Re: Is it safe for kthreadd to drain_all_pages?

[Hugh Dickins]

On Thu, 6 Apr 2017, Mel Gorman wrote:
> On Wed, Apr 05, 2017 at 01:59:49PM -0700, Hugh Dickins wrote:
> > Hi Mel,
> > 
> > I suspect that it's not safe for kthreadd to drain_all_pages();
> > but I haven't studied flush_work() etc, so don't really know what
> > I'm talking about: hoping that you will jump to a realization.
> > 
> 
> You're right, it's not safe. If kthreadd is creating the workqueue
> thread to do the drain and it'll recurse into itself.
> 
> > 4.11-rc has been giving me hangs after hours of swapping load.  At
> > first they looked like memory leaks ("fork: Cannot allocate memory");
> > but for no good reason I happened to do "cat /proc/sys/vm/stat_refresh"
> > before looking at /proc/meminfo one time, and the stat_refresh stuck
> > in D state, waiting for completion of flush_work like many kworkers.
> > kthreadd waiting for completion of flush_work in drain_all_pages().
> > 
> 
> It's asking itself to do work in all likelihood.
> 
> > Patch below has been running well for 36 hours now:
> > a bit too early to be sure, but I think it's time to turn to you.
> > 
> 
> I think the patch is valid but like Michal, would appreciate if you
> could run the patch he linked to see if it also side-steps the same
> problem. 
> 
> Good spot!

Thank you both for explanations, and direction to the two "drainging"
patches.  I've put those on to 4.11-rc5 (and double-checked that I've
taken mine off), and set it going.  Fine so far but much too soon to
tell - mine did 56 hours with clean /var/log/messages before I switched,
so I demand no less of Michal's :).  I'll report back tomorrow and the
day after (unless badness appears sooner once I'm home).

Hugh