On Fri, Nov 13, 2020 at 5:00 PM Andrew Morton <akpm@xxxxxxxxxxxxxxxxxxxx> wrote: > > On Fri, 13 Nov 2020 16:06:25 -0800 Suren Baghdasaryan <surenb@xxxxxxxxxx> wrote: > > > On Fri, Nov 13, 2020 at 3:55 PM Andrew Morton <akpm@xxxxxxxxxxxxxxxxxxxx> wrote: > > > > > > On Fri, 13 Nov 2020 09:34:48 -0800 Suren Baghdasaryan <surenb@xxxxxxxxxx> wrote: > > > > > > > When a process is being killed it might be in an uninterruptible sleep > > > > which leads to an unpredictable delay in its memory reclaim. In low memory > > > > situations, when it's important to free up memory quickly, such delay is > > > > problematic. Kernel solves this problem with oom-reaper thread which > > > > performs memory reclaim even when the victim process is not runnable. > > > > Userspace currently lacks such mechanisms and the need and potential > > > > solutions were discussed before (see links below). > > > > This patch provides a mechanism to perform memory reclaim in the context > > > > of the process that sends SIGKILL signal. New SYNC_REAP_MM flag for > > > > pidfd_send_signal syscall can be used only when sending SIGKILL signal > > > > and will lead to the caller synchronously reclaiming the memory that > > > > belongs to the victim and can be easily reclaimed. > > > > > > hm. > > > > > > Seems to me that the ability to reap another process's memory is a > > > generally useful one, and that it should not be tied to delivering a > > > signal in this fashion. > > > > > > And we do have the new process_madvise(MADV_PAGEOUT). It may need a > > > few changes and tweaks, but can't that be used to solve this problem? > > > > Thank you for the feedback, Andrew. process_madvise(MADV_DONTNEED) was > > one of the options recently discussed in > > https://lore.kernel.org/linux-api/CAJuCfpGz1kPM3G1gZH+09Z7aoWKg05QSAMMisJ7H5MdmRrRhNQ@xxxxxxxxxxxxxx > > . The thread describes some of the issues with that approach but if we > > limit it to processes with pending SIGKILL only then I think that > > would be doable. > > Why would it be necessary to read /proc/pid/maps? I'd have thought > that a starting effort would be > > madvise((void *)0, (void *)-1, MADV_PAGEOUT) > > (after translation into process_madvise() speak). Which is equivalent > to the proposed process_madvise(MADV_DONTNEED_MM)? Yep, this is very similar to option #3 in https://lore.kernel.org/linux-api/CAJuCfpGz1kPM3G1gZH+09Z7aoWKg05QSAMMisJ7H5MdmRrRhNQ@xxxxxxxxxxxxxx and I actually have a tested prototype for that. If that's the preferred method then I can post it quite quickly. > > There may be things which trip this up, such as mlocked regions or > whatever, but we could add another madvise `advice' mode to handle > this?
