On 20.10.2025 11:07, Roger Pau Monné wrote:
> On Mon, Oct 20, 2025 at 09:46:42AM +0200, Jan Beulich wrote:
>> On 17.10.2025 16:14, Roger Pau Monne wrote:
>>> Otherwise it's not possible for device models to map IRQs of devices on
>>> segments different than 0. Keep the same function prototype and pass the
>>> segment in the high 16bits of the bus parameter, like it's done for the
>>> hypercall itself.
>>
>> While easiest, that's an odd interface, though. Should, at the very least the
>> function parameter then be named e.g. "segbus", along the lines of "devfn"?
>
> I certainly don't mind using segbus instead of plain bus, will adjust
> now.
>
>>> Fixes: 7620c0cf9a4d ("PCI multi-seg: add new physdevop-s")
>>
>> This commit wasn't about tool stack uses of the interfaces at all.
>
> But there should have been a tools side change somewhere to make use
> of that interface, at the point that support for multi-segment was
> added to Xen? Otherwise the support feels like half done.
Perhaps. I was after kernel functionality only at that point, and I
would really have liked toolstack folks to have picked up any work
needed there.
> Would you prefer me to use the "Amends:" tag? Or no tag at all.
Both would be fine with me.
Jan
