On Tue, Apr 24, 2018 at 11:43:35PM -0700, Christoph Hellwig wrote:
> On Wed, Apr 25, 2018 at 08:23:15AM +0200, Daniel Vetter wrote:
> > For more fun:
> > 
> > https://www.spinics.net/lists/dri-devel/msg173630.html
> > 
> > Yeah, sometimes we want to disable the iommu because the on-gpu
> > pagetables are faster ...
> 
> I am not on this list, but remote NAK from here.  This needs an
> API from the iommu/dma-mapping code.  Drivers have no business poking
> into these details.

The interfaces that the above patch uses are all EXPORT_SYMBOL_GPL,
which is rather misleading if they are not meant to be used by drivers
directly.

> Thierry, please resend this with at least the iommu list and
> linux-arm-kernel in Cc to have a proper discussion on the right API.

I'm certainly open to help with finding a correct solution, but the
patch above was purposefully terse because this is something that I
hope we can get backported to v4.16 to unbreak Nouveau. Coordinating
such a backport between ARM and DRM trees does not sound like something
that would help getting this fixed in v4.16.

The fundamental issue here is that the DMA/IOMMU integration is
something that has caused a number of surprising regressions in the past
because it tends to sneak in unexpectedly. For example the current
regression shows up only if CONFIG_ARM_DMA_USE_IOMMU=y because the DMA
API will then transparently create a second mapping and mess things up.
Everything works fine if that option is disabled. This is ultimately why
we didn't notice, since we don't enable that option by default. I do
have a patch that I plan to apply to the Tegra tree that will always
enable CONFIG_ARM_DMA_USE_IOMMU=y on Tegra to avoid any such surprises
in the future, but I can obviously only apply that once the above patch
is applied to Nouveau, otherwise we'll break Nouveau unconditionally.

Granted, this issue could've been caught with a little more testing, but
in retrospect I think it would've been a lot better if ARM_DMA_USE_IOMMU
was just enabled unconditionally if it has side-effects that platforms
don't opt in to but have to explicitly opt out of.

Thierry