[prev in list] [next in list] [prev in thread] [next in thread]
List: illumos-discuss
Subject: [discuss] Re: [smartos-discuss] Illumos failure on EPYC at Hetzner
From: Sriram Narayanan <sriramnrn () gmail ! com>
Date: 2019-02-03 9:06:05
Message-ID: CANiY96aSB3kkMg6UDN=viXsoaPTFgnbfxKBHHOAtbU1ocdOG_w () mail ! gmail ! com
[Download RAW message or body]
On Wed, Jan 9, 2019 at 12:21 PM Robert Mustacchi <rm@joyent.com> wrote:
> On 1/8/19 20:13 , Sriram Narayanan wrote:
> > Hi all:
> >
> > I got access to an EPYC at Hetzner. When I booted the latest SmartOS on
> it,
> > I got the error "MCA is not available on core 0 (cmi_hdl_create returned
> > NULL)".
> >
> > This is specifically here:
> >
> https://github.com/joyent/illumos-joyent/blob/master/usr/src/uts/i86pc/os/cmi_hw.c#L1242
> >
> > Has anyone got SmartOS or another Illumos based distro running on an
> EPYC?
>
> Hi Sriram,
>
> This and a number of other improvements are in the works around this and
> other EPYC improvements. This particular issue is a multi-pronged
> problem of topology which I just finished putting a fix together for
> yesterday. Should hopefully wrap up a good chunk of this and other stuff
> over the next week or so and have it upstream in illumos shortly there
> after. This is https://smartos.org/bugview/OS-7486 and the fix I'm
> currently testing out is https://cr.joyent.us/#/c/5324/. There's some
> other stuff I have in the queue to improve the general AMD EPYC support.
>
Hi Robert:
I was able to install and boot SmartOS build joyent_20190131T012237Z on an
AMD EPYC at Hetzner. Thanks very much for your work in this space.
The prtconf-v output is here:
https://gist.github.com/sriramnrn/1b5326544fcecfea32e2d667369ab741
There is an error that "fmadm faulty" reports:
Fault class : fault.io.pciex.device-interr
Affects : dev:////pci@1e,0/pci1022,1454@8,1/pci1458,1000@0,2
faulted and taken out of service
FRU : "MB"
(hc://:product-id=MZ31-AR0-00:server-id=karna:chassis-id=01234567890123456789AB/motherboard=0)
faulty
Here is a screenshot of the server console: https://imgur.com/a/WHWfAOa
Is there a test suite that I could run to assess whether the CPU and the
various devices are ready for use with SmartOS?
-- Sriram
------------------------------------------
illumos: illumos-discuss
Permalink: https://illumos.topicbox.com/groups/discuss/Tf06a5f6adce2665e-M3785dbbe5b0bd80951eadf73
Delivery options: https://illumos.topicbox.com/groups/discuss/subscription
[Attachment #3 (text/html)]
<html><html><div dir="ltr"><div dir="ltr"><div dir="ltr"><br /></div><br /><div \
class="gmail_quote"><div class="gmail_attr" dir="ltr">On Wed, Jan 9, 2019 at 12:21 PM \
Robert Mustacchi <<a href="mailto:rm@joyent.com">rm@joyent.com</a>> wrote:<br \
/></div><blockquote class="gmail_quote" style="margin:0px 0px 0px \
0.8ex;border-left:1px solid rgb(204,204,204);padding-left:1ex">On 1/8/19 20:13 , \
Sriram Narayanan wrote:<br />> Hi all:<br />> <br />> I got access to an \
EPYC at Hetzner. When I booted the latest SmartOS on it,<br />> I got the error \
"MCA is not available on core 0 (cmi_hdl_create returned<br />> \
NULL)".<br />> <br />> This is specifically here:<br />> <a \
href="https://github.com/joyent/illumos-joyent/blob/master/usr/src/uts/i86pc/os/cmi_hw.c#L1242" \
rel="noreferrer" target="_blank">https://github.com/joyent/illumos-joyent/blob/master/usr/src/uts/i86pc/os/cmi_hw.c#L1242</a><br \
/>> <br />> Has anyone got SmartOS or another Illumos based distro running on \
an EPYC?<br /><br />Hi Sriram,<br /><br />This and a number of other improvements are \
in the works around this and<br />other EPYC improvements. This particular issue is a \
multi-pronged<br />problem of topology which I just finished putting a fix together \
for<br />yesterday. Should hopefully wrap up a good chunk of this and other stuff<br \
/>over the next week or so and have it upstream in illumos shortly there<br />after. \
This is <a href="https://smartos.org/bugview/OS-7486" rel="noreferrer" \
target="_blank">https://smartos.org/bugview/OS-7486</a> and the fix I'm<br \
/>currently testing out is <a href="https://cr.joyent.us/#/c/5324/" rel="noreferrer" \
target="_blank">https://cr.joyent.us/#/c/5324/</a>. There's some<br />other stuff \
I have in the queue to improve the general AMD EPYC support.<br \
/></blockquote><div><br /></div><div><div>Hi Robert:</div><div><br /></div><div>I was \
able to install and boot SmartOS build joyent_20190131T012237Z on an AMD EPYC at \
Hetzner. Thanks very much for your work in this space.</div><div><br /></div><div>The \
prtconf-v output is here: <a \
href="https://gist.github.com/sriramnrn/1b5326544fcecfea32e2d667369ab741">https://gist.github.com/sriramnrn/1b5326544fcecfea32e2d667369ab741</a></div><div><br \
/></div><div>There is an error that "fmadm faulty" reports:</div><div>Fault \
class : fault.io.pciex.device-interr</div><div>Affects : \
dev:////pci@1e,0/pci1022,1454@8,1/pci1458,1000@0,2</div><div> \
faulted and taken out of \
service</div><div>FRU : "MB" \
(hc://:product-id=MZ31-AR0-00:server-id=karna:chassis-id=01234567890123456789AB/motherboard=0)</div><div> \
faulty</div><div><span \
style="white-space:pre"> </span> </div><div>Here is a screenshot of the \
server console: <a href="https://imgur.com/a/WHWfAOa">https://imgur.com/a/WHWfAOa</a></div><div><br \
/></div><div>Is there a test suite that I could run to assess whether the CPU and the \
various devices are ready for use with SmartOS?</div><div><br /></div></div><div>-- \
Sriram</div><div><br /></div></div></div></div></html><div id="topicbox-footer" \
style="margin:10px 0 0;border-top:1px solid \
#ddd;border-color:rgba(0,0,0,.15);padding:7px 0;">
<strong><a href="https://illumos.topicbox.com/latest" \
style="color:inherit;text-decoration:none">illumos</a></strong> / illumos-discuss / \
see <a href="https://illumos.topicbox.com/groups/discuss">discussions</a>
+
<a href="https://illumos.topicbox.com/groups/discuss/members">participants</a>
+
<a href="https://illumos.topicbox.com/groups/discuss/subscription">delivery \
options</a> <a href="https://illumos.topicbox.com/groups/discuss/Tf06a5f6adce2665e-M3785dbbe5b0bd80951eadf73" \
style="float:right">Permalink</a> </div>
</html>
[prev in list] [next in list] [prev in thread] [next in thread]
Configure |
About |
News |
Add a list |
Sponsored by KoreLogic