Re: [PATCH v8 2/5] xen/arm: Enable the existing x86 virtual PCI support for ARM [and 1 more messages]

[Jan Beulich <jbeulich@xxxxxxxx>]

On 18.10.2021 12:38, Ian Jackson wrote:
> Julien Grall writes ("Re: [PATCH v8 2/5] xen/arm: Enable the existing x86 
> virtual PCI support for ARM"):
>> AFAICT, the code is not reachable on Arm (?). Therefore, one could argue 
>> we this can wait after the week-end as this is a latent bug. Yet, I am 
>> not really comfortable to see knowningly buggy code merged.
> 
> I agree that merging something that is known to be wrong would be
> quite irregular, at least without a compelling reason.
> 
> Jan Beulich writes ("Re: [PATCH v8 2/5] xen/arm: Enable the existing x86 
> virtual PCI support for ARM"):
>> On 16.10.2021 12:28, Roger Pau Monné wrote:
>>> Maybe I'm being pedantic, or there was some communication outside the
>>> mailing list, but I think strictly speaking you are missing an Ack
>>> from either Jan or Paul for the xen/drivers/passthrough/pci.c change.
>>>
>>> IMHO seeing how that chunk moved from 3 different places in just one
>>> afternoon also doesn't give me a lot of confidence. It's Arm only code
>>> at the end, so it's not going to effect the existing x86 support and
>>> I'm not specially worried, but I would like to avoid having to move it
>>> again.
>>
>> +1
>>
>> I'll be replying to the patch itself for the technical aspects. As per
>> context still visible above this code path is supposedly unreachable
>> right now, which makes me wonder even more: Why the rush? Depending on
>> the answer plus considering the __hwdom_init issue, Ian, I'm inclined
>> to suggest a revert.
> 
> I don't want to be waving hammers about at this stage, and I haven't
> looked at the technical details myself, but:
> 
> Can I ask the ARM folks to make sure that this situation is sorted out
> ASAP ?  Say, by the end of Thursday ?
> 
> By sorted out I mean that the __init_hwdom issue is fixed and that
> the overall changes to xen/drivers/passthrough/pci.c have been
> properly approved.
> 
> Furthermore, I think these followup patches should go in all in one
> go, as a small series, when everyone is OK with it, rather than
> dribbling in.  That will make it easier to see the wood for the trees
> (and it would also make a revert less complicated).
> 
> Jan, are you OK with this approach ?

Yes. FTR I'm not fussed about "all in one go" vs "dribbling in".

Jan