Re: [RFC] hw/nvme: Use irqfd to send interrupts

[Klaus Jensen <its@irrelevant.dk>]

[-- Attachment #1: Type: text/plain, Size: 1282 bytes --]

On Aug  2 12:03, Jinhao Fan wrote:
> at 6:21 PM, Stefan Hajnoczi <stefanha@gmail.com> wrote:
> 
> > What happens when the MSI-X vector is masked?
> > 
> > I remember the VIRTIO code having masking support. I'm on my phone and can't check now, but I think it registers a temporary eventfd and buffers irqs while the vector is masked.
> 
> Hi Stefan,
> 
> While implementing interrupt masking support, I found it hard to test this
> feature on the host. Keith told me that no NVMe drivers are currently using
> this feature. Do you remember how you tested interrupt masking?
> 

You can probably do this with qtest. I don't see a helper for masking
the vectors, but qpci_msix_masked() should be usable as a base for just
changing that readl to a writel and mask it out.

This should allow you to do a relatively simple test case where you

  1. bootstrap the device as simple as possible (forget about I/O
     queues) - I *think* you just need to use guest_alloc for the admin
     queue memory, use qpci_msix_enable() etc.
  2. setup a simple admin command in the queue
  3. mask the interrupt
  4. ring the doorbell (a writel)
  5. check that the vector remains in pending state
  (qpci_msix_pending()).


This *could* be a potential way to do this.

[-- Attachment #2: signature.asc --]
[-- Type: application/pgp-signature, Size: 488 bytes --]