ZTS: add regression test for #17180

[delan]

Motivation and Context

#17180 fixed an interesting bug that i believe i hit in one of my pools, but as far as i can tell, there was no test for it.

Description

this patch adds a regression test for #17180, minimised from my attempts to reproduce the bug in a way that resembled the history of my pool.

How Has This Been Tested?

tested using quiz, on linux 6.12.44 and zfs 2.3.1, with and without #17180 applied. without the patch, it kernel panics every time. with the patch, it passes every time. similarly on the latest zfs, it passes every time, but if the reverse patch is applied, it kernel panics every time.

Types of changes

• Bug fix (non-breaking change which fixes an issue)
• New feature (non-breaking change which adds functionality)
• Performance enhancement (non-breaking change which improves efficiency)
• Code cleanup (non-breaking change which makes code smaller or more readable)
• Quality assurance (non-breaking change which makes the code more robust against bugs)
• Breaking change (fix or feature that would cause existing functionality to change)
• Library ABI change (libzfs, libzfs_core, libnvpair, libuutil and libzfsbootenv)
• Documentation (a change to man pages or other documentation)

Checklist:

• My code follows the OpenZFS code style requirements.
• I have updated the documentation accordingly.
• I have read the contributing document.
• I have added tests to cover my changes.
• I have run the ZFS Test Suite with this change applied.
• All commit messages are properly formatted and contain Signed-off-by.

[delan]

original repro, as quiz user program

#!/usr/bin/env bash
# QUIZ_KERNEL_VERSION=6.12.44
# QUIZ_MEM=24G
# QUIZ_BLOCKDEV_COUNT=2
# QUIZ_BLOCKDEV_SIZE=5G
set -euvx
free -h
zpool create -o ashift=16 foo /dev/quizb0 /dev/quizb1
zfs create foo/home
dd if=/dev/urandom of=/foo/home/DSC0001.JPG bs=1M count=32
dd if=/dev/urandom of=/foo/home/DSC0002.JPG bs=1M count=32
snap0=$(date -u +\%TZ); zfs snapshot -r foo@$snap0; sleep 1
zfs create foo/photos
cp -av /foo/home/* /foo/photos/
snap1=$(date -u +\%TZ); zfs snapshot -r foo@$snap1; sleep 1
zpool remove -w foo /dev/quizb0
zfs send -Rv foo@$snap1 > /dev/null
dd if=/dev/urandom of=/foo/photos/DSC0002.JPG bs=1M count=33
snap2=$(date -u +\%TZ); zfs snapshot -r foo@$snap2; sleep 1
zfs send -Rv foo@$snap2 > /dev/null
rm -v /foo/home/*
snap3=$(date -u +\%TZ); zfs snapshot -r foo@$snap3; sleep 1
zfs destroy -vn foo/home@$snap0\%$snap2
zfs destroy -v foo/home@$snap0\%$snap2
zfs send -Rv foo@$snap2 > /dev/null || :
zfs send -Rv foo/photos@$snap2 > /dev/null
for i in {3..9}; do dd if=/dev/urandom of=/foo/photos/DSC000$i.JPG bs=1M count=129; done
zpool add foo /dev/quizb0
for i in {0..9}; do dd if=/dev/urandom of=/foo/photos/DSC001$i.JPG bs=1M count=129; done
snap4=$(date -u +\%TZ); zfs snapshot -r foo@$snap4; sleep 1
for i in {1..9}; do dd if=/dev/urandom of=/foo/photos/DSC000$i.JPG.xmp bs=1k count=129; done
for i in {0..9}; do dd if=/dev/urandom of=/foo/photos/DSC001$i.JPG.xmp bs=1k count=129; done
snap5=$(date -u +\%TZ); zfs snapshot -r foo@$snap5; sleep 1
zfs create foo/photos2
cp -av /foo/photos/* /foo/photos2/
snap6=$(date -u +\%TZ); zfs snapshot -r foo@$snap6; sleep 1
zpool remove -w foo /dev/quizb1
zfs create foo/photos/2024
zfs create foo/photos/2025
strace cp -a /foo/photos/DSC000* /foo/photos/2024/
cp -a /foo/photos/DSC001* /foo/photos/2025/
for i in {0..9}; do dd if=/dev/urandom of=/foo/photos/2025/DSC001$i.JPG.xmp bs=1k count=65 conv=notrunc; done
# this one needs a couple of tries, because it initially fails with “out of space”
snap7=$(date -u +\%TZ); while ! zfs snapshot -r foo@$snap7; do sleep 1; done; sleep 1
rm /foo/photos/DSC*
snap8=$(date -u +\%TZ); zfs snapshot -r foo@$snap8; sleep 1
zfs destroy foo/home@\%
zfs destroy -vn foo/photos@$snap1\%$snap7
zfs destroy -v foo/photos@$snap1\%$snap7
zfs send -Rv foo/photos2@$snap8 > /dev/null
zfs send -Rv foo/photos@$snap8 > /dev/null
zfs send -Rv foo@$snap8 > /dev/null || :
zfs snapshot foo/home@$snap8
zfs send -Rv foo@$snap8 > /dev/null
zfs send -Rv foo@$snap8 > /dev/null
zpool add foo /dev/quizb1
zfs destroy -vn foo/photos2 || :
zfs destroy -rvn foo/photos2
zfs destroy -rv foo/photos2
zpool remove -w foo /dev/quizb0
zfs send -Rv foo@$snap8 > /dev/null
zfs create foo/photos2
cp -a /foo/photos/202?/* /foo/photos2
# this one needs a couple of tries, because it initially fails with “out of space”
while ! zfs create foo/photos3; do sleep 1; done
cp -a /foo/photos/202?/* /foo/photos3
zpool add foo /dev/quizb0
zfs create foo/photos4
cp -a /foo/photos/202?/* /foo/photos4
zfs create foo/photos5
cp -a /foo/photos/202?/* /foo/photos5
zfs create foo/photos6
cp -a /foo/photos/202?/* /foo/photos6
snap9=$(date -u +\%TZ); zfs snapshot -r foo@$snap9; sleep 1
for i in {0..9}; do dd if=/dev/urandom of=/foo/photos3/DSC001$i.JPG bs=1M count=33 conv=notrunc; done
for i in {0..9}; do dd if=/dev/urandom of=/foo/photos5/DSC001$i.JPG bs=1M count=33 conv=notrunc; done
snap10=$(date -u +\%TZ); zfs snapshot -r foo@$snap10; sleep 1
zpool remove -w foo /dev/quizb1
for i in {0..9}; do dd if=/dev/urandom of=/foo/photos2/DSC001$i.JPG bs=1M count=33 conv=notrunc; done
for i in {0..9}; do dd if=/dev/urandom of=/foo/photos4/DSC001$i.JPG bs=1M count=33 conv=notrunc; done
snap11=$(date -u +\%TZ); zfs snapshot -r foo@$snap11 || :; sleep 1
zfs destroy -r foo/photos6
zfs destroy -r foo/photos5
zfs destroy -r foo/photos4

minimised repro, as quiz user program

#!/usr/bin/env bash
# QUIZ_KERNEL_VERSION=6.12.44
# QUIZ_BLOCKDEV_COUNT=2
# QUIZ_BLOCKDEV_SIZE=200M
set -euvx
zpool create foo /dev/quizb0 /dev/quizb1
dd if=/dev/urandom of=/foo/file bs=16M count=2
zpool remove -w foo /dev/quizb1
zfs create foo/d
cp /foo/file /foo/d
dd if=/dev/urandom of=/foo/file bs=16M count=2
zfs destroy -r foo/d

[adamdmoss]

Looks good to me. Glad to have a regression test for this.

[behlendorf]

Thanks for adding a test case for this. Just a couple of comments.

[amotin]

If you say without the patch this test causes a panic, then I have no objections.