Re: [PATCH v3 0/2] kvm: level irqfd and new eoifd

[Jan Kiszka <jan.kiszka@siemens.com>]

On 2012-07-11 12:49, Avi Kivity wrote:
> On 07/11/2012 01:18 PM, Jan Kiszka wrote:
>> On 2012-07-11 11:53, Avi Kivity wrote:
>>> On 07/03/2012 10:21 PM, Alex Williamson wrote:
>>>> Here's the latest iteration of adding an interface to assert and
>>>> de-assert level interrupts from external drivers like vfio.  These
>>>> apply on top of the previous argument cleanup, documentation, and
>>>> sanitization patches for irqfd.  It would be great to get this queued
>>>> in next for linux 3.6.
>>>>
>>>> I believe I've addressed all the previous comments, including fixing
>>>> the locking problems in eoifd.  I've run this with lockdep adding
>>>> and removing level irqfd/eoifd pairs without any problems.  Please
>>>> let me know if there are any further comments.  Thanks,
>>>
>>> Is there any performance justification for level irqfd?  Don't all
>>> new/high bandwidth devices support msi, and this is just a legacy path?
>>
>> I think we've been there before, but the situation hasn't improved much:
>>
>> Apparently, some GPUs still prefer INTx over MSI. Some wireless chipsets
>> too. 
> 
> I'd appreciate a couple of examples for formality's sake.

>From the top of my head: NVIDIA FX3700 (granted, legacy by now), Atheros
AR9287. For others, I need to check.

> 
>> And then there is not easily replaceable legacy hardware like old
>> telephony adapters or industrial I/O cards etc. that want this.
> 
> I imagine legacy hardware will live with the speed of routing through
> qemu, when running on modern platforms.

Just because it's "legacy" doesn't mean it's "low performance" and "low
interrupt rate".

We still have classic KVM device assignment to provide fast-path INTx.
But if we want to replace it midterm, I think it's necessary for VFIO to
be able to provide such a path as well.

Jan

-- 
Siemens AG, Corporate Technology, CT RTC ITP SDP-DE
Corporate Competence Center Embedded Linux