Comment 8 for bug 1847340

Regression test results with (x)fstests.

Bionic, data=journal: no regression.

Original:
---

$ ./fstests-check-logs.sh orig-journal/xfstests.log*
Kernel version: 5.4.0-84-generic #94-Ubuntu

- Number of Runs: 10

- Always fail (failed every time)
     10 ext4/024
     10 ext4/034
     10 ext4/045
     10 generic/082
     10 generic/219
     10 generic/223
     10 generic/230
     10 generic/231
     10 generic/232
     10 generic/233
     10 generic/235
     10 generic/270
     10 generic/381
     10 generic/382
     10 generic/397
     10 generic/398
     10 generic/399
     10 generic/419
     10 generic/421
     10 generic/429
     10 generic/435
     10 generic/440
     10 generic/441
     10 generic/537
     10 generic/548
     10 generic/549
     10 generic/550
     10 generic/566
     10 generic/600

- Flaky (failed, but not every time)
      2 generic/388
      2 generic/475
      9 generic/347

Modified:
---

$ ./fstests-check-logs.sh mod-journal/xfstests.log*
Kernel version: 5.4.0-84-generic #94+test20210906b1

- Number of Runs: 10

- Always fail (failed every time)
     10 ext4/024
     10 ext4/034
     10 ext4/045
     10 generic/082
     10 generic/219
     10 generic/223
     10 generic/230
     10 generic/231
     10 generic/232
     10 generic/233
     10 generic/235
     10 generic/270
     10 generic/381
     10 generic/382
     10 generic/397
     10 generic/398
     10 generic/399
     10 generic/419
     10 generic/421
     10 generic/429
     10 generic/435
     10 generic/440
     10 generic/441
     10 generic/537
     10 generic/548
     10 generic/549
     10 generic/550
     10 generic/566
     10 generic/600

- Flaky (failed, but not every time)
      1 generic/388
      6 generic/484 *
      9 generic/347

* This test has been confirmed to fail on the original kernel too,
with a larger sample size (more runs w/ only the generic/484 test.)
Even though it takes more than a few hundred runs to manifest a few
times, it demonstrates that the problematic behavior is also there,
and might be just a timing change that made it happen more often now
(but it's still flaky, doesn't fail every single time.)

$ ls -1 xfstests.log* | wc -l
518

$ grep -h ^Failures: xfstests.log* | sort | uniq -c
      6 Failures: generic/484