On Tuesday, January 12, 2016 07:17:19 PM Tetsuo Handa wrote: > Michal Hocko write: [cut] > > I am not sure I am following you here but how do you detect that the > > userspace has corrupted your image or accesses an already (half) > > suspended device or something similar? > > Can't we determine whether the OOM killer might have corrupted our image > by checking whether oom_killer_disabled is kept true until the point of > final decision? The freezing is really not about keeping the image consistent etc. It is not about hibernation specifically even. > To me, satisfying allocation requests by kernel threads by invoking the > OOM killer and aborting suspend operation if the OOM killer was invoked > sounds cleaner than forcing !__GFP_NOFAIL allocation requests to fail. What if the suspend is on emergency, like low battery or thermal? Thanks, Rafael -- 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>