https://bugs.freebsd.org/bugzilla/show_bug.cgi?id=235856
--- Comment #53 from Colin Percival ---
Charles: Yes, a stock FreeBSD 12.1-RELEASE kernel should have those sysctls
(they're in if_ena.ko technically but you should have that loaded...) and it
would be great to hear if those are non-zero an
https://bugs.freebsd.org/bugzilla/show_bug.cgi?id=235856
--- Comment #52 from Charles O'Donnell ---
Would "sysctl -a | grep mjum_alloc_fail" provide productive results on an
unpatched 12.1-R kernel? I am not running the patches, but I am getting these
same freezes, and I can run queries prior to
https://bugs.freebsd.org/bugzilla/show_bug.cgi?id=235856
--- Comment #51 from Colin Percival ---
(In reply to mail from comment #50)
Hi Ruben,
Thanks! This is consistent with my current understanding, which is that the
patch I provided fixes the data structure corruption issues (which showed u
On Mon., 20 Apr. 2020, 7:09 pm Felix Palmen, wrote:
>
> > Xeon are not affected, since they have vAPIC, so they should be already
> > running as fast as possible. Correct?
>
> I don't know the details, but on my Xeon E3-1240L v5, it did have a huge
> effect.
>
>
>
^^^ This.
Not all Xeon CPUs ha
* Andrea Venturoli [20200420 10:53]:
> First: this patch seems to apply cleanly to 12.1. Is it safe? At least worth
> testing?
I applied it to my 12.1-RELEASE before it was even committed to head --
the result is awesome.
> Xeon are not affected, since they have vAPIC, so they should b
On 2020-04-19 19:23, Michael Reifenberger wrote:
Hi,
r358848 has been MFC'd to stable/12
Hello.
Can I ask some questions about this?
First: this patch seems to apply cleanly to 12.1. Is it safe? At least
worth testing?
Second.
IIUIC it's useless to test this on an AMD system, since the pat
https://bugs.freebsd.org/bugzilla/show_bug.cgi?id=235856
--- Comment #50 from m...@rubenvos.com ---
(In reply to Colin Percival from comment #49)
Hi Colin,
The machine that was exhibiting the most freezes will be tested extensively
during the next couple of days. No errors on that instance yet.
