When running this test on 3 concurrent VMs with this kernel this runs fine. With the ZFS backing store of the VMs set to sync on writes, I can slow the I/O and I get a lot of timeout issues which result in tests running for a long time and probably would get nuked by the testing infrastructure -
see attached dmesg log for the timeouts.
I am currently speculating that the tests are failing because VFAT is sub-optimal compared to other file systems and with a slow backing store we may be just getting lockup warnings because of slow I/O write out and/or syncs.
At least I have a reproducer. I don't think this is a critical show stopping bug. I will try and figure out what particular stress case is causing the issue, but this may take a while to iterate through.
When running this test on 3 concurrent VMs with this kernel this runs fine. With the ZFS backing store of the VMs set to sync on writes, I can slow the I/O and I get a lot of timeout issues which result in tests running for a long time and probably would get nuked by the testing infrastructure -
see attached dmesg log for the timeouts.
I am currently speculating that the tests are failing because VFAT is sub-optimal compared to other file systems and with a slow backing store we may be just getting lockup warnings because of slow I/O write out and/or syncs.
At least I have a reproducer. I don't think this is a critical show stopping bug. I will try and figure out what particular stress case is causing the issue, but this may take a while to iterate through.