Merge branch 'drm-next-4.17' of git://people.freedesktop.org/~agd5f/linux into drm-next

- Expose thermal thresholds through hwmon properly
- Rework HDP flushing for rings and CPU
- Improved dual-link DVI handling in DC
- Lots of code clean up
- Additional DC clean up
- Allow scanout from system memory on CZ/BR/ST
- Improved PASID/VM integration
- Expose GPU voltage and power via hwmon
- Initial wattman-like support
- Initial power profiles for use-case optimized performance
- Rework GPUVM TLB flushing
- Rework IP offset handling for SOC15 asics
- Add CRC support in DC
- Fixes for mmhub powergating
- Initial regamma/degamma/CTM support in DC
- ttm cleanups and simplifications
- ttm OOM avoidance fixes

* 'drm-next-4.17' of git://people.freedesktop.org/~agd5f/linux: (348 commits)
  Revert "drm/radeon/pm: autoswitch power state when in balanced mode"
  drm/radeon: use drm_gem_private_object_init
  drm/amdgpu: use drm_gem_private_object_init
  drm/amdgpu: mitigate workaround for i915
  drm/amdgpu: implement amdgpu_gem_map_(attach/detach)
  drm/amdgpu/powerplay/smu7: drop refresh rate checks for mclk switching
  drm/amdgpu/cgs: add refresh rate checking to non-DC display code
  drm/amd/powerplay/smu7: allow mclk switching with no displays
  drm/amd/powerplay/vega10: allow mclk switching with no displays
  drm/amd/powerplay: use PP_CAP macro for disable_mclk_switching_for_frame_lock
  drm/amd/powerplay: remove unused headers
  drm/amdgpu_gem: fix error handling path in amdgpu_gem_va_update_vm
  drm/amdgpu: update the PASID mapping only on demand
  drm/amdgpu: separate PASID mapping from VM flush v2
  drm/amd/display: Fix increment when sampling OTF in DCE
  drm/amd/display: De PQ implementation
  drm/amd/display: Remove unused dm_pp_ interfaces
  drm/amd/display: Add logging for aux DPCD access
  drm/amd/display: Set vsc pack revision when DPCD revision is >= 1.2
  drm/amd/display: provide an interface to query firmware version
  ...

1 files changed, 118 insertions, 25 deletions

diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_prime.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_prime.c
index ae9c106979d7..8ce74a1d9966 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_prime.c
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_prime.c
@@ -26,9 +26,12 @@
 #include <drm/drmP.h>
 
 #include "amdgpu.h"
+#include "amdgpu_display.h"
 #include <drm/amdgpu_drm.h>
 #include <linux/dma-buf.h>
 
+static const struct dma_buf_ops amdgpu_dmabuf_ops;
+
 struct sg_table *amdgpu_gem_prime_get_sg_table(struct drm_gem_object *obj)
 {
 	struct amdgpu_bo *bo = gem_to_amdgpu_bo(obj);
@@ -103,7 +106,7 @@ amdgpu_gem_prime_import_sg_table(struct drm_device *dev,
 
 	ww_mutex_lock(&resv->lock, NULL);
 	ret = amdgpu_bo_create(adev, attach->dmabuf->size, PAGE_SIZE, false,
-			       AMDGPU_GEM_DOMAIN_GTT, 0, sg, resv, 0, &bo);
+			       AMDGPU_GEM_DOMAIN_GTT, 0, sg, resv, &bo);
 	ww_mutex_unlock(&resv->lock);
 	if (ret)
 		return ERR_PTR(ret);
@@ -112,49 +115,72 @@ amdgpu_gem_prime_import_sg_table(struct drm_device *dev,
 	return &bo->gem_base;
 }
 
-int amdgpu_gem_prime_pin(struct drm_gem_object *obj)
+static int amdgpu_gem_map_attach(struct dma_buf *dma_buf,
+				 struct device *target_dev,
+				 struct dma_buf_attachment *attach)
 {
+	struct drm_gem_object *obj = dma_buf->priv;
 	struct amdgpu_bo *bo = gem_to_amdgpu_bo(obj);
-	long ret = 0;
-
-	ret = amdgpu_bo_reserve(bo, false);
-	if (unlikely(ret != 0))
-		return ret;
-
-	/*
-	 * Wait for all shared fences to complete before we switch to future
-	 * use of exclusive fence on this prime shared bo.
-	 */
-	ret = reservation_object_wait_timeout_rcu(bo->tbo.resv, true, false,
-						  MAX_SCHEDULE_TIMEOUT);
-	if (unlikely(ret < 0)) {
-		DRM_DEBUG_PRIME("Fence wait failed: %li\n", ret);
-		amdgpu_bo_unreserve(bo);
-		return ret;
+	long r;
+
+	r = drm_gem_map_attach(dma_buf, target_dev, attach);
+	if (r)
+		return r;
+
+	r = amdgpu_bo_reserve(bo, false);
+	if (unlikely(r != 0))
+		goto error_detach;
+
+
+	if (dma_buf->ops != &amdgpu_dmabuf_ops) {
+		/*
+		 * Wait for all shared fences to complete before we switch to future
+		 * use of exclusive fence on this prime shared bo.
+		 */
+		r = reservation_object_wait_timeout_rcu(bo->tbo.resv,
+							true, false,
+							MAX_SCHEDULE_TIMEOUT);
+		if (unlikely(r < 0)) {
+			DRM_DEBUG_PRIME("Fence wait failed: %li\n", r);
+			goto error_unreserve;
+		}
 	}
 
 	/* pin buffer into GTT */
-	ret = amdgpu_bo_pin(bo, AMDGPU_GEM_DOMAIN_GTT, NULL);
-	if (likely(ret == 0))
+	r = amdgpu_bo_pin(bo, AMDGPU_GEM_DOMAIN_GTT, NULL);
+	if (r)
+		goto error_unreserve;
+
+	if (dma_buf->ops != &amdgpu_dmabuf_ops)
 		bo->prime_shared_count++;
 
+error_unreserve:
 	amdgpu_bo_unreserve(bo);
-	return ret;
+
+error_detach:
+	if (r)
+		drm_gem_map_detach(dma_buf, attach);
+	return r;
 }
 
-void amdgpu_gem_prime_unpin(struct drm_gem_object *obj)
+static void amdgpu_gem_map_detach(struct dma_buf *dma_buf,
+				  struct dma_buf_attachment *attach)
 {
+	struct drm_gem_object *obj = dma_buf->priv;
 	struct amdgpu_bo *bo = gem_to_amdgpu_bo(obj);
 	int ret = 0;
 
 	ret = amdgpu_bo_reserve(bo, true);
 	if (unlikely(ret != 0))
-		return;
+		goto error;
 
 	amdgpu_bo_unpin(bo);
-	if (bo->prime_shared_count)
+	if (dma_buf->ops != &amdgpu_dmabuf_ops && bo->prime_shared_count)
 		bo->prime_shared_count--;
 	amdgpu_bo_unreserve(bo);
+
+error:
+	drm_gem_map_detach(dma_buf, attach);
 }
 
 struct reservation_object *amdgpu_gem_prime_res_obj(struct drm_gem_object *obj)
@@ -164,6 +190,50 @@ struct reservation_object *amdgpu_gem_prime_res_obj(struct drm_gem_object *obj)
 	return bo->tbo.resv;
 }
 
+static int amdgpu_gem_begin_cpu_access(struct dma_buf *dma_buf,
+				       enum dma_data_direction direction)
+{
+	struct amdgpu_bo *bo = gem_to_amdgpu_bo(dma_buf->priv);
+	struct amdgpu_device *adev = amdgpu_ttm_adev(bo->tbo.bdev);
+	struct ttm_operation_ctx ctx = { true, false };
+	u32 domain = amdgpu_display_framebuffer_domains(adev);
+	int ret;
+	bool reads = (direction == DMA_BIDIRECTIONAL ||
+		      direction == DMA_FROM_DEVICE);
+
+	if (!reads || !(domain & AMDGPU_GEM_DOMAIN_GTT))
+		return 0;
+
+	/* move to gtt */
+	ret = amdgpu_bo_reserve(bo, false);
+	if (unlikely(ret != 0))
+		return ret;
+
+	if (!bo->pin_count && (bo->allowed_domains & AMDGPU_GEM_DOMAIN_GTT)) {
+		amdgpu_ttm_placement_from_domain(bo, AMDGPU_GEM_DOMAIN_GTT);
+		ret = ttm_bo_validate(&bo->tbo, &bo->placement, &ctx);
+	}
+
+	amdgpu_bo_unreserve(bo);
+	return ret;
+}
+
+static const struct dma_buf_ops amdgpu_dmabuf_ops = {
+	.attach = amdgpu_gem_map_attach,
+	.detach = amdgpu_gem_map_detach,
+	.map_dma_buf = drm_gem_map_dma_buf,
+	.unmap_dma_buf = drm_gem_unmap_dma_buf,
+	.release = drm_gem_dmabuf_release,
+	.begin_cpu_access = amdgpu_gem_begin_cpu_access,
+	.map = drm_gem_dmabuf_kmap,
+	.map_atomic = drm_gem_dmabuf_kmap_atomic,
+	.unmap = drm_gem_dmabuf_kunmap,
+	.unmap_atomic = drm_gem_dmabuf_kunmap_atomic,
+	.mmap = drm_gem_dmabuf_mmap,
+	.vmap = drm_gem_dmabuf_vmap,
+	.vunmap = drm_gem_dmabuf_vunmap,
+};
+
 struct dma_buf *amdgpu_gem_prime_export(struct drm_device *dev,
 					struct drm_gem_object *gobj,
 					int flags)
@@ -176,7 +246,30 @@ struct dma_buf *amdgpu_gem_prime_export(struct drm_device *dev,
 		return ERR_PTR(-EPERM);
 
 	buf = drm_gem_prime_export(dev, gobj, flags);
-	if (!IS_ERR(buf))
+	if (!IS_ERR(buf)) {
 		buf->file->f_mapping = dev->anon_inode->i_mapping;
+		buf->ops = &amdgpu_dmabuf_ops;
+	}
+
 	return buf;
 }
+
+struct drm_gem_object *amdgpu_gem_prime_import(struct drm_device *dev,
+					    struct dma_buf *dma_buf)
+{
+	struct drm_gem_object *obj;
+
+	if (dma_buf->ops == &amdgpu_dmabuf_ops) {
+		obj = dma_buf->priv;
+		if (obj->dev == dev) {
+			/*
+			 * Importing dmabuf exported from out own gem increases
+			 * refcount on gem itself instead of f_count of dmabuf.
+			 */
+			drm_gem_object_get(obj);
+			return obj;
+		}
+	}
+
+	return drm_gem_prime_import(dev, dma_buf);
+}