/**
* r600_dma_ring_ib_execute - Schedule an IB on the DMA engine
*
* @rdev: radeon_device pointer
* @ib: IB object to schedule
*
* Schedule an IB in the DMA ring (r6xx-r7xx).
*/
void r600_dma_ring_ib_execute(struct radeon_device *rdev, struct radeon_ib *ib)
{
struct radeon_ring *ring = &rdev->ring[ib->ring];
if (rdev->wb.enabled) {
u32 next_rptr = ring->wptr + 4;
while ((next_rptr & 7) != 5)
next_rptr++;
next_rptr += 3;
radeon_ring_write(ring, DMA_PACKET(DMA_PACKET_WRITE, 0, 0, 1));
radeon_ring_write(ring, ring->next_rptr_gpu_addr & 0xfffffffc);
radeon_ring_write(ring, upper_32_bits(ring->next_rptr_gpu_addr) & 0xff);
radeon_ring_write(ring, next_rptr);
}
/* The indirect buffer packet must end on an 8 DW boundary in the DMA ring.
* Pad as necessary with NOPs.
*/
while ((ring->wptr & 7) != 5)
radeon_ring_write(ring, DMA_PACKET(DMA_PACKET_NOP, 0, 0, 0));
radeon_ring_write(ring, DMA_PACKET(DMA_PACKET_INDIRECT_BUFFER, 0, 0, 0));
radeon_ring_write(ring, (ib->gpu_addr & 0xFFFFFFE0));
radeon_ring_write(ring, (ib->length_dw << 16) | (upper_32_bits(ib->gpu_addr) & 0xFF));
}
/**
* r600_dma_fence_ring_emit - emit a fence on the DMA ring
*
* @rdev: radeon_device pointer
* @fence: radeon fence object
*
* Add a DMA fence packet to the ring to write
* the fence seq number and DMA trap packet to generate
* an interrupt if needed (r6xx-r7xx).
*/
void r600_dma_fence_ring_emit(struct radeon_device *rdev,
struct radeon_fence *fence)
{
struct radeon_ring *ring = &rdev->ring[fence->ring];
u64 addr = rdev->fence_drv[fence->ring].gpu_addr;
/* write the fence */
radeon_ring_write(ring, DMA_PACKET(DMA_PACKET_FENCE, 0, 0, 0));
radeon_ring_write(ring, addr & 0xfffffffc);
radeon_ring_write(ring, (upper_32_bits(addr) & 0xff));
radeon_ring_write(ring, lower_32_bits(fence->seq));
/* generate an interrupt */
radeon_ring_write(ring, DMA_PACKET(DMA_PACKET_TRAP, 0, 0, 0));
}
/**
* si_dma_ring_test_ib - test an IB on the DMA engine
*
* @ring: amdgpu_ring structure holding ring information
*
* Test a simple IB in the DMA ring (VI).
* Returns 0 on success, error on failure.
*/
static int si_dma_ring_test_ib(struct amdgpu_ring *ring, long timeout)
{
struct amdgpu_device *adev = ring->adev;
struct amdgpu_ib ib;
struct dma_fence *f = NULL;
unsigned index;
u32 tmp = 0;
u64 gpu_addr;
long r;
r = amdgpu_wb_get(adev, &index);
if (r) {
dev_err(adev->dev, "(%ld) failed to allocate wb slot\n", r);
return r;
}
gpu_addr = adev->wb.gpu_addr + (index * 4);
tmp = 0xCAFEDEAD;
adev->wb.wb[index] = cpu_to_le32(tmp);
memset(&ib, 0, sizeof(ib));
r = amdgpu_ib_get(adev, NULL, 256, &ib);
if (r) {
DRM_ERROR("amdgpu: failed to get ib (%ld).\n", r);
goto err0;
}
ib.ptr[0] = DMA_PACKET(DMA_PACKET_WRITE, 0, 0, 0, 1);
ib.ptr[1] = lower_32_bits(gpu_addr);
ib.ptr[2] = upper_32_bits(gpu_addr) & 0xff;
ib.ptr[3] = 0xDEADBEEF;
ib.length_dw = 4;
r = amdgpu_ib_schedule(ring, 1, &ib, NULL, NULL, &f);
if (r)
goto err1;
r = dma_fence_wait_timeout(f, false, timeout);
if (r == 0) {
DRM_ERROR("amdgpu: IB test timed out\n");
r = -ETIMEDOUT;
goto err1;
} else if (r < 0) {
DRM_ERROR("amdgpu: fence wait failed (%ld).\n", r);
goto err1;
}
tmp = le32_to_cpu(adev->wb.wb[index]);
if (tmp == 0xDEADBEEF) {
DRM_INFO("ib test on ring %d succeeded\n", ring->idx);
r = 0;
} else {
DRM_ERROR("amdgpu: ib test failed (0x%08X)\n", tmp);
r = -EINVAL;
}
err1:
amdgpu_ib_free(adev, &ib, NULL);
dma_fence_put(f);
err0:
amdgpu_wb_free(adev, index);
return r;
}
/**
* r600_copy_dma - copy pages using the DMA engine
*
* @rdev: radeon_device pointer
* @src_offset: src GPU address
* @dst_offset: dst GPU address
* @num_gpu_pages: number of GPU pages to xfer
* @resv: reservation object to sync to
*
* Copy GPU paging using the DMA engine (r6xx).
* Used by the radeon ttm implementation to move pages if
* registered as the asic copy callback.
*/
struct radeon_fence *r600_copy_dma(struct radeon_device *rdev,
uint64_t src_offset, uint64_t dst_offset,
unsigned num_gpu_pages,
struct reservation_object *resv)
{
struct radeon_semaphore *sem = NULL;
struct radeon_fence *fence;
int ring_index = rdev->asic->copy.dma_ring_index;
struct radeon_ring *ring = &rdev->ring[ring_index];
u32 size_in_dw, cur_size_in_dw;
int i, num_loops;
int r = 0;
r = radeon_semaphore_create(rdev, &sem);
if (r) {
DRM_ERROR("radeon: moving bo (%d).\n", r);
return ERR_PTR(r);
}
size_in_dw = (num_gpu_pages << RADEON_GPU_PAGE_SHIFT) / 4;
num_loops = DIV_ROUND_UP(size_in_dw, 0xFFFE);
r = radeon_ring_lock(rdev, ring, num_loops * 4 + 8);
if (r) {
DRM_ERROR("radeon: moving bo (%d).\n", r);
radeon_semaphore_free(rdev, &sem, NULL);
return ERR_PTR(r);
}
radeon_semaphore_sync_resv(rdev, sem, resv, false);
radeon_semaphore_sync_rings(rdev, sem, ring->idx);
for (i = 0; i < num_loops; i++) {
cur_size_in_dw = size_in_dw;
if (cur_size_in_dw > 0xFFFE)
cur_size_in_dw = 0xFFFE;
size_in_dw -= cur_size_in_dw;
radeon_ring_write(ring, DMA_PACKET(DMA_PACKET_COPY, 0, 0, cur_size_in_dw));
radeon_ring_write(ring, dst_offset & 0xfffffffc);
radeon_ring_write(ring, src_offset & 0xfffffffc);
radeon_ring_write(ring, (((upper_32_bits(dst_offset) & 0xff) << 16) |
(upper_32_bits(src_offset) & 0xff)));
src_offset += cur_size_in_dw * 4;
dst_offset += cur_size_in_dw * 4;
}
r = radeon_fence_emit(rdev, &fence, ring->idx);
if (r) {
radeon_ring_unlock_undo(rdev, ring);
radeon_semaphore_free(rdev, &sem, NULL);
return ERR_PTR(r);
}
radeon_ring_unlock_commit(rdev, ring, false);
radeon_semaphore_free(rdev, &sem, fence);
return fence;
}
/**
* si_dma_emit_fill_buffer - fill buffer using the sDMA engine
*
* @ring: amdgpu_ring structure holding ring information
* @src_data: value to write to buffer
* @dst_offset: dst GPU address
* @byte_count: number of bytes to xfer
*
* Fill GPU buffers using the DMA engine (VI).
*/
static void si_dma_emit_fill_buffer(struct amdgpu_ib *ib,
uint32_t src_data,
uint64_t dst_offset,
uint32_t byte_count)
{
ib->ptr[ib->length_dw++] = DMA_PACKET(DMA_PACKET_CONSTANT_FILL,
0, 0, 0, byte_count / 4);
ib->ptr[ib->length_dw++] = lower_32_bits(dst_offset);
ib->ptr[ib->length_dw++] = src_data;
ib->ptr[ib->length_dw++] = upper_32_bits(dst_offset) << 16;
}
/**
* r600_dma_ib_test - test an IB on the DMA engine
*
* @rdev: radeon_device pointer
* @ring: radeon_ring structure holding ring information
*
* Test a simple IB in the DMA ring (r6xx-SI).
* Returns 0 on success, error on failure.
*/
int r600_dma_ib_test(struct radeon_device *rdev, struct radeon_ring *ring)
{
struct radeon_ib ib;
unsigned i;
int r;
void __iomem *ptr = (void *)rdev->vram_scratch.ptr;
u32 tmp = 0;
if (!ptr) {
DRM_ERROR("invalid vram scratch pointer\n");
return -EINVAL;
}
tmp = 0xCAFEDEAD;
writel(tmp, ptr);
r = radeon_ib_get(rdev, ring->idx, &ib, NULL, 256);
if (r) {
DRM_ERROR("radeon: failed to get ib (%d).\n", r);
return r;
}
ib.ptr[0] = DMA_PACKET(DMA_PACKET_WRITE, 0, 0, 1);
ib.ptr[1] = rdev->vram_scratch.gpu_addr & 0xfffffffc;
ib.ptr[2] = upper_32_bits(rdev->vram_scratch.gpu_addr) & 0xff;
ib.ptr[3] = 0xDEADBEEF;
ib.length_dw = 4;
r = radeon_ib_schedule(rdev, &ib, NULL);
if (r) {
radeon_ib_free(rdev, &ib);
DRM_ERROR("radeon: failed to schedule ib (%d).\n", r);
return r;
}
r = radeon_fence_wait(ib.fence, false);
if (r) {
DRM_ERROR("radeon: fence wait failed (%d).\n", r);
return r;
}
for (i = 0; i < rdev->usec_timeout; i++) {
tmp = readl(ptr);
if (tmp == 0xDEADBEEF)
break;
DRM_UDELAY(1);
}
if (i < rdev->usec_timeout) {
DRM_INFO("ib test on ring %d succeeded in %u usecs\n", ib.fence->ring, i);
} else {
DRM_ERROR("radeon: ib test failed (0x%08X)\n", tmp);
r = -EINVAL;
}
radeon_ib_free(rdev, &ib);
return r;
}
/**
* si_dma_ring_emit_fence - emit a fence on the DMA ring
*
* @ring: amdgpu ring pointer
* @fence: amdgpu fence object
*
* Add a DMA fence packet to the ring to write
* the fence seq number and DMA trap packet to generate
* an interrupt if needed (VI).
*/
static void si_dma_ring_emit_fence(struct amdgpu_ring *ring, u64 addr, u64 seq,
unsigned flags)
{
bool write64bit = flags & AMDGPU_FENCE_FLAG_64BIT;
/* write the fence */
amdgpu_ring_write(ring, DMA_PACKET(DMA_PACKET_FENCE, 0, 0, 0, 0));
amdgpu_ring_write(ring, addr & 0xfffffffc);
amdgpu_ring_write(ring, (upper_32_bits(addr) & 0xff));
amdgpu_ring_write(ring, seq);
/* optionally write high bits as well */
if (write64bit) {
addr += 4;
amdgpu_ring_write(ring, DMA_PACKET(DMA_PACKET_FENCE, 0, 0, 0, 0));
amdgpu_ring_write(ring, addr & 0xfffffffc);
amdgpu_ring_write(ring, (upper_32_bits(addr) & 0xff));
amdgpu_ring_write(ring, upper_32_bits(seq));
}
/* generate an interrupt */
amdgpu_ring_write(ring, DMA_PACKET(DMA_PACKET_TRAP, 0, 0, 0, 0));
}
/**
* si_dma_emit_copy_buffer - copy buffer using the sDMA engine
*
* @ring: amdgpu_ring structure holding ring information
* @src_offset: src GPU address
* @dst_offset: dst GPU address
* @byte_count: number of bytes to xfer
*
* Copy GPU buffers using the DMA engine (VI).
* Used by the amdgpu ttm implementation to move pages if
* registered as the asic copy callback.
*/
static void si_dma_emit_copy_buffer(struct amdgpu_ib *ib,
uint64_t src_offset,
uint64_t dst_offset,
uint32_t byte_count)
{
ib->ptr[ib->length_dw++] = DMA_PACKET(DMA_PACKET_COPY,
1, 0, 0, byte_count);
ib->ptr[ib->length_dw++] = lower_32_bits(dst_offset);
ib->ptr[ib->length_dw++] = lower_32_bits(src_offset);
ib->ptr[ib->length_dw++] = upper_32_bits(dst_offset) & 0xff;
ib->ptr[ib->length_dw++] = upper_32_bits(src_offset) & 0xff;
}
/**
* cik_dma_vm_copy_pte - update PTEs by copying them from the GART
*
* @ib: indirect buffer to fill with commands
* @pe: addr of the page entry
* @src: src addr to copy from
* @count: number of page entries to update
*
* Update PTEs by copying them from the GART using DMA (SI).
*/
static void si_dma_vm_copy_pte(struct amdgpu_ib *ib,
uint64_t pe, uint64_t src,
unsigned count)
{
unsigned bytes = count * 8;
ib->ptr[ib->length_dw++] = DMA_PACKET(DMA_PACKET_COPY,
1, 0, 0, bytes);
ib->ptr[ib->length_dw++] = lower_32_bits(pe);
ib->ptr[ib->length_dw++] = lower_32_bits(src);
ib->ptr[ib->length_dw++] = upper_32_bits(pe) & 0xff;
ib->ptr[ib->length_dw++] = upper_32_bits(src) & 0xff;
}
/**
* si_dma_ring_emit_vm_flush - cik vm flush using sDMA
*
* @ring: amdgpu_ring pointer
* @vm: amdgpu_vm pointer
*
* Update the page table base and flush the VM TLB
* using sDMA (VI).
*/
static void si_dma_ring_emit_vm_flush(struct amdgpu_ring *ring,
unsigned vm_id, uint64_t pd_addr)
{
amdgpu_ring_write(ring, DMA_PACKET(DMA_PACKET_SRBM_WRITE, 0, 0, 0, 0));
if (vm_id < 8)
amdgpu_ring_write(ring, (0xf << 16) | (VM_CONTEXT0_PAGE_TABLE_BASE_ADDR + vm_id));
else
amdgpu_ring_write(ring, (0xf << 16) | (VM_CONTEXT8_PAGE_TABLE_BASE_ADDR + (vm_id - 8)));
amdgpu_ring_write(ring, pd_addr >> 12);
/* bits 0-7 are the VM contexts0-7 */
amdgpu_ring_write(ring, DMA_PACKET(DMA_PACKET_SRBM_WRITE, 0, 0, 0, 0));
amdgpu_ring_write(ring, (0xf << 16) | (VM_INVALIDATE_REQUEST));
amdgpu_ring_write(ring, 1 << vm_id);
/* wait for invalidate to complete */
amdgpu_ring_write(ring, DMA_PACKET(DMA_PACKET_POLL_REG_MEM, 0, 0, 0, 0));
amdgpu_ring_write(ring, VM_INVALIDATE_REQUEST);
amdgpu_ring_write(ring, 0xff << 16); /* retry */
amdgpu_ring_write(ring, 1 << vm_id); /* mask */
amdgpu_ring_write(ring, 0); /* value */
amdgpu_ring_write(ring, (0 << 28) | 0x20); /* func(always) | poll interval */
}
/**
* cik_sdma_ring_emit_pipeline_sync - sync the pipeline
*
* @ring: amdgpu_ring pointer
*
* Make sure all previous operations are completed (CIK).
*/
static void si_dma_ring_emit_pipeline_sync(struct amdgpu_ring *ring)
{
uint32_t seq = ring->fence_drv.sync_seq;
uint64_t addr = ring->fence_drv.gpu_addr;
/* wait for idle */
amdgpu_ring_write(ring, DMA_PACKET(DMA_PACKET_POLL_REG_MEM, 0, 0, 0, 0) |
(1 << 27)); /* Poll memory */
amdgpu_ring_write(ring, lower_32_bits(addr));
amdgpu_ring_write(ring, (0xff << 16) | upper_32_bits(addr)); /* retry, addr_hi */
amdgpu_ring_write(ring, 0xffffffff); /* mask */
amdgpu_ring_write(ring, seq); /* value */
amdgpu_ring_write(ring, (3 << 28) | 0x20); /* func(equal) | poll interval */
}
/**
* si_dma_ring_test_ring - simple async dma engine test
*
* @ring: amdgpu_ring structure holding ring information
*
* Test the DMA engine by writing using it to write an
* value to memory. (VI).
* Returns 0 for success, error for failure.
*/
static int si_dma_ring_test_ring(struct amdgpu_ring *ring)
{
struct amdgpu_device *adev = ring->adev;
unsigned i;
unsigned index;
int r;
u32 tmp;
u64 gpu_addr;
r = amdgpu_wb_get(adev, &index);
if (r) {
dev_err(adev->dev, "(%d) failed to allocate wb slot\n", r);
return r;
}
gpu_addr = adev->wb.gpu_addr + (index * 4);
tmp = 0xCAFEDEAD;
adev->wb.wb[index] = cpu_to_le32(tmp);
r = amdgpu_ring_alloc(ring, 4);
if (r) {
DRM_ERROR("amdgpu: dma failed to lock ring %d (%d).\n", ring->idx, r);
amdgpu_wb_free(adev, index);
return r;
}
amdgpu_ring_write(ring, DMA_PACKET(DMA_PACKET_WRITE, 0, 0, 0, 1));
amdgpu_ring_write(ring, lower_32_bits(gpu_addr));
amdgpu_ring_write(ring, upper_32_bits(gpu_addr) & 0xff);
amdgpu_ring_write(ring, 0xDEADBEEF);
amdgpu_ring_commit(ring);
for (i = 0; i < adev->usec_timeout; i++) {
tmp = le32_to_cpu(adev->wb.wb[index]);
if (tmp == 0xDEADBEEF)
break;
DRM_UDELAY(1);
}
if (i < adev->usec_timeout) {
DRM_INFO("ring test on %d succeeded in %d usecs\n", ring->idx, i);
} else {
DRM_ERROR("amdgpu: ring %d test failed (0x%08X)\n",
ring->idx, tmp);
r = -EINVAL;
}
amdgpu_wb_free(adev, index);
return r;
}
static void si_dma_ring_emit_ib(struct amdgpu_ring *ring,
struct amdgpu_ib *ib,
unsigned vm_id, bool ctx_switch)
{
/* The indirect buffer packet must end on an 8 DW boundary in the DMA ring.
* Pad as necessary with NOPs.
*/
while ((ring->wptr & 7) != 5)
amdgpu_ring_write(ring, DMA_PACKET(DMA_PACKET_NOP, 0, 0, 0, 0));
amdgpu_ring_write(ring, DMA_IB_PACKET(DMA_PACKET_INDIRECT_BUFFER, vm_id, 0));
amdgpu_ring_write(ring, (ib->gpu_addr & 0xFFFFFFE0));
amdgpu_ring_write(ring, (ib->length_dw << 12) | (upper_32_bits(ib->gpu_addr) & 0xFF));
}
/**
* r600_dma_semaphore_ring_emit - emit a semaphore on the dma ring
*
* @rdev: radeon_device pointer
* @ring: radeon_ring structure holding ring information
* @semaphore: radeon semaphore object
* @emit_wait: wait or signal semaphore
*
* Add a DMA semaphore packet to the ring wait on or signal
* other rings (r6xx-SI).
*/
bool r600_dma_semaphore_ring_emit(struct radeon_device *rdev,
struct radeon_ring *ring,
struct radeon_semaphore *semaphore,
bool emit_wait)
{
u64 addr = semaphore->gpu_addr;
u32 s = emit_wait ? 0 : 1;
radeon_ring_write(ring, DMA_PACKET(DMA_PACKET_SEMAPHORE, 0, s, 0));
radeon_ring_write(ring, addr & 0xfffffffc);
radeon_ring_write(ring, upper_32_bits(addr) & 0xff);
return true;
}
/**
* si_dma_vm_write_pte - update PTEs by writing them manually
*
* @ib: indirect buffer to fill with commands
* @pe: addr of the page entry
* @value: dst addr to write into pe
* @count: number of page entries to update
* @incr: increase next addr by incr bytes
*
* Update PTEs by writing them manually using DMA (SI).
*/
static void si_dma_vm_write_pte(struct amdgpu_ib *ib, uint64_t pe,
uint64_t value, unsigned count,
uint32_t incr)
{
unsigned ndw = count * 2;
ib->ptr[ib->length_dw++] = DMA_PACKET(DMA_PACKET_WRITE, 0, 0, 0, ndw);
ib->ptr[ib->length_dw++] = lower_32_bits(pe);
ib->ptr[ib->length_dw++] = upper_32_bits(pe);
for (; ndw > 0; ndw -= 2) {
ib->ptr[ib->length_dw++] = lower_32_bits(value);
ib->ptr[ib->length_dw++] = upper_32_bits(value);
value += incr;
}
}
/**
* r600_dma_ring_test - simple async dma engine test
*
* @rdev: radeon_device pointer
* @ring: radeon_ring structure holding ring information
*
* Test the DMA engine by writing using it to write an
* value to memory. (r6xx-SI).
* Returns 0 for success, error for failure.
*/
int r600_dma_ring_test(struct radeon_device *rdev,
struct radeon_ring *ring)
{
unsigned i;
int r;
unsigned index;
u32 tmp;
u64 gpu_addr;
if (ring->idx == R600_RING_TYPE_DMA_INDEX)
index = R600_WB_DMA_RING_TEST_OFFSET;
else
index = CAYMAN_WB_DMA1_RING_TEST_OFFSET;
gpu_addr = rdev->wb.gpu_addr + index;
tmp = 0xCAFEDEAD;
rdev->wb.wb[index/4] = cpu_to_le32(tmp);
r = radeon_ring_lock(rdev, ring, 4);
if (r) {
DRM_ERROR("radeon: dma failed to lock ring %d (%d).\n", ring->idx, r);
return r;
}
radeon_ring_write(ring, DMA_PACKET(DMA_PACKET_WRITE, 0, 0, 1));
radeon_ring_write(ring, lower_32_bits(gpu_addr));
radeon_ring_write(ring, upper_32_bits(gpu_addr) & 0xff);
radeon_ring_write(ring, 0xDEADBEEF);
radeon_ring_unlock_commit(rdev, ring, false);
for (i = 0; i < rdev->usec_timeout; i++) {
tmp = le32_to_cpu(rdev->wb.wb[index/4]);
if (tmp == 0xDEADBEEF)
break;
DRM_UDELAY(1);
}
if (i < rdev->usec_timeout) {
DRM_INFO("ring test on %d succeeded in %d usecs\n", ring->idx, i);
} else {
DRM_ERROR("radeon: ring %d test failed (0x%08X)\n",
ring->idx, tmp);
r = -EINVAL;
}
return r;
}
/**
* r600_dma_ring_test - simple async dma engine test
*
* @rdev: radeon_device pointer
* @ring: radeon_ring structure holding ring information
*
* Test the DMA engine by writing using it to write an
* value to memory. (r6xx-SI).
* Returns 0 for success, error for failure.
*/
int r600_dma_ring_test(struct radeon_device *rdev,
struct radeon_ring *ring)
{
unsigned i;
int r;
void __iomem *ptr = (void *)rdev->vram_scratch.ptr;
u32 tmp;
if (!ptr) {
DRM_ERROR("invalid vram scratch pointer\n");
return -EINVAL;
}
tmp = 0xCAFEDEAD;
writel(tmp, ptr);
r = radeon_ring_lock(rdev, ring, 4);
if (r) {
DRM_ERROR("radeon: dma failed to lock ring %d (%d).\n", ring->idx, r);
return r;
}
radeon_ring_write(ring, DMA_PACKET(DMA_PACKET_WRITE, 0, 0, 1));
radeon_ring_write(ring, rdev->vram_scratch.gpu_addr & 0xfffffffc);
radeon_ring_write(ring, upper_32_bits(rdev->vram_scratch.gpu_addr) & 0xff);
radeon_ring_write(ring, 0xDEADBEEF);
radeon_ring_unlock_commit(rdev, ring);
for (i = 0; i < rdev->usec_timeout; i++) {
tmp = readl(ptr);
if (tmp == 0xDEADBEEF)
break;
DRM_UDELAY(1);
}
if (i < rdev->usec_timeout) {
DRM_INFO("ring test on %d succeeded in %d usecs\n", ring->idx, i);
} else {
DRM_ERROR("radeon: ring %d test failed (0x%08X)\n",
ring->idx, tmp);
r = -EINVAL;
}
return r;
}
void r600_dma_copy_buffer(struct r600_context *rctx,
struct pipe_resource *dst,
struct pipe_resource *src,
uint64_t dst_offset,
uint64_t src_offset,
uint64_t size)
{
struct radeon_winsys_cs *cs = rctx->b.dma.cs;
unsigned i, ncopy, csize;
struct r600_resource *rdst = (struct r600_resource*)dst;
struct r600_resource *rsrc = (struct r600_resource*)src;
/* Mark the buffer range of destination as valid (initialized),
* so that transfer_map knows it should wait for the GPU when mapping
* that range. */
util_range_add(&rdst->valid_buffer_range, dst_offset,
dst_offset + size);
size >>= 2; /* convert to dwords */
ncopy = (size / R600_DMA_COPY_MAX_SIZE_DW) + !!(size % R600_DMA_COPY_MAX_SIZE_DW);
r600_need_dma_space(&rctx->b, ncopy * 5, rdst, rsrc);
for (i = 0; i < ncopy; i++) {
csize = size < R600_DMA_COPY_MAX_SIZE_DW ? size : R600_DMA_COPY_MAX_SIZE_DW;
/* emit reloc before writing cs so that cs is always in consistent state */
radeon_add_to_buffer_list(&rctx->b, &rctx->b.dma, rsrc, RADEON_USAGE_READ,
RADEON_PRIO_SDMA_BUFFER);
radeon_add_to_buffer_list(&rctx->b, &rctx->b.dma, rdst, RADEON_USAGE_WRITE,
RADEON_PRIO_SDMA_BUFFER);
radeon_emit(cs, DMA_PACKET(DMA_PACKET_COPY, 0, 0, csize));
radeon_emit(cs, dst_offset & 0xfffffffc);
radeon_emit(cs, src_offset & 0xfffffffc);
radeon_emit(cs, (dst_offset >> 32UL) & 0xff);
radeon_emit(cs, (src_offset >> 32UL) & 0xff);
dst_offset += csize << 2;
src_offset += csize << 2;
size -= csize;
}
r600_dma_emit_wait_idle(&rctx->b);
}
.sw_fini = si_dma_sw_fini,
.hw_init = si_dma_hw_init,
.hw_fini = si_dma_hw_fini,
.suspend = si_dma_suspend,
.resume = si_dma_resume,
.is_idle = si_dma_is_idle,
.wait_for_idle = si_dma_wait_for_idle,
.soft_reset = si_dma_soft_reset,
.set_clockgating_state = si_dma_set_clockgating_state,
.set_powergating_state = si_dma_set_powergating_state,
};
static const struct amdgpu_ring_funcs si_dma_ring_funcs = {
.type = AMDGPU_RING_TYPE_SDMA,
.align_mask = 0xf,
.nop = DMA_PACKET(DMA_PACKET_NOP, 0, 0, 0, 0),
.get_rptr = si_dma_ring_get_rptr,
.get_wptr = si_dma_ring_get_wptr,
.set_wptr = si_dma_ring_set_wptr,
.emit_frame_size =
3 + /* si_dma_ring_emit_hdp_flush */
3 + /* si_dma_ring_emit_hdp_invalidate */
6 + /* si_dma_ring_emit_pipeline_sync */
12 + /* si_dma_ring_emit_vm_flush */
9 + 9 + 9, /* si_dma_ring_emit_fence x3 for user fence, vm fence */
.emit_ib_size = 7 + 3, /* si_dma_ring_emit_ib */
.emit_ib = si_dma_ring_emit_ib,
.emit_fence = si_dma_ring_emit_fence,
.emit_pipeline_sync = si_dma_ring_emit_pipeline_sync,
.emit_vm_flush = si_dma_ring_emit_vm_flush,
.emit_hdp_flush = si_dma_ring_emit_hdp_flush,
/**
* si_dma_pad_ib - pad the IB to the required number of dw
*
* @ib: indirect buffer to fill with padding
*
*/
static void si_dma_ring_pad_ib(struct amdgpu_ring *ring, struct amdgpu_ib *ib)
{
while (ib->length_dw & 0x7)
ib->ptr[ib->length_dw++] = DMA_PACKET(DMA_PACKET_NOP, 0, 0, 0, 0);
}
static void si_dma_ring_emit_hdp_flush(struct amdgpu_ring *ring)
{
amdgpu_ring_write(ring, DMA_PACKET(DMA_PACKET_SRBM_WRITE, 0, 0, 0, 0));
amdgpu_ring_write(ring, (0xf << 16) | (HDP_MEM_COHERENCY_FLUSH_CNTL));
amdgpu_ring_write(ring, 1);
}
static void si_dma_ring_emit_hdp_invalidate(struct amdgpu_ring *ring)
{
amdgpu_ring_write(ring, DMA_PACKET(DMA_PACKET_SRBM_WRITE, 0, 0, 0, 0));
amdgpu_ring_write(ring, (0xf << 16) | (HDP_DEBUG0));
amdgpu_ring_write(ring, 1);
}
