Re: [PATCH] mm: terminate shrink_slab loop if signal is pending

[Tetsuo Handa]

On 2017/12/07 17:34, Michal Hocko wrote:
> On Wed 06-12-17 11:20:26, Suren Baghdasaryan wrote:
>> Slab shrinkers can be quite time consuming and when signal
>> is pending they can delay handling of the signal. If fatal
>> signal is pending there is no point in shrinking that process
>> since it will be killed anyway. This change checks for pending
>> fatal signals inside shrink_slab loop and if one is detected
>> terminates this loop early.
> 
> This is not enough. You would have to make sure the direct reclaim will
> bail out immeditally which is not at all that simple. We do check fatal
> signals in throttle_direct_reclaim and conditionally in shrink_inactive_list
> so even if you bail out from shrinkers we could still finish the full
> reclaim cycle.
> 
> Besides that shrinkers shouldn't really take very long so this looks
> like it papers over a real bug somewhere else. I am not saying the patch
> is wrong but it would deserve much more details to judge wether this is
> the right way to go for your particular problem.
> 

I wish that normal threads do not invoke direct reclaim operation.
Only dedicated kernel threads (such as filesystem's writeback) invoke
direct reclaim operation. Then, we can implement __GFP_KILLABLE for
normal threads, and hopefully get rid of distinction between GFP_NOIO/
GFP_NOFS/GFP_KERNEL because reclaim (and locking) dependency becomes
simpler.