Comment 2 for bug 1934709

Revision history for this message
Matthew Ruffell (mruffell) wrote :

Performing verification for Bionic.

Firstly, I installed 4.15.0-151-generic from -updates, to get a baseline for all xfstests results.

I built and ran btrfs/* from xfstests, using the following local config:

export TEST_DEV=/dev/vdc
export TEST_DIR=/mnt/test
export SCRATCH_DEV_POOL="/dev/vdd /dev/vde /dev/vdf"
export SCRATCH_MNT=/mnt/scratch

The results are here:

https://paste.ubuntu.com/p/5TZTTHkvGs/

I then enabled -proposed and installed 4.15.0-152-generic, and executed the same testsuite.

The results are here:

https://paste.ubuntu.com/p/cpXBjP5bkF/

Doing a diff of the two, there are no differences, and they also match the output of 4.15.0-136-generic done earlier.

Now, I was a little concerned about btrfs/078 failing, so I decided to look into it some more.

btrfs/078 _check_btrfs_filesystem: filesystem on /dev/vdd is inconsistent
(see /home/ubuntu/xfstests/results//btrfs/078.full for details)

This test has failed with all Ubuntu kernels in the past, and is not new. If you run the testcase several times however, we see that sometimes it passes, and sometimes fails. If it fails, it almost always errors out at 399 seconds, which suggests a timeout? When it succeeds, it takes 430s.

4.15.0-151-generic:

https://paste.ubuntu.com/p/tPgyRHXtmb/

4.15.0-152-generic:

https://paste.ubuntu.com/p/P3xgJz5JnQ/

What the test does is fill the filesystem with files, and perform 100 read only snapshots, and then checks for consistency. The functionality doesn't appear to be related to balancing, and the tests seem to pass more on the kernel in -proposed then they did before, so I believe the kernel in -proposed is safe and unlikely to introduce any new regressions.

I am happy with the performance of the kernel in -proposed, testing various balances, and running the xfstest suite, that I will mark it as verified.