On Tue 20-03-18 21:52:33, Tetsuo Handa wrote: > Michal Hocko wrote: > > On Tue 20-03-18 20:57:55, Tetsuo Handa wrote: > > > I found that it is not difficult to hit "oom_reaper: unable to reap pid:" > > > messages if the victim thread is doing copy_process(). Since I noticed > > > that it is likely helpful to show trace of unable to reap victim thread > > > for finding locations which should use killable wait, this patch does so. > > > > > > [ 226.608508] oom_reaper: unable to reap pid:9261 (a.out) > > > [ 226.611971] a.out D13056 9261 6927 0x00100084 > > > [ 226.615879] Call Trace: > > > [ 226.617926] ? __schedule+0x25f/0x780 > > > [ 226.620559] schedule+0x2d/0x80 > > > [ 226.623356] rwsem_down_write_failed+0x2bb/0x440 > > > [ 226.626426] ? rwsem_down_write_failed+0x55/0x440 > > > [ 226.629458] ? anon_vma_fork+0x124/0x150 > > > [ 226.632679] call_rwsem_down_write_failed+0x13/0x20 > > > [ 226.635884] down_write+0x49/0x60 > > > [ 226.638867] ? copy_process.part.41+0x12f2/0x1fe0 > > > [ 226.642042] copy_process.part.41+0x12f2/0x1fe0 /* i_mmap_lock_write() in dup_mmap() */ > > > [ 226.645087] ? _do_fork+0xe6/0x560 > > > [ 226.647991] _do_fork+0xe6/0x560 > > > [ 226.650495] ? syscall_trace_enter+0x1a9/0x240 > > > [ 226.653443] ? retint_user+0x18/0x18 > > > [ 226.656601] ? page_fault+0x2f/0x50 > > > [ 226.659159] ? trace_hardirqs_on_caller+0x11f/0x1b0 > > > [ 226.662399] do_syscall_64+0x74/0x230 > > > [ 226.664989] entry_SYSCALL_64_after_hwframe+0x42/0xb7 > > > > A single stack trace in the changelog would be sufficient IMHO. > > Appart from that. What do you expect users will do about this trace? > > Sure they will see a path which holds mmap_sem, we will see a bug report > > but we can hardly do anything about that. We simply cannot drop the lock > > from that path in 99% of situations. So _why_ do we want to add more > > information to the log? > > This case is blocked at i_mmap_lock_write(). But why does i_mmap_lock_write matter for oom_reaping. We are not touching hugetlb mappings. dup_mmap holds mmap_sem for write which is the most probable source of the backoff. -- Michal Hocko SUSE Labs
