On 1/25/19 3:51 AM, Michal Hocko wrote: > On Thu 24-01-19 23:31:46, Qian Cai wrote: > [...] >> It looks like the put_and_wait commit just make the bug easier to reproduce, as >> it has finally been able to reproduce it (via a different path) after 50+ runs >> of migrate_pages03 on one of the affected machines even with the commit reverted. > > OK, great. This makes it a little bit less of a head scratcher then. > Could you confirm whether this is FS specific please? I will go and > check the migration path. Maybe we doing something wrong there but it > would be definitely good to know whether the underlying fs is really > relevant. Thanks! > So, I reinstalled everything using an ext4 rootfs, and then it becomes impossible to reproduce it anymore...
