On Thu, Apr 11, 2019 at 11:53 PM Michal Hocko <mhocko@xxxxxxxxxx> wrote:
On Thu 11-04-19 08:33:13, Matthew Wilcox wrote:
> On Wed, Apr 10, 2019 at 06:43:53PM -0700, Suren Baghdasaryan wrote:
> > Add new SS_EXPEDITE flag to be used when sending SIGKILL via
> > pidfd_send_signal() syscall to allow expedited memory reclaim of the
> > victim process. The usage of this flag is currently limited to SIGKILL
> > signal and only to privileged users.
>
> What is the downside of doing expedited memory reclaim? ie why not do it
> every time a process is going to die?
Well, you are tearing down an address space which might be still in use
because the task not fully dead yeat. So there are two downsides AFAICS.
Core dumping which will not see the reaped memory so the resulting
coredump might be incomplete. And unexpected #PF/gup on the reaped
memory will result in SIGBUS. These are things that we have closed our
eyes in the oom context because they likely do not matter. If we want to
use the same technique for other usecases then we have to think how much
that matter again.
After some internal discussions we realized that there is one additional downside of doing reaping unconditionally. If this async reaping is done on a more efficient and energy hungrier CPU irrespective of urgency of the kill it should end up costing more power overall (I’m referring here to assimetric architectures like ARM big.LITTLE). Obviously quantifying that cost is not easy as it depends on the usecase and a particular system but it won’t be zero. So I think we will need some gating condition after all.
--
Michal Hocko
SUSE Labs
--
You received this message because you are subscribed to the Google Groups "kernel-team" group.
To unsubscribe from this group and stop receiving emails from it, send an email to kernel-team+unsubscribe@xxxxxxxxxxx.
