Received-SPF: pass (google.com: best guess record for domain of linux-kernel-owner@vger.kernel.org designates 209.132.180.67 as permitted sender) client-ip=209.132.180.67;
Date:   Wed, 5 Feb 2020 06:58:36 -0700
From:   Alex Williamson <alex.williamson@redhat.com>
To:     Christoph Hellwig <hch@infradead.org>
Cc:     kvm@vger.kernel.org, linux-pci@vger.kernel.org,
        linux-kernel@vger.kernel.org, dev@dpdk.org, mtosatti@redhat.com,
        thomas@monjalon.net, bluca@debian.org, jerinjacobk@gmail.com,
        bruce.richardson@intel.com, cohuck@redhat.com
Subject: Re: [RFC PATCH 0/7] vfio/pci: SR-IOV support
Message-ID: <20200205065836.12308197@x1.home>
In-Reply-To: <20200205070109.GA18027@infradead.org>
References: <158085337582.9445.17682266437583505502.stgit@gimli.home>
        <20200205070109.GA18027@infradead.org>
Organization: Red Hat
MIME-Version: 1.0
Content-Type: text/plain; charset=US-ASCII
Content-Transfer-Encoding: 7bit
Sender: linux-kernel-owner@vger.kernel.org
Precedence: bulk

On Tue, 4 Feb 2020 23:01:09 -0800
Christoph Hellwig <hch@infradead.org> wrote:

> On Tue, Feb 04, 2020 at 04:05:34PM -0700, Alex Williamson wrote:
> > We address this in a few ways in this series.  First, we can use a bus
> > notifier and the driver_override facility to make sure VFs are bound
> > to the vfio-pci driver by default.  This should eliminate the chance
> > that a VF is accidentally bound and used by host drivers.  We don't
> > however remove the ability for a host admin to change this override.  
> 
> That is just such a bad idea.  Using VFs in the host is a perfectly
> valid use case that you are breaking.

vfio-pci currently does not allow binding to a PF with VFs enabled and
does not provide an sriov_configure callback, so it's not possible to
have VFs on a vfio-pci bound PF.  Therefore I'm not breaking any
existing use cases.  I'm also not preventing VFs from being used in the
host, I only set a default driver_override value, which can be replaced
if a different driver binding is desired.  So I also don't see that I'm
breaking a usage model here.  I do stand by the idea that VFs sourced
from a user owned PF should not by default be used in the host (ie.
autoprobed on device add).  There's a pci-pf-stub driver that can be
used to create VFs on a PF if no userspace access of the PF is required.
Thanks,

Alex