Re: XEN/ARM Cortex-A72: Unable to copy the kernel in the hwdom memory

[Anda-alexandra Dorneanu <anda-alexandra.dorneanu@xxxxxxx>]

> On 18/11/2020 14:56, Anda-alexandra Dorneanu wrote:
>> Hi Leo,
>
> Hi,
>
> I think you forgot to CC Leo :).

Yes, my bad.

>>> I am trying to set up XEN on a ARM Cortex-A72 (NXP LS1028A) and running 
>>> into some problems here.
>>> While some of them I could solve on my own, the following error is hard for 
>>> me to understand and solve.
>>> I would be grateful for some hints e.g. about some specific nodes in the 
>>> device tree that I might need to take care of.
>>>
>>> (XEN) *** LOADING DOMAIN 0 ***
>>> (XEN) Loading Domd0 kernel from boot module @ 0000000081200000
>>> (XEN) Allocating 1:1 mappings totalling 512MB for dom0:
>>> (XEN) BANK[0] 0x000000c0000000-0x000000e0000000 (512MB)
>>> (XEN) Grant table range: 0x00000081000000-0x00000081040000
>>> (XEN) Allocating PPI 16 for event channel interrupt
>>> (XEN) Loading zImage from 0000000081200000 to 
>>> 00000000c0080000-00000000c1597008
>>> (XEN)
>>> (XEN) ****************************************
>>> (XEN) Panic on CPU 0:
>>> (XEN) Unable to copy the kernel in the hwdom memory
>>> (XEN) ****************************************
>>>
>>> For me, the memory ranges given look good. Or is there something here I 
>>> overlooked?
>>>
>>> My XEN boot arguments: "console=dtuart dtuart=serial0 dom0_mem=512M 
>>> dom0_max_vcpus=1 bootscrub=0 vwfi=native sched=null"
>>> Full boot log is attached (does this work here?).
>>> If anything else is needed to at least make a best guess what the problem 
>>> is, let me know.
>>>
>>> I would really appreciate the help!
>>> Thanks & best regards,
>>> Leo
>>
>> I encountered the same issue. A possible workaround is to remove the gpu 
>> node from the dts.
>> The issue occurs due to the entire low memory bank being reserved for this 
>> device.
>
> Xen 4.12 and onwards should be able to discover reserved memory regions.
>
> How are the regions described in the device-tree?
>

There's a known issue with the GPU on our side and we're working on it. At the 
moment we recommend to apply this workaround.

Best regards,
Anda

> Cheers,
>
> --
> Julien Grall