[Intel-gfx] [PATCH] drm/i915: don't conflate is_dgfx with fake lmem

[Matthew Auld]

On Wed, 30 Sep 2020 at 16:30, Lucas De Marchi <lucas.demarchi at intel.com> wrote:
>
> When using fake lmem for tests, we are overriding the setting in
> device info for dgfx devices. Current users of IS_DGFX() are correct,
> but as we add support for DG1, we are going to use it in additional
> places to trigger dgfx-only code path.
>
> In future if we need we can use HAS_LMEM() instead of IS_DGFX() in the
> places that make sense to also contemplate fake lmem use.
>
> Signed-off-by: Lucas De Marchi <lucas.demarchi at intel.com>
> ---
>  drivers/gpu/drm/i915/i915_drv.c | 2 --
>  1 file changed, 2 deletions(-)
>
> diff --git a/drivers/gpu/drm/i915/i915_drv.c b/drivers/gpu/drm/i915/i915_drv.c
> index 45e719c79183..3f6ed142198c 100644
> --- a/drivers/gpu/drm/i915/i915_drv.c
> +++ b/drivers/gpu/drm/i915/i915_drv.c
> @@ -840,9 +840,7 @@ int i915_driver_probe(struct pci_dev *pdev, const struct pci_device_id *ent)
>                     i915->params.fake_lmem_start) {
>                         mkwrite_device_info(i915)->memory_regions =
>                                 REGION_SMEM | REGION_LMEM | REGION_STOLEN;
> -                       mkwrite_device_info(i915)->is_dgfx = true;
>                         GEM_BUG_ON(!HAS_LMEM(i915));
> -                       GEM_BUG_ON(!IS_DGFX(i915));

Hmm, I guess this could work as an option c) but I think we will also
need to add something like:

@@ -835,7 +835,7 @@ static int gen8_gmch_probe(struct i915_ggtt *ggtt)
        u16 snb_gmch_ctl;

        /* TODO: We're not aware of mappable constraints on gen8 yet */
-       if (!IS_DGFX(i915)) {
+       if (!IS_DGFX(i915) && !HAS_LMEM(i915)) {
                ggtt->gmadr = pci_resource(pdev, 2);
                ggtt->mappable_end = resource_size(&ggtt->gmadr);
        }

The point is that we need to steal the mappable aperture later(which
is fine since it doesn't exist on "DGFX"), and use it as a substitute
for LMEMBAR. The !mappable aperture property is also useful since it
exercises some other parts of the code too.

Assuming CI is happy and this unbocks you then,
Reviewed-by: Matthew Auld <matthew.auld at intel.com>