drm/radeon: introduce kernel modesetting for radeon hardware

Add kernel modesetting support to radeon driver, use the ttm memory
manager to manage memory and DRM/GEM to provide userspace API.
In order to avoid backward compatibility issue and to allow clean
design and code the radeon kernel modesetting use different code path
than old radeon/drm driver.

When kernel modesetting is enabled the IOCTL of radeon/drm
driver are considered as invalid and an error message is printed
in the log and they return failure.

KMS enabled userspace will use new API to talk with the radeon/drm
driver. The new API provide functions to create/destroy/share/mmap
buffer object which are then managed by the kernel memory manager
(here TTM). In order to submit command to the GPU the userspace
provide a buffer holding the command stream, along this buffer
userspace have to provide a list of buffer object used by the
command stream. The kernel radeon driver will then place buffer
in GPU accessible memory and will update command stream to reflect
the position of the different buffers.

The kernel will also perform security check on command stream
provided by the user, we want to catch and forbid any illegal use
of the GPU such as DMA into random system memory or into memory
not owned by the process supplying the command stream. This part
of the code is still incomplete and this why we propose that patch
as a staging driver addition, future security might forbid current
experimental userspace to run.

This code support the following hardware : R1XX,R2XX,R3XX,R4XX,R5XX
(radeon up to X1950). Works is underway to provide support for R6XX,
R7XX and newer hardware (radeon from HD2XXX to HD4XXX).

Authors:
    Jerome Glisse <jglisse@redhat.com>
    Dave Airlie <airlied@redhat.com>
    Alex Deucher <alexdeucher@gmail.com>

Signed-off-by: Jerome Glisse <jglisse@redhat.com>
Signed-off-by: Dave Airlie <airlied@redhat.com>
Signed-off-by: Alex Deucher <alexdeucher@gmail.com>
Signed-off-by: Dave Airlie <airlied@redhat.com>

1 files changed, 511 insertions, 0 deletions

diff --git a/drivers/gpu/drm/radeon/radeon_object.c b/drivers/gpu/drm/radeon/radeon_object.c
new file mode 100644
index 000000000000..983e8df5e000
--- /dev/null
+++ b/drivers/gpu/drm/radeon/radeon_object.c
@@ -0,0 +1,511 @@
+/*
+ * Copyright 2009 Jerome Glisse.
+ * All Rights Reserved.
+ *
+ * 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, sub license, 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 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 NON-INFRINGEMENT. IN NO EVENT SHALL
+ * THE COPYRIGHT HOLDERS, AUTHORS AND/OR ITS SUPPLIERS 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.
+ *
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ *
+ */
+/*
+ * Authors:
+ *    Jerome Glisse <glisse@freedesktop.org>
+ *    Thomas Hellstrom <thomas-at-tungstengraphics-dot-com>
+ *    Dave Airlie
+ */
+#include <linux/list.h>
+#include <drm/drmP.h>
+#include "radeon_drm.h"
+#include "radeon.h"
+
+struct radeon_object {
+	struct ttm_buffer_object	tobj;
+	struct list_head		list;
+	struct radeon_device		*rdev;
+	struct drm_gem_object		*gobj;
+	struct ttm_bo_kmap_obj		kmap;
+	unsigned			pin_count;
+	uint64_t			gpu_addr;
+	void				*kptr;
+	bool				is_iomem;
+};
+
+int radeon_ttm_init(struct radeon_device *rdev);
+void radeon_ttm_fini(struct radeon_device *rdev);
+
+/*
+ * To exclude mutual BO access we rely on bo_reserve exclusion, as all
+ * function are calling it.
+ */
+
+static int radeon_object_reserve(struct radeon_object *robj, bool interruptible)
+{
+	return ttm_bo_reserve(&robj->tobj, interruptible, false, false, 0);
+}
+
+static void radeon_object_unreserve(struct radeon_object *robj)
+{
+	ttm_bo_unreserve(&robj->tobj);
+}
+
+static void radeon_ttm_object_object_destroy(struct ttm_buffer_object *tobj)
+{
+	struct radeon_object *robj;
+
+	robj = container_of(tobj, struct radeon_object, tobj);
+	list_del_init(&robj->list);
+	kfree(robj);
+}
+
+static inline void radeon_object_gpu_addr(struct radeon_object *robj)
+{
+	/* Default gpu address */
+	robj->gpu_addr = 0xFFFFFFFFFFFFFFFFULL;
+	if (robj->tobj.mem.mm_node == NULL) {
+		return;
+	}
+	robj->gpu_addr = ((u64)robj->tobj.mem.mm_node->start) << PAGE_SHIFT;
+	switch (robj->tobj.mem.mem_type) {
+	case TTM_PL_VRAM:
+		robj->gpu_addr += (u64)robj->rdev->mc.vram_location;
+		break;
+	case TTM_PL_TT:
+		robj->gpu_addr += (u64)robj->rdev->mc.gtt_location;
+		break;
+	default:
+		DRM_ERROR("Unknown placement %d\n", robj->tobj.mem.mem_type);
+		robj->gpu_addr = 0xFFFFFFFFFFFFFFFFULL;
+		return;
+	}
+}
+
+static inline uint32_t radeon_object_flags_from_domain(uint32_t domain)
+{
+	uint32_t flags = 0;
+	if (domain & RADEON_GEM_DOMAIN_VRAM) {
+		flags |= TTM_PL_FLAG_VRAM;
+	}
+	if (domain & RADEON_GEM_DOMAIN_GTT) {
+		flags |= TTM_PL_FLAG_TT;
+	}
+	if (domain & RADEON_GEM_DOMAIN_CPU) {
+		flags |= TTM_PL_FLAG_SYSTEM;
+	}
+	if (!flags) {
+		flags |= TTM_PL_FLAG_SYSTEM;
+	}
+	return flags;
+}
+
+int radeon_object_create(struct radeon_device *rdev,
+			 struct drm_gem_object *gobj,
+			 unsigned long size,
+			 bool kernel,
+			 uint32_t domain,
+			 bool interruptible,
+			 struct radeon_object **robj_ptr)
+{
+	struct radeon_object *robj;
+	enum ttm_bo_type type;
+	uint32_t flags;
+	int r;
+
+	if (unlikely(rdev->mman.bdev.dev_mapping == NULL)) {
+		rdev->mman.bdev.dev_mapping = rdev->ddev->dev_mapping;
+	}
+	if (kernel) {
+		type = ttm_bo_type_kernel;
+	} else {
+		type = ttm_bo_type_device;
+	}
+	*robj_ptr = NULL;
+	robj = kzalloc(sizeof(struct radeon_object), GFP_KERNEL);
+	if (robj == NULL) {
+		return -ENOMEM;
+	}
+	robj->rdev = rdev;
+	robj->gobj = gobj;
+	INIT_LIST_HEAD(&robj->list);
+
+	flags = radeon_object_flags_from_domain(domain);
+	r = ttm_buffer_object_init(&rdev->mman.bdev, &robj->tobj, size, type, flags,
+				   0, 0, false, NULL, size,
+				   &radeon_ttm_object_object_destroy);
+	if (unlikely(r != 0)) {
+		/* ttm call radeon_ttm_object_object_destroy if error happen */
+		DRM_ERROR("Failed to allocate TTM object (%ld, 0x%08X, %u)\n",
+			  size, flags, 0);
+		return r;
+	}
+	*robj_ptr = robj;
+	if (gobj) {
+		list_add_tail(&robj->list, &rdev->gem.objects);
+	}
+	return 0;
+}
+
+int radeon_object_kmap(struct radeon_object *robj, void **ptr)
+{
+	int r;
+
+	spin_lock(&robj->tobj.lock);
+	if (robj->kptr) {
+		if (ptr) {
+			*ptr = robj->kptr;
+		}
+		spin_unlock(&robj->tobj.lock);
+		return 0;
+	}
+	spin_unlock(&robj->tobj.lock);
+	r = ttm_bo_kmap(&robj->tobj, 0, robj->tobj.num_pages, &robj->kmap);
+	if (r) {
+		return r;
+	}
+	spin_lock(&robj->tobj.lock);
+	robj->kptr = ttm_kmap_obj_virtual(&robj->kmap, &robj->is_iomem);
+	spin_unlock(&robj->tobj.lock);
+	if (ptr) {
+		*ptr = robj->kptr;
+	}
+	return 0;
+}
+
+void radeon_object_kunmap(struct radeon_object *robj)
+{
+	spin_lock(&robj->tobj.lock);
+	if (robj->kptr == NULL) {
+		spin_unlock(&robj->tobj.lock);
+		return;
+	}
+	robj->kptr = NULL;
+	spin_unlock(&robj->tobj.lock);
+	ttm_bo_kunmap(&robj->kmap);
+}
+
+void radeon_object_unref(struct radeon_object **robj)
+{
+	struct ttm_buffer_object *tobj;
+
+	if ((*robj) == NULL) {
+		return;
+	}
+	tobj = &((*robj)->tobj);
+	ttm_bo_unref(&tobj);
+	if (tobj == NULL) {
+		*robj = NULL;
+	}
+}
+
+int radeon_object_mmap(struct radeon_object *robj, uint64_t *offset)
+{
+	*offset = robj->tobj.addr_space_offset;
+	return 0;
+}
+
+int radeon_object_pin(struct radeon_object *robj, uint32_t domain,
+		      uint64_t *gpu_addr)
+{
+	uint32_t flags;
+	uint32_t tmp;
+	void *fbptr;
+	int r;
+
+	flags = radeon_object_flags_from_domain(domain);
+	spin_lock(&robj->tobj.lock);
+	if (robj->pin_count) {
+		robj->pin_count++;
+		if (gpu_addr != NULL) {
+			*gpu_addr = robj->gpu_addr;
+		}
+		spin_unlock(&robj->tobj.lock);
+		return 0;
+	}
+	spin_unlock(&robj->tobj.lock);
+	r = radeon_object_reserve(robj, false);
+	if (unlikely(r != 0)) {
+		DRM_ERROR("radeon: failed to reserve object for pinning it.\n");
+		return r;
+	}
+	if (robj->rdev->fbdev_robj == robj) {
+		mutex_lock(&robj->rdev->fbdev_info->lock);
+		radeon_object_kunmap(robj);
+	}
+	tmp = robj->tobj.mem.placement;
+	ttm_flag_masked(&tmp, flags, TTM_PL_MASK_MEM);
+	robj->tobj.proposed_placement = tmp | TTM_PL_FLAG_NO_EVICT | TTM_PL_MASK_CACHING;
+	r = ttm_buffer_object_validate(&robj->tobj,
+				       robj->tobj.proposed_placement,
+				       false, false);
+	radeon_object_gpu_addr(robj);
+	if (gpu_addr != NULL) {
+		*gpu_addr = robj->gpu_addr;
+	}
+	robj->pin_count = 1;
+	if (unlikely(r != 0)) {
+		DRM_ERROR("radeon: failed to pin object.\n");
+	}
+	radeon_object_unreserve(robj);
+	if (robj->rdev->fbdev_robj == robj) {
+		if (!r) {
+			r = radeon_object_kmap(robj, &fbptr);
+		}
+		if (!r) {
+			robj->rdev->fbdev_info->screen_base = fbptr;
+			robj->rdev->fbdev_info->fix.smem_start = (unsigned long)fbptr;
+		}
+		mutex_unlock(&robj->rdev->fbdev_info->lock);
+	}
+	return r;
+}
+
+void radeon_object_unpin(struct radeon_object *robj)
+{
+	uint32_t flags;
+	void *fbptr;
+	int r;
+
+	spin_lock(&robj->tobj.lock);
+	if (!robj->pin_count) {
+		spin_unlock(&robj->tobj.lock);
+		printk(KERN_WARNING "Unpin not necessary for %p !\n", robj);
+		return;
+	}
+	robj->pin_count--;
+	if (robj->pin_count) {
+		spin_unlock(&robj->tobj.lock);
+		return;
+	}
+	spin_unlock(&robj->tobj.lock);
+	r = radeon_object_reserve(robj, false);
+	if (unlikely(r != 0)) {
+		DRM_ERROR("radeon: failed to reserve object for unpinning it.\n");
+		return;
+	}
+	if (robj->rdev->fbdev_robj == robj) {
+		mutex_lock(&robj->rdev->fbdev_info->lock);
+		radeon_object_kunmap(robj);
+	}
+	flags = robj->tobj.mem.placement;
+	robj->tobj.proposed_placement = flags & ~TTM_PL_FLAG_NO_EVICT;
+	r = ttm_buffer_object_validate(&robj->tobj,
+				       robj->tobj.proposed_placement,
+				       false, false);
+	if (unlikely(r != 0)) {
+		DRM_ERROR("radeon: failed to unpin buffer.\n");
+	}
+	radeon_object_unreserve(robj);
+	if (robj->rdev->fbdev_robj == robj) {
+		if (!r) {
+			r = radeon_object_kmap(robj, &fbptr);
+		}
+		if (!r) {
+			robj->rdev->fbdev_info->screen_base = fbptr;
+			robj->rdev->fbdev_info->fix.smem_start = (unsigned long)fbptr;
+		}
+		mutex_unlock(&robj->rdev->fbdev_info->lock);
+	}
+}
+
+int radeon_object_wait(struct radeon_object *robj)
+{
+	int r = 0;
+
+	/* FIXME: should use block reservation instead */
+	r = radeon_object_reserve(robj, true);
+	if (unlikely(r != 0)) {
+		DRM_ERROR("radeon: failed to reserve object for waiting.\n");
+		return r;
+	}
+	spin_lock(&robj->tobj.lock);
+	if (robj->tobj.sync_obj) {
+		r = ttm_bo_wait(&robj->tobj, true, false, false);
+	}
+	spin_unlock(&robj->tobj.lock);
+	radeon_object_unreserve(robj);
+	return r;
+}
+
+int radeon_object_evict_vram(struct radeon_device *rdev)
+{
+	if (rdev->flags & RADEON_IS_IGP) {
+		/* Useless to evict on IGP chips */
+		return 0;
+	}
+	return ttm_bo_evict_mm(&rdev->mman.bdev, TTM_PL_VRAM);
+}
+
+void radeon_object_force_delete(struct radeon_device *rdev)
+{
+	struct radeon_object *robj, *n;
+	struct drm_gem_object *gobj;
+
+	if (list_empty(&rdev->gem.objects)) {
+		return;
+	}
+	DRM_ERROR("Userspace still has active objects !\n");
+	list_for_each_entry_safe(robj, n, &rdev->gem.objects, list) {
+		mutex_lock(&rdev->ddev->struct_mutex);
+		gobj = robj->gobj;
+		DRM_ERROR("Force free for (%p,%p,%lu,%lu)\n",
+			  gobj, robj, (unsigned long)gobj->size,
+			  *((unsigned long *)&gobj->refcount));
+		list_del_init(&robj->list);
+		radeon_object_unref(&robj);
+		gobj->driver_private = NULL;
+		drm_gem_object_unreference(gobj);
+		mutex_unlock(&rdev->ddev->struct_mutex);
+	}
+}
+
+int radeon_object_init(struct radeon_device *rdev)
+{
+	return radeon_ttm_init(rdev);
+}
+
+void radeon_object_fini(struct radeon_device *rdev)
+{
+	radeon_ttm_fini(rdev);
+}
+
+void radeon_object_list_add_object(struct radeon_object_list *lobj,
+				   struct list_head *head)
+{
+	if (lobj->wdomain) {
+		list_add(&lobj->list, head);
+	} else {
+		list_add_tail(&lobj->list, head);
+	}
+}
+
+int radeon_object_list_reserve(struct list_head *head)
+{
+	struct radeon_object_list *lobj;
+	struct list_head *i;
+	int r;
+
+	list_for_each(i, head) {
+		lobj = list_entry(i, struct radeon_object_list, list);
+		if (!lobj->robj->pin_count) {
+			r = radeon_object_reserve(lobj->robj, true);
+			if (unlikely(r != 0)) {
+				DRM_ERROR("radeon: failed to reserve object.\n");
+				return r;
+			}
+		} else {
+		}
+	}
+	return 0;
+}
+
+void radeon_object_list_unreserve(struct list_head *head)
+{
+	struct radeon_object_list *lobj;
+	struct list_head *i;
+
+	list_for_each(i, head) {
+		lobj = list_entry(i, struct radeon_object_list, list);
+		if (!lobj->robj->pin_count) {
+			radeon_object_unreserve(lobj->robj);
+		} else {
+		}
+	}
+}
+
+int radeon_object_list_validate(struct list_head *head, void *fence)
+{
+	struct radeon_object_list *lobj;
+	struct radeon_object *robj;
+	struct radeon_fence *old_fence = NULL;
+	struct list_head *i;
+	uint32_t flags;
+	int r;
+
+	r = radeon_object_list_reserve(head);
+	if (unlikely(r != 0)) {
+		radeon_object_list_unreserve(head);
+		return r;
+	}
+	list_for_each(i, head) {
+		lobj = list_entry(i, struct radeon_object_list, list);
+		robj = lobj->robj;
+		if (lobj->wdomain) {
+			flags = radeon_object_flags_from_domain(lobj->wdomain);
+			flags |= TTM_PL_FLAG_TT;
+		} else {
+			flags = radeon_object_flags_from_domain(lobj->rdomain);
+			flags |= TTM_PL_FLAG_TT;
+			flags |= TTM_PL_FLAG_VRAM;
+		}
+		if (!robj->pin_count) {
+			robj->tobj.proposed_placement = flags | TTM_PL_MASK_CACHING;
+			r = ttm_buffer_object_validate(&robj->tobj,
+						       robj->tobj.proposed_placement,
+						       true, false);
+			if (unlikely(r)) {
+				radeon_object_list_unreserve(head);
+				DRM_ERROR("radeon: failed to validate.\n");
+				return r;
+			}
+			radeon_object_gpu_addr(robj);
+		}
+		lobj->gpu_offset = robj->gpu_addr;
+		if (fence) {
+			old_fence = (struct radeon_fence *)robj->tobj.sync_obj;
+			robj->tobj.sync_obj = radeon_fence_ref(fence);
+			robj->tobj.sync_obj_arg = NULL;
+		}
+		if (old_fence) {
+			radeon_fence_unref(&old_fence);
+		}
+	}
+	return 0;
+}
+
+void radeon_object_list_unvalidate(struct list_head *head)
+{
+	struct radeon_object_list *lobj;
+	struct radeon_fence *old_fence = NULL;
+	struct list_head *i;
+
+	list_for_each(i, head) {
+		lobj = list_entry(i, struct radeon_object_list, list);
+		old_fence = (struct radeon_fence *)lobj->robj->tobj.sync_obj;
+		lobj->robj->tobj.sync_obj = NULL;
+		if (old_fence) {
+			radeon_fence_unref(&old_fence);
+		}
+	}
+	radeon_object_list_unreserve(head);
+}
+
+void radeon_object_list_clean(struct list_head *head)
+{
+	radeon_object_list_unreserve(head);
+}
+
+int radeon_object_fbdev_mmap(struct radeon_object *robj,
+			     struct vm_area_struct *vma)
+{
+	return ttm_fbdev_mmap(vma, &robj->tobj);
+}
+
+unsigned long radeon_object_size(struct radeon_object *robj)
+{
+	return robj->tobj.num_pages << PAGE_SHIFT;
+}