Re: [patch 21/32] NTB/msi: Convert to msi_on_each_desc()

[Thomas Gleixner]

dave,

On Wed, Dec 01 2021 at 15:53, Dave Jiang wrote:
> On 12/1/2021 3:03 PM, Thomas Gleixner wrote:
>> This still depends on how this overall discussion about representation
>> of all of this stuff is resolved.
>>
>>>> What needs a subdevice to expose?
>> Can you answer that too please?
>
> Sorry. So initial version of the IDXD sub-device is represented with a 
> single queue. It needs a command irq (emulated) and a completion irq 
> that is backed by a device vector (currently IMS).

thanks for clarification! Let me think about it some more.

       tglx