On Wed 25-07-18 21:23:11, Tetsuo Handa wrote: > Michal, I think that I hit WQ OOM lockup caused by lack of guaranteed schedule_timeout_*() for WQ. > I think we should immediately send https://marc.info/?l=linux-mm&m=153062798103081 > (or both http://lkml.kernel.org/r/20180709074706.30635-1-mhocko@xxxxxxxxxx and > https://marc.info/?l=linux-mm&m=152723708623015 ) to linux.git so that we can send > to stable kernels without waiting for "mm, oom: cgroup-aware OOM killer" patchset. Then do not pollute unrelated threads. Really! The patch to drop the sleep from the oom_lock is in mmotm tree already. I really do not see why you make it more important than it really is (ohhh it is the dubious CVE you have filed?). If you have hit a WQ OOM lockup beucase of a missing schedule_timeout then report it in a new email thread and we can discuss the proper way of handling it. Thanks! -- Michal Hocko SUSE Labs