[Sorry for spamming you] On Thu 19-05-16 09:17:36, Michal Hocko wrote: > On Thu 19-05-16 08:53:29, Michal Hocko wrote: > > On Wed 18-05-16 14:09:32, Andrew Morton wrote: > > > On Wed, 18 May 2016 16:15:45 +0200 Michal Hocko <mhocko@xxxxxxxxxx> wrote: > > > > > > > > This patch adds a counter to signal_struct for tracking how many > > > > > TIF_MEMDIE threads are in a given thread group, and check it at > > > > > oom_scan_process_thread() so that select_bad_process() can use > > > > > for_each_process() rather than for_each_process_thread(). > > > > > > > > OK, this looks correct. Strictly speaking the patch is missing any note > > > > on _why_ this is needed or an improvement. I would add something like > > > > the following: > > > > " > > > > Although the original code was correct it was quite inefficient because > > > > each thread group was scanned num_threads times which can be a lot > > > > especially with processes with many threads. Even though the OOM is > > > > extremely cold path it is always good to be as effective as possible > > > > when we are inside rcu_read_lock() - aka unpreemptible context. > > > > " > > > > > > This sounds quite rubbery to me. Lots of code calls > > > for_each_process_thread() and presumably that isn't causing problems. > > > > Yeah, many paths call for_each_process_thread but they are > > O(num_threads) while this is O(num_threads^2). > > And just to clarify the regular num_threads^2 is the absolute worst case > which doesn't happen normally. We would be closer to O(num_threads) but > there is no reason to risk pathological cases when we can simply use > for_each_process to achieve the same. Blee, fat fingers today... some vim-foo removed the rest of the paragraph which was: " Especially when calculating oom_badness for all threads in the same thread group is just pointless wasting of cycles (e.g. take task_lock etc.). " Tetsuo, btw. I guess you can safely drop /* Prefer thread group leaders for display purposes */ if (points == chosen_points && thread_group_leader(chosen)) continue; from select_bad_process because you are iterating group leaders. -- Michal Hocko SUSE Labs -- To unsubscribe, send a message with 'unsubscribe linux-mm' in the body to majordomo@xxxxxxxxx. For more info on Linux MM, see: http://www.linux-mm.org/ . Don't email: <a href=mailto:"dont@xxxxxxxxx";> email@xxxxxxxxx </a>