2025-02-06 Setup

The purpose of this test is to see if FreeBSD (which uses the same ZFS code base as Linux) is also subject to this corruption.

Plan

This work is being recorded in a separate git branch FreeBSD because it is not know if it is going to produce useful results.

Install FreeBSD on a Pi 4B and repeat tests using the scripts employed on Linux. The scripts required some tweaking and an additional script was added manage_snaps.sh to loop the snapshot script.

  • Pi 4B with 4GB RAM
  • Powered hub
  • 120GB Kingston SSD for system disk.
  • 1TB Samsung 850 EVO for pools.
  • reeBSD 15.0-CURRENT (hostname vulcan)

In order to boot it was necessary to apply a fix described in https://forums.freebsd.org/threads/bug-freebsd-14-2-raspberry-pi-4-with-4gb-wont-boot.96643. The pools that existed on the 1TB SSD could not easily be mounted so the drive was repartitioned and new pools created.

root@vulcan:~ # gpart show da1
=>        40  1953525088  da1  GPT  (932G)
          40   977272832    1  freebsd-zfs  (466G)
   977272872   976252256    2  freebsd-zfs  (466G)

root@vulcan:~ #

ZFS version is:

root@vulcan:~ # zfs --version
zfs-2.3.99-170-FreeBSD_g34205715e
zfs_version_kernel() failed: No such file or directory
root@vulcan:~ # 

dd if=/dev/urandom bs=32 count=1 of=/pool-key
dev=da1p1
zpool create -o ashift=12 \
      -O acltype=posixacl -O canmount=on -O compression=lz4 \
      -O dnodesize=auto -O normalization=formD -O relatime=on -O xattr=sa \
      -O encryption=aes-256-gcm  -O keylocation=file:///pool-key -O keyformat=raw \
      -O mountpoint=/mnt/send \
      send $dev
zfs load-key -a
chmod a+rwx /mnt/send/

dev=da1p2
zpool create -o ashift=12 \
      -O acltype=posixacl -O canmount=on -O compression=lz4 \
      -O dnodesize=auto -O normalization=formD -O relatime=on -O xattr=sa \
      -O mountpoint=/mnt/recv \
      recv $dev
chmod a+rwx /mnt/recv/

user=hbarta
zfs allow -u $user \
    compression,create,destroy,hold,mount,mountpoint,receive,send,snapshot,destroy,rollback \
    send
zfs allow -u $user \
    compression,create,destroy,hold,mount,mountpoint,receive,send,snapshot,destroy,rollback \
    recv

time -p syncoid --recursive --no-privilege-elevation send/test recv/test

First syncoid produced an error. Second syncoid (run as user) completed the mirroring of send to recv and the third completed w/out any reported issues.

2025-02-10 tweaking and testing proceeded.

Existing scripts were tweaked to run on FreeBSD (still using bash.) After the run was complete (and corruption was detected) a script was added manage_snaps.sh to terminate the snapshot processing after corruption was detected.

See Results for more detail on the results.