Re: Strange issues with epoll since 5.0

[Eric Wong]

Davidlohr Bueso <dave@xxxxxxxxxxxx> wrote:
> On Sun, 28 Apr 2019, Eric Wong wrote:
> 
> > Just running one test won't trigger since it needs a busy
> > machine; but:
> > 
> > 	make test/mgmt_auto_adjust.log
> > 	(and "rm make test/mgmt_auto_adjust.log" if you want to rerun)
> 
> fyi no luck reproducing on both either a large (280) or small (4 cpu)
> machine, I'm running it along with a kernel build overcommiting cpus x2.

Thanks for taking a look.

> Is there any workload in particular you are using to stress the box?

Just the "make -j$N check" I mentioned up-thread which spawns a
lot of tests in parallel.  For the small 4 CPU machine,
-j32 or -j16 ought to be reproducing the problem.

I'll try to set aside some time this week to get familiar with
kernel internals w.r.t. signal handling.