drm/radeon: add PRIME support (v2)

This adds prime->fd and fd->prime support to radeon.
It passes the sg object to ttm and then populates
the gart entries using it.

Compile tested only.

v2: stub kmap + use new helpers + add reimporting
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
Reviewed-by: Jerome Glisse <jglisse@redhat.com>
Reviewed-by: Alex Deucher <alexander.deucher@amd.com>
Signed-off-by: Dave Airlie <airlied@redhat.com>

drivers/gpu/drm/radeon/Makefile

@@ -72,7 +72,7 @@ radeon-y += radeon_device.o radeon_asic.o radeon_kms.o \
 	evergreen.o evergreen_cs.o evergreen_blit_shaders.o evergreen_blit_kms.o \
 	evergreen_hdmi.o radeon_trace_points.o ni.o cayman_blit_shaders.o \
 	atombios_encoders.o radeon_semaphore.o radeon_sa.o atombios_i2c.o si.o \
-	si_blit_shaders.o
+	si_blit_shaders.o radeon_prime.o
 
 radeon-$(CONFIG_COMPAT) += radeon_ioc32.o
 radeon-$(CONFIG_VGA_SWITCHEROO) += radeon_atpx_handler.o

drivers/gpu/drm/radeon/evergreen_blit_kms.c

@@ -668,7 +668,7 @@ int evergreen_blit_init(struct radeon_device *rdev)
 	obj_size = ALIGN(obj_size, 256);
 
 	r = radeon_bo_create(rdev, obj_size, PAGE_SIZE, true, RADEON_GEM_DOMAIN_VRAM,
-				&rdev->r600_blit.shader_obj);
+			     NULL, &rdev->r600_blit.shader_obj);
 	if (r) {
 		DRM_ERROR("evergreen failed to allocate shader\n");
 		return r;

drivers/gpu/drm/radeon/r600.c

@@ -1231,7 +1231,7 @@ int r600_vram_scratch_init(struct radeon_device *rdev)
 	if (rdev->vram_scratch.robj == NULL) {
 		r = radeon_bo_create(rdev, RADEON_GPU_PAGE_SIZE,
 				     PAGE_SIZE, true, RADEON_GEM_DOMAIN_VRAM,
-				     &rdev->vram_scratch.robj);
+				     NULL, &rdev->vram_scratch.robj);
 		if (r) {
 			return r;
 		}
@@ -2769,7 +2769,7 @@ int r600_ih_ring_alloc(struct radeon_device *rdev)
 		r = radeon_bo_create(rdev, rdev->ih.ring_size,
 				     PAGE_SIZE, true,
 				     RADEON_GEM_DOMAIN_GTT,
-				     &rdev->ih.ring_obj);
+				     NULL, &rdev->ih.ring_obj);
 		if (r) {
 			DRM_ERROR("radeon: failed to create ih ring buffer (%d).\n", r);
 			return r;

drivers/gpu/drm/radeon/r600_blit_kms.c

@@ -552,7 +552,7 @@ int r600_blit_init(struct radeon_device *rdev)
 	obj_size = ALIGN(obj_size, 256);
 
 	r = radeon_bo_create(rdev, obj_size, PAGE_SIZE, true, RADEON_GEM_DOMAIN_VRAM,
-				&rdev->r600_blit.shader_obj);
+			     NULL, &rdev->r600_blit.shader_obj);
 	if (r) {
 		DRM_ERROR("r600 failed to allocate shader\n");
 		return r;

drivers/gpu/drm/radeon/radeon_benchmark.c

@@ -103,7 +103,7 @@ static void radeon_benchmark_move(struct radeon_device *rdev, unsigned size,
 	int time;
 
 	n = RADEON_BENCHMARK_ITERATIONS;
-	r = radeon_bo_create(rdev, size, PAGE_SIZE, true, sdomain, &sobj);
+	r = radeon_bo_create(rdev, size, PAGE_SIZE, true, sdomain, NULL, &sobj);
 	if (r) {
 		goto out_cleanup;
 	}
@@ -115,7 +115,7 @@ static void radeon_benchmark_move(struct radeon_device *rdev, unsigned size,
 	if (r) {
 		goto out_cleanup;
 	}
-	r = radeon_bo_create(rdev, size, PAGE_SIZE, true, ddomain, &dobj);
+	r = radeon_bo_create(rdev, size, PAGE_SIZE, true, ddomain, NULL, &dobj);
 	if (r) {
 		goto out_cleanup;
 	}

drivers/gpu/drm/radeon/radeon_device.c

@@ -193,7 +193,7 @@ int radeon_wb_init(struct radeon_device *rdev)
 
 	if (rdev->wb.wb_obj == NULL) {
 		r = radeon_bo_create(rdev, RADEON_GPU_PAGE_SIZE, PAGE_SIZE, true,
-				RADEON_GEM_DOMAIN_GTT, &rdev->wb.wb_obj);
+				     RADEON_GEM_DOMAIN_GTT, NULL, &rdev->wb.wb_obj);
 		if (r) {
 			dev_warn(rdev->dev, "(%d) create WB bo failed\n", r);
 			return r;

drivers/gpu/drm/radeon/radeon_drv.c

@@ -105,6 +105,11 @@ int radeon_mode_dumb_create(struct drm_file *file_priv,
 int radeon_mode_dumb_destroy(struct drm_file *file_priv,
 			     struct drm_device *dev,
 			     uint32_t handle);
+struct dma_buf *radeon_gem_prime_export(struct drm_device *dev,
+					struct drm_gem_object *obj,
+					int flags);
+struct drm_gem_object *radeon_gem_prime_import(struct drm_device *dev,
+					       struct dma_buf *dma_buf);
 
 #if defined(CONFIG_DEBUG_FS)
 int radeon_debugfs_init(struct drm_minor *minor);
@@ -333,7 +338,8 @@ static const struct file_operations radeon_driver_kms_fops = {
 static struct drm_driver kms_driver = {
 	.driver_features =
 	    DRIVER_USE_AGP | DRIVER_USE_MTRR | DRIVER_PCI_DMA | DRIVER_SG |
-	    DRIVER_HAVE_IRQ | DRIVER_HAVE_DMA | DRIVER_IRQ_SHARED | DRIVER_GEM,
+	    DRIVER_HAVE_IRQ | DRIVER_HAVE_DMA | DRIVER_IRQ_SHARED | DRIVER_GEM |
+	    DRIVER_PRIME,
 	.dev_priv_size = 0,
 	.load = radeon_driver_load_kms,
 	.firstopen = radeon_driver_firstopen_kms,
@@ -368,6 +374,12 @@ static struct drm_driver kms_driver = {
 	.dumb_map_offset = radeon_mode_dumb_mmap,
 	.dumb_destroy = radeon_mode_dumb_destroy,
 	.fops = &radeon_driver_kms_fops,
+
+	.prime_handle_to_fd = drm_gem_prime_handle_to_fd,
+	.prime_fd_to_handle = drm_gem_prime_fd_to_handle,
+	.gem_prime_export = radeon_gem_prime_export,
+	.gem_prime_import = radeon_gem_prime_import,
+
 	.name = DRIVER_NAME,
 	.desc = DRIVER_DESC,
 	.date = DRIVER_DATE,

drivers/gpu/drm/radeon/radeon_gart.c

@@ -80,7 +80,7 @@ int radeon_gart_table_vram_alloc(struct radeon_device *rdev)
 	if (rdev->gart.robj == NULL) {
 		r = radeon_bo_create(rdev, rdev->gart.table_size,
 				     PAGE_SIZE, true, RADEON_GEM_DOMAIN_VRAM,
-				     &rdev->gart.robj);
+				     NULL, &rdev->gart.robj);
 		if (r) {
 			return r;
 		}

drivers/gpu/drm/radeon/radeon_gem.c

@@ -42,6 +42,8 @@ void radeon_gem_object_free(struct drm_gem_object *gobj)
 	struct radeon_bo *robj = gem_to_radeon_bo(gobj);
 
 	if (robj) {
+		if (robj->gem_base.import_attach)
+			drm_prime_gem_destroy(&robj->gem_base, robj->tbo.sg);
 		radeon_bo_unref(&robj);
 	}
 }
@@ -59,7 +61,7 @@ int radeon_gem_object_create(struct radeon_device *rdev, int size,
 	if (alignment < PAGE_SIZE) {
 		alignment = PAGE_SIZE;
 	}
-	r = radeon_bo_create(rdev, size, alignment, kernel, initial_domain, &robj);
+	r = radeon_bo_create(rdev, size, alignment, kernel, initial_domain, NULL, &robj);
 	if (r) {
 		if (r != -ERESTARTSYS)
 			DRM_ERROR("Failed to allocate GEM object (%d, %d, %u, %d)\n",

drivers/gpu/drm/radeon/radeon_object.c

@@ -104,7 +104,7 @@ void radeon_ttm_placement_from_domain(struct radeon_bo *rbo, u32 domain)
 
 int radeon_bo_create(struct radeon_device *rdev,
 		     unsigned long size, int byte_align, bool kernel, u32 domain,
-		     struct radeon_bo **bo_ptr)
+		     struct sg_table *sg, struct radeon_bo **bo_ptr)
 {
 	struct radeon_bo *bo;
 	enum ttm_bo_type type;
@@ -120,6 +120,8 @@ int radeon_bo_create(struct radeon_device *rdev,
 	}
 	if (kernel) {
 		type = ttm_bo_type_kernel;
+	} else if (sg) {
+		type = ttm_bo_type_sg;
 	} else {
 		type = ttm_bo_type_device;
 	}
@@ -155,7 +157,7 @@ int radeon_bo_create(struct radeon_device *rdev,
 	mutex_lock(&rdev->vram_mutex);
 	r = ttm_bo_init(&rdev->mman.bdev, &bo->tbo, size, type,
 			&bo->placement, page_align, 0, !kernel, NULL,
-			acc_size, NULL, &radeon_ttm_bo_destroy);
+			acc_size, sg, &radeon_ttm_bo_destroy);
 	mutex_unlock(&rdev->vram_mutex);
 	if (unlikely(r != 0)) {
 		if (r != -ERESTARTSYS) {

drivers/gpu/drm/radeon/radeon_object.h

@@ -111,9 +111,10 @@ extern int radeon_bo_wait(struct radeon_bo *bo, u32 *mem_type,
 			  bool no_wait);
 
 extern int radeon_bo_create(struct radeon_device *rdev,
-				unsigned long size, int byte_align,
-				bool kernel, u32 domain,
-				struct radeon_bo **bo_ptr);
+			    unsigned long size, int byte_align,
+			    bool kernel, u32 domain,
+			    struct sg_table *sg,
+			    struct radeon_bo **bo_ptr);
 extern int radeon_bo_kmap(struct radeon_bo *bo, void **ptr);
 extern void radeon_bo_kunmap(struct radeon_bo *bo);
 extern void radeon_bo_unref(struct radeon_bo **bo);

drivers/gpu/drm/radeon/radeon_prime.c

+/*
+ * Copyright 2012 Advanced Micro Devices, Inc.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice shall be included in
+ * all copies or substantial portions of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
+ * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR
+ * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE,
+ * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR
+ * OTHER DEALINGS IN THE SOFTWARE.
+ *
+ * based on nouveau_prime.c
+ *
+ * Authors: Alex Deucher
+ */
+#include "drmP.h"
+#include "drm.h"
+
+#include "radeon.h"
+#include "radeon_drm.h"
+
+#include <linux/dma-buf.h>
+
+static struct sg_table *radeon_gem_map_dma_buf(struct dma_buf_attachment *attachment,
+					       enum dma_data_direction dir)
+{
+	struct radeon_bo *bo = attachment->dmabuf->priv;
+	struct drm_device *dev = bo->rdev->ddev;
+	int npages = bo->tbo.num_pages;
+	struct sg_table *sg;
+	int nents;
+
+	mutex_lock(&dev->struct_mutex);
+	sg = drm_prime_pages_to_sg(bo->tbo.ttm->pages, npages);
+	nents = dma_map_sg(attachment->dev, sg->sgl, sg->nents, dir);
+	mutex_unlock(&dev->struct_mutex);
+	return sg;
+}
+
+static void radeon_gem_unmap_dma_buf(struct dma_buf_attachment *attachment,
+				     struct sg_table *sg, enum dma_data_direction dir)
+{
+	dma_unmap_sg(attachment->dev, sg->sgl, sg->nents, dir);
+	sg_free_table(sg);
+	kfree(sg);
+}
+
+static void radeon_gem_dmabuf_release(struct dma_buf *dma_buf)
+{
+	struct radeon_bo *bo = dma_buf->priv;
+
+	if (bo->gem_base.export_dma_buf == dma_buf) {
+		DRM_ERROR("unreference dmabuf %p\n", &bo->gem_base);
+		bo->gem_base.export_dma_buf = NULL;
+		drm_gem_object_unreference_unlocked(&bo->gem_base);
+	}
+}
+
+static void *radeon_gem_kmap_atomic(struct dma_buf *dma_buf, unsigned long page_num)
+{
+	return NULL;
+}
+
+static void radeon_gem_kunmap_atomic(struct dma_buf *dma_buf, unsigned long page_num, void *addr)
+{
+
+}
+static void *radeon_gem_kmap(struct dma_buf *dma_buf, unsigned long page_num)
+{
+	return NULL;
+}
+
+static void radeon_gem_kunmap(struct dma_buf *dma_buf, unsigned long page_num, void *addr)
+{
+
+}
+
+struct dma_buf_ops radeon_dmabuf_ops =  {
+	.map_dma_buf = radeon_gem_map_dma_buf,
+	.unmap_dma_buf = radeon_gem_unmap_dma_buf,
+	.release = radeon_gem_dmabuf_release,
+	.kmap = radeon_gem_kmap,
+	.kmap_atomic = radeon_gem_kmap_atomic,
+	.kunmap = radeon_gem_kunmap,
+	.kunmap_atomic = radeon_gem_kunmap_atomic,
+};
+
+static int radeon_prime_create(struct drm_device *dev,
+			       size_t size,
+			       struct sg_table *sg,
+			       struct radeon_bo **pbo)
+{
+	struct radeon_device *rdev = dev->dev_private;
+	struct radeon_bo *bo;
+	int ret;
+
+	ret = radeon_bo_create(rdev, size, PAGE_SIZE, false,
+			       RADEON_GEM_DOMAIN_GTT, sg, pbo);
+	if (ret)
+		return ret;
+	bo = *pbo;
+	bo->gem_base.driver_private = bo;
+
+	mutex_lock(&rdev->gem.mutex);
+	list_add_tail(&bo->list, &rdev->gem.objects);
+	mutex_unlock(&rdev->gem.mutex);
+
+	return 0;
+}
+
+struct dma_buf *radeon_gem_prime_export(struct drm_device *dev,
+					struct drm_gem_object *obj,
+					int flags)
+{
+	struct radeon_bo *bo = gem_to_radeon_bo(obj);
+	int ret = 0;
+
+	/* pin buffer into GTT */
+	ret = radeon_bo_pin(bo, RADEON_GEM_DOMAIN_GTT, NULL);
+	if (ret)
+		return ERR_PTR(ret);
+
+	return dma_buf_export(bo, &radeon_dmabuf_ops, obj->size, flags);
+}
+
+struct drm_gem_object *radeon_gem_prime_import(struct drm_device *dev,
+					       struct dma_buf *dma_buf)
+{
+	struct dma_buf_attachment *attach;
+	struct sg_table *sg;
+	struct radeon_bo *bo;
+	int ret;
+
+	if (dma_buf->ops == &radeon_dmabuf_ops) {
+		bo = dma_buf->priv;
+		if (bo->gem_base.dev == dev) {
+			drm_gem_object_reference(&bo->gem_base);
+			return &bo->gem_base;
+		}
+	}
+
+	/* need to attach */
+	attach = dma_buf_attach(dma_buf, dev->dev);
+	if (IS_ERR(attach))
+		return ERR_CAST(attach);
+
+	sg = dma_buf_map_attachment(attach, DMA_BIDIRECTIONAL);
+	if (IS_ERR(sg)) {
+		ret = PTR_ERR(sg);
+		goto fail_detach;
+	}
+
+	ret = radeon_prime_create(dev, dma_buf->size, sg, &bo);
+	if (ret)
+		goto fail_unmap;
+
+	bo->gem_base.import_attach = attach;
+
+	return &bo->gem_base;
+
+fail_unmap:
+	dma_buf_unmap_attachment(attach, sg, DMA_BIDIRECTIONAL);
+fail_detach:
+	dma_buf_detach(dma_buf, attach);
+	return ERR_PTR(ret);
+}

drivers/gpu/drm/radeon/radeon_ring.c

@@ -390,8 +390,8 @@ int radeon_ring_init(struct radeon_device *rdev, struct radeon_ring *ring, unsig
 	/* Allocate ring buffer */
 	if (ring->ring_obj == NULL) {
 		r = radeon_bo_create(rdev, ring->ring_size, PAGE_SIZE, true,
-					RADEON_GEM_DOMAIN_GTT,
-					&ring->ring_obj);
+				     RADEON_GEM_DOMAIN_GTT,
+				     NULL, &ring->ring_obj);
 		if (r) {
 			dev_err(rdev->dev, "(%d) ring create failed\n", r);
 			return r;

drivers/gpu/drm/radeon/radeon_sa.c

@@ -65,7 +65,7 @@ int radeon_sa_bo_manager_init(struct radeon_device *rdev,
 	}
 
 	r = radeon_bo_create(rdev, size, RADEON_GPU_PAGE_SIZE, true,
-			     RADEON_GEM_DOMAIN_CPU, &sa_manager->bo);
+			     RADEON_GEM_DOMAIN_CPU, NULL, &sa_manager->bo);
 	if (r) {
 		dev_err(rdev->dev, "(%d) failed to allocate bo for manager\n", r);
 		return r;

drivers/gpu/drm/radeon/radeon_test.c

@@ -59,7 +59,7 @@ void radeon_test_moves(struct radeon_device *rdev)
 	}
 
 	r = radeon_bo_create(rdev, size, PAGE_SIZE, true, RADEON_GEM_DOMAIN_VRAM,
-				&vram_obj);
+			     NULL, &vram_obj);
 	if (r) {
 		DRM_ERROR("Failed to create VRAM object\n");
 		goto out_cleanup;
@@ -78,7 +78,7 @@ void radeon_test_moves(struct radeon_device *rdev)
 		void **vram_start, **vram_end;
 
 		r = radeon_bo_create(rdev, size, PAGE_SIZE, true,
-					 RADEON_GEM_DOMAIN_GTT, gtt_obj + i);
+				     RADEON_GEM_DOMAIN_GTT, NULL, gtt_obj + i);
 		if (r) {
 			DRM_ERROR("Failed to create GTT object %d\n", i);
 			goto out_cleanup;

drivers/gpu/drm/radeon/radeon_ttm.c

@@ -611,10 +611,18 @@ static int radeon_ttm_tt_populate(struct ttm_tt *ttm)
 	struct radeon_ttm_tt *gtt = (void *)ttm;
 	unsigned i;
 	int r;
+	bool slave = !!(ttm->page_flags & TTM_PAGE_FLAG_SG);
 
 	if (ttm->state != tt_unpopulated)
 		return 0;
 
+	if (slave && ttm->sg) {
+		drm_prime_sg_to_page_addr_arrays(ttm->sg, ttm->pages,
+						 gtt->ttm.dma_address, ttm->num_pages);
+		ttm->state = tt_unbound;
+		return 0;
+	}
+
 	rdev = radeon_get_rdev(ttm->bdev);
 #if __OS_HAS_AGP
 	if (rdev->flags & RADEON_IS_AGP) {
@@ -655,6 +663,10 @@ static void radeon_ttm_tt_unpopulate(struct ttm_tt *ttm)
 	struct radeon_device *rdev;
 	struct radeon_ttm_tt *gtt = (void *)ttm;
 	unsigned i;
+	bool slave = !!(ttm->page_flags & TTM_PAGE_FLAG_SG);
+
+	if (slave)
+		return;
 
 	rdev = radeon_get_rdev(ttm->bdev);
 #if __OS_HAS_AGP
@@ -726,8 +738,8 @@ int radeon_ttm_init(struct radeon_device *rdev)
 		return r;
 	}
 	r = radeon_bo_create(rdev, 256 * 1024, PAGE_SIZE, true,
-				RADEON_GEM_DOMAIN_VRAM,
-				&rdev->stollen_vga_memory);
+			     RADEON_GEM_DOMAIN_VRAM,
+			     NULL, &rdev->stollen_vga_memory);
 	if (r) {
 		return r;
 	}

drivers/gpu/drm/radeon/si.c

@@ -2976,7 +2976,8 @@ int si_rlc_init(struct radeon_device *rdev)
 	/* save restore block */
 	if (rdev->rlc.save_restore_obj == NULL) {
 		r = radeon_bo_create(rdev, RADEON_GPU_PAGE_SIZE, PAGE_SIZE, true,
-				RADEON_GEM_DOMAIN_VRAM, &rdev->rlc.save_restore_obj);
+				     RADEON_GEM_DOMAIN_VRAM, NULL,
+				     &rdev->rlc.save_restore_obj);
 		if (r) {
 			dev_warn(rdev->dev, "(%d) create RLC sr bo failed\n", r);
 			return r;
@@ -3000,7 +3001,8 @@ int si_rlc_init(struct radeon_device *rdev)
 	/* clear state block */
 	if (rdev->rlc.clear_state_obj == NULL) {
 		r = radeon_bo_create(rdev, RADEON_GPU_PAGE_SIZE, PAGE_SIZE, true,
-				RADEON_GEM_DOMAIN_VRAM, &rdev->rlc.clear_state_obj);
+				     RADEON_GEM_DOMAIN_VRAM, NULL,
+				     &rdev->rlc.clear_state_obj);
 		if (r) {
 			dev_warn(rdev->dev, "(%d) create RLC c bo failed\n", r);
 			si_rlc_fini(rdev);