Re: [RFC PATCH 10/10] vfio/type1: Register device notifier

From: Jason Gunthorpe <jgg@nvidia.com>
To: Peter Xu <peterx@redhat.com>
Cc: Alex Williamson <alex.williamson@redhat.com>, <cohuck@redhat.com>,
	<kvm@vger.kernel.org>, <linux-kernel@vger.kernel.org>
Subject: Re: [RFC PATCH 10/10] vfio/type1: Register device notifier
Date: Thu, 25 Feb 2021 14:19:45 -0400	[thread overview]
Message-ID: <20210225181945.GT4247@nvidia.com> (raw)
In-Reply-To: <20210225175457.GD250483@xz-x1>

On Thu, Feb 25, 2021 at 12:54:57PM -0500, Peter Xu wrote:

> I can't say I fully understand the whole rational behind 5cbf3264bc71, but that
> commit still sounds reasonable to me, since I don't see why VFIO cannot do
> VFIO_IOMMU_MAP_DMA upon another memory range that's neither anonymous memory
> nor vfio mapped MMIO range.

It is not so much it can't, more that it doesn't and doesn't need to.

> In those cases, vm_pgoff namespace defined by vfio may not be true
> anymore, iiuc.

Since this series is proposing linking the VMA to an address_space all
the vm_pgoffs must be in the same namespace

> Or does it mean that we don't want to allow VFIO dma to those unknown memory
> backends, for some reason?

Correct. VFIO can map into the IOMMU PFNs it can get a reference
to. pin_user_pages() works for the majority, special VFIO VMAs cover
the rest, and everthing else must be blocked for security.

Jason