[Lima] [PATCH 2/2] drm/lima: driver for ARM Mali4xx GPUs

[Alex Deucher]

I typed this up yesterday, but it looks like I never hit send.


On Tue, Feb 12, 2019 at 3:05 PM Rob Herring <robh at kernel.org> wrote:
>
> On Tue, Feb 12, 2019 at 10:24 AM Alex Deucher <alexdeucher at gmail.com> wrote:
> >
> > On Tue, Feb 12, 2019 at 10:53 AM Rob Herring via dri-devel
> > <dri-devel at lists.freedesktop.org> wrote:
> > >
> > > On Wed, Feb 6, 2019 at 7:16 AM Qiang Yu <yuq825 at gmail.com> wrote:
> > > >
> > > > From: Lima Project Developers <lima at lists.freedesktop.org>
>
> [...]
>
> > > > +static int lima_ioctl_gem_va(struct drm_device *dev, void *data, struct drm_file *file)
> > > > +{
> > > > +       struct drm_lima_gem_va *args = data;
> > > > +
> > > > +       switch (args->op) {
> > > > +       case LIMA_VA_OP_MAP:
> > > > +               return lima_gem_va_map(file, args->handle, args->flags, args->va);
> > > > +       case LIMA_VA_OP_UNMAP:
> > > > +               return lima_gem_va_unmap(file, args->handle, args->va);
> > >
> > > These are mapping to GPU VA. Why not do that on GEM object creation or
> > > import or when the objects are submitted with cmd queue as other
> > > drivers do?
> > >
> > > To put it another way, These ioctls look different than what other
> > > drivers do. Why do you need to do things differently? My understanding
> > > is best practice is to map and return the GPU offset when the GEM
> > > object is created. This is what v3d does. I think Intel is moving to
> > > that. And panfrost will do that.
> >
> > I think it would be a good idea to look at the amdgpu driver.  This
> > driver is heavily modeled after it.  Basically the GEM VA ioctl allows
> > userspace to manage per process (per fd really) virtual addresses.
>
> Why do you want userspace to manage assigning VAs versus the kernel to
> do so? Exposing that detail to userspace means the driver must support
> a per process address space. Letting the kernel assign addresses means
> it can either be a single address space or be a per process address
> space. It seems to me more flexible to allow the kernel driver to
> evolve without that ABI.

Having it in userspace provides a lot more flexibility and makes it
easier to support things like unified address space between CPU and
GPU. I guess it depends on the hw as to what is the right choice.

>
> With any new driver in the kernel, the question is always which
> existing one is the best model to follow. I don't think Intel, AMD or
> Nouveau are good examples to follow because they have a lot of history
> and legacy, are both GPU and DC, and have separate graphics memory
> (except Intel I guess). The GPUs in ARM land have none of those
> really. Looking thru freedreno, etnaviv, and v3d mostly, I see they
> all have similar user ABIs. But they are all different based on what
> driver they copied and how they've evolved. I know it's a big can of
> worms, but it would be nice to have some alignment of ABIs. I know the
> reasons why there isn't, but it's frustrating that 11 out of 60K IGT
> tests will run. I don't think a common ABI matters much for the big 3,
> but in the ARM zoo I think it does. At least if the interfaces are
> kept similar, then having common code shared among the embedded GPUs
> would be easier and writing some IGT shim for each driver would be
> easier.

It also depends on what you are familiar with.  Qiang is familiar with
amdgpu so it makes sense to use that as a model for his own projects.
I don't know that similarity between drivers for hw that is mostly
part of ARM SoCs is any more important than the applicability to the
hw or what the maintainer likes.  I mean you can use amdgpu or nouveau
on ARM too.

Alex