From 0a6e7b06bdbead2e43d56a2274b7e0c9c86d536e Mon Sep 17 00:00:00 2001 From: David Rosca Date: Fri, 28 Feb 2025 14:12:10 +0100 Subject: [PATCH 01/16] drm/amdgpu: Remove JPEG from vega and carrizo video caps JPEG is only supported for VCN1+. Signed-off-by: David Rosca Acked-by: Alex Deucher Reviewed-by: Ruijing Dong Signed-off-by: Alex Deucher --- drivers/gpu/drm/amd/amdgpu/soc15.c | 1 - drivers/gpu/drm/amd/amdgpu/vi.c | 7 ------- 2 files changed, 8 deletions(-) diff --git a/drivers/gpu/drm/amd/amdgpu/soc15.c b/drivers/gpu/drm/amd/amdgpu/soc15.c index d6db75258e66..659eab9b90be 100644 --- a/drivers/gpu/drm/amd/amdgpu/soc15.c +++ b/drivers/gpu/drm/amd/amdgpu/soc15.c @@ -107,7 +107,6 @@ static const struct amdgpu_video_codec_info vega_video_codecs_decode_array[] = {codec_info_build(AMDGPU_INFO_VIDEO_CAPS_CODEC_IDX_MPEG4_AVC, 4096, 4096, 52)}, {codec_info_build(AMDGPU_INFO_VIDEO_CAPS_CODEC_IDX_VC1, 1920, 1088, 4)}, {codec_info_build(AMDGPU_INFO_VIDEO_CAPS_CODEC_IDX_HEVC, 4096, 4096, 186)}, - {codec_info_build(AMDGPU_INFO_VIDEO_CAPS_CODEC_IDX_JPEG, 4096, 4096, 0)}, }; static const struct amdgpu_video_codecs vega_video_codecs_decode = diff --git a/drivers/gpu/drm/amd/amdgpu/vi.c b/drivers/gpu/drm/amd/amdgpu/vi.c index 9b3510e53112..86d8bc10d90a 100644 --- a/drivers/gpu/drm/amd/amdgpu/vi.c +++ b/drivers/gpu/drm/amd/amdgpu/vi.c @@ -239,13 +239,6 @@ static const struct amdgpu_video_codec_info cz_video_codecs_decode_array[] = .max_pixels_per_frame = 4096 * 4096, .max_level = 186, }, - { - .codec_type = AMDGPU_INFO_VIDEO_CAPS_CODEC_IDX_JPEG, - .max_width = 4096, - .max_height = 4096, - .max_pixels_per_frame = 4096 * 4096, - .max_level = 0, - }, }; static const struct amdgpu_video_codecs cz_video_codecs_decode = -- 2.51.0 From 19478f2011f8b53dee401c91423c4e0b73753e4f Mon Sep 17 00:00:00 2001 From: David Rosca Date: Fri, 28 Feb 2025 13:44:32 +0100 Subject: [PATCH 02/16] drm/amdgpu: Update SRIOV video codec caps There have been multiple fixes to the video caps that are missing for SRIOV. Update the SRIOV caps with correct values. Signed-off-by: David Rosca Acked-by: Alex Deucher Reviewed-by: Ruijing Dong Signed-off-by: Alex Deucher --- drivers/gpu/drm/amd/amdgpu/nv.c | 16 ++++++++-------- drivers/gpu/drm/amd/amdgpu/soc21.c | 10 ++-------- 2 files changed, 10 insertions(+), 16 deletions(-) diff --git a/drivers/gpu/drm/amd/amdgpu/nv.c b/drivers/gpu/drm/amd/amdgpu/nv.c index 32b824c3658f..50e77d9b30af 100644 --- a/drivers/gpu/drm/amd/amdgpu/nv.c +++ b/drivers/gpu/drm/amd/amdgpu/nv.c @@ -141,23 +141,23 @@ static struct amdgpu_video_codec_info sriov_sc_video_codecs_encode_array[] = { }; static struct amdgpu_video_codec_info sriov_sc_video_codecs_decode_array_vcn0[] = { - {codec_info_build(AMDGPU_INFO_VIDEO_CAPS_CODEC_IDX_MPEG2, 4096, 4096, 3)}, - {codec_info_build(AMDGPU_INFO_VIDEO_CAPS_CODEC_IDX_MPEG4, 4096, 4096, 5)}, + {codec_info_build(AMDGPU_INFO_VIDEO_CAPS_CODEC_IDX_MPEG2, 1920, 1088, 3)}, + {codec_info_build(AMDGPU_INFO_VIDEO_CAPS_CODEC_IDX_MPEG4, 1920, 1088, 5)}, {codec_info_build(AMDGPU_INFO_VIDEO_CAPS_CODEC_IDX_MPEG4_AVC, 4096, 4096, 52)}, - {codec_info_build(AMDGPU_INFO_VIDEO_CAPS_CODEC_IDX_VC1, 4096, 4096, 4)}, + {codec_info_build(AMDGPU_INFO_VIDEO_CAPS_CODEC_IDX_VC1, 1920, 1088, 4)}, {codec_info_build(AMDGPU_INFO_VIDEO_CAPS_CODEC_IDX_HEVC, 8192, 4352, 186)}, - {codec_info_build(AMDGPU_INFO_VIDEO_CAPS_CODEC_IDX_JPEG, 4096, 4096, 0)}, + {codec_info_build(AMDGPU_INFO_VIDEO_CAPS_CODEC_IDX_JPEG, 16384, 16384, 0)}, {codec_info_build(AMDGPU_INFO_VIDEO_CAPS_CODEC_IDX_VP9, 8192, 4352, 0)}, {codec_info_build(AMDGPU_INFO_VIDEO_CAPS_CODEC_IDX_AV1, 8192, 4352, 0)}, }; static struct amdgpu_video_codec_info sriov_sc_video_codecs_decode_array_vcn1[] = { - {codec_info_build(AMDGPU_INFO_VIDEO_CAPS_CODEC_IDX_MPEG2, 4096, 4096, 3)}, - {codec_info_build(AMDGPU_INFO_VIDEO_CAPS_CODEC_IDX_MPEG4, 4096, 4096, 5)}, + {codec_info_build(AMDGPU_INFO_VIDEO_CAPS_CODEC_IDX_MPEG2, 1920, 1088, 3)}, + {codec_info_build(AMDGPU_INFO_VIDEO_CAPS_CODEC_IDX_MPEG4, 1920, 1088, 5)}, {codec_info_build(AMDGPU_INFO_VIDEO_CAPS_CODEC_IDX_MPEG4_AVC, 4096, 4096, 52)}, - {codec_info_build(AMDGPU_INFO_VIDEO_CAPS_CODEC_IDX_VC1, 4096, 4096, 4)}, + {codec_info_build(AMDGPU_INFO_VIDEO_CAPS_CODEC_IDX_VC1, 1920, 1088, 4)}, {codec_info_build(AMDGPU_INFO_VIDEO_CAPS_CODEC_IDX_HEVC, 8192, 4352, 186)}, - {codec_info_build(AMDGPU_INFO_VIDEO_CAPS_CODEC_IDX_JPEG, 4096, 4096, 0)}, + {codec_info_build(AMDGPU_INFO_VIDEO_CAPS_CODEC_IDX_JPEG, 16384, 16384, 0)}, {codec_info_build(AMDGPU_INFO_VIDEO_CAPS_CODEC_IDX_VP9, 8192, 4352, 0)}, }; diff --git a/drivers/gpu/drm/amd/amdgpu/soc21.c b/drivers/gpu/drm/amd/amdgpu/soc21.c index dd5d04c068f9..ad36c96478a8 100644 --- a/drivers/gpu/drm/amd/amdgpu/soc21.c +++ b/drivers/gpu/drm/amd/amdgpu/soc21.c @@ -117,23 +117,17 @@ static struct amdgpu_video_codecs sriov_vcn_4_0_0_video_codecs_encode_vcn1 = { }; static struct amdgpu_video_codec_info sriov_vcn_4_0_0_video_codecs_decode_array_vcn0[] = { - {codec_info_build(AMDGPU_INFO_VIDEO_CAPS_CODEC_IDX_MPEG2, 4096, 4096, 3)}, - {codec_info_build(AMDGPU_INFO_VIDEO_CAPS_CODEC_IDX_MPEG4, 4096, 4096, 5)}, {codec_info_build(AMDGPU_INFO_VIDEO_CAPS_CODEC_IDX_MPEG4_AVC, 4096, 4096, 52)}, - {codec_info_build(AMDGPU_INFO_VIDEO_CAPS_CODEC_IDX_VC1, 4096, 4096, 4)}, {codec_info_build(AMDGPU_INFO_VIDEO_CAPS_CODEC_IDX_HEVC, 8192, 4352, 186)}, - {codec_info_build(AMDGPU_INFO_VIDEO_CAPS_CODEC_IDX_JPEG, 4096, 4096, 0)}, + {codec_info_build(AMDGPU_INFO_VIDEO_CAPS_CODEC_IDX_JPEG, 16384, 16384, 0)}, {codec_info_build(AMDGPU_INFO_VIDEO_CAPS_CODEC_IDX_VP9, 8192, 4352, 0)}, {codec_info_build(AMDGPU_INFO_VIDEO_CAPS_CODEC_IDX_AV1, 8192, 4352, 0)}, }; static struct amdgpu_video_codec_info sriov_vcn_4_0_0_video_codecs_decode_array_vcn1[] = { - {codec_info_build(AMDGPU_INFO_VIDEO_CAPS_CODEC_IDX_MPEG2, 4096, 4096, 3)}, - {codec_info_build(AMDGPU_INFO_VIDEO_CAPS_CODEC_IDX_MPEG4, 4096, 4096, 5)}, {codec_info_build(AMDGPU_INFO_VIDEO_CAPS_CODEC_IDX_MPEG4_AVC, 4096, 4096, 52)}, - {codec_info_build(AMDGPU_INFO_VIDEO_CAPS_CODEC_IDX_VC1, 4096, 4096, 4)}, {codec_info_build(AMDGPU_INFO_VIDEO_CAPS_CODEC_IDX_HEVC, 8192, 4352, 186)}, - {codec_info_build(AMDGPU_INFO_VIDEO_CAPS_CODEC_IDX_JPEG, 4096, 4096, 0)}, + {codec_info_build(AMDGPU_INFO_VIDEO_CAPS_CODEC_IDX_JPEG, 16384, 16384, 0)}, {codec_info_build(AMDGPU_INFO_VIDEO_CAPS_CODEC_IDX_VP9, 8192, 4352, 0)}, }; -- 2.51.0 From 16fbc18cb07470cd33fb5f37ad181b51583e6dc0 Mon Sep 17 00:00:00 2001 From: Harish Kasiviswanathan Date: Tue, 11 Mar 2025 14:15:18 -0400 Subject: [PATCH 03/16] drm/amd/pm: add unique_id for gfx12 Expose unique_id for gfx12 Signed-off-by: Harish Kasiviswanathan Reviewed-by: Alex Deucher Signed-off-by: Alex Deucher --- drivers/gpu/drm/amd/pm/amdgpu_pm.c | 2 ++ 1 file changed, 2 insertions(+) diff --git a/drivers/gpu/drm/amd/pm/amdgpu_pm.c b/drivers/gpu/drm/amd/pm/amdgpu_pm.c index 1d04f1b79ded..2179344e78d9 100644 --- a/drivers/gpu/drm/amd/pm/amdgpu_pm.c +++ b/drivers/gpu/drm/amd/pm/amdgpu_pm.c @@ -2341,6 +2341,8 @@ static int default_attr_update(struct amdgpu_device *adev, struct amdgpu_device_ case IP_VERSION(11, 0, 1): case IP_VERSION(11, 0, 2): case IP_VERSION(11, 0, 3): + case IP_VERSION(12, 0, 0): + case IP_VERSION(12, 0, 1): *states = ATTR_STATE_SUPPORTED; break; default: -- 2.51.0 From b9e75bcb2b39e1202364d958ee4f27fd8a6f1313 Mon Sep 17 00:00:00 2001 From: Lijo Lazar Date: Thu, 6 Feb 2025 17:33:46 +0530 Subject: [PATCH 04/16] drm/amdgpu: Remove unsupported xgmi versions XGMI v4.8.0 is not used in any SOCs. Remove the associated functions. Also, ensure get_xgmi_info callback pointer is not NULL before calling the function. Signed-off-by: Lijo Lazar Reviewed-by: Hawking Zhang Signed-off-by: Alex Deucher --- drivers/gpu/drm/amd/amdgpu/amdgpu_discovery.c | 3 -- drivers/gpu/drm/amd/amdgpu/gfxhub_v2_1.c | 37 ------------------- 2 files changed, 40 deletions(-) diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_discovery.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_discovery.c index a4258127083d..967a992829bd 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_discovery.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_discovery.c @@ -2772,9 +2772,6 @@ int amdgpu_discovery_set_ip_blocks(struct amdgpu_device *adev) break; } - if (amdgpu_ip_version(adev, XGMI_HWIP, 0) == IP_VERSION(4, 8, 0)) - adev->gmc.xgmi.supported = true; - if (amdgpu_ip_version(adev, GC_HWIP, 0) == IP_VERSION(9, 4, 3) || amdgpu_ip_version(adev, GC_HWIP, 0) == IP_VERSION(9, 4, 4)) adev->ip_versions[XGMI_HWIP][0] = IP_VERSION(6, 4, 0); diff --git a/drivers/gpu/drm/amd/amdgpu/gfxhub_v2_1.c b/drivers/gpu/drm/amd/amdgpu/gfxhub_v2_1.c index 17509f32f61a..deb95fab02df 100644 --- a/drivers/gpu/drm/amd/amdgpu/gfxhub_v2_1.c +++ b/drivers/gpu/drm/amd/amdgpu/gfxhub_v2_1.c @@ -505,42 +505,6 @@ static void gfxhub_v2_1_init(struct amdgpu_device *adev) hub->vmhub_funcs = &gfxhub_v2_1_vmhub_funcs; } -static int gfxhub_v2_1_get_xgmi_info(struct amdgpu_device *adev) -{ - u32 xgmi_lfb_cntl = RREG32_SOC15(GC, 0, mmGCMC_VM_XGMI_LFB_CNTL); - u32 max_region = - REG_GET_FIELD(xgmi_lfb_cntl, GCMC_VM_XGMI_LFB_CNTL, PF_MAX_REGION); - u32 max_num_physical_nodes = 0; - u32 max_physical_node_id = 0; - - switch (amdgpu_ip_version(adev, XGMI_HWIP, 0)) { - case IP_VERSION(4, 8, 0): - max_num_physical_nodes = 4; - max_physical_node_id = 3; - break; - default: - return -EINVAL; - } - - /* PF_MAX_REGION=0 means xgmi is disabled */ - if (max_region) { - adev->gmc.xgmi.num_physical_nodes = max_region + 1; - if (adev->gmc.xgmi.num_physical_nodes > max_num_physical_nodes) - return -EINVAL; - - adev->gmc.xgmi.physical_node_id = - REG_GET_FIELD(xgmi_lfb_cntl, GCMC_VM_XGMI_LFB_CNTL, PF_LFB_REGION); - if (adev->gmc.xgmi.physical_node_id > max_physical_node_id) - return -EINVAL; - - adev->gmc.xgmi.node_segment_size = REG_GET_FIELD( - RREG32_SOC15(GC, 0, mmGCMC_VM_XGMI_LFB_SIZE), - GCMC_VM_XGMI_LFB_SIZE, PF_LFB_SIZE) << 24; - } - - return 0; -} - static void gfxhub_v2_1_utcl2_harvest(struct amdgpu_device *adev) { int i; @@ -696,7 +660,6 @@ const struct amdgpu_gfxhub_funcs gfxhub_v2_1_funcs = { .gart_disable = gfxhub_v2_1_gart_disable, .set_fault_enable_default = gfxhub_v2_1_set_fault_enable_default, .init = gfxhub_v2_1_init, - .get_xgmi_info = gfxhub_v2_1_get_xgmi_info, .utcl2_harvest = gfxhub_v2_1_utcl2_harvest, .mode2_save_regs = gfxhub_v2_1_save_regs, .mode2_restore_regs = gfxhub_v2_1_restore_regs, -- 2.51.0 From f844732e3ad9c4b78df7436232949b8d2096d1a6 Mon Sep 17 00:00:00 2001 From: Emily Deng Date: Thu, 6 Mar 2025 08:40:01 +0800 Subject: [PATCH 05/16] drm/amdgpu: Fix the race condition for draining retry fault Issue: In the scenario where svm_range_restore_pages is called, but svm->checkpoint_ts has not been set and the retry fault has not been drained, svm_range_unmap_from_cpu is triggered and calls svm_range_free. Meanwhile, svm_range_restore_pages continues execution and reaches svm_range_from_addr. This results in a "failed to find prange..." error, causing the page recovery to fail. How to fix: Move the timestamp check code under the protection of svm->lock. v2: Make sure all right locks are released before go out. v3: Directly goto out_unlock_svms, and return -EAGAIN. v4: Refine code. Signed-off-by: Emily Deng Reviewed-by: Felix Kuehling Signed-off-by: Alex Deucher --- drivers/gpu/drm/amd/amdkfd/kfd_svm.c | 31 +++++++++++++++------------- 1 file changed, 17 insertions(+), 14 deletions(-) diff --git a/drivers/gpu/drm/amd/amdkfd/kfd_svm.c b/drivers/gpu/drm/amd/amdkfd/kfd_svm.c index 1a38ac75abbd..63eeef2a75ba 100644 --- a/drivers/gpu/drm/amd/amdkfd/kfd_svm.c +++ b/drivers/gpu/drm/amd/amdkfd/kfd_svm.c @@ -3009,19 +3009,6 @@ svm_range_restore_pages(struct amdgpu_device *adev, unsigned int pasid, goto out; } - /* check if this page fault time stamp is before svms->checkpoint_ts */ - if (svms->checkpoint_ts[gpuidx] != 0) { - if (amdgpu_ih_ts_after_or_equal(ts, svms->checkpoint_ts[gpuidx])) { - pr_debug("draining retry fault, drop fault 0x%llx\n", addr); - r = 0; - goto out; - } else - /* ts is after svms->checkpoint_ts now, reset svms->checkpoint_ts - * to zero to avoid following ts wrap around give wrong comparing - */ - svms->checkpoint_ts[gpuidx] = 0; - } - if (!p->xnack_enabled) { pr_debug("XNACK not enabled for pasid 0x%x\n", pasid); r = -EFAULT; @@ -3041,6 +3028,21 @@ svm_range_restore_pages(struct amdgpu_device *adev, unsigned int pasid, mmap_read_lock(mm); retry_write_locked: mutex_lock(&svms->lock); + + /* check if this page fault time stamp is before svms->checkpoint_ts */ + if (svms->checkpoint_ts[gpuidx] != 0) { + if (amdgpu_ih_ts_after_or_equal(ts, svms->checkpoint_ts[gpuidx])) { + pr_debug("draining retry fault, drop fault 0x%llx\n", addr); + r = -EAGAIN; + goto out_unlock_svms; + } else { + /* ts is after svms->checkpoint_ts now, reset svms->checkpoint_ts + * to zero to avoid following ts wrap around give wrong comparing + */ + svms->checkpoint_ts[gpuidx] = 0; + } + } + prange = svm_range_from_addr(svms, addr, NULL); if (!prange) { pr_debug("failed to find prange svms 0x%p address [0x%llx]\n", @@ -3166,7 +3168,8 @@ out_unlock_svms: mutex_unlock(&svms->lock); mmap_read_unlock(mm); - svm_range_count_fault(node, p, gpuidx); + if (r != -EAGAIN) + svm_range_count_fault(node, p, gpuidx); mmput(mm); out: -- 2.51.0 From e27b36ea6ba5f29e91fcfb375ea29503708fcf43 Mon Sep 17 00:00:00 2001 From: Alex Deucher Date: Wed, 26 Feb 2025 15:55:33 -0500 Subject: [PATCH 06/16] drm/amdgpu/gfx11: don't read registers in mqd init Just use the default values. There's not need to get the value from hardware and it could cause problems if we do that at runtime and gfxoff is active. Reviewed-by: Mukul Joshi Signed-off-by: Alex Deucher --- drivers/gpu/drm/amd/amdgpu/gfx_v11_0.c | 47 ++++++++++++++++++-------- 1 file changed, 32 insertions(+), 15 deletions(-) diff --git a/drivers/gpu/drm/amd/amdgpu/gfx_v11_0.c b/drivers/gpu/drm/amd/amdgpu/gfx_v11_0.c index 2b3ba404955d..b9f6d89dafb2 100644 --- a/drivers/gpu/drm/amd/amdgpu/gfx_v11_0.c +++ b/drivers/gpu/drm/amd/amdgpu/gfx_v11_0.c @@ -62,6 +62,23 @@ #define regPC_CONFIG_CNTL_1 0x194d #define regPC_CONFIG_CNTL_1_BASE_IDX 1 +#define regCP_GFX_MQD_CONTROL_DEFAULT 0x00000100 +#define regCP_GFX_HQD_VMID_DEFAULT 0x00000000 +#define regCP_GFX_HQD_QUEUE_PRIORITY_DEFAULT 0x00000000 +#define regCP_GFX_HQD_QUANTUM_DEFAULT 0x00000a01 +#define regCP_GFX_HQD_CNTL_DEFAULT 0x00a00000 +#define regCP_RB_DOORBELL_CONTROL_DEFAULT 0x00000000 +#define regCP_GFX_HQD_RPTR_DEFAULT 0x00000000 + +#define regCP_HQD_EOP_CONTROL_DEFAULT 0x00000006 +#define regCP_HQD_PQ_DOORBELL_CONTROL_DEFAULT 0x00000000 +#define regCP_MQD_CONTROL_DEFAULT 0x00000100 +#define regCP_HQD_PQ_CONTROL_DEFAULT 0x00308509 +#define regCP_HQD_PQ_DOORBELL_CONTROL_DEFAULT 0x00000000 +#define regCP_HQD_PQ_RPTR_DEFAULT 0x00000000 +#define regCP_HQD_PERSISTENT_STATE_DEFAULT 0x0be05501 +#define regCP_HQD_IB_CONTROL_DEFAULT 0x00300000 + MODULE_FIRMWARE("amdgpu/gc_11_0_0_pfp.bin"); MODULE_FIRMWARE("amdgpu/gc_11_0_0_me.bin"); MODULE_FIRMWARE("amdgpu/gc_11_0_0_mec.bin"); @@ -3965,7 +3982,7 @@ static void gfx_v11_0_gfx_mqd_set_priority(struct amdgpu_device *adev, if (prop->hqd_pipe_priority == AMDGPU_GFX_PIPE_PRIO_HIGH) priority = 1; - tmp = RREG32_SOC15(GC, 0, regCP_GFX_HQD_QUEUE_PRIORITY); + tmp = regCP_GFX_HQD_QUEUE_PRIORITY_DEFAULT; tmp = REG_SET_FIELD(tmp, CP_GFX_HQD_QUEUE_PRIORITY, PRIORITY_LEVEL, priority); mqd->cp_gfx_hqd_queue_priority = tmp; } @@ -3987,14 +4004,14 @@ static int gfx_v11_0_gfx_mqd_init(struct amdgpu_device *adev, void *m, mqd->cp_mqd_base_addr_hi = upper_32_bits(prop->mqd_gpu_addr); /* set up mqd control */ - tmp = RREG32_SOC15(GC, 0, regCP_GFX_MQD_CONTROL); + tmp = regCP_GFX_MQD_CONTROL_DEFAULT; tmp = REG_SET_FIELD(tmp, CP_GFX_MQD_CONTROL, VMID, 0); tmp = REG_SET_FIELD(tmp, CP_GFX_MQD_CONTROL, PRIV_STATE, 1); tmp = REG_SET_FIELD(tmp, CP_GFX_MQD_CONTROL, CACHE_POLICY, 0); mqd->cp_gfx_mqd_control = tmp; /* set up gfx_hqd_vimd with 0x0 to indicate the ring buffer's vmid */ - tmp = RREG32_SOC15(GC, 0, regCP_GFX_HQD_VMID); + tmp = regCP_GFX_HQD_VMID_DEFAULT; tmp = REG_SET_FIELD(tmp, CP_GFX_HQD_VMID, VMID, 0); mqd->cp_gfx_hqd_vmid = 0; @@ -4002,7 +4019,7 @@ static int gfx_v11_0_gfx_mqd_init(struct amdgpu_device *adev, void *m, gfx_v11_0_gfx_mqd_set_priority(adev, mqd, prop); /* set up time quantum */ - tmp = RREG32_SOC15(GC, 0, regCP_GFX_HQD_QUANTUM); + tmp = regCP_GFX_HQD_QUANTUM_DEFAULT; tmp = REG_SET_FIELD(tmp, CP_GFX_HQD_QUANTUM, QUANTUM_EN, 1); mqd->cp_gfx_hqd_quantum = tmp; @@ -4024,7 +4041,7 @@ static int gfx_v11_0_gfx_mqd_init(struct amdgpu_device *adev, void *m, /* set up the gfx_hqd_control, similar as CP_RB0_CNTL */ rb_bufsz = order_base_2(prop->queue_size / 4) - 1; - tmp = RREG32_SOC15(GC, 0, regCP_GFX_HQD_CNTL); + tmp = regCP_GFX_HQD_CNTL_DEFAULT; tmp = REG_SET_FIELD(tmp, CP_GFX_HQD_CNTL, RB_BUFSZ, rb_bufsz); tmp = REG_SET_FIELD(tmp, CP_GFX_HQD_CNTL, RB_BLKSZ, rb_bufsz - 2); #ifdef __BIG_ENDIAN @@ -4033,7 +4050,7 @@ static int gfx_v11_0_gfx_mqd_init(struct amdgpu_device *adev, void *m, mqd->cp_gfx_hqd_cntl = tmp; /* set up cp_doorbell_control */ - tmp = RREG32_SOC15(GC, 0, regCP_RB_DOORBELL_CONTROL); + tmp = regCP_RB_DOORBELL_CONTROL_DEFAULT; if (prop->use_doorbell) { tmp = REG_SET_FIELD(tmp, CP_RB_DOORBELL_CONTROL, DOORBELL_OFFSET, prop->doorbell_index); @@ -4045,7 +4062,7 @@ static int gfx_v11_0_gfx_mqd_init(struct amdgpu_device *adev, void *m, mqd->cp_rb_doorbell_control = tmp; /* reset read and write pointers, similar to CP_RB0_WPTR/_RPTR */ - mqd->cp_gfx_hqd_rptr = RREG32_SOC15(GC, 0, regCP_GFX_HQD_RPTR); + mqd->cp_gfx_hqd_rptr = regCP_GFX_HQD_RPTR_DEFAULT; /* active the queue */ mqd->cp_gfx_hqd_active = 1; @@ -4131,14 +4148,14 @@ static int gfx_v11_0_compute_mqd_init(struct amdgpu_device *adev, void *m, mqd->cp_hqd_eop_base_addr_hi = upper_32_bits(eop_base_addr); /* set the EOP size, register value is 2^(EOP_SIZE+1) dwords */ - tmp = RREG32_SOC15(GC, 0, regCP_HQD_EOP_CONTROL); + tmp = regCP_HQD_EOP_CONTROL_DEFAULT; tmp = REG_SET_FIELD(tmp, CP_HQD_EOP_CONTROL, EOP_SIZE, (order_base_2(GFX11_MEC_HPD_SIZE / 4) - 1)); mqd->cp_hqd_eop_control = tmp; /* enable doorbell? */ - tmp = RREG32_SOC15(GC, 0, regCP_HQD_PQ_DOORBELL_CONTROL); + tmp = regCP_HQD_PQ_DOORBELL_CONTROL_DEFAULT; if (prop->use_doorbell) { tmp = REG_SET_FIELD(tmp, CP_HQD_PQ_DOORBELL_CONTROL, @@ -4167,7 +4184,7 @@ static int gfx_v11_0_compute_mqd_init(struct amdgpu_device *adev, void *m, mqd->cp_mqd_base_addr_hi = upper_32_bits(prop->mqd_gpu_addr); /* set MQD vmid to 0 */ - tmp = RREG32_SOC15(GC, 0, regCP_MQD_CONTROL); + tmp = regCP_MQD_CONTROL_DEFAULT; tmp = REG_SET_FIELD(tmp, CP_MQD_CONTROL, VMID, 0); mqd->cp_mqd_control = tmp; @@ -4177,7 +4194,7 @@ static int gfx_v11_0_compute_mqd_init(struct amdgpu_device *adev, void *m, mqd->cp_hqd_pq_base_hi = upper_32_bits(hqd_gpu_addr); /* set up the HQD, this is similar to CP_RB0_CNTL */ - tmp = RREG32_SOC15(GC, 0, regCP_HQD_PQ_CONTROL); + tmp = regCP_HQD_PQ_CONTROL_DEFAULT; tmp = REG_SET_FIELD(tmp, CP_HQD_PQ_CONTROL, QUEUE_SIZE, (order_base_2(prop->queue_size / 4) - 1)); tmp = REG_SET_FIELD(tmp, CP_HQD_PQ_CONTROL, RPTR_BLOCK_SIZE, @@ -4203,7 +4220,7 @@ static int gfx_v11_0_compute_mqd_init(struct amdgpu_device *adev, void *m, tmp = 0; /* enable the doorbell if requested */ if (prop->use_doorbell) { - tmp = RREG32_SOC15(GC, 0, regCP_HQD_PQ_DOORBELL_CONTROL); + tmp = regCP_HQD_PQ_DOORBELL_CONTROL_DEFAULT; tmp = REG_SET_FIELD(tmp, CP_HQD_PQ_DOORBELL_CONTROL, DOORBELL_OFFSET, prop->doorbell_index); @@ -4218,17 +4235,17 @@ static int gfx_v11_0_compute_mqd_init(struct amdgpu_device *adev, void *m, mqd->cp_hqd_pq_doorbell_control = tmp; /* reset read and write pointers, similar to CP_RB0_WPTR/_RPTR */ - mqd->cp_hqd_pq_rptr = RREG32_SOC15(GC, 0, regCP_HQD_PQ_RPTR); + mqd->cp_hqd_pq_rptr = regCP_HQD_PQ_RPTR_DEFAULT; /* set the vmid for the queue */ mqd->cp_hqd_vmid = 0; - tmp = RREG32_SOC15(GC, 0, regCP_HQD_PERSISTENT_STATE); + tmp = regCP_HQD_PERSISTENT_STATE_DEFAULT; tmp = REG_SET_FIELD(tmp, CP_HQD_PERSISTENT_STATE, PRELOAD_SIZE, 0x55); mqd->cp_hqd_persistent_state = tmp; /* set MIN_IB_AVAIL_SIZE */ - tmp = RREG32_SOC15(GC, 0, regCP_HQD_IB_CONTROL); + tmp = regCP_HQD_IB_CONTROL_DEFAULT; tmp = REG_SET_FIELD(tmp, CP_HQD_IB_CONTROL, MIN_IB_AVAIL_SIZE, 3); mqd->cp_hqd_ib_control = tmp; -- 2.51.0 From fc3c139cf0432b79fd08e23100a559ee51cd0be4 Mon Sep 17 00:00:00 2001 From: Alex Deucher Date: Wed, 26 Feb 2025 16:08:03 -0500 Subject: [PATCH 07/16] drm/amdgpu/gfx12: don't read registers in mqd init Just use the default values. There's not need to get the value from hardware and it could cause problems if we do that at runtime and gfxoff is active. Reviewed-by: Mukul Joshi Signed-off-by: Alex Deucher --- drivers/gpu/drm/amd/amdgpu/gfx_v12_0.c | 48 ++++++++++++++++++-------- 1 file changed, 33 insertions(+), 15 deletions(-) diff --git a/drivers/gpu/drm/amd/amdgpu/gfx_v12_0.c b/drivers/gpu/drm/amd/amdgpu/gfx_v12_0.c index 926fb536bbff..87dc4195192c 100644 --- a/drivers/gpu/drm/amd/amdgpu/gfx_v12_0.c +++ b/drivers/gpu/drm/amd/amdgpu/gfx_v12_0.c @@ -50,6 +50,24 @@ #define RLCG_UCODE_LOADING_START_ADDRESS 0x00002000L +#define regCP_GFX_MQD_CONTROL_DEFAULT 0x00000100 +#define regCP_GFX_HQD_VMID_DEFAULT 0x00000000 +#define regCP_GFX_HQD_QUEUE_PRIORITY_DEFAULT 0x00000000 +#define regCP_GFX_HQD_QUANTUM_DEFAULT 0x00000a01 +#define regCP_GFX_HQD_CNTL_DEFAULT 0x00f00000 +#define regCP_RB_DOORBELL_CONTROL_DEFAULT 0x00000000 +#define regCP_GFX_HQD_RPTR_DEFAULT 0x00000000 + +#define regCP_HQD_EOP_CONTROL_DEFAULT 0x00000006 +#define regCP_HQD_PQ_DOORBELL_CONTROL_DEFAULT 0x00000000 +#define regCP_MQD_CONTROL_DEFAULT 0x00000100 +#define regCP_HQD_PQ_CONTROL_DEFAULT 0x00308509 +#define regCP_HQD_PQ_DOORBELL_CONTROL_DEFAULT 0x00000000 +#define regCP_HQD_PQ_RPTR_DEFAULT 0x00000000 +#define regCP_HQD_PERSISTENT_STATE_DEFAULT 0x0be05501 +#define regCP_HQD_IB_CONTROL_DEFAULT 0x00300000 + + MODULE_FIRMWARE("amdgpu/gc_12_0_0_pfp.bin"); MODULE_FIRMWARE("amdgpu/gc_12_0_0_me.bin"); MODULE_FIRMWARE("amdgpu/gc_12_0_0_mec.bin"); @@ -2891,25 +2909,25 @@ static int gfx_v12_0_gfx_mqd_init(struct amdgpu_device *adev, void *m, mqd->cp_mqd_base_addr_hi = upper_32_bits(prop->mqd_gpu_addr); /* set up mqd control */ - tmp = RREG32_SOC15(GC, 0, regCP_GFX_MQD_CONTROL); + tmp = regCP_GFX_MQD_CONTROL_DEFAULT; tmp = REG_SET_FIELD(tmp, CP_GFX_MQD_CONTROL, VMID, 0); tmp = REG_SET_FIELD(tmp, CP_GFX_MQD_CONTROL, PRIV_STATE, 1); tmp = REG_SET_FIELD(tmp, CP_GFX_MQD_CONTROL, CACHE_POLICY, 0); mqd->cp_gfx_mqd_control = tmp; /* set up gfx_hqd_vimd with 0x0 to indicate the ring buffer's vmid */ - tmp = RREG32_SOC15(GC, 0, regCP_GFX_HQD_VMID); + tmp = regCP_GFX_HQD_VMID_DEFAULT; tmp = REG_SET_FIELD(tmp, CP_GFX_HQD_VMID, VMID, 0); mqd->cp_gfx_hqd_vmid = 0; /* set up default queue priority level * 0x0 = low priority, 0x1 = high priority */ - tmp = RREG32_SOC15(GC, 0, regCP_GFX_HQD_QUEUE_PRIORITY); + tmp = regCP_GFX_HQD_QUEUE_PRIORITY_DEFAULT; tmp = REG_SET_FIELD(tmp, CP_GFX_HQD_QUEUE_PRIORITY, PRIORITY_LEVEL, 0); mqd->cp_gfx_hqd_queue_priority = tmp; /* set up time quantum */ - tmp = RREG32_SOC15(GC, 0, regCP_GFX_HQD_QUANTUM); + tmp = regCP_GFX_HQD_QUANTUM_DEFAULT; tmp = REG_SET_FIELD(tmp, CP_GFX_HQD_QUANTUM, QUANTUM_EN, 1); mqd->cp_gfx_hqd_quantum = tmp; @@ -2931,7 +2949,7 @@ static int gfx_v12_0_gfx_mqd_init(struct amdgpu_device *adev, void *m, /* set up the gfx_hqd_control, similar as CP_RB0_CNTL */ rb_bufsz = order_base_2(prop->queue_size / 4) - 1; - tmp = RREG32_SOC15(GC, 0, regCP_GFX_HQD_CNTL); + tmp = regCP_GFX_HQD_CNTL_DEFAULT; tmp = REG_SET_FIELD(tmp, CP_GFX_HQD_CNTL, RB_BUFSZ, rb_bufsz); tmp = REG_SET_FIELD(tmp, CP_GFX_HQD_CNTL, RB_BLKSZ, rb_bufsz - 2); #ifdef __BIG_ENDIAN @@ -2940,7 +2958,7 @@ static int gfx_v12_0_gfx_mqd_init(struct amdgpu_device *adev, void *m, mqd->cp_gfx_hqd_cntl = tmp; /* set up cp_doorbell_control */ - tmp = RREG32_SOC15(GC, 0, regCP_RB_DOORBELL_CONTROL); + tmp = regCP_RB_DOORBELL_CONTROL_DEFAULT; if (prop->use_doorbell) { tmp = REG_SET_FIELD(tmp, CP_RB_DOORBELL_CONTROL, DOORBELL_OFFSET, prop->doorbell_index); @@ -2952,7 +2970,7 @@ static int gfx_v12_0_gfx_mqd_init(struct amdgpu_device *adev, void *m, mqd->cp_rb_doorbell_control = tmp; /* reset read and write pointers, similar to CP_RB0_WPTR/_RPTR */ - mqd->cp_gfx_hqd_rptr = RREG32_SOC15(GC, 0, regCP_GFX_HQD_RPTR); + mqd->cp_gfx_hqd_rptr = regCP_GFX_HQD_RPTR_DEFAULT; /* active the queue */ mqd->cp_gfx_hqd_active = 1; @@ -3047,14 +3065,14 @@ static int gfx_v12_0_compute_mqd_init(struct amdgpu_device *adev, void *m, mqd->cp_hqd_eop_base_addr_hi = upper_32_bits(eop_base_addr); /* set the EOP size, register value is 2^(EOP_SIZE+1) dwords */ - tmp = RREG32_SOC15(GC, 0, regCP_HQD_EOP_CONTROL); + tmp = regCP_HQD_EOP_CONTROL_DEFAULT; tmp = REG_SET_FIELD(tmp, CP_HQD_EOP_CONTROL, EOP_SIZE, (order_base_2(GFX12_MEC_HPD_SIZE / 4) - 1)); mqd->cp_hqd_eop_control = tmp; /* enable doorbell? */ - tmp = RREG32_SOC15(GC, 0, regCP_HQD_PQ_DOORBELL_CONTROL); + tmp = regCP_HQD_PQ_DOORBELL_CONTROL_DEFAULT; if (prop->use_doorbell) { tmp = REG_SET_FIELD(tmp, CP_HQD_PQ_DOORBELL_CONTROL, @@ -3083,7 +3101,7 @@ static int gfx_v12_0_compute_mqd_init(struct amdgpu_device *adev, void *m, mqd->cp_mqd_base_addr_hi = upper_32_bits(prop->mqd_gpu_addr); /* set MQD vmid to 0 */ - tmp = RREG32_SOC15(GC, 0, regCP_MQD_CONTROL); + tmp = regCP_MQD_CONTROL_DEFAULT; tmp = REG_SET_FIELD(tmp, CP_MQD_CONTROL, VMID, 0); mqd->cp_mqd_control = tmp; @@ -3093,7 +3111,7 @@ static int gfx_v12_0_compute_mqd_init(struct amdgpu_device *adev, void *m, mqd->cp_hqd_pq_base_hi = upper_32_bits(hqd_gpu_addr); /* set up the HQD, this is similar to CP_RB0_CNTL */ - tmp = RREG32_SOC15(GC, 0, regCP_HQD_PQ_CONTROL); + tmp = regCP_HQD_PQ_CONTROL_DEFAULT; tmp = REG_SET_FIELD(tmp, CP_HQD_PQ_CONTROL, QUEUE_SIZE, (order_base_2(prop->queue_size / 4) - 1)); tmp = REG_SET_FIELD(tmp, CP_HQD_PQ_CONTROL, RPTR_BLOCK_SIZE, @@ -3118,7 +3136,7 @@ static int gfx_v12_0_compute_mqd_init(struct amdgpu_device *adev, void *m, tmp = 0; /* enable the doorbell if requested */ if (prop->use_doorbell) { - tmp = RREG32_SOC15(GC, 0, regCP_HQD_PQ_DOORBELL_CONTROL); + tmp = regCP_HQD_PQ_DOORBELL_CONTROL_DEFAULT; tmp = REG_SET_FIELD(tmp, CP_HQD_PQ_DOORBELL_CONTROL, DOORBELL_OFFSET, prop->doorbell_index); @@ -3133,17 +3151,17 @@ static int gfx_v12_0_compute_mqd_init(struct amdgpu_device *adev, void *m, mqd->cp_hqd_pq_doorbell_control = tmp; /* reset read and write pointers, similar to CP_RB0_WPTR/_RPTR */ - mqd->cp_hqd_pq_rptr = RREG32_SOC15(GC, 0, regCP_HQD_PQ_RPTR); + mqd->cp_hqd_pq_rptr = regCP_HQD_PQ_RPTR_DEFAULT; /* set the vmid for the queue */ mqd->cp_hqd_vmid = 0; - tmp = RREG32_SOC15(GC, 0, regCP_HQD_PERSISTENT_STATE); + tmp = regCP_HQD_PERSISTENT_STATE_DEFAULT; tmp = REG_SET_FIELD(tmp, CP_HQD_PERSISTENT_STATE, PRELOAD_SIZE, 0x55); mqd->cp_hqd_persistent_state = tmp; /* set MIN_IB_AVAIL_SIZE */ - tmp = RREG32_SOC15(GC, 0, regCP_HQD_IB_CONTROL); + tmp = regCP_HQD_IB_CONTROL_DEFAULT; tmp = REG_SET_FIELD(tmp, CP_HQD_IB_CONTROL, MIN_IB_AVAIL_SIZE, 3); mqd->cp_hqd_ib_control = tmp; -- 2.51.0 From 02fc2f3c468d560c1daa17a55a88f33dc238f7ae Mon Sep 17 00:00:00 2001 From: Asad Kamal Date: Tue, 11 Mar 2025 18:17:33 +0800 Subject: [PATCH 08/16] drm/amd/pm: Update feature list for smu_v13_0_12 Update feature list for smu_v13_0_12 to show vcn & smu deep sleep feature enable status. Signed-off-by: Asad Kamal Reviewed-by: Hawking Zhang Reviewed-by: Lijo Lazar Signed-off-by: Alex Deucher --- drivers/gpu/drm/amd/pm/swsmu/inc/smu_types.h | 4 +++- drivers/gpu/drm/amd/pm/swsmu/smu13/smu_v13_0_12_ppt.c | 5 +++++ 2 files changed, 8 insertions(+), 1 deletion(-) diff --git a/drivers/gpu/drm/amd/pm/swsmu/inc/smu_types.h b/drivers/gpu/drm/amd/pm/swsmu/inc/smu_types.h index 9c8468fb203a..c9dee09395e3 100644 --- a/drivers/gpu/drm/amd/pm/swsmu/inc/smu_types.h +++ b/drivers/gpu/drm/amd/pm/swsmu/inc/smu_types.h @@ -356,6 +356,7 @@ enum smu_clk_type { __SMU_DUMMY_MAP(DS_FCLK), \ __SMU_DUMMY_MAP(DS_MP1CLK), \ __SMU_DUMMY_MAP(DS_MP0CLK), \ + __SMU_DUMMY_MAP(DS_MPIOCLK), \ __SMU_DUMMY_MAP(XGMI_PER_LINK_PWR_DWN), \ __SMU_DUMMY_MAP(DPM_GFX_PACE), \ __SMU_DUMMY_MAP(MEM_VDDCI_SCALING), \ @@ -452,7 +453,8 @@ enum smu_clk_type { __SMU_DUMMY_MAP(APT_PF_DCS), \ __SMU_DUMMY_MAP(GFX_EDC_XVMIN), \ __SMU_DUMMY_MAP(GFX_DIDT_XVMIN), \ - __SMU_DUMMY_MAP(FAN_ABNORMAL), + __SMU_DUMMY_MAP(FAN_ABNORMAL), \ + __SMU_DUMMY_MAP(PIT), #undef __SMU_DUMMY_MAP #define __SMU_DUMMY_MAP(feature) SMU_FEATURE_##feature##_BIT diff --git a/drivers/gpu/drm/amd/pm/swsmu/smu13/smu_v13_0_12_ppt.c b/drivers/gpu/drm/amd/pm/swsmu/smu13/smu_v13_0_12_ppt.c index 285dbfe10303..51078accc0f8 100644 --- a/drivers/gpu/drm/amd/pm/swsmu/smu13/smu_v13_0_12_ppt.c +++ b/drivers/gpu/drm/amd/pm/swsmu/smu13/smu_v13_0_12_ppt.c @@ -76,6 +76,11 @@ const struct cmn2asic_mapping smu_v13_0_12_feature_mask_map[SMU_FEATURE_COUNT] = SMU_13_0_12_FEA_MAP(SMU_FEATURE_THERMAL_BIT, FEATURE_THERMAL), SMU_13_0_12_FEA_MAP(SMU_FEATURE_SOC_PCC_BIT, FEATURE_SOC_PCC), SMU_13_0_12_FEA_MAP(SMU_FEATURE_XGMI_PER_LINK_PWR_DWN_BIT, FEATURE_XGMI_PER_LINK_PWR_DOWN), + SMU_13_0_12_FEA_MAP(SMU_FEATURE_DS_VCN_BIT, FEATURE_DS_VCN), + SMU_13_0_12_FEA_MAP(SMU_FEATURE_DS_MP1CLK_BIT, FEATURE_DS_MP1CLK), + SMU_13_0_12_FEA_MAP(SMU_FEATURE_DS_MPIOCLK_BIT, FEATURE_DS_MPIOCLK), + SMU_13_0_12_FEA_MAP(SMU_FEATURE_DS_MP0CLK_BIT, FEATURE_DS_MP0CLK), + SMU_13_0_12_FEA_MAP(SMU_FEATURE_PIT_BIT, FEATURE_PIT), }; // clang-format off -- 2.51.0 From 8a7820c07224bdae1ba704f295f9654cc2ed6691 Mon Sep 17 00:00:00 2001 From: Harish Kasiviswanathan Date: Tue, 25 Feb 2025 15:50:30 -0500 Subject: [PATCH 09/16] drm/amdgpu: Reduce dequeue retry timeout for gfx9 family Dequeue retry timeout controls the interval between checks for unmet conditions. On MI series, reduce this from 0x40 to 0x1 (~ 1 uS). The cost of additional bandwidth consumed by CP when polling memory shouldn't be substantial. Signed-off-by: Harish Kasiviswanathan Reviewed-by: Jonathan Kim Signed-off-by: Alex Deucher --- .../drm/amd/amdgpu/amdgpu_amdkfd_aldebaran.c | 2 +- .../drm/amd/amdgpu/amdgpu_amdkfd_arcturus.c | 2 +- .../drm/amd/amdgpu/amdgpu_amdkfd_gc_9_4_3.c | 4 +- .../drm/amd/amdgpu/amdgpu_amdkfd_gfx_v10.c | 28 ++++++------ .../drm/amd/amdgpu/amdgpu_amdkfd_gfx_v10.h | 5 ++- .../drm/amd/amdgpu/amdgpu_amdkfd_gfx_v10_3.c | 2 +- .../gpu/drm/amd/amdgpu/amdgpu_amdkfd_gfx_v9.c | 28 ++++++------ .../gpu/drm/amd/amdgpu/amdgpu_amdkfd_gfx_v9.h | 5 ++- .../drm/amd/amdkfd/kfd_packet_manager_v9.c | 43 +++++++++++++------ .../gpu/drm/amd/include/kgd_kfd_interface.h | 5 ++- 10 files changed, 72 insertions(+), 52 deletions(-) diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_aldebaran.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_aldebaran.c index 6e861d08d044..7e9f7a280c1b 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_aldebaran.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_aldebaran.c @@ -189,7 +189,7 @@ const struct kfd2kgd_calls aldebaran_kfd2kgd = { .set_address_watch = kgd_gfx_aldebaran_set_address_watch, .clear_address_watch = kgd_gfx_v9_clear_address_watch, .get_iq_wait_times = kgd_gfx_v9_get_iq_wait_times, - .build_grace_period_packet_info = kgd_gfx_v9_build_grace_period_packet_info, + .build_dequeue_wait_counts_packet_info = kgd_gfx_v9_build_dequeue_wait_counts_packet_info, .program_trap_handler_settings = kgd_gfx_v9_program_trap_handler_settings, .hqd_get_pq_addr = kgd_gfx_v9_hqd_get_pq_addr, .hqd_reset = kgd_gfx_v9_hqd_reset, diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_arcturus.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_arcturus.c index c820418e8ccd..ffbaa8bc5eea 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_arcturus.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_arcturus.c @@ -415,7 +415,7 @@ const struct kfd2kgd_calls arcturus_kfd2kgd = { .set_address_watch = kgd_gfx_v9_set_address_watch, .clear_address_watch = kgd_gfx_v9_clear_address_watch, .get_iq_wait_times = kgd_gfx_v9_get_iq_wait_times, - .build_grace_period_packet_info = kgd_gfx_v9_build_grace_period_packet_info, + .build_dequeue_wait_counts_packet_info = kgd_gfx_v9_build_dequeue_wait_counts_packet_info, .get_cu_occupancy = kgd_gfx_v9_get_cu_occupancy, .program_trap_handler_settings = kgd_gfx_v9_program_trap_handler_settings, .hqd_get_pq_addr = kgd_gfx_v9_hqd_get_pq_addr, diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gc_9_4_3.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gc_9_4_3.c index 0c0998477598..89a45a9218f3 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gc_9_4_3.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gc_9_4_3.c @@ -541,8 +541,8 @@ const struct kfd2kgd_calls gc_9_4_3_kfd2kgd = { .get_cu_occupancy = kgd_gfx_v9_get_cu_occupancy, .program_trap_handler_settings = kgd_gfx_v9_program_trap_handler_settings, - .build_grace_period_packet_info = - kgd_gfx_v9_build_grace_period_packet_info, + .build_dequeue_wait_counts_packet_info = + kgd_gfx_v9_build_dequeue_wait_counts_packet_info, .get_iq_wait_times = kgd_gfx_v9_get_iq_wait_times, .enable_debug_trap = kgd_aldebaran_enable_debug_trap, .disable_debug_trap = kgd_gfx_v9_4_3_disable_debug_trap, diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gfx_v10.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gfx_v10.c index 2887b6f3eaa2..04ef0ca10541 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gfx_v10.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gfx_v10.c @@ -1021,25 +1021,25 @@ void kgd_gfx_v10_get_iq_wait_times(struct amdgpu_device *adev, *wait_times = RREG32(SOC15_REG_OFFSET(GC, 0, mmCP_IQ_WAIT_TIME2)); } -void kgd_gfx_v10_build_grace_period_packet_info(struct amdgpu_device *adev, +void kgd_gfx_v10_build_dequeue_wait_counts_packet_info(struct amdgpu_device *adev, uint32_t wait_times, - uint32_t grace_period, + uint32_t sch_wave, + uint32_t que_sleep, uint32_t *reg_offset, uint32_t *reg_data) { *reg_data = wait_times; - /* - * The CP cannont handle a 0 grace period input and will result in - * an infinite grace period being set so set to 1 to prevent this. - */ - if (grace_period == 0) - grace_period = 1; - - *reg_data = REG_SET_FIELD(*reg_data, - CP_IQ_WAIT_TIME2, - SCH_WAVE, - grace_period); + if (sch_wave) + *reg_data = REG_SET_FIELD(*reg_data, + CP_IQ_WAIT_TIME2, + SCH_WAVE, + sch_wave); + if (que_sleep) + *reg_data = REG_SET_FIELD(*reg_data, + CP_IQ_WAIT_TIME2, + QUE_SLEEP, + que_sleep); *reg_offset = SOC15_REG_OFFSET(GC, 0, mmCP_IQ_WAIT_TIME2); } @@ -1115,7 +1115,7 @@ const struct kfd2kgd_calls gfx_v10_kfd2kgd = { .set_address_watch = kgd_gfx_v10_set_address_watch, .clear_address_watch = kgd_gfx_v10_clear_address_watch, .get_iq_wait_times = kgd_gfx_v10_get_iq_wait_times, - .build_grace_period_packet_info = kgd_gfx_v10_build_grace_period_packet_info, + .build_dequeue_wait_counts_packet_info = kgd_gfx_v10_build_dequeue_wait_counts_packet_info, .program_trap_handler_settings = program_trap_handler_settings, .hqd_get_pq_addr = kgd_gfx_v10_hqd_get_pq_addr, .hqd_reset = kgd_gfx_v10_hqd_reset, diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gfx_v10.h b/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gfx_v10.h index db577c2a847a..a4c607c88178 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gfx_v10.h +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gfx_v10.h @@ -51,9 +51,10 @@ uint32_t kgd_gfx_v10_clear_address_watch(struct amdgpu_device *adev, void kgd_gfx_v10_get_iq_wait_times(struct amdgpu_device *adev, uint32_t *wait_times, uint32_t inst); -void kgd_gfx_v10_build_grace_period_packet_info(struct amdgpu_device *adev, +void kgd_gfx_v10_build_dequeue_wait_counts_packet_info(struct amdgpu_device *adev, uint32_t wait_times, - uint32_t grace_period, + uint32_t sch_wave, + uint32_t que_sleep, uint32_t *reg_offset, uint32_t *reg_data); uint64_t kgd_gfx_v10_hqd_get_pq_addr(struct amdgpu_device *adev, diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gfx_v10_3.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gfx_v10_3.c index ac9ad505f9d7..6d08bc2781a3 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gfx_v10_3.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gfx_v10_3.c @@ -673,7 +673,7 @@ const struct kfd2kgd_calls gfx_v10_3_kfd2kgd = { .set_vm_context_page_table_base = set_vm_context_page_table_base_v10_3, .program_trap_handler_settings = program_trap_handler_settings_v10_3, .get_iq_wait_times = kgd_gfx_v10_get_iq_wait_times, - .build_grace_period_packet_info = kgd_gfx_v10_build_grace_period_packet_info, + .build_dequeue_wait_counts_packet_info = kgd_gfx_v10_build_dequeue_wait_counts_packet_info, .enable_debug_trap = kgd_gfx_v10_enable_debug_trap, .disable_debug_trap = kgd_gfx_v10_disable_debug_trap, .validate_trap_override_request = kgd_gfx_v10_validate_trap_override_request, diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gfx_v9.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gfx_v9.c index 84135eb90660..088d09cc7a72 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gfx_v9.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gfx_v9.c @@ -1077,25 +1077,25 @@ void kgd_gfx_v9_get_cu_occupancy(struct amdgpu_device *adev, adev->gfx.cu_info.max_waves_per_simd; } -void kgd_gfx_v9_build_grace_period_packet_info(struct amdgpu_device *adev, +void kgd_gfx_v9_build_dequeue_wait_counts_packet_info(struct amdgpu_device *adev, uint32_t wait_times, - uint32_t grace_period, + uint32_t sch_wave, + uint32_t que_sleep, uint32_t *reg_offset, uint32_t *reg_data) { *reg_data = wait_times; - /* - * The CP cannot handle a 0 grace period input and will result in - * an infinite grace period being set so set to 1 to prevent this. - */ - if (grace_period == 0) - grace_period = 1; - - *reg_data = REG_SET_FIELD(*reg_data, - CP_IQ_WAIT_TIME2, - SCH_WAVE, - grace_period); + if (sch_wave) + *reg_data = REG_SET_FIELD(*reg_data, + CP_IQ_WAIT_TIME2, + SCH_WAVE, + sch_wave); + if (que_sleep) + *reg_data = REG_SET_FIELD(*reg_data, + CP_IQ_WAIT_TIME2, + QUE_SLEEP, + que_sleep); *reg_offset = SOC15_REG_OFFSET(GC, 0, mmCP_IQ_WAIT_TIME2); } @@ -1255,7 +1255,7 @@ const struct kfd2kgd_calls gfx_v9_kfd2kgd = { .set_address_watch = kgd_gfx_v9_set_address_watch, .clear_address_watch = kgd_gfx_v9_clear_address_watch, .get_iq_wait_times = kgd_gfx_v9_get_iq_wait_times, - .build_grace_period_packet_info = kgd_gfx_v9_build_grace_period_packet_info, + .build_dequeue_wait_counts_packet_info = kgd_gfx_v9_build_dequeue_wait_counts_packet_info, .get_cu_occupancy = kgd_gfx_v9_get_cu_occupancy, .program_trap_handler_settings = kgd_gfx_v9_program_trap_handler_settings, .hqd_get_pq_addr = kgd_gfx_v9_hqd_get_pq_addr, diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gfx_v9.h b/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gfx_v9.h index 90c8fa13d519..704452ca62f8 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gfx_v9.h +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gfx_v9.h @@ -97,9 +97,10 @@ uint32_t kgd_gfx_v9_clear_address_watch(struct amdgpu_device *adev, void kgd_gfx_v9_get_iq_wait_times(struct amdgpu_device *adev, uint32_t *wait_times, uint32_t inst); -void kgd_gfx_v9_build_grace_period_packet_info(struct amdgpu_device *adev, +void kgd_gfx_v9_build_dequeue_wait_counts_packet_info(struct amdgpu_device *adev, uint32_t wait_times, - uint32_t grace_period, + uint32_t sch_wave, + uint32_t que_sleep, uint32_t *reg_offset, uint32_t *reg_data); uint64_t kgd_gfx_v9_hqd_get_pq_addr(struct amdgpu_device *adev, diff --git a/drivers/gpu/drm/amd/amdkfd/kfd_packet_manager_v9.c b/drivers/gpu/drm/amd/amdkfd/kfd_packet_manager_v9.c index b9c611b249e6..d440df602393 100644 --- a/drivers/gpu/drm/amd/amdkfd/kfd_packet_manager_v9.c +++ b/drivers/gpu/drm/amd/amdkfd/kfd_packet_manager_v9.c @@ -298,13 +298,14 @@ static int pm_map_queues_v9(struct packet_manager *pm, uint32_t *buffer, } static inline void pm_build_dequeue_wait_counts_packet_info(struct packet_manager *pm, - uint32_t sch_value, uint32_t *reg_offset, + uint32_t sch_value, uint32_t que_sleep, uint32_t *reg_offset, uint32_t *reg_data) { - pm->dqm->dev->kfd2kgd->build_grace_period_packet_info( + pm->dqm->dev->kfd2kgd->build_dequeue_wait_counts_packet_info( pm->dqm->dev->adev, pm->dqm->wait_times, sch_value, + que_sleep, reg_offset, reg_data); } @@ -319,27 +320,43 @@ static int pm_config_dequeue_wait_counts_v9(struct packet_manager *pm, uint32_t reg_data = 0; switch (cmd) { - case KFD_DEQUEUE_WAIT_INIT: - /* Set CWSR grace period to 1x1000 cycle for GFX9.4.3 APU */ - if (amdgpu_emu_mode == 0 && pm->dqm->dev->adev->gmc.is_app_apu && - (KFD_GC_VERSION(pm->dqm->dev) == IP_VERSION(9, 4, 3))) - pm_build_dequeue_wait_counts_packet_info(pm, 1, ®_offset, ®_data); - else + case KFD_DEQUEUE_WAIT_INIT: { + uint32_t sch_wave = 0, que_sleep = 0; + /* Reduce CP_IQ_WAIT_TIME2.QUE_SLEEP to 0x1 from default 0x40. + * On a 1GHz machine this is roughly 1 microsecond, which is + * about how long it takes to load data out of memory during + * queue connect + * QUE_SLEEP: Wait Count for Dequeue Retry. + */ + if (KFD_GC_VERSION(pm->dqm->dev) >= IP_VERSION(9, 4, 1) && + KFD_GC_VERSION(pm->dqm->dev) < IP_VERSION(10, 0, 0)) { + que_sleep = 1; + + /* Set CWSR grace period to 1x1000 cycle for GFX9.4.3 APU */ + if (amdgpu_emu_mode == 0 && pm->dqm->dev->adev->gmc.is_app_apu && + (KFD_GC_VERSION(pm->dqm->dev) == IP_VERSION(9, 4, 3))) + sch_wave = 1; + } else { return 0; + } + pm_build_dequeue_wait_counts_packet_info(pm, sch_wave, que_sleep, + ®_offset, ®_data); + break; + } case KFD_DEQUEUE_WAIT_RESET: - /* function called only to get reg_offset */ - pm_build_dequeue_wait_counts_packet_info(pm, 0, ®_offset, ®_data); - reg_data = pm->dqm->wait_times; + /* reg_data would be set to dqm->wait_times */ + pm_build_dequeue_wait_counts_packet_info(pm, 0, 0, ®_offset, ®_data); break; case KFD_DEQUEUE_WAIT_SET_SCH_WAVE: /* The CP cannot handle value 0 and it will result in - * an infinite grace period being set so set to 1 to prevent this. + * an infinite grace period being set so set to 1 to prevent this. Also + * avoid debugger API breakage as it sets 0 and expects a low value. */ if (!value) value = 1; - pm_build_dequeue_wait_counts_packet_info(pm, value, ®_offset, ®_data); + pm_build_dequeue_wait_counts_packet_info(pm, value, 0, ®_offset, ®_data); break; default: pr_err("Invalid dequeue wait cmd\n"); diff --git a/drivers/gpu/drm/amd/include/kgd_kfd_interface.h b/drivers/gpu/drm/amd/include/kgd_kfd_interface.h index 1e8dfa6c0dc8..9aba8596faa7 100644 --- a/drivers/gpu/drm/amd/include/kgd_kfd_interface.h +++ b/drivers/gpu/drm/amd/include/kgd_kfd_interface.h @@ -313,9 +313,10 @@ struct kfd2kgd_calls { void (*get_iq_wait_times)(struct amdgpu_device *adev, uint32_t *wait_times, uint32_t inst); - void (*build_grace_period_packet_info)(struct amdgpu_device *adev, + void (*build_dequeue_wait_counts_packet_info)(struct amdgpu_device *adev, uint32_t wait_times, - uint32_t grace_period, + uint32_t sch_wave, + uint32_t que_sleep, uint32_t *reg_offset, uint32_t *reg_data); void (*get_cu_occupancy)(struct amdgpu_device *adev, -- 2.51.0 From 90df6db62fa78a8ab0b705ec38db99c7973b95d6 Mon Sep 17 00:00:00 2001 From: Alex Deucher Date: Tue, 11 Mar 2025 10:34:36 -0400 Subject: [PATCH 10/16] drm/amdgpu/pm: wire up hwmon fan speed for smu 14.0.2 Add callbacks for fan speed fetching. Closes: https://gitlab.freedesktop.org/drm/amd/-/issues/4034 Reviewed-by: Kenneth Feng Signed-off-by: Alex Deucher --- .../drm/amd/pm/swsmu/smu14/smu_v14_0_2_ppt.c | 35 +++++++++++++++++++ 1 file changed, 35 insertions(+) diff --git a/drivers/gpu/drm/amd/pm/swsmu/smu14/smu_v14_0_2_ppt.c b/drivers/gpu/drm/amd/pm/swsmu/smu14/smu_v14_0_2_ppt.c index 5cad09c5f2ff..e4089fd58711 100644 --- a/drivers/gpu/drm/amd/pm/swsmu/smu14/smu_v14_0_2_ppt.c +++ b/drivers/gpu/drm/amd/pm/swsmu/smu14/smu_v14_0_2_ppt.c @@ -1627,6 +1627,39 @@ out: adev->unique_id = ((uint64_t)upper32 << 32) | lower32; } +static int smu_v14_0_2_get_fan_speed_pwm(struct smu_context *smu, + uint32_t *speed) +{ + int ret; + + if (!speed) + return -EINVAL; + + ret = smu_v14_0_2_get_smu_metrics_data(smu, + METRICS_CURR_FANPWM, + speed); + if (ret) { + dev_err(smu->adev->dev, "Failed to get fan speed(PWM)!"); + return ret; + } + + /* Convert the PMFW output which is in percent to pwm(255) based */ + *speed = min(*speed * 255 / 100, (uint32_t)255); + + return 0; +} + +static int smu_v14_0_2_get_fan_speed_rpm(struct smu_context *smu, + uint32_t *speed) +{ + if (!speed) + return -EINVAL; + + return smu_v14_0_2_get_smu_metrics_data(smu, + METRICS_CURR_FANSPEED, + speed); +} + static int smu_v14_0_2_get_power_limit(struct smu_context *smu, uint32_t *current_power_limit, uint32_t *default_power_limit, @@ -2804,6 +2837,8 @@ static const struct pptable_funcs smu_v14_0_2_ppt_funcs = { .set_performance_level = smu_v14_0_set_performance_level, .gfx_off_control = smu_v14_0_gfx_off_control, .get_unique_id = smu_v14_0_2_get_unique_id, + .get_fan_speed_pwm = smu_v14_0_2_get_fan_speed_pwm, + .get_fan_speed_rpm = smu_v14_0_2_get_fan_speed_rpm, .get_power_limit = smu_v14_0_2_get_power_limit, .set_power_limit = smu_v14_0_2_set_power_limit, .get_power_profile_mode = smu_v14_0_2_get_power_profile_mode, -- 2.51.0 From 2bc016737a115469eb19a9a2d5d8d60ed38f6f82 Mon Sep 17 00:00:00 2001 From: Alex Deucher Date: Tue, 11 Mar 2025 16:30:58 -0400 Subject: [PATCH 11/16] drm/amdgpu/pm: add VCN activity for renoir Wire up the query. Reviewed-by: Lijo Lazar Reviewed-by: Kenneth Feng Signed-off-by: Alex Deucher --- drivers/gpu/drm/amd/pm/swsmu/smu12/renoir_ppt.c | 6 ++++++ 1 file changed, 6 insertions(+) diff --git a/drivers/gpu/drm/amd/pm/swsmu/smu12/renoir_ppt.c b/drivers/gpu/drm/amd/pm/swsmu/smu12/renoir_ppt.c index 37d82a71a2d7..9481f897432d 100644 --- a/drivers/gpu/drm/amd/pm/swsmu/smu12/renoir_ppt.c +++ b/drivers/gpu/drm/amd/pm/swsmu/smu12/renoir_ppt.c @@ -1285,6 +1285,12 @@ static int renoir_read_sensor(struct smu_context *smu, (uint32_t *)data); *size = 4; break; + case AMDGPU_PP_SENSOR_VCN_LOAD: + ret = renoir_get_smu_metrics_data(smu, + METRICS_AVERAGE_VCNACTIVITY, + (uint32_t *)data); + *size = 4; + break; case AMDGPU_PP_SENSOR_EDGE_TEMP: ret = renoir_get_smu_metrics_data(smu, METRICS_TEMPERATURE_EDGE, -- 2.51.0 From 357506799bacb21ba1ed7d64e02ac9c4a1f45b31 Mon Sep 17 00:00:00 2001 From: Lijo Lazar Date: Thu, 6 Feb 2025 17:40:42 +0530 Subject: [PATCH 12/16] drm/amdgpu: Calculate IP specific xgmi bandwidth Use IP version specific xgmi speed/width for bandwidth calculation. Signed-off-by: Lijo Lazar Reviewed-by: Jonathan Kim Signed-off-by: Alex Deucher --- drivers/gpu/drm/amd/amdgpu/amdgpu_device.c | 3 +++ drivers/gpu/drm/amd/amdgpu/amdgpu_xgmi.c | 4 +++- 2 files changed, 6 insertions(+), 1 deletion(-) diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_device.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_device.c index 9e9fec5b52de..17ce1677378e 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_device.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_device.c @@ -2757,6 +2757,9 @@ static int amdgpu_device_ip_early_init(struct amdgpu_device *adev) if (!total) return -ENODEV; + if (adev->gmc.xgmi.supported) + amdgpu_xgmi_early_init(adev); + ip_block = amdgpu_device_ip_get_ip_block(adev, AMD_IP_BLOCK_TYPE_GFX); if (ip_block->status.valid != false) amdgpu_amdkfd_device_probe(adev); diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_xgmi.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_xgmi.c index 7fdf30f1161c..6029a799074d 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_xgmi.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_xgmi.c @@ -844,7 +844,9 @@ int amdgpu_xgmi_get_bandwidth(struct amdgpu_device *adev, struct amdgpu_device * { bool peer_mode = bw_mode == AMDGPU_XGMI_BW_MODE_PER_PEER; int unit_scale = bw_unit == AMDGPU_XGMI_BW_UNIT_MBYTES ? 1000 : 1; - int speed = 25, num_lanes = 16, num_links = !peer_mode ? 1 : -1; + int num_lanes = adev->gmc.xgmi.max_width; + int speed = adev->gmc.xgmi.max_speed; + int num_links = !peer_mode ? 1 : -1; if (!(min_bw && max_bw)) return -EINVAL; -- 2.51.0 From 6a87982b5897c387f5508fddfe2404726045517c Mon Sep 17 00:00:00 2001 From: Alex Hung Date: Tue, 11 Mar 2025 11:10:17 -0600 Subject: [PATCH 13/16] drm/amd/display: Remove incorrect macro guard MIME-Version: 1.0 Content-Type: text/plain; charset=utf8 Content-Transfer-Encoding: 8bit This macro guard "__cplusplus" is unnecessary and should not be there. Signed-off-by: Alex Hung Reviewed-by: Rodrigo Siqueira Reviewed-by: Christian König Signed-off-by: Alex Deucher --- drivers/gpu/drm/amd/display/dc/sspl/dc_spl.h | 3 --- 1 file changed, 3 deletions(-) diff --git a/drivers/gpu/drm/amd/display/dc/sspl/dc_spl.h b/drivers/gpu/drm/amd/display/dc/sspl/dc_spl.h index 145961803a92..d621c42a237e 100644 --- a/drivers/gpu/drm/amd/display/dc/sspl/dc_spl.h +++ b/drivers/gpu/drm/amd/display/dc/sspl/dc_spl.h @@ -17,9 +17,6 @@ #define SPL_EXPAND(a, b) SPL_EXPAND2(a, b) #define SPL_NAMESPACE(symbol) SPL_EXPAND(SPL_PFX_, symbol) -#ifdef __cplusplus -extern "C" { -#endif /* SPL interfaces */ -- 2.51.0 From 2393c1a907c2e79cfb4a1fd6ecee9dc7110ad3d1 Mon Sep 17 00:00:00 2001 From: Alex Deucher Date: Tue, 11 Mar 2025 15:00:12 -0400 Subject: [PATCH 14/16] drm/amdgpu/pm: add VCN activity for SMU 13.0.0/7 Wire up the query. Reviewed-by: Lijo Lazar Reviewed-by: Kenneth Feng Signed-off-by: Alex Deucher --- drivers/gpu/drm/amd/pm/swsmu/smu13/smu_v13_0_0_ppt.c | 10 ++++++++++ drivers/gpu/drm/amd/pm/swsmu/smu13/smu_v13_0_7_ppt.c | 10 ++++++++++ 2 files changed, 20 insertions(+) diff --git a/drivers/gpu/drm/amd/pm/swsmu/smu13/smu_v13_0_0_ppt.c b/drivers/gpu/drm/amd/pm/swsmu/smu13/smu_v13_0_0_ppt.c index 898487ad6cd2..5a9711e8cf68 100644 --- a/drivers/gpu/drm/amd/pm/swsmu/smu13/smu_v13_0_0_ppt.c +++ b/drivers/gpu/drm/amd/pm/swsmu/smu13/smu_v13_0_0_ppt.c @@ -836,6 +836,10 @@ static int smu_v13_0_0_get_smu_metrics_data(struct smu_context *smu, case METRICS_AVERAGE_MEMACTIVITY: *value = metrics->AverageUclkActivity; break; + case METRICS_AVERAGE_VCNACTIVITY: + *value = max(metrics->Vcn0ActivityPercentage, + metrics->Vcn1ActivityPercentage); + break; case METRICS_AVERAGE_SOCKETPOWER: *value = metrics->AverageSocketPower << 8; break; @@ -962,6 +966,12 @@ static int smu_v13_0_0_read_sensor(struct smu_context *smu, (uint32_t *)data); *size = 4; break; + case AMDGPU_PP_SENSOR_VCN_LOAD: + ret = smu_v13_0_0_get_smu_metrics_data(smu, + METRICS_AVERAGE_VCNACTIVITY, + (uint32_t *)data); + *size = 4; + break; case AMDGPU_PP_SENSOR_GPU_AVG_POWER: ret = smu_v13_0_0_get_smu_metrics_data(smu, METRICS_AVERAGE_SOCKETPOWER, diff --git a/drivers/gpu/drm/amd/pm/swsmu/smu13/smu_v13_0_7_ppt.c b/drivers/gpu/drm/amd/pm/swsmu/smu13/smu_v13_0_7_ppt.c index 19f47811f6db..c8f4f6fb4083 100644 --- a/drivers/gpu/drm/amd/pm/swsmu/smu13/smu_v13_0_7_ppt.c +++ b/drivers/gpu/drm/amd/pm/swsmu/smu13/smu_v13_0_7_ppt.c @@ -807,6 +807,10 @@ static int smu_v13_0_7_get_smu_metrics_data(struct smu_context *smu, else *value = metrics->AverageMemclkFrequencyPreDs; break; + case METRICS_AVERAGE_VCNACTIVITY: + *value = max(metrics->Vcn0ActivityPercentage, + metrics->Vcn1ActivityPercentage); + break; case METRICS_AVERAGE_VCLK: *value = metrics->AverageVclk0Frequency; break; @@ -951,6 +955,12 @@ static int smu_v13_0_7_read_sensor(struct smu_context *smu, (uint32_t *)data); *size = 4; break; + case AMDGPU_PP_SENSOR_VCN_LOAD: + ret = smu_v13_0_7_get_smu_metrics_data(smu, + METRICS_AVERAGE_VCNACTIVITY, + (uint32_t *)data); + *size = 4; + break; case AMDGPU_PP_SENSOR_GPU_AVG_POWER: ret = smu_v13_0_7_get_smu_metrics_data(smu, METRICS_AVERAGE_SOCKETPOWER, -- 2.51.0 From 1b81674e0baf6661b12bbd0beaf68d9ed1afb04c Mon Sep 17 00:00:00 2001 From: Alex Deucher Date: Tue, 11 Mar 2025 14:48:22 -0400 Subject: [PATCH 15/16] drm/amdgpu/pm: add VCN activity for SMU 14.0.2 Wire up the query. Reviewed-by: Lijo Lazar Reviewed-by: Kenneth Feng Signed-off-by: Alex Deucher --- drivers/gpu/drm/amd/pm/swsmu/smu14/smu_v14_0_2_ppt.c | 10 ++++++++++ 1 file changed, 10 insertions(+) diff --git a/drivers/gpu/drm/amd/pm/swsmu/smu14/smu_v14_0_2_ppt.c b/drivers/gpu/drm/amd/pm/swsmu/smu14/smu_v14_0_2_ppt.c index e4089fd58711..21d1b34ada05 100644 --- a/drivers/gpu/drm/amd/pm/swsmu/smu14/smu_v14_0_2_ppt.c +++ b/drivers/gpu/drm/amd/pm/swsmu/smu14/smu_v14_0_2_ppt.c @@ -756,6 +756,10 @@ static int smu_v14_0_2_get_smu_metrics_data(struct smu_context *smu, case METRICS_AVERAGE_MEMACTIVITY: *value = metrics->AverageUclkActivity; break; + case METRICS_AVERAGE_VCNACTIVITY: + *value = max(metrics->AverageVcn0ActivityPercentage, + metrics->Vcn1ActivityPercentage); + break; case METRICS_AVERAGE_SOCKETPOWER: *value = metrics->AverageSocketPower << 8; break; @@ -882,6 +886,12 @@ static int smu_v14_0_2_read_sensor(struct smu_context *smu, (uint32_t *)data); *size = 4; break; + case AMDGPU_PP_SENSOR_VCN_LOAD: + ret = smu_v14_0_2_get_smu_metrics_data(smu, + METRICS_AVERAGE_VCNACTIVITY, + (uint32_t *)data); + *size = 4; + break; case AMDGPU_PP_SENSOR_GPU_AVG_POWER: ret = smu_v14_0_2_get_smu_metrics_data(smu, METRICS_AVERAGE_SOCKETPOWER, -- 2.51.0 From 18537feb182d8506d4ee2778e11aad7c60e1453f Mon Sep 17 00:00:00 2001 From: Alex Deucher Date: Tue, 11 Mar 2025 15:37:30 -0400 Subject: [PATCH 16/16] drm/amdgpu/pm: enable vcn busy sysfs for additional GC 11.x Make it visible for the all GC 11.x chips that support it. Reviewed-by: Lijo Lazar Reviewed-by: Kenneth Feng Signed-off-by: Alex Deucher --- drivers/gpu/drm/amd/pm/amdgpu_pm.c | 18 ++++++++++++------ 1 file changed, 12 insertions(+), 6 deletions(-) diff --git a/drivers/gpu/drm/amd/pm/amdgpu_pm.c b/drivers/gpu/drm/amd/pm/amdgpu_pm.c index 2179344e78d9..c1f5ef7384b4 100644 --- a/drivers/gpu/drm/amd/pm/amdgpu_pm.c +++ b/drivers/gpu/drm/amd/pm/amdgpu_pm.c @@ -2315,12 +2315,18 @@ static int default_attr_update(struct amdgpu_device *adev, struct amdgpu_device_ *states = ATTR_STATE_UNSUPPORTED; } else if (DEVICE_ATTR_IS(vcn_busy_percent)) { if (!(gc_ver == IP_VERSION(10, 3, 1) || - gc_ver == IP_VERSION(10, 3, 3) || - gc_ver == IP_VERSION(10, 3, 6) || - gc_ver == IP_VERSION(10, 3, 7) || - gc_ver == IP_VERSION(11, 0, 1) || - gc_ver == IP_VERSION(11, 0, 4) || - gc_ver == IP_VERSION(11, 5, 0))) + gc_ver == IP_VERSION(10, 3, 3) || + gc_ver == IP_VERSION(10, 3, 6) || + gc_ver == IP_VERSION(10, 3, 7) || + gc_ver == IP_VERSION(11, 0, 0) || + gc_ver == IP_VERSION(11, 0, 1) || + gc_ver == IP_VERSION(11, 0, 2) || + gc_ver == IP_VERSION(11, 0, 3) || + gc_ver == IP_VERSION(11, 0, 4) || + gc_ver == IP_VERSION(11, 5, 0) || + gc_ver == IP_VERSION(11, 5, 1) || + gc_ver == IP_VERSION(11, 5, 2) || + gc_ver == IP_VERSION(11, 5, 3))) *states = ATTR_STATE_UNSUPPORTED; } else if (DEVICE_ATTR_IS(pcie_bw)) { /* PCIe Perf counters won't work on APU nodes */ -- 2.51.0