Merge tag 'drm-fixes-5.5-2019-12-12' of git://people.freedesktop.org/~agd5f/linux into drm-fixes

drm-fixes-5.5-2019-12-12:

amdgpu:
- DC fixes for renoir
- Gfx8 fence flush align with mesa
- Power profile fix for arcturus
- Freesync fix
- DC I2c over aux fix
- DC aux defer fix
- GPU reset fix
- GPUVM invalidation semaphore fixes for PCO and SR-IOV
- Golden settings updates for gfx10
Signed-off-by: Dave Airlie <airlied@redhat.com>
From: Alex Deucher <alexdeucher@gmail.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20191212223211.8034-1-alexander.deucher@amd.com

drivers/gpu/drm/amd/acp/Kconfig

-# SPDX-License-Identifier: GPL-2.0-only
+# SPDX-License-Identifier: MIT
 menu "ACP (Audio CoProcessor) Configuration"
 
 config DRM_AMD_ACP

drivers/gpu/drm/amd/amdgpu/Kconfig

-# SPDX-License-Identifier: GPL-2.0-only
+# SPDX-License-Identifier: MIT
 config DRM_AMDGPU_SI
 	bool "Enable amdgpu support for SI parts"
 	depends on DRM_AMDGPU

drivers/gpu/drm/amd/amdgpu/amdgpu_ctx.c

@@ -604,11 +604,8 @@ void amdgpu_ctx_mgr_entity_fini(struct amdgpu_ctx_mgr *mgr)
 			continue;
 		}
 
-		for (i = 0; i < num_entities; i++) {
-			mutex_lock(&ctx->adev->lock_reset);
+		for (i = 0; i < num_entities; i++)
 			drm_sched_entity_fini(&ctx->entities[0][i].entity);
-			mutex_unlock(&ctx->adev->lock_reset);
-		}
 	}
 }
 

drivers/gpu/drm/amd/amdgpu/df_v3_6.c

@@ -268,23 +268,29 @@ static void df_v3_6_update_medium_grain_clock_gating(struct amdgpu_device *adev,
 {
 	u32 tmp;
 
-	/* Put DF on broadcast mode */
-	adev->df_funcs->enable_broadcast_mode(adev, true);
-
-	if (enable && (adev->cg_flags & AMD_CG_SUPPORT_DF_MGCG)) {
-		tmp = RREG32_SOC15(DF, 0, mmDF_PIE_AON0_DfGlobalClkGater);
-		tmp &= ~DF_PIE_AON0_DfGlobalClkGater__MGCGMode_MASK;
-		tmp |= DF_V3_6_MGCG_ENABLE_15_CYCLE_DELAY;
-		WREG32_SOC15(DF, 0, mmDF_PIE_AON0_DfGlobalClkGater, tmp);
-	} else {
-		tmp = RREG32_SOC15(DF, 0, mmDF_PIE_AON0_DfGlobalClkGater);
-		tmp &= ~DF_PIE_AON0_DfGlobalClkGater__MGCGMode_MASK;
-		tmp |= DF_V3_6_MGCG_DISABLE;
-		WREG32_SOC15(DF, 0, mmDF_PIE_AON0_DfGlobalClkGater, tmp);
-	}
+	if (adev->cg_flags & AMD_CG_SUPPORT_DF_MGCG) {
+		/* Put DF on broadcast mode */
+		adev->df_funcs->enable_broadcast_mode(adev, true);
+
+		if (enable) {
+			tmp = RREG32_SOC15(DF, 0,
+					mmDF_PIE_AON0_DfGlobalClkGater);
+			tmp &= ~DF_PIE_AON0_DfGlobalClkGater__MGCGMode_MASK;
+			tmp |= DF_V3_6_MGCG_ENABLE_15_CYCLE_DELAY;
+			WREG32_SOC15(DF, 0,
+					mmDF_PIE_AON0_DfGlobalClkGater, tmp);
+		} else {
+			tmp = RREG32_SOC15(DF, 0,
+					mmDF_PIE_AON0_DfGlobalClkGater);
+			tmp &= ~DF_PIE_AON0_DfGlobalClkGater__MGCGMode_MASK;
+			tmp |= DF_V3_6_MGCG_DISABLE;
+			WREG32_SOC15(DF, 0,
+					mmDF_PIE_AON0_DfGlobalClkGater, tmp);
+		}
 
-	/* Exit broadcast mode */
-	adev->df_funcs->enable_broadcast_mode(adev, false);
+		/* Exit broadcast mode */
+		adev->df_funcs->enable_broadcast_mode(adev, false);
+	}
 }
 
 static void df_v3_6_get_clockgating_state(struct amdgpu_device *adev,

drivers/gpu/drm/amd/amdgpu/gfx_v10_0.c

@@ -117,10 +117,13 @@ static const struct soc15_reg_golden golden_settings_gc_10_1[] =
 	SOC15_REG_GOLDEN_VALUE(GC, 0, mmGL2C_CGTT_SCLK_CTRL, 0x10000000, 0x10000100),
 	SOC15_REG_GOLDEN_VALUE(GC, 0, mmGL2C_CTRL2, 0xffffffff, 0x1402002f),
 	SOC15_REG_GOLDEN_VALUE(GC, 0, mmGL2C_CTRL3, 0xffff9fff, 0x00001188),
+	SOC15_REG_GOLDEN_VALUE(GC, 0, mmPA_SC_BINNER_TIMEOUT_COUNTER, 0xffffffff, 0x00000800),
 	SOC15_REG_GOLDEN_VALUE(GC, 0, mmPA_SC_ENHANCE, 0x3fffffff, 0x08000009),
 	SOC15_REG_GOLDEN_VALUE(GC, 0, mmPA_SC_ENHANCE_1, 0x00400000, 0x04440000),
+	SOC15_REG_GOLDEN_VALUE(GC, 0, mmPA_SC_ENHANCE_2, 0x00000800, 0x00000820),
 	SOC15_REG_GOLDEN_VALUE(GC, 0, mmPA_SC_LINE_STIPPLE_STATE, 0x0000ff0f, 0x00000000),
 	SOC15_REG_GOLDEN_VALUE(GC, 0, mmRMI_SPARE, 0xffffffff, 0xffff3101),
+	SOC15_REG_GOLDEN_VALUE(GC, 0, mmSPI_CONFIG_CNTL, 0x001f0000, 0x00070104),
 	SOC15_REG_GOLDEN_VALUE(GC, 0, mmSQ_ALU_CLK_CTRL, 0xffffffff, 0xffffffff),
 	SOC15_REG_GOLDEN_VALUE(GC, 0, mmSQ_ARB_CONFIG, 0x00000100, 0x00000130),
 	SOC15_REG_GOLDEN_VALUE(GC, 0, mmSQ_LDS_CLK_CTRL, 0xffffffff, 0xffffffff),
@@ -162,10 +165,13 @@ static const struct soc15_reg_golden golden_settings_gc_10_1_1[] =
 	SOC15_REG_GOLDEN_VALUE(GC, 0, mmGL2C_CGTT_SCLK_CTRL, 0xffff0fff, 0x10000100),
 	SOC15_REG_GOLDEN_VALUE(GC, 0, mmGL2C_CTRL2, 0xffffffff, 0x1402002f),
 	SOC15_REG_GOLDEN_VALUE(GC, 0, mmGL2C_CTRL3, 0xffffbfff, 0x00000188),
+	SOC15_REG_GOLDEN_VALUE(GC, 0, mmPA_SC_BINNER_TIMEOUT_COUNTER, 0xffffffff, 0x00000800),
 	SOC15_REG_GOLDEN_VALUE(GC, 0, mmPA_SC_ENHANCE, 0x3fffffff, 0x08000009),
 	SOC15_REG_GOLDEN_VALUE(GC, 0, mmPA_SC_ENHANCE_1, 0x00400000, 0x04440000),
+	SOC15_REG_GOLDEN_VALUE(GC, 0, mmPA_SC_ENHANCE_2, 0x00000800, 0x00000820),
 	SOC15_REG_GOLDEN_VALUE(GC, 0, mmPA_SC_LINE_STIPPLE_STATE, 0x0000ff0f, 0x00000000),
 	SOC15_REG_GOLDEN_VALUE(GC, 0, mmRMI_SPARE, 0xffffffff, 0xffff3101),
+	SOC15_REG_GOLDEN_VALUE(GC, 0, mmSPI_CONFIG_CNTL, 0x001f0000, 0x00070105),
 	SOC15_REG_GOLDEN_VALUE(GC, 0, mmSQ_ALU_CLK_CTRL, 0xffffffff, 0xffffffff),
 	SOC15_REG_GOLDEN_VALUE(GC, 0, mmSQ_ARB_CONFIG, 0x00000133, 0x00000130),
 	SOC15_REG_GOLDEN_VALUE(GC, 0, mmSQ_LDS_CLK_CTRL, 0xffffffff, 0xffffffff),

drivers/gpu/drm/amd/amdgpu/gfx_v8_0.c

@@ -6146,7 +6146,23 @@ static void gfx_v8_0_ring_emit_fence_gfx(struct amdgpu_ring *ring, u64 addr,
 	bool write64bit = flags & AMDGPU_FENCE_FLAG_64BIT;
 	bool int_sel = flags & AMDGPU_FENCE_FLAG_INT;
 
-	/* EVENT_WRITE_EOP - flush caches, send int */
+	/* Workaround for cache flush problems. First send a dummy EOP
+	 * event down the pipe with seq one below.
+	 */
+	amdgpu_ring_write(ring, PACKET3(PACKET3_EVENT_WRITE_EOP, 4));
+	amdgpu_ring_write(ring, (EOP_TCL1_ACTION_EN |
+				 EOP_TC_ACTION_EN |
+				 EOP_TC_WB_ACTION_EN |
+				 EVENT_TYPE(CACHE_FLUSH_AND_INV_TS_EVENT) |
+				 EVENT_INDEX(5)));
+	amdgpu_ring_write(ring, addr & 0xfffffffc);
+	amdgpu_ring_write(ring, (upper_32_bits(addr) & 0xffff) |
+				DATA_SEL(1) | INT_SEL(0));
+	amdgpu_ring_write(ring, lower_32_bits(seq - 1));
+	amdgpu_ring_write(ring, upper_32_bits(seq - 1));
+
+	/* Then send the real EOP event down the pipe:
+	 * EVENT_WRITE_EOP - flush caches, send int */
 	amdgpu_ring_write(ring, PACKET3(PACKET3_EVENT_WRITE_EOP, 4));
 	amdgpu_ring_write(ring, (EOP_TCL1_ACTION_EN |
 				 EOP_TC_ACTION_EN |
@@ -6888,7 +6904,7 @@ static const struct amdgpu_ring_funcs gfx_v8_0_ring_funcs_gfx = {
 		5 +  /* COND_EXEC */
 		7 +  /* PIPELINE_SYNC */
 		VI_FLUSH_GPU_TLB_NUM_WREG * 5 + 9 + /* VM_FLUSH */
-		8 +  /* FENCE for VM_FLUSH */
+		12 +  /* FENCE for VM_FLUSH */
 		20 + /* GDS switch */
 		4 + /* double SWITCH_BUFFER,
 		       the first COND_EXEC jump to the place just
@@ -6900,7 +6916,7 @@ static const struct amdgpu_ring_funcs gfx_v8_0_ring_funcs_gfx = {
 		31 + /*	DE_META */
 		3 + /* CNTX_CTRL */
 		5 + /* HDP_INVL */
-		8 + 8 + /* FENCE x2 */
+		12 + 12 + /* FENCE x2 */
 		2, /* SWITCH_BUFFER */
 	.emit_ib_size =	4, /* gfx_v8_0_ring_emit_ib_gfx */
 	.emit_ib = gfx_v8_0_ring_emit_ib_gfx,

drivers/gpu/drm/amd/amdgpu/gmc_v10_0.c

@@ -219,6 +219,21 @@ static uint32_t gmc_v10_0_get_invalidate_req(unsigned int vmid,
 	return req;
 }
 
+/**
+ * gmc_v10_0_use_invalidate_semaphore - judge whether to use semaphore
+ *
+ * @adev: amdgpu_device pointer
+ * @vmhub: vmhub type
+ *
+ */
+static bool gmc_v10_0_use_invalidate_semaphore(struct amdgpu_device *adev,
+				       uint32_t vmhub)
+{
+	return ((vmhub == AMDGPU_MMHUB_0 ||
+		 vmhub == AMDGPU_MMHUB_1) &&
+		(!amdgpu_sriov_vf(adev)));
+}
+
 /*
  * GART
  * VMID 0 is the physical GPU addresses as used by the kernel.
@@ -229,6 +244,7 @@ static uint32_t gmc_v10_0_get_invalidate_req(unsigned int vmid,
 static void gmc_v10_0_flush_vm_hub(struct amdgpu_device *adev, uint32_t vmid,
 				   unsigned int vmhub, uint32_t flush_type)
 {
+	bool use_semaphore = gmc_v10_0_use_invalidate_semaphore(adev, vmhub);
 	struct amdgpu_vmhub *hub = &adev->vmhub[vmhub];
 	u32 tmp = gmc_v10_0_get_invalidate_req(vmid, flush_type);
 	/* Use register 17 for GART */
@@ -244,8 +260,7 @@ static void gmc_v10_0_flush_vm_hub(struct amdgpu_device *adev, uint32_t vmid,
 	 */
 
 	/* TODO: It needs to continue working on debugging with semaphore for GFXHUB as well. */
-	if (vmhub == AMDGPU_MMHUB_0 ||
-	    vmhub == AMDGPU_MMHUB_1) {
+	if (use_semaphore) {
 		for (i = 0; i < adev->usec_timeout; i++) {
 			/* a read return value of 1 means semaphore acuqire */
 			tmp = RREG32_NO_KIQ(hub->vm_inv_eng0_sem + eng);
@@ -278,8 +293,7 @@ static void gmc_v10_0_flush_vm_hub(struct amdgpu_device *adev, uint32_t vmid,
 	}
 
 	/* TODO: It needs to continue working on debugging with semaphore for GFXHUB as well. */
-	if (vmhub == AMDGPU_MMHUB_0 ||
-	    vmhub == AMDGPU_MMHUB_1)
+	if (use_semaphore)
 		/*
 		 * add semaphore release after invalidation,
 		 * write with 0 means semaphore release
@@ -369,6 +383,7 @@ static void gmc_v10_0_flush_gpu_tlb(struct amdgpu_device *adev, uint32_t vmid,
 static uint64_t gmc_v10_0_emit_flush_gpu_tlb(struct amdgpu_ring *ring,
 					     unsigned vmid, uint64_t pd_addr)
 {
+	bool use_semaphore = gmc_v10_0_use_invalidate_semaphore(ring->adev, ring->funcs->vmhub);
 	struct amdgpu_vmhub *hub = &ring->adev->vmhub[ring->funcs->vmhub];
 	uint32_t req = gmc_v10_0_get_invalidate_req(vmid, 0);
 	unsigned eng = ring->vm_inv_eng;
@@ -381,8 +396,7 @@ static uint64_t gmc_v10_0_emit_flush_gpu_tlb(struct amdgpu_ring *ring,
 	 */
 
 	/* TODO: It needs to continue working on debugging with semaphore for GFXHUB as well. */
-	if (ring->funcs->vmhub == AMDGPU_MMHUB_0 ||
-	    ring->funcs->vmhub == AMDGPU_MMHUB_1)
+	if (use_semaphore)
 		/* a read return value of 1 means semaphore acuqire */
 		amdgpu_ring_emit_reg_wait(ring,
 					  hub->vm_inv_eng0_sem + eng, 0x1, 0x1);
@@ -398,8 +412,7 @@ static uint64_t gmc_v10_0_emit_flush_gpu_tlb(struct amdgpu_ring *ring,
 					    req, 1 << vmid);
 
 	/* TODO: It needs to continue working on debugging with semaphore for GFXHUB as well. */
-	if (ring->funcs->vmhub == AMDGPU_MMHUB_0 ||
-	    ring->funcs->vmhub == AMDGPU_MMHUB_1)
+	if (use_semaphore)
 		/*
 		 * add semaphore release after invalidation,
 		 * write with 0 means semaphore release

drivers/gpu/drm/amd/amdgpu/gmc_v9_0.c

@@ -416,6 +416,24 @@ static uint32_t gmc_v9_0_get_invalidate_req(unsigned int vmid,
 	return req;
 }
 
+/**
+ * gmc_v9_0_use_invalidate_semaphore - judge whether to use semaphore
+ *
+ * @adev: amdgpu_device pointer
+ * @vmhub: vmhub type
+ *
+ */
+static bool gmc_v9_0_use_invalidate_semaphore(struct amdgpu_device *adev,
+				       uint32_t vmhub)
+{
+	return ((vmhub == AMDGPU_MMHUB_0 ||
+		 vmhub == AMDGPU_MMHUB_1) &&
+		(!amdgpu_sriov_vf(adev)) &&
+		(!(adev->asic_type == CHIP_RAVEN &&
+		   adev->rev_id < 0x8 &&
+		   adev->pdev->device == 0x15d8)));
+}
+
 /*
  * GART
  * VMID 0 is the physical GPU addresses as used by the kernel.
@@ -435,6 +453,7 @@ static uint32_t gmc_v9_0_get_invalidate_req(unsigned int vmid,
 static void gmc_v9_0_flush_gpu_tlb(struct amdgpu_device *adev, uint32_t vmid,
 					uint32_t vmhub, uint32_t flush_type)
 {
+	bool use_semaphore = gmc_v9_0_use_invalidate_semaphore(adev, vmhub);
 	const unsigned eng = 17;
 	u32 j, tmp;
 	struct amdgpu_vmhub *hub;
@@ -468,8 +487,7 @@ static void gmc_v9_0_flush_gpu_tlb(struct amdgpu_device *adev, uint32_t vmid,
 	 */
 
 	/* TODO: It needs to continue working on debugging with semaphore for GFXHUB as well. */
-	if (vmhub == AMDGPU_MMHUB_0 ||
-	    vmhub == AMDGPU_MMHUB_1) {
+	if (use_semaphore) {
 		for (j = 0; j < adev->usec_timeout; j++) {
 			/* a read return value of 1 means semaphore acuqire */
 			tmp = RREG32_NO_KIQ(hub->vm_inv_eng0_sem + eng);
@@ -499,8 +517,7 @@ static void gmc_v9_0_flush_gpu_tlb(struct amdgpu_device *adev, uint32_t vmid,
 	}
 
 	/* TODO: It needs to continue working on debugging with semaphore for GFXHUB as well. */
-	if (vmhub == AMDGPU_MMHUB_0 ||
-	    vmhub == AMDGPU_MMHUB_1)
+	if (use_semaphore)
 		/*
 		 * add semaphore release after invalidation,
 		 * write with 0 means semaphore release
@@ -518,6 +535,7 @@ static void gmc_v9_0_flush_gpu_tlb(struct amdgpu_device *adev, uint32_t vmid,
 static uint64_t gmc_v9_0_emit_flush_gpu_tlb(struct amdgpu_ring *ring,
 					    unsigned vmid, uint64_t pd_addr)
 {
+	bool use_semaphore = gmc_v9_0_use_invalidate_semaphore(ring->adev, ring->funcs->vmhub);
 	struct amdgpu_device *adev = ring->adev;
 	struct amdgpu_vmhub *hub = &adev->vmhub[ring->funcs->vmhub];
 	uint32_t req = gmc_v9_0_get_invalidate_req(vmid, 0);
@@ -531,8 +549,7 @@ static uint64_t gmc_v9_0_emit_flush_gpu_tlb(struct amdgpu_ring *ring,
 	 */
 
 	/* TODO: It needs to continue working on debugging with semaphore for GFXHUB as well. */
-	if (ring->funcs->vmhub == AMDGPU_MMHUB_0 ||
-	    ring->funcs->vmhub == AMDGPU_MMHUB_1)
+	if (use_semaphore)
 		/* a read return value of 1 means semaphore acuqire */
 		amdgpu_ring_emit_reg_wait(ring,
 					  hub->vm_inv_eng0_sem + eng, 0x1, 0x1);
@@ -548,8 +565,7 @@ static uint64_t gmc_v9_0_emit_flush_gpu_tlb(struct amdgpu_ring *ring,
 					    req, 1 << vmid);
 
 	/* TODO: It needs to continue working on debugging with semaphore for GFXHUB as well. */
-	if (ring->funcs->vmhub == AMDGPU_MMHUB_0 ||
-	    ring->funcs->vmhub == AMDGPU_MMHUB_1)
+	if (use_semaphore)
 		/*
 		 * add semaphore release after invalidation,
 		 * write with 0 means semaphore release

drivers/gpu/drm/amd/amdkfd/Kconfig

-# SPDX-License-Identifier: GPL-2.0-only
+# SPDX-License-Identifier: MIT
 #
 # Heterogenous system architecture configuration
 #

drivers/gpu/drm/amd/display/Kconfig

-# SPDX-License-Identifier: GPL-2.0-only
+# SPDX-License-Identifier: MIT
 menu "Display Engine Configuration"
 	depends on DRM && DRM_AMDGPU
 

drivers/gpu/drm/amd/display/dc/bios/bios_parser2.c

@@ -1625,6 +1625,7 @@ static enum bp_result construct_integrated_info(
 		/* Don't need to check major revision as they are all 1 */
 		switch (revision.minor) {
 		case 11:
+		case 12:
 			result = get_integrated_info_v11(bp, info);
 			break;
 		default:

drivers/gpu/drm/amd/display/dc/clk_mgr/dcn21/rn_clk_mgr.c

@@ -471,12 +471,28 @@ static void rn_notify_wm_ranges(struct clk_mgr *clk_mgr_base)
 
 }
 
+static bool rn_are_clock_states_equal(struct dc_clocks *a,
+		struct dc_clocks *b)
+{
+	if (a->dispclk_khz != b->dispclk_khz)
+		return false;
+	else if (a->dppclk_khz != b->dppclk_khz)
+		return false;
+	else if (a->dcfclk_khz != b->dcfclk_khz)
+		return false;
+	else if (a->dcfclk_deep_sleep_khz != b->dcfclk_deep_sleep_khz)
+		return false;
+
+	return true;
+}
+
+
 static struct clk_mgr_funcs dcn21_funcs = {
 	.get_dp_ref_clk_frequency = dce12_get_dp_ref_freq_khz,
 	.update_clocks = rn_update_clocks,
 	.init_clocks = rn_init_clocks,
 	.enable_pme_wa = rn_enable_pme_wa,
-	/* .dump_clk_registers = rn_dump_clk_registers, */
+	.are_clock_states_equal = rn_are_clock_states_equal,
 	.notify_wm_ranges = rn_notify_wm_ranges
 };
 
@@ -518,36 +534,83 @@ struct clk_bw_params rn_bw_params = {
 		.num_entries = 4,
 	},
 
-	.wm_table = {
-		.entries = {
-			{
-				.wm_inst = WM_A,
-				.wm_type = WM_TYPE_PSTATE_CHG,
-				.pstate_latency_us = 23.84,
-				.valid = true,
-			},
-			{
-				.wm_inst = WM_B,
-				.wm_type = WM_TYPE_PSTATE_CHG,
-				.pstate_latency_us = 23.84,
-				.valid = true,
-			},
-			{
-				.wm_inst = WM_C,
-				.wm_type = WM_TYPE_PSTATE_CHG,
-				.pstate_latency_us = 23.84,
-				.valid = true,
-			},
-			{
-				.wm_inst = WM_D,
-				.wm_type = WM_TYPE_PSTATE_CHG,
-				.pstate_latency_us = 23.84,
-				.valid = true,
-			},
+};
+
+struct wm_table ddr4_wm_table = {
+	.entries = {
+		{
+			.wm_inst = WM_A,
+			.wm_type = WM_TYPE_PSTATE_CHG,
+			.pstate_latency_us = 11.72,
+			.sr_exit_time_us = 6.09,
+			.sr_enter_plus_exit_time_us = 7.14,
+			.valid = true,
+		},
+		{
+			.wm_inst = WM_B,
+			.wm_type = WM_TYPE_PSTATE_CHG,
+			.pstate_latency_us = 11.72,
+			.sr_exit_time_us = 10.12,
+			.sr_enter_plus_exit_time_us = 11.48,
+			.valid = true,
+		},
+		{
+			.wm_inst = WM_C,
+			.wm_type = WM_TYPE_PSTATE_CHG,
+			.pstate_latency_us = 11.72,
+			.sr_exit_time_us = 10.12,
+			.sr_enter_plus_exit_time_us = 11.48,
+			.valid = true,
+		},
+		{
+			.wm_inst = WM_D,
+			.wm_type = WM_TYPE_PSTATE_CHG,
+			.pstate_latency_us = 11.72,
+			.sr_exit_time_us = 10.12,
+			.sr_enter_plus_exit_time_us = 11.48,
+			.valid = true,
 		},
 	}
 };
 
+struct wm_table lpddr4_wm_table = {
+	.entries = {
+		{
+			.wm_inst = WM_A,
+			.wm_type = WM_TYPE_PSTATE_CHG,
+			.pstate_latency_us = 23.84,
+			.sr_exit_time_us = 12.5,
+			.sr_enter_plus_exit_time_us = 17.0,
+			.valid = true,
+		},
+		{
+			.wm_inst = WM_B,
+			.wm_type = WM_TYPE_PSTATE_CHG,
+			.pstate_latency_us = 23.84,
+			.sr_exit_time_us = 12.5,
+			.sr_enter_plus_exit_time_us = 17.0,
+			.valid = true,
+		},
+		{
+			.wm_inst = WM_C,
+			.wm_type = WM_TYPE_PSTATE_CHG,
+			.pstate_latency_us = 23.84,
+			.sr_exit_time_us = 12.5,
+			.sr_enter_plus_exit_time_us = 17.0,
+			.valid = true,
+		},
+		{
+			.wm_inst = WM_D,
+			.wm_type = WM_TYPE_PSTATE_CHG,
+			.pstate_latency_us = 23.84,
+			.sr_exit_time_us = 12.5,
+			.sr_enter_plus_exit_time_us = 17.0,
+			.valid = true,
+		},
+	}
+};
+
+
 static unsigned int find_dcfclk_for_voltage(struct dpm_clocks *clock_table, unsigned int voltage)
 {
 	int i;
@@ -561,7 +624,7 @@ static unsigned int find_dcfclk_for_voltage(struct dpm_clocks *clock_table, unsi
 	return 0;
 }
 
-static void rn_clk_mgr_helper_populate_bw_params(struct clk_bw_params *bw_params, struct dpm_clocks *clock_table, struct hw_asic_id *asic_id)
+static void rn_clk_mgr_helper_populate_bw_params(struct clk_bw_params *bw_params, struct dpm_clocks *clock_table, struct integrated_info *bios_info)
 {
 	int i, j = 0;
 
@@ -593,8 +656,8 @@ static void rn_clk_mgr_helper_populate_bw_params(struct clk_bw_params *bw_params
 		bw_params->clk_table.entries[i].dcfclk_mhz = find_dcfclk_for_voltage(clock_table, clock_table->FClocks[j].Vol);
 	}
 
-	bw_params->vram_type = asic_id->vram_type;
-	bw_params->num_channels = asic_id->vram_width / DDR4_DRAM_WIDTH;
+	bw_params->vram_type = bios_info->memory_type;
+	bw_params->num_channels = bios_info->ma_channel_number;
 
 	for (i = 0; i < WM_SET_COUNT; i++) {
 		bw_params->wm_table.entries[i].wm_inst = i;
@@ -669,15 +732,24 @@ void rn_clk_mgr_construct(
 			ASSERT(clk_mgr->base.dprefclk_khz == 600000);
 			clk_mgr->base.dprefclk_khz = 600000;
 		}
+
+		if (ctx->dc_bios->integrated_info->memory_type == LpDdr4MemType) {
+			rn_bw_params.wm_table = lpddr4_wm_table;
+		} else {
+			rn_bw_params.wm_table = ddr4_wm_table;
+		}
 	}
 
 	dce_clock_read_ss_info(clk_mgr);
 
+
 	clk_mgr->base.bw_params = &rn_bw_params;
 
 	if (pp_smu && pp_smu->rn_funcs.get_dpm_clock_table) {
 		pp_smu->rn_funcs.get_dpm_clock_table(&pp_smu->rn_funcs.pp_smu, &clock_table);
-		rn_clk_mgr_helper_populate_bw_params(clk_mgr->base.bw_params, &clock_table, &ctx->asic_id);
+		if (ctx->dc_bios && ctx->dc_bios->integrated_info) {
+			rn_clk_mgr_helper_populate_bw_params (clk_mgr->base.bw_params, &clock_table, ctx->dc_bios->integrated_info);
+		}
 	}
 
 	if (!IS_FPGA_MAXIMUS_DC(ctx->dce_environment) && clk_mgr->smu_ver >= 0x00371500) {

drivers/gpu/drm/amd/display/dc/core/dc_link.c

@@ -372,7 +372,7 @@ bool dc_link_is_dp_sink_present(struct dc_link *link)
 
 	if (GPIO_RESULT_OK != dal_ddc_open(
 		ddc, GPIO_MODE_INPUT, GPIO_DDC_CONFIG_TYPE_MODE_I2C)) {
-		dal_gpio_destroy_ddc(&ddc);
+		dal_ddc_close(ddc);
 
 		return present;
 	}

drivers/gpu/drm/amd/display/dc/core/dc_link_ddc.c

@@ -586,7 +586,7 @@ bool dal_ddc_service_query_ddc_data(
 bool dal_ddc_submit_aux_command(struct ddc_service *ddc,
 		struct aux_payload *payload)
 {
-	uint8_t retrieved = 0;
+	uint32_t retrieved = 0;
 	bool ret = 0;
 
 	if (!ddc)

drivers/gpu/drm/amd/display/dc/core/dc_link_dp.c

@@ -3522,7 +3522,14 @@ void dp_set_fec_enable(struct dc_link *link, bool enable)
 	if (link_enc->funcs->fec_set_enable &&
 			link->dpcd_caps.fec_cap.bits.FEC_CAPABLE) {
 		if (link->fec_state == dc_link_fec_ready && enable) {
-			msleep(1);
+			/* Accord to DP spec, FEC enable sequence can first
+			 * be transmitted anytime after 1000 LL codes have
+			 * been transmitted on the link after link training
+			 * completion. Using 1 lane RBR should have the maximum
+			 * time for transmitting 1000 LL codes which is 6.173 us.
+			 * So use 7 microseconds delay instead.
+			 */
+			udelay(7);
 			link_enc->funcs->fec_set_enable(link_enc, true);
 			link->fec_state = dc_link_fec_enabled;
 		} else if (link->fec_state == dc_link_fec_enabled && !enable) {

drivers/gpu/drm/amd/display/dc/dce/dce_aux.c

@@ -583,6 +583,8 @@ bool dce_aux_transfer_with_retries(struct ddc_service *ddc,
 	uint8_t reply;
 	bool payload_reply = true;
 	enum aux_channel_operation_result operation_result;
+	bool retry_on_defer = false;
+
 	int aux_ack_retries = 0,
 		aux_defer_retries = 0,
 		aux_i2c_defer_retries = 0,
@@ -613,8 +615,10 @@ bool dce_aux_transfer_with_retries(struct ddc_service *ddc,
 			break;
 
 			case AUX_TRANSACTION_REPLY_AUX_DEFER:
-			case AUX_TRANSACTION_REPLY_I2C_OVER_AUX_NACK:
 			case AUX_TRANSACTION_REPLY_I2C_OVER_AUX_DEFER:
+				retry_on_defer = true;
+				/* fall through */
+			case AUX_TRANSACTION_REPLY_I2C_OVER_AUX_NACK:
 				if (++aux_defer_retries >= AUX_MAX_DEFER_RETRIES) {
 					goto fail;
 				} else {
@@ -647,15 +651,24 @@ bool dce_aux_transfer_with_retries(struct ddc_service *ddc,
 			break;
 
 		case AUX_CHANNEL_OPERATION_FAILED_TIMEOUT:
-			if (++aux_timeout_retries >= AUX_MAX_TIMEOUT_RETRIES)
-				goto fail;
-			else {
-				/*
-				 * DP 1.4, 2.8.2:  AUX Transaction Response/Reply Timeouts
-				 * According to the DP spec there should be 3 retries total
-				 * with a 400us wait inbetween each. Hardware already waits
-				 * for 550us therefore no wait is required here.
-				 */
+			// Check whether a DEFER had occurred before the timeout.
+			// If so, treat timeout as a DEFER.
+			if (retry_on_defer) {
+				if (++aux_defer_retries >= AUX_MAX_DEFER_RETRIES)
+					goto fail;
+				else if (payload->defer_delay > 0)
+					msleep(payload->defer_delay);
+			} else {
+				if (++aux_timeout_retries >= AUX_MAX_TIMEOUT_RETRIES)
+					goto fail;
+				else {
+					/*
+					 * DP 1.4, 2.8.2:  AUX Transaction Response/Reply Timeouts
+					 * According to the DP spec there should be 3 retries total
+					 * with a 400us wait inbetween each. Hardware already waits
+					 * for 550us therefore no wait is required here.
+					 */
+				}
 			}
 			break;
 

drivers/gpu/drm/amd/display/dc/dcn20/Makefile

+# SPDX-License-Identifier: MIT
 #
 # Makefile for DCN.
 

drivers/gpu/drm/amd/display/dc/dcn20/dcn20_resource.c

@@ -923,7 +923,9 @@ static const struct resource_caps res_cap_nv14 = {
 		.num_dwb = 1,
 		.num_ddc = 5,
 		.num_vmid = 16,
+#ifdef CONFIG_DRM_AMD_DC_DSC_SUPPORT
 		.num_dsc = 5,
+#endif
 };
 
 static const struct dc_debug_options debug_defaults_drv = {
@@ -1536,13 +1538,20 @@ enum dc_status dcn20_build_mapped_resource(const struct dc *dc, struct dc_state
 
 static void acquire_dsc(struct resource_context *res_ctx,
 			const struct resource_pool *pool,
-			struct display_stream_compressor **dsc)
+			struct display_stream_compressor **dsc,
+			int pipe_idx)
 {
 	int i;
 
 	ASSERT(*dsc == NULL);
 	*dsc = NULL;
 
+	if (pool->res_cap->num_dsc == pool->res_cap->num_opp) {
+		*dsc = pool->dscs[pipe_idx];
+		res_ctx->is_dsc_acquired[pipe_idx] = true;
+		return;
+	}
+
 	/* Find first free DSC */
 	for (i = 0; i < pool->res_cap->num_dsc; i++)
 		if (!res_ctx->is_dsc_acquired[i]) {
@@ -1585,7 +1594,7 @@ static enum dc_status add_dsc_to_stream_resource(struct dc *dc,
 		if (pipe_ctx->stream != dc_stream)
 			continue;
 
-		acquire_dsc(&dc_ctx->res_ctx, pool, &pipe_ctx->stream_res.dsc);
+		acquire_dsc(&dc_ctx->res_ctx, pool, &pipe_ctx->stream_res.dsc, i);
 
 		/* The number of DSCs can be less than the number of pipes */
 		if (!pipe_ctx->stream_res.dsc) {
@@ -1785,7 +1794,7 @@ bool dcn20_split_stream_for_odm(
 	next_odm_pipe->stream_res.opp = pool->opps[next_odm_pipe->pipe_idx];
 #ifdef CONFIG_DRM_AMD_DC_DSC_SUPPORT
 	if (next_odm_pipe->stream->timing.flags.DSC == 1) {
-		acquire_dsc(res_ctx, pool, &next_odm_pipe->stream_res.dsc);
+		acquire_dsc(res_ctx, pool, &next_odm_pipe->stream_res.dsc, next_odm_pipe->pipe_idx);
 		ASSERT(next_odm_pipe->stream_res.dsc);
 		if (next_odm_pipe->stream_res.dsc == NULL)
 			return false;

drivers/gpu/drm/amd/display/dc/dcn20/dcn20_stream_encoder.c

@@ -492,15 +492,23 @@ void enc2_stream_encoder_dp_unblank(
 				DP_VID_N_MUL, n_multiply);
 	}
 
-	/* set DIG_START to 0x1 to reset FIFO */
+	/* make sure stream is disabled before resetting steer fifo */
+	REG_UPDATE(DP_VID_STREAM_CNTL, DP_VID_STREAM_ENABLE, false);
+	REG_WAIT(DP_VID_STREAM_CNTL, DP_VID_STREAM_STATUS, 0, 10, 5000);
 
+	/* set DIG_START to 0x1 to reset FIFO */
 	REG_UPDATE(DIG_FE_CNTL, DIG_START, 1);
+	udelay(1);
 
 	/* write 0 to take the FIFO out of reset */
 
 	REG_UPDATE(DIG_FE_CNTL, DIG_START, 0);
 
-	/* switch DP encoder to CRTC data */
+	/* switch DP encoder to CRTC data, but reset it the fifo first. It may happen
+	 * that it overflows during mode transition, and sometimes doesn't recover.
+	 */
+	REG_UPDATE(DP_STEER_FIFO, DP_STEER_FIFO_RESET, 1);
+	udelay(10);
 
 	REG_UPDATE(DP_STEER_FIFO, DP_STEER_FIFO_RESET, 0);
 

drivers/gpu/drm/amd/display/dc/dcn21/Makefile

+# SPDX-License-Identifier: MIT
 #
 # Makefile for DCN21.
 

drivers/gpu/drm/amd/display/dc/dcn21/dcn21_resource.c

@@ -23,6 +23,8 @@
  *
  */
 
+#include <linux/slab.h>
+
 #include "dm_services.h"
 #include "dc.h"
 
@@ -257,7 +259,7 @@ struct _vcs_dpi_soc_bounding_box_st dcn2_1_soc = {
 	.vmm_page_size_bytes = 4096,
 	.dram_clock_change_latency_us = 23.84,
 	.return_bus_width_bytes = 64,
-	.dispclk_dppclk_vco_speed_mhz = 3550,
+	.dispclk_dppclk_vco_speed_mhz = 3600,
 	.xfc_bus_transport_time_us = 4,
 	.xfc_xbuf_latency_tolerance_us = 4,
 	.use_urgent_burst_bw = 1,
@@ -1000,6 +1002,8 @@ static void calculate_wm_set_for_vlevel(
 	pipes[0].clks_cfg.socclk_mhz = dml->soc.clock_limits[vlevel].socclk_mhz;
 
 	dml->soc.dram_clock_change_latency_us = table_entry->pstate_latency_us;
+	dml->soc.sr_exit_time_us = table_entry->sr_exit_time_us;
+	dml->soc.sr_enter_plus_exit_time_us = table_entry->sr_enter_plus_exit_time_us;
 
 	wm_set->urgent_ns = get_wm_urgent(dml, pipes, pipe_cnt) * 1000;
 	wm_set->cstate_pstate.cstate_enter_plus_exit_ns = get_wm_stutter_enter_exit(dml, pipes, pipe_cnt) * 1000;
@@ -1017,14 +1021,21 @@ static void calculate_wm_set_for_vlevel(
 
 static void patch_bounding_box(struct dc *dc, struct _vcs_dpi_soc_bounding_box_st *bb)
 {
+	int i;
+
 	kernel_fpu_begin();
 	if (dc->bb_overrides.sr_exit_time_ns) {
-		bb->sr_exit_time_us = dc->bb_overrides.sr_exit_time_ns / 1000.0;
+		for (i = 0; i < WM_SET_COUNT; i++) {
+			  dc->clk_mgr->bw_params->wm_table.entries[i].sr_exit_time_us =
+					  dc->bb_overrides.sr_exit_time_ns / 1000.0;
+		}
 	}
 
 	if (dc->bb_overrides.sr_enter_plus_exit_time_ns) {
-		bb->sr_enter_plus_exit_time_us =
-				dc->bb_overrides.sr_enter_plus_exit_time_ns / 1000.0;
+		for (i = 0; i < WM_SET_COUNT; i++) {
+			  dc->clk_mgr->bw_params->wm_table.entries[i].sr_enter_plus_exit_time_us =
+					  dc->bb_overrides.sr_enter_plus_exit_time_ns / 1000.0;
+		}
 	}
 
 	if (dc->bb_overrides.urgent_latency_ns) {
@@ -1032,9 +1043,12 @@ static void patch_bounding_box(struct dc *dc, struct _vcs_dpi_soc_bounding_box_s
 	}
 
 	if (dc->bb_overrides.dram_clock_change_latency_ns) {
-		bb->dram_clock_change_latency_us =
+		for (i = 0; i < WM_SET_COUNT; i++) {
+			dc->clk_mgr->bw_params->wm_table.entries[i].pstate_latency_us =
 				dc->bb_overrides.dram_clock_change_latency_ns / 1000.0;
+		}
 	}
+
 	kernel_fpu_end();
 }
 

drivers/gpu/drm/amd/display/dc/dsc/Makefile

+# SPDX-License-Identifier: MIT
 #
 # Makefile for the 'dsc' sub-component of DAL.
 

drivers/gpu/drm/amd/display/dc/inc/hw/clk_mgr.h

@@ -69,6 +69,8 @@ struct wm_range_table_entry {
 	unsigned int wm_inst;
 	unsigned int wm_type;
 	double pstate_latency_us;
+	double sr_exit_time_us;
+	double sr_enter_plus_exit_time_us;
 	bool valid;
 };
 

drivers/gpu/drm/amd/display/include/i2caux_interface.h

@@ -42,7 +42,7 @@ struct aux_payload {
 	bool write;
 	bool mot;
 	uint32_t address;
-	uint8_t length;
+	uint32_t length;
 	uint8_t *data;
 	/*
 	 * used to return the reply type of the transaction

drivers/gpu/drm/amd/display/modules/freesync/freesync.c

@@ -37,8 +37,8 @@
 #define STATIC_SCREEN_RAMP_DELTA_REFRESH_RATE_PER_FRAME ((1000 / 60) * 65)
 /* Number of elements in the render times cache array */
 #define RENDER_TIMES_MAX_COUNT 10
-/* Threshold to exit/exit BTR (to avoid frequent enter-exits at the lower limit) */
-#define BTR_MAX_MARGIN 2500
+/* Threshold to exit BTR (to avoid frequent enter-exits at the lower limit) */
+#define BTR_EXIT_MARGIN 2000
 /* Threshold to change BTR multiplier (to avoid frequent changes) */
 #define BTR_DRIFT_MARGIN 2000
 /*Threshold to exit fixed refresh rate*/
@@ -254,22 +254,24 @@ static void apply_below_the_range(struct core_freesync *core_freesync,
 	unsigned int delta_from_mid_point_in_us_1 = 0xFFFFFFFF;
 	unsigned int delta_from_mid_point_in_us_2 = 0xFFFFFFFF;
 	unsigned int frames_to_insert = 0;
+	unsigned int min_frame_duration_in_ns = 0;
+	unsigned int max_render_time_in_us = in_out_vrr->max_duration_in_us;
 	unsigned int delta_from_mid_point_delta_in_us;
-	unsigned int max_render_time_in_us =
-			in_out_vrr->max_duration_in_us - in_out_vrr->btr.margin_in_us;
+
+	min_frame_duration_in_ns = ((unsigned int) (div64_u64(
+		(1000000000ULL * 1000000),
+		in_out_vrr->max_refresh_in_uhz)));
 
 	/* Program BTR */
-	if ((last_render_time_in_us + in_out_vrr->btr.margin_in_us / 2) < max_render_time_in_us) {
+	if (last_render_time_in_us + BTR_EXIT_MARGIN < max_render_time_in_us) {
 		/* Exit Below the Range */
 		if (in_out_vrr->btr.btr_active) {
 			in_out_vrr->btr.frame_counter = 0;
 			in_out_vrr->btr.btr_active = false;
 		}
-	} else if (last_render_time_in_us > (max_render_time_in_us + in_out_vrr->btr.margin_in_us / 2)) {
+	} else if (last_render_time_in_us > max_render_time_in_us) {
 		/* Enter Below the Range */
-		if (!in_out_vrr->btr.btr_active) {
-			in_out_vrr->btr.btr_active = true;
-		}
+		in_out_vrr->btr.btr_active = true;
 	}
 
 	/* BTR set to "not active" so disengage */
@@ -325,9 +327,7 @@ static void apply_below_the_range(struct core_freesync *core_freesync,
 		/* Choose number of frames to insert based on how close it
 		 * can get to the mid point of the variable range.
 		 */
-		if ((frame_time_in_us / mid_point_frames_ceil) > in_out_vrr->min_duration_in_us &&
-				(delta_from_mid_point_in_us_1 < delta_from_mid_point_in_us_2 ||
-						mid_point_frames_floor < 2)) {
+		if (delta_from_mid_point_in_us_1 < delta_from_mid_point_in_us_2) {
 			frames_to_insert = mid_point_frames_ceil;
 			delta_from_mid_point_delta_in_us = delta_from_mid_point_in_us_2 -
 					delta_from_mid_point_in_us_1;
@@ -343,7 +343,7 @@ static void apply_below_the_range(struct core_freesync *core_freesync,
 		if (in_out_vrr->btr.frames_to_insert != 0 &&
 				delta_from_mid_point_delta_in_us < BTR_DRIFT_MARGIN) {
 			if (((last_render_time_in_us / in_out_vrr->btr.frames_to_insert) <
-					max_render_time_in_us) &&
+					in_out_vrr->max_duration_in_us) &&
 				((last_render_time_in_us / in_out_vrr->btr.frames_to_insert) >
 					in_out_vrr->min_duration_in_us))
 				frames_to_insert = in_out_vrr->btr.frames_to_insert;
@@ -796,11 +796,6 @@ void mod_freesync_build_vrr_params(struct mod_freesync *mod_freesync,
 		refresh_range = in_out_vrr->max_refresh_in_uhz -
 				in_out_vrr->min_refresh_in_uhz;
 
-		in_out_vrr->btr.margin_in_us = in_out_vrr->max_duration_in_us -
-				2 * in_out_vrr->min_duration_in_us;
-		if (in_out_vrr->btr.margin_in_us > BTR_MAX_MARGIN)
-			in_out_vrr->btr.margin_in_us = BTR_MAX_MARGIN;
-
 		in_out_vrr->supported = true;
 	}
 
@@ -816,7 +811,6 @@ void mod_freesync_build_vrr_params(struct mod_freesync *mod_freesync,
 	in_out_vrr->btr.inserted_duration_in_us = 0;
 	in_out_vrr->btr.frames_to_insert = 0;
 	in_out_vrr->btr.frame_counter = 0;
-
 	in_out_vrr->btr.mid_point_in_us =
 				(in_out_vrr->min_duration_in_us +
 				 in_out_vrr->max_duration_in_us) / 2;

drivers/gpu/drm/amd/display/modules/inc/mod_freesync.h

@@ -92,7 +92,6 @@ struct mod_vrr_params_btr {
 	uint32_t inserted_duration_in_us;
 	uint32_t frames_to_insert;
 	uint32_t frame_counter;
-	uint32_t margin_in_us;
 };
 
 struct mod_vrr_params_fixed_refresh {

drivers/gpu/drm/amd/powerplay/arcturus_ppt.c

@@ -1313,12 +1313,17 @@ static int arcturus_get_power_profile_mode(struct smu_context *smu,
 					"VR",
 					"COMPUTE",
 					"CUSTOM"};
+	static const char *title[] = {
+			"PROFILE_INDEX(NAME)"};
 	uint32_t i, size = 0;
 	int16_t workload_type = 0;
 
 	if (!smu->pm_enabled || !buf)
 		return -EINVAL;
 
+	size += sprintf(buf + size, "%16s\n",
+			title[0]);
+
 	for (i = 0; i <= PP_SMC_POWER_PROFILE_CUSTOM; i++) {
 		/*
 		 * Conv PP_SMC_POWER_PROFILE* to WORKLOAD_PPLIB_*_BIT