Re: [PATCH v8 02/12] iommu: Introduce a replace API for device pasid

[Nicolin Chen <nicolinc@nvidia.com>]

On Fri, Feb 28, 2025 at 10:12:32PM +0800, Yi Liu wrote:
> On 2025/2/28 00:04, Nicolin Chen wrote:
> > On Thu, Feb 27, 2025 at 09:43:01AM +0800, Yi Liu wrote:
> > > On 2025/2/27 07:11, Nicolin Chen wrote:
> > > > > +	/* Not a replace case */
> > > > > +	if (!curr) {
> > > > 
> > > > Mind elaborating this?
> > > 
> > > If curr is null, this is more an attach op.
> > 
> > Does it mean that this function doesn't allow to replace a domain
> > that has no handle? But how can that be an attach op? I mean there
> > still could be a use case of replacing a domain (no handle) with
> > another domain (no handle or has handle)?
> 
> I think you misunderstood me here. The core stores either domain or handle
> now after the handle series. So curr==NULL means no prior attached
> domain(either with or w/o handle). That's why I said it is more like an
> attach and should use attach op.
> 
> replace a domain w/o handle to another domain with handle is supported.
> vice versa.
> 
> But replace a domain w/o handle to another domain w/o handle is not
> supported as I didn't add it. But it can be supported if there is
> use case.

OK. Do you mind sparing a few more words in the comment line,
to clarify that !curr means the domain has not been attached
to anything so it shouldn't be a replace case?

Thanks
Nicolin