Re: [PATCH v1 3/8] iommu: Extend iommu_at[de]tach_device() for multi-device groups

From: Jason Gunthorpe via iommu <iommu@lists.linux-foundation.org>
To: Joerg Roedel <joro@8bytes.org>
Cc: Stuart Yoder <stuyoder@gmail.com>,
	rafael@kernel.org, David Airlie <airlied@linux.ie>,
	linux-pci@vger.kernel.org,
	Thierry Reding <thierry.reding@gmail.com>,
	Diana Craciun <diana.craciun@oss.nxp.com>,
	Dmitry Osipenko <digetx@gmail.com>, Will Deacon <will@kernel.org>,
	Ashok Raj <ashok.raj@intel.com>,
	Jonathan Hunter <jonathanh@nvidia.com>,
	Christoph Hellwig <hch@infradead.org>,
	Kevin Tian <kevin.tian@intel.com>,
	Chaitanya Kulkarni <kch@nvidia.com>,
	Alex Williamson <alex.williamson@redhat.com>,
	kvm@vger.kernel.org, Bjorn Helgaas <bhelgaas@google.com>,
	Dan Williams <dan.j.williams@intel.com>,
	Greg Kroah-Hartman <gregkh@linuxfoundation.org>,
	Cornelia Huck <cohuck@redhat.com>,
	linux-kernel@vger.kernel.org, Li Yang <leoyang.li@nxp.com>,
	iommu@lists.linux-foundation.org,
	Jacob jun Pan <jacob.jun.pan@intel.com>,
	Daniel Vetter <daniel@ffwll.ch>,
	Robin Murphy <robin.murphy@arm.com>
Subject: Re: [PATCH v1 3/8] iommu: Extend iommu_at[de]tach_device() for multi-device groups
Date: Mon, 14 Feb 2022 09:03:13 -0400	[thread overview]
Message-ID: <20220214130313.GV4160@nvidia.com> (raw)
In-Reply-To: <Ygo/eCRFnraY01WA@8bytes.org>

On Mon, Feb 14, 2022 at 12:39:36PM +0100, Joerg Roedel wrote:

> This extends iommu_attach_device() to behave as iommu_attach_group(),
> changing the domain for the whole group. 

Of course, the only action to take is to change the domain of a
group..

> Wouldn't it be better to scrap the iommu_attach_device() interface
> instead and only rely on iommu_attach_group()? This way it is clear
> that a call changes the whole group.

From an API design perspective drivers should never touch groups -
they have struct devices, they should have a clean struct device based
API.

Groups should disappear into an internal implementation detail, not be
so prominent in the API.

> IIUC this work is heading towards allowing multiple domains in one group
> as long as the group is owned by one entity.

No, it isn't. This work is only about properly arbitrating which
single domain is attached to an entire group.

> 	1) Introduce a concept of a sub-group (or whatever we want to
> 	   call it), which groups devices together which must be in the
> 	   same domain because they use the same request ID and thus
> 	   look all the same to the IOMMU.
>
> 	2) Keep todays IOMMU groups to group devices together which can
> 	   bypass the IOMMU when talking to each other, like
> 	   multi-function devices and devices behind a no-ACS bridge.

We've talked about all these details before and nobody has thought
they are important enough to implement. This distinction is not the
goal of this series.

I think if someone did want to do this there is room in the API to
allow the distinction between 1 (must share) and 2 (sharing is
insecure). eg by checking owner and blocking mixing user/kernel. 

This is another reason to stick with the device centric API as if we
did someday want multi-domain groups then the device input is still
the correct input and the iommu code can figure out what sub-groups or
whatever transparently.

Jason
_______________________________________________
iommu mailing list
iommu@lists.linux-foundation.org
https://lists.linuxfoundation.org/mailman/listinfo/iommu