Re: [PATCH v3 1/3] xen/ns16550: Make ns16550 driver usable on ARM with HAS_PCI enabled.

[Jan Beulich <jbeulich@xxxxxxxx>]

On 24.11.2020 11:22, Julien Grall wrote:
> On 24/11/2020 09:47, Jan Beulich wrote:
>> On 23.11.2020 18:13, Julien Grall wrote:
>>> My view on 2) can change if Jan provides enough information into why one
>>> would want NS1650 PCI enabled by default on Arm but disable MSI.
>>
>> Because, like it was on x86, initially there may be no support for
>> MSI?
> 
> "no support for MSI" implies that there will be at least support for 
> NS16550 PCI.
> 
>>  I have no idea what the plans are ...
> 
> There are no such plan on Arm for the forseeable future (read as we 
> haven't seen any interested from the Arm community).

Okay, so you're question wasn't so much about the "but" in there,
but the "PCI" in the first place.

> The NS16550 PCI code will stay unusable until someone effectively send a 
> patch to plumb it correctly.
> 
> While I agree that disabling MSI may be nice to have to do in the 
> future, this doesn't address the need for Arm. I don't want to get in 
> our way the NS16550 PCI code in our way when implementing PCI (with or 
> without MSI) on Arm.
> 
> Even if there were an interest, I would still expect some users (e.g. 
> embedded folks) to want to compile-out unused feature (you may have a 
> platform with a embedded NS16550).
> 
> So the path forward will stay either 1) or 3) for me.

Well, as said elsewhere - 3) it is then afaic, for making more
obvious that this is really a hack.

Jan