Re: Programming PASID in IMS entries

[Jason Gunthorpe]

On Wed, Jul 07, 2021 at 10:50:52AM +0200, Thomas Gleixner wrote:
> Megha,
> 
> On Wed, Jul 07 2021 at 09:49, Megha Dey wrote:
> > Per your suggestions during the last meeting, we wanted to confirm the 
> > sequence to program the PASID into the IMS entries:
> >
> > 1. Add a PASID member to struct msi_desc (Add as part of a union. Other 
> > source-id's such as Jason's vm-id can be added to it)
> 
> Yes. Though we also discussed storing the default PASID in struct device
> to begin with which is then copied to the msi_desc entries during
> allocation.

This seems like a pretty good idea, though it requires that the
device's IRQ code cast the struct device to some driver specific
subtype, like mdev in this case.

> > 2. Create an API which device drivers can call, to program the PASID 
> > (PASID provided by the driver) on a per-irq basis. This API is to be 
> > called after msi_domain_alloc_irqs and will write to the corresponding 
> > msi_desc->pasid entry. (Assumption: For now, all devices will have the 
> > same IMS format). for e.g:
> >
> > msi_desc_set_pasid (irq, pasid) {
> >
> > struct msi_desc *desc = irq_get_msi_desc(irq);
> >
> > desc->pasid = pasid;
> >
> > }
> 
> That interface should be opaque probably with an u64 argument so it can
> be reused for Jason's VM-id. Jason?

Well, I certainly wouldn't put any IDXD specific words like PASID in
the general API. The comingling of PASID with the rest of the IRQ
registers is entirely a device specific choice.

Most likely something like mlx5 is going to want to associate a
pointer with the irq, and I believe it could use a struct device just
fine.

Jason