The only thing that patch changes is that /dev/xen/evtchn and
/dev/xen/gntdev no longer get auto-created by the toolstack. So just make
sure you have those device nodes before starting the Xen tools. They ought
to be created by udev really but you could try mknod'ing them instead. You
can find the minor numbers in /proc/misc and then, for example 'mknod
/dev/xen/evtchn c 10 <minor_from_proc_misc>'.
-- Keir
On 14/07/2010 18:32, "Kathy Hadley" <Kathy.Hadley@xxxxxxxxxxxxxxx> wrote:
> We've tried the latest changesets in xen-unstable and the xen-unstable
> staging area, and still have the same issue. To re-cap, we are using
> the 32-bit hypervisor with the credit scheduler. Dom0 appears to boot
> successfully, but when we attempt to start unprivileged domains they
> hang and appear to be stuck in an idle loop.
>
> This appears to have been introduced in changeset 21507.
>
> I'd appreciate suggestions for how to fix this issue.
>
> Thank you,
> Kathy Hadley
>
>> -----Original Message-----
>> From: Keir Fraser [mailto:keir.fraser@xxxxxxxxxxxxx]
>> Sent: Wednesday, June 30, 2010 4:54 PM
>> To: Kathy Hadley; George Dunlap
>> Cc: xen-devel@xxxxxxxxxxxxxxxxxxx; Ian Campbell
>> Subject: Re: [Xen-devel] [PATCH 1/1] Xen ARINC 653 Scheduler (updated
>> to add support for CPU pools)
>>
>> On 30/06/2010 21:44, "Kathy Hadley" <Kathy.Hadley@xxxxxxxxxxxxxxx>
>> wrote:
>>
>>> Good afternoon,
>>> We have determined that changeset 21507 introduced the error that
>>> prevents unprivileged domains from running on our machine using the
>>> 32-bit hypervisor.
>>>
>>> With changeset 21506, we are able to run unprivileged domains
> using
>>> the credit scheduler. We cannot do so with changeset 21507 (or
>>> subsequent changesets) -- the unprivileged domains appear to be
> stuck
>> in
>>> an idle loop (as indicated by the call trace below).
>>>
>>> I'd appreciate help addressing this issue.
>>
>> The tools no longer automatically create /dev/xen/evtchn and expect it
>> to
>> already be created by the distro (e.g., via a udev rule) My guess
> would
>> be
>> that you are missing /dev/xen/evtchn. Ccing the patch author.
>>
>> -- Keir
>>
>>> Thanks,
>>> Kathy Hadley
>>>
>>>> -----Original Message-----
>>>> From: Keir Fraser [mailto:keir.fraser@xxxxxxxxxxxxx]
>>>> Sent: Wednesday, June 23, 2010 6:36 PM
>>>> To: Kathy Hadley; George Dunlap
>>>> Cc: xen-devel@xxxxxxxxxxxxxxxxxxx
>>>> Subject: Re: [Xen-devel] [PATCH 1/1] Xen ARINC 653 Scheduler
>> (updated
>>>> to add support for CPU pools)
>>>>
>>>> I've just built latest xen-unstable.hg and linux-2.6.18-xen.hg and
>>>> booted a
>>>> domU just fine. All my builds are 64-bit though whereas yours are
>> 32-
>>>> bit. I
>>>> suppose that could cause a difference (in particular, 32-bit
>>> hypervisor
>>>> is
>>>> less tested by people).
>>>>
>>>> -- Keir
>>>>
>>>> On 23/06/2010 22:16, "Kathy Hadley" <Kathy.Hadley@xxxxxxxxxxxxxxx>
>>>> wrote:
>>>>
>>>>> Keir,
>>>>> I see this same behavior when I run the credit scheduler. It
>>>> doesn't
>>>>> look like it's localized to the scheduler I'm working on. I
> pulled
>>>> the
>>>>> latest code from http://xenbits.xensource.com/linux-2.6.18-xen.hg
>>> and
>>>>> rebuilt the kernel earlier today, with no effect.
>>>>>
>>>>> Note that I can successfully start the domain with Xen-3.4.1 and
>>>>> Xen-4.0.0, using the same configuration file as I am using with
>>>>> xen-unstable.
>>>>>
>>>>> Kathy
>>>>>
>>>>>> -----Original Message-----
>>>>>> From: Keir Fraser [mailto:keir.fraser@xxxxxxxxxxxxx]
>>>>>> Sent: Wednesday, June 23, 2010 4:23 PM
>>>>>> To: Kathy Hadley; George Dunlap
>>>>>> Cc: xen-devel@xxxxxxxxxxxxxxxxxxx
>>>>>> Subject: Re: [Xen-devel] [PATCH 1/1] Xen ARINC 653 Scheduler
>>>> (updated
>>>>>> to add support for CPU pools)
>>>>>>
>>>>>> On 23/06/2010 20:57, "Kathy Hadley"
> <Kathy.Hadley@xxxxxxxxxxxxxxx>
>>>>>> wrote:
>>>>>>
>>>>>>> Call Trace:
>>>>>>> [<c01013a7>] hypercall_page+0x3a7 <--
>>>>>>> [<c0109005>] raw_safe_halt+0xa5
>>>>>>> [<c0104789>] xen_idle+0x49
>>>>>>> [<c010482d>] cpu_idle+0x8d
>>>>>>> [<c0404895>] start_kernel+0x3f5
>>>>>>> [<c04041d0>] do_early_param+0x80
>>>>>>>
>>>>>>> Does this shed any light on the situation?
>>>>>>
>>>>>> Looks like you're in the idle loop. So, no, it doesn't really
> shed
>>>>> much
>>>>>> useful light.
>>>>>>
>>>>>> -- Keir
>>>>>>
>>>>>
>>>>
>>>
>>
>
_______________________________________________
Xen-devel mailing list
Xen-devel@xxxxxxxxxxxxxxxxxxx
http://lists.xensource.com/xen-devel
|
Home