@ -1681,7 +1696,7 @@ zvol_os_create_minor(const char *name)
 	if (zv) {
 		ASSERT(MUTEX_HELD(&zv->zv_state_lock));
 		mutex_exit(&zv->zv_state_lock);
-		ida_simple_remove(&zvol_ida, idx);
+		ida_free(&zvol_ida, idx);
 		return (SET_ERROR(EEXIST));
 	}
 
@@ -1783,7 +1798,7 @@ out_doi:
 		rw_exit(&zvol_state_lock);
 		error = zvol_os_add_disk(zv->zv_zso->zvo_disk);
 	} else {
-		ida_simple_remove(&zvol_ida, idx);
+		ida_free(&zvol_ida, idx);
 	}
 
 	return (error);

module/zfs/abd.c

@@ -1111,13 +1111,6 @@ abd_raidz_gen_iterate(abd_t **cabds, abd_t *dabd, size_t off,
 
 		func_raidz_gen(caddrs, daddr, len, dlen);
 
-		for (i = parity-1; i >= 0; i--) {
-			abd_iter_unmap(&caiters[i]);
-			c_cabds[i] =
-			    abd_advance_abd_iter(cabds[i], c_cabds[i],
-			    &caiters[i], len);
-		}
-
 		if (dsize > 0) {
 			abd_iter_unmap(&daiter);
 			c_dabd =
@@ -1126,6 +1119,13 @@ abd_raidz_gen_iterate(abd_t **cabds, abd_t *dabd, size_t off,
 			dsize -= dlen;
 		}
 
+		for (i = parity - 1; i >= 0; i--) {
+			abd_iter_unmap(&caiters[i]);
+			c_cabds[i] =
+			    abd_advance_abd_iter(cabds[i], c_cabds[i],
+			    &caiters[i], len);
+		}
+
 		csize -= len;
 	}
 	abd_exit_critical(flags);
@@ -1194,7 +1194,7 @@ abd_raidz_rec_iterate(abd_t **cabds, abd_t **tabds,
 
 		func_raidz_rec(xaddrs, len, caddrs, mul);
 
-		for (i = parity-1; i >= 0; i--) {
+		for (i = parity - 1; i >= 0; i--) {
 			abd_iter_unmap(&xiters[i]);
 			abd_iter_unmap(&citers[i]);
 			c_tabds[i] =

module/zfs/dataset_kstats.c

@@ -212,7 +212,7 @@ dataset_kstats_rename(dataset_kstats_t *dk, const char *name)
 	char *ds_name;
 
 	ds_name = KSTAT_NAMED_STR_PTR(&dkv->dkv_ds_name);
-	ASSERT3S(ds_name, !=, NULL);
+	ASSERT3P(ds_name, !=, NULL);
 	(void) strlcpy(ds_name, name,
 	    KSTAT_NAMED_STR_BUFLEN(&dkv->dkv_ds_name));
 }

module/zfs/ddt.c

@@ -1508,7 +1508,7 @@ ddt_configure(ddt_t *ddt, boolean_t new)
 		    DMU_POOL_DIRECTORY_OBJECT, name, sizeof (uint64_t), 1,
 		    &ddt->ddt_dir_object);
 		if (error == 0) {
-			ASSERT3U(spa->spa_meta_objset, ==, ddt->ddt_os);
+			ASSERT3P(spa->spa_meta_objset, ==, ddt->ddt_os);
 
 			error = zap_lookup(ddt->ddt_os, ddt->ddt_dir_object,
 			    DDT_DIR_VERSION, sizeof (uint64_t), 1,

module/zfs/ddt_log.c

@@ -262,7 +262,7 @@ ddt_log_update_entry(ddt_t *ddt, ddt_log_t *ddl, ddt_lightweight_entry_t *ddlwe)
 void
 ddt_log_entry(ddt_t *ddt, ddt_lightweight_entry_t *ddlwe, ddt_log_update_t *dlu)
 {
-	ASSERT3U(dlu->dlu_dbp, !=, NULL);
+	ASSERT3P(dlu->dlu_dbp, !=, NULL);
 
 	ddt_log_update_entry(ddt, ddt->ddt_log_active, ddlwe);
 	ddt_histogram_add_entry(ddt, &ddt->ddt_log_histogram, ddlwe);
@@ -312,7 +312,7 @@ ddt_log_entry(ddt_t *ddt, ddt_lightweight_entry_t *ddlwe, ddt_log_update_t *dlu)
 void
 ddt_log_commit(ddt_t *ddt, ddt_log_update_t *dlu)
 {
-	ASSERT3U(dlu->dlu_dbp, !=, NULL);
+	ASSERT3P(dlu->dlu_dbp, !=, NULL);
 	ASSERT3U(dlu->dlu_block+1, ==, dlu->dlu_ndbp);
 	ASSERT3U(dlu->dlu_offset, >, 0);
 

module/zfs/mmp.c

@@ -145,6 +145,15 @@
  * Additionally, the duration is then extended by a random 25% to attempt to to
  * detect simultaneous imports.  For example, if both partner hosts are rebooted
  * at the same time and automatically attempt to import the pool.
+ *
+ * Once the read-only activity check completes and the pool is determined to
+ * be inactive a second check is performed to claim the pool.  During this
+ * phase the host writes out MMP uberblocks to each of the devices which are
+ * identical to the best uberblock but with a randomly selected sequence id.
+ * The "best" uberblock is then read back and it must contain this new sequence
+ * number.  This check is performed multiple times to ensure that there is
+ * no window where a concurrently importing system can incorrectly determine
+ * the pool to be inactive.
  */
 
 /*
@@ -237,8 +246,8 @@ mmp_thread_start(spa_t *spa)
 		if (!mmp->mmp_thread) {
 			mmp->mmp_thread = thread_create(NULL, 0, mmp_thread,
 			    spa, 0, &p0, TS_RUN, defclsyspri);
-			zfs_dbgmsg("MMP thread started pool '%s' "
+			zfs_dbgmsg("mmp: mmp thread started spa=%s "
-			    "gethrtime %llu", spa_name(spa), gethrtime());
+			    "gethrtime=%llu", spa_name(spa), gethrtime());
 		}
 		mutex_exit(&mmp->mmp_thread_lock);
 	}
@@ -257,7 +266,7 @@ mmp_thread_stop(spa_t *spa)
 		cv_wait(&mmp->mmp_thread_cv, &mmp->mmp_thread_lock);
 	}
 	mutex_exit(&mmp->mmp_thread_lock);
-	zfs_dbgmsg("MMP thread stopped pool '%s' gethrtime %llu",
+	zfs_dbgmsg("mmp: mmp thread stopped spa=%s gethrtime=%llu",
 	    spa_name(spa), gethrtime());
 
 	ASSERT(mmp->mmp_thread == NULL);
@@ -449,9 +458,9 @@ mmp_write_uberblock(spa_t *spa)
 	spa_config_enter_mmp(spa, SCL_STATE, mmp_tag, RW_READER);
 	lock_acquire_time = gethrtime() - lock_acquire_time;
 	if (lock_acquire_time > (MSEC2NSEC(MMP_MIN_INTERVAL) / 10))
-		zfs_dbgmsg("MMP SCL_STATE acquisition pool '%s' took %llu ns "
+		zfs_dbgmsg("mmp: long SCL_STATE acquisition, spa=%s "
-		    "gethrtime %llu", spa_name(spa), lock_acquire_time,
+		    "acquire_time=%llu gethrtime=%llu", spa_name(spa),
-		    gethrtime());
+		    lock_acquire_time, gethrtime());
 
 	mutex_enter(&mmp->mmp_io_lock);
 
@@ -474,8 +483,8 @@ mmp_write_uberblock(spa_t *spa)
 			spa_mmp_history_add(spa, mmp->mmp_ub.ub_txg,
 			    gethrestime_sec(), mmp->mmp_delay, NULL, 0,
 			    mmp->mmp_kstat_id++, error);
-			zfs_dbgmsg("MMP error choosing leaf pool '%s' "
+			zfs_dbgmsg("mmp: error choosing leaf, spa=%s "
-			    "gethrtime %llu fail_mask %#x", spa_name(spa),
+			    "gethrtime=%llu fail_mask=%#x", spa_name(spa),
 			    gethrtime(), error);
 		}
 		mutex_exit(&mmp->mmp_io_lock);
@@ -485,11 +494,11 @@ mmp_write_uberblock(spa_t *spa)
 
 	vd = spa->spa_mmp.mmp_last_leaf;
 	if (mmp->mmp_skip_error != 0) {
-		mmp->mmp_skip_error = 0;
+		zfs_dbgmsg("mmp: write after skipping due to unavailable "
-		zfs_dbgmsg("MMP write after skipping due to unavailable "
+		    "leaves, spa=%s gethrtime=%llu vdev=%llu error=%d",
-		    "leaves, pool '%s' gethrtime %llu leaf %llu",
 		    spa_name(spa), (u_longlong_t)gethrtime(),
-		    (u_longlong_t)vd->vdev_guid);
+		    (u_longlong_t)vd->vdev_guid, mmp->mmp_skip_error);
+		mmp->mmp_skip_error = 0;
 	}
 
 	if (mmp->mmp_zio_root == NULL)
@@ -540,6 +549,108 @@ mmp_write_uberblock(spa_t *spa)
 	zio_nowait(zio);
 }
 
+static void
+mmp_claim_uberblock_sync_done(zio_t *zio)
+{
+	uint64_t *good_writes = zio->io_private;
+
+	if (zio->io_error == 0 && zio->io_vd->vdev_top->vdev_ms_array != 0)
+		atomic_inc_64(good_writes);
+}
+
+/*
+ * Write the uberblock to the first label of all leaves of the specified vdev.
+ * Two writes required for each mirror, one for a singleton, and parity+1 for
+ * raidz or draid vdevs.
+ */
+static void
+mmp_claim_uberblock_sync(zio_t *zio, uint64_t *good_writes,
+    uint64_t *req_writes, uberblock_t *ub, vdev_t *vd, int flags)
+{
+	for (uint64_t c = 0; c < vd->vdev_children; c++) {
+		vdev_t *cvd = vd->vdev_child[c];
+
+		if (cvd->vdev_islog || cvd->vdev_isspare || cvd->vdev_isl2cache)
+			continue;
+
+		if (cvd->vdev_top == cvd) {
+			uint64_t nparity = vdev_get_nparity(cvd);
+			if (nparity) {
+				*req_writes += nparity + 1;
+			} else {
+				*req_writes +=
+				    MIN(MAX(cvd->vdev_children, 1), 2);
+			}
+		}
+
+		mmp_claim_uberblock_sync(zio, good_writes, req_writes,
+		    ub, cvd, flags);
+	}
+
+	if (!vd->vdev_ops->vdev_op_leaf)
+		return;
+
+	if (!vdev_writeable(vd))
+		return;
+
+	if (vd->vdev_ops == &vdev_draid_spare_ops)
+		return;
+
+	abd_t *ub_abd = abd_alloc_for_io(VDEV_UBERBLOCK_SIZE(vd), B_TRUE);
+	abd_copy_from_buf(ub_abd, ub, sizeof (uberblock_t));
+	abd_zero_off(ub_abd, sizeof (uberblock_t),
+	    VDEV_UBERBLOCK_SIZE(vd) - sizeof (uberblock_t));
+
+	vdev_label_write(zio, vd, 0, ub_abd,
+	    VDEV_UBERBLOCK_OFFSET(vd, VDEV_UBERBLOCK_COUNT(vd) -
+	    MMP_BLOCKS_PER_LABEL), VDEV_UBERBLOCK_SIZE(vd),
+	    mmp_claim_uberblock_sync_done, good_writes,
+	    flags | ZIO_FLAG_DONT_PROPAGATE);
+
+	abd_free(ub_abd);
+}
+
+int
+mmp_claim_uberblock(spa_t *spa, vdev_t *vd, uberblock_t *ub)
+{
+	int flags = ZIO_FLAG_CONFIG_WRITER | ZIO_FLAG_CANFAIL;
+	uint64_t good_writes = 0;
+	uint64_t req_writes = 0;
+	zio_t *zio;
+
+	ASSERT(MMP_VALID(ub));
+	ASSERT(MMP_SEQ_VALID(ub));
+
+	spa_config_enter(spa, SCL_ALL, mmp_tag, RW_WRITER);
+
+	/* Sync the uberblock to all writeable leaves */
+	zio = zio_root(spa, NULL, NULL, flags);
+	mmp_claim_uberblock_sync(zio, &good_writes, &req_writes, ub, vd, flags);
+	(void) zio_wait(zio);
+
+	/* Flush the new uberblocks so they're immediately visible */
+	zio = zio_root(spa, NULL, NULL, flags);
+	zio_flush(zio, vd);
+	(void) zio_wait(zio);
+
+	spa_config_exit(spa, SCL_ALL, mmp_tag);
+
+	zfs_dbgmsg("mmp: claiming uberblock, spa=%s txg=%llu seq=%llu "
+	    "req_writes=%llu good_writes=%llu", spa_load_name(spa),
+	    (u_longlong_t)ub->ub_txg, (u_longlong_t)MMP_SEQ(ub),
+	    (u_longlong_t)req_writes, (u_longlong_t)good_writes);
+
+	/*
+	 * To guarantee visibility from a remote host we require a minimum
+	 * number of good writes. For raidz/draid vdevs parity+1 writes, for
+	 * mirrors 2 writes, and for singletons 1 write.
+	 */
+	if (req_writes == 0 || good_writes < req_writes)
+		return (SET_ERROR(EIO));
+
+	return (0);
+}
+
 static __attribute__((noreturn)) void
 mmp_thread(void *arg)
 {
@@ -616,11 +727,11 @@ mmp_thread(void *arg)
 			next_time = gethrtime() + mmp_interval / leaves;
 
 		if (mmp_fail_ns != last_mmp_fail_ns) {
-			zfs_dbgmsg("MMP interval change pool '%s' "
+			zfs_dbgmsg("mmp: interval change, spa=%s "
-			    "gethrtime %llu last_mmp_interval %llu "
+			    "gethrtime=%llu last_mmp_interval=%llu "
-			    "mmp_interval %llu last_mmp_fail_intervals %u "
+			    "mmp_interval=%llu last_mmp_fail_intervals=%u "
-			    "mmp_fail_intervals %u mmp_fail_ns %llu "
+			    "mmp_fail_intervals=%u mmp_fail_ns=%llu "
-			    "skip_wait %d leaves %d next_time %llu",
+			    "skip_wait=%d leaves=%d next_time=%llu",
 			    spa_name(spa), (u_longlong_t)gethrtime(),
 			    (u_longlong_t)last_mmp_interval,
 			    (u_longlong_t)mmp_interval, last_mmp_fail_intervals,
@@ -635,9 +746,9 @@ mmp_thread(void *arg)
 		 */
 		if ((!last_spa_multihost && multihost) ||
 		    (last_spa_suspended && !suspended)) {
-			zfs_dbgmsg("MMP state change pool '%s': gethrtime %llu "
+			zfs_dbgmsg("mmp: state change spa=%s: gethrtime=%llu "
-			    "last_spa_multihost %u multihost %u "
+			    "last_spa_multihost=%u multihost=%u "
-			    "last_spa_suspended %u suspended %u",
+			    "last_spa_suspended=%u suspended=%u",
 			    spa_name(spa), (u_longlong_t)gethrtime(),
 			    last_spa_multihost, multihost, last_spa_suspended,
 			    suspended);
@@ -663,9 +774,10 @@ mmp_thread(void *arg)
 		 */
 		if (multihost && !suspended && mmp_fail_intervals &&
 		    (gethrtime() - mmp->mmp_last_write) > mmp_fail_ns) {
-			zfs_dbgmsg("MMP suspending pool '%s': gethrtime %llu "
+			zfs_dbgmsg("mmp: suspending pool, spa=%s "
-			    "mmp_last_write %llu mmp_interval %llu "
+			    "gethrtime=%llu mmp_last_write=%llu "
-			    "mmp_fail_intervals %llu mmp_fail_ns %llu txg %llu",
+			    "mmp_interval=%llu mmp_fail_intervals=%llu "
+			    "mmp_fail_ns=%llu txg=%llu",
 			    spa_name(spa), (u_longlong_t)gethrtime(),
 			    (u_longlong_t)mmp->mmp_last_write,
 			    (u_longlong_t)mmp_interval,

module/zfs/spa_misc.c

@@ -413,7 +413,7 @@ spa_load_failed(spa_t *spa, const char *fmt, ...)
 	(void) vsnprintf(buf, sizeof (buf), fmt, adx);
 	va_end(adx);
 
-	zfs_dbgmsg("spa_load(%s, config %s): FAILED: %s", spa->spa_name,
+	zfs_dbgmsg("spa_load(%s, config %s): FAILED: %s", spa_load_name(spa),
 	    spa->spa_trust_config ? "trusted" : "untrusted", buf);
 }
 
@@ -427,7 +427,7 @@ spa_load_note(spa_t *spa, const char *fmt, ...)
 	(void) vsnprintf(buf, sizeof (buf), fmt, adx);
 	va_end(adx);
 
-	zfs_dbgmsg("spa_load(%s, config %s): %s", spa->spa_name,
+	zfs_dbgmsg("spa_load(%s, config %s): %s", spa_load_name(spa),
 	    spa->spa_trust_config ? "trusted" : "untrusted", buf);
 
 	spa_import_progress_set_notes_nolog(spa, "%s", buf);
@@ -857,6 +857,9 @@ spa_remove(spa_t *spa)
 	if (spa->spa_root)
 		spa_strfree(spa->spa_root);
 
+	if (spa->spa_load_name)
+		spa_strfree(spa->spa_load_name);
+
 	while ((dp = list_remove_head(&spa->spa_config_list)) != NULL) {
 		if (dp->scd_path != NULL)
 			spa_strfree(dp->scd_path);
@@ -1242,7 +1245,7 @@ spa_vdev_enter(spa_t *spa)
 	mutex_enter(&spa->spa_vdev_top_lock);
 	mutex_enter(&spa_namespace_lock);
 
-	ASSERT0(spa->spa_export_thread);
+	ASSERT0P(spa->spa_export_thread);
 
 	vdev_autotrim_stop_all(spa);
 
@@ -1261,7 +1264,7 @@ spa_vdev_detach_enter(spa_t *spa, uint64_t guid)
 	mutex_enter(&spa->spa_vdev_top_lock);
 	mutex_enter(&spa_namespace_lock);
 
-	ASSERT0(spa->spa_export_thread);
+	ASSERT0P(spa->spa_export_thread);
 
 	vdev_autotrim_stop_all(spa);
 
@@ -1777,6 +1780,19 @@ spa_name(spa_t *spa)
 	return (spa->spa_name);
 }
 
+char *
+spa_load_name(spa_t *spa)
+{
+	/*
+	 * During spa_tryimport() the pool name includes a unique prefix.
+	 * Returns the original name which can be used for log messages.
+	 */
+	if (spa->spa_load_name)
+		return (spa->spa_load_name);
+
+	return (spa->spa_name);
+}
+
 uint64_t
 spa_guid(spa_t *spa)
 {
@@ -3089,6 +3105,7 @@ EXPORT_SYMBOL(spa_set_rootblkptr);
 EXPORT_SYMBOL(spa_altroot);
 EXPORT_SYMBOL(spa_sync_pass);
 EXPORT_SYMBOL(spa_name);
+EXPORT_SYMBOL(spa_load_name);
 EXPORT_SYMBOL(spa_guid);
 EXPORT_SYMBOL(spa_last_synced_txg);
 EXPORT_SYMBOL(spa_first_txg);

module/zfs/vdev_file.c

@@ -228,7 +228,8 @@ vdev_file_io_strategy(void *arg)
 		abd_return_buf_copy(zio->io_abd, buf, size);
 	} else {
 		buf = abd_borrow_buf_copy(zio->io_abd, zio->io_size);
-		err = zfs_file_pwrite(vf->vf_file, buf, size, off, &resid);
+		err = zfs_file_pwrite(vf->vf_file, buf, size, off,
+		    vd->vdev_ashift, &resid);
 		abd_return_buf(zio->io_abd, buf, size);
 	}
 	zio->io_error = err;

module/zfs/vdev_label.c

@@ -1500,7 +1500,7 @@ retry:
  * conflicting uberblocks on disk with the same txg.  The solution is simple:
  * among uberblocks with equal txg, choose the one with the latest timestamp.
  */
-static int
+int
 vdev_uberblock_compare(const uberblock_t *ub1, const uberblock_t *ub2)
 {
 	int cmp = TREE_CMP(ub1->ub_txg, ub2->ub_txg);
@@ -1631,8 +1631,10 @@ vdev_uberblock_load(vdev_t *rvd, uberblock_t *ub, nvlist_t **config)
 	 * matches the txg for our uberblock.
 	 */
 	if (cb.ubl_vd != NULL) {
-		vdev_dbgmsg(cb.ubl_vd, "best uberblock found for spa %s. "
+		vdev_dbgmsg(cb.ubl_vd, "best uberblock found for spa %s, "
-		    "txg %llu", spa->spa_name, (u_longlong_t)ub->ub_txg);
+		    "txg=%llu seq=%llu", spa_load_name(spa),
+		    (u_longlong_t)ub->ub_txg,
+		    (u_longlong_t)(MMP_SEQ_VALID(ub) ? MMP_SEQ(ub) : 0));
 
 		if (ub->ub_raidz_reflow_info !=
 		    cb.ubl_latest.ub_raidz_reflow_info) {
@@ -1640,7 +1642,7 @@ vdev_uberblock_load(vdev_t *rvd, uberblock_t *ub, nvlist_t **config)
 			    "spa=%s best uberblock (txg=%llu info=0x%llx) "
 			    "has different raidz_reflow_info than latest "
 			    "uberblock (txg=%llu info=0x%llx)",
-			    spa->spa_name,
+			    spa_load_name(spa),
 			    (u_longlong_t)ub->ub_txg,
 			    (u_longlong_t)ub->ub_raidz_reflow_info,
 			    (u_longlong_t)cb.ubl_latest.ub_txg,

module/zfs/zfs_chksum.c

@@ -155,11 +155,11 @@ chksum_run(chksum_stat_t *cs, abd_t *abd, void *ctx, int round,
 	switch (round) {
 	case 1: /* 1k */
 		size = 1<<10; loops = 128; break;
-	case 2: /* 2k */
+	case 2: /* 4k */
 		size = 1<<12; loops = 64; break;
-	case 3: /* 4k */
+	case 3: /* 16k */
 		size = 1<<14; loops = 32; break;
-	case 4: /* 16k */
+	case 4: /* 64k */
 		size = 1<<16; loops = 16; break;
 	case 5: /* 256k */
 		size = 1<<18; loops = 8; break;
@@ -212,6 +212,7 @@ chksum_benchit(chksum_stat_t *cs)
 	chksum_run(cs, abd, ctx, 2, &cs->bs4k);
 	chksum_run(cs, abd, ctx, 3, &cs->bs16k);
 	chksum_run(cs, abd, ctx, 4, &cs->bs64k);
+	chksum_run(cs, abd, ctx, 5, &cs->bs256k);
 	chksum_run(cs, abd, ctx, 6, &cs->bs1m);
 	abd_free(abd);
 
@@ -249,15 +250,16 @@ chksum_benchmark(void)
 	if (chksum_stat_limit == AT_DONE)
 		return;
 
-
 	/* count implementations */
-	chksum_stat_cnt = 1;  /* edonr */
+	if (chksum_stat_limit == AT_STARTUP) {
-	chksum_stat_cnt += 1; /* skein */
+		chksum_stat_cnt = 1;  /* edonr */
-	chksum_stat_cnt += sha256->getcnt();
+		chksum_stat_cnt += 1; /* skein */
-	chksum_stat_cnt += sha512->getcnt();
+		chksum_stat_cnt += sha256->getcnt();
-	chksum_stat_cnt += blake3->getcnt();
+		chksum_stat_cnt += sha512->getcnt();
-	chksum_stat_data = kmem_zalloc(
+		chksum_stat_cnt += blake3->getcnt();
-	    sizeof (chksum_stat_t) * chksum_stat_cnt, KM_SLEEP);
+		chksum_stat_data = kmem_zalloc(
+		    sizeof (chksum_stat_t) * chksum_stat_cnt, KM_SLEEP);
+	}
 
 	/* edonr - needs to be the first one here (slow CPU check) */
 	cs = &chksum_stat_data[cbid++];

module/zfs/zio_compress.c

@@ -128,7 +128,7 @@ zio_compress_data(enum zio_compress c, abd_t *src, abd_t **dst, size_t s_len,
 	uint8_t complevel;
 	zio_compress_info_t *ci = &zio_compress_table[c];
 
-	ASSERT3U(ci->ci_compress, !=, NULL);
+	ASSERT3P(ci->ci_compress, !=, NULL);
 	ASSERT3U(s_len, >, 0);
 
 	complevel = ci->ci_level;

scripts/commitcheck.sh

@@ -7,7 +7,9 @@ REF="HEAD"
 test_commit_bodylength()
 {
     length="72"
-    body=$(git log --no-show-signature -n 1 --pretty=%b "$REF" | grep -Ev "http(s)*://" | grep -E -m 1 ".{$((length + 1))}")
+    body=$(git log --no-show-signature -n 1 --pretty=%b "$REF" |
+        grep -Evi -e "http(s)*://" -e "signed-off-by:" -e "reviewed-by:" |
+        grep -E -m 1 ".{$((length + 1))}")
     if [ -n "$body" ]; then
         echo "error: commit message body contains line over ${length} characters"
         return 1

scripts/objtool-wrapper.in

@@ -22,10 +22,10 @@
 # FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS
 # IN THE SOFTWARE.
 
-# Filter out objtools '--Werror' flag.
+# Filter out objtools '--Werror or --werror' flag.
 
 objtool="@abs_objtool_binary@"
-args=$(echo "$*" | sed s/--Werror//)
+args=$(echo "$*" | sed 's/--Werror\|--werror//')
 
 if [ -z "$objtool" ]; then
 	echo "$(basename "$0"): No objtool binary configured" 1>&2

tests/runfiles/common.run

@@ -378,7 +378,7 @@ tags = ['functional', 'cli_root', 'zfs_wait']
 
 [tests/functional/cli_root/zhack]
 tests = ['zhack_label_repair_001', 'zhack_label_repair_002',
-    'zhack_label_repair_003', 'zhack_label_repair_004']
+    'zhack_label_repair_003', 'zhack_label_repair_004', 'zhack_metaslab_leak']
 pre =
 post =
 tags = ['functional', 'cli_root', 'zhack']

tests/runfiles/linux.run

@@ -161,11 +161,13 @@ tags = ['functional', 'mmap']
 tests = ['mmp_on_thread', 'mmp_on_uberblocks', 'mmp_on_off', 'mmp_interval',
     'mmp_active_import', 'mmp_inactive_import', 'mmp_exported_import',
     'mmp_write_uberblocks', 'mmp_reset_interval', 'multihost_history',
-    'mmp_on_zdb', 'mmp_write_distribution', 'mmp_hostid', 'mmp_write_slow_disk']
+    'mmp_on_zdb', 'mmp_write_distribution', 'mmp_hostid', 'mmp_write_slow_disk',
+    'mmp_concurrent_import']
 tags = ['functional', 'mmp']
+timeout = 1200
 
 [tests/functional/mount:Linux]
-tests = ['umount_unlinked_drain']
+tests = ['umount_unlinked_drain', 'mount_loopback']
 tags = ['functional', 'mount']
 
 [tests/functional/pam:Linux]

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

@@ -247,7 +247,6 @@ maybe = {
     'l2arc/persist_l2arc_005_pos': ['FAIL', known_reason],
     'largest_pool/largest_pool_001_pos': ['FAIL', known_reason],
     'mmap/mmap_sync_001_pos': ['FAIL', known_reason],
-    'mmp/mmp_on_uberblocks': ['FAIL', known_reason],
     'pam/setup': ['SKIP', "pamtester might be not available"],
     'pool_checkpoint/checkpoint_discard_busy': ['FAIL', 11946],
     'projectquota/setup': ['SKIP', exec_reason],
@@ -366,9 +365,6 @@ elif sys.platform.startswith('linux'):
         'io/io_uring': ['SKIP', 'io_uring support required'],
         'limits/filesystem_limit': ['SKIP', known_reason],
         'limits/snapshot_limit': ['SKIP', known_reason],
-        'mmp/mmp_active_import': ['FAIL', known_reason],
-        'mmp/mmp_exported_import': ['FAIL', known_reason],
-        'mmp/mmp_inactive_import': ['FAIL', known_reason],
         'stat/statx_dioalign': ['SKIP', 'statx_reason'],
     })
 

tests/zfs-tests/callbacks/zfs_dbgmsg.ksh

@@ -26,7 +26,7 @@ echo "================================================================="
 sudo tail -n $lines /proc/spl/kstat/zfs/dbgmsg
 
 # reset dbgmsg
-sudo bash -c "echo > /proc/spl/kstat/zfs/dbgmsg"
+sudo sh -c "echo > /proc/spl/kstat/zfs/dbgmsg"
 
 echo "================================================================="
 echo " End of zfs_dbgmsg log"

tests/zfs-tests/callbacks/zfs_mmp.ksh

@@ -31,7 +31,7 @@ for f in /proc/spl/kstat/zfs/*/multihost; do
 	echo "================================================================="
 
 	sudo tail -n $lines $f
-	sudo bash -c "echo > $f"
+	sudo sh -c "echo > $f"
 done
 
 echo "================================================================="

tests/zfs-tests/include/commands.cfg

@@ -100,6 +100,7 @@ export SYSTEM_FILES_COMMON='awk
     uniq
     vmstat
     wc
+    which
     xargs
     xxh128sum'
 
@@ -146,6 +147,7 @@ export SYSTEM_FILES_LINUX='attr
     lscpu
     lsmod
     lsscsi
+    mkfs.xfs
     mkswap
     modprobe
     mountpoint

tests/zfs-tests/include/libtest.shlib

@@ -559,7 +559,7 @@ function default_cleanup_noexit
 		# Here, we loop through the pools we're allowed to
 		# destroy, only destroying them if it's safe to do
 		# so.
-		while [ ! -z ${ALL_POOLS} ]
+		while [ -n "${ALL_POOLS}" ]
 		do
 			for pool in ${ALL_POOLS}
 			do
@@ -3803,8 +3803,6 @@ function directory_diff # dir_a dir_b
 	# do not match there is a "c" entry in one of the columns).
 	if rsync --version | grep -q "[, ] crtimes"; then
 		args+=("--crtimes")
-	else
-		log_note "This rsync package does not support --crtimes (-N)."
 	fi
 
 	# If we are testing a ZIL replay, we need to ignore timestamp changes.

tests/zfs-tests/tests/Makefile.am

@@ -1004,6 +1004,7 @@ nobase_dist_datadir_zfs_tests_tests_SCRIPTS += \
 	functional/cli_root/zhack/zhack_label_repair_002.ksh \
 	functional/cli_root/zhack/zhack_label_repair_003.ksh \
 	functional/cli_root/zhack/zhack_label_repair_004.ksh \
+	functional/cli_root/zhack/zhack_metaslab_leak.ksh \
 	functional/cli_root/zpool_add/add_nested_replacing_spare.ksh \
 	functional/cli_root/zpool_add/add-o_ashift.ksh \
 	functional/cli_root/zpool_add/add_prop_ashift.ksh \
@@ -1666,6 +1667,7 @@ nobase_dist_datadir_zfs_tests_tests_SCRIPTS += \
 	functional/mmap/setup.ksh \
 	functional/mmp/cleanup.ksh \
 	functional/mmp/mmp_active_import.ksh \
+	functional/mmp/mmp_concurrent_import.ksh \
 	functional/mmp/mmp_exported_import.ksh \
 	functional/mmp/mmp_hostid.ksh \
 	functional/mmp/mmp_inactive_import.ksh \
@@ -1682,6 +1684,7 @@ nobase_dist_datadir_zfs_tests_tests_SCRIPTS += \
 	functional/mmp/setup.ksh \
 	functional/mount/cleanup.ksh \
 	functional/mount/setup.ksh \
+	functional/mount/mount_loopback.ksh \
 	functional/mount/umount_001.ksh \
 	functional/mount/umountall_001.ksh \
 	functional/mount/umount_unlinked_drain.ksh \

tests/zfs-tests/tests/functional/cli_root/zhack/zhack_metaslab_leak.ksh

@@ -0,0 +1,70 @@
+#!/bin/ksh
+# SPDX-License-Identifier: CDDL-1.0
+
+#
+# This file and its contents are supplied under the terms of the
+# Common Development and Distribution License ("CDDL"), version 1.0.
+# You may only use this file in accordance with the terms of version
+# 1.0 of the CDDL.
+#
+
+#
+# Description:
+#
+# Test whether zhack metaslab leak functions correctly
+#
+# Strategy:
+#
+# 1. Create pool on a loopback device with some test data
+# 2. Gather pool capacity stats
+# 3. Generate fragmentation data with zdb
+# 4. Destroy the pool
+# 5. Create a new pool with the same configuration
+# 6. Export the pool
+# 7. Apply the fragmentation information with zhack metaslab leak
+# 8. Import the pool
+# 9. Verify that pool capacity stats match
+
+. "$STF_SUITE"/include/libtest.shlib
+
+verify_runnable "global"
+
+function cleanup
+{
+	zpool destroy $TESTPOOL
+	rm $tmp
+}
+
+log_onexit cleanup
+log_assert "zhack metaslab leak leaks the right amount of space"
+
+typeset tmp=$(mktemp)
+
+log_must zpool create $TESTPOOL $DISKS
+for i in `seq 1 16`; do
+	log_must dd if=/dev/urandom of=/$TESTPOOL/f$i bs=1M count=16
+	log_must zpool sync $TESTPOOL
+done
+for i in `seq 2 2 16`; do
+	log_must rm /$TESTPOOL/f$i
+done
+for i in `seq 1 16`; do
+	log_must touch /$TESTPOOL/g$i
+	log_must zpool sync $TESTPOOL
+done
+
+alloc=$(zpool get -Hpo value alloc $TESTPOOL)
+log_must eval "zdb -m --allocated-map $TESTPOOL > $tmp"
+log_must zpool destroy $TESTPOOL
+
+log_must zpool create $TESTPOOL $DISKS
+log_must zpool export $TESTPOOL
+log_must eval "zhack metaslab leak $TESTPOOL < $tmp"
+log_must zpool import $TESTPOOL
+
+alloc2=$(zpool get -Hpo value alloc $TESTPOOL)
+
+[[ $((alloc * 1.05)) -gt $alloc2 ]] && [[ $alloc -lt $alloc2 ]] || \
+	log_fail "space usage changed too much: $alloc to $alloc2"
+
+log_pass "zhack metaslab leak behaved correctly"

tests/zfs-tests/tests/functional/mmp/mmp.cfg

@@ -25,6 +25,8 @@ export DISK=${DISKS%% *}
 export HOSTID_FILE="/etc/hostid"
 export HOSTID1=01234567
 export HOSTID2=89abcdef
+export HOSTID3=aaaabbbb
+export HOSTID4=ccccdddd
 
 export TXG_TIMEOUT_LONG=5000
 export TXG_TIMEOUT_DEFAULT=5
@@ -32,7 +34,7 @@ export TXG_TIMEOUT_DEFAULT=5
 export MMP_POOL=mmppool
 export MMP_DIR=$TEST_BASE_DIR/mmp
 export MMP_CACHE=$MMP_DIR/zpool.cache
-export MMP_ZTEST_LOG=$MMP_DIR/ztest.log
+export MMP_ZHACK_LOG=$MMP_DIR/zhack.log
 export MMP_HISTORY=100
 export MMP_HISTORY_OFF=0
 
@@ -43,5 +45,3 @@ export MMP_INTERVAL_MIN=100
 export MMP_IMPORT_INTERVALS=20
 export MMP_FAIL_INTERVALS_DEFAULT=10
 export MMP_FAIL_INTERVALS_MIN=2
-
-export MMP_TEST_DURATION_DEFAULT=$((MMP_IMPORT_INTERVALS*MMP_INTERVAL_DEFAULT/1000))

tests/zfs-tests/tests/functional/mmp/mmp.kshlib

@@ -99,11 +99,11 @@ function mmp_pool_create_simple # pool dir
 	log_must zpool set multihost=on $pool
 }
 
-function mmp_pool_create # pool dir
+function mmp_pool_create_zhack # pool dir
 {
 	typeset pool=${1:-$MMP_POOL}
 	typeset dir=${2:-$MMP_DIR}
-	typeset opts="-VVVVV -T120 -M -k0 -f $dir -E -p $pool"
+	typeset opts="-d $dir action idle -t120 $pool"
 
 	mmp_pool_create_simple $pool $dir
 
@@ -112,11 +112,11 @@ function mmp_pool_create # pool dir
 	log_must mmp_clear_hostid
 	log_must mmp_set_hostid $HOSTID2
 
-	log_note "Starting ztest in the background as hostid $HOSTID1"
+	log_note "Starting zhack in the background as hostid $HOSTID1"
-	log_must eval "ZFS_HOSTID=$HOSTID1 ztest $opts >$MMP_ZTEST_LOG 2>&1 &"
+	log_must eval "ZFS_HOSTID=$HOSTID1 zhack $opts >$MMP_ZHACK_LOG 2>&1 &"
 
 	while ! is_pool_imported "$pool" "-d $dir"; do
-		log_must pgrep ztest
+		log_must pgrep zhack
 		log_must sleep 5
 	done
 }
@@ -126,10 +126,10 @@ function mmp_pool_destroy # pool dir
 	typeset pool=${1:-$MMP_POOL}
 	typeset dir=${2:-$MMP_DIR}
 
-	ZTESTPID=$(pgrep ztest)
+	ZHACKPID=$(pgrep zhack)
-	if [ -n "$ZTESTPID" ]; then
+	if [ -n "$ZHACKPID" ]; then
-		log_must kill $ZTESTPID
+		log_must kill $ZHACKPID
-		wait $ZTESTPID
+		wait $ZHACKPID
 	fi
 
 	if poolexists $pool; then
@@ -158,33 +158,34 @@ function import_no_activity_check # pool opts
 	typeset pool=$1
 	typeset opts=$2
 
-	typeset max_duration=$((MMP_TEST_DURATION_DEFAULT-1))
+	RESULT=$(ZFS_LOAD_INFO_DEBUG=1 zpool import $opts $pool)
-
-	SECONDS=0
-	zpool import $opts $pool
 	typeset rc=$?
 
-	if [[ $SECONDS -gt $max_duration ]]; then
+	# mmp_result: 3 (ESRCH) no activity check was run not required
-		log_fail "ERROR: import_no_activity_check unexpected activity \
+	# mmp_result: 6 (ENXIO) no activity check was run hostid not set
-check (${SECONDS}s gt $max_duration)"
+	if ! echo "$RESULT" | grep -q "mmp_result: 3" &&
+	   ! echo "$RESULT" | grep -q "mmp_result: 6"; then
+		log_note "ERROR: $RESULT"
+		log_fail "ERROR: import_no_activity_check unexpected activity check"
 	fi
 
 	return $rc
 }
 
-function import_activity_check # pool opts act_test_duration
+function import_activity_check # pool opts
 {
 	typeset pool=$1
 	typeset opts=$2
-	typeset min_duration=${3:-$MMP_TEST_DURATION_DEFAULT}
 
-	SECONDS=0
+	RESULT=$(ZFS_LOAD_INFO_DEBUG=1 zpool import $opts $pool)
-	zpool import $opts $pool
 	typeset rc=$?
 
-	if [[ $SECONDS -le $min_duration ]]; then
+	# mmp_result: 0 (Success) check was run no activity detected
-		log_fail "ERROR: import_activity_check expected activity check \
+	# mmp_result: 121 (EREMOTEIO) check was run activity detected
-(${SECONDS}s le min_duration $min_duration)"
+	# mmp_result: 4 (EINTR) check was run but interrupted by user
+	if ! echo "$RESULT" | grep -q "mmp_result: 0"; then
+		log_note "ERROR: $RESULT"
+		log_fail "ERROR: import_activity_check expected activity check"
 	fi
 
 	return $rc

tests/zfs-tests/tests/functional/mmp/mmp_active_import.ksh

@@ -24,10 +24,10 @@
 #	with one hostid be importable by a host with a different hostid.
 #
 # STRATEGY:
-#	 1. Simulate an active pool on another host with ztest.
+#	 1. Simulate an active pool on another host with zhack.
 #	 2. Verify 'zpool import' reports an active pool.
 #	 3. Verify 'zpool import [-f] $MMP_POOL' cannot import the pool.
-#	 4. Kill ztest to make pool eligible for import.
+#	 4. Kill zhack to make pool eligible for import.
 #	 5. Verify 'zpool import' fails with the expected error message.
 #	 6. Verify 'zpool import $MMP_POOL' fails with the expected message.
 #	 7. Verify 'zpool import -f $MMP_POOL' can now import the pool.
@@ -44,16 +44,16 @@ function cleanup
 {
 	mmp_pool_destroy $MMP_POOL $MMP_DIR
 	log_must mmp_clear_hostid
-	ZTESTPID=$(pgrep ztest)
+	ZHACKPID=$(pgrep zhack)
-	if [ -n "$ZTESTPID" ]; then
+	if [ -n "$ZHACKPID" ]; then
-		for pid in $ZTESTPID; do
+		for pid in $ZHACKPID; do
 			log_must kill -9 $pid
 		done
 	else
-		# if ztest not running and log present, ztest crashed
+		# if zhack is not running and log present, zhack crashed
-		if [ -f $MMP_ZTEST_LOG ]; then
+		if [ -f $MMP_ZHACK_LOG ]; then
-			log_note "ztest appears to have crashed.  Tail of log:"
+			log_note "zhack appears to have crashed.  Tail of log:"
-			tail -n 50 $MMP_ZTEST_LOG
+			tail -n 50 $MMP_ZHACK_LOG
 		fi
 	fi
 }
@@ -61,15 +61,18 @@ function cleanup
 log_assert "multihost=on|off active pool activity checks"
 log_onexit cleanup
 
-# 1. Simulate an active pool on another host with ztest.
+# 1. Simulate an active pool on another host with zhack.
+log_note "Simulate an active pool on another host with zhack"
 mmp_pool_destroy $MMP_POOL $MMP_DIR
-mmp_pool_create $MMP_POOL $MMP_DIR
+mmp_pool_create_zhack $MMP_POOL $MMP_DIR
 
 # 2. Verify 'zpool import' reports an active pool.
+log_note "Verify 'zpool import' reports an active pool"
 log_must mmp_set_hostid $HOSTID2
 log_must is_pool_imported $MMP_POOL "-d $MMP_DIR"
 
 # 3. Verify 'zpool import [-f] $MMP_POOL' cannot import the pool.
+log_note "Verify 'zpool import [-f] $MMP_POOL' cannot import the pool"
 MMP_IMPORTED_MSG="Cannot import '$MMP_POOL': pool is imported"
 
 log_must try_pool_import $MMP_POOL "-d $MMP_DIR" "$MMP_IMPORTED_MSG"
@@ -84,14 +87,15 @@ for i in {1..10}; do
 	    "$MMP_IMPORTED_MSG"
 done
 
-# 4. Kill ztest to make pool eligible for import.  Poll with 'zpool status'.
+# 4. Kill zhack to make pool eligible for import.  Poll with 'zpool status'.
-ZTESTPID=$(pgrep ztest)
+log_note "Kill zhack to make pool eligible for import.  Poll with 'zpool status'"
-if [ -n "$ZTESTPID" ]; then
+ZHACKPID=$(pgrep zhack)
-	log_must kill -9 $ZTESTPID
+if [ -n "$ZHACKPID" ]; then
+	log_must kill -9 $ZHACKPID
 fi
 log_must wait_pool_imported $MMP_POOL "-d $MMP_DIR"
-if [ -f $MMP_ZTEST_LOG ]; then
+if [ -f $MMP_ZHACK_LOG ]; then
-	log_must rm $MMP_ZTEST_LOG
+	log_must rm $MMP_ZHACK_LOG
 fi
 
 # 5. Verify 'zpool import' fails with the expected error message, when
@@ -99,6 +103,7 @@ fi
 #    - hostid=matches   - safe to import the pool
 #    - hostid=different - previously imported on a different system
 #
+log_note "Verify 'zpool import' fails with the expected error message"
 log_must mmp_clear_hostid
 MMP_IMPORTED_MSG="Set a unique system hostid"
 log_must check_pool_import $MMP_POOL "-d $MMP_DIR" "action" "$MMP_IMPORTED_MSG"
@@ -113,13 +118,16 @@ MMP_IMPORTED_MSG="The pool was last accessed by another system."
 log_must check_pool_import $MMP_POOL "-d $MMP_DIR" "status" "$MMP_IMPORTED_MSG"
 
 # 6. Verify 'zpool import $MMP_POOL' fails with the expected message.
+log_note "Verify 'zpool import $MMP_POOL' fails with the expected message"
 MMP_IMPORTED_MSG="pool was previously in use from another system."
 log_must try_pool_import $MMP_POOL "-d $MMP_DIR" "$MMP_IMPORTED_MSG"
 
 # 7. Verify 'zpool import -f $MMP_POOL' can now import the pool.
+log_note "Verify 'zpool import -f $MMP_POOL' can now import the pool"
 log_must import_activity_check $MMP_POOL "-f -d $MMP_DIR"
 
 # 8 Verify pool may be exported/imported without -f argument.
+log_note "Verify pool may be exported/imported without -f argument"
 log_must zpool export $MMP_POOL
 log_must import_no_activity_check $MMP_POOL "-d $MMP_DIR"
 

tests/zfs-tests/tests/functional/mmp/mmp_concurrent_import.ksh

@@ -0,0 +1,133 @@
+#!/bin/ksh -p
+# SPDX-License-Identifier: CDDL-1.0
+#
+# CDDL HEADER START
+#
+# This file and its contents are supplied under the terms of the
+# Common Development and Distribution License ("CDDL"), version 1.0.
+# You may only use this file in accordance with the terms of version
+# 1.0 of the CDDL.
+#
+# A full copy of the text of the CDDL should have accompanied this
+# source.  A copy of the CDDL is also available via the Internet at
+# http://www.illumos.org/license/CDDL.
+#
+# CDDL HEADER END
+#
+
+#
+# Copyright (c) 2026 by Lawrence Livermore National Security, LLC.
+#
+
+# DESCRIPTION:
+#	Verify that even when importing a shared pool simultaneously
+#	on systems with different host ids at most one will succeed.
+#
+# STRATEGY:
+# 1. Create an multihost enabled pool
+# 2. zhack imports: $HOSTID1 (matching) and $HOSTID1 (matching)
+# 3. zhack imports: $HOSTID1 (matching) and $HOSTID2 (different)
+# 4. zhack imports: $HOSTID3 (different) and $HOSTID4 (different)
+#
+
+. $STF_SUITE/include/libtest.shlib
+. $STF_SUITE/tests/functional/mmp/mmp.cfg
+. $STF_SUITE/tests/functional/mmp/mmp.kshlib
+
+verify_runnable "both"
+
+function cleanup
+{
+	ZHACKPIDS=$(pgrep zhack)
+	if [ -n "$ZHACKPIDS" ]; then
+		for pid in $ZHACKPIDS; do
+			log_must kill -9 $pid
+		done
+	fi
+
+	log_must rm -f $MMP_ZHACK_LOG.1 $MMP_ZHACK_LOG.2
+
+	mmp_pool_destroy $MMP_POOL $MMP_DIR
+	log_must mmp_clear_hostid
+}
+
+# Verify that pool was imported by at most one of the zhack processes.
+# Check both the return code and expected import message.
+function verify_zhack
+{
+	IMPORT_COUNT=0
+	IMPORT_MSGS=0
+
+	ZHACKPIDS=$(pgrep zhack)
+	for pid in $ZHACKPIDS; do
+		wait $pid
+		STATUS=$?
+		if [[ $STATUS -eq 0 ]]; then
+			(( IMPORT_COUNT++ ))
+		fi
+		log_note "PID $pid exited with status $STATUS"
+	done
+
+	grep -H "Imported pool $MMP_POOL" $MMP_ZHACK_LOG.1 && (( IMPORT_MSGS++ ))
+	grep -H "Imported pool $MMP_POOL" $MMP_ZHACK_LOG.2 && (( IMPORT_MSGS++ ))
+
+	if [[ $IMPORT_MSGS -gt 1 ]]; then
+		cat $MMP_ZHACK_LOG.*
+		log_fail "Multiple import success messages"
+	fi
+
+	if [[ $IMPORT_COUNT -gt 1 ]]; then
+		cat $MMP_ZHACK_LOG.*
+		log_fail "Multiple import success return codes"
+	fi
+
+	if [[ $IMPORT_MSGS -ne $IMPORT_COUNT ]]; then
+		cat $MMP_ZHACK_LOG.*
+		log_fail "Messages ($IMPORT_MSGS) differs from count ($IMPORT_COUNT)"
+	fi
+}
+
+OPTS="-d $MMP_DIR action idle -t5 $MMP_POOL"
+
+log_assert "multihost=on concurrent imports"
+log_onexit cleanup
+
+# 1. Create a multihost enabled pool with HOSTID1
+mmp_pool_create_simple $MMP_POOL $MMP_DIR
+log_must zpool export -F $MMP_POOL
+
+# 2. zhack imports: $HOSTID1 (matching) and $HOSTID1 (matching)
+# Activity check required because the pool was exported with -F above, the
+# claim phase will detect the double import despite matching hostids.
+log_note "zhack import with $HOSTID1 (matching) and $HOSTID1 (matching)"
+log_must eval "ZFS_HOSTID=$HOSTID1 zhack $OPTS >$MMP_ZHACK_LOG.1 2>&1 &"
+log_must eval "ZFS_HOSTID=$HOSTID1 zhack $OPTS >$MMP_ZHACK_LOG.2 2>&1 &"
+log_must verify_zhack
+
+mmp_clear_hostid
+mmp_set_hostid $HOSTID1
+log_must import_activity_check $MMP_POOL "-d $MMP_DIR"
+log_must zpool export $MMP_POOL
+
+# 3. zhack imports: $HOSTID1 (matching) and $HOSTID2 (different)
+# Activity check skipped for HOSTID1 it is expected to import successfully.
+# zhack with HOSTID2 will run the activity check and detect the active pool.
+log_note "zhack import with $HOSTID1 (matching) and $HOSTID2 (different)"
+log_must eval "ZFS_HOSTID=$HOSTID1 zhack $OPTS >$MMP_ZHACK_LOG.1 2>&1 &"
+log_must eval "ZFS_HOSTID=$HOSTID2 zhack $OPTS >$MMP_ZHACK_LOG.2 2>&1 &"
+log_must verify_zhack
+
+mmp_clear_hostid
+mmp_set_hostid $HOSTID3
+log_must import_activity_check $MMP_POOL "-d $MMP_DIR"
+log_must zpool export $MMP_POOL
+
+# 4. zhack imports: $HOSTID1 (different) and $HOSTID2 (different)
+# Both zhacks will run the activity checks, depending on the exact timing
+# one may succeed and the other fail, or both may fail.
+log_note "zhack import with $HOSTID1 (different) and $HOSTID2 (different)"
+log_must eval "ZFS_HOSTID=$HOSTID1 zhack $OPTS >$MMP_ZHACK_LOG.1 2>&1 &"
+log_must eval "ZFS_HOSTID=$HOSTID2 zhack $OPTS >$MMP_ZHACK_LOG.2 2>&1 &"
+log_must verify_zhack
+
+log_pass "multihost=on concurrent imports"

tests/zfs-tests/tests/functional/mmp/mmp_exported_import.ksh

@@ -28,7 +28,7 @@
 #	3. Verify multihost=off and hostids differ (no activity check)
 #	4. Verify multihost=off and hostid zero allowed (no activity check)
 #	5. Verify multihost=on and hostids match (no activity check)
-#	6. Verify multihost=on and hostids differ (no activity check)
+#	6. Verify multihost=on and hostids differ (activity check)
 #	7. Verify multihost=on and hostid zero fails (no activity check)
 #
 
@@ -40,7 +40,7 @@ verify_runnable "both"
 
 function cleanup
 {
-	default_cleanup_noexit
+	datasetexists $TESTPOOL && destroy_pool $TESTPOOL
 	log_must mmp_clear_hostid
 }
 
@@ -49,9 +49,10 @@ log_onexit cleanup
 
 # 1. Create a zpool
 log_must mmp_set_hostid $HOSTID1
-default_setup_noexit $DISK
+log_must zpool create -f $TESTPOOL $DISK
 
 # 2. Verify multihost=off and hostids match (no activity check)
+log_note "Verify multihost=off and hostids match (no activity check)"
 log_must zpool set multihost=off $TESTPOOL
 
 for opt in "" "-f"; do
@@ -60,6 +61,7 @@ for opt in "" "-f"; do
 done
 
 # 3. Verify multihost=off and hostids differ (no activity check)
+log_note "Verify multihost=off and hostids differ (no activity check)"
 for opt in "" "-f"; do
 	log_must mmp_pool_set_hostid $TESTPOOL $HOSTID1
 	log_must zpool export $TESTPOOL
@@ -69,6 +71,7 @@ for opt in "" "-f"; do
 done
 
 # 4. Verify multihost=off and hostid zero allowed (no activity check)
+log_note "Verify multihost=off and hostid zero allowed (no activity check)"
 log_must mmp_clear_hostid
 
 for opt in "" "-f"; do
@@ -77,6 +80,7 @@ for opt in "" "-f"; do
 done
 
 # 5. Verify multihost=on and hostids match (no activity check)
+log_note "Verify multihost=on and hostids match (no activity check)"
 log_must mmp_pool_set_hostid $TESTPOOL $HOSTID1
 log_must zpool set multihost=on $TESTPOOL
 
@@ -85,16 +89,18 @@ for opt in "" "-f"; do
 	log_must import_no_activity_check $TESTPOOL $opt
 done
 
-# 6. Verify multihost=on and hostids differ (no activity check)
+# 6. Verify multihost=on and hostids differ (activity check)
+log_note "Verify multihost=on and hostids differ (activity check)"
 for opt in "" "-f"; do
 	log_must mmp_pool_set_hostid $TESTPOOL $HOSTID1
 	log_must zpool export $TESTPOOL
 	log_must mmp_clear_hostid
 	log_must mmp_set_hostid $HOSTID2
-	log_must import_no_activity_check $TESTPOOL $opt
+	log_must import_activity_check $TESTPOOL $opt
 done
 
 # 7. Verify multihost=on and hostid zero fails (no activity check)
+log_note "Verify multihost=on and hostid zero fails (no activity check)"
 log_must zpool export $TESTPOOL
 log_must mmp_clear_hostid
 

tests/zfs-tests/tests/functional/mmp/mmp_hostid.ksh

@@ -40,7 +40,7 @@ verify_runnable "both"
 
 function cleanup
 {
-	default_cleanup_noexit
+	datasetexists $MMP_POOL && destroy_pool $MMP_POOL
 	log_must rm $MMP_DIR/file.{0,1,2,3,4,5}
 	log_must rmdir $MMP_DIR
 	log_must mmp_clear_hostid
@@ -56,7 +56,7 @@ log_must mkdir -p $MMP_DIR
 log_must truncate -s $MINVDEVSIZE $MMP_DIR/file.{0,1,2,3,4,5}
 
 # 1. Create a non-redundant pool
-log_must zpool create $MMP_POOL $MMP_DIR/file.0
+log_must zpool create -f $MMP_POOL $MMP_DIR/file.0
 
 # 2. Create an 'etc' dataset containing a valid hostid file; caching is
 #    disabled on the dataset to force the hostid to be read from disk.
@@ -71,16 +71,19 @@ mntpnt_fs=$(get_prop mountpoint $MMP_POOL/fs)
 log_must mkfile 1M $mntpnt_fs/file
 
 # 4. Verify multihost cannot be enabled until the /etc/hostid is linked
+log_note "Verify multihost cannot be enabled until the /etc/hostid is linked"
 log_mustnot zpool set multihost=on $MMP_POOL
 log_mustnot ls -l $HOSTID_FILE
 log_must ln -s $mntpnt_etc/hostid $HOSTID_FILE
 log_must zpool set multihost=on $MMP_POOL
 
 # 5. Verify vdevs may be attached and detached
+log_note "Verify vdevs may be attached and detached"
 log_must zpool attach $MMP_POOL $MMP_DIR/file.0 $MMP_DIR/file.1
 log_must zpool detach $MMP_POOL $MMP_DIR/file.1
 
 # 6. Verify normal, cache, log and special vdevs can be added
+log_note "Verify normal, cache, log and special vdevs can be added"
 log_must zpool add $MMP_POOL $MMP_DIR/file.1
 log_must zpool add $MMP_POOL $MMP_DIR/file.2
 log_must zpool add $MMP_POOL cache $MMP_DIR/file.3
@@ -88,6 +91,7 @@ log_must zpool add $MMP_POOL log $MMP_DIR/file.4
 log_must zpool add $MMP_POOL special $MMP_DIR/file.5
 
 # 7. Verify normal, cache, and log vdevs can be removed
+log_note "Verify normal, cache, and log vdevs can be removed"
 log_must zpool remove $MMP_POOL $MMP_DIR/file.2
 log_must zpool remove $MMP_POOL $MMP_DIR/file.3
 log_must zpool remove $MMP_POOL $MMP_DIR/file.4

tests/zfs-tests/tests/functional/mmp/mmp_inactive_import.ksh

@@ -27,7 +27,7 @@
 #	2. Verify multihost=off and hostids match (no activity check)
 #	3. Verify multihost=off and hostids differ (no activity check)
 #	4. Verify multihost=off and hostid allowed (no activity check)
-#	5. Verify multihost=on and hostids match (no activity check)
+#	5. Verify multihost=on and hostids match (activity check)
 #	6. Verify multihost=on and hostids differ (activity check)
 #	7. Verify mmp_write and mmp_fail are set correctly
 #	8. Verify multihost=on and hostid zero fails (no activity check)
@@ -42,7 +42,7 @@ verify_runnable "both"
 
 function cleanup
 {
-	default_cleanup_noexit
+	datasetexists $TESTPOOL && destroy_pool $TESTPOOL
 	log_must mmp_clear_hostid
 	log_must set_tunable64 MULTIHOST_INTERVAL $MMP_INTERVAL_DEFAULT
 }
@@ -52,9 +52,10 @@ log_onexit cleanup
 
 # 1. Create a zpool
 log_must mmp_set_hostid $HOSTID1
-default_setup_noexit $DISK
+log_must zpool create -f $TESTPOOL $DISK
 
 # 2. Verify multihost=off and hostids match (no activity check)
+log_note "Verify multihost=off and hostids match (no activity check)"
 log_must zpool set multihost=off $TESTPOOL
 
 for opt in "" "-f"; do
@@ -63,6 +64,7 @@ for opt in "" "-f"; do
 done
 
 # 3. Verify multihost=off and hostids differ (no activity check)
+log_note "Verify multihost=off and hostids differ (no activity check)"
 log_must zpool export -F $TESTPOOL
 log_must mmp_clear_hostid
 log_must mmp_set_hostid $HOSTID2
@@ -70,21 +72,24 @@ log_mustnot import_no_activity_check $TESTPOOL ""
 log_must import_no_activity_check $TESTPOOL "-f"
 
 # 4. Verify multihost=off and hostid zero allowed (no activity check)
+log_note "Verify multihost=off and hostid zero allowed (no activity check)"
 log_must zpool export -F $TESTPOOL
 log_must mmp_clear_hostid
 log_mustnot import_no_activity_check $TESTPOOL ""
 log_must import_no_activity_check $TESTPOOL "-f"
 
-# 5. Verify multihost=on and hostids match (no activity check)
+# 5. Verify multihost=on and hostids match (activity check)
+log_note "Verify multihost=on and hostids match (activity check)"
 log_must mmp_pool_set_hostid $TESTPOOL $HOSTID1
 log_must zpool set multihost=on $TESTPOOL
 
 for opt in "" "-f"; do
 	log_must zpool export -F $TESTPOOL
-	log_must import_no_activity_check $TESTPOOL $opt
+	log_must import_activity_check $TESTPOOL $opt
 done
 
 # 6. Verify multihost=on and hostids differ (activity check)
+log_note "Verify multihost=on and hostids differ (activity check)"
 log_must zpool export -F $TESTPOOL
 log_must mmp_clear_hostid
 log_must mmp_set_hostid $HOSTID2
@@ -92,10 +97,12 @@ log_mustnot import_activity_check $TESTPOOL ""
 log_must import_activity_check $TESTPOOL "-f"
 
 # 7. Verify mmp_write and mmp_fail are set correctly
+log_note "Verify mmp_write and mmp_fail are set correctly"
 log_must zpool export -F $TESTPOOL
 log_must verify_mmp_write_fail_present ${DISK[0]}
 
 # 8. Verify multihost=on and hostid zero fails (no activity check)
+log_note "Verify multihost=on and hostid zero fails (no activity check)"
 log_must mmp_clear_hostid
 MMP_IMPORTED_MSG="Set a unique system hostid"
 log_must check_pool_import $TESTPOOL "-f" "action" "$MMP_IMPORTED_MSG"
@@ -104,9 +111,10 @@ log_mustnot import_no_activity_check $TESTPOOL "-f"
 # 9. Verify activity check duration based on mmp_write and mmp_fail
 # Specify a short test via tunables but import pool imported while
 # tunables set to default duration.
+log_note "Verify activity check duration based on mmp_write and mmp_fail"
 log_must set_tunable64 MULTIHOST_INTERVAL $MMP_INTERVAL_MIN
 log_must mmp_clear_hostid
 log_must mmp_set_hostid $HOSTID1
-log_must import_activity_check $TESTPOOL "-f" $MMP_TEST_DURATION_DEFAULT
+log_must import_activity_check $TESTPOOL "-f"
 
 log_pass "multihost=on|off inactive pool activity checks passed"