[dpdk-dev] VFIO no-iommu

[Alex Williamson]

On Mon, 2015-12-21 at 11:46 +0000, Yigit, Ferruh wrote:
> On Fri, Dec 18, 2015 at 02:50:17PM -0700, Alex Williamson wrote:
> > On Fri, 2015-12-18 at 07:38 -0700, Alex Williamson wrote:
> > > On Fri, 2015-12-18 at 10:43 +0000, Yigit, Ferruh wrote:
> > > > On Thu, Dec 17, 2015 at 09:43:59AM -0700, Alex Williamson
> > > > wrote:
> > > > <...>
> > > > > > > > > > 
> > > > > > > > > > Also I need to disable VFIO_CHECK_EXTENSION ioctl,
> > > > > > > > > > because in
> > > > > > > > > > vfio
> > > > > > > > > > module,
> > > > > > > > > > container->noiommu is not set before doing a
> > > > > > > > > > vfio_group_set_container()
> > > > > > > > > > and vfio_for_each_iommu_driver selects wrong
> > > > > > > > > > driver.
> > > > > > > > > 
> > > > > > > > > Running CHECK_EXTENSION on a container without the
> > > > > > > > > group
> > > > > > > > > attached is
> > > > > > > > > only going to tell you what extensions vfio is
> > > > > > > > > capable
> > > > > > > > > of,
> > > > > > > > > not
> > > > > > > > > necessarily what extensions are available to you with
> > > > > > > > > that
> > > > > > > > > group.
> > > > > > > > > Is this just a general dpdk- vfio ordering bug?
> > > > > > > > 
> > > > > > > > Yes, that is how VFIO was implemented in DPDK. I was
> > > > > > > > under
> > > > > > > > the
> > > > > > > > impression that checking extension before assigning
> > > > > > > > devices
> > > > > > > > was
> > > > > > > > the
> > > > > > > > correct way to do things, so as to not to try anything
> > > > > > > > we
> > > > > > > > know
> > > > > > > > would
> > > > > > > > fail anyway. Does this imply that CHECK_EXTENSION needs
> > > > > > > > to
> > > > > > > > be
> > > > > > > > called
> > > > > > > > on both container and groups (or just on groups)?
> > > > > > > 
> > > > > > > Hmm, in Documentation/vfio.txt we do give the following
> > > > > > > algorithm:
> > > > > > > 
> > > > > > >         if (ioctl(container, VFIO_GET_API_VERSION) !=
> > > > > > > VFIO_API_VERSION)
> > > > > > >                 /* Unknown API version */
> > > > > > > 
> > > > > > >         if (!ioctl(container, VFIO_CHECK_EXTENSION,
> > > > > > > VFIO_TYPE1_IOMMU))
> > > > > > >                 /* Doesn't support the IOMMU driver we
> > > > > > > want.
> > > > > > > */
> > > > > > >         ...
> > > > > > > 
> > > > > > > That's just going to query each iommu driver and we can't
> > > > > > > yet
> > > > > > > say
> > > > > > > whether
> > > > > > > the group the user attaches to the container later will
> > > > > > > actually
> > > > > > > support that
> > > > > > > extension until we try to do it, that would come at
> > > > > > > VFIO_SET_IOMMU.
> > > > > > >  So is
> > > > > > > it perhaps a vfio bug that we're not advertising no-iommu
> > > > > > > until
> > > > > > > the
> > > > > > > group is
> > > > > > > attached?  After all, we are capable of it with just an
> > > > > > > empty
> > > > > > > container, just
> > > > > > > like we are with type1, but we're going to fail SET_IOMMU
> > > > > > > for
> > > > > > > the
> > > > > > > wrong
> > > > > > > combination.
> > > > > > >  This is exactly the sort of thing that makes me glad we
> > > > > > > reverted
> > > > > > > it without
> > > > > > > feedback from a working user driver.  Thanks,
> > > > > > 
> > > > > > Whether it should be considered a "bug" in VFIO or "by
> > > > > > design"
> > > > > > is
> > > > > > up
> > > > > > to you, of course, but at least according to the VFIO
> > > > > > documentation,
> > > > > > we are meant to check for type 1 extension and then attach
> > > > > > devices,
> > > > > > so it would be expected to get VFIO_NOIOMMU_IOMMU marked as
> > > > > > supported
> > > > > > even without any devices attached to the container (just
> > > > > > like
> > > > > > we
> > > > > > get
> > > > > > type 1 as supported without any devices attached). Having
> > > > > > said
> > > > > > that,
> > > > > > if it was meant to attach devices first and then check the
> > > > > > extensions, then perhaps the documentation should also
> > > > > > point
> > > > > > out
> > > > > > that
> > > > > > fact (or perhaps I missed that detail in my readings of the
> > > > > > docs,
> > > > > > in
> > > > > > which case my apologies).
> > > > > 
> > > > > Hi Anatoly,
> > > > > 
> > > > > Does the below patch make it behave more like you'd expect.
> > > > >  This
> > > > > applies to v4.4-rc4, I'd fold this into the base patch if we
> > > > > reincorporate it to a future kernel.  Thanks,
> > > > > 
> > > > > Alex
> > > > > 
> > > > > commit 88d4dcb6b77624965f0b45b5cd305a2b4a105c94
> > > > > Author: Alex Williamson <alex.williamson at redhat.com>
> > > > > Date:   Wed Dec 16 19:02:01 2015 -0700
> > > > > 
> > > > >     vfio: Fix no-iommu CHECK_EXTENSION
> > > > >     
> > > > >     Previously the no-iommu iommu driver was only visible
> > > > > when
> > > > > the
> > > > >     container had an attached no-iommu group.  This means
> > > > > that
> > > > >     CHECK_EXTENSION on and empty container couldn't report
> > > > > the
> > > > > possibility
> > > > >     of using VFIO_NOIOMMU_IOMMU.  We report TYPE1 whether or
> > > > > not
> > > > > the user
> > > > >     can make use of it with the group, so this is
> > > > > inconsistent.  Add the
> > > > >     no-iommu iommu to the list of iommu drivers when enabled
> > > > > via
> > > > > module
> > > > >     option, but skip all the others if the container is
> > > > > attached
> > > > > to
> > > > > a
> > > > >     no-iommu groups.  Note that tainting is now done with the
> > > > > "unsafe"
> > > > >     module callback rather than explictly within vfio.
> > > > >     
> > > > >     Also fixes module option and module description name
> > > > > inconsistency.
> > > > >     
> > > > >     Also make vfio_noiommu_ops const.
> > > > >     
> > > > >     Signed-off-by: Alex Williamson <alex.williamson at redhat.co
> > > > > m>
> > > > 
> > > > Hi Alex,
> > > > 
> > > > I got following crash with this update:
> > 
> > Let's try this one:
> > 
> > commit 8ff839c6ffe9f3b50b50f1cc87e7afbf23171f05
> > Author: Alex Williamson <alex.williamson at redhat.com>
> > Date:   Fri Dec 18 14:45:55 2015 -0700
> > 
> >     v2 vfio fix no-iommu CHECK_EXTENSION
> >     
> >     Register and unregister the no-iommu iommu backend at module
> >     initialization and exit, but disable it unless enabled via
> > module
> >     option.  Rather than modify the iommu driver walk, selectively
> > skip
> >     combinations that aren't supported.  CHECK_EXTENSION on a
> > container
> >     without any groups attached exposes all possible
> > extensions.  Once a
> >     group is attached, the no-iommu backend is skipped for regular
> > groups
> >     and regular iommu backends are skipped for no-iommu groups.
> >     
> >     This would be folded into a single patch to re-propose vfio no-
> > iommu
> >     mode upstream.
> >     
> >     Signed-off-by: Alex Williamson <alex.williamson at redhat.com>
> > 
> Hi Alex,
> 
> Thank you for the update. I have tested this on both no-iommu and
> iommu environment
> and worked successfully. I believe this approach is better because it
> is simpler.
> 
> From DPDK point of view, only update to support vfio no-iommu is: to
> use new group names
> and disable DMA mapping.
> 
> If VFIO module compiled with "CONFIG_VFIO_NOIOMMU=y" by default, that
> makes things easier
> for DPDK, in no-iommu environment inserting vfio module with proper
> parameter makes it
> available for DPDK.

Thanks for the update Ferruh.  Also note that the module option is
dynamically settable so that it can support running with statically
compiled modules where you may not know whether or not to enable no-
iommu until after boot, or where unloading and re-loading a module
might not be an option.  It will be up to each distro to decide whether
to enable the config option, but I think we at least highlight the
existing support issue for non-iommu protected userspace drivers, which
is something that was not at all clear with the uio driver approach.
 Thanks,

Alex