Re: [PATCH RFC PREVIEW 0/6] iommu: enable user space iopfs in non-nested and non-svm cases

From: Joel Granados
Date: Tue Sep 03 2024 - 09:20:34 EST


On Mon, Sep 02, 2024 at 08:47:21PM +0800, Baolu Lu wrote:
> On 2024/9/2 18:48, Joel Granados wrote:
> >> I definitely expect PRI to work outside PASID and SVA cases, so this
> >> is going in a good direction
> > This touches on a detail (at least in Intel's vtd-io spec) that is not
> > 100% clear to me. Second paragraph of section "3.4.3 Scalable Mode
> > Address Translation" reads:
> > "
> > ... Scalable-mode context-entries support both requests-without-PASID
> > and requests-with-PASID. However unlike legacy mode, in scalable-mode,
> > requests-without-PASID obtain a PASID value from the RID_PASID field of
> > the scalable-mode context- entry and are processed similarly to
> > requests-with-PASID.Implementations not supporting RID_PASID capability
> > (ECAP_REG.RPS is 0b), use a PASID value of 0 to perform address
> > translation for requests without PASID.
> > "
> > This basically means that a default PASID is used even though the
> > request is without PASID. Right? Therefore "outside PASID" means with
> > the default PASID (at least in Intels case). Right?
>
> Kind of yes.
>
> The PCI specification defines the concept of PASID and its role in
> transaction routing. We refer to PCI transactions with a PASID prefix as
> "request-with-PASID" and those without a PASID prefix as "request-
> without-PASID." Consequently, I understand 'outside PASID' to mean
> transactions that do not have a PASID prefix.
>
> The VT-d specification describes how the IOMMU hardware handles request-
> without-PASID. It uses a reserved PASID for its internal routing and
> handling purposes. If RID_PASID is supported (ECAP_REG.RPS=1), software
> can select its own reserved PASID. Otherwise, the IOMMU hardware will
> use a default value of 0.
>
Thx for getting back to me. This generates another doubt in my head
regarding the published capabilities from the intel IOMMU Hardware:

So ecap_pasid [1] does not have to be set in scalable-mode. Right? This
allows hardware supporting scalable-mode to reject transactions with
PASID whenever ecap_pasid is *NOT* set; even though internally things
are handled with a PASID. This question is directly related to the two
last patches in the set.5/6 and 6/6.

Best

[1] I use ecap_pasid like in the kernel `ecap_pasid(e) (e>>40) & 0x1`

--

Joel Granados