[vfio-users] Passing intel_iommu=on in grub causes boot to immediately lock on two different X58 LGA1366 MB.

Eddie Yen missile0407 at gmail.com
Wed May 18 09:12:28 UTC 2016


As discussed in another recent thread, LGA2011 SandBridge-EP processors are
cheap, but motherboard prices are verging on absurd.


Indeed, X79 motherboard price just like gold, especially EOF. It's still
hard to find a good price in second-hand market.

But the other way, maybe can try Haswell-EP with X99 motherboard. Since
Broadwell-EP prepare to come out, it may a good opportunity to pick one
processor in second-hand market.

2016-05-18 4:52 GMT+08:00 Alex Williamson <alex.l.williamson at gmail.com>:

>
>
> On Tue, May 17, 2016 at 1:19 PM, Brian Yglesias <
> brian at atlanticdigitalsolutions.com> wrote:
>
>> The first is an Asus Rampage II Gene.  While the manual claims VT-d
>> support, Asus boards are known to break PCI pass through in linux.  The CPU
>> is a ci7 920.
>>
>> I went out and purchased a ga-ex58-ud5, partly because I'd heard reports
>> of it working in Linux with a few different configuratio
>>
>> The first is an Asus Rampage II Gene.  While the manual claims VT-d
>> support, Asus boards are known to break PCI pass through in linux.  The CPU
>> is a ci7 920.
>>
>> I went out and purchased a ga-ex58-ud5, partly because I'd heard reports
>> of it working in Linux with a few different configurations, and also
>> because I know GB released a beta BIOS in response to community requests
>> for proper VT-d support.
>>
>> The spec page for the 930 does not specifically list VT-d, and I could
>> use another CPU anyway, so I bought a Xeon E5670.
>>
>> I enable VT-d and once again, rcu_shed detected HardLock on cpu0.
>> Kthread starved for 769 jiffies!
>>
>> It is essentially the exact same error as with the Rampage II Gene.
>>
>> At this point I start looking for a software problem.  I'm using the
>> latest Debian, and the kernel is 4.4.X.  I installed a 4.2 kernel and tried
>> again.
>>
>> The same thing happened, except that now rcu_shed detects "stalls", not
>> HardLock, on cpu0.   The system does not boot beyond that point (this
>> happens right as the kernel is loaded at the beginning of boot).
>>
>> I'm not sure if this distinction is helpful.  The system hangs well
>> before logging, and there is no kernel dump.
>>
>> I took a picture of my screen with the older kernel (stalls not
>> hardlocks), and attached it.
>>
>
> Still not attached.
>
>
>> I'm using Proxmox 4.2 VE (latest), which I think I recall reading uses an
>> Ubuntu kernel.
>>
>> I attempted a couple of LiveCDs, only to find they don't use grub, and
>> the Intel_iommu parameter does nothing in isolinux.
>>
>
> Hmm, usually just a matter of hitting <tab> and adding the option, it's
> definitely possible to do this.
>
>
>> I'm not sure how to proceed from here, aside from buying new hardware.
>> X58 is my only option for LGA1366, and comparable LGA2011 stuff is not
>> cheap.
>>
> As discussed in another recent thread, LGA2011 SandBridge-EP processors
> are cheap, but motherboard prices are verging on absurd.
>
> _______________________________________________
> vfio-users mailing list
> vfio-users at redhat.com
> https://www.redhat.com/mailman/listinfo/vfio-users
>
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://listman.redhat.com/archives/vfio-users/attachments/20160518/c68772b4/attachment.htm>


More information about the vfio-users mailing list