> > Note also the note about "p == current" check. it should be fixed too. > > I am resending the fixes above plus the new one. > > David, Kosaki, what do you think? Oleg, could you please give me some testing time? Now TIF_MEMDIE has really unclear and nasty meanings. mainly 1) to prevent multiple oom-killer (see select_bad_process) and 2) to allow to use zone's last resort reserved memory (see gfp_to_alloc_flags). The latter has really problematic even if apply or not apply your patch. If no apply your patch, multi thread application have a lot of risk to fail to exit when oom killed. because sub threads can't use reserved memory and may get stucked exiting path. if apply your patch, multi thread application also have a lot of risk to fail to exit when oom killed. because if all thread used a little reserved memory but it is not enough successful exit. It become deadlock. The optimal way is, take a bonus one thread and successfull thread exiting pass a bonus to sibling thread. But, who want to put performance overhead into thread exiting path for only really really rare oom events? this is the problem. therefore, I can't put ack until I've finished some test. Thanks. -- To unsubscribe, send a message with 'unsubscribe linux-mm' in the body to majordomo@xxxxxxxxxx For more info on Linux MM, see: http://www.linux-mm.org/ . Fight unfair telecom internet charges in Canada: sign http://stopthemeter.ca/ Don't email: <a href=mailto:"dont@xxxxxxxxx";> email@xxxxxxxxx </a>