Re: [PATCH] mm: warn only once if page table misaccounting is detected

From: Michal Hocko
Date: Tue Nov 27 2018 - 11:29:21 EST


On Tue 27-11-18 15:36:38, Heiko Carstens wrote:
> On Tue, Nov 27, 2018 at 02:19:16PM +0100, Michal Hocko wrote:
> > On Tue 27-11-18 09:36:03, Heiko Carstens wrote:
> > > Use pr_alert_once() instead of pr_alert() if page table misaccounting
> > > has been detected.
> > >
> > > If this happens once it is very likely that there will be numerous
> > > other occurrence as well, which would flood dmesg and the console with
> > > hardly any added information. Therefore print the warning only once.
> >
> > Have you actually experience a flood of these messages? Is one per mm
> > message really that much?
>
> Yes, I did. Since in this case all compat processes caused the message
> to appear, I saw thousands of these messages.

This means something went colossally wrong and seeing an avalanche of
messages might be actually helpful because you can at least see the
pattern. I wonder whether the underlying issue would be obvious from a
single instance.

Maybe we want ratelimit instead?

> > If yes why rss counters do not exhibit the same problem?
>
> No rss counter messages appeared. Or do you suggest that the other
> pr_alert() within check_mm() should also be changed?

Whatever we go with (and I do not have a strong opinion here) we should
be consistent I believe.

--
Michal Hocko
SUSE Labs