[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: [PATCH v7 0/5] Fix ARM domcreate



On 10/13/26 2:28 PM, Orzel, Michal wrote:
> 
> On 12-Oct-26 15:36, Julian Vetter wrote:
>> Hello Michal and Anthony,
>> thank you for the reviews and acks! Michal, I have addressed your
>> remaining remarks on the "xen/arm: report clock_frequency via sysctl
>> physinfo, not createdomain" patch. The golang bindings are regenerated
>> for the new arm_clock_frequency_hz physinfo field, and
>> alloc_domaininfo() no longer references the removed clock_frequency
>> field. Apart from that I have collected the Acked-by/Reviewed-by tags.
>>
>> Signed-off-by: Julian Vetter <julian.vetter@xxxxxxxxxx>
> I was about to commit the series but I found out that some tests like
> test-mem-claim and possibly others still pass gic_version=0 which will cause 
> the
> domain creation to fail. You need to update the tests.

Yes, thank you. I didn't see this. I have fixed the tests, and ran them 
on a QEMU setup. 2 of them now pass the domain creation at least just 
like on master (1 doesn't work on QEMU because there is no IOMMU 
emulation?!):

```
@@@ RUN test-mem-claim
Memory claims tests
Free pages: 254208, Oustanding claims: 0
   Created d1
   Failed to populate physmap domain: 16 - Resource busy
@@@ RESULT test-mem-claim rc=1
@@@ RUN test-paging-mempool
Paging mempool tests
(XEN) [   28.386587] common/domain.c:845: IOMMU requested but not available
   Skip: 22 - Invalid argument
@@@ RESULT test-paging-mempool rc=0
@@@ RUN test-resource
XENMEM_acquire_resource tests
Test ARM
   Created d3
   Failed to populate physmap domain: 16 - Resource busy
@@@ RESULT test-resource rc=1
@@@ ALL DONE
```

But, unfortunately, these tests seem broken on ARM even on master, 
because on ARM domains start with an empty P2M pool, and since these 
tests call xc_domain_create() without setting the pool size first, they 
fail on the populate physmap call. I'm not sure how to proceed on this. 
Should I fix these tests in their entirety or leave this for another 
patch set? Thank you for your advice.

> 
>>
>> Andrew Cooper (1):
>>    ARM/sysctl: Expose the supported guest GIC modes in physinfo
>>
>> Julian Vetter (4):
>>    tools/arm: choose GIC version explicitly instead of relying on
>>      GIC_NATIVE
>>    xen/arm: remove XEN_DOMCTL_CONFIG_GIC_NATIVE from the ABI
>>    xen/arm: report clock_frequency via sysctl physinfo, not createdomain
>>    xen: make config argument const
> Here you are still missing tags from e.g. RISCV I think.
> 
> ~Michal
> 

Thank you again for pointing this out! Unfortunately neither PPC nor 
RISC-V have a formal maintainer defined in the MAINTAINERS file, so it's 
actually "THE REST" responsible(s), which is also you ;-) For RISC-V I 
have asked internally to do a review, but this will only add a 
'Reviewed-By' not a full 'Ack'. So, not sure what to do about it.


--
Julian Vetter | Vates Hypervisor & Kernel Developer

XCP-ng & Xen Orchestra - Vates solutions

web: https://vates.tech

 


Rackspace

Lists.xenproject.org is hosted with RackSpace, monitoring our
servers 24x7x365 and backed by RackSpace's Fanatical Support®.