1 files changed, 334 insertions, 91 deletions

diff --git a/drivers/gpu/drm/amd/amdgpu/gfx_v7_0.c b/drivers/gpu/drm/amd/amdgpu/gfx_v7_0.c
index 7f18a53ab53a..71116da9e782 100644
--- a/drivers/gpu/drm/amd/amdgpu/gfx_v7_0.c
+++ b/drivers/gpu/drm/amd/amdgpu/gfx_v7_0.c
@@ -991,6 +991,22 @@ out:
 	return err;
 }
 
+static void gfx_v7_0_free_microcode(struct amdgpu_device *adev)
+{
+	release_firmware(adev->gfx.pfp_fw);
+	adev->gfx.pfp_fw = NULL;
+	release_firmware(adev->gfx.me_fw);
+	adev->gfx.me_fw = NULL;
+	release_firmware(adev->gfx.ce_fw);
+	adev->gfx.ce_fw = NULL;
+	release_firmware(adev->gfx.mec_fw);
+	adev->gfx.mec_fw = NULL;
+	release_firmware(adev->gfx.mec2_fw);
+	adev->gfx.mec2_fw = NULL;
+	release_firmware(adev->gfx.rlc_fw);
+	adev->gfx.rlc_fw = NULL;
+}
+
 /**
  * gfx_v7_0_tiling_mode_table_init - init the hw tiling table
  *
@@ -1567,9 +1583,15 @@ static void gfx_v7_0_tiling_mode_table_init(struct amdgpu_device *adev)
  * registers are instanced per SE or SH.  0xffffffff means
  * broadcast to all SEs or SHs (CIK).
  */
-void gfx_v7_0_select_se_sh(struct amdgpu_device *adev, u32 se_num, u32 sh_num)
+static void gfx_v7_0_select_se_sh(struct amdgpu_device *adev,
+				  u32 se_num, u32 sh_num, u32 instance)
 {
-	u32 data = GRBM_GFX_INDEX__INSTANCE_BROADCAST_WRITES_MASK;
+	u32 data;
+
+	if (instance == 0xffffffff)
+		data = REG_SET_FIELD(0, GRBM_GFX_INDEX, INSTANCE_BROADCAST_WRITES, 1);
+	else
+		data = REG_SET_FIELD(0, GRBM_GFX_INDEX, INSTANCE_INDEX, instance);
 
 	if ((se_num == 0xffffffff) && (sh_num == 0xffffffff))
 		data |= GRBM_GFX_INDEX__SH_BROADCAST_WRITES_MASK |
@@ -1623,6 +1645,147 @@ static u32 gfx_v7_0_get_rb_active_bitmap(struct amdgpu_device *adev)
 	return (~data) & mask;
 }
 
+static void
+gfx_v7_0_raster_config(struct amdgpu_device *adev, u32 *rconf, u32 *rconf1)
+{
+	switch (adev->asic_type) {
+	case CHIP_BONAIRE:
+		*rconf |= RB_MAP_PKR0(2) | RB_XSEL2(1) | SE_MAP(2) |
+			  SE_XSEL(1) | SE_YSEL(1);
+		*rconf1 |= 0x0;
+		break;
+	case CHIP_HAWAII:
+		*rconf |= RB_MAP_PKR0(2) | RB_MAP_PKR1(2) |
+			  RB_XSEL2(1) | PKR_MAP(2) | PKR_XSEL(1) |
+			  PKR_YSEL(1) | SE_MAP(2) | SE_XSEL(2) |
+			  SE_YSEL(3);
+		*rconf1 |= SE_PAIR_MAP(2) | SE_PAIR_XSEL(3) |
+			   SE_PAIR_YSEL(2);
+		break;
+	case CHIP_KAVERI:
+		*rconf |= RB_MAP_PKR0(2);
+		*rconf1 |= 0x0;
+		break;
+	case CHIP_KABINI:
+	case CHIP_MULLINS:
+		*rconf |= 0x0;
+		*rconf1 |= 0x0;
+		break;
+	default:
+		DRM_ERROR("unknown asic: 0x%x\n", adev->asic_type);
+		break;
+	}
+}
+
+static void
+gfx_v7_0_write_harvested_raster_configs(struct amdgpu_device *adev,
+					u32 raster_config, u32 raster_config_1,
+					unsigned rb_mask, unsigned num_rb)
+{
+	unsigned sh_per_se = max_t(unsigned, adev->gfx.config.max_sh_per_se, 1);
+	unsigned num_se = max_t(unsigned, adev->gfx.config.max_shader_engines, 1);
+	unsigned rb_per_pkr = min_t(unsigned, num_rb / num_se / sh_per_se, 2);
+	unsigned rb_per_se = num_rb / num_se;
+	unsigned se_mask[4];
+	unsigned se;
+
+	se_mask[0] = ((1 << rb_per_se) - 1) & rb_mask;
+	se_mask[1] = (se_mask[0] << rb_per_se) & rb_mask;
+	se_mask[2] = (se_mask[1] << rb_per_se) & rb_mask;
+	se_mask[3] = (se_mask[2] << rb_per_se) & rb_mask;
+
+	WARN_ON(!(num_se == 1 || num_se == 2 || num_se == 4));
+	WARN_ON(!(sh_per_se == 1 || sh_per_se == 2));
+	WARN_ON(!(rb_per_pkr == 1 || rb_per_pkr == 2));
+
+	if ((num_se > 2) && ((!se_mask[0] && !se_mask[1]) ||
+			     (!se_mask[2] && !se_mask[3]))) {
+		raster_config_1 &= ~SE_PAIR_MAP_MASK;
+
+		if (!se_mask[0] && !se_mask[1]) {
+			raster_config_1 |=
+				SE_PAIR_MAP(RASTER_CONFIG_SE_PAIR_MAP_3);
+		} else {
+			raster_config_1 |=
+				SE_PAIR_MAP(RASTER_CONFIG_SE_PAIR_MAP_0);
+		}
+	}
+
+	for (se = 0; se < num_se; se++) {
+		unsigned raster_config_se = raster_config;
+		unsigned pkr0_mask = ((1 << rb_per_pkr) - 1) << (se * rb_per_se);
+		unsigned pkr1_mask = pkr0_mask << rb_per_pkr;
+		int idx = (se / 2) * 2;
+
+		if ((num_se > 1) && (!se_mask[idx] || !se_mask[idx + 1])) {
+			raster_config_se &= ~SE_MAP_MASK;
+
+			if (!se_mask[idx]) {
+				raster_config_se |= SE_MAP(RASTER_CONFIG_SE_MAP_3);
+			} else {
+				raster_config_se |= SE_MAP(RASTER_CONFIG_SE_MAP_0);
+			}
+		}
+
+		pkr0_mask &= rb_mask;
+		pkr1_mask &= rb_mask;
+		if (rb_per_se > 2 && (!pkr0_mask || !pkr1_mask)) {
+			raster_config_se &= ~PKR_MAP_MASK;
+
+			if (!pkr0_mask) {
+				raster_config_se |= PKR_MAP(RASTER_CONFIG_PKR_MAP_3);
+			} else {
+				raster_config_se |= PKR_MAP(RASTER_CONFIG_PKR_MAP_0);
+			}
+		}
+
+		if (rb_per_se >= 2) {
+			unsigned rb0_mask = 1 << (se * rb_per_se);
+			unsigned rb1_mask = rb0_mask << 1;
+
+			rb0_mask &= rb_mask;
+			rb1_mask &= rb_mask;
+			if (!rb0_mask || !rb1_mask) {
+				raster_config_se &= ~RB_MAP_PKR0_MASK;
+
+				if (!rb0_mask) {
+					raster_config_se |=
+						RB_MAP_PKR0(RASTER_CONFIG_RB_MAP_3);
+				} else {
+					raster_config_se |=
+						RB_MAP_PKR0(RASTER_CONFIG_RB_MAP_0);
+				}
+			}
+
+			if (rb_per_se > 2) {
+				rb0_mask = 1 << (se * rb_per_se + rb_per_pkr);
+				rb1_mask = rb0_mask << 1;
+				rb0_mask &= rb_mask;
+				rb1_mask &= rb_mask;
+				if (!rb0_mask || !rb1_mask) {
+					raster_config_se &= ~RB_MAP_PKR1_MASK;
+
+					if (!rb0_mask) {
+						raster_config_se |=
+							RB_MAP_PKR1(RASTER_CONFIG_RB_MAP_3);
+					} else {
+						raster_config_se |=
+							RB_MAP_PKR1(RASTER_CONFIG_RB_MAP_0);
+					}
+				}
+			}
+		}
+
+		/* GRBM_GFX_INDEX has a different offset on CI+ */
+		gfx_v7_0_select_se_sh(adev, se, 0xffffffff, 0xffffffff);
+		WREG32(mmPA_SC_RASTER_CONFIG, raster_config_se);
+		WREG32(mmPA_SC_RASTER_CONFIG_1, raster_config_1);
+	}
+
+	/* GRBM_GFX_INDEX has a different offset on CI+ */
+	gfx_v7_0_select_se_sh(adev, 0xffffffff, 0xffffffff, 0xffffffff);
+}
+
 /**
  * gfx_v7_0_setup_rb - setup the RBs on the asic
  *
@@ -1636,24 +1799,41 @@ static void gfx_v7_0_setup_rb(struct amdgpu_device *adev)
 {
 	int i, j;
 	u32 data;
+	u32 raster_config = 0, raster_config_1 = 0;
 	u32 active_rbs = 0;
 	u32 rb_bitmap_width_per_sh = adev->gfx.config.max_backends_per_se /
 					adev->gfx.config.max_sh_per_se;
+	unsigned num_rb_pipes;
 
 	mutex_lock(&adev->grbm_idx_mutex);
 	for (i = 0; i < adev->gfx.config.max_shader_engines; i++) {
 		for (j = 0; j < adev->gfx.config.max_sh_per_se; j++) {
-			gfx_v7_0_select_se_sh(adev, i, j);
+			gfx_v7_0_select_se_sh(adev, i, j, 0xffffffff);
 			data = gfx_v7_0_get_rb_active_bitmap(adev);
 			active_rbs |= data << ((i * adev->gfx.config.max_sh_per_se + j) *
 					       rb_bitmap_width_per_sh);
 		}
 	}
-	gfx_v7_0_select_se_sh(adev, 0xffffffff, 0xffffffff);
-	mutex_unlock(&adev->grbm_idx_mutex);
+	gfx_v7_0_select_se_sh(adev, 0xffffffff, 0xffffffff, 0xffffffff);
 
 	adev->gfx.config.backend_enable_mask = active_rbs;
 	adev->gfx.config.num_rbs = hweight32(active_rbs);
+
+	num_rb_pipes = min_t(unsigned, adev->gfx.config.max_backends_per_se *
+			     adev->gfx.config.max_shader_engines, 16);
+
+	gfx_v7_0_raster_config(adev, &raster_config, &raster_config_1);
+
+	if (!adev->gfx.config.backend_enable_mask ||
+			adev->gfx.config.num_rbs >= num_rb_pipes) {
+		WREG32(mmPA_SC_RASTER_CONFIG, raster_config);
+		WREG32(mmPA_SC_RASTER_CONFIG_1, raster_config_1);
+	} else {
+		gfx_v7_0_write_harvested_raster_configs(adev, raster_config, raster_config_1,
+							adev->gfx.config.backend_enable_mask,
+							num_rb_pipes);
+	}
+	mutex_unlock(&adev->grbm_idx_mutex);
 }
 
 /**
@@ -1730,7 +1910,7 @@ static void gfx_v7_0_gpu_init(struct amdgpu_device *adev)
 	 * making sure that the following register writes will be broadcasted
 	 * to all the shaders
 	 */
-	gfx_v7_0_select_se_sh(adev, 0xffffffff, 0xffffffff);
+	gfx_v7_0_select_se_sh(adev, 0xffffffff, 0xffffffff, 0xffffffff);
 
 	/* XXX SH_MEM regs */
 	/* where to put LDS, scratch, GPUVM in FSA64 space */
@@ -2034,17 +2214,6 @@ static void gfx_v7_0_ring_emit_ib_gfx(struct amdgpu_ring *ring,
 				      unsigned vm_id, bool ctx_switch)
 {
 	u32 header, control = 0;
-	u32 next_rptr = ring->wptr + 5;
-
-	if (ctx_switch)
-		next_rptr += 2;
-
-	next_rptr += 4;
-	amdgpu_ring_write(ring, PACKET3(PACKET3_WRITE_DATA, 3));
-	amdgpu_ring_write(ring, WRITE_DATA_DST_SEL(5) | WR_CONFIRM);
-	amdgpu_ring_write(ring, ring->next_rptr_gpu_addr & 0xfffffffc);
-	amdgpu_ring_write(ring, upper_32_bits(ring->next_rptr_gpu_addr) & 0xffffffff);
-	amdgpu_ring_write(ring, next_rptr);
 
 	/* insert SWITCH_BUFFER packet before first IB in the ring frame */
 	if (ctx_switch) {
@@ -2073,22 +2242,9 @@ static void gfx_v7_0_ring_emit_ib_compute(struct amdgpu_ring *ring,
 					  struct amdgpu_ib *ib,
 					  unsigned vm_id, bool ctx_switch)
 {
-	u32 header, control = 0;
-	u32 next_rptr = ring->wptr + 5;
+	u32 control = INDIRECT_BUFFER_VALID | ib->length_dw | (vm_id << 24);
 
-	control |= INDIRECT_BUFFER_VALID;
-	next_rptr += 4;
-	amdgpu_ring_write(ring, PACKET3(PACKET3_WRITE_DATA, 3));
-	amdgpu_ring_write(ring, WRITE_DATA_DST_SEL(5) | WR_CONFIRM);
-	amdgpu_ring_write(ring, ring->next_rptr_gpu_addr & 0xfffffffc);
-	amdgpu_ring_write(ring, upper_32_bits(ring->next_rptr_gpu_addr) & 0xffffffff);
-	amdgpu_ring_write(ring, next_rptr);
-
-	header = PACKET3(PACKET3_INDIRECT_BUFFER, 2);
-
-	control |= ib->length_dw | (vm_id << 24);
-
-	amdgpu_ring_write(ring, header);
+	amdgpu_ring_write(ring, PACKET3(PACKET3_INDIRECT_BUFFER, 2));
 	amdgpu_ring_write(ring,
 #ifdef __BIG_ENDIAN
 					  (2 << 0) |
@@ -2098,6 +2254,25 @@ static void gfx_v7_0_ring_emit_ib_compute(struct amdgpu_ring *ring,
 	amdgpu_ring_write(ring, control);
 }
 
+static void gfx_v7_ring_emit_cntxcntl(struct amdgpu_ring *ring, uint32_t flags)
+{
+	uint32_t dw2 = 0;
+
+	dw2 |= 0x80000000; /* set load_enable otherwise this package is just NOPs */
+	if (flags & AMDGPU_HAVE_CTX_SWITCH) {
+		/* set load_global_config & load_global_uconfig */
+		dw2 |= 0x8001;
+		/* set load_cs_sh_regs */
+		dw2 |= 0x01000000;
+		/* set load_per_context_state & load_gfx_sh_regs */
+		dw2 |= 0x10002;
+	}
+
+	amdgpu_ring_write(ring, PACKET3(PACKET3_CONTEXT_CONTROL, 1));
+	amdgpu_ring_write(ring, dw2);
+	amdgpu_ring_write(ring, 0);
+}
+
 /**
  * gfx_v7_0_ring_test_ib - basic ring IB test
  *
@@ -2107,26 +2282,25 @@ static void gfx_v7_0_ring_emit_ib_compute(struct amdgpu_ring *ring,
  * Provides a basic gfx ring test to verify that IBs are working.
  * Returns 0 on success, error on failure.
  */
-static int gfx_v7_0_ring_test_ib(struct amdgpu_ring *ring)
+static int gfx_v7_0_ring_test_ib(struct amdgpu_ring *ring, long timeout)
 {
 	struct amdgpu_device *adev = ring->adev;
 	struct amdgpu_ib ib;
 	struct fence *f = NULL;
 	uint32_t scratch;
 	uint32_t tmp = 0;
-	unsigned i;
-	int r;
+	long r;
 
 	r = amdgpu_gfx_scratch_get(adev, &scratch);
 	if (r) {
-		DRM_ERROR("amdgpu: failed to get scratch reg (%d).\n", r);
+		DRM_ERROR("amdgpu: failed to get scratch reg (%ld).\n", r);
 		return r;
 	}
 	WREG32(scratch, 0xCAFEDEAD);
 	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 err1;
 	}
 	ib.ptr[0] = PACKET3(PACKET3_SET_UCONFIG_REG, 1);
@@ -2138,21 +2312,19 @@ static int gfx_v7_0_ring_test_ib(struct amdgpu_ring *ring)
 	if (r)
 		goto err2;
 
-	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 err2;
-	}
-	for (i = 0; i < adev->usec_timeout; i++) {
-		tmp = RREG32(scratch);
-		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 err2;
+	}
+	tmp = RREG32(scratch);
+	if (tmp == 0xDEADBEEF) {
+		DRM_INFO("ib test on ring %d succeeded\n", ring->idx);
+		r = 0;
 	} else {
 		DRM_ERROR("amdgpu: ib test failed (scratch(0x%04X)=0x%08X)\n",
 			  scratch, tmp);
@@ -2160,7 +2332,6 @@ static int gfx_v7_0_ring_test_ib(struct amdgpu_ring *ring)
 	}
 
 err2:
-	fence_put(f);
 	amdgpu_ib_free(adev, &ib, NULL);
 	fence_put(f);
 err1:
@@ -2449,7 +2620,7 @@ static int gfx_v7_0_cp_gfx_resume(struct amdgpu_device *adev)
 	return 0;
 }
 
-static u32 gfx_v7_0_ring_get_rptr_gfx(struct amdgpu_ring *ring)
+static u32 gfx_v7_0_ring_get_rptr(struct amdgpu_ring *ring)
 {
 	return ring->adev->wb.wb[ring->rptr_offs];
 }
@@ -2469,11 +2640,6 @@ static void gfx_v7_0_ring_set_wptr_gfx(struct amdgpu_ring *ring)
 	(void)RREG32(mmCP_RB0_WPTR);
 }
 
-static u32 gfx_v7_0_ring_get_rptr_compute(struct amdgpu_ring *ring)
-{
-	return ring->adev->wb.wb[ring->rptr_offs];
-}
-
 static u32 gfx_v7_0_ring_get_wptr_compute(struct amdgpu_ring *ring)
 {
 	/* XXX check if swapping is necessary on BE */
@@ -2761,8 +2927,7 @@ static int gfx_v7_0_cp_compute_resume(struct amdgpu_device *adev)
 	u64 wb_gpu_addr;
 	u32 *buf;
 	struct bonaire_mqd *mqd;
-
-	gfx_v7_0_cp_compute_enable(adev, true);
+	struct amdgpu_ring *ring;
 
 	/* fix up chicken bits */
 	tmp = RREG32(mmCP_CPF_DEBUG);
@@ -2797,7 +2962,7 @@ static int gfx_v7_0_cp_compute_resume(struct amdgpu_device *adev)
 
 	/* init the queues.  Just two for now. */
 	for (i = 0; i < adev->gfx.num_compute_rings; i++) {
-		struct amdgpu_ring *ring = &adev->gfx.compute_ring[i];
+		ring = &adev->gfx.compute_ring[i];
 
 		if (ring->mqd_obj == NULL) {
 			r = amdgpu_bo_create(adev,
@@ -2976,6 +3141,13 @@ static int gfx_v7_0_cp_compute_resume(struct amdgpu_device *adev)
 		amdgpu_bo_unreserve(ring->mqd_obj);
 
 		ring->ready = true;
+	}
+
+	gfx_v7_0_cp_compute_enable(adev, true);
+
+	for (i = 0; i < adev->gfx.num_compute_rings; i++) {
+		ring = &adev->gfx.compute_ring[i];
+
 		r = amdgpu_ring_test_ring(ring);
 		if (r)
 			ring->ready = false;
@@ -3205,7 +3377,8 @@ static int gfx_v7_0_rlc_init(struct amdgpu_device *adev)
 		}
 	}
 	adev->gfx.rlc.cs_data = ci_cs_data;
-	adev->gfx.rlc.cp_table_size = CP_ME_TABLE_SIZE * 5 * 4;
+	adev->gfx.rlc.cp_table_size = ALIGN(CP_ME_TABLE_SIZE * 5 * 4, 2048); /* CP JT */
+	adev->gfx.rlc.cp_table_size += 64 * 1024; /* GDS */
 
 	src_ptr = adev->gfx.rlc.reg_list;
 	dws = adev->gfx.rlc.reg_list_size;
@@ -3363,7 +3536,7 @@ static void gfx_v7_0_wait_for_rlc_serdes(struct amdgpu_device *adev)
 	mutex_lock(&adev->grbm_idx_mutex);
 	for (i = 0; i < adev->gfx.config.max_shader_engines; i++) {
 		for (j = 0; j < adev->gfx.config.max_sh_per_se; j++) {
-			gfx_v7_0_select_se_sh(adev, i, j);
+			gfx_v7_0_select_se_sh(adev, i, j, 0xffffffff);
 			for (k = 0; k < adev->usec_timeout; k++) {
 				if (RREG32(mmRLC_SERDES_CU_MASTER_BUSY) == 0)
 					break;
@@ -3371,7 +3544,7 @@ static void gfx_v7_0_wait_for_rlc_serdes(struct amdgpu_device *adev)
 			}
 		}
 	}
-	gfx_v7_0_select_se_sh(adev, 0xffffffff, 0xffffffff);
+	gfx_v7_0_select_se_sh(adev, 0xffffffff, 0xffffffff, 0xffffffff);
 	mutex_unlock(&adev->grbm_idx_mutex);
 
 	mask = RLC_SERDES_NONCU_MASTER_BUSY__SE_MASTER_BUSY_MASK |
@@ -3418,7 +3591,7 @@ static u32 gfx_v7_0_halt_rlc(struct amdgpu_device *adev)
 	return orig;
 }
 
-void gfx_v7_0_enter_rlc_safe_mode(struct amdgpu_device *adev)
+static void gfx_v7_0_enter_rlc_safe_mode(struct amdgpu_device *adev)
 {
 	u32 tmp, i, mask;
 
@@ -3440,7 +3613,7 @@ void gfx_v7_0_enter_rlc_safe_mode(struct amdgpu_device *adev)
 	}
 }
 
-void gfx_v7_0_exit_rlc_safe_mode(struct amdgpu_device *adev)
+static void gfx_v7_0_exit_rlc_safe_mode(struct amdgpu_device *adev)
 {
 	u32 tmp;
 
@@ -3455,7 +3628,7 @@ void gfx_v7_0_exit_rlc_safe_mode(struct amdgpu_device *adev)
  *
  * Halt the RLC ME (MicroEngine) (CIK).
  */
-void gfx_v7_0_rlc_stop(struct amdgpu_device *adev)
+static void gfx_v7_0_rlc_stop(struct amdgpu_device *adev)
 {
 	WREG32(mmRLC_CNTL, 0);
 
@@ -3531,7 +3704,7 @@ static int gfx_v7_0_rlc_resume(struct amdgpu_device *adev)
 	WREG32(mmRLC_LB_CNTR_MAX, 0x00008000);
 
 	mutex_lock(&adev->grbm_idx_mutex);
-	gfx_v7_0_select_se_sh(adev, 0xffffffff, 0xffffffff);
+	gfx_v7_0_select_se_sh(adev, 0xffffffff, 0xffffffff, 0xffffffff);
 	WREG32(mmRLC_LB_INIT_CU_MASK, 0xffffffff);
 	WREG32(mmRLC_LB_PARAMS, 0x00600408);
 	WREG32(mmRLC_LB_CNTL, 0x80000004);
@@ -3571,7 +3744,7 @@ static void gfx_v7_0_enable_cgcg(struct amdgpu_device *adev, bool enable)
 		tmp = gfx_v7_0_halt_rlc(adev);
 
 		mutex_lock(&adev->grbm_idx_mutex);
-		gfx_v7_0_select_se_sh(adev, 0xffffffff, 0xffffffff);
+		gfx_v7_0_select_se_sh(adev, 0xffffffff, 0xffffffff, 0xffffffff);
 		WREG32(mmRLC_SERDES_WR_CU_MASTER_MASK, 0xffffffff);
 		WREG32(mmRLC_SERDES_WR_NONCU_MASTER_MASK, 0xffffffff);
 		tmp2 = RLC_SERDES_WR_CTRL__BPM_ADDR_MASK |
@@ -3622,7 +3795,7 @@ static void gfx_v7_0_enable_mgcg(struct amdgpu_device *adev, bool enable)
 		tmp = gfx_v7_0_halt_rlc(adev);
 
 		mutex_lock(&adev->grbm_idx_mutex);
-		gfx_v7_0_select_se_sh(adev, 0xffffffff, 0xffffffff);
+		gfx_v7_0_select_se_sh(adev, 0xffffffff, 0xffffffff, 0xffffffff);
 		WREG32(mmRLC_SERDES_WR_CU_MASTER_MASK, 0xffffffff);
 		WREG32(mmRLC_SERDES_WR_NONCU_MASTER_MASK, 0xffffffff);
 		data = RLC_SERDES_WR_CTRL__BPM_ADDR_MASK |
@@ -3673,7 +3846,7 @@ static void gfx_v7_0_enable_mgcg(struct amdgpu_device *adev, bool enable)
 		tmp = gfx_v7_0_halt_rlc(adev);
 
 		mutex_lock(&adev->grbm_idx_mutex);
-		gfx_v7_0_select_se_sh(adev, 0xffffffff, 0xffffffff);
+		gfx_v7_0_select_se_sh(adev, 0xffffffff, 0xffffffff, 0xffffffff);
 		WREG32(mmRLC_SERDES_WR_CU_MASTER_MASK, 0xffffffff);
 		WREG32(mmRLC_SERDES_WR_NONCU_MASTER_MASK, 0xffffffff);
 		data = RLC_SERDES_WR_CTRL__BPM_ADDR_MASK | RLC_SERDES_WR_CTRL__MGCG_OVERRIDE_1_MASK;
@@ -3851,6 +4024,20 @@ static void gfx_v7_0_enable_gfx_cgpg(struct amdgpu_device *adev,
 	}
 }
 
+static void gfx_v7_0_set_user_cu_inactive_bitmap(struct amdgpu_device *adev,
+						 u32 bitmap)
+{
+	u32 data;
+
+	if (!bitmap)
+		return;
+
+	data = bitmap << GC_USER_SHADER_ARRAY_CONFIG__INACTIVE_CUS__SHIFT;
+	data &= GC_USER_SHADER_ARRAY_CONFIG__INACTIVE_CUS_MASK;
+
+	WREG32(mmGC_USER_SHADER_ARRAY_CONFIG, data);
+}
+
 static u32 gfx_v7_0_get_cu_active_bitmap(struct amdgpu_device *adev)
 {
 	u32 data, mask;
@@ -4107,7 +4294,7 @@ static void gfx_v7_0_fini_pg(struct amdgpu_device *adev)
  * Fetches a GPU clock counter snapshot (SI).
  * Returns the 64 bit clock counter snapshot.
  */
-uint64_t gfx_v7_0_get_gpu_clock_counter(struct amdgpu_device *adev)
+static uint64_t gfx_v7_0_get_gpu_clock_counter(struct amdgpu_device *adev)
 {
 	uint64_t clock;
 
@@ -4167,12 +4354,59 @@ static void gfx_v7_0_ring_emit_gds_switch(struct amdgpu_ring *ring,
 	amdgpu_ring_write(ring, (1 << (oa_size + oa_base)) - (1 << oa_base));
 }
 
+static unsigned gfx_v7_0_ring_get_emit_ib_size_gfx(struct amdgpu_ring *ring)
+{
+	return
+		4; /* gfx_v7_0_ring_emit_ib_gfx */
+}
+
+static unsigned gfx_v7_0_ring_get_dma_frame_size_gfx(struct amdgpu_ring *ring)
+{
+	return
+		20 + /* gfx_v7_0_ring_emit_gds_switch */
+		7 + /* gfx_v7_0_ring_emit_hdp_flush */
+		5 + /* gfx_v7_0_ring_emit_hdp_invalidate */
+		12 + 12 + 12 + /* gfx_v7_0_ring_emit_fence_gfx x3 for user fence, vm fence */
+		7 + 4 + /* gfx_v7_0_ring_emit_pipeline_sync */
+		17 + 6 + /* gfx_v7_0_ring_emit_vm_flush */
+		3; /* gfx_v7_ring_emit_cntxcntl */
+}
+
+static unsigned gfx_v7_0_ring_get_emit_ib_size_compute(struct amdgpu_ring *ring)
+{
+	return
+		4; /* gfx_v7_0_ring_emit_ib_compute */
+}
+
+static unsigned gfx_v7_0_ring_get_dma_frame_size_compute(struct amdgpu_ring *ring)
+{
+	return
+		20 + /* gfx_v7_0_ring_emit_gds_switch */
+		7 + /* gfx_v7_0_ring_emit_hdp_flush */
+		5 + /* gfx_v7_0_ring_emit_hdp_invalidate */
+		7 + /* gfx_v7_0_ring_emit_pipeline_sync */
+		17 + /* gfx_v7_0_ring_emit_vm_flush */
+		7 + 7 + 7; /* gfx_v7_0_ring_emit_fence_compute x3 for user fence, vm fence */
+}
+
+static const struct amdgpu_gfx_funcs gfx_v7_0_gfx_funcs = {
+	.get_gpu_clock_counter = &gfx_v7_0_get_gpu_clock_counter,
+	.select_se_sh = &gfx_v7_0_select_se_sh,
+};
+
+static const struct amdgpu_rlc_funcs gfx_v7_0_rlc_funcs = {
+	.enter_safe_mode = gfx_v7_0_enter_rlc_safe_mode,
+	.exit_safe_mode = gfx_v7_0_exit_rlc_safe_mode
+};
+
 static int gfx_v7_0_early_init(void *handle)
 {
 	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
 
 	adev->gfx.num_gfx_rings = GFX7_NUM_GFX_RINGS;
 	adev->gfx.num_compute_rings = GFX7_NUM_COMPUTE_RINGS;
+	adev->gfx.funcs = &gfx_v7_0_gfx_funcs;
+	adev->gfx.rlc.funcs = &gfx_v7_0_rlc_funcs;
 	gfx_v7_0_set_ring_funcs(adev);
 	gfx_v7_0_set_irq_funcs(adev);
 	gfx_v7_0_set_gds_init(adev);
@@ -4444,24 +4678,21 @@ static int gfx_v7_0_sw_init(void *handle)
 	}
 
 	/* reserve GDS, GWS and OA resource for gfx */
-	r = amdgpu_bo_create(adev, adev->gds.mem.gfx_partition_size,
-			PAGE_SIZE, true,
-			AMDGPU_GEM_DOMAIN_GDS, 0,
-			NULL, NULL, &adev->gds.gds_gfx_bo);
+	r = amdgpu_bo_create_kernel(adev, adev->gds.mem.gfx_partition_size,
+				    PAGE_SIZE, AMDGPU_GEM_DOMAIN_GDS,
+				    &adev->gds.gds_gfx_bo, NULL, NULL);
 	if (r)
 		return r;
 
-	r = amdgpu_bo_create(adev, adev->gds.gws.gfx_partition_size,
-		PAGE_SIZE, true,
-		AMDGPU_GEM_DOMAIN_GWS, 0,
-		NULL, NULL, &adev->gds.gws_gfx_bo);
+	r = amdgpu_bo_create_kernel(adev, adev->gds.gws.gfx_partition_size,
+				    PAGE_SIZE, AMDGPU_GEM_DOMAIN_GWS,
+				    &adev->gds.gws_gfx_bo, NULL, NULL);
 	if (r)
 		return r;
 
-	r = amdgpu_bo_create(adev, adev->gds.oa.gfx_partition_size,
-			PAGE_SIZE, true,
-			AMDGPU_GEM_DOMAIN_OA, 0,
-			NULL, NULL, &adev->gds.oa_gfx_bo);
+	r = amdgpu_bo_create_kernel(adev, adev->gds.oa.gfx_partition_size,
+				    PAGE_SIZE, AMDGPU_GEM_DOMAIN_OA,
+				    &adev->gds.oa_gfx_bo, NULL, NULL);
 	if (r)
 		return r;
 
@@ -4477,9 +4708,9 @@ static int gfx_v7_0_sw_fini(void *handle)
 	int i;
 	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
 
-	amdgpu_bo_unref(&adev->gds.oa_gfx_bo);
-	amdgpu_bo_unref(&adev->gds.gws_gfx_bo);
-	amdgpu_bo_unref(&adev->gds.gds_gfx_bo);
+	amdgpu_bo_free_kernel(&adev->gds.oa_gfx_bo, NULL, NULL);
+	amdgpu_bo_free_kernel(&adev->gds.gws_gfx_bo, NULL, NULL);
+	amdgpu_bo_free_kernel(&adev->gds.gds_gfx_bo, NULL, NULL);
 
 	for (i = 0; i < adev->gfx.num_gfx_rings; i++)
 		amdgpu_ring_fini(&adev->gfx.gfx_ring[i]);
@@ -4489,6 +4720,7 @@ static int gfx_v7_0_sw_fini(void *handle)
 	gfx_v7_0_cp_compute_fini(adev);
 	gfx_v7_0_rlc_fini(adev);
 	gfx_v7_0_mec_fini(adev);
+	gfx_v7_0_free_microcode(adev);
 
 	return 0;
 }
@@ -4816,7 +5048,7 @@ static int gfx_v7_0_eop_irq(struct amdgpu_device *adev,
 	case 2:
 		for (i = 0; i < adev->gfx.num_compute_rings; i++) {
 			ring = &adev->gfx.compute_ring[i];
-			if ((ring->me == me_id) & (ring->pipe == pipe_id))
+			if ((ring->me == me_id) && (ring->pipe == pipe_id))
 				amdgpu_fence_process(ring);
 		}
 		break;
@@ -4909,7 +5141,7 @@ const struct amd_ip_funcs gfx_v7_0_ip_funcs = {
 };
 
 static const struct amdgpu_ring_funcs gfx_v7_0_ring_funcs_gfx = {
-	.get_rptr = gfx_v7_0_ring_get_rptr_gfx,
+	.get_rptr = gfx_v7_0_ring_get_rptr,
 	.get_wptr = gfx_v7_0_ring_get_wptr_gfx,
 	.set_wptr = gfx_v7_0_ring_set_wptr_gfx,
 	.parse_cs = NULL,
@@ -4924,10 +5156,13 @@ static const struct amdgpu_ring_funcs gfx_v7_0_ring_funcs_gfx = {
 	.test_ib = gfx_v7_0_ring_test_ib,
 	.insert_nop = amdgpu_ring_insert_nop,
 	.pad_ib = amdgpu_ring_generic_pad_ib,
+	.emit_cntxcntl = gfx_v7_ring_emit_cntxcntl,
+	.get_emit_ib_size = gfx_v7_0_ring_get_emit_ib_size_gfx,
+	.get_dma_frame_size = gfx_v7_0_ring_get_dma_frame_size_gfx,
 };
 
 static const struct amdgpu_ring_funcs gfx_v7_0_ring_funcs_compute = {
-	.get_rptr = gfx_v7_0_ring_get_rptr_compute,
+	.get_rptr = gfx_v7_0_ring_get_rptr,
 	.get_wptr = gfx_v7_0_ring_get_wptr_compute,
 	.set_wptr = gfx_v7_0_ring_set_wptr_compute,
 	.parse_cs = NULL,
@@ -4942,6 +5177,8 @@ static const struct amdgpu_ring_funcs gfx_v7_0_ring_funcs_compute = {
 	.test_ib = gfx_v7_0_ring_test_ib,
 	.insert_nop = amdgpu_ring_insert_nop,
 	.pad_ib = amdgpu_ring_generic_pad_ib,
+	.get_emit_ib_size = gfx_v7_0_ring_get_emit_ib_size_compute,
+	.get_dma_frame_size = gfx_v7_0_ring_get_dma_frame_size_compute,
 };
 
 static void gfx_v7_0_set_ring_funcs(struct amdgpu_device *adev)
@@ -5015,16 +5252,22 @@ static void gfx_v7_0_get_cu_info(struct amdgpu_device *adev)
 	int i, j, k, counter, active_cu_number = 0;
 	u32 mask, bitmap, ao_bitmap, ao_cu_mask = 0;
 	struct amdgpu_cu_info *cu_info = &adev->gfx.cu_info;
+	unsigned disable_masks[4 * 2];
 
 	memset(cu_info, 0, sizeof(*cu_info));
 
+	amdgpu_gfx_parse_disable_cu(disable_masks, 4, 2);
+
 	mutex_lock(&adev->grbm_idx_mutex);
 	for (i = 0; i < adev->gfx.config.max_shader_engines; i++) {
 		for (j = 0; j < adev->gfx.config.max_sh_per_se; j++) {
 			mask = 1;
 			ao_bitmap = 0;
 			counter = 0;
-			gfx_v7_0_select_se_sh(adev, i, j);
+			gfx_v7_0_select_se_sh(adev, i, j, 0xffffffff);
+			if (i < 4 && j < 2)
+				gfx_v7_0_set_user_cu_inactive_bitmap(
+					adev, disable_masks[i * 2 + j]);
 			bitmap = gfx_v7_0_get_cu_active_bitmap(adev);
 			cu_info->bitmap[i][j] = bitmap;
 
@@ -5040,7 +5283,7 @@ static void gfx_v7_0_get_cu_info(struct amdgpu_device *adev)
 			ao_cu_mask |= (ao_bitmap << (i * 16 + j * 8));
 		}
 	}
-	gfx_v7_0_select_se_sh(adev, 0xffffffff, 0xffffffff);
+	gfx_v7_0_select_se_sh(adev, 0xffffffff, 0xffffffff, 0xffffffff);
 	mutex_unlock(&adev->grbm_idx_mutex);
 
 	cu_info->number = active_cu_number;