[Intel-gfx] [PATCH v3 5/8] drm/i915/dg1: Reserve first 1MB of local memory

[Chris Wilson]

Quoting Chris Wilson (2021-01-27 12:19:38)
> Quoting Matthew Auld (2021-01-27 12:03:13)
> > From: Imre Deak <imre.deak at intel.com>
> > 
> > On DG1 A0/B0 steppings the first 1MB of local memory must be reserved.
> > One reason for this is that the 0xA0000-0xB0000 range is not accessible
> > by the display, probably since this region is redirected to another
> > memory location for legacy VGA compatibility.
> > 
> > BSpec: 50586
> > Testcase: igt/kms_big_fb/linear-64bpp-rotate-0
> > 
> > v2:
> > - Reserve the memory on B0 as well.
> > 
> > v3: replace DRM_DEBUG/DRM_ERROR with drm_dbg/drm_err
> > 
> > Signed-off-by: Imre Deak <imre.deak at intel.com>
> > Signed-off-by: Matthew Auld <matthew.auld at intel.com>
> > ---
> >  drivers/gpu/drm/i915/gt/intel_region_lmem.c | 56 +++++++++++++++++++++
> >  1 file changed, 56 insertions(+)
> > 
> > diff --git a/drivers/gpu/drm/i915/gt/intel_region_lmem.c b/drivers/gpu/drm/i915/gt/intel_region_lmem.c
> > index 71bb38706dbf..f5c12cbbaa86 100644
> > --- a/drivers/gpu/drm/i915/gt/intel_region_lmem.c
> > +++ b/drivers/gpu/drm/i915/gt/intel_region_lmem.c
> > @@ -143,6 +143,52 @@ intel_gt_setup_fake_lmem(struct intel_gt *gt)
> >         return mem;
> >  }
> >  
> > +static bool get_legacy_lowmem_region(struct intel_uncore *uncore,
> > +                                    u64 *start, u32 *size)
> > +{
> > +       *start = 0;
> > +       *size = 0;
> 
> Redundant now with the return indicating not to trust the values.
> 
> > +
> > +       if (!IS_DG1_REVID(uncore->i915, DG1_REVID_A0, DG1_REVID_B0))
> > +               return false;
> > +
> > +       *size = SZ_1M;
> > +
> > +       drm_dbg(&uncore->i915->drm, "LMEM: reserved legacy low-memory [0x%llx-0x%llx]\n",
> > +               *start, *start + *size);
> > +
> > +       return true;
> > +}
> > +
> > +static int reserve_lowmem_region(struct intel_uncore *uncore,
> > +                                struct intel_memory_region *mem)
> > +{
> > +       u64 reserve_start;
> > +       u64 reserve_end;
> > +       u64 region_start;
> > +       u32 region_size;
> > +       int ret;
> > +
> > +       if (!get_legacy_lowmem_region(uncore, &region_start, &region_size))
> > +               return 0;
> > +
> > +       reserve_start = region_start;
> > +       reserve_end = region_start + region_size;
> > +
> > +       if (!reserve_end)
> > +               return 0;
> > +
> > +       drm_dbg(&uncore->i915->drm, "LMEM: reserving low-memory region [0x%llx-0x%llx]\n",
> > +               reserve_start, reserve_end);
> > +       ret = intel_memory_region_reserve(mem,
> > +                                         reserve_start,
> > +                                         reserve_end - reserve_start);
> 
> You are doing this on purpose!
> u32 start, u64 size -> u64 start, end and then back to u64 start, size
> 
> The two drm_dbg() are functionally identical (other than giving a
> telltale for the reserve_end overflow escape).

So other than this patch, the series is
Reviewed-by: Chris Wilson <chris at chris-wilson.co.uk>

I've a question for Tvrtko about how we want the driver_probe callchain
to look, but that's a minor nit.
-Chris