On 17.3.2022 13.55, Matthew Auld wrote:
On Wed, 16 Mar 2022 at 22:23, Juha-Pekka Heikkila
<juhapekka.heikkila@xxxxxxxxx> wrote:
Add fallback smem allocation for dpt if stolen memory
allocation failed.
Signed-off-by: Juha-Pekka Heikkila <juhapekka.heikkila@xxxxxxxxx>
---
drivers/gpu/drm/i915/display/intel_dpt.c | 18 ++++++++++++++----
1 file changed, 14 insertions(+), 4 deletions(-)
diff --git a/drivers/gpu/drm/i915/display/intel_dpt.c b/drivers/gpu/drm/i915/display/intel_dpt.c
index fb0e7e79e0cd..c8b66433d4db 100644
--- a/drivers/gpu/drm/i915/display/intel_dpt.c
+++ b/drivers/gpu/drm/i915/display/intel_dpt.c
@@ -10,6 +10,7 @@
#include "intel_display_types.h"
#include "intel_dpt.h"
#include "intel_fb.h"
+#include "gem/i915_gem_internal.h"
Nit: these should be kept sorted
struct i915_dpt {
struct i915_address_space vm;
@@ -128,6 +129,10 @@ struct i915_vma *intel_dpt_pin(struct i915_address_space *vm)
void __iomem *iomem;
struct i915_gem_ww_ctx ww;
int err;
+ u64 pin_flags = 0;
+
+ if (i915_gem_object_is_stolen(dpt->obj))
+ pin_flags |= PIN_MAPPABLE; /* for i915_vma_pin_iomap(stolen) */
wakeref = intel_runtime_pm_get(&i915->runtime_pm);
atomic_inc(&i915->gpu_error.pending_fb_pin);
@@ -138,7 +143,7 @@ struct i915_vma *intel_dpt_pin(struct i915_address_space *vm)
continue;
vma = i915_gem_object_ggtt_pin_ww(dpt->obj, &ww, NULL, 0, 4096,
- HAS_LMEM(i915) ? 0 : PIN_MAPPABLE);
+ pin_flags);
if (IS_ERR(vma)) {
err = PTR_ERR(vma);
continue;
@@ -248,10 +253,15 @@ intel_dpt_create(struct intel_framebuffer *fb)
size = round_up(size * sizeof(gen8_pte_t), I915_GTT_PAGE_SIZE);
- if (HAS_LMEM(i915))
- dpt_obj = i915_gem_object_create_lmem(i915, size, I915_BO_ALLOC_CONTIGUOUS);
- else
+ dpt_obj = i915_gem_object_create_lmem(i915, size, I915_BO_ALLOC_CONTIGUOUS);
+ if (IS_ERR(dpt_obj) && i915_ggtt_has_aperture(to_gt(i915)->ggtt))
dpt_obj = i915_gem_object_create_stolen(i915, size);
+ if (IS_ERR(dpt_obj) && !HAS_LMEM(i915)) {
+ drm_dbg_kms(&i915->drm, "fb: [FB:%d] Allocating dpt from smem\n",
+ fb->base.base.id);
+
+ dpt_obj = i915_gem_object_create_internal(i915, size);
Looks like we are missing some prerequisite patch to be able to
directly map such memory in vma_pin_iomap?
For these functions I'm more like a consumer, I was following
suggestions from Chris on this. Is there something extra that should be
considered in this regard when use it like this?
+ }
if (IS_ERR(dpt_obj))
return ERR_CAST(dpt_obj);
--
2.28.0
