drm/radeon: add userptr support v8

This patch adds an IOCTL for turning a pointer supplied by
userspace into a buffer object.

It imposes several restrictions upon the memory being mapped:

1. It must be page aligned (both start/end addresses, i.e ptr and size).

2. It must be normal system memory, not a pointer into another map of IO
space (e.g. it must not be a GTT mmapping of another object).

3. The BO is mapped into GTT, so the maximum amount of memory mapped at
all times is still the GTT limit.

4. The BO is only mapped readonly for now, so no write support.

5. List of backing pages is only acquired once, so they represent a
snapshot of the first use.

Exporting and sharing as well as mapping of buffer objects created by
this function is forbidden and results in an -EPERM.

v2: squash all previous changes into first public version
v3: fix tabs, map readonly, don't use MM callback any more
v4: set TTM_PAGE_FLAG_SG so that TTM never messes with the pages,
    pin/unpin pages on bind/unbind instead of populate/unpopulate
v5: rebased on 3.17-wip, IOCTL renamed to userptr, reject any unknown
    flags, better handle READONLY flag, improve permission check
v6: fix ptr cast warning, use set_page_dirty/mark_page_accessed on unpin
v7: add warning about it's availability in the API definition
v8: drop access_ok check, fix VM mapping bits

Signed-off-by: Christian König <christian.koenig@amd.com>
Reviewed-by: Alex Deucher <alexander.deucher@amd.com> (v4)
Reviewed-by: Jérôme Glisse <jglisse@redhat.com> (v4)
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

1 files changed, 145 insertions, 0 deletions

diff --git a/drivers/gpu/drm/radeon/radeon_ttm.c b/drivers/gpu/drm/radeon/radeon_ttm.c
index 72afe82a95c9..b20933fa35c6 100644
--- a/drivers/gpu/drm/radeon/radeon_ttm.c
+++ b/drivers/gpu/drm/radeon/radeon_ttm.c
@@ -39,6 +39,8 @@
 #include <linux/seq_file.h>
 #include <linux/slab.h>
 #include <linux/swiotlb.h>
+#include <linux/swap.h>
+#include <linux/pagemap.h>
 #include <linux/debugfs.h>
 #include "radeon_reg.h"
 #include "radeon.h"
@@ -515,8 +517,92 @@ struct radeon_ttm_tt {
 	struct ttm_dma_tt		ttm;
 	struct radeon_device		*rdev;
 	u64				offset;
+
+	uint64_t			userptr;
+	struct mm_struct		*usermm;
+	uint32_t			userflags;
 };
 
+/* prepare the sg table with the user pages */
+static int radeon_ttm_tt_pin_userptr(struct ttm_tt *ttm)
+{
+	struct radeon_device *rdev = radeon_get_rdev(ttm->bdev);
+	struct radeon_ttm_tt *gtt = (void *)ttm;
+	unsigned pinned = 0, nents;
+	int r;
+
+	int write = !(gtt->userflags & RADEON_GEM_USERPTR_READONLY);
+	enum dma_data_direction direction = write ?
+		DMA_BIDIRECTIONAL : DMA_TO_DEVICE;
+
+	if (current->mm != gtt->usermm)
+		return -EPERM;
+
+	do {
+		unsigned num_pages = ttm->num_pages - pinned;
+		uint64_t userptr = gtt->userptr + pinned * PAGE_SIZE;
+		struct page **pages = ttm->pages + pinned;
+
+		r = get_user_pages(current, current->mm, userptr, num_pages,
+				   write, 0, pages, NULL);
+		if (r < 0)
+			goto release_pages;
+
+		pinned += r;
+
+	} while (pinned < ttm->num_pages);
+
+	r = sg_alloc_table_from_pages(ttm->sg, ttm->pages, ttm->num_pages, 0,
+				      ttm->num_pages << PAGE_SHIFT,
+				      GFP_KERNEL);
+	if (r)
+		goto release_sg;
+
+	r = -ENOMEM;
+	nents = dma_map_sg(rdev->dev, ttm->sg->sgl, ttm->sg->nents, direction);
+	if (nents != ttm->sg->nents)
+		goto release_sg;
+
+	drm_prime_sg_to_page_addr_arrays(ttm->sg, ttm->pages,
+					 gtt->ttm.dma_address, ttm->num_pages);
+
+	return 0;
+
+release_sg:
+	kfree(ttm->sg);
+
+release_pages:
+	release_pages(ttm->pages, pinned, 0);
+	return r;
+}
+
+static void radeon_ttm_tt_unpin_userptr(struct ttm_tt *ttm)
+{
+	struct radeon_device *rdev = radeon_get_rdev(ttm->bdev);
+	struct radeon_ttm_tt *gtt = (void *)ttm;
+	struct scatterlist *sg;
+	int i;
+
+	int write = !(gtt->userflags & RADEON_GEM_USERPTR_READONLY);
+	enum dma_data_direction direction = write ?
+		DMA_BIDIRECTIONAL : DMA_TO_DEVICE;
+
+	/* free the sg table and pages again */
+	dma_unmap_sg(rdev->dev, ttm->sg->sgl, ttm->sg->nents, direction);
+
+	for_each_sg(ttm->sg->sgl, sg, ttm->sg->nents, i) {
+		struct page *page = sg_page(sg);
+
+		if (!(gtt->userflags & RADEON_GEM_USERPTR_READONLY))
+			set_page_dirty(page);
+
+		mark_page_accessed(page);
+		page_cache_release(page);
+	}
+
+	sg_free_table(ttm->sg);
+}
+
 static int radeon_ttm_backend_bind(struct ttm_tt *ttm,
 				   struct ttm_mem_reg *bo_mem)
 {
@@ -525,6 +611,11 @@ static int radeon_ttm_backend_bind(struct ttm_tt *ttm,
 		RADEON_GART_PAGE_WRITE;
 	int r;
 
+	if (gtt->userptr) {
+		radeon_ttm_tt_pin_userptr(ttm);
+		flags &= ~RADEON_GART_PAGE_WRITE;
+	}
+
 	gtt->offset = (unsigned long)(bo_mem->start << PAGE_SHIFT);
 	if (!ttm->num_pages) {
 		WARN(1, "nothing to bind %lu pages for mreg %p back %p!\n",
@@ -547,6 +638,10 @@ static int radeon_ttm_backend_unbind(struct ttm_tt *ttm)
 	struct radeon_ttm_tt *gtt = (void *)ttm;
 
 	radeon_gart_unbind(gtt->rdev, gtt->offset, ttm->num_pages);
+
+	if (gtt->userptr)
+		radeon_ttm_tt_unpin_userptr(ttm);
+
 	return 0;
 }
 
@@ -603,6 +698,16 @@ static int radeon_ttm_tt_populate(struct ttm_tt *ttm)
 	if (ttm->state != tt_unpopulated)
 		return 0;
 
+	if (gtt->userptr) {
+		ttm->sg = kcalloc(1, sizeof(struct sg_table), GFP_KERNEL);
+		if (!ttm->sg)
+			return -ENOMEM;
+
+		ttm->page_flags |= TTM_PAGE_FLAG_SG;
+		ttm->state = tt_unbound;
+		return 0;
+	}
+
 	if (slave && ttm->sg) {
 		drm_prime_sg_to_page_addr_arrays(ttm->sg, ttm->pages,
 						 gtt->ttm.dma_address, ttm->num_pages);
@@ -652,6 +757,12 @@ static void radeon_ttm_tt_unpopulate(struct ttm_tt *ttm)
 	unsigned i;
 	bool slave = !!(ttm->page_flags & TTM_PAGE_FLAG_SG);
 
+	if (gtt->userptr) {
+		kfree(ttm->sg);
+		ttm->page_flags &= ~TTM_PAGE_FLAG_SG;
+		return;
+	}
+
 	if (slave)
 		return;
 
@@ -680,6 +791,40 @@ static void radeon_ttm_tt_unpopulate(struct ttm_tt *ttm)
 	ttm_pool_unpopulate(ttm);
 }
 
+int radeon_ttm_tt_set_userptr(struct ttm_tt *ttm, uint64_t addr,
+			      uint32_t flags)
+{
+	struct radeon_ttm_tt *gtt = (void *)ttm;
+
+	if (gtt == NULL)
+		return -EINVAL;
+
+	gtt->userptr = addr;
+	gtt->usermm = current->mm;
+	gtt->userflags = flags;
+	return 0;
+}
+
+bool radeon_ttm_tt_has_userptr(struct ttm_tt *ttm)
+{
+	struct radeon_ttm_tt *gtt = (void *)ttm;
+
+	if (gtt == NULL)
+		return false;
+
+	return !!gtt->userptr;
+}
+
+bool radeon_ttm_tt_is_readonly(struct ttm_tt *ttm)
+{
+	struct radeon_ttm_tt *gtt = (void *)ttm;
+
+	if (gtt == NULL)
+		return false;
+
+	return !!(gtt->userflags & RADEON_GEM_USERPTR_READONLY);
+}
+
 static struct ttm_bo_driver radeon_bo_driver = {
 	.ttm_tt_create = &radeon_ttm_tt_create,
 	.ttm_tt_populate = &radeon_ttm_tt_populate,