On Mon 24-08-15 14:10:10, David Rientjes wrote: > On Fri, 21 Aug 2015, Tetsuo Handa wrote: > > > Why can't we think about choosing more OOM victims instead of granting > > access > > to memory reserves? > > > > We have no indication of which thread is holding a mutex that would need > to be killed, so we'd be randomly killing processes waiting for forward > progress. A worst-case scenario would be the thread is OOM_DISABLE and we > kill every process on the system needlessly. This problem obviously > occurs often enough that killing all userspace isnt going to be a viable > solution. > > > Also, SysRq might not be usable under OOM because workqueues can get stuck. > > The panic_on_oom_timeout was first proposed using a workqueue but was > > updated to use a timer because there is no guarantee that workqueues work > > as expected under OOM. > > > > I don't know anything about a panic_on_oom_timeout,
You were CCed on the discussion http://lkml.kernel.org/r/20150609170310.GA8990%40dhcp22.suse.cz > but panicking would > only be a reasonable action if memory reserves were fully depleted. That > could easily be dealt with in the page allocator so there's no timeout > involved. As noted in other email. Just depletion is not a good indicator. The system can still make a forward progress even when reserves are depleted. -- Michal Hocko SUSE Labs -- To unsubscribe from this list: send the line "unsubscribe linux-kernel" in the body of a message to majord...@vger.kernel.org More majordomo info at http://vger.kernel.org/majordomo-info.html Please read the FAQ at http://www.tux.org/lkml/