<div dir="ltr">On 21 November 2017 at 11:09, Richard Purdie <span dir="ltr"><<a href="mailto:richard.purdie@linuxfoundation.org" target="_blank">richard.purdie@linuxfoundation.org</a>></span> wrote:<br><div class="gmail_extra"><div class="gmail_quote"><blockquote class="gmail_quote" style="margin:0 0 0 .8ex;border-left:1px #ccc solid;padding-left:1ex"><div class="HOEnZb"><div class="h5"><span style="color:rgb(34,34,34)">> > o Issues with 4.13.10 host kernels booting kvm x86 guests on</span><br></div></div><span class="">
> > Tumbleweed (Suse) and Fedora 26 (attempting to see if 4.13.12<br>
> > helps)<br>
> <br>
> Robert, can you test Fedora 26. It would help to have a defect open<br>
> with steps to reproduce or<br>
> something about the typical workflow/ build time/ day of the week/<br>
> phase of the moon. <br>
<br>
</span>FWIW we have noticed that the choice of kernel timers seems to vary in<br>
the x86_64 boots but not with a pattern that matches the hangs.</blockquote><div><br></div><div><br></div><div>Also of note is that last week the cluster was hitting this once in every nightly run, but after a complete reboot and kernel upgrade we haven't seen it so far. So either it's gone with the latest kernels, or it's a slow accumulation of non-movable memory that eventually breaks kvm (my money is on the latter).</div><div> <br></div><blockquote class="gmail_quote" style="margin:0 0 0 .8ex;border-left:1px #ccc solid;padding-left:1ex"><span class="">> > o a perf build race (bug 12302)<br>
> I'll take a look to:<br>
> - see if I can duplicate the bug on a fast build host<br>
> - check upstream to see if the bug is known/fixed<br>
> - see if I can spot a race in the build rules.<br>
<br>
</span>Sounds good, thanks!</blockquote><div><br></div><div>It looks and sounds like a classic makefile dependency bug, should be reproducible on demand with careful explicit target steps. The perf makefiles are not exactly easy to hack though...</div><div><br></div><div>Ross</div></div></div></div>