Re: [patch 2/4] task isolation: sync vmstats on return to userspace

[Marcelo Tosatti]

On Tue, Aug 03, 2021 at 05:13:03PM +0200, nsaenzju@xxxxxxxxxx wrote:
> On Fri, 2021-07-30 at 17:18 -0300, Marcelo Tosatti wrote:
> > The logic to disable vmstat worker thread, when entering
> > nohz full, does not cover all scenarios. For example, it is possible
> > for the following to happen:
> > 
> > 1) enter nohz_full, which calls refresh_cpu_vm_stats, syncing the stats.
> > 2) app runs mlock, which increases counters for mlock'ed pages.
> > 3) start -RT loop
> > 
> > Since refresh_cpu_vm_stats from nohz_full logic can happen _before_
> > the mlock, vmstat shepherd can restart vmstat worker thread on
> > the CPU in question.
> > 
> > To fix this, use the task isolation prctl interface to quiesce 
> > deferred actions when returning to userspace.
> 
> Even though this is mostly targeted at nohz_full users, I believe I haven't
> seen anything in this series that forces the feature to be run on nohz_full
> CPUs (this is a good thing IMO). 

I don't think there is such a dependency either.

> So, I'd suggest to reword the patch
> description so it doesn't imply nohz_full is necessary to use this.

Its describing a fact from nohz_full where it can't guarantee entering
userspace with vmstat turned off (which is a reply to Christopher's
earlier comment that "this should just work with nohz_full and
logic to shutdown the vmstat delayed work timer").

Will add a comment to make it explicit that the series does not depend
on nohz_full.

Thanks.