aboutsummaryrefslogtreecommitdiff
path: root/drivers/gpu/drm/radeon/radeon_test.c
diff options
context:
space:
mode:
Diffstat (limited to 'drivers/gpu/drm/radeon/radeon_test.c')
-rw-r--r--drivers/gpu/drm/radeon/radeon_test.c270
1 files changed, 171 insertions, 99 deletions
diff --git a/drivers/gpu/drm/radeon/radeon_test.c b/drivers/gpu/drm/radeon/radeon_test.c
index efff929ea49..3a13e0d1055 100644
--- a/drivers/gpu/drm/radeon/radeon_test.c
+++ b/drivers/gpu/drm/radeon/radeon_test.c
@@ -26,16 +26,30 @@
#include "radeon_reg.h"
#include "radeon.h"
+#define RADEON_TEST_COPY_BLIT 1
+#define RADEON_TEST_COPY_DMA 0
+
/* Test BO GTT->VRAM and VRAM->GTT GPU copies across the whole GTT aperture */
-void radeon_test_moves(struct radeon_device *rdev)
+static void radeon_do_test_moves(struct radeon_device *rdev, int flag)
{
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;
+ 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;
@@ -66,37 +80,38 @@ void radeon_test_moves(struct radeon_device *rdev)
}
r = radeon_bo_reserve(vram_obj, false);
if (unlikely(r != 0))
- goto out_cleanup;
+ 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_cleanup;
+ 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_cleanup;
+ goto out_lclean;
}
r = radeon_bo_reserve(gtt_obj[i], false);
if (unlikely(r != 0))
- goto out_cleanup;
+ goto out_lclean_unref;
r = radeon_bo_pin(gtt_obj[i], RADEON_GEM_DOMAIN_GTT, &gtt_addr);
if (r) {
DRM_ERROR("Failed to pin GTT object %d\n", i);
- goto out_cleanup;
+ goto out_lclean_unres;
}
r = radeon_bo_kmap(gtt_obj[i], &gtt_map);
if (r) {
DRM_ERROR("Failed to map GTT object %d\n", i);
- goto out_cleanup;
+ goto out_lclean_unpin;
}
for (gtt_start = gtt_map, gtt_end = gtt_map + size;
@@ -106,22 +121,19 @@ void radeon_test_moves(struct radeon_device *rdev)
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 (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_cleanup;
+ 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_cleanup;
+ goto out_lclean_unpin;
}
radeon_fence_unref(&fence);
@@ -129,7 +141,7 @@ void radeon_test_moves(struct radeon_device *rdev)
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;
+ goto out_lclean_unpin;
}
for (gtt_start = gtt_map, gtt_end = gtt_map + size,
@@ -148,29 +160,26 @@ void radeon_test_moves(struct radeon_device *rdev)
(vram_addr - rdev->mc.vram_start +
(void*)gtt_start - gtt_map));
radeon_bo_kunmap(vram_obj);
- goto out_cleanup;
+ goto out_lclean_unpin;
}
*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 (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_cleanup;
+ 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_cleanup;
+ goto out_lclean_unpin;
}
radeon_fence_unref(&fence);
@@ -178,7 +187,7 @@ void radeon_test_moves(struct radeon_device *rdev)
r = radeon_bo_kmap(gtt_obj[i], &gtt_map);
if (r) {
DRM_ERROR("Failed to map GTT object after copy %d\n", i);
- goto out_cleanup;
+ goto out_lclean_unpin;
}
for (gtt_start = gtt_map, gtt_end = gtt_map + size,
@@ -197,7 +206,7 @@ void radeon_test_moves(struct radeon_device *rdev)
(gtt_addr - rdev->mc.gtt_start +
(void*)vram_start - vram_map));
radeon_bo_kunmap(gtt_obj[i]);
- goto out_cleanup;
+ goto out_lclean_unpin;
}
}
@@ -205,74 +214,129 @@ void radeon_test_moves(struct radeon_device *rdev)
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(&gtt_obj[i]);
+out_lclean:
+ for (--i; i >= 0; --i) {
+ radeon_bo_unpin(gtt_obj[i]);
+ radeon_bo_unreserve(gtt_obj[i]);
+ radeon_bo_unref(&gtt_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:
- 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(&gtt_obj[i]);
- }
- }
- kfree(gtt_obj);
- }
- if (fence) {
- radeon_fence_unref(&fence);
- }
+ kfree(gtt_obj);
if (r) {
printk(KERN_WARNING "Error while testing BO move.\n");
}
}
+void radeon_test_moves(struct radeon_device *rdev)
+{
+ if (rdev->asic->copy.dma)
+ radeon_do_test_moves(rdev, RADEON_TEST_COPY_DMA);
+ if (rdev->asic->copy.blit)
+ radeon_do_test_moves(rdev, RADEON_TEST_COPY_BLIT);
+}
+
+static int radeon_test_create_and_emit_fence(struct radeon_device *rdev,
+ struct radeon_ring *ring,
+ struct radeon_fence **fence)
+{
+ uint32_t handle = ring->idx ^ 0xdeafbeef;
+ int r;
+
+ if (ring->idx == R600_RING_TYPE_UVD_INDEX) {
+ r = radeon_uvd_get_create_msg(rdev, ring->idx, handle, NULL);
+ if (r) {
+ DRM_ERROR("Failed to get dummy create msg\n");
+ return r;
+ }
+
+ r = radeon_uvd_get_destroy_msg(rdev, ring->idx, handle, fence);
+ if (r) {
+ DRM_ERROR("Failed to get dummy destroy msg\n");
+ return r;
+ }
+
+ } else if (ring->idx == TN_RING_TYPE_VCE1_INDEX ||
+ ring->idx == TN_RING_TYPE_VCE2_INDEX) {
+ r = radeon_vce_get_create_msg(rdev, ring->idx, handle, NULL);
+ if (r) {
+ DRM_ERROR("Failed to get dummy create msg\n");
+ return r;
+ }
+
+ r = radeon_vce_get_destroy_msg(rdev, ring->idx, handle, fence);
+ if (r) {
+ DRM_ERROR("Failed to get dummy destroy msg\n");
+ return r;
+ }
+
+ } else {
+ r = radeon_ring_lock(rdev, ring, 64);
+ if (r) {
+ DRM_ERROR("Failed to lock ring A %d\n", ring->idx);
+ return r;
+ }
+ radeon_fence_emit(rdev, fence, ring->idx);
+ radeon_ring_unlock_commit(rdev, ring);
+ }
+ return 0;
+}
+
void radeon_test_ring_sync(struct radeon_device *rdev,
struct radeon_ring *ringA,
struct radeon_ring *ringB)
{
struct radeon_fence *fence1 = NULL, *fence2 = NULL;
struct radeon_semaphore *semaphore = NULL;
- int ridxA = radeon_ring_index(rdev, ringA);
- int ridxB = radeon_ring_index(rdev, ringB);
int r;
- r = radeon_fence_create(rdev, &fence1, ridxA);
+ r = radeon_semaphore_create(rdev, &semaphore);
if (r) {
- DRM_ERROR("Failed to create sync fence 1\n");
+ DRM_ERROR("Failed to create semaphore\n");
goto out_cleanup;
}
- r = radeon_fence_create(rdev, &fence2, ridxA);
+
+ r = radeon_ring_lock(rdev, ringA, 64);
if (r) {
- DRM_ERROR("Failed to create sync fence 2\n");
+ DRM_ERROR("Failed to lock ring A %d\n", ringA->idx);
goto out_cleanup;
}
+ radeon_semaphore_emit_wait(rdev, ringA->idx, semaphore);
+ radeon_ring_unlock_commit(rdev, ringA);
- r = radeon_semaphore_create(rdev, &semaphore);
- if (r) {
- DRM_ERROR("Failed to create semaphore\n");
+ r = radeon_test_create_and_emit_fence(rdev, ringA, &fence1);
+ if (r)
goto out_cleanup;
- }
r = radeon_ring_lock(rdev, ringA, 64);
if (r) {
- DRM_ERROR("Failed to lock ring A %d\n", ridxA);
+ DRM_ERROR("Failed to lock ring A %d\n", ringA->idx);
goto out_cleanup;
}
- radeon_semaphore_emit_wait(rdev, ridxA, semaphore);
- radeon_fence_emit(rdev, fence1);
- radeon_semaphore_emit_wait(rdev, ridxA, semaphore);
- radeon_fence_emit(rdev, fence2);
+ radeon_semaphore_emit_wait(rdev, ringA->idx, semaphore);
radeon_ring_unlock_commit(rdev, ringA);
+ r = radeon_test_create_and_emit_fence(rdev, ringA, &fence2);
+ if (r)
+ goto out_cleanup;
+
mdelay(1000);
if (radeon_fence_signaled(fence1)) {
@@ -285,7 +349,7 @@ void radeon_test_ring_sync(struct radeon_device *rdev,
DRM_ERROR("Failed to lock ring B %p\n", ringB);
goto out_cleanup;
}
- radeon_semaphore_emit_signal(rdev, ridxB, semaphore);
+ radeon_semaphore_emit_signal(rdev, ringB->idx, semaphore);
radeon_ring_unlock_commit(rdev, ringB);
r = radeon_fence_wait(fence1, false);
@@ -306,7 +370,7 @@ void radeon_test_ring_sync(struct radeon_device *rdev,
DRM_ERROR("Failed to lock ring B %p\n", ringB);
goto out_cleanup;
}
- radeon_semaphore_emit_signal(rdev, ridxB, semaphore);
+ radeon_semaphore_emit_signal(rdev, ringB->idx, semaphore);
radeon_ring_unlock_commit(rdev, ringB);
r = radeon_fence_wait(fence2, false);
@@ -316,8 +380,7 @@ void radeon_test_ring_sync(struct radeon_device *rdev,
}
out_cleanup:
- if (semaphore)
- radeon_semaphore_free(rdev, semaphore, NULL);
+ radeon_semaphore_free(rdev, &semaphore, NULL);
if (fence1)
radeon_fence_unref(&fence1);
@@ -329,30 +392,16 @@ out_cleanup:
printk(KERN_WARNING "Error while testing ring sync (%d).\n", r);
}
-void radeon_test_ring_sync2(struct radeon_device *rdev,
+static void radeon_test_ring_sync2(struct radeon_device *rdev,
struct radeon_ring *ringA,
struct radeon_ring *ringB,
struct radeon_ring *ringC)
{
struct radeon_fence *fenceA = NULL, *fenceB = NULL;
struct radeon_semaphore *semaphore = NULL;
- int ridxA = radeon_ring_index(rdev, ringA);
- int ridxB = radeon_ring_index(rdev, ringB);
- int ridxC = radeon_ring_index(rdev, ringC);
bool sigA, sigB;
int i, r;
- r = radeon_fence_create(rdev, &fenceA, ridxA);
- if (r) {
- DRM_ERROR("Failed to create sync fence 1\n");
- goto out_cleanup;
- }
- r = radeon_fence_create(rdev, &fenceB, ridxB);
- if (r) {
- DRM_ERROR("Failed to create sync fence 2\n");
- goto out_cleanup;
- }
-
r = radeon_semaphore_create(rdev, &semaphore);
if (r) {
DRM_ERROR("Failed to create semaphore\n");
@@ -361,21 +410,26 @@ void radeon_test_ring_sync2(struct radeon_device *rdev,
r = radeon_ring_lock(rdev, ringA, 64);
if (r) {
- DRM_ERROR("Failed to lock ring A %d\n", ridxA);
+ DRM_ERROR("Failed to lock ring A %d\n", ringA->idx);
goto out_cleanup;
}
- radeon_semaphore_emit_wait(rdev, ridxA, semaphore);
- radeon_fence_emit(rdev, fenceA);
+ radeon_semaphore_emit_wait(rdev, ringA->idx, semaphore);
radeon_ring_unlock_commit(rdev, ringA);
+ r = radeon_test_create_and_emit_fence(rdev, ringA, &fenceA);
+ if (r)
+ goto out_cleanup;
+
r = radeon_ring_lock(rdev, ringB, 64);
if (r) {
- DRM_ERROR("Failed to lock ring B %d\n", ridxB);
+ DRM_ERROR("Failed to lock ring B %d\n", ringB->idx);
goto out_cleanup;
}
- radeon_semaphore_emit_wait(rdev, ridxB, semaphore);
- radeon_fence_emit(rdev, fenceB);
+ radeon_semaphore_emit_wait(rdev, ringB->idx, semaphore);
radeon_ring_unlock_commit(rdev, ringB);
+ r = radeon_test_create_and_emit_fence(rdev, ringB, &fenceB);
+ if (r)
+ goto out_cleanup;
mdelay(1000);
@@ -384,7 +438,7 @@ void radeon_test_ring_sync2(struct radeon_device *rdev,
goto out_cleanup;
}
if (radeon_fence_signaled(fenceB)) {
- DRM_ERROR("Fence A signaled without waiting for semaphore.\n");
+ DRM_ERROR("Fence B signaled without waiting for semaphore.\n");
goto out_cleanup;
}
@@ -393,7 +447,7 @@ void radeon_test_ring_sync2(struct radeon_device *rdev,
DRM_ERROR("Failed to lock ring B %p\n", ringC);
goto out_cleanup;
}
- radeon_semaphore_emit_signal(rdev, ridxC, semaphore);
+ radeon_semaphore_emit_signal(rdev, ringC->idx, semaphore);
radeon_ring_unlock_commit(rdev, ringC);
for (i = 0; i < 30; ++i) {
@@ -419,7 +473,7 @@ void radeon_test_ring_sync2(struct radeon_device *rdev,
DRM_ERROR("Failed to lock ring B %p\n", ringC);
goto out_cleanup;
}
- radeon_semaphore_emit_signal(rdev, ridxC, semaphore);
+ radeon_semaphore_emit_signal(rdev, ringC->idx, semaphore);
radeon_ring_unlock_commit(rdev, ringC);
mdelay(1000);
@@ -436,8 +490,7 @@ void radeon_test_ring_sync2(struct radeon_device *rdev,
}
out_cleanup:
- if (semaphore)
- radeon_semaphore_free(rdev, semaphore, NULL);
+ radeon_semaphore_free(rdev, &semaphore, NULL);
if (fenceA)
radeon_fence_unref(&fenceA);
@@ -449,6 +502,16 @@ out_cleanup:
printk(KERN_WARNING "Error while testing ring sync (%d).\n", r);
}
+static bool radeon_test_sync_possible(struct radeon_ring *ringA,
+ struct radeon_ring *ringB)
+{
+ if (ringA->idx == TN_RING_TYPE_VCE2_INDEX &&
+ ringB->idx == TN_RING_TYPE_VCE1_INDEX)
+ return false;
+
+ return true;
+}
+
void radeon_test_syncing(struct radeon_device *rdev)
{
int i, j, k;
@@ -463,6 +526,9 @@ void radeon_test_syncing(struct radeon_device *rdev)
if (!ringB->ready)
continue;
+ if (!radeon_test_sync_possible(ringA, ringB))
+ continue;
+
DRM_INFO("Testing syncing between rings %d and %d...\n", i, j);
radeon_test_ring_sync(rdev, ringA, ringB);
@@ -474,6 +540,12 @@ void radeon_test_syncing(struct radeon_device *rdev)
if (!ringC->ready)
continue;
+ if (!radeon_test_sync_possible(ringA, ringC))
+ continue;
+
+ if (!radeon_test_sync_possible(ringB, ringC))
+ continue;
+
DRM_INFO("Testing syncing between rings %d, %d and %d...\n", i, j, k);
radeon_test_ring_sync2(rdev, ringA, ringB, ringC);
generated by cgit v1.2.3-70-g09d2 (git 2.46.0) at 2025-11-10 13:16:32 +0000