Merge tag 'amd-drm-fixes-6.11-2024-09-05' of https://gitlab.freedesktop.org/agd5f/linux into drm-fixes

amd-drm-fixes-6.11-2024-09-05:

amdgpu:
- IPS workaround
- Fix compatibility with older MES firmware
- Fix CPU spikes when clearing VRAM
- Backlight fix
- PMO fix
- Revert SWSMU change to fix regression

Signed-off-by: Dave Airlie <airlied@redhat.com>

From: Alex Deucher <alexander.deucher@amd.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20240905190533.854116-1-alexander.deucher@amd.com

8 files changed, 89 insertions, 25 deletions

diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_gem.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_gem.c
index aad2027e5c7c..0e617dff8765 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_gem.c
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_gem.c
@@ -348,6 +348,9 @@ int amdgpu_gem_create_ioctl(struct drm_device *dev, void *data,
 		return -EINVAL;
 	}
 
+	/* always clear VRAM */
+	flags |= AMDGPU_GEM_CREATE_VRAM_CLEARED;
+
 	/* create a gem object to contain this object in */
 	if (args->in.domains & (AMDGPU_GEM_DOMAIN_GDS |
 	    AMDGPU_GEM_DOMAIN_GWS | AMDGPU_GEM_DOMAIN_OA)) {
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_gfx.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_gfx.c
index c770cb201e64..1849510a308a 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_gfx.c
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_gfx.c
@@ -657,7 +657,7 @@ int amdgpu_gfx_enable_kcq(struct amdgpu_device *adev, int xcc_id)
 	uint64_t queue_mask = 0;
 	int r, i, j;
 
-	if (adev->enable_mes)
+	if (adev->mes.enable_legacy_queue_map)
 		return amdgpu_gfx_mes_enable_kcq(adev, xcc_id);
 
 	if (!kiq->pmf || !kiq->pmf->kiq_map_queues || !kiq->pmf->kiq_set_resources)
@@ -719,7 +719,7 @@ int amdgpu_gfx_enable_kgq(struct amdgpu_device *adev, int xcc_id)
 
 	amdgpu_device_flush_hdp(adev, NULL);
 
-	if (adev->enable_mes) {
+	if (adev->mes.enable_legacy_queue_map) {
 		for (i = 0; i < adev->gfx.num_gfx_rings; i++) {
 			j = i + xcc_id * adev->gfx.num_gfx_rings;
 			r = amdgpu_mes_map_legacy_queue(adev,
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_mes.h b/drivers/gpu/drm/amd/amdgpu/amdgpu_mes.h
index 0bc837dab578..bcce1add4ef6 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_mes.h
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_mes.h
@@ -75,6 +75,7 @@ struct amdgpu_mes {
 
 	uint32_t			sched_version;
 	uint32_t			kiq_version;
+	bool                            enable_legacy_queue_map;
 
 	uint32_t                        total_max_queue;
 	uint32_t                        max_doorbell_slices;
diff --git a/drivers/gpu/drm/amd/amdgpu/mes_v11_0.c b/drivers/gpu/drm/amd/amdgpu/mes_v11_0.c
index 2ea8223eb969..8aded0a67037 100644
--- a/drivers/gpu/drm/amd/amdgpu/mes_v11_0.c
+++ b/drivers/gpu/drm/amd/amdgpu/mes_v11_0.c
@@ -693,6 +693,28 @@ static void mes_v11_0_free_ucode_buffers(struct amdgpu_device *adev,
 			      (void **)&adev->mes.ucode_fw_ptr[pipe]);
 }
 
+static void mes_v11_0_get_fw_version(struct amdgpu_device *adev)
+{
+	int pipe;
+
+	/* get MES scheduler/KIQ versions */
+	mutex_lock(&adev->srbm_mutex);
+
+	for (pipe = 0; pipe < AMDGPU_MAX_MES_PIPES; pipe++) {
+		soc21_grbm_select(adev, 3, pipe, 0, 0);
+
+		if (pipe == AMDGPU_MES_SCHED_PIPE)
+			adev->mes.sched_version =
+				RREG32_SOC15(GC, 0, regCP_MES_GP3_LO);
+		else if (pipe == AMDGPU_MES_KIQ_PIPE && adev->enable_mes_kiq)
+			adev->mes.kiq_version =
+				RREG32_SOC15(GC, 0, regCP_MES_GP3_LO);
+	}
+
+	soc21_grbm_select(adev, 0, 0, 0, 0);
+	mutex_unlock(&adev->srbm_mutex);
+}
+
 static void mes_v11_0_enable(struct amdgpu_device *adev, bool enable)
 {
 	uint64_t ucode_addr;
@@ -1062,18 +1084,6 @@ static int mes_v11_0_queue_init(struct amdgpu_device *adev,
 		mes_v11_0_queue_init_register(ring);
 	}
 
-	/* get MES scheduler/KIQ versions */
-	mutex_lock(&adev->srbm_mutex);
-	soc21_grbm_select(adev, 3, pipe, 0, 0);
-
-	if (pipe == AMDGPU_MES_SCHED_PIPE)
-		adev->mes.sched_version = RREG32_SOC15(GC, 0, regCP_MES_GP3_LO);
-	else if (pipe == AMDGPU_MES_KIQ_PIPE && adev->enable_mes_kiq)
-		adev->mes.kiq_version = RREG32_SOC15(GC, 0, regCP_MES_GP3_LO);
-
-	soc21_grbm_select(adev, 0, 0, 0, 0);
-	mutex_unlock(&adev->srbm_mutex);
-
 	return 0;
 }
 
@@ -1320,15 +1330,24 @@ static int mes_v11_0_kiq_hw_init(struct amdgpu_device *adev)
 
 	mes_v11_0_enable(adev, true);
 
+	mes_v11_0_get_fw_version(adev);
+
 	mes_v11_0_kiq_setting(&adev->gfx.kiq[0].ring);
 
 	r = mes_v11_0_queue_init(adev, AMDGPU_MES_KIQ_PIPE);
 	if (r)
 		goto failure;
 
-	r = mes_v11_0_hw_init(adev);
-	if (r)
-		goto failure;
+	if ((adev->mes.sched_version & AMDGPU_MES_VERSION_MASK) >= 0x47)
+		adev->mes.enable_legacy_queue_map = true;
+	else
+		adev->mes.enable_legacy_queue_map = false;
+
+	if (adev->mes.enable_legacy_queue_map) {
+		r = mes_v11_0_hw_init(adev);
+		if (r)
+			goto failure;
+	}
 
 	return r;
 
diff --git a/drivers/gpu/drm/amd/amdgpu/mes_v12_0.c b/drivers/gpu/drm/amd/amdgpu/mes_v12_0.c
index e39a58d262c9..a79a8adc3aa5 100644
--- a/drivers/gpu/drm/amd/amdgpu/mes_v12_0.c
+++ b/drivers/gpu/drm/amd/amdgpu/mes_v12_0.c
@@ -1266,6 +1266,7 @@ static int mes_v12_0_sw_init(void *handle)
 	adev->mes.funcs = &mes_v12_0_funcs;
 	adev->mes.kiq_hw_init = &mes_v12_0_kiq_hw_init;
 	adev->mes.kiq_hw_fini = &mes_v12_0_kiq_hw_fini;
+	adev->mes.enable_legacy_queue_map = true;
 
 	adev->mes.event_log_size = AMDGPU_MES_LOG_BUFFER_SIZE;
 
@@ -1422,9 +1423,11 @@ static int mes_v12_0_kiq_hw_init(struct amdgpu_device *adev)
 		mes_v12_0_set_hw_resources_1(&adev->mes, AMDGPU_MES_KIQ_PIPE);
 	}
 
-	r = mes_v12_0_hw_init(adev);
-	if (r)
-		goto failure;
+	if (adev->mes.enable_legacy_queue_map) {
+		r = mes_v12_0_hw_init(adev);
+		if (r)
+			goto failure;
+	}
 
 	return r;
 
diff --git a/drivers/gpu/drm/amd/display/amdgpu_dm/amdgpu_dm.c b/drivers/gpu/drm/amd/display/amdgpu_dm/amdgpu_dm.c
index 983a977632ff..5942fc4e1c86 100644
--- a/drivers/gpu/drm/amd/display/amdgpu_dm/amdgpu_dm.c
+++ b/drivers/gpu/drm/amd/display/amdgpu_dm/amdgpu_dm.c
@@ -1752,6 +1752,30 @@ static struct dml2_soc_bb *dm_dmub_get_vbios_bounding_box(struct amdgpu_device *
 	return bb;
 }
 
+static enum dmub_ips_disable_type dm_get_default_ips_mode(
+	struct amdgpu_device *adev)
+{
+	/*
+	 * On DCN35 systems with Z8 enabled, it's possible for IPS2 + Z8 to
+	 * cause a hard hang. A fix exists for newer PMFW.
+	 *
+	 * As a workaround, for non-fixed PMFW, force IPS1+RCG as the deepest
+	 * IPS state in all cases, except for s0ix and all displays off (DPMS),
+	 * where IPS2 is allowed.
+	 *
+	 * When checking pmfw version, use the major and minor only.
+	 */
+	if (amdgpu_ip_version(adev, DCE_HWIP, 0) == IP_VERSION(3, 5, 0) &&
+	    (adev->pm.fw_version & 0x00FFFF00) < 0x005D6300)
+		return DMUB_IPS_RCG_IN_ACTIVE_IPS2_IN_OFF;
+
+	if (amdgpu_ip_version(adev, DCE_HWIP, 0) >= IP_VERSION(3, 5, 0))
+		return DMUB_IPS_ENABLE;
+
+	/* ASICs older than DCN35 do not have IPSs */
+	return DMUB_IPS_DISABLE_ALL;
+}
+
 static int amdgpu_dm_init(struct amdgpu_device *adev)
 {
 	struct dc_init_data init_data;
@@ -1863,7 +1887,7 @@ static int amdgpu_dm_init(struct amdgpu_device *adev)
 	if (amdgpu_dc_debug_mask & DC_DISABLE_IPS)
 		init_data.flags.disable_ips = DMUB_IPS_DISABLE_ALL;
 	else
-		init_data.flags.disable_ips = DMUB_IPS_ENABLE;
+		init_data.flags.disable_ips = dm_get_default_ips_mode(adev);
 
 	init_data.flags.disable_ips_in_vpb = 0;
 
@@ -4492,7 +4516,7 @@ static void amdgpu_dm_backlight_set_level(struct amdgpu_display_manager *dm,
 	struct amdgpu_dm_backlight_caps caps;
 	struct dc_link *link;
 	u32 brightness;
-	bool rc;
+	bool rc, reallow_idle = false;
 
 	amdgpu_dm_update_backlight_caps(dm, bl_idx);
 	caps = dm->backlight_caps[bl_idx];
@@ -4505,6 +4529,12 @@ static void amdgpu_dm_backlight_set_level(struct amdgpu_display_manager *dm,
 	link = (struct dc_link *)dm->backlight_link[bl_idx];
 
 	/* Change brightness based on AUX property */
+	mutex_lock(&dm->dc_lock);
+	if (dm->dc->caps.ips_support && dm->dc->ctx->dmub_srv->idle_allowed) {
+		dc_allow_idle_optimizations(dm->dc, false);
+		reallow_idle = true;
+	}
+
 	if (caps.aux_support) {
 		rc = dc_link_set_backlight_level_nits(link, true, brightness,
 						      AUX_BL_DEFAULT_TRANSITION_TIME_MS);
@@ -4516,6 +4546,11 @@ static void amdgpu_dm_backlight_set_level(struct amdgpu_display_manager *dm,
 			DRM_DEBUG("DM: Failed to update backlight on eDP[%d]\n", bl_idx);
 	}
 
+	if (dm->dc->caps.ips_support && reallow_idle)
+		dc_allow_idle_optimizations(dm->dc, true);
+
+	mutex_unlock(&dm->dc_lock);
+
 	if (rc)
 		dm->actual_brightness[bl_idx] = user_brightness;
 }
diff --git a/drivers/gpu/drm/amd/display/dc/dml2/dml21/src/dml2_pmo/dml2_pmo_dcn4_fams2.c b/drivers/gpu/drm/amd/display/dc/dml2/dml21/src/dml2_pmo/dml2_pmo_dcn4_fams2.c
index 603036df68ba..6547cc2c2a77 100644
--- a/drivers/gpu/drm/amd/display/dc/dml2/dml21/src/dml2_pmo/dml2_pmo_dcn4_fams2.c
+++ b/drivers/gpu/drm/amd/display/dc/dml2/dml21/src/dml2_pmo/dml2_pmo_dcn4_fams2.c
@@ -811,7 +811,8 @@ static void build_synchronized_timing_groups(
 		for (j = i + 1; j < display_config->display_config.num_streams; j++) {
 			if (memcmp(master_timing,
 				&display_config->display_config.stream_descriptors[j].timing,
-				sizeof(struct dml2_timing_cfg)) == 0) {
+				sizeof(struct dml2_timing_cfg)) == 0 &&
+				display_config->display_config.stream_descriptors[i].output.output_encoder == display_config->display_config.stream_descriptors[j].output.output_encoder) {
 				set_bit_in_bitfield(&pmo->scratch.pmo_dcn4.synchronized_timing_group_masks[timing_group_idx], j);
 				set_bit_in_bitfield(&stream_mapped_mask, j);
 			}
diff --git a/drivers/gpu/drm/amd/pm/swsmu/amdgpu_smu.c b/drivers/gpu/drm/amd/pm/swsmu/amdgpu_smu.c
index 74e35f8ddefc..2cf951184561 100644
--- a/drivers/gpu/drm/amd/pm/swsmu/amdgpu_smu.c
+++ b/drivers/gpu/drm/amd/pm/swsmu/amdgpu_smu.c
@@ -2266,7 +2266,8 @@ static int smu_adjust_power_state_dynamic(struct smu_context *smu,
 		smu_dpm_ctx->dpm_level = level;
 	}
 
-	if (smu_dpm_ctx->dpm_level != AMD_DPM_FORCED_LEVEL_PERF_DETERMINISM) {
+	if (smu_dpm_ctx->dpm_level != AMD_DPM_FORCED_LEVEL_MANUAL &&
+		smu_dpm_ctx->dpm_level != AMD_DPM_FORCED_LEVEL_PERF_DETERMINISM) {
 		index = fls(smu->workload_mask);
 		index = index > 0 && index <= WORKLOAD_POLICY_MAX ? index - 1 : 0;
 		workload[0] = smu->workload_setting[index];
@@ -2345,7 +2346,8 @@ static int smu_switch_power_profile(void *handle,
 		workload[0] = smu->workload_setting[index];
 	}
 
-	if (smu_dpm_ctx->dpm_level != AMD_DPM_FORCED_LEVEL_PERF_DETERMINISM)
+	if (smu_dpm_ctx->dpm_level != AMD_DPM_FORCED_LEVEL_MANUAL &&
+		smu_dpm_ctx->dpm_level != AMD_DPM_FORCED_LEVEL_PERF_DETERMINISM)
 		smu_bump_power_profile_mode(smu, workload, 0);
 
 	return 0;