Xen killing my e1000s?

Xen killing my e1000s?

[Jeremy Fitzhardinge]

I have this Xeon 3430-based system with a pair of on-board PCI-e 
e1000s.  When I'm running under Xen, they disappear in use.  That is, 
the PCI devices themselves seem to give up the ghost, and lspci -x shows 
all 0xff for their config spaces. They come back to life when I reset 
the machine.

When I boot under native Linux all seems fine.

I'm running with VT-d and all the other features enabled; I haven't 
tried turning them off yet.  There are no obvious messages from the 
kernel or Xen complaining about anything.

This is a new machine, so it could be an outright hardware problem, but 
it seems like an odd failure-mode for a hardware fault (I'd expect 
on-board devices to either work or not work, and not come back on reset).

Anyone got any ideas?

Thanks,
     J

Re: Xen killing my e1000s?

[Nathan Stratton]

On Fri, 11 Dec 2009, Jeremy Fitzhardinge wrote:

> I have this Xeon 3430-based system with a pair of on-board PCI-e e1000s. 
> When I'm running under Xen, they disappear in use.  That is, the PCI devices 
> themselves seem to give up the ghost, and lspci -x shows all 0xff for their 
> config spaces. They come back to life when I reset the machine.
>
> When I boot under native Linux all seems fine.
>
> I'm running with VT-d and all the other features enabled; I haven't tried 
> turning them off yet.  There are no obvious messages from the kernel or Xen 
> complaining about anything.
>
> This is a new machine, so it could be an outright hardware problem, but it 
> seems like an odd failure-mode for a hardware fault (I'd expect on-board 
> devices to either work or not work, and not come back on reset).
>
> Anyone got any ideas?

I am seeing something similar with my Xeon systems and igb (Intel 82576) 
cards. I see the ethernet ports die on my switch and the whole xen box 
crashes. Nothing in the logs at all, only clue I have is that the ethernet 
ports bounce on the switch.


><>
Nathan Stratton                                CTO, BlinkMind, Inc.
nathan at robotics.net                         nathan at blinkmind.com
http://www.robotics.net                        http://www.blinkmind.com

Re: Xen killing my e1000s?

[Jeremy Fitzhardinge]

On 12/11/2009 04:11 PM, Nathan Stratton wrote:
> I am seeing something similar with my Xeon systems and igb (Intel 
> 82576) cards. I see the ethernet ports die on my switch and the whole 
> xen box crashes. Nothing in the logs at all, only clue I have is that 
> the ethernet ports bounce on the switch.

I don't think my case is very similar.  The e1000 devices go away, but 
the system seems perfectly happy otherwise (net traffic just stops).  
The leds on the ports look normal too, so the phys appear to be acting 
normally.

     J

Re: Xen killing my e1000s?

[Andrew Lyon]

On Fri, Dec 11, 2009 at 11:36 PM, Jeremy Fitzhardinge <jeremy@goop.org> wrote:
> I have this Xeon 3430-based system with a pair of on-board PCI-e e1000s.
>  When I'm running under Xen, they disappear in use.  That is, the PCI
> devices themselves seem to give up the ghost, and lspci -x shows all 0xff
> for their config spaces. They come back to life when I reset the machine.
>
> When I boot under native Linux all seems fine.
>
> I'm running with VT-d and all the other features enabled; I haven't tried
> turning them off yet.  There are no obvious messages from the kernel or Xen
> complaining about anything.
>
> This is a new machine, so it could be an outright hardware problem, but it
> seems like an odd failure-mode for a hardware fault (I'd expect on-board
> devices to either work or not work, and not come back on reset).
>
> Anyone got any ideas?
>
> Thanks,
>    J
>
> _______________________________________________
> Xen-devel mailing list
> Xen-devel@lists.xensource.com
> http://lists.xensource.com/xen-devel
>

Have you tried using the latest version of the e1000 driver from intel
rather than the in-kernel one? I've had the same problem when using
older kernels on systems with e1000 and Xen, I used the intel driver
for a while but once I switched to using a newer dom0 kernel the
problem went away.

Andy

Pygrub cannot detect kernel on ISO image at Xen 4.0-rc1-pre Dom0

[Boris Derzhavets]

[-- Attachment #1.1: Type: text/plain, Size: 591 bytes --]

Profile bellow works fine to start OSOL DomU install at Xen 3.4.2 Dom0 (2.6.31.6 pv)

root@ServerKoala:~# cat osol.install
name="OSOL127"
memory=2048
bootloader="/usr/local/bin/pygrub"
kernel="/platform/i86xpv/kernel/amd64/unix"
ramdisk="/boot/amd64/x86.microroot"
disk=['phy:/dev/loop0,xvdc:cdrom,r','phy:/dev/sda9,xvda,w']
vif=['bridge=eth0']

Same profile cannot find kernel on ISO image at Xen 4.0-rc1-pre Dom0(2.6.31.6 pv)

I just uninstalled old Xen 3.4.2 and installed freshly been built 4.0-rc1-pre on top of Ubuntu 9.10 Server.

Boris.





________________________________


      

[-- Attachment #1.2: Type: text/html, Size: 1143 bytes --]

[-- Attachment #2: Type: text/plain, Size: 138 bytes --]

_______________________________________________
Xen-devel mailing list
Xen-devel@lists.xensource.com
http://lists.xensource.com/xen-devel

Re: Xen killing my e1000s?

[Jeremy Fitzhardinge]

On 12/12/09 06:16, Andrew Lyon wrote:
> Have you tried using the latest version of the e1000 driver from intel
> rather than the in-kernel one? I've had the same problem when using
> older kernels on systems with e1000 and Xen, I used the intel driver
> for a while but once I switched to using a newer dom0 kernel the
> problem went away.
>    

Which drivers are those?

I've been reproducing this fairly reliably with the stock F12 kernel, so 
I think Xen is off the hook.  Still unsure whether its a driver problem 
or something to do with the platform (hardware/BIOS/etc).  It's an Ibex 
Peak-based system, which I think it relatively new.  The BIOS is 1.0a, 
so is probably due for an update...

     J

Re: Xen killing my e1000s?

[Andrew Lyon]

On Sun, Dec 13, 2009 at 7:09 PM, Jeremy Fitzhardinge <jeremy@goop.org> wrote:
> On 12/12/09 06:16, Andrew Lyon wrote:
>>
>> Have you tried using the latest version of the e1000 driver from intel
>> rather than the in-kernel one? I've had the same problem when using
>> older kernels on systems with e1000 and Xen, I used the intel driver
>> for a while but once I switched to using a newer dom0 kernel the
>> problem went away.
>>
>
> Which drivers are those?

http://downloadcenter.intel.com/Detail_Desc.aspx?agr=Y&DwnldID=9180&lang=eng
http://downloadcenter.intel.com/Detail_Desc.aspx?agr=Y&DwnldID=15817&lang=eng
http://downloadcenter.intel.com/Detail_Desc.aspx?agr=Y&DwnldID=13663&lang=eng



>
> I've been reproducing this fairly reliably with the stock F12 kernel, so I
> think Xen is off the hook.  Still unsure whether its a driver problem or
> something to do with the platform (hardware/BIOS/etc).  It's an Ibex
> Peak-based system, which I think it relatively new.  The BIOS is 1.0a, so is
> probably due for an update...
>
>    J
>