On Sat, May 9, 2020 at 4:15 AM Andrew Pam <andrew@xxxxxxxxxxxxxx> wrote: > > $ uname -a > Linux RainbowDash 5.7.0-050700rc4-generic #202005051752 SMP Tue May 5 > 21:56:32 UTC 2020 x86_64 x86_64 x86_64 GNU/Linux > > After a couple of cancel/resumes here we are again: OK I didn't previously pick up on the fact that you're cancelling a scrub. Suspending. Then resuming. Things to verify: 1. That a scrub started, and allowed to finish without any interruption, does finish. This is the only experience I have, but I'm wondering whether you've tried this just to eliminate the possibility of some weird multiple device scrub bug; or possibly something unique about your file system state. 2. That a scrub started, then cancelled, then resumed, also does finish (or not). 3. That a scrub started, then cancelled, then suspend-to-RAM, then resumed, also does finish (or not). The distinction between (2) and (3) is system sleep. Is this strictly a resume scrub bug. Or is there something about suspend that causing state to be lost? And a low priority (3b) variation might be suspend-to-disk; but this is not very well supported anyway, in particular UEFI Secure Boot enabled systems due to hibernation lockdown policy being common these days. -- Chris Murphy
