1 files changed, 105 insertions, 129 deletions

diff --git a/drivers/gpu/drm/amd/amdgpu/sdma_v2_4.c b/drivers/gpu/drm/amd/amdgpu/sdma_v2_4.c
index f4c3130d3fdb..565dab3c7218 100644
--- a/drivers/gpu/drm/amd/amdgpu/sdma_v2_4.c
+++ b/drivers/gpu/drm/amd/amdgpu/sdma_v2_4.c
@@ -105,6 +105,15 @@ static void sdma_v2_4_init_golden_registers(struct amdgpu_device *adev)
 	}
 }
 
+static void sdma_v2_4_free_microcode(struct amdgpu_device *adev)
+{
+	int i;
+	for (i = 0; i < adev->sdma.num_instances; i++) {
+		release_firmware(adev->sdma.instance[i].fw);
+		adev->sdma.instance[i].fw = NULL;
+	}
+}
+
 /**
  * sdma_v2_4_init_microcode - load ucode images from disk
  *
@@ -181,12 +190,8 @@ out:
  */
 static uint32_t sdma_v2_4_ring_get_rptr(struct amdgpu_ring *ring)
 {
-	u32 rptr;
-
 	/* XXX check if swapping is necessary on BE */
-	rptr = ring->adev->wb.wb[ring->rptr_offs] >> 2;
-
-	return rptr;
+	return ring->adev->wb.wb[ring->rptr_offs] >> 2;
 }
 
 /**
@@ -246,19 +251,6 @@ static void sdma_v2_4_ring_emit_ib(struct amdgpu_ring *ring,
 				   unsigned vm_id, bool ctx_switch)
 {
 	u32 vmid = vm_id & 0xf;
-	u32 next_rptr = ring->wptr + 5;
-
-	while ((next_rptr & 7) != 2)
-		next_rptr++;
-
-	next_rptr += 6;
-
-	amdgpu_ring_write(ring, SDMA_PKT_HEADER_OP(SDMA_OP_WRITE) |
-			  SDMA_PKT_HEADER_SUB_OP(SDMA_SUBOP_WRITE_LINEAR));
-	amdgpu_ring_write(ring, lower_32_bits(ring->next_rptr_gpu_addr) & 0xfffffffc);
-	amdgpu_ring_write(ring, upper_32_bits(ring->next_rptr_gpu_addr));
-	amdgpu_ring_write(ring, SDMA_PKT_WRITE_UNTILED_DW_3_COUNT(1));
-	amdgpu_ring_write(ring, next_rptr);
 
 	/* IB packet must end on a 8 DW boundary */
 	sdma_v2_4_ring_insert_nop(ring, (10 - (ring->wptr & 7)) % 8);
@@ -397,7 +389,7 @@ static void sdma_v2_4_enable(struct amdgpu_device *adev, bool enable)
 	u32 f32_cntl;
 	int i;
 
-	if (enable == false) {
+	if (!enable) {
 		sdma_v2_4_gfx_stop(adev);
 		sdma_v2_4_rlc_stop(adev);
 	}
@@ -461,6 +453,8 @@ static int sdma_v2_4_gfx_resume(struct amdgpu_device *adev)
 		/* Initialize the ring buffer's read and write pointers */
 		WREG32(mmSDMA0_GFX_RB_RPTR + sdma_offsets[i], 0);
 		WREG32(mmSDMA0_GFX_RB_WPTR + sdma_offsets[i], 0);
+		WREG32(mmSDMA0_GFX_IB_RPTR + sdma_offsets[i], 0);
+		WREG32(mmSDMA0_GFX_IB_OFFSET + sdma_offsets[i], 0);
 
 		/* set the wb address whether it's enabled or not */
 		WREG32(mmSDMA0_GFX_RB_RPTR_ADDR_HI + sdma_offsets[i],
@@ -489,7 +483,11 @@ static int sdma_v2_4_gfx_resume(struct amdgpu_device *adev)
 		WREG32(mmSDMA0_GFX_IB_CNTL + sdma_offsets[i], ib_cntl);
 
 		ring->ready = true;
+	}
 
+	sdma_v2_4_enable(adev, true);
+	for (i = 0; i < adev->sdma.num_instances; i++) {
+		ring = &adev->sdma.instance[i].ring;
 		r = amdgpu_ring_test_ring(ring);
 		if (r) {
 			ring->ready = false;
@@ -565,23 +563,25 @@ static int sdma_v2_4_start(struct amdgpu_device *adev)
 {
 	int r;
 
-	if (!adev->firmware.smu_load) {
-		r = sdma_v2_4_load_microcode(adev);
-		if (r)
-			return r;
-	} else {
-		r = adev->smu.smumgr_funcs->check_fw_load_finish(adev,
-						AMDGPU_UCODE_ID_SDMA0);
-		if (r)
-			return -EINVAL;
-		r = adev->smu.smumgr_funcs->check_fw_load_finish(adev,
-						AMDGPU_UCODE_ID_SDMA1);
-		if (r)
-			return -EINVAL;
+	if (!adev->pp_enabled) {
+		if (!adev->firmware.smu_load) {
+			r = sdma_v2_4_load_microcode(adev);
+			if (r)
+				return r;
+		} else {
+			r = adev->smu.smumgr_funcs->check_fw_load_finish(adev,
+							AMDGPU_UCODE_ID_SDMA0);
+			if (r)
+				return -EINVAL;
+			r = adev->smu.smumgr_funcs->check_fw_load_finish(adev,
+							AMDGPU_UCODE_ID_SDMA1);
+			if (r)
+				return -EINVAL;
+		}
 	}
 
-	/* unhalt the MEs */
-	sdma_v2_4_enable(adev, true);
+	/* halt the engine before programing */
+	sdma_v2_4_enable(adev, false);
 
 	/* start the gfx rings and rlc compute queues */
 	r = sdma_v2_4_gfx_resume(adev);
@@ -664,20 +664,19 @@ static int sdma_v2_4_ring_test_ring(struct amdgpu_ring *ring)
  * Test a simple IB in the DMA ring (VI).
  * Returns 0 on success, error on failure.
  */
-static int sdma_v2_4_ring_test_ib(struct amdgpu_ring *ring)
+static int sdma_v2_4_ring_test_ib(struct amdgpu_ring *ring, long timeout)
 {
 	struct amdgpu_device *adev = ring->adev;
 	struct amdgpu_ib ib;
 	struct fence *f = NULL;
-	unsigned i;
 	unsigned index;
-	int r;
 	u32 tmp = 0;
 	u64 gpu_addr;
+	long r;
 
 	r = amdgpu_wb_get(adev, &index);
 	if (r) {
-		dev_err(adev->dev, "(%d) failed to allocate wb slot\n", r);
+		dev_err(adev->dev, "(%ld) failed to allocate wb slot\n", r);
 		return r;
 	}
 
@@ -687,7 +686,7 @@ static int sdma_v2_4_ring_test_ib(struct amdgpu_ring *ring)
 	memset(&ib, 0, sizeof(ib));
 	r = amdgpu_ib_get(adev, NULL, 256, &ib);
 	if (r) {
-		DRM_ERROR("amdgpu: failed to get ib (%d).\n", r);
+		DRM_ERROR("amdgpu: failed to get ib (%ld).\n", r);
 		goto err0;
 	}
 
@@ -706,28 +705,25 @@ static int sdma_v2_4_ring_test_ib(struct amdgpu_ring *ring)
 	if (r)
 		goto err1;
 
-	r = fence_wait(f, false);
-	if (r) {
-		DRM_ERROR("amdgpu: fence wait failed (%d).\n", r);
+	r = fence_wait_timeout(f, false, timeout);
+	if (r == 0) {
+		DRM_ERROR("amdgpu: IB test timed out\n");
+		r = -ETIMEDOUT;
 		goto err1;
-	}
-	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("ib test on ring %d succeeded in %u usecs\n",
-			 ring->idx, i);
+	} 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:
-	fence_put(f);
 	amdgpu_ib_free(adev, &ib, NULL);
 	fence_put(f);
 err0:
@@ -749,24 +745,16 @@ static void sdma_v2_4_vm_copy_pte(struct amdgpu_ib *ib,
 				  uint64_t pe, uint64_t src,
 				  unsigned count)
 {
-	while (count) {
-		unsigned bytes = count * 8;
-		if (bytes > 0x1FFFF8)
-			bytes = 0x1FFFF8;
-
-		ib->ptr[ib->length_dw++] = SDMA_PKT_HEADER_OP(SDMA_OP_COPY) |
-			SDMA_PKT_HEADER_SUB_OP(SDMA_SUBOP_COPY_LINEAR);
-		ib->ptr[ib->length_dw++] = bytes;
-		ib->ptr[ib->length_dw++] = 0; /* src/dst endian swap */
-		ib->ptr[ib->length_dw++] = lower_32_bits(src);
-		ib->ptr[ib->length_dw++] = upper_32_bits(src);
-		ib->ptr[ib->length_dw++] = lower_32_bits(pe);
-		ib->ptr[ib->length_dw++] = upper_32_bits(pe);
-
-		pe += bytes;
-		src += bytes;
-		count -= bytes / 8;
-	}
+	unsigned bytes = count * 8;
+
+	ib->ptr[ib->length_dw++] = SDMA_PKT_HEADER_OP(SDMA_OP_COPY) |
+		SDMA_PKT_HEADER_SUB_OP(SDMA_SUBOP_COPY_LINEAR);
+	ib->ptr[ib->length_dw++] = bytes;
+	ib->ptr[ib->length_dw++] = 0; /* src/dst endian swap */
+	ib->ptr[ib->length_dw++] = lower_32_bits(src);
+	ib->ptr[ib->length_dw++] = upper_32_bits(src);
+	ib->ptr[ib->length_dw++] = lower_32_bits(pe);
+	ib->ptr[ib->length_dw++] = upper_32_bits(pe);
 }
 
 /**
@@ -774,39 +762,27 @@ static void sdma_v2_4_vm_copy_pte(struct amdgpu_ib *ib,
  *
  * @ib: indirect buffer to fill with commands
  * @pe: addr of the page entry
- * @addr: dst addr to write into pe
+ * @value: dst addr to write into pe
  * @count: number of page entries to update
  * @incr: increase next addr by incr bytes
- * @flags: access flags
  *
  * Update PTEs by writing them manually using sDMA (CIK).
  */
-static void sdma_v2_4_vm_write_pte(struct amdgpu_ib *ib,
-				   const dma_addr_t *pages_addr, uint64_t pe,
-				   uint64_t addr, unsigned count,
-				   uint32_t incr, uint32_t flags)
+static void sdma_v2_4_vm_write_pte(struct amdgpu_ib *ib, uint64_t pe,
+				   uint64_t value, unsigned count,
+				   uint32_t incr)
 {
-	uint64_t value;
-	unsigned ndw;
-
-	while (count) {
-		ndw = count * 2;
-		if (ndw > 0xFFFFE)
-			ndw = 0xFFFFE;
-
-		/* for non-physically contiguous pages (system) */
-		ib->ptr[ib->length_dw++] = SDMA_PKT_HEADER_OP(SDMA_OP_WRITE) |
-			SDMA_PKT_HEADER_SUB_OP(SDMA_SUBOP_COPY_LINEAR);
-		ib->ptr[ib->length_dw++] = pe;
-		ib->ptr[ib->length_dw++] = upper_32_bits(pe);
-		ib->ptr[ib->length_dw++] = ndw;
-		for (; ndw > 0; ndw -= 2, --count, pe += 8) {
-			value = amdgpu_vm_map_gart(pages_addr, addr);
-			addr += incr;
-			value |= flags;
-			ib->ptr[ib->length_dw++] = value;
-			ib->ptr[ib->length_dw++] = upper_32_bits(value);
-		}
+	unsigned ndw = count * 2;
+
+	ib->ptr[ib->length_dw++] = SDMA_PKT_HEADER_OP(SDMA_OP_WRITE) |
+		SDMA_PKT_HEADER_SUB_OP(SDMA_SUBOP_COPY_LINEAR);
+	ib->ptr[ib->length_dw++] = pe;
+	ib->ptr[ib->length_dw++] = upper_32_bits(pe);
+	ib->ptr[ib->length_dw++] = ndw;
+	for (; ndw > 0; ndw -= 2, --count, pe += 8) {
+		ib->ptr[ib->length_dw++] = lower_32_bits(value);
+		ib->ptr[ib->length_dw++] = upper_32_bits(value);
+		value += incr;
 	}
 }
 
@@ -822,40 +798,21 @@ static void sdma_v2_4_vm_write_pte(struct amdgpu_ib *ib,
  *
  * Update the page tables using sDMA (CIK).
  */
-static void sdma_v2_4_vm_set_pte_pde(struct amdgpu_ib *ib,
-				     uint64_t pe,
+static void sdma_v2_4_vm_set_pte_pde(struct amdgpu_ib *ib, uint64_t pe,
 				     uint64_t addr, unsigned count,
 				     uint32_t incr, uint32_t flags)
 {
-	uint64_t value;
-	unsigned ndw;
-
-	while (count) {
-		ndw = count;
-		if (ndw > 0x7FFFF)
-			ndw = 0x7FFFF;
-
-		if (flags & AMDGPU_PTE_VALID)
-			value = addr;
-		else
-			value = 0;
-
-		/* for physically contiguous pages (vram) */
-		ib->ptr[ib->length_dw++] = SDMA_PKT_HEADER_OP(SDMA_OP_GEN_PTEPDE);
-		ib->ptr[ib->length_dw++] = pe; /* dst addr */
-		ib->ptr[ib->length_dw++] = upper_32_bits(pe);
-		ib->ptr[ib->length_dw++] = flags; /* mask */
-		ib->ptr[ib->length_dw++] = 0;
-		ib->ptr[ib->length_dw++] = value; /* value */
-		ib->ptr[ib->length_dw++] = upper_32_bits(value);
-		ib->ptr[ib->length_dw++] = incr; /* increment size */
-		ib->ptr[ib->length_dw++] = 0;
-		ib->ptr[ib->length_dw++] = ndw; /* number of entries */
-
-		pe += ndw * 8;
-		addr += ndw * incr;
-		count -= ndw;
-	}
+	/* for physically contiguous pages (vram) */
+	ib->ptr[ib->length_dw++] = SDMA_PKT_HEADER_OP(SDMA_OP_GEN_PTEPDE);
+	ib->ptr[ib->length_dw++] = lower_32_bits(pe); /* dst addr */
+	ib->ptr[ib->length_dw++] = upper_32_bits(pe);
+	ib->ptr[ib->length_dw++] = flags; /* mask */
+	ib->ptr[ib->length_dw++] = 0;
+	ib->ptr[ib->length_dw++] = lower_32_bits(addr); /* value */
+	ib->ptr[ib->length_dw++] = upper_32_bits(addr);
+	ib->ptr[ib->length_dw++] = incr; /* increment size */
+	ib->ptr[ib->length_dw++] = 0;
+	ib->ptr[ib->length_dw++] = count; /* number of entries */
 }
 
 /**
@@ -945,6 +902,22 @@ static void sdma_v2_4_ring_emit_vm_flush(struct amdgpu_ring *ring,
 			  SDMA_PKT_POLL_REGMEM_DW5_INTERVAL(10)); /* retry count, poll interval */
 }
 
+static unsigned sdma_v2_4_ring_get_emit_ib_size(struct amdgpu_ring *ring)
+{
+	return
+		7 + 6; /* sdma_v2_4_ring_emit_ib */
+}
+
+static unsigned sdma_v2_4_ring_get_dma_frame_size(struct amdgpu_ring *ring)
+{
+	return
+		6 + /* sdma_v2_4_ring_emit_hdp_flush */
+		3 + /* sdma_v2_4_ring_emit_hdp_invalidate */
+		6 + /* sdma_v2_4_ring_emit_pipeline_sync */
+		12 + /* sdma_v2_4_ring_emit_vm_flush */
+		10 + 10 + 10; /* sdma_v2_4_ring_emit_fence x3 for user fence, vm fence */
+}
+
 static int sdma_v2_4_early_init(void *handle)
 {
 	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
@@ -1012,6 +985,7 @@ static int sdma_v2_4_sw_fini(void *handle)
 	for (i = 0; i < adev->sdma.num_instances; i++)
 		amdgpu_ring_fini(&adev->sdma.instance[i].ring);
 
+	sdma_v2_4_free_microcode(adev);
 	return 0;
 }
 
@@ -1262,6 +1236,8 @@ static const struct amdgpu_ring_funcs sdma_v2_4_ring_funcs = {
 	.test_ib = sdma_v2_4_ring_test_ib,
 	.insert_nop = sdma_v2_4_ring_insert_nop,
 	.pad_ib = sdma_v2_4_ring_pad_ib,
+	.get_emit_ib_size = sdma_v2_4_ring_get_emit_ib_size,
+	.get_dma_frame_size = sdma_v2_4_ring_get_dma_frame_size,
 };
 
 static void sdma_v2_4_set_ring_funcs(struct amdgpu_device *adev)