Re: interaction of MADV_PAGEOUT with CoW anonymous mappings?

[Michal Hocko <mhocko@kernel.org>]

On Thu 12-03-20 19:08:51, Minchan Kim wrote:
> On Thu, Mar 12, 2020 at 09:41:55PM +0100, Michal Hocko wrote:
> > On Thu 12-03-20 13:16:02, Minchan Kim wrote:
> > > On Thu, Mar 12, 2020 at 09:22:48AM +0100, Michal Hocko wrote:
> > [...]
> > > > From eca97990372679c097a88164ff4b3d7879b0e127 Mon Sep 17 00:00:00 2001
> > > > From: Michal Hocko <mhocko@suse.com>
> > > > Date: Thu, 12 Mar 2020 09:04:35 +0100
> > > > Subject: [PATCH] mm: do not allow MADV_PAGEOUT for CoW pages
> > > > 
> > > > Jann has brought up a very interesting point [1]. While shared pages are
> > > > excluded from MADV_PAGEOUT normally, CoW pages can be easily reclaimed
> > > > that way. This can lead to all sorts of hard to debug problems. E.g.
> > > > performance problems outlined by Daniel [2]. There are runtime
> > > > environments where there is a substantial memory shared among security
> > > > domains via CoW memory and a easy to reclaim way of that memory, which
> > > > MADV_{COLD,PAGEOUT} offers, can lead to either performance degradation
> > > > in for the parent process which might be more privileged or even open
> > > > side channel attacks. The feasibility of the later is not really clear
> > > 
> > > I am not sure it's a good idea to mention performance stuff because
> > > it's rather arguble. You and Johannes already pointed it out when I sbumit
> > > early draft which had shared page filtering out logic due to performance
> > > reason. You guys suggested the shared pages has higher chance to be touched
> > > so that if it's really hot pages, that whould keep in the memory. I agree.
> > 
> > Yes, the hot memory is likely to be referenced but the point was an
> > unexpected latency because of the major fault. I have to say that I have
> 
> I don't understand your point here. If it's likely to be referenced
> among several processes, it doesn't have the major fault latency.
> What's your point here?

a) the particular CoW page might be cold enough to be reclaimed and b)
nothing really prevents the MADV_PAGEOUT to be called faster than the
reference bit being readded.

> > underestimated the issue because I was not aware of runtimes mentioned
> > in the referenced links. Essentially a lot of CoW memory shared over
> > security domains.
> 
> I tend to agree about security part in the description, but still don't
> agree with performance concern in the description so I'd like to remove
> it in the description. Current situation is caused by security concern
> unfortunately, not performance reason.

Well, I have to admit that I haven't seen any actual numbers here but
considering zygote like workload I would rather not even risk it. Even
if the risk is theoretical I would rather put the restriction and
mention it in the changelog. If somebody would like to drop this
restriction it is at least more clear what to test for.

-- 
Michal Hocko
SUSE Labs