[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]
Message-ID: <ZuFHGHd2S+r6XzyN@nvidia.com>
Date: Wed, 11 Sep 2024 00:30:32 -0700
From: Nicolin Chen <nicolinc@...dia.com>
To: "Tian, Kevin" <kevin.tian@...el.com>
CC: "jgg@...dia.com" <jgg@...dia.com>, "corbet@....net" <corbet@....net>,
"iommu@...ts.linux.dev" <iommu@...ts.linux.dev>, "linux-doc@...r.kernel.org"
<linux-doc@...r.kernel.org>, "linux-kernel@...r.kernel.org"
<linux-kernel@...r.kernel.org>, "Liu, Yi L" <yi.l.liu@...el.com>,
"eric.auger@...hat.com" <eric.auger@...hat.com>,
"shameerali.kolothum.thodi@...wei.com"
<shameerali.kolothum.thodi@...wei.com>, "smostafa@...gle.com"
<smostafa@...gle.com>, "baolu.lu@...ux.intel.com" <baolu.lu@...ux.intel.com>
Subject: Re: [PATCH] Documentation: userspace-api: iommufd: Update
HWPT_PAGING and HWPT_NESTED
On Wed, Sep 11, 2024 at 07:09:15AM +0000, Tian, Kevin wrote:
> > From: Nicolin Chen <nicolinc@...dia.com>
> > Sent: Wednesday, September 11, 2024 4:41 AM
> >
> > + feature flag. This can be either an UNMANAGED stage-1 domain for a
> > device
> > + running in the user space, or a nesting parent stage-2 domain for
> > mappings
> > + from guest-level physical addresses to host-level physical addresses.
>
> the former part is inaccurate. It could be an UNMANAGED stage-2 domain.
>
> > +
> > +- IOMMUFD_OBJ_HWPT_NESTED, representing an actual hardware I/O
> > page table
> > + (i.e. a single struct iommu_domain) managed by user space (e.g. guest OS).
> > + "NESTED" indicates that this type of HWPT can be linked to an
> > HWPT_PAGING.
>
> s/can be/should be/
Was thinking of the 2nd choice: HWPT_NESTED->vIOMMU (HWPT_PAGING)
Yet, I think "should" could fit that narrative too. Will change.
> >
> > -3. IOMMUFD_OBJ_HW_PAGETABLE is created when an external driver calls
> > the IOMMUFD
> > +3. IOMMUFD_OBJ_HWPT_PAGING can be created in two ways:
> > +
> > + IOMMUFD_OBJ_HWPT_PAGING is created when an external driver calls
> > the IOMMUFD
> > kAPI to attach a bound device to an IOAS. Similarly the external driver uAPI
> > allows userspace to initiate the attaching operation. If a compatible
> > pagetable already exists then it is reused for the attachment. Otherwise a
> > new pagetable object and iommu_domain is created. Successful
> > completion of
> > this operation sets up the linkages among IOAS, device and
> > iommu_domain. Once
> > - this completes the device could do DMA.
> > -
> > - Every iommu_domain inside the IOAS is also represented to userspace as
> > a
> > - HW_PAGETABLE object.
> > + this completes the device could do DMA. Note that every iommu_domain
> > inside
> > + the IOAS is also represented to userspace as an
> > IOMMUFD_OBJ_HWPT_PAGING.
>
> the last sentence is redundant. here we are talking about how HWPT_PAGING
> is created so it's implied. probably you can state that HWPT_PAGING object
> is created when talking about "otherwise a new page table object and
> iommu_domain is created".
I don't quite get this part...where is the redundancy? And where
is "otherwise a new page table object .."?
> > +
> > +4. IOMMUFD_OBJ_HWPT_NESTED can be only manually created via the
> > IOMMU_HWPT_ALLOC
> > + uAPI, provided an hwpt_id via @pt_id to associate the new
> > HWPT_NESTED object
> > + to the corresponding HWPT_PAGING object. The associating
> > HWPT_PAGING object
> > + must be a nesting parent manually allocated via the same uAPI previously
> > with
> > + an IOMMU_HWPT_ALLOC_NEST_PARENT flag, otherwise the allocation
> > will fail. The
> > + allocation will be further validated by the IOMMU driver of an IOMMU
> > hardware
> > + that the given device (via @dev_id) is physically linked to, to ensure that
> > + the nesting parent domain and a nested domain being allocated are
> > compatible.
>
> just "validated by the IOMMU driver to ensure that ..."
OK.
> >
> > .. note::
> >
> > - Future IOMMUFD updates will provide an API to create and manipulate
> > the
> > - HW_PAGETABLE directly.
> > + Either a manual IOMMUFD_OBJ_HWPT_PAGING or an
> > IOMMUFD_OBJ_HWPT_NESTED is
> > + created via the same IOMMU_HWPT_ALLOC uAPI. The difference is at
> > the type
> > + of the object passed in via the @pt_id field of struct
> > iommufd_hwpt_alloc:
> > + When @pt_id carries an ioas_id to an IOAS object, the
> > IOMMU_HWPT_ALLOC
> > + call is instructed to allocate an HWPT_PAGING object only.
> > + When @pt_id carries an hwpt_id to an HWPT_PAGING object, the uAPI
> > call
> > + is instructed to allocate an HWPT_NESTED object only.
> > + If any other type of object is passed in via the @pt_id, the uAPI call
> > + will fail.
> >
>
> I'm not sure whether this note is still required. probably just one
> sentence to highlight that it's @pt_id field to mark out the
> object type? most descriptions duplicate with the earlier words.
I feel it's a nice summary though... I'll see how I can simplify it.
Thanks
Nicolin
Powered by blists - more mailing lists