Quoting Mika Kuoppala (2019-06-11 18:27:30)
> Swapping a pd entry is same across the page directories, if
> we succeed we need to increment the count and write the phys page
> entry. Make a common function for it.
>
> Signed-off-by: Mika Kuoppala <mika.kuoppala@xxxxxxxxxxxxxxx>
> ---
> drivers/gpu/drm/i915/i915_gem_gtt.c | 41 +++++++++++++++++++----------
> 1 file changed, 27 insertions(+), 14 deletions(-)
>
> diff --git a/drivers/gpu/drm/i915/i915_gem_gtt.c b/drivers/gpu/drm/i915/i915_gem_gtt.c
> index f1d7874834e2..9b0d0a077e31 100644
> --- a/drivers/gpu/drm/i915/i915_gem_gtt.c
> +++ b/drivers/gpu/drm/i915/i915_gem_gtt.c
> @@ -752,6 +752,27 @@ static void __set_pd_entry(struct i915_page_directory * const pd,
> gen8_pde_encode(px_dma(to), I915_CACHE_LLC)); \
> })
>
> +static void *__swap_pd_entry(struct i915_page_directory * const pd,
> + const unsigned short pde,
> + void * const old_val,
> + void * const new_val,
> + const u64 encoded_entry)
Mark this as inline and pass in the encode function, the compiler should
do the rest.
> +{
> + void * const old = cmpxchg(&pd->entry[pde], old_val, new_val);
> +
> + if (likely(old == old_val)) {
> + atomic_inc(&pd->used);
Hmm, looking at this again, this would be safer if atomic_inc was before
the cmpxchg, with an atomic_dec on the fail path.
> + if (likely(pd_has_phys_page(pd)))
> + __set_pd_entry(pd, pde, encoded_entry);
> + }
> +
> + return old;
> +}
-Chris
_______________________________________________
Intel-gfx mailing list
Intel-gfx@xxxxxxxxxxxxxxxxxxxxx
https://lists.freedesktop.org/mailman/listinfo/intel-gfx
