static cairo_surface_t *
radeon_surface_create_internal (cairo_drm_device_t *device,
cairo_format_t format,
int width, int height)
{
radeon_surface_t *surface;
cairo_status_t status;
surface = malloc (sizeof (radeon_surface_t));
if (unlikely (surface == NULL))
return _cairo_surface_create_in_error (_cairo_error (CAIRO_STATUS_NO_MEMORY));
radeon_surface_init (surface, device, format, width, height);
if (width && height) {
surface->base.stride =
cairo_format_stride_for_width (surface->base.format, width);
surface->base.bo = radeon_bo_create (to_radeon_device (&device->base),
surface->base.stride * height,
RADEON_GEM_DOMAIN_GTT);
if (unlikely (surface->base.bo == NULL)) {
status = _cairo_drm_surface_finish (&surface->base);
free (surface);
return _cairo_surface_create_in_error (_cairo_error (CAIRO_STATUS_NO_MEMORY));
}
}
return &surface->base.base;
}
int radeon_gem_object_create(struct radeon_device *rdev, int size,
int alignment, int initial_domain,
bool discardable, bool kernel,
struct drm_gem_object **obj)
{
struct radeon_bo *robj;
int r;
*obj = NULL;
/* At least align on page size */
if (alignment < PAGE_SIZE) {
alignment = PAGE_SIZE;
}
r = radeon_bo_create(rdev, size, alignment, kernel, initial_domain, &robj);
if (r) {
if (r != -ERESTARTSYS)
DRM_ERROR("Failed to allocate GEM object (%d, %d, %u, %d)\n",
size, initial_domain, alignment, r);
return r;
}
*obj = &robj->gem_base;
mutex_lock(&rdev->gem.mutex);
list_add_tail(&robj->list, &rdev->gem.objects);
mutex_unlock(&rdev->gem.mutex);
return 0;
}
int radeon_gem_object_create(struct radeon_device *rdev, int size,
int alignment, int initial_domain,
bool discardable, bool kernel,
struct drm_gem_object **obj)
{
struct drm_gem_object *gobj;
struct radeon_bo *robj;
int r;
*obj = NULL;
gobj = drm_gem_object_alloc(rdev->ddev, size);
if (!gobj) {
return -ENOMEM;
}
/* At least align on page size */
if (alignment < PAGE_SIZE) {
alignment = PAGE_SIZE;
}
r = radeon_bo_create(rdev, gobj, size, kernel, initial_domain, &robj);
if (r) {
if (r != -ERESTARTSYS)
DRM_ERROR("Failed to allocate GEM object (%d, %d, %u, %d)\n",
size, initial_domain, alignment, r);
drm_gem_object_unreference_unlocked(gobj);
return r;
}
gobj->driver_private = robj;
*obj = gobj;
return 0;
}
int radeon_ttm_init(struct radeon_device *rdev)
{
int r;
r = radeon_ttm_global_init(rdev);
if (r) {
return r;
}
/* No others user of address space so set it to 0 */
r = ttm_bo_device_init(&rdev->mman.bdev,
rdev->mman.bo_global_ref.ref.object,
&radeon_bo_driver, DRM_FILE_PAGE_OFFSET,
rdev->need_dma32);
if (r) {
DRM_ERROR("failed initializing buffer object driver(%d).\n", r);
return r;
}
rdev->mman.initialized = true;
r = ttm_bo_init_mm(&rdev->mman.bdev, TTM_PL_VRAM,
rdev->mc.real_vram_size >> PAGE_SHIFT);
if (r) {
DRM_ERROR("Failed initializing VRAM heap.\n");
return r;
}
r = radeon_bo_create(rdev, 256 * 1024, PAGE_SIZE, true,
RADEON_GEM_DOMAIN_VRAM,
&rdev->stollen_vga_memory);
if (r) {
return r;
}
r = radeon_bo_reserve(rdev->stollen_vga_memory, false);
if (r)
return r;
r = radeon_bo_pin(rdev->stollen_vga_memory, RADEON_GEM_DOMAIN_VRAM, NULL);
radeon_bo_unreserve(rdev->stollen_vga_memory);
if (r) {
radeon_bo_unref(&rdev->stollen_vga_memory);
return r;
}
DRM_INFO("radeon: %uM of VRAM memory ready\n",
(unsigned)rdev->mc.real_vram_size / (1024 * 1024));
r = ttm_bo_init_mm(&rdev->mman.bdev, TTM_PL_TT,
rdev->mc.gtt_size >> PAGE_SHIFT);
if (r) {
DRM_ERROR("Failed initializing GTT heap.\n");
return r;
}
DRM_INFO("radeon: %uM of GTT memory ready.\n",
(unsigned)(rdev->mc.gtt_size / (1024 * 1024)));
if (unlikely(rdev->mman.bdev.dev_mapping == NULL)) {
rdev->mman.bdev.dev_mapping = rdev->ddev->dev_mapping;
}
r = radeon_ttm_debugfs_init(rdev);
if (r) {
DRM_ERROR("Failed to init debugfs\n");
return r;
}
return 0;
}
int create_bitmap(bitmap_t **pbitmap, int width, int height)
{
size_t size;
size_t pitch;
bitmap_t *bitmap;
uint64_t gaddr;
void *uaddr;
struct radeon_device *rdev = main_device;
struct radeon_bo *sobj = NULL;
int r;
bitmap = CreateObject(GetPid(), sizeof(bitmap_t));
if( bitmap == NULL)
{
*pbitmap = NULL;
return -1;
}
pitch = radeon_align_pitch(rdev, width, 32, false) * 4;
size = pitch * height;
r = radeon_bo_create(rdev, size, PAGE_SIZE, true,
RADEON_GEM_DOMAIN_GTT, &sobj);
if (r) {
goto fail;
}
r = radeon_bo_reserve(sobj, false);
if (unlikely(r != 0))
goto fail;
r = radeon_bo_pin(sobj, RADEON_GEM_DOMAIN_GTT, &gaddr);
if (r) {
goto fail;
}
r = radeon_bo_user_map(sobj, &uaddr);
if (r) {
goto fail;
}
bitmap->width = width;
bitmap->height = height;
bitmap->stride = pitch;
bitmap->gaddr = gaddr;
bitmap->uaddr = uaddr;
bitmap->robj = sobj;
*pbitmap = bitmap;
return 0;
fail:
DestroyObject(bitmap);
return -1;
};
/**
* radeon_ring_init - init driver ring struct.
*
* @rdev: radeon_device pointer
* @ring: radeon_ring structure holding ring information
* @ring_size: size of the ring
* @rptr_offs: offset of the rptr writeback location in the WB buffer
* @rptr_reg: MMIO offset of the rptr register
* @wptr_reg: MMIO offset of the wptr register
* @ptr_reg_shift: bit offset of the rptr/wptr values
* @ptr_reg_mask: bit mask of the rptr/wptr values
* @nop: nop packet for this ring
*
* Initialize the driver information for the selected ring (all asics).
* Returns 0 on success, error on failure.
*/
int radeon_ring_init(struct radeon_device *rdev, struct radeon_ring *ring, unsigned ring_size,
unsigned rptr_offs, unsigned rptr_reg, unsigned wptr_reg,
u32 ptr_reg_shift, u32 ptr_reg_mask, u32 nop)
{
int r;
ring->ring_size = ring_size;
ring->rptr_offs = rptr_offs;
ring->rptr_reg = rptr_reg;
ring->wptr_reg = wptr_reg;
ring->ptr_reg_shift = ptr_reg_shift;
ring->ptr_reg_mask = ptr_reg_mask;
ring->nop = nop;
/* Allocate ring buffer */
if (ring->ring_obj == NULL) {
r = radeon_bo_create(rdev, ring->ring_size, PAGE_SIZE, true,
RADEON_GEM_DOMAIN_GTT,
NULL, &ring->ring_obj);
if (r) {
dev_err(rdev->dev, "(%d) ring create failed\n", r);
return r;
}
r = radeon_bo_reserve(ring->ring_obj, false);
if (unlikely(r != 0))
return r;
r = radeon_bo_pin(ring->ring_obj, RADEON_GEM_DOMAIN_GTT,
&ring->gpu_addr);
if (r) {
radeon_bo_unreserve(ring->ring_obj);
dev_err(rdev->dev, "(%d) ring pin failed\n", r);
return r;
}
r = radeon_bo_kmap(ring->ring_obj,
(void **)&ring->ring);
radeon_bo_unreserve(ring->ring_obj);
if (r) {
dev_err(rdev->dev, "(%d) ring map failed\n", r);
return r;
}
}
ring->ptr_mask = (ring->ring_size / 4) - 1;
ring->ring_free_dw = ring->ring_size / 4;
if (rdev->wb.enabled) {
u32 index = RADEON_WB_RING0_NEXT_RPTR + (ring->idx * 4);
ring->next_rptr_gpu_addr = rdev->wb.gpu_addr + index;
ring->next_rptr_cpu_addr = &rdev->wb.wb[index/4];
}
if (radeon_debugfs_ring_init(rdev, ring)) {
DRM_ERROR("Failed to register debugfs file for rings !\n");
}
radeon_ring_lockup_update(ring);
return 0;
}
struct drm_gem_object *radeon_gem_prime_import_sg_table(struct drm_device *dev,
size_t size,
struct sg_table *sg)
{
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, &bo);
if (ret)
return ERR_PTR(ret);
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 &bo->gem_base;
}
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 drm_gem_object *radeon_gem_prime_import_sg_table(struct drm_device *dev,
struct dma_buf_attachment *attach,
struct sg_table *sg)
{
struct reservation_object *resv = attach->dmabuf->resv;
struct radeon_device *rdev = dev->dev_private;
struct radeon_bo *bo;
int ret;
ww_mutex_lock(&resv->lock, NULL);
ret = radeon_bo_create(rdev, attach->dmabuf->size, PAGE_SIZE, false,
RADEON_GEM_DOMAIN_GTT, 0, sg, resv, &bo);
ww_mutex_unlock(&resv->lock);
if (ret)
return ERR_PTR(ret);
mutex_lock(&rdev->gem.mutex);
list_add_tail(&bo->list, &rdev->gem.objects);
mutex_unlock(&rdev->gem.mutex);
bo->prime_shared_count = 1;
return &bo->gem_base;
}
int radeon_uvd_init(struct radeon_device *rdev)
{
struct platform_device *pdev;
unsigned long bo_size;
const char *fw_name;
int i, r;
INIT_DELAYED_WORK(&rdev->uvd.idle_work, radeon_uvd_idle_work_handler);
pdev = platform_device_register_simple("radeon_uvd", 0, NULL, 0);
r = IS_ERR(pdev);
if (r) {
dev_err(rdev->dev, "radeon_uvd: Failed to register firmware\n");
return -EINVAL;
}
switch (rdev->family) {
case CHIP_RV710:
case CHIP_RV730:
case CHIP_RV740:
fw_name = FIRMWARE_RV710;
break;
case CHIP_CYPRESS:
case CHIP_HEMLOCK:
case CHIP_JUNIPER:
case CHIP_REDWOOD:
case CHIP_CEDAR:
fw_name = FIRMWARE_CYPRESS;
break;
case CHIP_SUMO:
case CHIP_SUMO2:
case CHIP_PALM:
case CHIP_CAYMAN:
case CHIP_BARTS:
case CHIP_TURKS:
case CHIP_CAICOS:
fw_name = FIRMWARE_SUMO;
break;
case CHIP_TAHITI:
case CHIP_VERDE:
case CHIP_PITCAIRN:
case CHIP_ARUBA:
fw_name = FIRMWARE_TAHITI;
break;
default:
return -EINVAL;
}
r = request_firmware(&rdev->uvd_fw, fw_name, &pdev->dev);
if (r) {
dev_err(rdev->dev, "radeon_uvd: Can't load firmware \"%s\"\n",
fw_name);
platform_device_unregister(pdev);
return r;
}
platform_device_unregister(pdev);
bo_size = RADEON_GPU_PAGE_ALIGN(rdev->uvd_fw->size + 8) +
RADEON_UVD_STACK_SIZE + RADEON_UVD_HEAP_SIZE;
r = radeon_bo_create(rdev, bo_size, PAGE_SIZE, true,
RADEON_GEM_DOMAIN_VRAM, NULL, &rdev->uvd.vcpu_bo);
if (r) {
dev_err(rdev->dev, "(%d) failed to allocate UVD bo\n", r);
return r;
}
r = radeon_uvd_resume(rdev);
if (r)
return r;
memset(rdev->uvd.cpu_addr, 0, bo_size);
memcpy(rdev->uvd.cpu_addr, rdev->uvd_fw->data, rdev->uvd_fw->size);
r = radeon_uvd_suspend(rdev);
if (r)
return r;
for (i = 0; i < RADEON_MAX_UVD_HANDLES; ++i) {
atomic_set(&rdev->uvd.handles[i], 0);
rdev->uvd.filp[i] = NULL;
}
return 0;
}
/* Test BO GTT->VRAM and VRAM->GTT GPU copies across the whole GTT aperture */
void radeon_test_moves(struct radeon_device *rdev)
{
struct radeon_bo *vram_obj = NULL;
struct radeon_bo **gtt_obj = NULL;
struct radeon_fence *fence = NULL;
uint64_t gtt_addr, vram_addr;
unsigned i, n, size;
int r;
size = 1024 * 1024;
/* Number of tests =
* (Total GTT - IB pool - writeback page - ring buffers) / test size
*/
n = rdev->mc.gtt_size - RADEON_IB_POOL_SIZE*64*1024;
for (i = 0; i < RADEON_NUM_RINGS; ++i)
n -= rdev->ring[i].ring_size;
if (rdev->wb.wb_obj)
n -= RADEON_GPU_PAGE_SIZE;
if (rdev->ih.ring_obj)
n -= rdev->ih.ring_size;
n /= size;
gtt_obj = kzalloc(n * sizeof(*gtt_obj), GFP_KERNEL);
if (!gtt_obj) {
DRM_ERROR("Failed to allocate %d pointers\n", n);
r = 1;
goto out_cleanup;
}
r = radeon_bo_create(rdev, size, PAGE_SIZE, true, RADEON_GEM_DOMAIN_VRAM,
&vram_obj);
if (r) {
DRM_ERROR("Failed to create VRAM object\n");
goto out_cleanup;
}
r = radeon_bo_reserve(vram_obj, false);
if (unlikely(r != 0))
goto out_cleanup;
r = radeon_bo_pin(vram_obj, RADEON_GEM_DOMAIN_VRAM, &vram_addr);
if (r) {
DRM_ERROR("Failed to pin VRAM object\n");
goto out_cleanup;
}
for (i = 0; i < n; i++) {
void *gtt_map, *vram_map;
void **gtt_start, **gtt_end;
void **vram_start, **vram_end;
r = radeon_bo_create(rdev, size, PAGE_SIZE, true,
RADEON_GEM_DOMAIN_GTT, gtt_obj + i);
if (r) {
DRM_ERROR("Failed to create GTT object %d\n", i);
goto out_cleanup;
}
r = radeon_bo_reserve(gtt_obj[i], false);
if (unlikely(r != 0))
goto out_cleanup;
r = radeon_bo_pin(gtt_obj[i], RADEON_GEM_DOMAIN_GTT, >t_addr);
if (r) {
DRM_ERROR("Failed to pin GTT object %d\n", i);
goto out_cleanup;
}
r = radeon_bo_kmap(gtt_obj[i], >t_map);
if (r) {
DRM_ERROR("Failed to map GTT object %d\n", i);
goto out_cleanup;
}
for (gtt_start = gtt_map, gtt_end = gtt_map + size;
gtt_start < gtt_end;
gtt_start++)
*gtt_start = gtt_start;
radeon_bo_kunmap(gtt_obj[i]);
r = radeon_fence_create(rdev, &fence, RADEON_RING_TYPE_GFX_INDEX);
if (r) {
DRM_ERROR("Failed to create GTT->VRAM fence %d\n", i);
goto out_cleanup;
}
r = radeon_copy(rdev, gtt_addr, vram_addr, size / RADEON_GPU_PAGE_SIZE, fence);
if (r) {
DRM_ERROR("Failed GTT->VRAM copy %d\n", i);
goto out_cleanup;
}
r = radeon_fence_wait(fence, false);
if (r) {
DRM_ERROR("Failed to wait for GTT->VRAM fence %d\n", i);
goto out_cleanup;
}
radeon_fence_unref(&fence);
r = radeon_bo_kmap(vram_obj, &vram_map);
if (r) {
DRM_ERROR("Failed to map VRAM object after copy %d\n", i);
goto out_cleanup;
}
for (gtt_start = gtt_map, gtt_end = gtt_map + size,
vram_start = vram_map, vram_end = vram_map + size;
vram_start < vram_end;
gtt_start++, vram_start++) {
if (*vram_start != gtt_start) {
DRM_ERROR("Incorrect GTT->VRAM copy %d: Got 0x%p, "
"expected 0x%p (GTT/VRAM offset "
"0x%16llx/0x%16llx)\n",
i, *vram_start, gtt_start,
(unsigned long long)
(gtt_addr - rdev->mc.gtt_start +
(void*)gtt_start - gtt_map),
(unsigned long long)
(vram_addr - rdev->mc.vram_start +
(void*)gtt_start - gtt_map));
radeon_bo_kunmap(vram_obj);
goto out_cleanup;
}
*vram_start = vram_start;
}
radeon_bo_kunmap(vram_obj);
r = radeon_fence_create(rdev, &fence, RADEON_RING_TYPE_GFX_INDEX);
if (r) {
DRM_ERROR("Failed to create VRAM->GTT fence %d\n", i);
goto out_cleanup;
}
r = radeon_copy(rdev, vram_addr, gtt_addr, size / RADEON_GPU_PAGE_SIZE, fence);
if (r) {
DRM_ERROR("Failed VRAM->GTT copy %d\n", i);
goto out_cleanup;
}
r = radeon_fence_wait(fence, false);
if (r) {
DRM_ERROR("Failed to wait for VRAM->GTT fence %d\n", i);
goto out_cleanup;
}
radeon_fence_unref(&fence);
r = radeon_bo_kmap(gtt_obj[i], >t_map);
if (r) {
DRM_ERROR("Failed to map GTT object after copy %d\n", i);
goto out_cleanup;
}
for (gtt_start = gtt_map, gtt_end = gtt_map + size,
vram_start = vram_map, vram_end = vram_map + size;
gtt_start < gtt_end;
gtt_start++, vram_start++) {
if (*gtt_start != vram_start) {
DRM_ERROR("Incorrect VRAM->GTT copy %d: Got 0x%p, "
"expected 0x%p (VRAM/GTT offset "
"0x%16llx/0x%16llx)\n",
i, *gtt_start, vram_start,
(unsigned long long)
(vram_addr - rdev->mc.vram_start +
(void*)vram_start - vram_map),
(unsigned long long)
(gtt_addr - rdev->mc.gtt_start +
(void*)vram_start - vram_map));
radeon_bo_kunmap(gtt_obj[i]);
goto out_cleanup;
}
}
radeon_bo_kunmap(gtt_obj[i]);
DRM_INFO("Tested GTT->VRAM and VRAM->GTT copy for GTT offset 0x%llx\n",
gtt_addr - rdev->mc.gtt_start);
}
out_cleanup:
if (vram_obj) {
if (radeon_bo_is_reserved(vram_obj)) {
radeon_bo_unpin(vram_obj);
radeon_bo_unreserve(vram_obj);
}
radeon_bo_unref(&vram_obj);
}
if (gtt_obj) {
for (i = 0; i < n; i++) {
if (gtt_obj[i]) {
if (radeon_bo_is_reserved(gtt_obj[i])) {
radeon_bo_unpin(gtt_obj[i]);
radeon_bo_unreserve(gtt_obj[i]);
}
radeon_bo_unref(>t_obj[i]);
}
}
kfree(gtt_obj);
}
if (fence) {
radeon_fence_unref(&fence);
}
if (r) {
printk(KERN_WARNING "Error while testing BO move.\n");
}
}
/**
* radeon_vce_init - allocate memory, load vce firmware
*
* @rdev: radeon_device pointer
*
* First step to get VCE online, allocate memory and load the firmware
*/
int radeon_vce_init(struct radeon_device *rdev)
{
static const char *fw_version = "[ATI LIB=VCEFW,";
static const char *fb_version = "[ATI LIB=VCEFWSTATS,";
unsigned long size;
const char *fw_name, *c;
uint8_t start, mid, end;
int i, r;
INIT_DELAYED_WORK(&rdev->vce.idle_work, radeon_vce_idle_work_handler);
switch (rdev->family) {
case CHIP_BONAIRE:
case CHIP_KAVERI:
case CHIP_KABINI:
fw_name = FIRMWARE_BONAIRE;
break;
default:
return -EINVAL;
}
r = request_firmware(&rdev->vce_fw, fw_name, rdev->dev);
if (r) {
dev_err(rdev->dev, "radeon_vce: Can't load firmware \"%s\"\n",
fw_name);
return r;
}
/* search for firmware version */
size = rdev->vce_fw->size - strlen(fw_version) - 9;
c = rdev->vce_fw->data;
for (;size > 0; --size, ++c)
if (strncmp(c, fw_version, strlen(fw_version)) == 0)
break;
if (size == 0)
return -EINVAL;
c += strlen(fw_version);
if (sscanf(c, "%2hhd.%2hhd.%2hhd]", &start, &mid, &end) != 3)
return -EINVAL;
/* search for feedback version */
size = rdev->vce_fw->size - strlen(fb_version) - 3;
c = rdev->vce_fw->data;
for (;size > 0; --size, ++c)
if (strncmp(c, fb_version, strlen(fb_version)) == 0)
break;
if (size == 0)
return -EINVAL;
c += strlen(fb_version);
if (sscanf(c, "%2u]", &rdev->vce.fb_version) != 1)
return -EINVAL;
DRM_INFO("Found VCE firmware/feedback version %hhd.%hhd.%hhd / %d!\n",
start, mid, end, rdev->vce.fb_version);
rdev->vce.fw_version = (start << 24) | (mid << 16) | (end << 8);
/* we can only work with this fw version for now */
if (rdev->vce.fw_version != ((40 << 24) | (2 << 16) | (2 << 8)))
return -EINVAL;
/* allocate firmware, stack and heap BO */
size = RADEON_GPU_PAGE_ALIGN(rdev->vce_fw->size) +
RADEON_VCE_STACK_SIZE + RADEON_VCE_HEAP_SIZE;
r = radeon_bo_create(rdev, size, PAGE_SIZE, true,
RADEON_GEM_DOMAIN_VRAM, NULL, &rdev->vce.vcpu_bo);
if (r) {
dev_err(rdev->dev, "(%d) failed to allocate VCE bo\n", r);
return r;
}
r = radeon_bo_reserve(rdev->vce.vcpu_bo, false);
if (r) {
radeon_bo_unref(&rdev->vce.vcpu_bo);
dev_err(rdev->dev, "(%d) failed to reserve VCE bo\n", r);
return r;
}
r = radeon_bo_pin(rdev->vce.vcpu_bo, RADEON_GEM_DOMAIN_VRAM,
&rdev->vce.gpu_addr);
radeon_bo_unreserve(rdev->vce.vcpu_bo);
if (r) {
radeon_bo_unref(&rdev->vce.vcpu_bo);
dev_err(rdev->dev, "(%d) VCE bo pin failed\n", r);
return r;
}
for (i = 0; i < RADEON_MAX_VCE_HANDLES; ++i) {
atomic_set(&rdev->vce.handles[i], 0);
rdev->vce.filp[i] = NULL;
}
return 0;
}
/* Test BO GTT->VRAM and VRAM->GTT GPU copies across the whole GTT aperture */
static void radeon_do_test_moves(struct radeon_device *rdev, int flag)
{
struct radeon_bo *vram_obj = NULL;
struct radeon_bo **gtt_obj = NULL;
uint64_t gtt_addr, vram_addr;
unsigned n, size;
int i, r, ring;
switch (flag) {
case RADEON_TEST_COPY_DMA:
ring = radeon_copy_dma_ring_index(rdev);
break;
case RADEON_TEST_COPY_BLIT:
ring = radeon_copy_blit_ring_index(rdev);
break;
default:
DRM_ERROR("Unknown copy method\n");
return;
}
size = 1024 * 1024;
/* Number of tests =
* (Total GTT - IB pool - writeback page - ring buffers) / test size
*/
n = rdev->mc.gtt_size - RADEON_IB_POOL_SIZE*64*1024;
for (i = 0; i < RADEON_NUM_RINGS; ++i)
n -= rdev->ring[i].ring_size;
if (rdev->wb.wb_obj)
n -= RADEON_GPU_PAGE_SIZE;
if (rdev->ih.ring_obj)
n -= rdev->ih.ring_size;
n /= size;
gtt_obj = kzalloc(n * sizeof(*gtt_obj), GFP_KERNEL);
if (!gtt_obj) {
DRM_ERROR("Failed to allocate %d pointers\n", n);
r = 1;
goto out_cleanup;
}
r = radeon_bo_create(rdev, size, PAGE_SIZE, true, RADEON_GEM_DOMAIN_VRAM,
NULL, &vram_obj);
if (r) {
DRM_ERROR("Failed to create VRAM object\n");
goto out_cleanup;
}
r = radeon_bo_reserve(vram_obj, false);
if (unlikely(r != 0))
goto out_unref;
r = radeon_bo_pin(vram_obj, RADEON_GEM_DOMAIN_VRAM, &vram_addr);
if (r) {
DRM_ERROR("Failed to pin VRAM object\n");
goto out_unres;
}
for (i = 0; i < n; i++) {
void *gtt_map, *vram_map;
void **gtt_start, **gtt_end;
void **vram_start, **vram_end;
struct radeon_fence *fence = NULL;
r = radeon_bo_create(rdev, size, PAGE_SIZE, true,
RADEON_GEM_DOMAIN_GTT, NULL, gtt_obj + i);
if (r) {
DRM_ERROR("Failed to create GTT object %d\n", i);
goto out_lclean;
}
r = radeon_bo_reserve(gtt_obj[i], false);
if (unlikely(r != 0))
goto out_lclean_unref;
r = radeon_bo_pin(gtt_obj[i], RADEON_GEM_DOMAIN_GTT, >t_addr);
if (r) {
DRM_ERROR("Failed to pin GTT object %d\n", i);
goto out_lclean_unres;
}
r = radeon_bo_kmap(gtt_obj[i], >t_map);
if (r) {
DRM_ERROR("Failed to map GTT object %d\n", i);
goto out_lclean_unpin;
}
for (gtt_start = gtt_map, gtt_end = gtt_map + size;
gtt_start < gtt_end;
gtt_start++)
*gtt_start = gtt_start;
radeon_bo_kunmap(gtt_obj[i]);
if (ring == R600_RING_TYPE_DMA_INDEX)
r = radeon_copy_dma(rdev, gtt_addr, vram_addr, size / RADEON_GPU_PAGE_SIZE, &fence);
else
r = radeon_copy_blit(rdev, gtt_addr, vram_addr, size / RADEON_GPU_PAGE_SIZE, &fence);
if (r) {
DRM_ERROR("Failed GTT->VRAM copy %d\n", i);
goto out_lclean_unpin;
}
r = radeon_fence_wait(fence, false);
if (r) {
DRM_ERROR("Failed to wait for GTT->VRAM fence %d\n", i);
goto out_lclean_unpin;
}
radeon_fence_unref(&fence);
r = radeon_bo_kmap(vram_obj, &vram_map);
if (r) {
DRM_ERROR("Failed to map VRAM object after copy %d\n", i);
goto out_lclean_unpin;
}
for (gtt_start = gtt_map, gtt_end = gtt_map + size,
vram_start = vram_map, vram_end = vram_map + size;
vram_start < vram_end;
gtt_start++, vram_start++) {
if (*vram_start != gtt_start) {
DRM_ERROR("Incorrect GTT->VRAM copy %d: Got 0x%p, "
"expected 0x%p (GTT/VRAM offset "
"0x%16llx/0x%16llx)\n",
i, *vram_start, gtt_start,
(unsigned long long)
(gtt_addr - rdev->mc.gtt_start +
(void*)gtt_start - gtt_map),
(unsigned long long)
(vram_addr - rdev->mc.vram_start +
(void*)gtt_start - gtt_map));
radeon_bo_kunmap(vram_obj);
goto out_lclean_unpin;
}
*vram_start = vram_start;
}
radeon_bo_kunmap(vram_obj);
if (ring == R600_RING_TYPE_DMA_INDEX)
r = radeon_copy_dma(rdev, vram_addr, gtt_addr, size / RADEON_GPU_PAGE_SIZE, &fence);
else
r = radeon_copy_blit(rdev, vram_addr, gtt_addr, size / RADEON_GPU_PAGE_SIZE, &fence);
if (r) {
DRM_ERROR("Failed VRAM->GTT copy %d\n", i);
goto out_lclean_unpin;
}
r = radeon_fence_wait(fence, false);
if (r) {
DRM_ERROR("Failed to wait for VRAM->GTT fence %d\n", i);
goto out_lclean_unpin;
}
radeon_fence_unref(&fence);
r = radeon_bo_kmap(gtt_obj[i], >t_map);
if (r) {
DRM_ERROR("Failed to map GTT object after copy %d\n", i);
goto out_lclean_unpin;
}
for (gtt_start = gtt_map, gtt_end = gtt_map + size,
vram_start = vram_map, vram_end = vram_map + size;
gtt_start < gtt_end;
gtt_start++, vram_start++) {
if (*gtt_start != vram_start) {
DRM_ERROR("Incorrect VRAM->GTT copy %d: Got 0x%p, "
"expected 0x%p (VRAM/GTT offset "
"0x%16llx/0x%16llx)\n",
i, *gtt_start, vram_start,
(unsigned long long)
(vram_addr - rdev->mc.vram_start +
(void*)vram_start - vram_map),
(unsigned long long)
(gtt_addr - rdev->mc.gtt_start +
(void*)vram_start - vram_map));
radeon_bo_kunmap(gtt_obj[i]);
goto out_lclean_unpin;
}
}
radeon_bo_kunmap(gtt_obj[i]);
DRM_INFO("Tested GTT->VRAM and VRAM->GTT copy for GTT offset 0x%llx\n",
gtt_addr - rdev->mc.gtt_start);
continue;
out_lclean_unpin:
radeon_bo_unpin(gtt_obj[i]);
out_lclean_unres:
radeon_bo_unreserve(gtt_obj[i]);
out_lclean_unref:
radeon_bo_unref(>t_obj[i]);
out_lclean:
for (--i; i >= 0; --i) {
radeon_bo_unpin(gtt_obj[i]);
radeon_bo_unreserve(gtt_obj[i]);
radeon_bo_unref(>t_obj[i]);
}
if (fence)
radeon_fence_unref(&fence);
break;
}
radeon_bo_unpin(vram_obj);
out_unres:
radeon_bo_unreserve(vram_obj);
out_unref:
radeon_bo_unref(&vram_obj);
out_cleanup:
kfree(gtt_obj);
if (r) {
printk(KERN_WARNING "Error while testing BO move.\n");
}
}
int radeon_uvd_init(struct radeon_device *rdev)
{
unsigned long bo_size;
const char *fw_name;
int i, r;
INIT_DELAYED_WORK(&rdev->uvd.idle_work, radeon_uvd_idle_work_handler);
switch (rdev->family) {
case CHIP_RV710:
case CHIP_RV730:
case CHIP_RV740:
fw_name = FIRMWARE_RV710;
break;
case CHIP_CYPRESS:
case CHIP_HEMLOCK:
case CHIP_JUNIPER:
case CHIP_REDWOOD:
case CHIP_CEDAR:
fw_name = FIRMWARE_CYPRESS;
break;
case CHIP_SUMO:
case CHIP_SUMO2:
case CHIP_PALM:
case CHIP_CAYMAN:
case CHIP_BARTS:
case CHIP_TURKS:
case CHIP_CAICOS:
fw_name = FIRMWARE_SUMO;
break;
case CHIP_TAHITI:
case CHIP_VERDE:
case CHIP_PITCAIRN:
case CHIP_ARUBA:
fw_name = FIRMWARE_TAHITI;
break;
case CHIP_BONAIRE:
case CHIP_KABINI:
case CHIP_KAVERI:
fw_name = FIRMWARE_BONAIRE;
break;
default:
return -EINVAL;
}
r = request_firmware(&rdev->uvd_fw, fw_name, rdev->dev);
if (r) {
dev_err(rdev->dev, "radeon_uvd: Can't load firmware \"%s\"\n",
fw_name);
return r;
}
bo_size = RADEON_GPU_PAGE_ALIGN(rdev->uvd_fw->size + 8) +
RADEON_UVD_STACK_SIZE + RADEON_UVD_HEAP_SIZE;
r = radeon_bo_create(rdev, bo_size, PAGE_SIZE, true,
RADEON_GEM_DOMAIN_VRAM, NULL, &rdev->uvd.vcpu_bo);
if (r) {
dev_err(rdev->dev, "(%d) failed to allocate UVD bo\n", r);
return r;
}
r = radeon_bo_reserve(rdev->uvd.vcpu_bo, false);
if (r) {
radeon_bo_unref(&rdev->uvd.vcpu_bo);
dev_err(rdev->dev, "(%d) failed to reserve UVD bo\n", r);
return r;
}
r = radeon_bo_pin(rdev->uvd.vcpu_bo, RADEON_GEM_DOMAIN_VRAM,
&rdev->uvd.gpu_addr);
if (r) {
radeon_bo_unreserve(rdev->uvd.vcpu_bo);
radeon_bo_unref(&rdev->uvd.vcpu_bo);
dev_err(rdev->dev, "(%d) UVD bo pin failed\n", r);
return r;
}
r = radeon_bo_kmap(rdev->uvd.vcpu_bo, &rdev->uvd.cpu_addr);
if (r) {
dev_err(rdev->dev, "(%d) UVD map failed\n", r);
return r;
}
radeon_bo_unreserve(rdev->uvd.vcpu_bo);
for (i = 0; i < RADEON_MAX_UVD_HANDLES; ++i) {
atomic_set(&rdev->uvd.handles[i], 0);
rdev->uvd.filp[i] = NULL;
}
return 0;
}
static void radeon_benchmark_move(struct radeon_device *rdev, unsigned size,
unsigned sdomain, unsigned ddomain)
{
struct radeon_bo *dobj = NULL;
struct radeon_bo *sobj = NULL;
uint64_t saddr, daddr;
int r, n;
int time;
n = RADEON_BENCHMARK_ITERATIONS;
r = radeon_bo_create(rdev, size, PAGE_SIZE, true, sdomain, NULL, &sobj);
if (r) {
goto out_cleanup;
}
r = radeon_bo_reserve(sobj, false);
if (unlikely(r != 0))
goto out_cleanup;
r = radeon_bo_pin(sobj, sdomain, &saddr);
radeon_bo_unreserve(sobj);
if (r) {
goto out_cleanup;
}
r = radeon_bo_create(rdev, size, PAGE_SIZE, true, ddomain, NULL, &dobj);
if (r) {
goto out_cleanup;
}
r = radeon_bo_reserve(dobj, false);
if (unlikely(r != 0))
goto out_cleanup;
r = radeon_bo_pin(dobj, ddomain, &daddr);
radeon_bo_unreserve(dobj);
if (r) {
goto out_cleanup;
}
/* r100 doesn't have dma engine so skip the test */
/* also, VRAM-to-VRAM test doesn't make much sense for DMA */
/* skip it as well if domains are the same */
if ((rdev->asic->copy.dma) && (sdomain != ddomain)) {
time = radeon_benchmark_do_move(rdev, size, saddr, daddr,
RADEON_BENCHMARK_COPY_DMA, n);
if (time < 0)
goto out_cleanup;
if (time > 0)
radeon_benchmark_log_results(n, size, time,
sdomain, ddomain, "dma");
}
time = radeon_benchmark_do_move(rdev, size, saddr, daddr,
RADEON_BENCHMARK_COPY_BLIT, n);
if (time < 0)
goto out_cleanup;
if (time > 0)
radeon_benchmark_log_results(n, size, time,
sdomain, ddomain, "blit");
out_cleanup:
if (sobj) {
r = radeon_bo_reserve(sobj, false);
if (likely(r == 0)) {
radeon_bo_unpin(sobj);
radeon_bo_unreserve(sobj);
}
radeon_bo_unref(&sobj);
}
if (dobj) {
r = radeon_bo_reserve(dobj, false);
if (likely(r == 0)) {
radeon_bo_unpin(dobj);
radeon_bo_unreserve(dobj);
}
radeon_bo_unref(&dobj);
}
if (r) {
DRM_ERROR("Error while benchmarking BO move.\n");
}
}
int radeon_ttm_init(struct radeon_device *rdev)
{
int r;
r = radeon_ttm_global_init(rdev);
if (r) {
return r;
}
/* No others user of address space so set it to 0 */
r = ttm_bo_device_init(&rdev->mman.bdev,
rdev->mman.bo_global_ref.ref.object,
&radeon_bo_driver, DRM_FILE_PAGE_OFFSET,
rdev->need_dma32);
if (r) {
DRM_ERROR("failed initializing buffer object driver(%d).\n", r);
return r;
}
rdev->mman.bdev.iot = rdev->iot;
rdev->mman.bdev.memt = rdev->memt;
rdev->mman.bdev.dmat = rdev->dmat;
rdev->mman.initialized = true;
r = ttm_bo_init_mm(&rdev->mman.bdev, TTM_PL_VRAM,
rdev->mc.real_vram_size >> PAGE_SHIFT);
if (r) {
DRM_ERROR("Failed initializing VRAM heap.\n");
return r;
}
/* Change the size here instead of the init above so only lpfn is affected */
radeon_ttm_set_active_vram_size(rdev, rdev->mc.visible_vram_size);
#ifdef __sparc64__
r = radeon_bo_create(rdev, rdev->fb_offset, PAGE_SIZE, true,
RADEON_GEM_DOMAIN_VRAM,
NULL, &rdev->stollen_vga_memory);
#else
r = radeon_bo_create(rdev, 256 * 1024, PAGE_SIZE, true,
RADEON_GEM_DOMAIN_VRAM,
NULL, &rdev->stollen_vga_memory);
#endif
if (r) {
return r;
}
r = radeon_bo_reserve(rdev->stollen_vga_memory, false);
if (r)
return r;
r = radeon_bo_pin(rdev->stollen_vga_memory, RADEON_GEM_DOMAIN_VRAM, NULL);
radeon_bo_unreserve(rdev->stollen_vga_memory);
if (r) {
radeon_bo_unref(&rdev->stollen_vga_memory);
return r;
}
DRM_INFO("radeon: %uM of VRAM memory ready\n",
(unsigned) (rdev->mc.real_vram_size / (1024 * 1024)));
r = ttm_bo_init_mm(&rdev->mman.bdev, TTM_PL_TT,
rdev->mc.gtt_size >> PAGE_SHIFT);
if (r) {
DRM_ERROR("Failed initializing GTT heap.\n");
return r;
}
DRM_INFO("radeon: %uM of GTT memory ready.\n",
(unsigned)(rdev->mc.gtt_size / (1024 * 1024)));
#ifdef notyet
rdev->mman.bdev.dev_mapping = rdev->ddev->dev_mapping;
#endif
r = radeon_ttm_debugfs_init(rdev);
if (r) {
DRM_ERROR("Failed to init debugfs\n");
return r;
}
return 0;
}
void radeon_benchmark_move(struct radeon_device *rdev, unsigned bsize,
unsigned sdomain, unsigned ddomain)
{
struct radeon_bo *dobj = NULL;
struct radeon_bo *sobj = NULL;
struct radeon_fence *fence = NULL;
uint64_t saddr, daddr;
unsigned long start_jiffies;
unsigned long end_jiffies;
unsigned long time;
unsigned i, n, size;
int r;
size = bsize;
n = 1024;
r = radeon_bo_create(rdev, NULL, size, PAGE_SIZE, true, sdomain, &sobj);
if (r) {
goto out_cleanup;
}
r = radeon_bo_reserve(sobj, false);
if (unlikely(r != 0))
goto out_cleanup;
r = radeon_bo_pin(sobj, sdomain, &saddr);
radeon_bo_unreserve(sobj);
if (r) {
goto out_cleanup;
}
r = radeon_bo_create(rdev, NULL, size, PAGE_SIZE, true, ddomain, &dobj);
if (r) {
goto out_cleanup;
}
r = radeon_bo_reserve(dobj, false);
if (unlikely(r != 0))
goto out_cleanup;
r = radeon_bo_pin(dobj, ddomain, &daddr);
radeon_bo_unreserve(dobj);
if (r) {
goto out_cleanup;
}
/* r100 doesn't have dma engine so skip the test */
if (rdev->asic->copy_dma) {
start_jiffies = jiffies;
for (i = 0; i < n; i++) {
r = radeon_fence_create(rdev, &fence);
if (r) {
goto out_cleanup;
}
r = radeon_copy_dma(rdev, saddr, daddr,
size / RADEON_GPU_PAGE_SIZE, fence);
if (r) {
goto out_cleanup;
}
r = radeon_fence_wait(fence, false);
if (r) {
goto out_cleanup;
}
radeon_fence_unref(&fence);
}
end_jiffies = jiffies;
time = end_jiffies - start_jiffies;
time = jiffies_to_msecs(time);
if (time > 0) {
i = ((n * size) >> 10) / time;
printk(KERN_INFO "radeon: dma %u bo moves of %ukb from"
" %d to %d in %lums (%ukb/ms %ukb/s %uM/s)\n",
n, size >> 10,
sdomain, ddomain, time,
i, i * 1000, (i * 1000) / 1024);
}
static void radeon_benchmark_move(struct radeon_device *rdev, unsigned size,
unsigned sdomain, unsigned ddomain)
{
struct radeon_bo *dobj = NULL;
struct radeon_bo *sobj = NULL;
uint64_t saddr, daddr;
int r, n;
int time;
n = RADEON_BENCHMARK_ITERATIONS;
r = radeon_bo_create(rdev, size, PAGE_SIZE, true, sdomain, 0, NULL, &sobj);
if (r) {
goto out_cleanup;
}
r = radeon_bo_reserve(sobj, false);
if (unlikely(r != 0))
goto out_cleanup;
r = radeon_bo_pin(sobj, sdomain, &saddr);
radeon_bo_unreserve(sobj);
if (r) {
goto out_cleanup;
}
r = radeon_bo_create(rdev, size, PAGE_SIZE, true, ddomain, 0, NULL, &dobj);
if (r) {
goto out_cleanup;
}
r = radeon_bo_reserve(dobj, false);
if (unlikely(r != 0))
goto out_cleanup;
r = radeon_bo_pin(dobj, ddomain, &daddr);
radeon_bo_unreserve(dobj);
if (r) {
goto out_cleanup;
}
if (rdev->asic->copy.dma) {
time = radeon_benchmark_do_move(rdev, size, saddr, daddr,
RADEON_BENCHMARK_COPY_DMA, n);
if (time < 0)
goto out_cleanup;
if (time > 0)
radeon_benchmark_log_results(n, size, time,
sdomain, ddomain, "dma");
}
if (rdev->asic->copy.blit) {
time = radeon_benchmark_do_move(rdev, size, saddr, daddr,
RADEON_BENCHMARK_COPY_BLIT, n);
if (time < 0)
goto out_cleanup;
if (time > 0)
radeon_benchmark_log_results(n, size, time,
sdomain, ddomain, "blit");
}
out_cleanup:
if (sobj) {
r = radeon_bo_reserve(sobj, false);
if (likely(r == 0)) {
radeon_bo_unpin(sobj);
radeon_bo_unreserve(sobj);
}
radeon_bo_unref(&sobj);
}
if (dobj) {
r = radeon_bo_reserve(dobj, false);
if (likely(r == 0)) {
radeon_bo_unpin(dobj);
radeon_bo_unreserve(dobj);
}
radeon_bo_unref(&dobj);
}
if (r) {
DRM_ERROR("Error while benchmarking BO move.\n");
}
}
