[Intel-gfx] [PATCH] drm/i915/gtt: Replace struct_mutex serialisation for allocation

[Joonas Lahtinen]

Quoting Chris Wilson (2019-06-03 20:11:30)
> Instead of relying on the caller holding struct_mutex across the
> allocation, push the allocation under a tree of spinlocks stored inside
> the page tables. Not only should this allow us to avoid struct_mutex
> here, but it will allow multiple users to lock independent ranges for
> concurrent allocations, and operate independently. This is vital for
> pushing the GTT manipulation into a background thread where dependency
> on struct_mutex is verboten, and for allowing other callers to avoid
> struct_mutex altogether.
> 
> Signed-off-by: Chris Wilson <chris at chris-wilson.co.uk>
> Cc: Matthew Auld <matthew.auld at intel.com>
> Cc: Mika Kuoppala <mika.kuoppala at intel.com>

<SNIP>

> @@ -1684,9 +1752,7 @@ static void gen6_ppgtt_clear_range(struct i915_address_space *vm,
>  
>                 num_entries -= count;
>  
> -               GEM_BUG_ON(count > pt->used_ptes);

This seems to be lost, and it's definitely a valid check, still.

With that retained, this is:

Reviewed-by: Joonas Lahtinen <joonas.lahtinen at linux.intel.com>

Operations *_ppgtt_set_* + atomic_inc(used_*) and *_ppgtt_set_*(scratch) +
atomic_dec() appear repetitive, but as they're for each different level,
a helper might or might not make it cleaner.

Regards, Joonas