Comment 4 for bug 1902254

2) stress-ng --class filesystem,io

Compared the stress-ng output and kernel log between original/patched kernels.
No regressions.

Actually, in Bionic the original kernel didn't even finish stress-ng in ~20 hours, while the patched kernel finished it in ~1.5 hours (in line w/ other the releases.)
Thus it's apparently an improvement.

For documentation purposes, this seems to be due to the 'fiemap' stressor in the 'filesystem' class, which had defunct processes when I checked before cancelling the tests:

 $ ps axf
 ...
  2659 ? Ss 0:00 tmux new
  2661 pts/2 Ss 0:00 \_ -bash
  2692 pts/2 S+ 0:00 | \_ /bin/sh ./run-stressng.sh
  2743 pts/2 S+ 0:00 | \_ sudo stress-ng --sequential 2 --class filesystem,io -t 2m
  2745 pts/2 SL+ 0:00 | \_ stress-ng --sequential 2 --class filesystem io -t 2m
  2852 pts/2 D+ 0:00 | \_ stress-ng --sequential 2 --class filesystem io -t 2m
  2854 pts/2 D+ 0:01 | | \_ stress-ng --sequential 2 --class filesystem io -t 2m
  2856 pts/2 D+ 0:01 | | \_ stress-ng --sequential 2 --class filesystem io -t 2m
  2859 pts/2 D+ 0:01 | | \_ stress-ng --sequential 2 --class filesystem io -t 2m
  2860 pts/2 D+ 0:01 | | \_ stress-ng --sequential 2 --class filesystem io -t 2m
  2853 pts/2 D+ 0:00 | \_ stress-ng --sequential 2 --class filesystem io -t 2m
  2855 pts/2 Z+ 322:29 | \_ [stress-ng-fiema] <defunct>
  2857 pts/2 Z+ 323:55 | \_ [stress-ng-fiema] <defunct>
  2858 pts/2 Z+ 321:50 | \_ [stress-ng-fiema] <defunct>
  2861 pts/2 Z+ 322:51 | \_ [stress-ng-fiema] <defunct>
 11823 pts/4 Ss 0:00 \_ -bash
 ...