Comment 9 for bug 1780137

I am no longer able to reproduce after applying Ted's patch. I was able to run my unit test 20 times w/o failing on upstream + patch on d05-6. I then switched over to the Ubuntu kernel + patch, and it has now passed 64 times (and counting).

I then looked to see why Ike is still observing a failure. My theory is that the filesystem Ike was testing was already corrupted by a previous *unpatched* run, so the kernel is finding pre-existing corruption. Evidence follows.

The last record of "sudo mkfs.ext4" running in /var/log/auth.log:

Jul 5 06:43:39 d05-4 sudo: ubuntu : TTY=ttyAMA0 ; PWD=/home/ubuntu ; USER=root ; COMMAND=/sbin/mkfs.ext4 /dev/sda2

While the kernel w/ the fix wasn't built until Jul 9:
[ 0.000000] Linux version 4.15.0-25-generic (root@recht) (gcc version
 7.3.0 (Ubuntu/Linaro 7.3.0-16ubuntu3)) #27+ext4msg61578.1 SMP Mon Jul 9 08:28:49 UTC 2018 (Ubuntu 4.1
5.0-25.27+ext4msg61578.1-generic 4.15.18)

The first time /dev/sda2 was mounted after booting this kernel, it reported known errors:
Jul 9 05:43:33 d05-4 kernel: [ 138.522140] EXT4-fs (sda2): warning: mounting fs with errors, running e2fsck is recommended

Looking at the conserver log (logs all console activity on this system), it looks like the test used did not reformat the disk between iterations:
root@d05-4:~# ^G^G^G^G^G^G^G^G^G^G^G^Gwhile true; do sudo /usr/lib/plainbox-provider-checkbox/bin/disk
_st^Mtress_ng sda --base-time 240 --really-run; done

Finally, I manually ran mkfs.ext4 on /dev/sda2. Afterwards, both my unit test (https://bugs.launchpad.net/ubuntu/+source/linux/+bug/1780137/comments/5) and the full disk_stress_ng cert test Ike was running passed without error.