void rv770_pcie_gart_disable(struct radeon_device *rdev)
{
u32 tmp;
int i;
for (i = 0; i < 7; i++)
WREG32(VM_CONTEXT0_CNTL + (i * 4), 0);
WREG32(VM_L2_CNTL, ENABLE_L2_FRAGMENT_PROCESSING |
EFFECTIVE_L2_QUEUE_SIZE(7));
WREG32(VM_L2_CNTL2, 0);
WREG32(VM_L2_CNTL3, BANK_SELECT(0) | CACHE_UPDATE_MODE(2));
tmp = EFFECTIVE_L1_TLB_SIZE(5) | EFFECTIVE_L1_QUEUE_SIZE(5);
WREG32(MC_VM_MD_L1_TLB0_CNTL, tmp);
WREG32(MC_VM_MD_L1_TLB1_CNTL, tmp);
WREG32(MC_VM_MD_L1_TLB2_CNTL, tmp);
WREG32(MC_VM_MB_L1_TLB0_CNTL, tmp);
WREG32(MC_VM_MB_L1_TLB1_CNTL, tmp);
WREG32(MC_VM_MB_L1_TLB2_CNTL, tmp);
WREG32(MC_VM_MB_L1_TLB3_CNTL, tmp);
if (rdev->gart.table.vram.robj) {
radeon_object_kunmap(rdev->gart.table.vram.robj);
radeon_object_unpin(rdev->gart.table.vram.robj);
}
}
void r600_pcie_gart_disable(struct radeon_device *rdev)
{
u32 tmp;
int i;
for (i = 0; i < 7; i++)
WREG32(VM_CONTEXT0_CNTL + (i * 4), 0);
WREG32(VM_L2_CNTL, ENABLE_L2_FRAGMENT_PROCESSING |
EFFECTIVE_L2_QUEUE_SIZE(7));
WREG32(VM_L2_CNTL3, BANK_SELECT_0(0) | BANK_SELECT_1(1));
tmp = EFFECTIVE_L1_TLB_SIZE(5) | EFFECTIVE_L1_QUEUE_SIZE(5) |
ENABLE_WAIT_L2_QUERY;
WREG32(MC_VM_L1_TLB_MCD_RD_A_CNTL, tmp);
WREG32(MC_VM_L1_TLB_MCD_WR_A_CNTL, tmp);
WREG32(MC_VM_L1_TLB_MCD_RD_B_CNTL, tmp);
WREG32(MC_VM_L1_TLB_MCD_WR_B_CNTL, tmp);
WREG32(MC_VM_L1_TLB_MCB_RD_GFX_CNTL, tmp);
WREG32(MC_VM_L1_TLB_MCB_WR_GFX_CNTL, tmp);
WREG32(MC_VM_L1_TLB_MCB_RD_PDMA_CNTL, tmp);
WREG32(MC_VM_L1_TLB_MCB_WR_PDMA_CNTL, tmp);
WREG32(MC_VM_L1_TLB_MCB_RD_SEM_CNTL, tmp);
WREG32(MC_VM_L1_TLB_MCB_WR_SEM_CNTL, tmp);
WREG32(MC_VM_L1_TLB_MCB_RD_SYS_CNTL, tmp);
WREG32(MC_VM_L1_TLB_MCB_WR_SYS_CNTL, tmp);
WREG32(MC_VM_L1_TLB_MCB_RD_HDP_CNTL, tmp);
WREG32(MC_VM_L1_TLB_MCB_WR_HDP_CNTL, tmp);
if (rdev->gart.table.vram.robj) {
radeon_object_kunmap(rdev->gart.table.vram.robj);
radeon_object_unpin(rdev->gart.table.vram.robj);
}
}
void rs600_gart_disable(struct radeon_device *rdev)
{
uint32_t tmp;
/* FIXME: disable out of gart access */
WREG32_MC(RS600_MC_PT0_CNTL, 0);
tmp = RREG32_MC(RS600_MC_CNTL1);
tmp &= ~RS600_ENABLE_PAGE_TABLES;
WREG32_MC(RS600_MC_CNTL1, tmp);
radeon_object_kunmap(rdev->gart.table.vram.robj);
radeon_object_unpin(rdev->gart.table.vram.robj);
}
void radeon_test_moves(struct radeon_device *rdev)
{
struct radeon_object *vram_obj = NULL;
struct radeon_object **gtt_obj = NULL;
struct radeon_fence *fence = NULL;
uint64_t gtt_addr, vram_addr;
unsigned i, n, size;
int r;
size = 1024 * 1024;
n = ((u32)(rdev->mc.gtt_size - RADEON_IB_POOL_SIZE*64*1024 - RADEON_GPU_PAGE_SIZE -
rdev->cp.ring_size)) / 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_object_create(rdev, NULL, size, true, RADEON_GEM_DOMAIN_VRAM,
false, &vram_obj);
if (r) {
DRM_ERROR("Failed to create VRAM object\n");
goto out_cleanup;
}
r = radeon_object_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_object_create(rdev, NULL, size, true,
RADEON_GEM_DOMAIN_GTT, false, gtt_obj + i);
if (r) {
DRM_ERROR("Failed to create GTT object %d\n", i);
goto out_cleanup;
}
r = radeon_object_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_object_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_object_kunmap(gtt_obj[i]);
r = radeon_fence_create(rdev, &fence);
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_object_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 map 0x%p-0x%p)\n",
i, *vram_start, gtt_start, gtt_map,
gtt_end);
radeon_object_kunmap(vram_obj);
goto out_cleanup;
}
*vram_start = vram_start;
}
radeon_object_kunmap(vram_obj);
r = radeon_fence_create(rdev, &fence);
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_object_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 map 0x%p-0x%p)\n",
i, *gtt_start, vram_start, vram_map,
vram_end);
radeon_object_kunmap(gtt_obj[i]);
goto out_cleanup;
}
}
radeon_object_kunmap(gtt_obj[i]);
DRM_INFO("Tested GTT->VRAM and VRAM->GTT copy for GTT offset 0x%llx\n",
gtt_addr - rdev->mc.gtt_location);
}
out_cleanup:
if (vram_obj) {
radeon_object_unpin(vram_obj);
radeon_object_unref(&vram_obj);
}
if (gtt_obj) {
for (i = 0; i < n; i++) {
if (gtt_obj[i]) {
radeon_object_unpin(gtt_obj[i]);
radeon_object_unref(>t_obj[i]);
}
}
kfree(gtt_obj);
}
if (fence) {
radeon_fence_unref(&fence);
}
if (r) {
printk(KERN_WARNING "Error while testing BO move.\n");
}
}
