From eed6a6b2264078806163424b7cb70a3b47cb8e97 Mon Sep 17 00:00:00 2001 From: Andrey Vatoropin Date: Wed, 2 Apr 2025 14:12:25 +0000 Subject: drm/amdkfd: Change svm_range_get_info return type Static analysis shows that pointer "svms" cannot be NULL because it points to the object "struct svm_range_list". Remove the extra NULL check. It is meaningless and harms the readability of the code. In the function svm_range_get_info() there is no possibility of failure. Therefore, the caller of the function svm_range_get_info() does not need a return value. Change the function svm_range_get_info() return type from "int" to "void". Since the function svm_range_get_info() has a return type of "void". The caller of the function svm_range_get_info() does not need a return value. Delete extra code. Found by Linux Verification Center (linuxtesting.org) with SVACE. Signed-off-by: Andrey Vatoropin Signed-off-by: Felix Kuehling Reviewed-by: Felix Kuehling Signed-off-by: Alex Deucher --- drivers/gpu/drm/amd/amdkfd/kfd_chardev.c | 4 +--- drivers/gpu/drm/amd/amdkfd/kfd_svm.c | 9 ++------- drivers/gpu/drm/amd/amdkfd/kfd_svm.h | 11 +++++------ 3 files changed, 8 insertions(+), 16 deletions(-) (limited to 'drivers/gpu') diff --git a/drivers/gpu/drm/amd/amdkfd/kfd_chardev.c b/drivers/gpu/drm/amd/amdkfd/kfd_chardev.c index 1e9dd00620bf..a2149afa5803 100644 --- a/drivers/gpu/drm/amd/amdkfd/kfd_chardev.c +++ b/drivers/gpu/drm/amd/amdkfd/kfd_chardev.c @@ -2039,9 +2039,7 @@ static int criu_get_process_object_info(struct kfd_process *p, num_events = kfd_get_num_events(p); - ret = svm_range_get_info(p, &num_svm_ranges, &svm_priv_data_size); - if (ret) - return ret; + svm_range_get_info(p, &num_svm_ranges, &svm_priv_data_size); *num_objects = num_queues + num_events + num_svm_ranges; diff --git a/drivers/gpu/drm/amd/amdkfd/kfd_svm.c b/drivers/gpu/drm/amd/amdkfd/kfd_svm.c index 72be6e152e88..865dca2547de 100644 --- a/drivers/gpu/drm/amd/amdkfd/kfd_svm.c +++ b/drivers/gpu/drm/amd/amdkfd/kfd_svm.c @@ -4075,8 +4075,8 @@ exit: return ret; } -int svm_range_get_info(struct kfd_process *p, uint32_t *num_svm_ranges, - uint64_t *svm_priv_data_size) +void svm_range_get_info(struct kfd_process *p, uint32_t *num_svm_ranges, + uint64_t *svm_priv_data_size) { uint64_t total_size, accessibility_size, common_attr_size; int nattr_common = 4, nattr_accessibility = 1; @@ -4088,8 +4088,6 @@ int svm_range_get_info(struct kfd_process *p, uint32_t *num_svm_ranges, *svm_priv_data_size = 0; svms = &p->svms; - if (!svms) - return -EINVAL; mutex_lock(&svms->lock); list_for_each_entry(prange, &svms->list, list) { @@ -4131,7 +4129,6 @@ int svm_range_get_info(struct kfd_process *p, uint32_t *num_svm_ranges, pr_debug("num_svm_ranges %u total_priv_size %llu\n", *num_svm_ranges, *svm_priv_data_size); - return 0; } int kfd_criu_checkpoint_svm(struct kfd_process *p, @@ -4148,8 +4145,6 @@ int kfd_criu_checkpoint_svm(struct kfd_process *p, struct mm_struct *mm; svms = &p->svms; - if (!svms) - return -EINVAL; mm = get_task_mm(p->lead_thread); if (!mm) { diff --git a/drivers/gpu/drm/amd/amdkfd/kfd_svm.h b/drivers/gpu/drm/amd/amdkfd/kfd_svm.h index 6ea23c78009c..01c7a4877904 100644 --- a/drivers/gpu/drm/amd/amdkfd/kfd_svm.h +++ b/drivers/gpu/drm/amd/amdkfd/kfd_svm.h @@ -184,8 +184,8 @@ void schedule_deferred_list_work(struct svm_range_list *svms); void svm_range_dma_unmap_dev(struct device *dev, dma_addr_t *dma_addr, unsigned long offset, unsigned long npages); void svm_range_dma_unmap(struct svm_range *prange); -int svm_range_get_info(struct kfd_process *p, uint32_t *num_svm_ranges, - uint64_t *svm_priv_data_size); +void svm_range_get_info(struct kfd_process *p, uint32_t *num_svm_ranges, + uint64_t *svm_priv_data_size); int kfd_criu_checkpoint_svm(struct kfd_process *p, uint8_t __user *user_priv_data, uint64_t *priv_offset); @@ -237,13 +237,12 @@ static inline int svm_range_schedule_evict_svm_bo( return -EINVAL; } -static inline int svm_range_get_info(struct kfd_process *p, - uint32_t *num_svm_ranges, - uint64_t *svm_priv_data_size) +static inline void svm_range_get_info(struct kfd_process *p, + uint32_t *num_svm_ranges, + uint64_t *svm_priv_data_size) { *num_svm_ranges = 0; *svm_priv_data_size = 0; - return 0; } static inline int kfd_criu_checkpoint_svm(struct kfd_process *p, -- cgit v1.2.3 From b2c11e27080d8556664c20c30ca3527ffa99bec4 Mon Sep 17 00:00:00 2001 From: Prike Liang Date: Wed, 30 Apr 2025 10:34:14 +0800 Subject: drm/amdgpu: lock the eviction fence for wq signals it MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit Lock and refer to the eviction fence before the eviction fence schedules work queue tries to signal it. Suggested-by: Christian König Signed-off-by: Prike Liang Acked-by: Alex Deucher Reviewed-by: Arvind Yadav Reviewed-by: Christian König Signed-off-by: Alex Deucher --- drivers/gpu/drm/amd/amdgpu/amdgpu_eviction_fence.c | 11 ++++++++++- 1 file changed, 10 insertions(+), 1 deletion(-) (limited to 'drivers/gpu') diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_eviction_fence.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_eviction_fence.c index 73b629b5f56f..8b919ad3af29 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_eviction_fence.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_eviction_fence.c @@ -108,13 +108,22 @@ amdgpu_eviction_fence_suspend_worker(struct work_struct *work) struct amdgpu_eviction_fence *ev_fence; mutex_lock(&uq_mgr->userq_mutex); + spin_lock(&evf_mgr->ev_fence_lock); ev_fence = evf_mgr->ev_fence; - if (!ev_fence) + if (ev_fence) + dma_fence_get(&ev_fence->base); + else goto unlock; + spin_unlock(&evf_mgr->ev_fence_lock); amdgpu_userq_evict(uq_mgr, ev_fence); + mutex_unlock(&uq_mgr->userq_mutex); + dma_fence_put(&ev_fence->base); + return; + unlock: + spin_unlock(&evf_mgr->ev_fence_lock); mutex_unlock(&uq_mgr->userq_mutex); } -- cgit v1.2.3 From 0132ba7ff0f613915d332a30fcf14cf66e317f98 Mon Sep 17 00:00:00 2001 From: "Jesse.Zhang" Date: Thu, 15 May 2025 15:02:13 +0800 Subject: drm/amdgpu: Fix eviction fence worker race during fd close MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit The current cleanup order during file descriptor close can lead to a race condition where the eviction fence worker attempts to access a destroyed mutex from the user queue manager: [ 517.294055] DEBUG_LOCKS_WARN_ON(lock->magic != lock) [ 517.294060] WARNING: CPU: 8 PID: 2030 at kernel/locking/mutex.c:564 [ 517.294094] Workqueue: events amdgpu_eviction_fence_suspend_worker [amdgpu] The issue occurs because: 1. We destroy the user queue manager (including its mutex) first 2. Then try to destroy eviction fences which may have pending work 3. The eviction fence worker may try to access the already-destroyed mutex Fix this by reordering the cleanup to: 1. First mark the fd as closing and destroy eviction fences, which flushes any pending work 2. Then safely destroy the user queue manager after we're certain no more fence work will be executed The copy in amdgpu_driver_postclose_kms() needs to be removed (Christian) Reviewed-by: Christian König Reviewed-by: Prike Liang Reviewed-by: Arvind Yadav Signed-off-by: Jesse Zhang Signed-off-by: Alex Deucher --- drivers/gpu/drm/amd/amdgpu/amdgpu_drv.c | 2 +- drivers/gpu/drm/amd/amdgpu/amdgpu_kms.c | 5 ----- 2 files changed, 1 insertion(+), 6 deletions(-) (limited to 'drivers/gpu') diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_drv.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_drv.c index 4ddd08ce8885..4db92e0a60da 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_drv.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_drv.c @@ -2913,8 +2913,8 @@ static int amdgpu_drm_release(struct inode *inode, struct file *filp) if (fpriv) { fpriv->evf_mgr.fd_closing = true; - amdgpu_userq_mgr_fini(&fpriv->userq_mgr); amdgpu_eviction_fence_destroy(&fpriv->evf_mgr); + amdgpu_userq_mgr_fini(&fpriv->userq_mgr); } return drm_release(inode, filp); diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_kms.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_kms.c index 9fbb04aee97b..d2ce7d86dbc8 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_kms.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_kms.c @@ -1502,11 +1502,6 @@ void amdgpu_driver_postclose_kms(struct drm_device *dev, amdgpu_bo_unreserve(pd); } - if (!fpriv->evf_mgr.fd_closing) { - fpriv->evf_mgr.fd_closing = true; - amdgpu_userq_mgr_fini(&fpriv->userq_mgr); - amdgpu_eviction_fence_destroy(&fpriv->evf_mgr); - } amdgpu_ctx_mgr_fini(&fpriv->ctx_mgr); amdgpu_vm_fini(adev, &fpriv->vm); -- cgit v1.2.3 From 7f30f8ef6547d67c1932697d6b5b971665b43e5b Mon Sep 17 00:00:00 2001 From: Asad Kamal Date: Wed, 30 Apr 2025 00:25:01 +0800 Subject: drm/amd/pm: Update pmfw headers for smu_v_13_0_6 Update pmfw headers for smu_v_13_0_6 to include pldm version as part of statics metrics table Signed-off-by: Asad Kamal Signed-off-by: Lijo Lazar Reviewed-by: Hawking Zhang Signed-off-by: Alex Deucher --- drivers/gpu/drm/amd/pm/swsmu/inc/pmfw_if/smu_v13_0_6_pmfw.h | 4 +++- 1 file changed, 3 insertions(+), 1 deletion(-) (limited to 'drivers/gpu') diff --git a/drivers/gpu/drm/amd/pm/swsmu/inc/pmfw_if/smu_v13_0_6_pmfw.h b/drivers/gpu/drm/amd/pm/swsmu/inc/pmfw_if/smu_v13_0_6_pmfw.h index 3d9e5e967c94..01790a927930 100644 --- a/drivers/gpu/drm/amd/pm/swsmu/inc/pmfw_if/smu_v13_0_6_pmfw.h +++ b/drivers/gpu/drm/amd/pm/swsmu/inc/pmfw_if/smu_v13_0_6_pmfw.h @@ -127,7 +127,7 @@ typedef enum { VOLTAGE_GUARDBAND_COUNT } GFX_GUARDBAND_e; -#define SMU_METRICS_TABLE_VERSION 0x10 +#define SMU_METRICS_TABLE_VERSION 0x11 // Unified metrics table for smu_v13_0_6 typedef struct __attribute__((packed, aligned(4))) { @@ -463,6 +463,8 @@ typedef struct __attribute__((packed, aligned(4))) { typedef struct { // Telemetry uint32_t InputTelemetryVoltageInmV; + // General info + uint32_t pldmVersion[2]; } StaticMetricsTable_t; #pragma pack(pop) -- cgit v1.2.3 From 2ed4fd99690ba1efab204f03be80fabd6c0ec3cf Mon Sep 17 00:00:00 2001 From: Asad Kamal Date: Wed, 30 Apr 2025 00:40:50 +0800 Subject: drm/amd/pm: Fill pldm version for SMU v13.0.6 SOCs Fetch pldm version from static metrics table for SMU v13.0.6 SOCs Signed-off-by: Asad Kamal Signed-off-by: Lijo Lazar Reviewed-by: Hawking Zhang Signed-off-by: Alex Deucher --- drivers/gpu/drm/amd/pm/swsmu/smu13/smu_v13_0_6_ppt.c | 7 +++++++ drivers/gpu/drm/amd/pm/swsmu/smu13/smu_v13_0_6_ppt.h | 1 + 2 files changed, 8 insertions(+) (limited to 'drivers/gpu') diff --git a/drivers/gpu/drm/amd/pm/swsmu/smu13/smu_v13_0_6_ppt.c b/drivers/gpu/drm/amd/pm/swsmu/smu13/smu_v13_0_6_ppt.c index 7d4ff09be7e8..cd9ed3b5e9fa 100644 --- a/drivers/gpu/drm/amd/pm/swsmu/smu13/smu_v13_0_6_ppt.c +++ b/drivers/gpu/drm/amd/pm/swsmu/smu13/smu_v13_0_6_ppt.c @@ -396,6 +396,8 @@ static void smu_v13_0_6_init_caps(struct smu_context *smu) smu_v13_0_6_cap_set(smu, SMU_CAP(STATIC_METRICS)); smu_v13_0_6_cap_set(smu, SMU_CAP(BOARD_VOLTAGE)); } + if (fw_ver >= 0x00558000) + smu_v13_0_6_cap_set(smu, SMU_CAP(PLDM_VERSION)); } if (((pgm == 7) && (fw_ver >= 0x7550700)) || ((pgm == 0) && (fw_ver >= 0x00557900)) || @@ -752,6 +754,11 @@ static void smu_v13_0_6_fill_static_metrics_table(struct smu_context *smu, } dpm_context->board_volt = static_metrics->InputTelemetryVoltageInmV; + + if (smu_v13_0_6_cap_supported(smu, SMU_CAP(PLDM_VERSION)) && + static_metrics->pldmVersion[0] != 0xFFFFFFFF) + smu->adev->firmware.pldm_version = + static_metrics->pldmVersion[0]; } int smu_v13_0_6_get_static_metrics_table(struct smu_context *smu) diff --git a/drivers/gpu/drm/amd/pm/swsmu/smu13/smu_v13_0_6_ppt.h b/drivers/gpu/drm/amd/pm/swsmu/smu13/smu_v13_0_6_ppt.h index d151bcd0cca7..1ccc150882eb 100644 --- a/drivers/gpu/drm/amd/pm/swsmu/smu13/smu_v13_0_6_ppt.h +++ b/drivers/gpu/drm/amd/pm/swsmu/smu13/smu_v13_0_6_ppt.h @@ -67,6 +67,7 @@ enum smu_v13_0_6_caps { SMU_CAP(STATIC_METRICS), SMU_CAP(HST_LIMIT_METRICS), SMU_CAP(BOARD_VOLTAGE), + SMU_CAP(PLDM_VERSION), SMU_CAP(ALL), }; -- cgit v1.2.3 From 1327d8f4061c08c29ea8ce7bb89e209d3c1e8b29 Mon Sep 17 00:00:00 2001 From: Lijo Lazar Date: Tue, 29 Apr 2025 08:37:38 +0530 Subject: drm/amd/pm: Use macro to initialize metrics table Helps to keep a build time check about usage of right datatype and avoids maintenance as new versions get added. Signed-off-by: Lijo Lazar Reviewed-by: Hawking Zhang Reviewed-by: Yang Wang Reviewed-by: Asad Kamal Signed-off-by: Alex Deucher --- drivers/gpu/drm/amd/pm/swsmu/smu_cmn.c | 67 ---------------------------------- drivers/gpu/drm/amd/pm/swsmu/smu_cmn.h | 14 ++++++- 2 files changed, 12 insertions(+), 69 deletions(-) (limited to 'drivers/gpu') diff --git a/drivers/gpu/drm/amd/pm/swsmu/smu_cmn.c b/drivers/gpu/drm/amd/pm/swsmu/smu_cmn.c index 80eb1a03b3ca..7eaf58fd7f9a 100644 --- a/drivers/gpu/drm/amd/pm/swsmu/smu_cmn.c +++ b/drivers/gpu/drm/amd/pm/swsmu/smu_cmn.c @@ -1051,73 +1051,6 @@ int smu_cmn_get_combo_pptable(struct smu_context *smu) false); } -void smu_cmn_init_soft_gpu_metrics(void *table, uint8_t frev, uint8_t crev) -{ - struct metrics_table_header *header = (struct metrics_table_header *)table; - uint16_t structure_size; - -#define METRICS_VERSION(a, b) ((a << 16) | b) - - switch (METRICS_VERSION(frev, crev)) { - case METRICS_VERSION(1, 0): - structure_size = sizeof(struct gpu_metrics_v1_0); - break; - case METRICS_VERSION(1, 1): - structure_size = sizeof(struct gpu_metrics_v1_1); - break; - case METRICS_VERSION(1, 2): - structure_size = sizeof(struct gpu_metrics_v1_2); - break; - case METRICS_VERSION(1, 3): - structure_size = sizeof(struct gpu_metrics_v1_3); - break; - case METRICS_VERSION(1, 4): - structure_size = sizeof(struct gpu_metrics_v1_4); - break; - case METRICS_VERSION(1, 5): - structure_size = sizeof(struct gpu_metrics_v1_5); - break; - case METRICS_VERSION(1, 6): - structure_size = sizeof(struct gpu_metrics_v1_6); - break; - case METRICS_VERSION(1, 7): - structure_size = sizeof(struct gpu_metrics_v1_7); - break; - case METRICS_VERSION(1, 8): - structure_size = sizeof(struct gpu_metrics_v1_8); - break; - case METRICS_VERSION(2, 0): - structure_size = sizeof(struct gpu_metrics_v2_0); - break; - case METRICS_VERSION(2, 1): - structure_size = sizeof(struct gpu_metrics_v2_1); - break; - case METRICS_VERSION(2, 2): - structure_size = sizeof(struct gpu_metrics_v2_2); - break; - case METRICS_VERSION(2, 3): - structure_size = sizeof(struct gpu_metrics_v2_3); - break; - case METRICS_VERSION(2, 4): - structure_size = sizeof(struct gpu_metrics_v2_4); - break; - case METRICS_VERSION(3, 0): - structure_size = sizeof(struct gpu_metrics_v3_0); - break; - default: - return; - } - -#undef METRICS_VERSION - - memset(header, 0xFF, structure_size); - - header->format_revision = frev; - header->content_revision = crev; - header->structure_size = structure_size; - -} - int smu_cmn_set_mp1_state(struct smu_context *smu, enum pp_mp1_state mp1_state) { diff --git a/drivers/gpu/drm/amd/pm/swsmu/smu_cmn.h b/drivers/gpu/drm/amd/pm/swsmu/smu_cmn.h index a020277dec3e..ade36a8ffa66 100644 --- a/drivers/gpu/drm/amd/pm/swsmu/smu_cmn.h +++ b/drivers/gpu/drm/amd/pm/swsmu/smu_cmn.h @@ -40,6 +40,18 @@ #define SMU_IH_INTERRUPT_CONTEXT_ID_FAN_ABNORMAL 0x8 #define SMU_IH_INTERRUPT_CONTEXT_ID_FAN_RECOVERY 0x9 +#define smu_cmn_init_soft_gpu_metrics(ptr, frev, crev) \ + do { \ + typecheck(struct gpu_metrics_v##frev##_##crev, \ + typeof(*(ptr))); \ + struct metrics_table_header *header = \ + (struct metrics_table_header *)(ptr); \ + memset(header, 0xFF, sizeof(*(ptr))); \ + header->format_revision = frev; \ + header->content_revision = crev; \ + header->structure_size = sizeof(*(ptr)); \ + } while (0) + extern const int link_speed[]; /* Helper to Convert from PCIE Gen 1/2/3/4/5/6 to 0.1 GT/s speed units */ @@ -125,8 +137,6 @@ int smu_cmn_get_metrics_table(struct smu_context *smu, int smu_cmn_get_combo_pptable(struct smu_context *smu); -void smu_cmn_init_soft_gpu_metrics(void *table, uint8_t frev, uint8_t crev); - int smu_cmn_set_mp1_state(struct smu_context *smu, enum pp_mp1_state mp1_state); -- cgit v1.2.3 From 4c4a89149608c5b006570b933e893a06c45edfe1 Mon Sep 17 00:00:00 2001 From: "Stanley.Yang" Date: Tue, 13 May 2025 19:46:08 +0800 Subject: drm/amdgpu: Register aqua vanjaram vcn poison irq Register aqua vanjaram vcn poison irq, add vcn poison handle. Signed-off-by: Stanley.Yang Reviewed-by: Hawking Zhang Signed-off-by: Alex Deucher --- drivers/gpu/drm/amd/amdgpu/vcn_v4_0_3.c | 65 +++++++++++++++++++++++++++++++++ drivers/gpu/drm/amd/amdgpu/vcn_v4_0_3.h | 6 +++ 2 files changed, 71 insertions(+) (limited to 'drivers/gpu') diff --git a/drivers/gpu/drm/amd/amdgpu/vcn_v4_0_3.c b/drivers/gpu/drm/amd/amdgpu/vcn_v4_0_3.c index 712e1fba33ce..764b3ff09f1e 100644 --- a/drivers/gpu/drm/amd/amdgpu/vcn_v4_0_3.c +++ b/drivers/gpu/drm/amd/amdgpu/vcn_v4_0_3.c @@ -169,6 +169,10 @@ static int vcn_v4_0_3_sw_init(struct amdgpu_ip_block *ip_block) if (r) return r; + /* VCN POISON TRAP */ + r = amdgpu_irq_add_id(adev, SOC15_IH_CLIENTID_VCN, + VCN_4_0__SRCID_UVD_POISON, &adev->vcn.inst->ras_poison_irq); + for (i = 0; i < adev->vcn.num_vcn_inst; i++) { r = amdgpu_vcn_sw_init(adev, i); @@ -387,6 +391,9 @@ static int vcn_v4_0_3_hw_fini(struct amdgpu_ip_block *ip_block) vinst->set_pg_state(vinst, AMD_PG_STATE_GATE); } + if (amdgpu_ras_is_supported(adev, AMDGPU_RAS_BLOCK__VCN)) + amdgpu_irq_put(adev, &adev->vcn.inst->ras_poison_irq, 0); + return 0; } @@ -1814,11 +1821,24 @@ static int vcn_v4_0_3_process_interrupt(struct amdgpu_device *adev, return 0; } +static int vcn_v4_0_3_set_ras_interrupt_state(struct amdgpu_device *adev, + struct amdgpu_irq_src *source, + unsigned int type, + enum amdgpu_interrupt_state state) +{ + return 0; +} + static const struct amdgpu_irq_src_funcs vcn_v4_0_3_irq_funcs = { .set = vcn_v4_0_3_set_interrupt_state, .process = vcn_v4_0_3_process_interrupt, }; +static const struct amdgpu_irq_src_funcs vcn_v4_0_3_ras_irq_funcs = { + .set = vcn_v4_0_3_set_ras_interrupt_state, + .process = amdgpu_vcn_process_poison_irq, +}; + /** * vcn_v4_0_3_set_irq_funcs - set VCN block interrupt irq functions * @@ -1834,6 +1854,9 @@ static void vcn_v4_0_3_set_irq_funcs(struct amdgpu_device *adev) adev->vcn.inst->irq.num_types++; } adev->vcn.inst->irq.funcs = &vcn_v4_0_3_irq_funcs; + + adev->vcn.inst->ras_poison_irq.num_types = 1; + adev->vcn.inst->ras_poison_irq.funcs = &vcn_v4_0_3_ras_irq_funcs; } static void vcn_v4_0_3_print_ip_state(struct amdgpu_ip_block *ip_block, struct drm_printer *p) @@ -1981,9 +2004,44 @@ static void vcn_v4_0_3_reset_ras_error_count(struct amdgpu_device *adev) vcn_v4_0_3_inst_reset_ras_error_count(adev, i); } +static uint32_t vcn_v4_0_3_query_poison_by_instance(struct amdgpu_device *adev, + uint32_t instance, uint32_t sub_block) +{ + uint32_t poison_stat = 0, reg_value = 0; + + switch (sub_block) { + case AMDGPU_VCN_V4_0_3_VCPU_VCODEC: + reg_value = RREG32_SOC15(VCN, instance, regUVD_RAS_VCPU_VCODEC_STATUS); + poison_stat = REG_GET_FIELD(reg_value, UVD_RAS_VCPU_VCODEC_STATUS, POISONED_PF); + break; + default: + break; + } + + if (poison_stat) + dev_info(adev->dev, "Poison detected in VCN%d, sub_block%d\n", + instance, sub_block); + + return poison_stat; +} + +static bool vcn_v4_0_3_query_poison_status(struct amdgpu_device *adev) +{ + uint32_t inst, sub; + uint32_t poison_stat = 0; + + for (inst = 0; inst < adev->vcn.num_vcn_inst; inst++) + for (sub = 0; sub < AMDGPU_VCN_V4_0_3_MAX_SUB_BLOCK; sub++) + poison_stat += + vcn_v4_0_3_query_poison_by_instance(adev, inst, sub); + + return !!poison_stat; +} + static const struct amdgpu_ras_block_hw_ops vcn_v4_0_3_ras_hw_ops = { .query_ras_error_count = vcn_v4_0_3_query_ras_error_count, .reset_ras_error_count = vcn_v4_0_3_reset_ras_error_count, + .query_poison_status = vcn_v4_0_3_query_poison_status, }; static int vcn_v4_0_3_aca_bank_parser(struct aca_handle *handle, struct aca_bank *bank, @@ -2059,6 +2117,13 @@ static int vcn_v4_0_3_ras_late_init(struct amdgpu_device *adev, struct ras_commo if (r) return r; + if (amdgpu_ras_is_supported(adev, ras_block->block) && + adev->vcn.inst->ras_poison_irq.funcs) { + r = amdgpu_irq_get(adev, &adev->vcn.inst->ras_poison_irq, 0); + if (r) + goto late_fini; + } + r = amdgpu_ras_bind_aca(adev, AMDGPU_RAS_BLOCK__VCN, &vcn_v4_0_3_aca_info, NULL); if (r) diff --git a/drivers/gpu/drm/amd/amdgpu/vcn_v4_0_3.h b/drivers/gpu/drm/amd/amdgpu/vcn_v4_0_3.h index 03572a1d0c9c..aeab89853a92 100644 --- a/drivers/gpu/drm/amd/amdgpu/vcn_v4_0_3.h +++ b/drivers/gpu/drm/amd/amdgpu/vcn_v4_0_3.h @@ -24,6 +24,12 @@ #ifndef __VCN_V4_0_3_H__ #define __VCN_V4_0_3_H__ +enum amdgpu_vcn_v4_0_3_sub_block { + AMDGPU_VCN_V4_0_3_VCPU_VCODEC = 0, + + AMDGPU_VCN_V4_0_3_MAX_SUB_BLOCK, +}; + extern const struct amdgpu_ip_block_version vcn_v4_0_3_ip_block; void vcn_v4_0_3_enc_ring_emit_reg_wait(struct amdgpu_ring *ring, uint32_t reg, -- cgit v1.2.3 From 1b2231de416312f96e29b9761bd06de1d637492e Mon Sep 17 00:00:00 2001 From: "Stanley.Yang" Date: Tue, 13 May 2025 20:10:05 +0800 Subject: drm/amdgpu: Register aqua vanjaram jpeg poison irq Register aqua vanjaram jpeg poison irq, add jpeg poison handle. Signed-off-by: Stanley.Yang Reviewed-by: Hawking Zhang Signed-off-by: Alex Deucher --- drivers/gpu/drm/amd/amdgpu/jpeg_v4_0_3.c | 76 ++++++++++++++++++++++++++++++++ drivers/gpu/drm/amd/amdgpu/jpeg_v4_0_3.h | 7 +++ 2 files changed, 83 insertions(+) (limited to 'drivers/gpu') diff --git a/drivers/gpu/drm/amd/amdgpu/jpeg_v4_0_3.c b/drivers/gpu/drm/amd/amdgpu/jpeg_v4_0_3.c index a8ccae361ec7..79e342d5ab28 100644 --- a/drivers/gpu/drm/amd/amdgpu/jpeg_v4_0_3.c +++ b/drivers/gpu/drm/amd/amdgpu/jpeg_v4_0_3.c @@ -149,6 +149,18 @@ static int jpeg_v4_0_3_sw_init(struct amdgpu_ip_block *ip_block) return r; } + /* JPEG DJPEG POISON EVENT */ + r = amdgpu_irq_add_id(adev, SOC15_IH_CLIENTID_VCN, + VCN_4_0__SRCID_DJPEG0_POISON, &adev->jpeg.inst->ras_poison_irq); + if (r) + return r; + + /* JPEG EJPEG POISON EVENT */ + r = amdgpu_irq_add_id(adev, SOC15_IH_CLIENTID_VCN, + VCN_4_0__SRCID_EJPEG0_POISON, &adev->jpeg.inst->ras_poison_irq); + if (r) + return r; + r = amdgpu_jpeg_sw_init(adev); if (r) return r; @@ -434,6 +446,9 @@ static int jpeg_v4_0_3_hw_fini(struct amdgpu_ip_block *ip_block) ret = jpeg_v4_0_3_set_powergating_state(ip_block, AMD_PG_STATE_GATE); } + if (amdgpu_ras_is_supported(adev, AMDGPU_RAS_BLOCK__JPEG)) + amdgpu_irq_put(adev, &adev->jpeg.inst->ras_poison_irq, 0); + return ret; } @@ -1041,6 +1056,14 @@ static int jpeg_v4_0_3_set_interrupt_state(struct amdgpu_device *adev, return 0; } +static int jpeg_v4_0_3_set_ras_interrupt_state(struct amdgpu_device *adev, + struct amdgpu_irq_src *source, + unsigned int type, + enum amdgpu_interrupt_state state) +{ + return 0; +} + static int jpeg_v4_0_3_process_interrupt(struct amdgpu_device *adev, struct amdgpu_irq_src *source, struct amdgpu_iv_entry *entry) @@ -1200,6 +1223,11 @@ static const struct amdgpu_irq_src_funcs jpeg_v4_0_3_irq_funcs = { .process = jpeg_v4_0_3_process_interrupt, }; +static const struct amdgpu_irq_src_funcs jpeg_v4_0_3_ras_irq_funcs = { + .set = jpeg_v4_0_3_set_ras_interrupt_state, + .process = amdgpu_jpeg_process_poison_irq, +}; + static void jpeg_v4_0_3_set_irq_funcs(struct amdgpu_device *adev) { int i; @@ -1208,6 +1236,9 @@ static void jpeg_v4_0_3_set_irq_funcs(struct amdgpu_device *adev) adev->jpeg.inst->irq.num_types += adev->jpeg.num_jpeg_rings; } adev->jpeg.inst->irq.funcs = &jpeg_v4_0_3_irq_funcs; + + adev->jpeg.inst->ras_poison_irq.num_types = 1; + adev->jpeg.inst->ras_poison_irq.funcs = &jpeg_v4_0_3_ras_irq_funcs; } const struct amdgpu_ip_block_version jpeg_v4_0_3_ip_block = { @@ -1304,9 +1335,47 @@ static void jpeg_v4_0_3_reset_ras_error_count(struct amdgpu_device *adev) jpeg_v4_0_3_inst_reset_ras_error_count(adev, i); } +static uint32_t jpeg_v4_0_3_query_poison_by_instance(struct amdgpu_device *adev, + uint32_t instance, uint32_t sub_block) +{ + uint32_t poison_stat = 0, reg_value = 0; + + switch (sub_block) { + case AMDGPU_JPEG_V4_0_3_JPEG0: + reg_value = RREG32_SOC15(JPEG, instance, regUVD_RAS_JPEG0_STATUS); + poison_stat = REG_GET_FIELD(reg_value, UVD_RAS_JPEG0_STATUS, POISONED_PF); + break; + case AMDGPU_JPEG_V4_0_3_JPEG1: + reg_value = RREG32_SOC15(JPEG, instance, regUVD_RAS_JPEG1_STATUS); + poison_stat = REG_GET_FIELD(reg_value, UVD_RAS_JPEG1_STATUS, POISONED_PF); + break; + default: + break; + } + + if (poison_stat) + dev_info(adev->dev, "Poison detected in JPEG%d sub_block%d\n", + instance, sub_block); + + return poison_stat; +} + +static bool jpeg_v4_0_3_query_ras_poison_status(struct amdgpu_device *adev) +{ + uint32_t inst = 0, sub = 0, poison_stat = 0; + + for (inst = 0; inst < adev->jpeg.num_jpeg_inst; inst++) + for (sub = 0; sub < AMDGPU_JPEG_V4_0_3_MAX_SUB_BLOCK; sub++) + poison_stat += + jpeg_v4_0_3_query_poison_by_instance(adev, inst, sub); + + return !!poison_stat; +} + static const struct amdgpu_ras_block_hw_ops jpeg_v4_0_3_ras_hw_ops = { .query_ras_error_count = jpeg_v4_0_3_query_ras_error_count, .reset_ras_error_count = jpeg_v4_0_3_reset_ras_error_count, + .query_poison_status = jpeg_v4_0_3_query_ras_poison_status, }; static int jpeg_v4_0_3_aca_bank_parser(struct aca_handle *handle, struct aca_bank *bank, @@ -1383,6 +1452,13 @@ static int jpeg_v4_0_3_ras_late_init(struct amdgpu_device *adev, struct ras_comm if (r) return r; + if (amdgpu_ras_is_supported(adev, ras_block->block) && + adev->jpeg.inst->ras_poison_irq.funcs) { + r = amdgpu_irq_get(adev, &adev->jpeg.inst->ras_poison_irq, 0); + if (r) + goto late_fini; + } + r = amdgpu_ras_bind_aca(adev, AMDGPU_RAS_BLOCK__JPEG, &jpeg_v4_0_3_aca_info, NULL); if (r) diff --git a/drivers/gpu/drm/amd/amdgpu/jpeg_v4_0_3.h b/drivers/gpu/drm/amd/amdgpu/jpeg_v4_0_3.h index a90bf370a002..2e110d04af84 100644 --- a/drivers/gpu/drm/amd/amdgpu/jpeg_v4_0_3.h +++ b/drivers/gpu/drm/amd/amdgpu/jpeg_v4_0_3.h @@ -46,6 +46,13 @@ #define JRBC_DEC_EXTERNAL_REG_WRITE_ADDR 0x18000 +enum amdgpu_jpeg_v4_0_3_sub_block { + AMDGPU_JPEG_V4_0_3_JPEG0 = 0, + AMDGPU_JPEG_V4_0_3_JPEG1, + + AMDGPU_JPEG_V4_0_3_MAX_SUB_BLOCK, +}; + extern const struct amdgpu_ip_block_version jpeg_v4_0_3_ip_block; void jpeg_v4_0_3_dec_ring_emit_ib(struct amdgpu_ring *ring, -- cgit v1.2.3 From 54a01f775194a8b5acc6bd735aa0d092469dcff7 Mon Sep 17 00:00:00 2001 From: Lijo Lazar Date: Mon, 5 May 2025 16:53:29 +0530 Subject: drm/amd/pm: Add support to query partition metrics Add interfaces to query compute partition related metrics data. Signed-off-by: Lijo Lazar Reviewed-by: Asad Kamal Reviewed-by: Hawking Zhang Signed-off-by: Alex Deucher --- drivers/gpu/drm/amd/include/kgd_pp_interface.h | 24 +++++++++++++++++++ drivers/gpu/drm/amd/pm/amdgpu_dpm.c | 32 ++++++++++++++++++++++++++ drivers/gpu/drm/amd/pm/inc/amdgpu_dpm.h | 2 ++ drivers/gpu/drm/amd/pm/swsmu/amdgpu_smu.c | 14 +++++++++++ drivers/gpu/drm/amd/pm/swsmu/inc/amdgpu_smu.h | 6 +++++ drivers/gpu/drm/amd/pm/swsmu/smu_cmn.h | 12 ++++++++++ 6 files changed, 90 insertions(+) (limited to 'drivers/gpu') diff --git a/drivers/gpu/drm/amd/include/kgd_pp_interface.h b/drivers/gpu/drm/amd/include/kgd_pp_interface.h index 0f7542d7074b..f4d914dc731f 100644 --- a/drivers/gpu/drm/amd/include/kgd_pp_interface.h +++ b/drivers/gpu/drm/amd/include/kgd_pp_interface.h @@ -494,6 +494,7 @@ struct amd_pm_funcs { int (*set_df_cstate)(void *handle, enum pp_df_cstate state); int (*set_xgmi_pstate)(void *handle, uint32_t pstate); ssize_t (*get_gpu_metrics)(void *handle, void **table); + ssize_t (*get_xcp_metrics)(void *handle, int xcp_id, void *table); ssize_t (*get_pm_metrics)(void *handle, void *pmmetrics, size_t size); int (*set_watermarks_for_clock_ranges)(void *handle, struct pp_smu_wm_range_sets *ranges); @@ -1592,4 +1593,27 @@ struct amdgpu_pm_metrics { uint8_t data[]; }; +struct amdgpu_partition_metrics_v1_0 { + struct metrics_table_header common_header; + /* Current clocks (Mhz) */ + uint16_t current_gfxclk[MAX_XCC]; + uint16_t current_socclk[MAX_CLKS]; + uint16_t current_vclk0[MAX_CLKS]; + uint16_t current_dclk0[MAX_CLKS]; + uint16_t current_uclk; + uint16_t padding; + + /* Utilization Instantaneous (%) */ + uint32_t gfx_busy_inst[MAX_XCC]; + uint16_t jpeg_busy[NUM_JPEG_ENG_V1]; + uint16_t vcn_busy[NUM_VCN]; + /* Utilization Accumulated (%) */ + uint64_t gfx_busy_acc[MAX_XCC]; + /* Total App Clock Counter Accumulated */ + uint64_t gfx_below_host_limit_ppt_acc[MAX_XCC]; + uint64_t gfx_below_host_limit_thm_acc[MAX_XCC]; + uint64_t gfx_low_utilization_acc[MAX_XCC]; + uint64_t gfx_below_host_limit_total_acc[MAX_XCC]; +}; + #endif diff --git a/drivers/gpu/drm/amd/pm/amdgpu_dpm.c b/drivers/gpu/drm/amd/pm/amdgpu_dpm.c index 2148c8db5a59..d98c95d1ed83 100644 --- a/drivers/gpu/drm/amd/pm/amdgpu_dpm.c +++ b/drivers/gpu/drm/amd/pm/amdgpu_dpm.c @@ -2019,3 +2019,35 @@ int amdgpu_dpm_get_dpm_clock_table(struct amdgpu_device *adev, return ret; } + +/** + * amdgpu_dpm_get_xcp_metrics - Retrieve metrics for a specific compute + * partition + * @adev: Pointer to the device. + * @xcp_id: Identifier of the XCP for which metrics are to be retrieved. + * @table: Pointer to a buffer where the metrics will be stored. If NULL, the + * function returns the size of the metrics structure. + * + * This function retrieves metrics for a specific XCP, including details such as + * VCN/JPEG activity, clock frequencies, and other performance metrics. If the + * table parameter is NULL, the function returns the size of the metrics + * structure without populating it. + * + * Return: Size of the metrics structure on success, or a negative error code on failure. + */ +ssize_t amdgpu_dpm_get_xcp_metrics(struct amdgpu_device *adev, int xcp_id, + void *table) +{ + const struct amd_pm_funcs *pp_funcs = adev->powerplay.pp_funcs; + int ret = 0; + + if (!pp_funcs->get_xcp_metrics) + return 0; + + mutex_lock(&adev->pm.mutex); + ret = pp_funcs->get_xcp_metrics(adev->powerplay.pp_handle, xcp_id, + table); + mutex_unlock(&adev->pm.mutex); + + return ret; +} diff --git a/drivers/gpu/drm/amd/pm/inc/amdgpu_dpm.h b/drivers/gpu/drm/amd/pm/inc/amdgpu_dpm.h index 2c3c97587dd5..c0f9ecb97fcc 100644 --- a/drivers/gpu/drm/amd/pm/inc/amdgpu_dpm.h +++ b/drivers/gpu/drm/amd/pm/inc/amdgpu_dpm.h @@ -524,6 +524,8 @@ int amdgpu_dpm_get_power_profile_mode(struct amdgpu_device *adev, int amdgpu_dpm_set_power_profile_mode(struct amdgpu_device *adev, long *input, uint32_t size); int amdgpu_dpm_get_gpu_metrics(struct amdgpu_device *adev, void **table); +ssize_t amdgpu_dpm_get_xcp_metrics(struct amdgpu_device *adev, int xcp_id, + void *table); /** * @get_pm_metrics: Get one snapshot of power management metrics from PMFW. The diff --git a/drivers/gpu/drm/amd/pm/swsmu/amdgpu_smu.c b/drivers/gpu/drm/amd/pm/swsmu/amdgpu_smu.c index f24a1d8c77db..d79a1d94661a 100644 --- a/drivers/gpu/drm/amd/pm/swsmu/amdgpu_smu.c +++ b/drivers/gpu/drm/amd/pm/swsmu/amdgpu_smu.c @@ -3758,6 +3758,19 @@ int smu_set_pm_policy(struct smu_context *smu, enum pp_pm_policy p_type, return ret; } +static ssize_t smu_sys_get_xcp_metrics(void *handle, int xcp_id, void *table) +{ + struct smu_context *smu = handle; + + if (!smu->pm_enabled || !smu->adev->pm.dpm_enabled) + return -EOPNOTSUPP; + + if (!smu->adev->xcp_mgr || !smu->ppt_funcs->get_xcp_metrics) + return -EOPNOTSUPP; + + return smu->ppt_funcs->get_xcp_metrics(smu, xcp_id, table); +} + static const struct amd_pm_funcs swsmu_pm_funcs = { /* export for sysfs */ .set_fan_control_mode = smu_set_fan_control_mode, @@ -3816,6 +3829,7 @@ static const struct amd_pm_funcs swsmu_pm_funcs = { .get_uclk_dpm_states = smu_get_uclk_dpm_states, .get_dpm_clock_table = smu_get_dpm_clock_table, .get_smu_prv_buf_details = smu_get_prv_buffer_details, + .get_xcp_metrics = smu_sys_get_xcp_metrics, }; int smu_wait_for_event(struct smu_context *smu, enum smu_event_type event, diff --git a/drivers/gpu/drm/amd/pm/swsmu/inc/amdgpu_smu.h b/drivers/gpu/drm/amd/pm/swsmu/inc/amdgpu_smu.h index d47e32ae4671..9aacc7bc1c69 100644 --- a/drivers/gpu/drm/amd/pm/swsmu/inc/amdgpu_smu.h +++ b/drivers/gpu/drm/amd/pm/swsmu/inc/amdgpu_smu.h @@ -1466,6 +1466,12 @@ struct pptable_funcs { */ int (*set_wbrf_exclusion_ranges)(struct smu_context *smu, struct freq_band_range *exclusion_ranges); + /** + * @get_xcp_metrics: Get a copy of the partition metrics table from SMU. + * Return: Size of table + */ + ssize_t (*get_xcp_metrics)(struct smu_context *smu, int xcp_id, + void *table); }; typedef enum { diff --git a/drivers/gpu/drm/amd/pm/swsmu/smu_cmn.h b/drivers/gpu/drm/amd/pm/swsmu/smu_cmn.h index ade36a8ffa66..7473672abd2a 100644 --- a/drivers/gpu/drm/amd/pm/swsmu/smu_cmn.h +++ b/drivers/gpu/drm/amd/pm/swsmu/smu_cmn.h @@ -52,6 +52,18 @@ header->structure_size = sizeof(*(ptr)); \ } while (0) +#define smu_cmn_init_partition_metrics(ptr, frev, crev) \ + do { \ + typecheck(struct amdgpu_partition_metrics_v##frev##_##crev, \ + typeof(*(ptr))); \ + struct metrics_table_header *header = \ + (struct metrics_table_header *)(ptr); \ + memset(header, 0xFF, sizeof(*(ptr))); \ + header->format_revision = frev; \ + header->content_revision = crev; \ + header->structure_size = sizeof(*(ptr)); \ + } while (0) + extern const int link_speed[]; /* Helper to Convert from PCIE Gen 1/2/3/4/5/6 to 0.1 GT/s speed units */ -- cgit v1.2.3 From cbbab29246c8b22c3d6285756dd14992cbaa2d5e Mon Sep 17 00:00:00 2001 From: Lijo Lazar Date: Mon, 5 May 2025 15:18:06 +0530 Subject: drm/amdgpu: Add sysfs nodes for partition Add sysfs nodes to provide compute paritition specific data. Signed-off-by: Lijo Lazar Reviewed-by: Asad Kamal Reviewed-by: Hawking Zhang Signed-off-by: Alex Deucher --- drivers/gpu/drm/amd/amdgpu/amdgpu_device.c | 4 +- drivers/gpu/drm/amd/amdgpu/amdgpu_xcp.c | 146 +++++++++++++++++++++++++++-- drivers/gpu/drm/amd/amdgpu/amdgpu_xcp.h | 6 +- 3 files changed, 146 insertions(+), 10 deletions(-) (limited to 'drivers/gpu') diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_device.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_device.c index 4d1b54f58495..908e39dae4b4 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_device.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_device.c @@ -4728,7 +4728,7 @@ fence_driver_init: amdgpu_fru_sysfs_init(adev); amdgpu_reg_state_sysfs_init(adev); - amdgpu_xcp_cfg_sysfs_init(adev); + amdgpu_xcp_sysfs_init(adev); if (IS_ENABLED(CONFIG_PERF_EVENTS)) r = amdgpu_pmu_init(adev); @@ -4858,7 +4858,7 @@ void amdgpu_device_fini_hw(struct amdgpu_device *adev) amdgpu_fru_sysfs_fini(adev); amdgpu_reg_state_sysfs_fini(adev); - amdgpu_xcp_cfg_sysfs_fini(adev); + amdgpu_xcp_sysfs_fini(adev); /* disable ras feature must before hw fini */ amdgpu_ras_pre_fini(adev); diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_xcp.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_xcp.c index b03c3895897b..322816805bfb 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_xcp.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_xcp.c @@ -27,6 +27,9 @@ #include #include "../amdxcp/amdgpu_xcp_drv.h" +static void amdgpu_xcp_sysfs_entries_init(struct amdgpu_xcp_mgr *xcp_mgr); +static void amdgpu_xcp_sysfs_entries_update(struct amdgpu_xcp_mgr *xcp_mgr); + static int __amdgpu_xcp_run(struct amdgpu_xcp_mgr *xcp_mgr, struct amdgpu_xcp_ip *xcp_ip, int xcp_state) { @@ -189,7 +192,7 @@ static int __amdgpu_xcp_switch_partition_mode(struct amdgpu_xcp_mgr *xcp_mgr, goto out; } - + amdgpu_xcp_sysfs_entries_update(xcp_mgr); out: mutex_unlock(&xcp_mgr->xcp_lock); @@ -263,9 +266,10 @@ static int amdgpu_xcp_dev_alloc(struct amdgpu_device *adev) if (ret == -ENOSPC) { dev_warn(adev->dev, "Skip xcp node #%d when out of drm node resource.", i); - return 0; + ret = 0; + goto out; } else if (ret) { - return ret; + goto out; } /* Redirect all IOCTLs to the primary device */ @@ -278,9 +282,14 @@ static int amdgpu_xcp_dev_alloc(struct amdgpu_device *adev) p_ddev->vma_offset_manager = ddev->vma_offset_manager; p_ddev->driver = &amdgpu_partition_driver; adev->xcp_mgr->xcp[i].ddev = p_ddev; + + dev_set_drvdata(p_ddev->dev, &adev->xcp_mgr->xcp[i]); } + ret = 0; +out: + amdgpu_xcp_sysfs_entries_init(adev->xcp_mgr); - return 0; + return ret; } int amdgpu_xcp_mgr_init(struct amdgpu_device *adev, int init_mode, @@ -288,6 +297,7 @@ int amdgpu_xcp_mgr_init(struct amdgpu_device *adev, int init_mode, struct amdgpu_xcp_mgr_funcs *xcp_funcs) { struct amdgpu_xcp_mgr *xcp_mgr; + int i; if (!xcp_funcs || !xcp_funcs->get_ip_details) return -EINVAL; @@ -306,6 +316,8 @@ int amdgpu_xcp_mgr_init(struct amdgpu_device *adev, int init_mode, amdgpu_xcp_init(xcp_mgr, init_num_xcps, init_mode); adev->xcp_mgr = xcp_mgr; + for (i = 0; i < MAX_XCP; ++i) + xcp_mgr->xcp[i].xcp_mgr = xcp_mgr; return amdgpu_xcp_dev_alloc(adev); } @@ -433,6 +445,7 @@ void amdgpu_xcp_release_sched(struct amdgpu_device *adev, } } +/*====================== xcp sysfs - configuration ======================*/ #define XCP_CFG_SYSFS_RES_ATTR_SHOW(_name) \ static ssize_t amdgpu_xcp_res_sysfs_##_name##_show( \ struct amdgpu_xcp_res_details *xcp_res, char *buf) \ @@ -635,7 +648,7 @@ static const struct attribute *xcp_attrs[] = { NULL, }; -void amdgpu_xcp_cfg_sysfs_init(struct amdgpu_device *adev) +static void amdgpu_xcp_cfg_sysfs_init(struct amdgpu_device *adev) { struct amdgpu_xcp_res_details *xcp_res; struct amdgpu_xcp_cfg *xcp_cfg; @@ -703,7 +716,7 @@ err1: kobject_put(&xcp_cfg->kobj); } -void amdgpu_xcp_cfg_sysfs_fini(struct amdgpu_device *adev) +static void amdgpu_xcp_cfg_sysfs_fini(struct amdgpu_device *adev) { struct amdgpu_xcp_res_details *xcp_res; struct amdgpu_xcp_cfg *xcp_cfg; @@ -722,3 +735,124 @@ void amdgpu_xcp_cfg_sysfs_fini(struct amdgpu_device *adev) sysfs_remove_files(&xcp_cfg->kobj, xcp_attrs); kobject_put(&xcp_cfg->kobj); } + +/*====================== xcp sysfs - data entries ======================*/ + +#define to_xcp(x) container_of(x, struct amdgpu_xcp, kobj) + +static ssize_t xcp_metrics_show(struct kobject *kobj, + struct kobj_attribute *attr, char *buf) +{ + struct amdgpu_xcp *xcp = to_xcp(kobj); + struct amdgpu_xcp_mgr *xcp_mgr; + ssize_t size; + + xcp_mgr = xcp->xcp_mgr; + size = amdgpu_dpm_get_xcp_metrics(xcp_mgr->adev, xcp->id, NULL); + if (size <= 0) + return size; + + if (size > PAGE_SIZE) + return -ENOSPC; + + return amdgpu_dpm_get_xcp_metrics(xcp_mgr->adev, xcp->id, buf); +} + +static umode_t amdgpu_xcp_attrs_is_visible(struct kobject *kobj, + struct attribute *attr, int n) +{ + struct amdgpu_xcp *xcp = to_xcp(kobj); + + if (!xcp || !xcp->valid) + return 0; + + return attr->mode; +} + +static struct kobj_attribute xcp_sysfs_metrics = __ATTR_RO(xcp_metrics); + +static struct attribute *amdgpu_xcp_attrs[] = { + &xcp_sysfs_metrics.attr, + NULL, +}; + +static const struct attribute_group amdgpu_xcp_attrs_group = { + .attrs = amdgpu_xcp_attrs, + .is_visible = amdgpu_xcp_attrs_is_visible +}; + +static const struct kobj_type xcp_sysfs_ktype = { + .sysfs_ops = &kobj_sysfs_ops, +}; + +static void amdgpu_xcp_sysfs_entries_fini(struct amdgpu_xcp_mgr *xcp_mgr, int n) +{ + struct amdgpu_xcp *xcp; + + for (n--; n >= 0; n--) { + xcp = &xcp_mgr->xcp[n]; + if (!xcp->ddev || !xcp->valid) + continue; + sysfs_remove_group(&xcp->kobj, &amdgpu_xcp_attrs_group); + kobject_put(&xcp->kobj); + } +} + +static void amdgpu_xcp_sysfs_entries_init(struct amdgpu_xcp_mgr *xcp_mgr) +{ + struct amdgpu_xcp *xcp; + int i, r; + + for (i = 0; i < MAX_XCP; i++) { + /* Redirect all IOCTLs to the primary device */ + xcp = &xcp_mgr->xcp[i]; + if (!xcp->ddev) + break; + r = kobject_init_and_add(&xcp->kobj, &xcp_sysfs_ktype, + &xcp->ddev->dev->kobj, "xcp"); + if (r) + goto out; + + r = sysfs_create_group(&xcp->kobj, &amdgpu_xcp_attrs_group); + if (r) + goto out; + } + + return; +out: + kobject_put(&xcp->kobj); +} + +static void amdgpu_xcp_sysfs_entries_update(struct amdgpu_xcp_mgr *xcp_mgr) +{ + struct amdgpu_xcp *xcp; + int i; + + for (i = 0; i < MAX_XCP; i++) { + /* Redirect all IOCTLs to the primary device */ + xcp = &xcp_mgr->xcp[i]; + if (!xcp->ddev) + continue; + sysfs_update_group(&xcp->kobj, &amdgpu_xcp_attrs_group); + } + + return; +} + +void amdgpu_xcp_sysfs_init(struct amdgpu_device *adev) +{ + if (!adev->xcp_mgr) + return; + + amdgpu_xcp_cfg_sysfs_init(adev); + + return; +} + +void amdgpu_xcp_sysfs_fini(struct amdgpu_device *adev) +{ + if (!adev->xcp_mgr) + return; + amdgpu_xcp_sysfs_entries_fini(adev->xcp_mgr, MAX_XCP); + amdgpu_xcp_cfg_sysfs_fini(adev); +} diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_xcp.h b/drivers/gpu/drm/amd/amdgpu/amdgpu_xcp.h index b63f53242c57..454b33f889fb 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_xcp.h +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_xcp.h @@ -108,6 +108,8 @@ struct amdgpu_xcp { struct drm_driver *driver; struct drm_vma_offset_manager *vma_offset_manager; struct amdgpu_sched gpu_sched[AMDGPU_HW_IP_NUM][AMDGPU_RING_PRIO_MAX]; + struct amdgpu_xcp_mgr *xcp_mgr; + struct kobject kobj; }; struct amdgpu_xcp_mgr { @@ -175,8 +177,8 @@ int amdgpu_xcp_open_device(struct amdgpu_device *adev, void amdgpu_xcp_release_sched(struct amdgpu_device *adev, struct amdgpu_ctx_entity *entity); -void amdgpu_xcp_cfg_sysfs_init(struct amdgpu_device *adev); -void amdgpu_xcp_cfg_sysfs_fini(struct amdgpu_device *adev); +void amdgpu_xcp_sysfs_init(struct amdgpu_device *adev); +void amdgpu_xcp_sysfs_fini(struct amdgpu_device *adev); #define amdgpu_xcp_select_scheds(adev, e, c, d, x, y) \ ((adev)->xcp_mgr && (adev)->xcp_mgr->funcs && \ -- cgit v1.2.3 From 4cd694fdbf453a6fe875ad6236b04d7b9a5e49db Mon Sep 17 00:00:00 2001 From: Lijo Lazar Date: Mon, 5 May 2025 16:55:34 +0530 Subject: drm/amd/pm: Fetch partition metrics on SMUv13.0.6 Add support to fetch compute partition related metrics in SMUv13.0.6 SOCs. Signed-off-by: Lijo Lazar Reviewed-by: Asad Kamal Reviewed-by: Hawking Zhang Signed-off-by: Alex Deucher --- .../gpu/drm/amd/pm/swsmu/smu13/smu_v13_0_6_ppt.c | 112 +++++++++++++++++++++ 1 file changed, 112 insertions(+) (limited to 'drivers/gpu') diff --git a/drivers/gpu/drm/amd/pm/swsmu/smu13/smu_v13_0_6_ppt.c b/drivers/gpu/drm/amd/pm/swsmu/smu13/smu_v13_0_6_ppt.c index cd9ed3b5e9fa..615fd3771ae3 100644 --- a/drivers/gpu/drm/amd/pm/swsmu/smu13/smu_v13_0_6_ppt.c +++ b/drivers/gpu/drm/amd/pm/swsmu/smu13/smu_v13_0_6_ppt.c @@ -2536,6 +2536,117 @@ static int smu_v13_0_6_get_current_pcie_link_speed(struct smu_context *smu) return pcie_gen_to_speed(speed_level + 1); } +static ssize_t smu_v13_0_6_get_xcp_metrics(struct smu_context *smu, int xcp_id, + void *table) +{ + const u8 num_jpeg_rings = AMDGPU_MAX_JPEG_RINGS_4_0_3; + int version = smu_v13_0_6_get_metrics_version(smu); + struct amdgpu_partition_metrics_v1_0 *xcp_metrics; + struct amdgpu_device *adev = smu->adev; + int ret, inst, i, j, k, idx; + MetricsTableV0_t *metrics_v0; + MetricsTableV1_t *metrics_v1; + MetricsTableV2_t *metrics_v2; + struct amdgpu_xcp *xcp; + u32 inst_mask; + bool per_inst; + + if (!table) + return sizeof(*xcp_metrics); + + for_each_xcp(adev->xcp_mgr, xcp, i) { + if (xcp->id == xcp_id) + break; + } + if (i == adev->xcp_mgr->num_xcps) + return -EINVAL; + + xcp_metrics = (struct amdgpu_partition_metrics_v1_0 *)table; + smu_cmn_init_partition_metrics(xcp_metrics, 1, 0); + + metrics_v0 = kzalloc(METRICS_TABLE_SIZE, GFP_KERNEL); + if (!metrics_v0) + return -ENOMEM; + + ret = smu_v13_0_6_get_metrics_table(smu, metrics_v0, false); + if (ret) { + kfree(metrics_v0); + return ret; + } + metrics_v1 = (MetricsTableV1_t *)metrics_v0; + metrics_v2 = (MetricsTableV2_t *)metrics_v0; + + per_inst = smu_v13_0_6_cap_supported(smu, SMU_CAP(PER_INST_METRICS)); + + amdgpu_xcp_get_inst_details(xcp, AMDGPU_XCP_VCN, &inst_mask); + idx = 0; + for_each_inst(k, inst_mask) { + /* Both JPEG and VCN has same instances */ + inst = GET_INST(VCN, k); + + for (j = 0; j < num_jpeg_rings; ++j) { + xcp_metrics->jpeg_busy[(idx * num_jpeg_rings) + j] = + SMUQ10_ROUND(GET_METRIC_FIELD( + JpegBusy, + version)[(inst * num_jpeg_rings) + j]); + } + xcp_metrics->vcn_busy[idx] = + SMUQ10_ROUND(GET_METRIC_FIELD(VcnBusy, version)[inst]); + + xcp_metrics->current_vclk0[idx] = SMUQ10_ROUND( + GET_METRIC_FIELD(VclkFrequency, version)[inst]); + xcp_metrics->current_dclk0[idx] = SMUQ10_ROUND( + GET_METRIC_FIELD(DclkFrequency, version)[inst]); + xcp_metrics->current_socclk[idx] = SMUQ10_ROUND( + GET_METRIC_FIELD(SocclkFrequency, version)[inst]); + + idx++; + } + + xcp_metrics->current_uclk = + SMUQ10_ROUND(GET_METRIC_FIELD(UclkFrequency, version)); + + if (per_inst) { + amdgpu_xcp_get_inst_details(xcp, AMDGPU_XCP_GFX, &inst_mask); + idx = 0; + for_each_inst(k, inst_mask) { + inst = GET_INST(GC, k); + xcp_metrics->current_gfxclk[idx] = + SMUQ10_ROUND(GET_METRIC_FIELD(GfxclkFrequency, + version)[inst]); + + xcp_metrics->gfx_busy_inst[idx] = SMUQ10_ROUND( + GET_GPU_METRIC_FIELD(GfxBusy, version)[inst]); + xcp_metrics->gfx_busy_acc[idx] = SMUQ10_ROUND( + GET_GPU_METRIC_FIELD(GfxBusyAcc, + version)[inst]); + if (smu_v13_0_6_cap_supported( + smu, SMU_CAP(HST_LIMIT_METRICS))) { + xcp_metrics->gfx_below_host_limit_ppt_acc + [idx] = SMUQ10_ROUND( + metrics_v0->GfxclkBelowHostLimitPptAcc + [inst]); + xcp_metrics->gfx_below_host_limit_thm_acc + [idx] = SMUQ10_ROUND( + metrics_v0->GfxclkBelowHostLimitThmAcc + [inst]); + xcp_metrics->gfx_low_utilization_acc + [idx] = SMUQ10_ROUND( + metrics_v0 + ->GfxclkLowUtilizationAcc[inst]); + xcp_metrics->gfx_below_host_limit_total_acc + [idx] = SMUQ10_ROUND( + metrics_v0->GfxclkBelowHostLimitTotalAcc + [inst]); + } + idx++; + } + } + kfree(metrics_v0); + + return sizeof(*xcp_metrics); +} + static ssize_t smu_v13_0_6_get_gpu_metrics(struct smu_context *smu, void **table) { struct smu_table_context *smu_table = &smu->smu_table; @@ -3680,6 +3791,7 @@ static const struct pptable_funcs smu_v13_0_6_ppt_funcs = { .get_pp_feature_mask = smu_cmn_get_pp_feature_mask, .get_gpu_metrics = smu_v13_0_6_get_gpu_metrics, .get_pm_metrics = smu_v13_0_6_get_pm_metrics, + .get_xcp_metrics = smu_v13_0_6_get_xcp_metrics, .get_thermal_temperature_range = smu_v13_0_6_get_thermal_temperature_range, .mode1_reset_is_support = smu_v13_0_6_is_mode1_reset_supported, .link_reset_is_support = smu_v13_0_6_is_link_reset_supported, -- cgit v1.2.3 From f0e9c6f67a02eea72aab9f598b884395876b971b Mon Sep 17 00:00:00 2001 From: Leonardo Gomes Date: Sat, 17 May 2025 11:20:54 -0300 Subject: drm/amd/display: Adjust get_value function with prefix to help in ftrace Adjust get_value function in hw_hpd.c file to have prefix to help in ftrace, the name change from 'get_value' to 'dal_hw_hpd_get_value' Signed-off-by: Leonardo da Silva Gomes Co-developed-by: Derick Frias Signed-off-by: Derick Frias Signed-off-by: Alex Deucher --- drivers/gpu/drm/amd/display/dc/gpio/hw_hpd.c | 4 ++-- 1 file changed, 2 insertions(+), 2 deletions(-) (limited to 'drivers/gpu') diff --git a/drivers/gpu/drm/amd/display/dc/gpio/hw_hpd.c b/drivers/gpu/drm/amd/display/dc/gpio/hw_hpd.c index 3f13a744d07d..886dd05b012f 100644 --- a/drivers/gpu/drm/amd/display/dc/gpio/hw_hpd.c +++ b/drivers/gpu/drm/amd/display/dc/gpio/hw_hpd.c @@ -62,7 +62,7 @@ static void dal_hw_hpd_destroy( *ptr = NULL; } -static enum gpio_result get_value( +static enum gpio_result dal_hw_hpd_get_value( const struct hw_gpio_pin *ptr, uint32_t *value) { @@ -104,7 +104,7 @@ static enum gpio_result set_config( static const struct hw_gpio_pin_funcs funcs = { .destroy = dal_hw_hpd_destroy, .open = dal_hw_gpio_open, - .get_value = get_value, + .get_value = dal_hw_hpd_get_value, .set_value = dal_hw_gpio_set_value, .set_config = set_config, .change_mode = dal_hw_gpio_change_mode, -- cgit v1.2.3 From 57d4100cc94fe5a8dcabd72a66ed1d5c6f14291a Mon Sep 17 00:00:00 2001 From: Leonardo Gomes Date: Sat, 17 May 2025 11:20:55 -0300 Subject: drm/amd/display: Adjust set_value function with prefix to help in ftrace Adjust set_value function in hw_hpd.c file to have prefix to help in ftrace, the name change from 'set_value' to 'dal_hw_hpd_set_value' Signed-off-by: Leonardo da Silva Gomes Co-developed-by: Derick Frias Signed-off-by: Derick Frias Signed-off-by: Alex Deucher --- drivers/gpu/drm/amd/display/dc/gpio/hw_hpd.c | 4 ++-- 1 file changed, 2 insertions(+), 2 deletions(-) (limited to 'drivers/gpu') diff --git a/drivers/gpu/drm/amd/display/dc/gpio/hw_hpd.c b/drivers/gpu/drm/amd/display/dc/gpio/hw_hpd.c index 886dd05b012f..01ec451004f7 100644 --- a/drivers/gpu/drm/amd/display/dc/gpio/hw_hpd.c +++ b/drivers/gpu/drm/amd/display/dc/gpio/hw_hpd.c @@ -85,7 +85,7 @@ static enum gpio_result dal_hw_hpd_get_value( return dal_hw_gpio_get_value(ptr, value); } -static enum gpio_result set_config( +static enum gpio_result dal_hw_hpd_set_config( struct hw_gpio_pin *ptr, const struct gpio_config_data *config_data) { @@ -106,7 +106,7 @@ static const struct hw_gpio_pin_funcs funcs = { .open = dal_hw_gpio_open, .get_value = dal_hw_hpd_get_value, .set_value = dal_hw_gpio_set_value, - .set_config = set_config, + .set_config = dal_hw_hpd_set_config, .change_mode = dal_hw_gpio_change_mode, .close = dal_hw_gpio_close, }; -- cgit v1.2.3 From 05daa7c739f43aec4358f5a4c291e1b2d0fc7885 Mon Sep 17 00:00:00 2001 From: Daniil Ryabov Date: Sat, 17 May 2025 02:35:16 +0300 Subject: drm/amd/display: fix typo in comments Fix double 'u' in 'frequuency' Signed-off-by: Daniil Ryabov Signed-off-by: Alex Deucher --- drivers/gpu/drm/amd/display/dc/basics/dce_calcs.c | 4 ++-- 1 file changed, 2 insertions(+), 2 deletions(-) (limited to 'drivers/gpu') diff --git a/drivers/gpu/drm/amd/display/dc/basics/dce_calcs.c b/drivers/gpu/drm/amd/display/dc/basics/dce_calcs.c index 681799468487..d897f8a30ede 100644 --- a/drivers/gpu/drm/amd/display/dc/basics/dce_calcs.c +++ b/drivers/gpu/drm/amd/display/dc/basics/dce_calcs.c @@ -1393,7 +1393,7 @@ static void calculate_bandwidth( if ((bw_mtn(data->dram_speed_change_margin, bw_int_to_fixed(0)) && bw_ltn(data->dram_speed_change_margin, bw_int_to_fixed(9999)))) { /*determine the minimum dram clock change margin for each set of clock frequencies*/ data->min_dram_speed_change_margin[i][j] = bw_min2(data->min_dram_speed_change_margin[i][j], data->dram_speed_change_margin); - /*compute the maximum clock frequuency required for the dram clock change at each set of clock frequencies*/ + /*compute the maximum clock frequency required for the dram clock change at each set of clock frequencies*/ data->dispclk_required_for_dram_speed_change_pipe[i][j] = bw_max2(bw_div(bw_div(bw_mul(data->src_pixels_for_first_output_pixel[k], dceip->display_pipe_throughput_factor), dceip->lb_write_pixels_per_dispclk), (bw_sub(bw_sub(bw_sub(data->maximum_latency_hiding_with_cursor[k], vbios->nbp_state_change_latency), data->dmif_burst_time[i][j]), data->dram_speed_change_line_source_transfer_time[k][i][j]))), bw_div(bw_div(bw_mul(data->src_pixels_for_last_output_pixel[k], dceip->display_pipe_throughput_factor), dceip->lb_write_pixels_per_dispclk), (bw_add(bw_sub(bw_sub(bw_sub(data->maximum_latency_hiding_with_cursor[k], vbios->nbp_state_change_latency), data->dmif_burst_time[i][j]), data->dram_speed_change_line_source_transfer_time[k][i][j]), data->active_time[k])))); if ((bw_ltn(data->dispclk_required_for_dram_speed_change_pipe[i][j], vbios->high_voltage_max_dispclk))) { data->display_pstate_change_enable[k] = 1; @@ -1407,7 +1407,7 @@ static void calculate_bandwidth( if ((bw_mtn(data->dram_speed_change_margin, bw_int_to_fixed(0)) && bw_ltn(data->dram_speed_change_margin, bw_int_to_fixed(9999)))) { /*determine the minimum dram clock change margin for each display pipe*/ data->min_dram_speed_change_margin[i][j] = bw_min2(data->min_dram_speed_change_margin[i][j], data->dram_speed_change_margin); - /*compute the maximum clock frequuency required for the dram clock change at each set of clock frequencies*/ + /*compute the maximum clock frequency required for the dram clock change at each set of clock frequencies*/ data->dispclk_required_for_dram_speed_change_pipe[i][j] = bw_max2(bw_div(bw_div(bw_mul(data->src_pixels_for_first_output_pixel[k], dceip->display_pipe_throughput_factor), dceip->lb_write_pixels_per_dispclk), (bw_sub(bw_sub(bw_sub(bw_sub(data->maximum_latency_hiding_with_cursor[k], vbios->nbp_state_change_latency), data->dmif_burst_time[i][j]), data->dram_speed_change_line_source_transfer_time[k][i][j]), data->mcifwr_burst_time[i][j]))), bw_div(bw_div(bw_mul(data->src_pixels_for_last_output_pixel[k], dceip->display_pipe_throughput_factor), dceip->lb_write_pixels_per_dispclk), (bw_add(bw_sub(bw_sub(bw_sub(bw_sub(data->maximum_latency_hiding_with_cursor[k], vbios->nbp_state_change_latency), data->dmif_burst_time[i][j]), data->dram_speed_change_line_source_transfer_time[k][i][j]), data->mcifwr_burst_time[i][j]), data->active_time[k])))); if ((bw_ltn(data->dispclk_required_for_dram_speed_change_pipe[i][j], vbios->high_voltage_max_dispclk))) { data->display_pstate_change_enable[k] = 1; -- cgit v1.2.3 From ae6f4ab96be19395b8977f65670ae60d01096a83 Mon Sep 17 00:00:00 2001 From: Jihed Chaibi Date: Sat, 17 May 2025 05:06:09 +0200 Subject: drm/amdgpu: fixing typo in macro name "ENABLE" is currently misspelled in SYS_INFO_GPUCAPS__ENABEL_DFS_BYPASS PS: checkpatch.pl is complaining about the presence of a space at the start of drivers/gpu/drm/amd/include/atomfirmware.h line: 1716 This is propably because this file uses (two) spaces and not tabs. Signed-off-by: Jihed Chaibi Signed-off-by: Alex Deucher --- drivers/gpu/drm/amd/display/include/grph_object_ctrl_defs.h | 2 +- drivers/gpu/drm/amd/include/atombios.h | 4 ++-- drivers/gpu/drm/amd/include/atomfirmware.h | 2 +- drivers/gpu/drm/amd/pm/legacy-dpm/kv_dpm.c | 2 +- drivers/gpu/drm/amd/pm/powerplay/hwmgr/smu8_hwmgr.c | 2 +- 5 files changed, 6 insertions(+), 6 deletions(-) (limited to 'drivers/gpu') diff --git a/drivers/gpu/drm/amd/display/include/grph_object_ctrl_defs.h b/drivers/gpu/drm/amd/display/include/grph_object_ctrl_defs.h index 813463ffe15c..cc467031651d 100644 --- a/drivers/gpu/drm/amd/display/include/grph_object_ctrl_defs.h +++ b/drivers/gpu/drm/amd/display/include/grph_object_ctrl_defs.h @@ -424,7 +424,7 @@ struct integrated_info { /* * DFS-bypass flag */ -/* Copy of SYS_INFO_GPUCAPS__ENABEL_DFS_BYPASS from atombios.h */ +/* Copy of SYS_INFO_GPUCAPS__ENABLE_DFS_BYPASS from atombios.h */ enum { DFS_BYPASS_ENABLE = 0x10 }; diff --git a/drivers/gpu/drm/amd/include/atombios.h b/drivers/gpu/drm/amd/include/atombios.h index 52bac19fb404..b344acefc606 100644 --- a/drivers/gpu/drm/amd/include/atombios.h +++ b/drivers/gpu/drm/amd/include/atombios.h @@ -6017,7 +6017,7 @@ typedef struct _ATOM_INTEGRATED_SYSTEM_INFO_V1_7 #define SYS_INFO_GPUCAPS__TMDSHDMI_COHERENT_SINGLEPLL_MODE 0x01 #define SYS_INFO_GPUCAPS__DP_SINGLEPLL_MODE 0x02 #define SYS_INFO_GPUCAPS__DISABLE_AUX_MODE_DETECT 0x08 -#define SYS_INFO_GPUCAPS__ENABEL_DFS_BYPASS 0x10 +#define SYS_INFO_GPUCAPS__ENABLE_DFS_BYPASS 0x10 //ulGPUCapInfo[16]=1 indicate SMC firmware is able to support GNB fast resume function, so that driver can call SMC to program most of GNB register during resuming, from ML #define SYS_INFO_GPUCAPS__GNB_FAST_RESUME_CAPABLE 0x00010000 @@ -6460,7 +6460,7 @@ typedef struct _ATOM_INTEGRATED_SYSTEM_INFO_V1_9 // ulGPUCapInfo #define SYS_INFO_V1_9_GPUCAPSINFO_DISABLE_AUX_MODE_DETECT 0x08 -#define SYS_INFO_V1_9_GPUCAPSINFO_ENABEL_DFS_BYPASS 0x10 +#define SYS_INFO_V1_9_GPUCAPSINFO_ENABLE_DFS_BYPASS 0x10 //ulGPUCapInfo[16]=1 indicate SMC firmware is able to support GNB fast resume function, so that driver can call SMC to program most of GNB register during resuming, from ML #define SYS_INFO_V1_9_GPUCAPSINFO_GNB_FAST_RESUME_CAPABLE 0x00010000 //ulGPUCapInfo[18]=1 indicate the IOMMU is not available diff --git a/drivers/gpu/drm/amd/include/atomfirmware.h b/drivers/gpu/drm/amd/include/atomfirmware.h index 2d1135bdc4b9..5c86423c2e92 100644 --- a/drivers/gpu/drm/amd/include/atomfirmware.h +++ b/drivers/gpu/drm/amd/include/atomfirmware.h @@ -1714,7 +1714,7 @@ enum atom_system_vbiosmisc_def{ // gpucapinfo enum atom_system_gpucapinf_def{ - SYS_INFO_GPUCAPS__ENABEL_DFS_BYPASS = 0x10, + SYS_INFO_GPUCAPS__ENABLE_DFS_BYPASS = 0x10, }; //dpphy_override diff --git a/drivers/gpu/drm/amd/pm/legacy-dpm/kv_dpm.c b/drivers/gpu/drm/amd/pm/legacy-dpm/kv_dpm.c index 59fae668dc3f..34e71727b27d 100644 --- a/drivers/gpu/drm/amd/pm/legacy-dpm/kv_dpm.c +++ b/drivers/gpu/drm/amd/pm/legacy-dpm/kv_dpm.c @@ -2594,7 +2594,7 @@ static int kv_parse_sys_info_table(struct amdgpu_device *adev) le32_to_cpu(igp_info->info_8.ulNbpStateNClkFreq[i]); } if (le32_to_cpu(igp_info->info_8.ulGPUCapInfo) & - SYS_INFO_GPUCAPS__ENABEL_DFS_BYPASS) + SYS_INFO_GPUCAPS__ENABLE_DFS_BYPASS) pi->caps_enable_dfs_bypass = true; sumo_construct_sclk_voltage_mapping_table(adev, diff --git a/drivers/gpu/drm/amd/pm/powerplay/hwmgr/smu8_hwmgr.c b/drivers/gpu/drm/amd/pm/powerplay/hwmgr/smu8_hwmgr.c index 9d3b33446adc..9b20076e26c0 100644 --- a/drivers/gpu/drm/amd/pm/powerplay/hwmgr/smu8_hwmgr.c +++ b/drivers/gpu/drm/amd/pm/powerplay/hwmgr/smu8_hwmgr.c @@ -394,7 +394,7 @@ static int smu8_get_system_info_data(struct pp_hwmgr *hwmgr) } if (le32_to_cpu(info->ulGPUCapInfo) & - SYS_INFO_GPUCAPS__ENABEL_DFS_BYPASS) { + SYS_INFO_GPUCAPS__ENABLE_DFS_BYPASS) { phm_cap_set(hwmgr->platform_descriptor.platformCaps, PHM_PlatformCaps_EnableDFSBypass); } -- cgit v1.2.3 From 8f9f3854a14efe34824d278a6127aed54b1294fc Mon Sep 17 00:00:00 2001 From: Jihed Chaibi Date: Sat, 17 May 2025 05:09:35 +0200 Subject: drm/radeon: fixing typo in macro name "ENABLE" is currently misspelled in SYS_INFO_GPUCAPS__ENABEL_DFS_BYPASS Signed-off-by: Jihed Chaibi Signed-off-by: Alex Deucher --- drivers/gpu/drm/radeon/atombios.h | 2 +- drivers/gpu/drm/radeon/kv_dpm.c | 2 +- 2 files changed, 2 insertions(+), 2 deletions(-) (limited to 'drivers/gpu') diff --git a/drivers/gpu/drm/radeon/atombios.h b/drivers/gpu/drm/radeon/atombios.h index a7caac5b8ac8..1afa70566985 100644 --- a/drivers/gpu/drm/radeon/atombios.h +++ b/drivers/gpu/drm/radeon/atombios.h @@ -5071,7 +5071,7 @@ typedef struct _ATOM_INTEGRATED_SYSTEM_INFO_V1_7 #define SYS_INFO_GPUCAPS__TMDSHDMI_COHERENT_SINGLEPLL_MODE 0x01 #define SYS_INFO_GPUCAPS__DP_SINGLEPLL_MODE 0x02 #define SYS_INFO_GPUCAPS__DISABLE_AUX_MODE_DETECT 0x08 -#define SYS_INFO_GPUCAPS__ENABEL_DFS_BYPASS 0x10 +#define SYS_INFO_GPUCAPS__ENABLE_DFS_BYPASS 0x10 /********************************************************************************************************************** ATOM_INTEGRATED_SYSTEM_INFO_V1_7 Description diff --git a/drivers/gpu/drm/radeon/kv_dpm.c b/drivers/gpu/drm/radeon/kv_dpm.c index 55dbf450bd9c..4aa050385284 100644 --- a/drivers/gpu/drm/radeon/kv_dpm.c +++ b/drivers/gpu/drm/radeon/kv_dpm.c @@ -2329,7 +2329,7 @@ static int kv_parse_sys_info_table(struct radeon_device *rdev) le32_to_cpu(igp_info->info_8.ulNbpStateNClkFreq[i]); } if (le32_to_cpu(igp_info->info_8.ulGPUCapInfo) & - SYS_INFO_GPUCAPS__ENABEL_DFS_BYPASS) + SYS_INFO_GPUCAPS__ENABLE_DFS_BYPASS) pi->caps_enable_dfs_bypass = true; sumo_construct_sclk_voltage_mapping_table(rdev, -- cgit v1.2.3 From 74956242a0dfe4ef7fef0f9a4e8f7ea7415be97b Mon Sep 17 00:00:00 2001 From: Lijo Lazar Date: Fri, 16 May 2025 20:16:57 +0530 Subject: drm/amd/pm: Use external link order for xgmi data xgmi_port_num interface reports external link number for port number. To be consistent, use the external link number for reporting other XGMI link data also. v2: For invalid link number return -EINVAL (Kevin) Signed-off-by: Lijo Lazar Acked-by: Yang Wang Acked-by: Alex Deucher Signed-off-by: Alex Deucher --- drivers/gpu/drm/amd/amdgpu/amdgpu_xgmi.c | 17 +++++++++++++++++ drivers/gpu/drm/amd/amdgpu/amdgpu_xgmi.h | 1 + drivers/gpu/drm/amd/pm/swsmu/smu13/smu_v13_0_12_ppt.c | 9 ++++++--- drivers/gpu/drm/amd/pm/swsmu/smu13/smu_v13_0_6_ppt.c | 13 ++++++++----- 4 files changed, 32 insertions(+), 8 deletions(-) (limited to 'drivers/gpu') diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_xgmi.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_xgmi.c index f51ef4cf16e0..d9ad37711c3e 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_xgmi.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_xgmi.c @@ -294,6 +294,23 @@ static const struct amdgpu_pcs_ras_field xgmi3x16_pcs_ras_fields[] = { SOC15_REG_FIELD(PCS_XGMI3X16_PCS_ERROR_STATUS, RxCMDPktErr)}, }; +int amdgpu_xgmi_get_ext_link(struct amdgpu_device *adev, int link_num) +{ + int link_map_6_4_x[8] = { 0, 3, 1, 2, 7, 6, 4, 5 }; + + switch (amdgpu_ip_version(adev, XGMI_HWIP, 0)) { + case IP_VERSION(6, 4, 0): + case IP_VERSION(6, 4, 1): + if (link_num < ARRAY_SIZE(link_map_6_4_x)) + return link_map_6_4_x[link_num]; + break; + default: + return -EINVAL; + } + + return -EINVAL; +} + static u32 xgmi_v6_4_get_link_status(struct amdgpu_device *adev, int global_link_num) { const u32 smn_xgmi_6_4_pcs_state_hist1[2] = { 0x11a00070, 0x11b00070 }; diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_xgmi.h b/drivers/gpu/drm/amd/amdgpu/amdgpu_xgmi.h index 32dabba4062f..f994be985f42 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_xgmi.h +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_xgmi.h @@ -125,6 +125,7 @@ int amdgpu_xgmi_request_nps_change(struct amdgpu_device *adev, int req_nps_mode); int amdgpu_get_xgmi_link_status(struct amdgpu_device *adev, int global_link_num); +int amdgpu_xgmi_get_ext_link(struct amdgpu_device *adev, int link_num); void amdgpu_xgmi_early_init(struct amdgpu_device *adev); uint32_t amdgpu_xgmi_get_max_bandwidth(struct amdgpu_device *adev); 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 533d58e57d05..5a8824cc1c63 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 @@ -416,13 +416,16 @@ ssize_t smu_v13_0_12_get_gpu_metrics(struct smu_context *smu, void **table) gpu_metrics->mem_activity_acc = SMUQ10_ROUND(metrics->DramBandwidthUtilizationAcc); for (i = 0; i < NUM_XGMI_LINKS; i++) { - gpu_metrics->xgmi_read_data_acc[i] = + j = amdgpu_xgmi_get_ext_link(adev, i); + if (j < 0 || j >= NUM_XGMI_LINKS) + continue; + gpu_metrics->xgmi_read_data_acc[j] = SMUQ10_ROUND(metrics->XgmiReadDataSizeAcc[i]); - gpu_metrics->xgmi_write_data_acc[i] = + gpu_metrics->xgmi_write_data_acc[j] = SMUQ10_ROUND(metrics->XgmiWriteDataSizeAcc[i]); ret = amdgpu_get_xgmi_link_status(adev, i); if (ret >= 0) - gpu_metrics->xgmi_link_status[i] = ret; + gpu_metrics->xgmi_link_status[j] = ret; } gpu_metrics->num_partition = adev->xcp_mgr->num_xcps; diff --git a/drivers/gpu/drm/amd/pm/swsmu/smu13/smu_v13_0_6_ppt.c b/drivers/gpu/drm/amd/pm/swsmu/smu13/smu_v13_0_6_ppt.c index 615fd3771ae3..78d831c20768 100644 --- a/drivers/gpu/drm/amd/pm/swsmu/smu13/smu_v13_0_6_ppt.c +++ b/drivers/gpu/drm/amd/pm/swsmu/smu13/smu_v13_0_6_ppt.c @@ -2788,13 +2788,16 @@ static ssize_t smu_v13_0_6_get_gpu_metrics(struct smu_context *smu, void **table SMUQ10_ROUND(GET_METRIC_FIELD(DramBandwidthUtilizationAcc, version)); for (i = 0; i < NUM_XGMI_LINKS; i++) { - gpu_metrics->xgmi_read_data_acc[i] = - SMUQ10_ROUND(GET_METRIC_FIELD(XgmiReadDataSizeAcc, version)[i]); - gpu_metrics->xgmi_write_data_acc[i] = - SMUQ10_ROUND(GET_METRIC_FIELD(XgmiWriteDataSizeAcc, version)[i]); + j = amdgpu_xgmi_get_ext_link(adev, i); + if (j < 0 || j >= NUM_XGMI_LINKS) + continue; + gpu_metrics->xgmi_read_data_acc[j] = SMUQ10_ROUND( + GET_METRIC_FIELD(XgmiReadDataSizeAcc, version)[i]); + gpu_metrics->xgmi_write_data_acc[j] = SMUQ10_ROUND( + GET_METRIC_FIELD(XgmiWriteDataSizeAcc, version)[i]); ret = amdgpu_get_xgmi_link_status(adev, i); if (ret >= 0) - gpu_metrics->xgmi_link_status[i] = ret; + gpu_metrics->xgmi_link_status[j] = ret; } gpu_metrics->num_partition = adev->xcp_mgr->num_xcps; -- cgit v1.2.3 From f55fcf15a9c585d0a3f294307f1499d3759459c6 Mon Sep 17 00:00:00 2001 From: Mangesh Gadre Date: Wed, 14 May 2025 12:31:36 +0800 Subject: drm/amdgpu: Add vcn poison status reg added register to enable vcn ras Signed-off-by: Mangesh Gadre Reviewed-by: Stanley.Yang Reviewed-by: Tao Zhou Signed-off-by: Alex Deucher --- drivers/gpu/drm/amd/include/asic_reg/vcn/vcn_5_0_0_offset.h | 2 ++ drivers/gpu/drm/amd/include/asic_reg/vcn/vcn_5_0_0_sh_mask.h | 6 ++++++ 2 files changed, 8 insertions(+) (limited to 'drivers/gpu') diff --git a/drivers/gpu/drm/amd/include/asic_reg/vcn/vcn_5_0_0_offset.h b/drivers/gpu/drm/amd/include/asic_reg/vcn/vcn_5_0_0_offset.h index c4aaa86a95e2..f45155280ff5 100644 --- a/drivers/gpu/drm/amd/include/asic_reg/vcn/vcn_5_0_0_offset.h +++ b/drivers/gpu/drm/amd/include/asic_reg/vcn/vcn_5_0_0_offset.h @@ -1067,6 +1067,8 @@ #define regVCN_FEATURES_BASE_IDX 1 #define regUVD_GPUIOV_STATUS 0x0055 #define regUVD_GPUIOV_STATUS_BASE_IDX 1 +#define regUVD_RAS_VCPU_VCODEC_STATUS 0x0057 +#define regUVD_RAS_VCPU_VCODEC_STATUS_BASE_IDX 1 #define regUVD_SCRATCH15 0x005c #define regUVD_SCRATCH15_BASE_IDX 1 #define regUVD_VERSION 0x005d diff --git a/drivers/gpu/drm/amd/include/asic_reg/vcn/vcn_5_0_0_sh_mask.h b/drivers/gpu/drm/amd/include/asic_reg/vcn/vcn_5_0_0_sh_mask.h index bd7242e4e9c6..eb8ff9de5826 100644 --- a/drivers/gpu/drm/amd/include/asic_reg/vcn/vcn_5_0_0_sh_mask.h +++ b/drivers/gpu/drm/amd/include/asic_reg/vcn/vcn_5_0_0_sh_mask.h @@ -5714,6 +5714,12 @@ //UVD_GPUIOV_STATUS #define UVD_GPUIOV_STATUS__UVD_GPUIOV_STATUS_VF_ENABLE__SHIFT 0x0 #define UVD_GPUIOV_STATUS__UVD_GPUIOV_STATUS_VF_ENABLE_MASK 0x00000001L +//UVD_RAS_VCPU_VCODEC_STATUS +#define UVD_RAS_VCPU_VCODEC_STATUS__POISONED_VF__SHIFT 0x0 +#define UVD_RAS_VCPU_VCODEC_STATUS__POISONED_PF__SHIFT 0x1f +#define UVD_RAS_VCPU_VCODEC_STATUS__POISONED_VF_MASK 0x7FFFFFFFL +#define UVD_RAS_VCPU_VCODEC_STATUS__POISONED_PF_MASK 0x80000000L + //UVD_SCRATCH15 #define UVD_SCRATCH15__SCRATCH15_DATA__SHIFT 0x0 #define UVD_SCRATCH15__SCRATCH15_DATA_MASK 0xFFFFFFFFL -- cgit v1.2.3 From e90bd6d898165c281b5582361857e7bc8e0f917d Mon Sep 17 00:00:00 2001 From: Alex Deucher Date: Wed, 7 May 2025 16:51:38 -0400 Subject: drm/amdgpu: Update runtime pm checks Don't enable BACO when in passthrough. PCI resets don't work correctly when in BACO. Reviewed-by: Hawking Zhang Signed-off-by: Alex Deucher --- drivers/gpu/drm/amd/amdgpu/amdgpu_device.c | 5 +++-- 1 file changed, 3 insertions(+), 2 deletions(-) (limited to 'drivers/gpu') diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_device.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_device.c index 908e39dae4b4..e1bab6a96cb6 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_device.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_device.c @@ -512,12 +512,13 @@ void amdgpu_device_detect_runtime_pm_mode(struct amdgpu_device *adev) break; case CHIP_VEGA10: /* enable BACO as runpm mode if noretry=0 */ - if (!adev->gmc.noretry) + if (!adev->gmc.noretry && !amdgpu_passthrough(adev)) adev->pm.rpm_mode = AMDGPU_RUNPM_BACO; break; default: /* enable BACO as runpm mode on CI+ */ - adev->pm.rpm_mode = AMDGPU_RUNPM_BACO; + if (!amdgpu_passthrough(adev)) + adev->pm.rpm_mode = AMDGPU_RUNPM_BACO; break; } -- cgit v1.2.3 From 16f2c942b6e44a0c3970134c8c460c7fd465ac9d Mon Sep 17 00:00:00 2001 From: Tvrtko Ursulin Date: Mon, 19 May 2025 17:37:11 +0100 Subject: drm/amdgpu: Make amdgpu_ctx_mgr_entity_fini static MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit Function amdgpu_ctx_mgr_entity_fini() only has a single local caller so lets make it local. Reviewed-by: Christian König Signed-off-by: Tvrtko Ursulin Cc: Alex Deucher Cc: Christian König Signed-off-by: Alex Deucher --- drivers/gpu/drm/amd/amdgpu/amdgpu_ctx.c | 2 +- drivers/gpu/drm/amd/amdgpu/amdgpu_ctx.h | 1 - 2 files changed, 1 insertion(+), 2 deletions(-) (limited to 'drivers/gpu') diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_ctx.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_ctx.c index c43d1b6e5d66..4ff8552e872d 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_ctx.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_ctx.c @@ -919,7 +919,7 @@ long amdgpu_ctx_mgr_entity_flush(struct amdgpu_ctx_mgr *mgr, long timeout) return timeout; } -void amdgpu_ctx_mgr_entity_fini(struct amdgpu_ctx_mgr *mgr) +static void amdgpu_ctx_mgr_entity_fini(struct amdgpu_ctx_mgr *mgr) { struct amdgpu_ctx *ctx; struct idr *idp; diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_ctx.h b/drivers/gpu/drm/amd/amdgpu/amdgpu_ctx.h index 85376baaa92f..090dfe86f75b 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_ctx.h +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_ctx.h @@ -92,7 +92,6 @@ int amdgpu_ctx_wait_prev_fence(struct amdgpu_ctx *ctx, void amdgpu_ctx_mgr_init(struct amdgpu_ctx_mgr *mgr, struct amdgpu_device *adev); -void amdgpu_ctx_mgr_entity_fini(struct amdgpu_ctx_mgr *mgr); long amdgpu_ctx_mgr_entity_flush(struct amdgpu_ctx_mgr *mgr, long timeout); void amdgpu_ctx_mgr_fini(struct amdgpu_ctx_mgr *mgr); void amdgpu_ctx_mgr_usage(struct amdgpu_ctx_mgr *mgr, -- cgit v1.2.3 From dd64956685fa48358c4152d952070c8c073e5f89 Mon Sep 17 00:00:00 2001 From: Tvrtko Ursulin Date: Mon, 19 May 2025 17:37:12 +0100 Subject: drm/amdgpu: Remove duplicated "context still alive" check MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit When amdgpu_ctx_mgr_fini() calls amdgpu_ctx_mgr_entity_fini() it contains the exact same "context still alive" check as it will do next. Remove the duplicated copy. Reviewed-by: Christian König Signed-off-by: Tvrtko Ursulin Cc: Alex Deucher Cc: Christian König Signed-off-by: Alex Deucher --- drivers/gpu/drm/amd/amdgpu/amdgpu_ctx.c | 12 ------------ 1 file changed, 12 deletions(-) (limited to 'drivers/gpu') diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_ctx.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_ctx.c index 4ff8552e872d..85567d0d9545 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_ctx.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_ctx.c @@ -949,19 +949,7 @@ static void amdgpu_ctx_mgr_entity_fini(struct amdgpu_ctx_mgr *mgr) void amdgpu_ctx_mgr_fini(struct amdgpu_ctx_mgr *mgr) { - struct amdgpu_ctx *ctx; - struct idr *idp; - uint32_t id; - amdgpu_ctx_mgr_entity_fini(mgr); - - idp = &mgr->ctx_handles; - - idr_for_each_entry(idp, ctx, id) { - if (kref_put(&ctx->refcount, amdgpu_ctx_fini) != 1) - DRM_ERROR("ctx %p is still alive\n", ctx); - } - idr_destroy(&mgr->ctx_handles); mutex_destroy(&mgr->lock); } -- cgit v1.2.3 From f9f403218e1ca947a1a017be01da6ae7fb637eba Mon Sep 17 00:00:00 2001 From: Melissa Wen Date: Tue, 13 May 2025 18:38:17 -0300 Subject: drm/amd/display: only collect data if debug gamut_remap is available Color gamut_remap state log may be not available for some hw versions, so prevent null pointer dereference by checking if there is a function to collect data for this hw version. Signed-off-by: Melissa Wen Reviewed-by: Alex Hung Signed-off-by: Alex Deucher --- .../drm/amd/display/dc/hwss/dcn20/dcn20_hwseq.c | 26 +++++++++++++++------- .../drm/amd/display/dc/hwss/dcn30/dcn30_hwseq.c | 24 ++++++++++++++------ 2 files changed, 35 insertions(+), 15 deletions(-) (limited to 'drivers/gpu') diff --git a/drivers/gpu/drm/amd/display/dc/hwss/dcn20/dcn20_hwseq.c b/drivers/gpu/drm/amd/display/dc/hwss/dcn20/dcn20_hwseq.c index 858288c3b1ac..c277df12c817 100644 --- a/drivers/gpu/drm/amd/display/dc/hwss/dcn20/dcn20_hwseq.c +++ b/drivers/gpu/drm/amd/display/dc/hwss/dcn20/dcn20_hwseq.c @@ -76,6 +76,7 @@ void dcn20_log_color_state(struct dc *dc, { struct dc_context *dc_ctx = dc->ctx; struct resource_pool *pool = dc->res_pool; + bool is_gamut_remap_available = false; int i; DTN_INFO("DPP: DGAM mode SHAPER mode 3DLUT mode 3DLUT bit depth" @@ -89,15 +90,15 @@ void dcn20_log_color_state(struct dc *dc, struct dcn_dpp_state s = {0}; dpp->funcs->dpp_read_state(dpp, &s); - dpp->funcs->dpp_get_gamut_remap(dpp, &s.gamut_remap); + if (dpp->funcs->dpp_get_gamut_remap) { + dpp->funcs->dpp_get_gamut_remap(dpp, &s.gamut_remap); + is_gamut_remap_available = true; + } if (!s.is_enabled) continue; - DTN_INFO("[%2d]: %8s %11s %10s %15s %10s %9s %12s " - "%010lld %010lld %010lld %010lld " - "%010lld %010lld %010lld %010lld " - "%010lld %010lld %010lld %010lld", + DTN_INFO("[%2d]: %8s %11s %10s %15s %10s %9s", dpp->inst, (s.dgam_lut_mode == 0) ? "Bypass" : ((s.dgam_lut_mode == 1) ? "sRGB" : @@ -114,10 +115,17 @@ void dcn20_log_color_state(struct dc *dc, (s.lut3d_bit_depth <= 0) ? "12-bit" : "10-bit", (s.lut3d_size == 0) ? "17x17x17" : "9x9x9", (s.rgam_lut_mode == 1) ? "RAM A" : - ((s.rgam_lut_mode == 1) ? "RAM B" : "Bypass"), + ((s.rgam_lut_mode == 1) ? "RAM B" : "Bypass")); + + if (is_gamut_remap_available) { + DTN_INFO(" %12s " + "%010lld %010lld %010lld %010lld " + "%010lld %010lld %010lld %010lld " + "%010lld %010lld %010lld %010lld", + (s.gamut_remap.gamut_adjust_type == 0) ? "Bypass" : - ((s.gamut_remap.gamut_adjust_type == 1) ? "HW" : - "SW"), + ((s.gamut_remap.gamut_adjust_type == 1) ? "HW" : + "SW"), s.gamut_remap.temperature_matrix[0].value, s.gamut_remap.temperature_matrix[1].value, s.gamut_remap.temperature_matrix[2].value, @@ -130,6 +138,8 @@ void dcn20_log_color_state(struct dc *dc, s.gamut_remap.temperature_matrix[9].value, s.gamut_remap.temperature_matrix[10].value, s.gamut_remap.temperature_matrix[11].value); + } + DTN_INFO("\n"); } DTN_INFO("\n"); diff --git a/drivers/gpu/drm/amd/display/dc/hwss/dcn30/dcn30_hwseq.c b/drivers/gpu/drm/amd/display/dc/hwss/dcn30/dcn30_hwseq.c index e89ebfda4873..37a239219dfe 100644 --- a/drivers/gpu/drm/amd/display/dc/hwss/dcn30/dcn30_hwseq.c +++ b/drivers/gpu/drm/amd/display/dc/hwss/dcn30/dcn30_hwseq.c @@ -74,6 +74,7 @@ void dcn30_log_color_state(struct dc *dc, { struct dc_context *dc_ctx = dc->ctx; struct resource_pool *pool = dc->res_pool; + bool is_gamut_remap_available = false; int i; DTN_INFO("DPP: DGAM ROM DGAM ROM type DGAM LUT SHAPER mode" @@ -88,16 +89,16 @@ void dcn30_log_color_state(struct dc *dc, struct dcn_dpp_state s = {0}; dpp->funcs->dpp_read_state(dpp, &s); - dpp->funcs->dpp_get_gamut_remap(dpp, &s.gamut_remap); + + if (dpp->funcs->dpp_get_gamut_remap) { + dpp->funcs->dpp_get_gamut_remap(dpp, &s.gamut_remap); + is_gamut_remap_available = true; + } if (!s.is_enabled) continue; - DTN_INFO("[%2d]: %7x %13s %8s %11s %10s %15s %10s %9s" - " %12s " - "%010lld %010lld %010lld %010lld " - "%010lld %010lld %010lld %010lld " - "%010lld %010lld %010lld %010lld", + DTN_INFO("[%2d]: %7x %13s %8s %11s %10s %15s %10s %9s", dpp->inst, s.pre_dgam_mode, (s.pre_dgam_select == 0) ? "sRGB" : @@ -121,7 +122,14 @@ void dcn30_log_color_state(struct dc *dc, (s.lut3d_size == 0) ? "17x17x17" : "9x9x9", (s.rgam_lut_mode == 0) ? "Bypass" : ((s.rgam_lut_mode == 1) ? "RAM A" : - "RAM B"), + "RAM B")); + + if (is_gamut_remap_available) { + DTN_INFO(" %12s " + "%010lld %010lld %010lld %010lld " + "%010lld %010lld %010lld %010lld " + "%010lld %010lld %010lld %010lld", + (s.gamut_remap.gamut_adjust_type == 0) ? "Bypass" : ((s.gamut_remap.gamut_adjust_type == 1) ? "HW" : "SW"), @@ -137,6 +145,8 @@ void dcn30_log_color_state(struct dc *dc, s.gamut_remap.temperature_matrix[9].value, s.gamut_remap.temperature_matrix[10].value, s.gamut_remap.temperature_matrix[11].value); + } + DTN_INFO("\n"); } DTN_INFO("\n"); -- cgit v1.2.3 From 02fd27e6cf853e55cebcefb4aa7a9828a6848510 Mon Sep 17 00:00:00 2001 From: Melissa Wen Date: Fri, 25 Apr 2025 17:52:30 -0300 Subject: drm/amd/display: no 3D and blnd LUT as DPP color caps for DCN401 Match what is declared as DPP color caps with hw caps. DCN401 has MPC shaper + 3D LUTs that are movable before and after blending (get from plane or stream), but no DPP blend LUTs. Signed-off-by: Melissa Wen Reviewed-by: Alex Hung Signed-off-by: Alex Deucher --- drivers/gpu/drm/amd/display/dc/resource/dcn401/dcn401_resource.c | 4 ++-- 1 file changed, 2 insertions(+), 2 deletions(-) (limited to 'drivers/gpu') diff --git a/drivers/gpu/drm/amd/display/dc/resource/dcn401/dcn401_resource.c b/drivers/gpu/drm/amd/display/dc/resource/dcn401/dcn401_resource.c index e0e32975ca34..f420c4dafa03 100644 --- a/drivers/gpu/drm/amd/display/dc/resource/dcn401/dcn401_resource.c +++ b/drivers/gpu/drm/amd/display/dc/resource/dcn401/dcn401_resource.c @@ -1938,8 +1938,8 @@ static bool dcn401_resource_construct( dc->caps.color.dpp.gamma_corr = 1; dc->caps.color.dpp.dgam_rom_for_yuv = 0; - dc->caps.color.dpp.hw_3d_lut = 1; - dc->caps.color.dpp.ogam_ram = 1; + dc->caps.color.dpp.hw_3d_lut = 0; + dc->caps.color.dpp.ogam_ram = 0; // no OGAM ROM on DCN2 and later ASICs dc->caps.color.dpp.ogam_rom_caps.srgb = 0; dc->caps.color.dpp.ogam_rom_caps.bt2020 = 0; -- cgit v1.2.3 From 076873e5b360ccd91687e23c6ca0042a0356b9eb Mon Sep 17 00:00:00 2001 From: Wayne Lin Date: Tue, 20 May 2025 09:34:42 +0800 Subject: drm/amd/display: Add a new dcdebugmask to allow skip detection LT Under specific embedded scenarios, we might still use DP interface rather than eDP interface. Under such case, detection link training is unnecessary. Add a new dcdebugmask value that can be used to skip the detection LT Reviewed-by: Tom Chung Link: https://lore.kernel.org/amd-gfx/20250521063934.2111323-1-Wayne.Lin@amd.com/ Signed-off-by: Wayne Lin Signed-off-by: Alex Deucher --- drivers/gpu/drm/amd/display/amdgpu_dm/amdgpu_dm.c | 3 +++ drivers/gpu/drm/amd/include/amd_shared.h | 5 +++++ 2 files changed, 8 insertions(+) (limited to 'drivers/gpu') 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 4b4e9241619f..2bb347771aa1 100644 --- a/drivers/gpu/drm/amd/display/amdgpu_dm/amdgpu_dm.c +++ b/drivers/gpu/drm/amd/display/amdgpu_dm/amdgpu_dm.c @@ -2020,6 +2020,9 @@ static int amdgpu_dm_init(struct amdgpu_device *adev) if (amdgpu_dc_debug_mask & DC_HDCP_LC_ENABLE_SW_FALLBACK) adev->dm.dc->debug.hdcp_lc_enable_sw_fallback = true; + if (amdgpu_dc_debug_mask & DC_SKIP_DETECTION_LT) + adev->dm.dc->debug.skip_detection_link_training = true; + adev->dm.dc->debug.visual_confirm = amdgpu_dc_visual_confirm; /* TODO: Remove after DP2 receiver gets proper support of Cable ID feature */ diff --git a/drivers/gpu/drm/amd/include/amd_shared.h b/drivers/gpu/drm/amd/include/amd_shared.h index c8eccee9b023..31de36c9156f 100644 --- a/drivers/gpu/drm/amd/include/amd_shared.h +++ b/drivers/gpu/drm/amd/include/amd_shared.h @@ -370,6 +370,11 @@ enum DC_DEBUG_MASK { * path failure, retry using legacy SW path. */ DC_HDCP_LC_ENABLE_SW_FALLBACK = 0x100000, + + /** + * @DC_SKIP_DETECTION_LT: If set, skip detection link training + */ + DC_SKIP_DETECTION_LT = 0x200000, }; enum amd_dpm_forced_level; -- cgit v1.2.3 From 5035caf18d88cde5a1fd661dbb45b1229816c07e Mon Sep 17 00:00:00 2001 From: Mangesh Gadre Date: Fri, 16 May 2025 07:52:14 +0800 Subject: drm/amdgpu: Enable RAS for vcn 5.0.1 Enable vcn ras posion processing and aca error logging Signed-off-by: Mangesh Gadre Reviewed-by: Stanley.Yang Reviewed-by: Tao Zhou Signed-off-by: Alex Deucher --- drivers/gpu/drm/amd/amdgpu/vcn_v5_0_1.c | 165 +++++++++++++++++++++++++++++++- drivers/gpu/drm/amd/amdgpu/vcn_v5_0_1.h | 7 ++ 2 files changed, 171 insertions(+), 1 deletion(-) (limited to 'drivers/gpu') diff --git a/drivers/gpu/drm/amd/amdgpu/vcn_v5_0_1.c b/drivers/gpu/drm/amd/amdgpu/vcn_v5_0_1.c index 8e843011703c..1e9d2aedf279 100644 --- a/drivers/gpu/drm/amd/amdgpu/vcn_v5_0_1.c +++ b/drivers/gpu/drm/amd/amdgpu/vcn_v5_0_1.c @@ -46,7 +46,7 @@ static void vcn_v5_0_1_set_irq_funcs(struct amdgpu_device *adev); static int vcn_v5_0_1_set_pg_state(struct amdgpu_vcn_inst *vinst, enum amd_powergating_state state); static void vcn_v5_0_1_unified_ring_set_wptr(struct amdgpu_ring *ring); - +static void vcn_v5_0_1_set_ras_funcs(struct amdgpu_device *adev); /** * vcn_v5_0_1_early_init - set function pointers and load microcode * @@ -66,6 +66,7 @@ static int vcn_v5_0_1_early_init(struct amdgpu_ip_block *ip_block) vcn_v5_0_1_set_unified_ring_funcs(adev); vcn_v5_0_1_set_irq_funcs(adev); + vcn_v5_0_1_set_ras_funcs(adev); for (i = 0; i < adev->vcn.num_vcn_inst; ++i) { adev->vcn.inst[i].set_pg_state = vcn_v5_0_1_set_pg_state; @@ -113,6 +114,10 @@ static int vcn_v5_0_1_sw_init(struct amdgpu_ip_block *ip_block) if (r) return r; + /* VCN POISON TRAP */ + r = amdgpu_irq_add_id(adev, SOC15_IH_CLIENTID_VCN, + VCN_5_0__SRCID_UVD_POISON, &adev->vcn.inst->ras_poison_irq); + for (i = 0; i < adev->vcn.num_vcn_inst; i++) { vcn_inst = GET_INST(VCN, i); @@ -279,6 +284,9 @@ static int vcn_v5_0_1_hw_fini(struct amdgpu_ip_block *ip_block) vinst->set_pg_state(vinst, AMD_PG_STATE_GATE); } + if (amdgpu_ras_is_supported(adev, AMDGPU_RAS_BLOCK__VCN)) + amdgpu_irq_put(adev, &adev->vcn.inst->ras_poison_irq, 0); + return 0; } @@ -1391,10 +1399,24 @@ static int vcn_v5_0_1_process_interrupt(struct amdgpu_device *adev, struct amdgp return 0; } +static int vcn_v5_0_1_set_ras_interrupt_state(struct amdgpu_device *adev, + struct amdgpu_irq_src *source, + unsigned int type, + enum amdgpu_interrupt_state state) +{ + return 0; +} + static const struct amdgpu_irq_src_funcs vcn_v5_0_1_irq_funcs = { .process = vcn_v5_0_1_process_interrupt, }; +static const struct amdgpu_irq_src_funcs vcn_v5_0_1_ras_irq_funcs = { + .set = vcn_v5_0_1_set_ras_interrupt_state, + .process = amdgpu_vcn_process_poison_irq, +}; + + /** * vcn_v5_0_1_set_irq_funcs - set VCN block interrupt irq functions * @@ -1408,7 +1430,12 @@ static void vcn_v5_0_1_set_irq_funcs(struct amdgpu_device *adev) for (i = 0; i < adev->vcn.num_vcn_inst; ++i) adev->vcn.inst->irq.num_types++; + adev->vcn.inst->irq.funcs = &vcn_v5_0_1_irq_funcs; + + adev->vcn.inst->ras_poison_irq.num_types = 1; + adev->vcn.inst->ras_poison_irq.funcs = &vcn_v5_0_1_ras_irq_funcs; + } static const struct amd_ip_funcs vcn_v5_0_1_ip_funcs = { @@ -1440,3 +1467,139 @@ const struct amdgpu_ip_block_version vcn_v5_0_1_ip_block = { .rev = 1, .funcs = &vcn_v5_0_1_ip_funcs, }; + +static uint32_t vcn_v5_0_1_query_poison_by_instance(struct amdgpu_device *adev, + uint32_t instance, uint32_t sub_block) +{ + uint32_t poison_stat = 0, reg_value = 0; + + switch (sub_block) { + case AMDGPU_VCN_V5_0_1_VCPU_VCODEC: + reg_value = RREG32_SOC15(VCN, instance, regUVD_RAS_VCPU_VCODEC_STATUS); + poison_stat = REG_GET_FIELD(reg_value, UVD_RAS_VCPU_VCODEC_STATUS, POISONED_PF); + break; + default: + break; + } + + if (poison_stat) + dev_info(adev->dev, "Poison detected in VCN%d, sub_block%d\n", + instance, sub_block); + + return poison_stat; +} + +static bool vcn_v5_0_1_query_poison_status(struct amdgpu_device *adev) +{ + uint32_t inst, sub; + uint32_t poison_stat = 0; + + for (inst = 0; inst < adev->vcn.num_vcn_inst; inst++) + for (sub = 0; sub < AMDGPU_VCN_V5_0_1_MAX_SUB_BLOCK; sub++) + poison_stat += + vcn_v5_0_1_query_poison_by_instance(adev, inst, sub); + + return !!poison_stat; +} + +static const struct amdgpu_ras_block_hw_ops vcn_v5_0_1_ras_hw_ops = { + .query_poison_status = vcn_v5_0_1_query_poison_status, +}; + +static int vcn_v5_0_1_aca_bank_parser(struct aca_handle *handle, struct aca_bank *bank, + enum aca_smu_type type, void *data) +{ + struct aca_bank_info info; + u64 misc0; + int ret; + + ret = aca_bank_info_decode(bank, &info); + if (ret) + return ret; + + misc0 = bank->regs[ACA_REG_IDX_MISC0]; + switch (type) { + case ACA_SMU_TYPE_UE: + bank->aca_err_type = ACA_ERROR_TYPE_UE; + ret = aca_error_cache_log_bank_error(handle, &info, ACA_ERROR_TYPE_UE, + 1ULL); + break; + case ACA_SMU_TYPE_CE: + bank->aca_err_type = ACA_ERROR_TYPE_CE; + ret = aca_error_cache_log_bank_error(handle, &info, bank->aca_err_type, + ACA_REG__MISC0__ERRCNT(misc0)); + break; + default: + return -EINVAL; + } + + return ret; +} + +/* reference to smu driver if header file */ +static int vcn_v5_0_1_err_codes[] = { + 14, 15, /* VCN */ +}; + +static bool vcn_v5_0_1_aca_bank_is_valid(struct aca_handle *handle, struct aca_bank *bank, + enum aca_smu_type type, void *data) +{ + u32 instlo; + + instlo = ACA_REG__IPID__INSTANCEIDLO(bank->regs[ACA_REG_IDX_IPID]); + instlo &= GENMASK(31, 1); + + if (instlo != mmSMNAID_AID0_MCA_SMU) + return false; + + if (aca_bank_check_error_codes(handle->adev, bank, + vcn_v5_0_1_err_codes, + ARRAY_SIZE(vcn_v5_0_1_err_codes))) + return false; + + return true; +} + +static const struct aca_bank_ops vcn_v5_0_1_aca_bank_ops = { + .aca_bank_parser = vcn_v5_0_1_aca_bank_parser, + .aca_bank_is_valid = vcn_v5_0_1_aca_bank_is_valid, +}; + +static const struct aca_info vcn_v5_0_1_aca_info = { + .hwip = ACA_HWIP_TYPE_SMU, + .mask = ACA_ERROR_UE_MASK, + .bank_ops = &vcn_v5_0_1_aca_bank_ops, +}; + +static int vcn_v5_0_1_ras_late_init(struct amdgpu_device *adev, struct ras_common_if *ras_block) +{ + int r; + + r = amdgpu_ras_block_late_init(adev, ras_block); + if (r) + return r; + + r = amdgpu_ras_bind_aca(adev, AMDGPU_RAS_BLOCK__VCN, + &vcn_v5_0_1_aca_info, NULL); + if (r) + goto late_fini; + + return 0; + +late_fini: + amdgpu_ras_block_late_fini(adev, ras_block); + + return r; +} + +static struct amdgpu_vcn_ras vcn_v5_0_1_ras = { + .ras_block = { + .hw_ops = &vcn_v5_0_1_ras_hw_ops, + .ras_late_init = vcn_v5_0_1_ras_late_init, + }, +}; + +static void vcn_v5_0_1_set_ras_funcs(struct amdgpu_device *adev) +{ + adev->vcn.ras = &vcn_v5_0_1_ras; +} diff --git a/drivers/gpu/drm/amd/amdgpu/vcn_v5_0_1.h b/drivers/gpu/drm/amd/amdgpu/vcn_v5_0_1.h index 8fd90bd10807..b72e4da68317 100644 --- a/drivers/gpu/drm/amd/amdgpu/vcn_v5_0_1.h +++ b/drivers/gpu/drm/amd/amdgpu/vcn_v5_0_1.h @@ -27,6 +27,13 @@ #define regVCN_RRMT_CNTL 0x0940 #define regVCN_RRMT_CNTL_BASE_IDX 1 + +enum amdgpu_vcn_v5_0_1_sub_block { + AMDGPU_VCN_V5_0_1_VCPU_VCODEC = 0, + + AMDGPU_VCN_V5_0_1_MAX_SUB_BLOCK, +}; + extern const struct amdgpu_ip_block_version vcn_v5_0_1_ip_block; #endif /* __VCN_v5_0_1_H__ */ -- cgit v1.2.3 From 8d74ce4e5524b39e991bfa025f1382e54c5f710a Mon Sep 17 00:00:00 2001 From: Mangesh Gadre Date: Wed, 14 May 2025 13:17:02 +0800 Subject: drm/amdgpu: Add jpeg poison status reg added registers to enable jpeg ras Signed-off-by: Mangesh Gadre Reviewed-by: Stanley.Yang Reviewed-by: Tao Zhou Signed-off-by: Alex Deucher --- drivers/gpu/drm/amd/include/asic_reg/vcn/vcn_5_0_0_offset.h | 4 ++++ drivers/gpu/drm/amd/include/asic_reg/vcn/vcn_5_0_0_sh_mask.h | 10 ++++++++++ 2 files changed, 14 insertions(+) (limited to 'drivers/gpu') diff --git a/drivers/gpu/drm/amd/include/asic_reg/vcn/vcn_5_0_0_offset.h b/drivers/gpu/drm/amd/include/asic_reg/vcn/vcn_5_0_0_offset.h index f45155280ff5..72a118b2af69 100644 --- a/drivers/gpu/drm/amd/include/asic_reg/vcn/vcn_5_0_0_offset.h +++ b/drivers/gpu/drm/amd/include/asic_reg/vcn/vcn_5_0_0_offset.h @@ -1070,6 +1070,10 @@ #define regUVD_RAS_VCPU_VCODEC_STATUS 0x0057 #define regUVD_RAS_VCPU_VCODEC_STATUS_BASE_IDX 1 #define regUVD_SCRATCH15 0x005c +#define regUVD_RAS_JPEG0_STATUS 0x0059 +#define regUVD_RAS_JPEG0_STATUS_BASE_IDX 1 +#define regUVD_RAS_JPEG1_STATUS 0x005a +#define regUVD_RAS_JPEG1_STATUS_BASE_IDX 1 #define regUVD_SCRATCH15_BASE_IDX 1 #define regUVD_VERSION 0x005d #define regUVD_VERSION_BASE_IDX 1 diff --git a/drivers/gpu/drm/amd/include/asic_reg/vcn/vcn_5_0_0_sh_mask.h b/drivers/gpu/drm/amd/include/asic_reg/vcn/vcn_5_0_0_sh_mask.h index eb8ff9de5826..c78b09d6fbae 100644 --- a/drivers/gpu/drm/amd/include/asic_reg/vcn/vcn_5_0_0_sh_mask.h +++ b/drivers/gpu/drm/amd/include/asic_reg/vcn/vcn_5_0_0_sh_mask.h @@ -5720,6 +5720,16 @@ #define UVD_RAS_VCPU_VCODEC_STATUS__POISONED_VF_MASK 0x7FFFFFFFL #define UVD_RAS_VCPU_VCODEC_STATUS__POISONED_PF_MASK 0x80000000L +//UVD_RAS_JPEG0_STATUS +#define UVD_RAS_JPEG0_STATUS__POISONED_VF__SHIFT 0x0 +#define UVD_RAS_JPEG0_STATUS__POISONED_PF__SHIFT 0x1f +#define UVD_RAS_JPEG0_STATUS__POISONED_VF_MASK 0x7FFFFFFFL +#define UVD_RAS_JPEG0_STATUS__POISONED_PF_MASK 0x80000000L +//UVD_RAS_JPEG1_STATUS +#define UVD_RAS_JPEG1_STATUS__POISONED_VF__SHIFT 0x0 +#define UVD_RAS_JPEG1_STATUS__POISONED_PF__SHIFT 0x1f +#define UVD_RAS_JPEG1_STATUS__POISONED_VF_MASK 0x7FFFFFFFL +#define UVD_RAS_JPEG1_STATUS__POISONED_PF_MASK 0x80000000L //UVD_SCRATCH15 #define UVD_SCRATCH15__SCRATCH15_DATA__SHIFT 0x0 #define UVD_SCRATCH15__SCRATCH15_DATA_MASK 0xFFFFFFFFL -- cgit v1.2.3 From 25e9fb6e3ad885a83437f4aab1039bbbaab13d92 Mon Sep 17 00:00:00 2001 From: Mangesh Gadre Date: Fri, 16 May 2025 08:06:00 +0800 Subject: drm/amdgpu: Enable RAS for jpeg 5.0.1 Enable jpeg ras posion processing and aca error logging Signed-off-by: Mangesh Gadre Reviewed-by: Stanley.Yang Reviewed-by: Tao Zhou Signed-off-by: Alex Deucher --- drivers/gpu/drm/amd/amdgpu/jpeg_v5_0_1.c | 182 +++++++++++++++++++++++++++++++ drivers/gpu/drm/amd/amdgpu/jpeg_v5_0_1.h | 10 ++ 2 files changed, 192 insertions(+) (limited to 'drivers/gpu') diff --git a/drivers/gpu/drm/amd/amdgpu/jpeg_v5_0_1.c b/drivers/gpu/drm/amd/amdgpu/jpeg_v5_0_1.c index cb94bd71300f..3b6f65a25646 100644 --- a/drivers/gpu/drm/amd/amdgpu/jpeg_v5_0_1.c +++ b/drivers/gpu/drm/amd/amdgpu/jpeg_v5_0_1.c @@ -39,6 +39,7 @@ static void jpeg_v5_0_1_set_dec_ring_funcs(struct amdgpu_device *adev); static void jpeg_v5_0_1_set_irq_funcs(struct amdgpu_device *adev); static int jpeg_v5_0_1_set_powergating_state(struct amdgpu_ip_block *ip_block, enum amd_powergating_state state); +static void jpeg_v5_0_1_set_ras_funcs(struct amdgpu_device *adev); static void jpeg_v5_0_1_dec_ring_set_wptr(struct amdgpu_ring *ring); static int amdgpu_ih_srcid_jpeg[] = { @@ -120,6 +121,7 @@ static int jpeg_v5_0_1_early_init(struct amdgpu_ip_block *ip_block) adev->jpeg.num_jpeg_rings = AMDGPU_MAX_JPEG_RINGS; jpeg_v5_0_1_set_dec_ring_funcs(adev); jpeg_v5_0_1_set_irq_funcs(adev); + jpeg_v5_0_1_set_ras_funcs(adev); return 0; } @@ -144,6 +146,17 @@ static int jpeg_v5_0_1_sw_init(struct amdgpu_ip_block *ip_block) if (r) return r; } + /* JPEG DJPEG POISON EVENT */ + r = amdgpu_irq_add_id(adev, SOC15_IH_CLIENTID_VCN, + VCN_5_0__SRCID_DJPEG0_POISON, &adev->jpeg.inst->ras_poison_irq); + if (r) + return r; + + /* JPEG EJPEG POISON EVENT */ + r = amdgpu_irq_add_id(adev, SOC15_IH_CLIENTID_VCN, + VCN_5_0__SRCID_EJPEG0_POISON, &adev->jpeg.inst->ras_poison_irq); + if (r) + return r; r = amdgpu_jpeg_sw_init(adev); if (r) @@ -296,6 +309,9 @@ static int jpeg_v5_0_1_hw_fini(struct amdgpu_ip_block *ip_block) ret = jpeg_v5_0_1_set_powergating_state(ip_block, AMD_PG_STATE_GATE); } + if (amdgpu_ras_is_supported(adev, AMDGPU_RAS_BLOCK__JPEG)) + amdgpu_irq_put(adev, &adev->jpeg.inst->ras_poison_irq, 0); + return ret; } @@ -723,6 +739,16 @@ static int jpeg_v5_0_1_set_interrupt_state(struct amdgpu_device *adev, return 0; } +static int jpeg_v5_0_1_set_ras_interrupt_state(struct amdgpu_device *adev, + struct amdgpu_irq_src *source, + unsigned int type, + enum amdgpu_interrupt_state state) +{ + return 0; +} + + + static int jpeg_v5_0_1_process_interrupt(struct amdgpu_device *adev, struct amdgpu_irq_src *source, struct amdgpu_iv_entry *entry) @@ -892,6 +918,11 @@ static const struct amdgpu_irq_src_funcs jpeg_v5_0_1_irq_funcs = { .process = jpeg_v5_0_1_process_interrupt, }; +static const struct amdgpu_irq_src_funcs jpeg_v5_0_1_ras_irq_funcs = { + .set = jpeg_v5_0_1_set_ras_interrupt_state, + .process = amdgpu_jpeg_process_poison_irq, +}; + static void jpeg_v5_0_1_set_irq_funcs(struct amdgpu_device *adev) { int i; @@ -900,6 +931,10 @@ static void jpeg_v5_0_1_set_irq_funcs(struct amdgpu_device *adev) adev->jpeg.inst->irq.num_types += adev->jpeg.num_jpeg_rings; adev->jpeg.inst->irq.funcs = &jpeg_v5_0_1_irq_funcs; + + adev->jpeg.inst->ras_poison_irq.num_types = 1; + adev->jpeg.inst->ras_poison_irq.funcs = &jpeg_v5_0_1_ras_irq_funcs; + } const struct amdgpu_ip_block_version jpeg_v5_0_1_ip_block = { @@ -909,3 +944,150 @@ const struct amdgpu_ip_block_version jpeg_v5_0_1_ip_block = { .rev = 1, .funcs = &jpeg_v5_0_1_ip_funcs, }; + +static uint32_t jpeg_v5_0_1_query_poison_by_instance(struct amdgpu_device *adev, + uint32_t instance, uint32_t sub_block) +{ + uint32_t poison_stat = 0, reg_value = 0; + + switch (sub_block) { + case AMDGPU_JPEG_V5_0_1_JPEG0: + reg_value = RREG32_SOC15(JPEG, instance, regUVD_RAS_JPEG0_STATUS); + poison_stat = REG_GET_FIELD(reg_value, UVD_RAS_JPEG0_STATUS, POISONED_PF); + break; + case AMDGPU_JPEG_V5_0_1_JPEG1: + reg_value = RREG32_SOC15(JPEG, instance, regUVD_RAS_JPEG1_STATUS); + poison_stat = REG_GET_FIELD(reg_value, UVD_RAS_JPEG1_STATUS, POISONED_PF); + break; + default: + break; + } + + if (poison_stat) + dev_info(adev->dev, "Poison detected in JPEG%d sub_block%d\n", + instance, sub_block); + + return poison_stat; +} + +static bool jpeg_v5_0_1_query_ras_poison_status(struct amdgpu_device *adev) +{ + uint32_t inst = 0, sub = 0, poison_stat = 0; + + for (inst = 0; inst < adev->jpeg.num_jpeg_inst; inst++) + for (sub = 0; sub < AMDGPU_JPEG_V5_0_1_MAX_SUB_BLOCK; sub++) + poison_stat += + jpeg_v5_0_1_query_poison_by_instance(adev, inst, sub); + + return !!poison_stat; +} + +static const struct amdgpu_ras_block_hw_ops jpeg_v5_0_1_ras_hw_ops = { + .query_poison_status = jpeg_v5_0_1_query_ras_poison_status, +}; + +static int jpeg_v5_0_1_aca_bank_parser(struct aca_handle *handle, struct aca_bank *bank, + enum aca_smu_type type, void *data) +{ + struct aca_bank_info info; + u64 misc0; + int ret; + + ret = aca_bank_info_decode(bank, &info); + if (ret) + return ret; + + misc0 = bank->regs[ACA_REG_IDX_MISC0]; + switch (type) { + case ACA_SMU_TYPE_UE: + bank->aca_err_type = ACA_ERROR_TYPE_UE; + ret = aca_error_cache_log_bank_error(handle, &info, ACA_ERROR_TYPE_UE, + 1ULL); + break; + case ACA_SMU_TYPE_CE: + bank->aca_err_type = ACA_ERROR_TYPE_CE; + ret = aca_error_cache_log_bank_error(handle, &info, bank->aca_err_type, + ACA_REG__MISC0__ERRCNT(misc0)); + break; + default: + return -EINVAL; + } + + return ret; +} + +/* reference to smu driver if header file */ +static int jpeg_v5_0_1_err_codes[] = { + 16, 17, 18, 19, 20, 21, 22, 23, /* JPEG[0-7][S|D] */ + 24, 25, 26, 27, 28, 29, 30, 31 +}; + +static bool jpeg_v5_0_1_aca_bank_is_valid(struct aca_handle *handle, struct aca_bank *bank, + enum aca_smu_type type, void *data) +{ + u32 instlo; + + instlo = ACA_REG__IPID__INSTANCEIDLO(bank->regs[ACA_REG_IDX_IPID]); + instlo &= GENMASK(31, 1); + + if (instlo != mmSMNAID_AID0_MCA_SMU) + return false; + + if (aca_bank_check_error_codes(handle->adev, bank, + jpeg_v5_0_1_err_codes, + ARRAY_SIZE(jpeg_v5_0_1_err_codes))) + return false; + + return true; +} + +static const struct aca_bank_ops jpeg_v5_0_1_aca_bank_ops = { + .aca_bank_parser = jpeg_v5_0_1_aca_bank_parser, + .aca_bank_is_valid = jpeg_v5_0_1_aca_bank_is_valid, +}; + +static const struct aca_info jpeg_v5_0_1_aca_info = { + .hwip = ACA_HWIP_TYPE_SMU, + .mask = ACA_ERROR_UE_MASK, + .bank_ops = &jpeg_v5_0_1_aca_bank_ops, +}; + +static int jpeg_v5_0_1_ras_late_init(struct amdgpu_device *adev, struct ras_common_if *ras_block) +{ + int r; + + r = amdgpu_ras_block_late_init(adev, ras_block); + if (r) + return r; + + if (amdgpu_ras_is_supported(adev, ras_block->block) && + adev->jpeg.inst->ras_poison_irq.funcs) { + r = amdgpu_irq_get(adev, &adev->jpeg.inst->ras_poison_irq, 0); + if (r) + goto late_fini; + } + + r = amdgpu_ras_bind_aca(adev, AMDGPU_RAS_BLOCK__JPEG, + &jpeg_v5_0_1_aca_info, NULL); + if (r) + goto late_fini; + + return 0; + +late_fini: + amdgpu_ras_block_late_fini(adev, ras_block); + + return r; +} + +static struct amdgpu_jpeg_ras jpeg_v5_0_1_ras = { + .ras_block = { + .hw_ops = &jpeg_v5_0_1_ras_hw_ops, + .ras_late_init = jpeg_v5_0_1_ras_late_init, + }, +}; + +static void jpeg_v5_0_1_set_ras_funcs(struct amdgpu_device *adev) +{ + adev->jpeg.ras = &jpeg_v5_0_1_ras; +} diff --git a/drivers/gpu/drm/amd/amdgpu/jpeg_v5_0_1.h b/drivers/gpu/drm/amd/amdgpu/jpeg_v5_0_1.h index efdab57324e4..a7e58d5fb246 100644 --- a/drivers/gpu/drm/amd/amdgpu/jpeg_v5_0_1.h +++ b/drivers/gpu/drm/amd/amdgpu/jpeg_v5_0_1.h @@ -26,6 +26,9 @@ extern const struct amdgpu_ip_block_version jpeg_v5_0_1_ip_block; +#define regUVD_JRBC0_UVD_JRBC_SCRATCH0_INTERNAL_OFFSET 0x4094 +#define regUVD_JRBC_EXTERNAL_MCM_ADDR_INTERNAL_OFFSET 0x1bffe + #define regUVD_JRBC0_UVD_JRBC_RB_WPTR 0x0640 #define regUVD_JRBC0_UVD_JRBC_RB_WPTR_BASE_IDX 1 #define regUVD_JRBC0_UVD_JRBC_STATUS 0x0649 @@ -98,4 +101,11 @@ extern const struct amdgpu_ip_block_version jpeg_v5_0_1_ip_block; #define regVCN_RRMT_CNTL 0x0940 #define regVCN_RRMT_CNTL_BASE_IDX 1 +enum amdgpu_jpeg_v5_0_1_sub_block { + AMDGPU_JPEG_V5_0_1_JPEG0 = 0, + AMDGPU_JPEG_V5_0_1_JPEG1, + + AMDGPU_JPEG_V5_0_1_MAX_SUB_BLOCK, +}; + #endif /* __JPEG_V5_0_1_H__ */ -- cgit v1.2.3 From b758667f55a09bc86ac0cb230c4e0e5ca931a3a5 Mon Sep 17 00:00:00 2001 From: Mangesh Gadre Date: Tue, 13 May 2025 13:20:20 +0800 Subject: drm/amdgpu: update ras support check update ras support check for vcn 5.0.1 Signed-off-by: Mangesh Gadre Reviewed-by: Stanley.Yang Reviewed-by: Tao Zhou Signed-off-by: Alex Deucher --- drivers/gpu/drm/amd/amdgpu/amdgpu_ras.c | 3 ++- 1 file changed, 2 insertions(+), 1 deletion(-) (limited to 'drivers/gpu') diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_ras.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_ras.c index dc07936d2fcb..8a13a0c7b9cd 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_ras.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_ras.c @@ -3708,7 +3708,8 @@ static void amdgpu_ras_query_ras_capablity_from_vbios(struct amdgpu_device *adev */ if (amdgpu_ip_version(adev, VCN_HWIP, 0) == IP_VERSION(2, 6, 0) || amdgpu_ip_version(adev, VCN_HWIP, 0) == IP_VERSION(4, 0, 0) || - amdgpu_ip_version(adev, VCN_HWIP, 0) == IP_VERSION(4, 0, 3)) + amdgpu_ip_version(adev, VCN_HWIP, 0) == IP_VERSION(4, 0, 3) || + amdgpu_ip_version(adev, VCN_HWIP, 0) == IP_VERSION(5, 0, 1)) adev->ras_hw_enabled |= (1 << AMDGPU_RAS_BLOCK__VCN | 1 << AMDGPU_RAS_BLOCK__JPEG); else -- cgit v1.2.3 From a359288ccb4dd8edb086e7de8fdf6e36f544c922 Mon Sep 17 00:00:00 2001 From: Philip Yang Date: Wed, 14 May 2025 11:13:52 -0400 Subject: drm/amdgpu: seq64 memory unmap uses uninterruptible lock MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit To unmap and free seq64 memory when drm node close to free vm, if there is signal accepted, then taking vm lock failed and leaking seq64 va mapping, and then dmesg has error log "still active bo inside vm". Change to use uninterruptible lock fix the mapping leaking and no dmesg error log. Signed-off-by: Philip Yang Reviewed-by: Christian König Signed-off-by: Alex Deucher --- drivers/gpu/drm/amd/amdgpu/amdgpu_seq64.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) (limited to 'drivers/gpu') diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_seq64.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_seq64.c index 3939761be31c..d45ebfb642ca 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_seq64.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_seq64.c @@ -139,7 +139,7 @@ void amdgpu_seq64_unmap(struct amdgpu_device *adev, struct amdgpu_fpriv *fpriv) vm = &fpriv->vm; - drm_exec_init(&exec, DRM_EXEC_INTERRUPTIBLE_WAIT, 0); + drm_exec_init(&exec, 0, 0); drm_exec_until_all_locked(&exec) { r = amdgpu_vm_lock_pd(vm, &exec, 0); if (likely(!r)) -- cgit v1.2.3 From fa75a9680b381647999f6a6329a993a99a6fd1fa Mon Sep 17 00:00:00 2001 From: Aurabindo Pillai Date: Wed, 21 May 2025 10:22:46 -0400 Subject: Revert "drm/amd/display: [FW Promotion] Release 0.1.11.0" This reverts commit 81fc9ca25f02c53c055b842a40f2a915bd0bd5e0 since it introduces incompatbility with older firmware Signed-off-by: Aurabindo Pillai Acked-by: Alex Deucher Signed-off-by: Alex Deucher --- drivers/gpu/drm/amd/display/dmub/inc/dmub_cmd.h | 34 ++----------------------- 1 file changed, 2 insertions(+), 32 deletions(-) (limited to 'drivers/gpu') diff --git a/drivers/gpu/drm/amd/display/dmub/inc/dmub_cmd.h b/drivers/gpu/drm/amd/display/dmub/inc/dmub_cmd.h index 57fa05bddb45..b66bd10cdc9b 100644 --- a/drivers/gpu/drm/amd/display/dmub/inc/dmub_cmd.h +++ b/drivers/gpu/drm/amd/display/dmub/inc/dmub_cmd.h @@ -2139,11 +2139,6 @@ union dmub_cmd_fams2_config { } stream_v1; //v1 }; -struct dmub_fams2_config_v2 { - struct dmub_cmd_fams2_global_config global; - struct dmub_fams2_stream_static_state_v1 stream_v1[DMUB_MAX_STREAMS]; //v1 -}; - /** * DMUB rb command definition for FAMS2 (merged SubVP, FPO, Legacy) */ @@ -2152,22 +2147,6 @@ struct dmub_rb_cmd_fams2 { union dmub_cmd_fams2_config config; }; -/** - * Indirect buffer descriptor - */ -struct dmub_ib_data { - union dmub_addr src; // location of indirect buffer in memory - uint16_t size; // indirect buffer size in bytes -}; - -/** - * DMUB rb command definition for commands passed over indirect buffer - */ -struct dmub_rb_cmd_ib { - struct dmub_cmd_header header; - struct dmub_ib_data ib_data; -}; - /** * enum dmub_cmd_idle_opt_type - Idle optimization command type. */ @@ -2191,11 +2170,6 @@ enum dmub_cmd_idle_opt_type { * DCN hardware notify power state. */ DMUB_CMD__IDLE_OPT_SET_DC_POWER_STATE = 3, - - /** - * DCN notify to release HW. - */ - DMUB_CMD__IDLE_OPT_RELEASE_HW = 4, }; /** @@ -2957,9 +2931,8 @@ enum dmub_cmd_fams_type { */ DMUB_CMD__FAMS_SET_MANUAL_TRIGGER = 3, DMUB_CMD__FAMS2_CONFIG = 4, - DMUB_CMD__FAMS2_IB_CONFIG = 5, - DMUB_CMD__FAMS2_DRR_UPDATE = 6, - DMUB_CMD__FAMS2_FLIP = 7, + DMUB_CMD__FAMS2_DRR_UPDATE = 5, + DMUB_CMD__FAMS2_FLIP = 6, }; /** @@ -5953,11 +5926,8 @@ union dmub_rb_cmd { * Definition of a DMUB_CMD__PSP_ASSR_ENABLE command. */ struct dmub_rb_cmd_assr_enable assr_enable; - struct dmub_rb_cmd_fams2 fams2_config; - struct dmub_rb_cmd_ib ib_fams2_config; - struct dmub_rb_cmd_fams2_drr_update fams2_drr_update; struct dmub_rb_cmd_fams2_flip fams2_flip; -- cgit v1.2.3 From 42f520f70d6ca066808026d0b6163040ee1d9b9d Mon Sep 17 00:00:00 2001 From: Lijo Lazar Date: Tue, 13 May 2025 18:57:21 +0530 Subject: drm/amd/pm: Fetch partition metrics on SMUv13.0.12 Add support to fetch compute partition related metrics in SMUv13.0.12 SOCs. Signed-off-by: Lijo Lazar Reviewed-by: Asad Kamal Reviewed-by: Hawking Zhang Signed-off-by: Alex Deucher --- .../gpu/drm/amd/pm/swsmu/smu13/smu_v13_0_12_ppt.c | 56 ++++++++++++++++++++++ .../gpu/drm/amd/pm/swsmu/smu13/smu_v13_0_6_ppt.c | 9 ++++ .../gpu/drm/amd/pm/swsmu/smu13/smu_v13_0_6_ppt.h | 3 ++ 3 files changed, 68 insertions(+) (limited to 'drivers/gpu') 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 5a8824cc1c63..69f92bd35bf2 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 @@ -322,6 +322,62 @@ int smu_v13_0_12_get_smu_metrics_data(struct smu_context *smu, return ret; } +ssize_t smu_v13_0_12_get_xcp_metrics(struct smu_context *smu, struct amdgpu_xcp *xcp, void *table, void *smu_metrics) +{ + const u8 num_jpeg_rings = NUM_JPEG_RINGS_FW; + struct amdgpu_partition_metrics_v1_0 *xcp_metrics; + struct amdgpu_device *adev = smu->adev; + MetricsTable_t *metrics; + int inst, j, k, idx; + u32 inst_mask; + + metrics = (MetricsTable_t *)smu_metrics; + xcp_metrics = (struct amdgpu_partition_metrics_v1_0 *) table; + smu_cmn_init_partition_metrics(xcp_metrics, 1, 0); + amdgpu_xcp_get_inst_details(xcp, AMDGPU_XCP_VCN, &inst_mask); + idx = 0; + for_each_inst(k, inst_mask) { + /* Both JPEG and VCN has same instance */ + inst = GET_INST(VCN, k); + for (j = 0; j < num_jpeg_rings; ++j) { + xcp_metrics->jpeg_busy[(idx * num_jpeg_rings) + j] = + SMUQ10_ROUND(metrics-> + JpegBusy[(inst * num_jpeg_rings) + j]); + } + xcp_metrics->vcn_busy[idx] = + SMUQ10_ROUND(metrics->VcnBusy[inst]); + xcp_metrics->current_vclk0[idx] = SMUQ10_ROUND( + metrics->VclkFrequency[inst]); + xcp_metrics->current_dclk0[idx] = SMUQ10_ROUND( + metrics->DclkFrequency[inst]); + xcp_metrics->current_socclk[idx] = SMUQ10_ROUND( + metrics->SocclkFrequency[inst]); + + idx++; + } + + xcp_metrics->current_uclk = + SMUQ10_ROUND(metrics->UclkFrequency); + + amdgpu_xcp_get_inst_details(xcp, AMDGPU_XCP_GFX, &inst_mask); + idx = 0; + for_each_inst(k, inst_mask) { + inst = GET_INST(GC, k); + xcp_metrics->current_gfxclk[idx] = SMUQ10_ROUND(metrics->GfxclkFrequency[inst]); + xcp_metrics->gfx_busy_inst[idx] = SMUQ10_ROUND(metrics->GfxBusy[inst]); + xcp_metrics->gfx_busy_acc[idx] = SMUQ10_ROUND(metrics->GfxBusyAcc[inst]); + if (smu_v13_0_6_cap_supported(smu, SMU_CAP(HST_LIMIT_METRICS))) { + xcp_metrics->gfx_below_host_limit_ppt_acc[idx] = SMUQ10_ROUND(metrics->GfxclkBelowHostLimitPptAcc[inst]); + xcp_metrics->gfx_below_host_limit_thm_acc[idx] = SMUQ10_ROUND(metrics->GfxclkBelowHostLimitThmAcc[inst]); + xcp_metrics->gfx_low_utilization_acc[idx] = SMUQ10_ROUND(metrics->GfxclkLowUtilizationAcc[inst]); + xcp_metrics->gfx_below_host_limit_total_acc[idx] = SMUQ10_ROUND(metrics->GfxclkBelowHostLimitTotalAcc[inst]); + } + idx++; + } + + return sizeof(*xcp_metrics); +} + ssize_t smu_v13_0_12_get_gpu_metrics(struct smu_context *smu, void **table) { struct smu_table_context *smu_table = &smu->smu_table; diff --git a/drivers/gpu/drm/amd/pm/swsmu/smu13/smu_v13_0_6_ppt.c b/drivers/gpu/drm/amd/pm/swsmu/smu13/smu_v13_0_6_ppt.c index 78d831c20768..0a9488576a4e 100644 --- a/drivers/gpu/drm/amd/pm/swsmu/smu13/smu_v13_0_6_ppt.c +++ b/drivers/gpu/drm/amd/pm/swsmu/smu13/smu_v13_0_6_ppt.c @@ -2573,6 +2573,14 @@ static ssize_t smu_v13_0_6_get_xcp_metrics(struct smu_context *smu, int xcp_id, kfree(metrics_v0); return ret; } + + if (amdgpu_ip_version(smu->adev, MP1_HWIP, 0) == + IP_VERSION(13, 0, 12) && + smu_v13_0_6_cap_supported(smu, SMU_CAP(STATIC_METRICS))) { + ret = smu_v13_0_12_get_xcp_metrics(smu, xcp, table, metrics_v0); + goto out; + } + metrics_v1 = (MetricsTableV1_t *)metrics_v0; metrics_v2 = (MetricsTableV2_t *)metrics_v0; @@ -2642,6 +2650,7 @@ static ssize_t smu_v13_0_6_get_xcp_metrics(struct smu_context *smu, int xcp_id, idx++; } } +out: kfree(metrics_v0); return sizeof(*xcp_metrics); diff --git a/drivers/gpu/drm/amd/pm/swsmu/smu13/smu_v13_0_6_ppt.h b/drivers/gpu/drm/amd/pm/swsmu/smu13/smu_v13_0_6_ppt.h index 1ccc150882eb..1a54675c576f 100644 --- a/drivers/gpu/drm/amd/pm/swsmu/smu13/smu_v13_0_6_ppt.h +++ b/drivers/gpu/drm/amd/pm/swsmu/smu13/smu_v13_0_6_ppt.h @@ -81,6 +81,9 @@ int smu_v13_0_12_setup_driver_pptable(struct smu_context *smu); int smu_v13_0_12_get_smu_metrics_data(struct smu_context *smu, MetricsMember_t member, uint32_t *value); ssize_t smu_v13_0_12_get_gpu_metrics(struct smu_context *smu, void **table); +ssize_t smu_v13_0_12_get_xcp_metrics(struct smu_context *smu, + struct amdgpu_xcp *xcp, void *table, + void *smu_metrics); extern const struct cmn2asic_mapping smu_v13_0_12_feature_mask_map[]; extern const struct cmn2asic_msg_mapping smu_v13_0_12_message_map[]; #endif -- cgit v1.2.3 From e485502c37b097b0bd773baa7e2741bf7bd2909a Mon Sep 17 00:00:00 2001 From: Mario Limonciello Date: Thu, 22 May 2025 09:13:28 -0500 Subject: Revert "drm/amd: Keep display off while going into S4" commit 68bfdc8dc0a1a ("drm/amd: Keep display off while going into S4") attempted to keep displays off during the S4 sequence by not resuming display IP. This however leads to hangs because DRM clients such as the console can try to access registers and cause a hang. Closes: https://gitlab.freedesktop.org/drm/amd/-/issues/4155 Fixes: 68bfdc8dc0a1a ("drm/amd: Keep display off while going into S4") Reviewed-by: Alex Deucher Link: https://lore.kernel.org/r/20250522141328.115095-1-mario.limonciello@amd.com Signed-off-by: Mario Limonciello Signed-off-by: Alex Deucher --- drivers/gpu/drm/amd/display/amdgpu_dm/amdgpu_dm.c | 5 ----- 1 file changed, 5 deletions(-) (limited to 'drivers/gpu') 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 2bb347771aa1..90889f6867aa 100644 --- a/drivers/gpu/drm/amd/display/amdgpu_dm/amdgpu_dm.c +++ b/drivers/gpu/drm/amd/display/amdgpu_dm/amdgpu_dm.c @@ -3383,11 +3383,6 @@ static int dm_resume(struct amdgpu_ip_block *ip_block) return 0; } - - /* leave display off for S4 sequence */ - if (adev->in_s4) - return 0; - /* Recreate dc_state - DC invalidates it when setting power state to S3. */ dc_state_release(dm_state->context); dm_state->context = dc_state_create(dm->dc, NULL); -- cgit v1.2.3 From 022546cdbf26ec6f42af7e2706702d5703df42f6 Mon Sep 17 00:00:00 2001 From: Arnd Bergmann Date: Fri, 23 May 2025 14:11:22 +0200 Subject: drm/panel: nt37801: select CONFIG_DRM_DISPLAY_DSC_HELPER The newly added driver uses the DSC helper module, but does not select its Kconfig symbol, so configurations are possible that cause a link failure: ERROR: modpost: "drm_dsc_pps_payload_pack" [drivers/gpu/drm/panel/panel-novatek-nt37801.ko] undefined! Fixes: 4fca6849864d ("drm/panel: Add Novatek NT37801 panel driver") Signed-off-by: Arnd Bergmann Reviewed-by: Dmitry Baryshkov Reviewed-by: Krzysztof Kozlowski Link: https://lore.kernel.org/r/20250523121127.2269693-1-arnd@kernel.org Signed-off-by: Dmitry Baryshkov --- drivers/gpu/drm/panel/Kconfig | 2 ++ 1 file changed, 2 insertions(+) (limited to 'drivers/gpu') diff --git a/drivers/gpu/drm/panel/Kconfig b/drivers/gpu/drm/panel/Kconfig index 721581d425b4..cfebb08e8a62 100644 --- a/drivers/gpu/drm/panel/Kconfig +++ b/drivers/gpu/drm/panel/Kconfig @@ -522,6 +522,8 @@ config DRM_PANEL_NOVATEK_NT37801 depends on OF depends on DRM_MIPI_DSI depends on BACKLIGHT_CLASS_DEVICE + select DRM_DISPLAY_DSC_HELPER + select DRM_DISPLAY_HELPER help Say Y here if you want to enable support for Novatek NT37801 (or NT37810) AMOLED DSI Video Mode LCD panel module with 1440x3200 -- cgit v1.2.3 From 4673dec88da803fa23f1af9e04761683a30dd6aa Mon Sep 17 00:00:00 2001 From: Dan Carpenter Date: Fri, 23 May 2025 19:09:03 +0300 Subject: drm/panel: nt37801: Fix IS_ERR() vs NULL check in probe() The devm_drm_panel_alloc() function returns error pointers, it doesn't return NULL. Update the check to match. Fixes: 4fca6849864d ("drm/panel: Add Novatek NT37801 panel driver") Signed-off-by: Dan Carpenter Reviewed-by: Krzysztof Kozlowski Reviewed-by: Jessica Zhang Link: https://lore.kernel.org/r/aDCdn9r_ZAUTRpWn@stanley.mountain Signed-off-by: Dmitry Baryshkov --- drivers/gpu/drm/panel/panel-novatek-nt37801.c | 4 ++-- 1 file changed, 2 insertions(+), 2 deletions(-) (limited to 'drivers/gpu') diff --git a/drivers/gpu/drm/panel/panel-novatek-nt37801.c b/drivers/gpu/drm/panel/panel-novatek-nt37801.c index 84d367eab058..d6a37d7e0cc6 100644 --- a/drivers/gpu/drm/panel/panel-novatek-nt37801.c +++ b/drivers/gpu/drm/panel/panel-novatek-nt37801.c @@ -257,8 +257,8 @@ static int novatek_nt37801_probe(struct mipi_dsi_device *dsi) ctx = devm_drm_panel_alloc(dev, struct novatek_nt37801, panel, &novatek_nt37801_panel_funcs, DRM_MODE_CONNECTOR_DSI); - if (!ctx) - return -ENOMEM; + if (IS_ERR(ctx)) + return PTR_ERR(ctx); ret = devm_regulator_bulk_get_const(dev, ARRAY_SIZE(novatek_nt37801_supplies), -- cgit v1.2.3 From 7d476f18abeba3241cd3a2b712b2666a23b6f6dd Mon Sep 17 00:00:00 2001 From: Imre Deak Date: Tue, 20 May 2025 17:22:19 +0300 Subject: drm/i915/dp: Fix the enabling/disabling of audio SDP splitting Adjust the enabling/disabling steps of the DP audio SDP splitting according to a recent Bspec update. This moves the enabling to the audio codec enable sequence after the transcoder is enabled and disables SDP splitting explicitly during the audio disable sequence. Bspec requires waiting for a vblank event after the transcoder is enabled and before SDP splitting is enabled. There is no need for an explicit wait for this, since after the transcoder is enabled this vblank event is guaranteed to have happened via a flip done wait (see intel_atomic_commit_tail() -> drm_atomic_helper_wait_for_flip_done()). The bspec update is for LNL+ only, but the HW team clarified that this has been always the intended sequence on all platforms and bspec will be updated everywhere accordingly. The way SDP splitting was originally enabled matched the version of bspec at that time. Adding here the Fixes: line still, since this change fixes a FIFO underrun on PTL during output enabling when DSC is enabled. Bspec: 49283, 68943 Fixes: 8853750dbad8 ("drm/i915: Enable SDP split for DP2.0") Cc: Vinod Govindapillai Acked-by: Jani Nikula Reviewed-by: Vinod Govindapillai Signed-off-by: Imre Deak Link: https://lore.kernel.org/r/20250520142219.1688401-1-imre.deak@intel.com (cherry picked from commit 56764c845aa5be14cd53702fc9f2da23e25857de) Signed-off-by: Joonas Lahtinen --- drivers/gpu/drm/i915/display/intel_audio.c | 27 +++++++++++++++++---------- drivers/gpu/drm/i915/display/intel_audio.h | 1 - drivers/gpu/drm/i915/display/intel_ddi.c | 3 --- drivers/gpu/drm/i915/display/intel_dp_mst.c | 2 -- 4 files changed, 17 insertions(+), 16 deletions(-) (limited to 'drivers/gpu') diff --git a/drivers/gpu/drm/i915/display/intel_audio.c b/drivers/gpu/drm/i915/display/intel_audio.c index 40d8bbd8107d..55af3a553c58 100644 --- a/drivers/gpu/drm/i915/display/intel_audio.c +++ b/drivers/gpu/drm/i915/display/intel_audio.c @@ -397,6 +397,19 @@ hsw_audio_config_update(struct intel_encoder *encoder, hsw_hdmi_audio_config_update(encoder, crtc_state); } +static void intel_audio_sdp_split_update(const struct intel_crtc_state *crtc_state, + bool enable) +{ + struct intel_display *display = to_intel_display(crtc_state); + enum transcoder trans = crtc_state->cpu_transcoder; + + if (!HAS_DP20(display)) + return; + + intel_de_rmw(display, AUD_DP_2DOT0_CTRL(trans), AUD_ENABLE_SDP_SPLIT, + enable && crtc_state->sdp_split_enable ? AUD_ENABLE_SDP_SPLIT : 0); +} + static void hsw_audio_codec_disable(struct intel_encoder *encoder, const struct intel_crtc_state *old_crtc_state, const struct drm_connector_state *old_conn_state) @@ -430,6 +443,8 @@ static void hsw_audio_codec_disable(struct intel_encoder *encoder, if (needs_wa_14020863754(display)) intel_de_rmw(display, AUD_CHICKENBIT_REG3, DACBE_DISABLE_MIN_HBLANK_FIX, 0); + intel_audio_sdp_split_update(old_crtc_state, false); + mutex_unlock(&display->audio.mutex); } @@ -555,6 +570,8 @@ static void hsw_audio_codec_enable(struct intel_encoder *encoder, if (intel_crtc_has_type(crtc_state, INTEL_OUTPUT_DP)) enable_audio_dsc_wa(encoder, crtc_state); + intel_audio_sdp_split_update(crtc_state, true); + if (needs_wa_14020863754(display)) intel_de_rmw(display, AUD_CHICKENBIT_REG3, 0, DACBE_DISABLE_MIN_HBLANK_FIX); @@ -681,16 +698,6 @@ static void ibx_audio_codec_enable(struct intel_encoder *encoder, mutex_unlock(&display->audio.mutex); } -void intel_audio_sdp_split_update(const struct intel_crtc_state *crtc_state) -{ - struct intel_display *display = to_intel_display(crtc_state); - enum transcoder trans = crtc_state->cpu_transcoder; - - if (HAS_DP20(display)) - intel_de_rmw(display, AUD_DP_2DOT0_CTRL(trans), AUD_ENABLE_SDP_SPLIT, - crtc_state->sdp_split_enable ? AUD_ENABLE_SDP_SPLIT : 0); -} - bool intel_audio_compute_config(struct intel_encoder *encoder, struct intel_crtc_state *crtc_state, struct drm_connector_state *conn_state) diff --git a/drivers/gpu/drm/i915/display/intel_audio.h b/drivers/gpu/drm/i915/display/intel_audio.h index ad49eefa7182..42cf886f3d24 100644 --- a/drivers/gpu/drm/i915/display/intel_audio.h +++ b/drivers/gpu/drm/i915/display/intel_audio.h @@ -31,6 +31,5 @@ int intel_audio_min_cdclk(const struct intel_crtc_state *crtc_state); void intel_audio_init(struct intel_display *display); void intel_audio_register(struct intel_display *display); void intel_audio_deinit(struct intel_display *display); -void intel_audio_sdp_split_update(const struct intel_crtc_state *crtc_state); #endif /* __INTEL_AUDIO_H__ */ diff --git a/drivers/gpu/drm/i915/display/intel_ddi.c b/drivers/gpu/drm/i915/display/intel_ddi.c index 74132c1d6385..d58f8fc37326 100644 --- a/drivers/gpu/drm/i915/display/intel_ddi.c +++ b/drivers/gpu/drm/i915/display/intel_ddi.c @@ -3507,9 +3507,6 @@ static void intel_ddi_enable(struct intel_atomic_state *state, intel_vrr_transcoder_enable(crtc_state); - /* Enable/Disable DP2.0 SDP split config before transcoder */ - intel_audio_sdp_split_update(crtc_state); - /* 128b/132b SST */ if (!is_hdmi && intel_dp_is_uhbr(crtc_state)) { struct intel_dp *intel_dp = enc_to_intel_dp(encoder); diff --git a/drivers/gpu/drm/i915/display/intel_dp_mst.c b/drivers/gpu/drm/i915/display/intel_dp_mst.c index c1fd6aceec2c..7335af92cfb0 100644 --- a/drivers/gpu/drm/i915/display/intel_dp_mst.c +++ b/drivers/gpu/drm/i915/display/intel_dp_mst.c @@ -1328,8 +1328,6 @@ static void mst_stream_enable(struct intel_atomic_state *state, FECSTALL_DIS_DPTSTREAM_DPTTG, pipe_config->fec_enable ? FECSTALL_DIS_DPTSTREAM_DPTTG : 0); - intel_audio_sdp_split_update(pipe_config); - intel_enable_transcoder(pipe_config); for_each_pipe_crtc_modeset_enable(display, pipe_crtc, pipe_config, i) { -- cgit v1.2.3 From 25eeba495b2fc16037647c1a51bcdf6fc157af5c Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Ville=20Syrj=C3=A4l=C3=A4?= Date: Mon, 12 May 2025 21:22:15 +0200 Subject: drm/i915/gem: Allow EXEC_CAPTURE on recoverable contexts on DG1 MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit The intel-media-driver is currently broken on DG1 because it uses EXEC_CAPTURE with recovarable contexts. Relax the check to allow that. I've also submitted a fix for the intel-media-driver: https://github.com/intel/media-driver/pull/1920 Cc: stable@vger.kernel.org # v6.0+ Cc: Matthew Auld Cc: Thomas Hellström Testcase: igt/gem_exec_capture/capture-invisible Fixes: 71b1669ea9bd ("drm/i915/uapi: tweak error capture on recoverable contexts") Reviewed-by: Andi Shyti Signed-off-by: Ville Syrjälä Signed-off-by: Andi Shyti Link: https://lore.kernel.org/r/20250411144313.11660-2-ville.syrjala@linux.intel.com (cherry picked from commit d6e020819612a4a06207af858e0978be4d3e3140) Signed-off-by: Joonas Lahtinen --- drivers/gpu/drm/i915/gem/i915_gem_execbuffer.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) (limited to 'drivers/gpu') diff --git a/drivers/gpu/drm/i915/gem/i915_gem_execbuffer.c b/drivers/gpu/drm/i915/gem/i915_gem_execbuffer.c index ca7e9216934a..ea9d5063ce78 100644 --- a/drivers/gpu/drm/i915/gem/i915_gem_execbuffer.c +++ b/drivers/gpu/drm/i915/gem/i915_gem_execbuffer.c @@ -2013,7 +2013,7 @@ static int eb_capture_stage(struct i915_execbuffer *eb) continue; if (i915_gem_context_is_recoverable(eb->gem_context) && - (IS_DGFX(eb->i915) || GRAPHICS_VER_FULL(eb->i915) > IP_VER(12, 0))) + GRAPHICS_VER_FULL(eb->i915) > IP_VER(12, 10)) return -EINVAL; for_each_batch_create_order(eb, j) { -- cgit v1.2.3 From ed5915cfce2abb9a553c3737badebd4a11d6c9c7 Mon Sep 17 00:00:00 2001 From: Joonas Lahtinen Date: Thu, 22 May 2025 09:41:27 +0300 Subject: Revert "drm/i915/gem: Allow EXEC_CAPTURE on recoverable contexts on DG1" MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit This reverts commit d6e020819612a4a06207af858e0978be4d3e3140. The IS_DGFX check was put in place because error capture of buffer objects is expected to be broken on devices with VRAM. Userspace fix[1] to the impacted media driver has been submitted, merged and a new driver release is out as 25.2.3 where the capture flag is dropped on DG1 thus unblocking the usage of media driver on DG1. [1] https://github.com/intel/media-driver/commit/93c07d9b4b96a78bab21f6acd4eb863f4313ea4a Cc: stable@vger.kernel.org # v6.0+ Cc: Ville Syrjälä Cc: Andi Shyti Cc: Matthew Auld Cc: Thomas Hellström Cc: Tvrtko Ursulin Acked-by: Tvrtko Ursulin Reviewed-by: Andi Shyti Link: https://lore.kernel.org/r/20250522064127.24293-1-joonas.lahtinen@linux.intel.com [Joonas: Update message to point out the merged userspace fix] Signed-off-by: Joonas Lahtinen (cherry picked from commit d2dc30e0aa252830f908c8e793d3139d51321370) Signed-off-by: Joonas Lahtinen --- drivers/gpu/drm/i915/gem/i915_gem_execbuffer.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) (limited to 'drivers/gpu') diff --git a/drivers/gpu/drm/i915/gem/i915_gem_execbuffer.c b/drivers/gpu/drm/i915/gem/i915_gem_execbuffer.c index ea9d5063ce78..ca7e9216934a 100644 --- a/drivers/gpu/drm/i915/gem/i915_gem_execbuffer.c +++ b/drivers/gpu/drm/i915/gem/i915_gem_execbuffer.c @@ -2013,7 +2013,7 @@ static int eb_capture_stage(struct i915_execbuffer *eb) continue; if (i915_gem_context_is_recoverable(eb->gem_context) && - GRAPHICS_VER_FULL(eb->i915) > IP_VER(12, 10)) + (IS_DGFX(eb->i915) || GRAPHICS_VER_FULL(eb->i915) > IP_VER(12, 0))) return -EINVAL; for_each_batch_create_order(eb, j) { -- cgit v1.2.3 From be9b3f9a54101c19226c25ba7163d291183777a0 Mon Sep 17 00:00:00 2001 From: Nicolas Frattaroli Date: Tue, 27 May 2025 19:57:08 +0200 Subject: drm/connector: only call HDMI audio helper plugged cb if non-null On driver remove, sound/soc/codecs/hdmi-codec.c calls the plugged_cb with NULL as the callback function and codec_dev, as seen in its hdmi_remove function. The HDMI audio helper then happily tries calling said null function pointer, and produces an Oops as a result. Fix this by only executing the callback if fn is non-null. This means the .plugged_cb and .plugged_cb_dev members still get appropriately cleared. Fixes: baf616647fe6 ("drm/connector: implement generic HDMI audio helpers") Signed-off-by: Nicolas Frattaroli Reviewed-by: Dmitry Baryshkov Link: https://lore.kernel.org/r/20250527-hdmi-audio-helper-remove-fix-v1-1-6cf77de364d8@collabora.com Signed-off-by: Dmitry Baryshkov --- drivers/gpu/drm/display/drm_hdmi_audio_helper.c | 3 ++- 1 file changed, 2 insertions(+), 1 deletion(-) (limited to 'drivers/gpu') diff --git a/drivers/gpu/drm/display/drm_hdmi_audio_helper.c b/drivers/gpu/drm/display/drm_hdmi_audio_helper.c index 05afc9f0bdd6..ae8a0cf595fc 100644 --- a/drivers/gpu/drm/display/drm_hdmi_audio_helper.c +++ b/drivers/gpu/drm/display/drm_hdmi_audio_helper.c @@ -103,7 +103,8 @@ static int drm_connector_hdmi_audio_hook_plugged_cb(struct device *dev, connector->hdmi_audio.plugged_cb = fn; connector->hdmi_audio.plugged_cb_dev = codec_dev; - fn(codec_dev, connector->hdmi_audio.last_state); + if (fn) + fn(codec_dev, connector->hdmi_audio.last_state); mutex_unlock(&connector->hdmi_audio.lock); -- cgit v1.2.3 From 6579a03e68ffa5feb2d2823dea16ca7466f6de16 Mon Sep 17 00:00:00 2001 From: Damon Ding Date: Sun, 2 Mar 2025 16:30:43 +0800 Subject: drm/bridge: analogix_dp: Remove the unnecessary calls to clk_disable_unprepare() during probing With the commit f37952339cc2 ("drm/bridge: analogix_dp: handle clock via runtime PM"), the PM operations can help enable/disable the clock. The err_disable_clk label and clk_disable_unprepare() operations are no longer necessary because the analogix_dp_resume() will not be called during probing. Fixes: f37952339cc2 ("drm/bridge: analogix_dp: handle clock via runtime PM") Suggested-by: Douglas Anderson Reviewed-by: Douglas Anderson Signed-off-by: Damon Ding Reviewed-by: Heiko Stuebner Link: https://lore.kernel.org/r/20250302083043.3197235-1-damon.ding@rock-chips.com Signed-off-by: Dmitry Baryshkov --- drivers/gpu/drm/bridge/analogix/analogix_dp_core.c | 18 +++++------------- 1 file changed, 5 insertions(+), 13 deletions(-) (limited to 'drivers/gpu') diff --git a/drivers/gpu/drm/bridge/analogix/analogix_dp_core.c b/drivers/gpu/drm/bridge/analogix/analogix_dp_core.c index a761941bc3c2..01201fff59a6 100644 --- a/drivers/gpu/drm/bridge/analogix/analogix_dp_core.c +++ b/drivers/gpu/drm/bridge/analogix/analogix_dp_core.c @@ -1531,10 +1531,8 @@ analogix_dp_probe(struct device *dev, struct analogix_dp_plat_data *plat_data) } dp->reg_base = devm_platform_ioremap_resource(pdev, 0); - if (IS_ERR(dp->reg_base)) { - ret = PTR_ERR(dp->reg_base); - goto err_disable_clk; - } + if (IS_ERR(dp->reg_base)) + return ERR_CAST(dp->reg_base); dp->force_hpd = of_property_read_bool(dev->of_node, "force-hpd"); @@ -1546,8 +1544,7 @@ analogix_dp_probe(struct device *dev, struct analogix_dp_plat_data *plat_data) if (IS_ERR(dp->hpd_gpiod)) { dev_err(dev, "error getting HDP GPIO: %ld\n", PTR_ERR(dp->hpd_gpiod)); - ret = PTR_ERR(dp->hpd_gpiod); - goto err_disable_clk; + return ERR_CAST(dp->hpd_gpiod); } if (dp->hpd_gpiod) { @@ -1567,8 +1564,7 @@ analogix_dp_probe(struct device *dev, struct analogix_dp_plat_data *plat_data) if (dp->irq == -ENXIO) { dev_err(&pdev->dev, "failed to get irq\n"); - ret = -ENODEV; - goto err_disable_clk; + return ERR_PTR(-ENODEV); } ret = devm_request_threaded_irq(&pdev->dev, dp->irq, @@ -1577,7 +1573,7 @@ analogix_dp_probe(struct device *dev, struct analogix_dp_plat_data *plat_data) irq_flags, "analogix-dp", dp); if (ret) { dev_err(&pdev->dev, "failed to request irq\n"); - goto err_disable_clk; + return ERR_PTR(ret); } dp->aux.name = "DP-AUX"; @@ -1593,10 +1589,6 @@ analogix_dp_probe(struct device *dev, struct analogix_dp_plat_data *plat_data) goto err_disable_clk; return dp; - -err_disable_clk: - clk_disable_unprepare(dp->clock); - return ERR_PTR(ret); } EXPORT_SYMBOL_GPL(analogix_dp_probe); -- cgit v1.2.3 From fd03f82a026cc03cb8051a8c6487c99f96c9029f Mon Sep 17 00:00:00 2001 From: Heiko Stuebner Date: Wed, 28 May 2025 00:51:19 +0200 Subject: drm/bridge: analogix_dp: Fix clk-disable removal Commit 6579a03e68ff ("drm/bridge: analogix_dp: Remove the unnecessary calls to clk_disable_unprepare() during probing") removed the mismatched clock_disable calls from analogix_dp_probe. But that patch was created and sent before commit e5e9fa9f7aad ("drm/bridge: analogix_dp: Add support to get panel from the DP AUX bus") was merged, so couldn't know about this change. So in the original patch the last change is if (ret) { dev_err(&pdev->dev, "failed to request irq\n"); - goto err_disable_clk; + return ERR_PTR(ret); } disable_irq(dp->irq); return dp; - -err_disable_clk: - clk_disable_unprepare(dp->clock); - return ERR_PTR(ret); } EXPORT_SYMBOL_GPL(analogix_dp_probe); the analogix_dp_core.c actually now has the runtime-pm handling between disable_irq() and return do introducing another goto err_clk_disable there. So remove that one too and return an error pointer, to not create build breakage. Fixes: 6579a03e68ff ("drm/bridge: analogix_dp: Remove the unnecessary calls to clk_disable_unprepare() during probing") Signed-off-by: Heiko Stuebner Reviewed-by: Dmitry Baryshkov Link: https://lore.kernel.org/r/20250527225120.3361663-1-heiko@sntech.de Signed-off-by: Dmitry Baryshkov --- drivers/gpu/drm/bridge/analogix/analogix_dp_core.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) (limited to 'drivers/gpu') diff --git a/drivers/gpu/drm/bridge/analogix/analogix_dp_core.c b/drivers/gpu/drm/bridge/analogix/analogix_dp_core.c index 01201fff59a6..505eec6b819b 100644 --- a/drivers/gpu/drm/bridge/analogix/analogix_dp_core.c +++ b/drivers/gpu/drm/bridge/analogix/analogix_dp_core.c @@ -1586,7 +1586,7 @@ analogix_dp_probe(struct device *dev, struct analogix_dp_plat_data *plat_data) pm_runtime_set_autosuspend_delay(dp->dev, 100); ret = devm_pm_runtime_enable(dp->dev); if (ret) - goto err_disable_clk; + return ERR_PTR(ret); return dp; } -- cgit v1.2.3 From 55f8aa083604ce098c9d6a0911c6bcde15d03a80 Mon Sep 17 00:00:00 2001 From: Rodrigo Vivi Date: Wed, 21 May 2025 12:51:47 -0400 Subject: drm/xe: Make xe_gt_freq part of the Documentation MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit The documentation was created with the creation of the component, however it has never been actually shown in the actual Documentation. While doing this, fixes the identation style, to avoid new warnings while building htmldocs. Fixes: bef52b5c7a19 ("drm/xe: Create a xe_gt_freq component for raw management and sysfs") Reviewed-by: Lucas De Marchi Link: https://lore.kernel.org/r/20250521165146.39616-3-rodrigo.vivi@intel.com Signed-off-by: Rodrigo Vivi (cherry picked from commit af53f0fd99c3bbb3afd29f1612c9e88c5a92cc01) Signed-off-by: Thomas Hellström --- drivers/gpu/drm/xe/xe_gt_freq.c | 2 ++ 1 file changed, 2 insertions(+) (limited to 'drivers/gpu') diff --git a/drivers/gpu/drm/xe/xe_gt_freq.c b/drivers/gpu/drm/xe/xe_gt_freq.c index 868a5d2c1a52..3293d89f1386 100644 --- a/drivers/gpu/drm/xe/xe_gt_freq.c +++ b/drivers/gpu/drm/xe/xe_gt_freq.c @@ -32,6 +32,7 @@ * Xe's Freq provides a sysfs API for frequency management: * * device/tile#/gt#/freq0/_freq *read-only* files: + * * - act_freq: The actual resolved frequency decided by PCODE. * - cur_freq: The current one requested by GuC PC to the PCODE. * - rpn_freq: The Render Performance (RP) N level, which is the minimal one. @@ -39,6 +40,7 @@ * - rp0_freq: The Render Performance (RP) 0 level, which is the maximum one. * * device/tile#/gt#/freq0/_freq *read-write* files: + * * - min_freq: Min frequency request. * - max_freq: Max frequency request. * If max <= min, then freq_min becomes a fixed frequency request. -- cgit v1.2.3 From 40493d97b329f8185c0f04dc0ef2b9ffc58e7f3b Mon Sep 17 00:00:00 2001 From: Rodrigo Vivi Date: Wed, 21 May 2025 12:51:48 -0400 Subject: drm/xe: Add missing documentation of rpa_freq MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit While at it, already adjust the rpe_freq frequency, to highlight that both are calculated by PCODE at runtime. Fixes: c6aac2fa77a3 ("drm/xe: Introduce the RPa information") Cc: Vinay Belgaumkar Cc: Lucas De Marchi Reviewed-by: Lucas De Marchi Reviewed-by: Vinay Belgaumkar Link: https://lore.kernel.org/r/20250521165146.39616-4-rodrigo.vivi@intel.com Signed-off-by: Rodrigo Vivi (cherry picked from commit 39578fa40420fb11dbe4f42225a347e945d8fd0e) Signed-off-by: Thomas Hellström --- drivers/gpu/drm/xe/xe_gt_freq.c | 3 +++ 1 file changed, 3 insertions(+) (limited to 'drivers/gpu') diff --git a/drivers/gpu/drm/xe/xe_gt_freq.c b/drivers/gpu/drm/xe/xe_gt_freq.c index 3293d89f1386..60d9354e7dbf 100644 --- a/drivers/gpu/drm/xe/xe_gt_freq.c +++ b/drivers/gpu/drm/xe/xe_gt_freq.c @@ -36,7 +36,10 @@ * - act_freq: The actual resolved frequency decided by PCODE. * - cur_freq: The current one requested by GuC PC to the PCODE. * - rpn_freq: The Render Performance (RP) N level, which is the minimal one. + * - rpa_freq: The Render Performance (RP) A level, which is the achiveable one. + * Calculated by PCODE at runtime based on multiple running conditions * - rpe_freq: The Render Performance (RP) E level, which is the efficient one. + * Calculated by PCODE at runtime based on multiple running conditions * - rp0_freq: The Render Performance (RP) 0 level, which is the maximum one. * * device/tile#/gt#/freq0/_freq *read-write* files: -- cgit v1.2.3 From 3e9d9df850f3f93261a3259751394643f100a52a Mon Sep 17 00:00:00 2001 From: Pratap Nirujogi Date: Sun, 9 Mar 2025 16:53:01 -0400 Subject: drm/amd/amdgpu: Add GPIO resources required for amdisp ISP is a child device to GFX, and its device specific information is not available in ACPI. Adding the 2 GPIO resources required for ISP_v4_1_1 in amdgpu_isp driver. - GPIO 0 to allow sensor driver to enable and disable sensor module. - GPIO 85 to allow ISP driver to enable and disable ISP RGB streaming mode. Signed-off-by: Pratap Nirujogi Reviewed-by: Mario Limonciello Signed-off-by: Alex Deucher --- drivers/gpu/drm/amd/amdgpu/Kconfig | 2 +- drivers/gpu/drm/amd/amdgpu/amdgpu.h | 4 ++++ drivers/gpu/drm/amd/amdgpu/amdgpu_acpi.c | 32 +++++++++++++++++++++++++++++++- drivers/gpu/drm/amd/amdgpu/isp_v4_1_1.c | 31 +++++++++++++++++++++++++++++++ 4 files changed, 67 insertions(+), 2 deletions(-) (limited to 'drivers/gpu') diff --git a/drivers/gpu/drm/amd/amdgpu/Kconfig b/drivers/gpu/drm/amd/amdgpu/Kconfig index 1a11cab741ac..6bd9aa5dbef7 100644 --- a/drivers/gpu/drm/amd/amdgpu/Kconfig +++ b/drivers/gpu/drm/amd/amdgpu/Kconfig @@ -77,7 +77,7 @@ config DRM_AMDGPU_USERPTR config DRM_AMD_ISP bool "Enable AMD Image Signal Processor IP support" - depends on DRM_AMDGPU + depends on DRM_AMDGPU && ACPI select MFD_CORE select PM_GENERIC_DOMAINS if PM help diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu.h b/drivers/gpu/drm/amd/amdgpu/amdgpu.h index 836ea081088a..a5ccd0ada16a 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu.h +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu.h @@ -1713,6 +1713,10 @@ static inline bool amdgpu_acpi_is_s0ix_active(struct amdgpu_device *adev) { retu static inline bool amdgpu_acpi_is_s3_active(struct amdgpu_device *adev) { return false; } #endif +#if defined(CONFIG_DRM_AMD_ISP) +int amdgpu_acpi_get_isp4_dev_hid(u8 (*hid)[ACPI_ID_LEN]); +#endif + void amdgpu_register_gpu_instance(struct amdgpu_device *adev); void amdgpu_unregister_gpu_instance(struct amdgpu_device *adev); diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_acpi.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_acpi.c index 707e131f89d2..f5466c592d94 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_acpi.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_acpi.c @@ -1532,5 +1532,35 @@ bool amdgpu_acpi_is_s0ix_active(struct amdgpu_device *adev) return true; #endif /* CONFIG_AMD_PMC */ } - #endif /* CONFIG_SUSPEND */ + +#if IS_ENABLED(CONFIG_DRM_AMD_ISP) +static const struct acpi_device_id isp_sensor_ids[] = { + { "OMNI5C10" }, + { } +}; + +static int isp_match_acpi_device_ids(struct device *dev, const void *data) +{ + return acpi_match_device(data, dev) ? 1 : 0; +} + +int amdgpu_acpi_get_isp4_dev_hid(u8 (*hid)[ACPI_ID_LEN]) +{ + struct device *pdev __free(put_device) = NULL; + struct acpi_device *acpi_pdev; + + pdev = bus_find_device(&platform_bus_type, NULL, isp_sensor_ids, + isp_match_acpi_device_ids); + if (!pdev) + return -EINVAL; + + acpi_pdev = ACPI_COMPANION(pdev); + if (!acpi_pdev) + return -ENODEV; + + strscpy(*hid, acpi_device_hid(acpi_pdev)); + + return 0; +} +#endif /* CONFIG_DRM_AMD_ISP */ diff --git a/drivers/gpu/drm/amd/amdgpu/isp_v4_1_1.c b/drivers/gpu/drm/amd/amdgpu/isp_v4_1_1.c index 69dd92f6e86d..574880d67009 100644 --- a/drivers/gpu/drm/amd/amdgpu/isp_v4_1_1.c +++ b/drivers/gpu/drm/amd/amdgpu/isp_v4_1_1.c @@ -25,6 +25,7 @@ * */ +#include #include "amdgpu.h" #include "isp_v4_1_1.h" @@ -39,15 +40,45 @@ static const unsigned int isp_4_1_1_int_srcid[MAX_ISP411_INT_SRC] = { ISP_4_1__SRCID__ISP_RINGBUFFER_WPT16 }; +static struct gpiod_lookup_table isp_gpio_table = { + .dev_id = "amd_isp_capture", + .table = { + GPIO_LOOKUP("AMDI0030:00", 85, "enable_isp", GPIO_ACTIVE_HIGH), + { } + }, +}; + +static struct gpiod_lookup_table isp_sensor_gpio_table = { + .dev_id = "i2c-ov05c10", + .table = { + GPIO_LOOKUP("amdisp-pinctrl", 0, "enable", GPIO_ACTIVE_HIGH), + { } + }, +}; + static int isp_v4_1_1_hw_init(struct amdgpu_isp *isp) { struct amdgpu_device *adev = isp->adev; int idx, int_idx, num_res, r; + u8 isp_dev_hid[ACPI_ID_LEN]; u64 isp_base; if (adev->rmmio_size == 0 || adev->rmmio_size < 0x5289) return -EINVAL; + r = amdgpu_acpi_get_isp4_dev_hid(&isp_dev_hid); + if (r) { + drm_dbg(&adev->ddev, "Invalid isp platform detected (%d)", r); + /* allow GPU init to progress */ + return 0; + } + + /* add GPIO resources required for OMNI5C10 sensor */ + if (!strcmp("OMNI5C10", isp_dev_hid)) { + gpiod_add_lookup_table(&isp_gpio_table); + gpiod_add_lookup_table(&isp_sensor_gpio_table); + } + isp_base = adev->rmmio_base; isp->isp_cell = kcalloc(3, sizeof(struct mfd_cell), GFP_KERNEL); -- cgit v1.2.3 From d78eb800f8f5169db89a28380631aefc224a76bb Mon Sep 17 00:00:00 2001 From: Aurabindo Pillai Date: Wed, 21 May 2025 15:59:56 -0400 Subject: drm/amd/display: Add some missing register headers for DCN401 Add some HDCP related register headers for future use. Signed-off-by: Aurabindo Pillai Reviewed-by: Leo Li Signed-off-by: Alex Deucher --- .../amd/include/asic_reg/dcn/dcn_4_1_0_offset.h | 26 ++++++++++++++++++++++ .../amd/include/asic_reg/dcn/dcn_4_1_0_sh_mask.h | 16 +++++++++++++ 2 files changed, 42 insertions(+) (limited to 'drivers/gpu') diff --git a/drivers/gpu/drm/amd/include/asic_reg/dcn/dcn_4_1_0_offset.h b/drivers/gpu/drm/amd/include/asic_reg/dcn/dcn_4_1_0_offset.h index 15e5a65cf492..70ee6be94a9b 100644 --- a/drivers/gpu/drm/amd/include/asic_reg/dcn/dcn_4_1_0_offset.h +++ b/drivers/gpu/drm/amd/include/asic_reg/dcn/dcn_4_1_0_offset.h @@ -9776,6 +9776,14 @@ #define regDIG0_DIG_BE_CNTL_BASE_IDX 2 #define regDIG0_DIG_BE_EN_CNTL 0x20bd #define regDIG0_DIG_BE_EN_CNTL_BASE_IDX 2 +#define regDIG0_HDCP_INT_CONTROL 0x20c0 +#define regDIG0_HDCP_INT_CONTROL_BASE_IDX 2 +#define regDIG0_HDCP_LINK0_STATUS 0x20c1 +#define regDIG0_HDCP_LINK0_STATUS_BASE_IDX 2 +#define regDIG0_HDCP_I2C_CONTROL_0 0x20c2 +#define regDIG0_HDCP_I2C_CONTROL_0_BASE_IDX 2 +#define regDIG0_HDCP_I2C_CONTROL_1 0x20c3 +#define regDIG0_HDCP_I2C_CONTROL_1_BASE_IDX 2 #define regDIG0_TMDS_CNTL 0x20e4 #define regDIG0_TMDS_CNTL_BASE_IDX 2 #define regDIG0_TMDS_CONTROL_CHAR 0x20e5 @@ -10081,6 +10089,12 @@ #define regDIG1_DIG_BE_CNTL_BASE_IDX 2 #define regDIG1_DIG_BE_EN_CNTL 0x21e1 #define regDIG1_DIG_BE_EN_CNTL_BASE_IDX 2 +#define regDIG1_HDCP_INT_CONTROL 0x21e4 +#define regDIG1_HDCP_INT_CONTROL_BASE_IDX 2 +#define regDIG1_HDCP_I2C_CONTROL_0 0x21e6 +#define regDIG1_HDCP_I2C_CONTROL_0_BASE_IDX 2 +#define regDIG1_HDCP_I2C_CONTROL_1 0x21e7 +#define regDIG1_HDCP_I2C_CONTROL_1_BASE_IDX 2 #define regDIG1_TMDS_CNTL 0x2208 #define regDIG1_TMDS_CNTL_BASE_IDX 2 #define regDIG1_TMDS_CONTROL_CHAR 0x2209 @@ -10386,6 +10400,12 @@ #define regDIG2_DIG_BE_CNTL_BASE_IDX 2 #define regDIG2_DIG_BE_EN_CNTL 0x2305 #define regDIG2_DIG_BE_EN_CNTL_BASE_IDX 2 +#define regDIG2_HDCP_INT_CONTROL 0x2308 +#define regDIG2_HDCP_INT_CONTROL_BASE_IDX 2 +#define regDIG2_HDCP_I2C_CONTROL_0 0x230a +#define regDIG2_HDCP_I2C_CONTROL_0_BASE_IDX 2 +#define regDIG2_HDCP_I2C_CONTROL_1 0x230b +#define regDIG2_HDCP_I2C_CONTROL_1_BASE_IDX 2 #define regDIG2_TMDS_CNTL 0x232c #define regDIG2_TMDS_CNTL_BASE_IDX 2 #define regDIG2_TMDS_CONTROL_CHAR 0x232d @@ -10691,6 +10711,12 @@ #define regDIG3_DIG_BE_CNTL_BASE_IDX 2 #define regDIG3_DIG_BE_EN_CNTL 0x2429 #define regDIG3_DIG_BE_EN_CNTL_BASE_IDX 2 +#define regDIG3_HDCP_INT_CONTROL 0x242c +#define regDIG3_HDCP_INT_CONTROL_BASE_IDX 2 +#define regDIG3_HDCP_I2C_CONTROL_0 0x242e +#define regDIG3_HDCP_I2C_CONTROL_0_BASE_IDX 2 +#define regDIG3_HDCP_I2C_CONTROL_1 0x242f +#define regDIG3_HDCP_I2C_CONTROL_1_BASE_IDX 2 #define regDIG3_TMDS_CNTL 0x2450 #define regDIG3_TMDS_CNTL_BASE_IDX 2 #define regDIG3_TMDS_CONTROL_CHAR 0x2451 diff --git a/drivers/gpu/drm/amd/include/asic_reg/dcn/dcn_4_1_0_sh_mask.h b/drivers/gpu/drm/amd/include/asic_reg/dcn/dcn_4_1_0_sh_mask.h index 5d9d5fea6e06..e3d841b2e9af 100644 --- a/drivers/gpu/drm/amd/include/asic_reg/dcn/dcn_4_1_0_sh_mask.h +++ b/drivers/gpu/drm/amd/include/asic_reg/dcn/dcn_4_1_0_sh_mask.h @@ -2847,6 +2847,14 @@ #define HDCP_INTERRUPT_DEST__DOUT_IHC_HDCP0_AUTH_FAIL_INTERRUPT_DEST__SHIFT 0x1 #define HDCP_INTERRUPT_DEST__DOUT_IHC_HDCP0_I2C_XFER_REQ_INTERRUPT_DEST__SHIFT 0x2 #define HDCP_INTERRUPT_DEST__DOUT_IHC_HDCP0_I2C_XFER_DONE_INTERRUPT_DEST__SHIFT 0x3 +#define HDCP_INTERRUPT_DEST__DOUT_IHC_HDCP1_AUTH_SUCCESS_INTERRUPT_DEST__SHIFT 0x4 +#define HDCP_INTERRUPT_DEST__DOUT_IHC_HDCP1_AUTH_FAIL_INTERRUPT_DEST__SHIFT 0x5 +#define HDCP_INTERRUPT_DEST__DOUT_IHC_HDCP1_I2C_XFER_REQ_INTERRUPT_DEST__SHIFT 0x6 +#define HDCP_INTERRUPT_DEST__DOUT_IHC_HDCP1_I2C_XFER_DONE_INTERRUPT_DEST__SHIFT 0x7 +#define HDCP_INTERRUPT_DEST__DOUT_IHC_HDCP2_AUTH_SUCCESS_INTERRUPT_DEST__SHIFT 0x8 +#define HDCP_INTERRUPT_DEST__DOUT_IHC_HDCP2_AUTH_FAIL_INTERRUPT_DEST__SHIFT 0x9 +#define HDCP_INTERRUPT_DEST__DOUT_IHC_HDCP2_I2C_XFER_REQ_INTERRUPT_DEST__SHIFT 0xa +#define HDCP_INTERRUPT_DEST__DOUT_IHC_HDCP2_I2C_XFER_DONE_INTERRUPT_DEST__SHIFT 0xb #define HDCP_INTERRUPT_DEST__DOUT_IHC_HDCP3_AUTH_SUCCESS_INTERRUPT_DEST__SHIFT 0xc #define HDCP_INTERRUPT_DEST__DOUT_IHC_HDCP3_AUTH_FAIL_INTERRUPT_DEST__SHIFT 0xd #define HDCP_INTERRUPT_DEST__DOUT_IHC_HDCP3_I2C_XFER_REQ_INTERRUPT_DEST__SHIFT 0xe @@ -2871,6 +2879,14 @@ #define HDCP_INTERRUPT_DEST__DOUT_IHC_HDCP0_AUTH_FAIL_INTERRUPT_DEST_MASK 0x00000002L #define HDCP_INTERRUPT_DEST__DOUT_IHC_HDCP0_I2C_XFER_REQ_INTERRUPT_DEST_MASK 0x00000004L #define HDCP_INTERRUPT_DEST__DOUT_IHC_HDCP0_I2C_XFER_DONE_INTERRUPT_DEST_MASK 0x00000008L +#define HDCP_INTERRUPT_DEST__DOUT_IHC_HDCP1_AUTH_SUCCESS_INTERRUPT_DEST_MASK 0x00000010L +#define HDCP_INTERRUPT_DEST__DOUT_IHC_HDCP1_AUTH_FAIL_INTERRUPT_DEST_MASK 0x00000020L +#define HDCP_INTERRUPT_DEST__DOUT_IHC_HDCP1_I2C_XFER_REQ_INTERRUPT_DEST_MASK 0x00000040L +#define HDCP_INTERRUPT_DEST__DOUT_IHC_HDCP1_I2C_XFER_DONE_INTERRUPT_DEST_MASK 0x00000080L +#define HDCP_INTERRUPT_DEST__DOUT_IHC_HDCP2_AUTH_SUCCESS_INTERRUPT_DEST_MASK 0x00000100L +#define HDCP_INTERRUPT_DEST__DOUT_IHC_HDCP2_AUTH_FAIL_INTERRUPT_DEST_MASK 0x00000200L +#define HDCP_INTERRUPT_DEST__DOUT_IHC_HDCP2_I2C_XFER_REQ_INTERRUPT_DEST_MASK 0x00000400L +#define HDCP_INTERRUPT_DEST__DOUT_IHC_HDCP2_I2C_XFER_DONE_INTERRUPT_DEST_MASK 0x00000800L #define HDCP_INTERRUPT_DEST__DOUT_IHC_HDCP3_AUTH_SUCCESS_INTERRUPT_DEST_MASK 0x00001000L #define HDCP_INTERRUPT_DEST__DOUT_IHC_HDCP3_AUTH_FAIL_INTERRUPT_DEST_MASK 0x00002000L #define HDCP_INTERRUPT_DEST__DOUT_IHC_HDCP3_I2C_XFER_REQ_INTERRUPT_DEST_MASK 0x00004000L -- cgit v1.2.3 From 1b824eef269db44d068bbc0de74c94a8e8f9ce02 Mon Sep 17 00:00:00 2001 From: Aurabindo Pillai Date: Wed, 21 May 2025 16:05:39 -0400 Subject: Revert "drm/amd/display: more liberal vmin/vmax update for freesync" This reverts commit cfb2d41831ee5647a4ae0ea7c24971a92d5dfa0d since it causes regressions on certain configs. Revert until the issue can be isolated and debugged. Closes: https://gitlab.freedesktop.org/drm/amd/-/issues/4238 Signed-off-by: Aurabindo Pillai Acked-by: Alex Deucher Signed-off-by: Alex Deucher Cc: stable@vger.kernel.org --- drivers/gpu/drm/amd/display/amdgpu_dm/amdgpu_dm.c | 16 +++++----------- 1 file changed, 5 insertions(+), 11 deletions(-) (limited to 'drivers/gpu') 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 90889f6867aa..9f2e26336ccc 100644 --- a/drivers/gpu/drm/amd/display/amdgpu_dm/amdgpu_dm.c +++ b/drivers/gpu/drm/amd/display/amdgpu_dm/amdgpu_dm.c @@ -676,21 +676,15 @@ static void dm_crtc_high_irq(void *interrupt_params) spin_lock_irqsave(&adev_to_drm(adev)->event_lock, flags); if (acrtc->dm_irq_params.stream && - acrtc->dm_irq_params.vrr_params.supported) { - bool replay_en = acrtc->dm_irq_params.stream->link->replay_settings.replay_feature_enabled; - bool psr_en = acrtc->dm_irq_params.stream->link->psr_settings.psr_feature_enabled; - bool fs_active_var_en = acrtc->dm_irq_params.freesync_config.state == VRR_STATE_ACTIVE_VARIABLE; - + acrtc->dm_irq_params.vrr_params.supported && + acrtc->dm_irq_params.freesync_config.state == + VRR_STATE_ACTIVE_VARIABLE) { mod_freesync_handle_v_update(adev->dm.freesync_module, acrtc->dm_irq_params.stream, &acrtc->dm_irq_params.vrr_params); - /* update vmin_vmax only if freesync is enabled, or only if PSR and REPLAY are disabled */ - if (fs_active_var_en || (!fs_active_var_en && !replay_en && !psr_en)) { - dc_stream_adjust_vmin_vmax(adev->dm.dc, - acrtc->dm_irq_params.stream, - &acrtc->dm_irq_params.vrr_params.adjust); - } + dc_stream_adjust_vmin_vmax(adev->dm.dc, acrtc->dm_irq_params.stream, + &acrtc->dm_irq_params.vrr_params.adjust); } /* -- cgit v1.2.3 From 040585df957d45ebec0297bad525f39471229987 Mon Sep 17 00:00:00 2001 From: Aurabindo Pillai Date: Thu, 22 May 2025 10:34:50 -0400 Subject: drm/amd/display: Reuse Subvp debug option for FAMS FAMS is the successor to SubVP starting with DCN4x. Reuse the same debug option to disable FAMS for debugging purposes. Signed-off-by: Aurabindo Pillai Reviewed-by: Harry Wentland Signed-off-by: Alex Deucher --- drivers/gpu/drm/amd/display/amdgpu_dm/amdgpu_dm.c | 4 +++- drivers/gpu/drm/amd/include/amd_shared.h | 5 +++-- 2 files changed, 6 insertions(+), 3 deletions(-) (limited to 'drivers/gpu') 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 9f2e26336ccc..1797fa85fac6 100644 --- a/drivers/gpu/drm/amd/display/amdgpu_dm/amdgpu_dm.c +++ b/drivers/gpu/drm/amd/display/amdgpu_dm/amdgpu_dm.c @@ -2000,8 +2000,10 @@ static int amdgpu_dm_init(struct amdgpu_device *adev) if (amdgpu_dc_debug_mask & DC_FORCE_SUBVP_MCLK_SWITCH) adev->dm.dc->debug.force_subvp_mclk_switch = true; - if (amdgpu_dc_debug_mask & DC_DISABLE_SUBVP) + if (amdgpu_dc_debug_mask & DC_DISABLE_SUBVP_FAMS) { adev->dm.dc->debug.force_disable_subvp = true; + adev->dm.dc->debug.fams2_config.bits.enable = false; + } if (amdgpu_dc_debug_mask & DC_ENABLE_DML2) { adev->dm.dc->debug.using_dml2 = true; diff --git a/drivers/gpu/drm/amd/include/amd_shared.h b/drivers/gpu/drm/amd/include/amd_shared.h index 31de36c9156f..11374a2cbab8 100644 --- a/drivers/gpu/drm/amd/include/amd_shared.h +++ b/drivers/gpu/drm/amd/include/amd_shared.h @@ -351,9 +351,10 @@ enum DC_DEBUG_MASK { DC_DISABLE_HDMI_CEC = 0x10000, /** - * @DC_DISABLE_SUBVP: If set, disable DCN Sub-Viewport feature in amdgpu driver. + * @DC_DISABLE_SUBVP_FAMS: If set, disable DCN Sub-Viewport & Firmware Assisted + * Memory Clock Switching (FAMS) feature in amdgpu driver. */ - DC_DISABLE_SUBVP = 0x20000, + DC_DISABLE_SUBVP_FAMS = 0x20000, /** * @DC_DISABLE_CUSTOM_BRIGHTNESS_CURVE: If set, disable support for custom brightness curves */ -- cgit v1.2.3 From 0ef2803173f180fe31b5083b1dcf0f8fa6e5608f Mon Sep 17 00:00:00 2001 From: "David (Ming Qiang) Wu" Date: Wed, 14 May 2025 18:50:15 -0400 Subject: drm/amdgpu/vcn1: read back register after written V3: drop changes where readbacks have implemented. This patch set is to add readbacks only. V2: use common register UVD_STATUS for readback (standard PCI MMIO behavior, i.e. readback post all writes to let the writes hit the hardware) add readback in ..._stop() for more coverage. Similar to the changes made for VCN v4.0.5 where readback to post the writes to avoid race with the doorbell, the addition of register readback support in other VCN versions is intended to prevent potential race conditions, even though such issues have not been observed yet. This change ensures consistency across different VCN variants and helps avoid similar issues. The overhead introduced is negligible. Reviewed-by: Ruijing Dong Signed-off-by: David (Ming Qiang) Wu Signed-off-by: Alex Deucher --- drivers/gpu/drm/amd/amdgpu/vcn_v1_0.c | 21 +++++++++++++++++++++ 1 file changed, 21 insertions(+) (limited to 'drivers/gpu') diff --git a/drivers/gpu/drm/amd/amdgpu/vcn_v1_0.c b/drivers/gpu/drm/amd/amdgpu/vcn_v1_0.c index 21b57c29bf7d..c74947705d77 100644 --- a/drivers/gpu/drm/amd/amdgpu/vcn_v1_0.c +++ b/drivers/gpu/drm/amd/amdgpu/vcn_v1_0.c @@ -1009,6 +1009,11 @@ static int vcn_v1_0_start_spg_mode(struct amdgpu_vcn_inst *vinst) jpeg_v1_0_start(adev, 0); + /* Keeping one read-back to ensure all register writes are done, + * otherwise it may introduce race conditions. + */ + RREG32_SOC15(UVD, 0, mmUVD_STATUS); + return 0; } @@ -1154,6 +1159,11 @@ static int vcn_v1_0_start_dpg_mode(struct amdgpu_vcn_inst *vinst) jpeg_v1_0_start(adev, 1); + /* Keeping one read-back to ensure all register writes are done, + * otherwise it may introduce race conditions. + */ + RREG32_SOC15(UVD, 0, mmUVD_STATUS); + return 0; } @@ -1216,6 +1226,12 @@ static int vcn_v1_0_stop_spg_mode(struct amdgpu_vcn_inst *vinst) vcn_v1_0_enable_clock_gating(vinst); vcn_1_0_enable_static_power_gating(vinst); + + /* Keeping one read-back to ensure all register writes are done, + * otherwise it may introduce race conditions. + */ + RREG32_SOC15(UVD, 0, mmUVD_STATUS); + return 0; } @@ -1250,6 +1266,11 @@ static int vcn_v1_0_stop_dpg_mode(struct amdgpu_vcn_inst *vinst) WREG32_P(SOC15_REG_OFFSET(UVD, 0, mmUVD_POWER_STATUS), 0, ~UVD_POWER_STATUS__UVD_PG_MODE_MASK); + /* Keeping one read-back to ensure all register writes are done, + * otherwise it may introduce race conditions. + */ + RREG32_SOC15(UVD, 0, mmUVD_STATUS); + return 0; } -- cgit v1.2.3 From 988b1d2164a1cbe56616a1ded0c877a3ae753558 Mon Sep 17 00:00:00 2001 From: Fangzhi Zuo Date: Tue, 20 May 2025 15:30:25 -0400 Subject: Revert "drm/amd/display: pause the workload setting in dm" This reverts commit 50f29ead1f1ba48983b6c5e3813b15e497714f55. Reason for revert: cause corruption on Dell U3224KB DP2 display. Signed-off-by: Fangzhi Zuo Reviewed-by: Kenneth Feng Acked-by: Alex Deucher Signed-off-by: Alex Deucher --- drivers/gpu/drm/amd/display/amdgpu_dm/amdgpu_dm_crtc.c | 11 +---------- 1 file changed, 1 insertion(+), 10 deletions(-) (limited to 'drivers/gpu') diff --git a/drivers/gpu/drm/amd/display/amdgpu_dm/amdgpu_dm_crtc.c b/drivers/gpu/drm/amd/display/amdgpu_dm/amdgpu_dm_crtc.c index e8bdd7f0c460..87058271b00c 100644 --- a/drivers/gpu/drm/amd/display/amdgpu_dm/amdgpu_dm_crtc.c +++ b/drivers/gpu/drm/amd/display/amdgpu_dm/amdgpu_dm_crtc.c @@ -246,8 +246,6 @@ static void amdgpu_dm_crtc_vblank_control_worker(struct work_struct *work) struct vblank_control_work *vblank_work = container_of(work, struct vblank_control_work, work); struct amdgpu_display_manager *dm = vblank_work->dm; - struct amdgpu_device *adev = drm_to_adev(dm->ddev); - int r; mutex_lock(&dm->dc_lock); @@ -275,15 +273,8 @@ static void amdgpu_dm_crtc_vblank_control_worker(struct work_struct *work) vblank_work->acrtc->dm_irq_params.allow_sr_entry); } - if (dm->active_vblank_irq_count == 0) { - r = amdgpu_dpm_pause_power_profile(adev, true); - if (r) - dev_warn(adev->dev, "failed to set default power profile mode\n"); + if (dm->active_vblank_irq_count == 0) dc_allow_idle_optimizations(dm->dc, true); - r = amdgpu_dpm_pause_power_profile(adev, false); - if (r) - dev_warn(adev->dev, "failed to restore the power profile mode\n"); - } mutex_unlock(&dm->dc_lock); -- cgit v1.2.3 From 8c5ed7f5abe2cae3b0bf33d4f51f6fb03aadf3ff Mon Sep 17 00:00:00 2001 From: "David (Ming Qiang) Wu" Date: Wed, 14 May 2025 18:52:13 -0400 Subject: drm/amdgpu/vcn2: read back register after written The addition of register read-back in VCN v2.0 is intended to prevent potential race conditions. Reviewed-by: Ruijing Dong Signed-off-by: David (Ming Qiang) Wu Signed-off-by: Alex Deucher --- drivers/gpu/drm/amd/amdgpu/vcn_v2_0.c | 21 +++++++++++++++++++++ 1 file changed, 21 insertions(+) (limited to 'drivers/gpu') diff --git a/drivers/gpu/drm/amd/amdgpu/vcn_v2_0.c b/drivers/gpu/drm/amd/amdgpu/vcn_v2_0.c index b8d835c9e17e..148b651be7ca 100644 --- a/drivers/gpu/drm/amd/amdgpu/vcn_v2_0.c +++ b/drivers/gpu/drm/amd/amdgpu/vcn_v2_0.c @@ -978,6 +978,12 @@ static int vcn_v2_0_start_dpg_mode(struct amdgpu_vcn_inst *vinst, bool indirect) /* Unstall DPG */ WREG32_P(SOC15_REG_OFFSET(UVD, 0, mmUVD_POWER_STATUS), 0, ~UVD_POWER_STATUS__STALL_DPG_POWER_UP_MASK); + + /* Keeping one read-back to ensure all register writes are done, + * otherwise it may introduce race conditions. + */ + RREG32_SOC15(UVD, 0, mmUVD_STATUS); + return 0; } @@ -1152,6 +1158,11 @@ static int vcn_v2_0_start(struct amdgpu_vcn_inst *vinst) WREG32_SOC15(UVD, 0, mmUVD_RB_SIZE2, ring->ring_size / 4); fw_shared->multi_queue.encode_lowlatency_queue_mode &= ~FW_QUEUE_RING_RESET; + /* Keeping one read-back to ensure all register writes are done, + * otherwise it may introduce race conditions. + */ + RREG32_SOC15(UVD, 0, mmUVD_STATUS); + return 0; } @@ -1183,6 +1194,11 @@ static int vcn_v2_0_stop_dpg_mode(struct amdgpu_vcn_inst *vinst) WREG32_P(SOC15_REG_OFFSET(UVD, 0, mmUVD_POWER_STATUS), 0, ~UVD_POWER_STATUS__UVD_PG_MODE_MASK); + /* Keeping one read-back to ensure all register writes are done, + * otherwise it may introduce race conditions. + */ + RREG32_SOC15(UVD, 0, mmUVD_STATUS); + return 0; } @@ -1248,6 +1264,11 @@ static int vcn_v2_0_stop(struct amdgpu_vcn_inst *vinst) vcn_v2_0_enable_clock_gating(vinst); vcn_v2_0_enable_static_power_gating(vinst); + /* Keeping one read-back to ensure all register writes are done, + * otherwise it may introduce race conditions. + */ + RREG32_SOC15(VCN, 0, mmUVD_STATUS); + power_off: if (adev->pm.dpm_enabled) amdgpu_dpm_enable_vcn(adev, false, 0); -- cgit v1.2.3 From d9e688b9148bb23629d32017344888dd67ec2ab1 Mon Sep 17 00:00:00 2001 From: "David (Ming Qiang) Wu" Date: Wed, 14 May 2025 18:54:01 -0400 Subject: drm/amdgpu/vcn2.5: read back register after written The addition of register read-back in VCN v2.5 is intended to prevent potential race conditions. Reviewed-by: Ruijing Dong Signed-off-by: David (Ming Qiang) Wu Signed-off-by: Alex Deucher --- drivers/gpu/drm/amd/amdgpu/vcn_v2_5.c | 19 +++++++++++++++++++ 1 file changed, 19 insertions(+) (limited to 'drivers/gpu') diff --git a/drivers/gpu/drm/amd/amdgpu/vcn_v2_5.c b/drivers/gpu/drm/amd/amdgpu/vcn_v2_5.c index 3eec1b8feaee..58b527a6b795 100644 --- a/drivers/gpu/drm/amd/amdgpu/vcn_v2_5.c +++ b/drivers/gpu/drm/amd/amdgpu/vcn_v2_5.c @@ -1158,6 +1158,11 @@ static int vcn_v2_5_start_dpg_mode(struct amdgpu_vcn_inst *vinst, bool indirect) WREG32_P(SOC15_REG_OFFSET(VCN, inst_idx, mmUVD_POWER_STATUS), 0, ~UVD_POWER_STATUS__STALL_DPG_POWER_UP_MASK); + /* Keeping one read-back to ensure all register writes are done, + * otherwise it may introduce race conditions. + */ + RREG32_SOC15(VCN, inst_idx, mmUVD_STATUS); + return 0; } @@ -1343,6 +1348,11 @@ static int vcn_v2_5_start(struct amdgpu_vcn_inst *vinst) WREG32_SOC15(VCN, i, mmUVD_RB_SIZE2, ring->ring_size / 4); fw_shared->multi_queue.encode_lowlatency_queue_mode &= ~FW_QUEUE_RING_RESET; + /* Keeping one read-back to ensure all register writes are done, + * otherwise it may introduce race conditions. + */ + RREG32_SOC15(VCN, i, mmUVD_STATUS); + return 0; } @@ -1569,6 +1579,11 @@ static int vcn_v2_5_stop_dpg_mode(struct amdgpu_vcn_inst *vinst) WREG32_P(SOC15_REG_OFFSET(VCN, inst_idx, mmUVD_POWER_STATUS), 0, ~UVD_POWER_STATUS__UVD_PG_MODE_MASK); + /* Keeping one read-back to ensure all register writes are done, + * otherwise it may introduce race conditions. + */ + RREG32_SOC15(VCN, inst_idx, mmUVD_STATUS); + return 0; } @@ -1635,6 +1650,10 @@ static int vcn_v2_5_stop(struct amdgpu_vcn_inst *vinst) UVD_POWER_STATUS__UVD_POWER_STATUS_MASK, ~UVD_POWER_STATUS__UVD_POWER_STATUS_MASK); + /* Keeping one read-back to ensure all register writes are done, + * otherwise it may introduce race conditions. + */ + RREG32_SOC15(VCN, i, mmUVD_STATUS); done: if (adev->pm.dpm_enabled) amdgpu_dpm_enable_vcn(adev, false, i); -- cgit v1.2.3 From b7a4842a917e3a251b5a6aa1a21a5daf6d396ef3 Mon Sep 17 00:00:00 2001 From: "David (Ming Qiang) Wu" Date: Wed, 14 May 2025 18:54:39 -0400 Subject: drm/amdgpu/vcn3: read back register after written The addition of register read-back in VCN v3.0 is intended to prevent potential race conditions. Reviewed-by: Ruijing Dong Signed-off-by: David (Ming Qiang) Wu Signed-off-by: Alex Deucher --- drivers/gpu/drm/amd/amdgpu/vcn_v3_0.c | 20 ++++++++++++++++++++ 1 file changed, 20 insertions(+) (limited to 'drivers/gpu') diff --git a/drivers/gpu/drm/amd/amdgpu/vcn_v3_0.c b/drivers/gpu/drm/amd/amdgpu/vcn_v3_0.c index 0b19f0ab4480..9fb0d5380589 100644 --- a/drivers/gpu/drm/amd/amdgpu/vcn_v3_0.c +++ b/drivers/gpu/drm/amd/amdgpu/vcn_v3_0.c @@ -1173,6 +1173,11 @@ static int vcn_v3_0_start_dpg_mode(struct amdgpu_vcn_inst *vinst, bool indirect) WREG32_P(SOC15_REG_OFFSET(VCN, inst_idx, mmUVD_POWER_STATUS), 0, ~UVD_POWER_STATUS__STALL_DPG_POWER_UP_MASK); + /* Keeping one read-back to ensure all register writes are done, + * otherwise it may introduce race conditions. + */ + RREG32_SOC15(VCN, inst_idx, mmUVD_STATUS); + return 0; } @@ -1360,6 +1365,11 @@ static int vcn_v3_0_start(struct amdgpu_vcn_inst *vinst) fw_shared->multi_queue.encode_lowlatency_queue_mode &= cpu_to_le32(~FW_QUEUE_RING_RESET); } + /* Keeping one read-back to ensure all register writes are done, + * otherwise it may introduce race conditions. + */ + RREG32_SOC15(VCN, i, mmUVD_STATUS); + return 0; } @@ -1602,6 +1612,11 @@ static int vcn_v3_0_stop_dpg_mode(struct amdgpu_vcn_inst *vinst) WREG32_P(SOC15_REG_OFFSET(VCN, inst_idx, mmUVD_POWER_STATUS), 0, ~UVD_POWER_STATUS__UVD_PG_MODE_MASK); + /* Keeping one read-back to ensure all register writes are done, + * otherwise it may introduce race conditions. + */ + RREG32_SOC15(VCN, inst_idx, mmUVD_STATUS); + return 0; } @@ -1674,6 +1689,11 @@ static int vcn_v3_0_stop(struct amdgpu_vcn_inst *vinst) /* enable VCN power gating */ vcn_v3_0_enable_static_power_gating(vinst); + /* Keeping one read-back to ensure all register writes are done, + * otherwise it may introduce race conditions. + */ + RREG32_SOC15(VCN, i, mmUVD_STATUS); + done: if (adev->pm.dpm_enabled) amdgpu_dpm_enable_vcn(adev, false, i); -- cgit v1.2.3 From a3810a5e37c58329aa2c7992f3172a423f4ae194 Mon Sep 17 00:00:00 2001 From: "David (Ming Qiang) Wu" Date: Wed, 14 May 2025 18:55:27 -0400 Subject: drm/amdgpu/vcn4: read back register after written The addition of register read-back in VCN v4.0.0 is intended to prevent potential race conditions. Reviewed-by: Ruijing Dong Signed-off-by: David (Ming Qiang) Wu Signed-off-by: Alex Deucher --- drivers/gpu/drm/amd/amdgpu/vcn_v4_0.c | 20 ++++++++++++++++++++ 1 file changed, 20 insertions(+) (limited to 'drivers/gpu') diff --git a/drivers/gpu/drm/amd/amdgpu/vcn_v4_0.c b/drivers/gpu/drm/amd/amdgpu/vcn_v4_0.c index 8fff470bce87..b5071f77f78d 100644 --- a/drivers/gpu/drm/amd/amdgpu/vcn_v4_0.c +++ b/drivers/gpu/drm/amd/amdgpu/vcn_v4_0.c @@ -1122,6 +1122,11 @@ static int vcn_v4_0_start_dpg_mode(struct amdgpu_vcn_inst *vinst, bool indirect) ring->doorbell_index << VCN_RB1_DB_CTRL__OFFSET__SHIFT | VCN_RB1_DB_CTRL__EN_MASK); + /* Keeping one read-back to ensure all register writes are done, + * otherwise it may introduce race conditions. + */ + RREG32_SOC15(VCN, inst_idx, regUVD_STATUS); + return 0; } @@ -1303,6 +1308,11 @@ static int vcn_v4_0_start(struct amdgpu_vcn_inst *vinst) WREG32_SOC15(VCN, i, regVCN_RB_ENABLE, tmp); fw_shared->sq.queue_mode &= ~(FW_QUEUE_RING_RESET | FW_QUEUE_DPG_HOLD_OFF); + /* Keeping one read-back to ensure all register writes are done, + * otherwise it may introduce race conditions. + */ + RREG32_SOC15(VCN, i, regUVD_STATUS); + return 0; } @@ -1583,6 +1593,11 @@ static void vcn_v4_0_stop_dpg_mode(struct amdgpu_vcn_inst *vinst) /* disable dynamic power gating mode */ WREG32_P(SOC15_REG_OFFSET(VCN, inst_idx, regUVD_POWER_STATUS), 0, ~UVD_POWER_STATUS__UVD_PG_MODE_MASK); + + /* Keeping one read-back to ensure all register writes are done, + * otherwise it may introduce race conditions. + */ + RREG32_SOC15(VCN, inst_idx, regUVD_STATUS); } /** @@ -1666,6 +1681,11 @@ static int vcn_v4_0_stop(struct amdgpu_vcn_inst *vinst) /* enable VCN power gating */ vcn_v4_0_enable_static_power_gating(vinst); + /* Keeping one read-back to ensure all register writes are done, + * otherwise it may introduce race conditions. + */ + RREG32_SOC15(VCN, i, regUVD_STATUS); + done: if (adev->pm.dpm_enabled) amdgpu_dpm_enable_vcn(adev, false, i); -- cgit v1.2.3 From 5b4c6413c89613cb46669b576d83658fe6e734da Mon Sep 17 00:00:00 2001 From: "David (Ming Qiang) Wu" Date: Wed, 14 May 2025 18:55:59 -0400 Subject: drm/amdgpu/vcn4.0.3: read back register after written The addition of register read-back in VCN v4.0.3 is intended to prevent potential race conditions. Reviewed-by: Ruijing Dong Signed-off-by: David (Ming Qiang) Wu Signed-off-by: Alex Deucher --- drivers/gpu/drm/amd/amdgpu/vcn_v4_0_3.c | 16 ++++++++++++++++ 1 file changed, 16 insertions(+) (limited to 'drivers/gpu') diff --git a/drivers/gpu/drm/amd/amdgpu/vcn_v4_0_3.c b/drivers/gpu/drm/amd/amdgpu/vcn_v4_0_3.c index 764b3ff09f1e..5a33140f5723 100644 --- a/drivers/gpu/drm/amd/amdgpu/vcn_v4_0_3.c +++ b/drivers/gpu/drm/amd/amdgpu/vcn_v4_0_3.c @@ -977,6 +977,11 @@ static int vcn_v4_0_3_start_dpg_mode(struct amdgpu_vcn_inst *vinst, /*resetting done, fw can check RB ring */ fw_shared->sq.queue_mode &= cpu_to_le32(~FW_QUEUE_RING_RESET); + /* Keeping one read-back to ensure all register writes are done, + * otherwise it may introduce race conditions. + */ + RREG32_SOC15(VCN, vcn_inst, regUVD_STATUS); + return 0; } @@ -1370,6 +1375,12 @@ static int vcn_v4_0_3_stop_dpg_mode(struct amdgpu_vcn_inst *vinst) /* disable dynamic power gating mode */ WREG32_P(SOC15_REG_OFFSET(VCN, vcn_inst, regUVD_POWER_STATUS), 0, ~UVD_POWER_STATUS__UVD_PG_MODE_MASK); + + /* Keeping one read-back to ensure all register writes are done, + * otherwise it may introduce race conditions. + */ + RREG32_SOC15(VCN, vcn_inst, regUVD_STATUS); + return 0; } @@ -1453,6 +1464,11 @@ static int vcn_v4_0_3_stop(struct amdgpu_vcn_inst *vinst) /* apply HW clock gating */ vcn_v4_0_3_enable_clock_gating(vinst); + /* Keeping one read-back to ensure all register writes are done, + * otherwise it may introduce race conditions. + */ + RREG32_SOC15(VCN, vcn_inst, regUVD_STATUS); + Done: return 0; } -- cgit v1.2.3 From 4d4275a0387790fbe0a016c075dcbb39d58b63ab Mon Sep 17 00:00:00 2001 From: "David (Ming Qiang) Wu" Date: Wed, 14 May 2025 18:56:35 -0400 Subject: drm/amdgpu/vcn4.0.5: read back register after written The addition of register read-back in VCN v4.0.5 is intended to prevent potential race conditions. Reviewed-by: Ruijing Dong Signed-off-by: David (Ming Qiang) Wu Signed-off-by: Alex Deucher --- drivers/gpu/drm/amd/amdgpu/vcn_v4_0_5.c | 10 ++++++++++ 1 file changed, 10 insertions(+) (limited to 'drivers/gpu') diff --git a/drivers/gpu/drm/amd/amdgpu/vcn_v4_0_5.c b/drivers/gpu/drm/amd/amdgpu/vcn_v4_0_5.c index a09f9a2dd471..16ade84facc7 100644 --- a/drivers/gpu/drm/amd/amdgpu/vcn_v4_0_5.c +++ b/drivers/gpu/drm/amd/amdgpu/vcn_v4_0_5.c @@ -1254,6 +1254,11 @@ static void vcn_v4_0_5_stop_dpg_mode(struct amdgpu_vcn_inst *vinst) /* disable dynamic power gating mode */ WREG32_P(SOC15_REG_OFFSET(VCN, inst_idx, regUVD_POWER_STATUS), 0, ~UVD_POWER_STATUS__UVD_PG_MODE_MASK); + + /* Keeping one read-back to ensure all register writes are done, + * otherwise it may introduce race conditions. + */ + RREG32_SOC15(VCN, inst_idx, regUVD_STATUS); } /** @@ -1337,6 +1342,11 @@ static int vcn_v4_0_5_stop(struct amdgpu_vcn_inst *vinst) /* enable VCN power gating */ vcn_v4_0_5_enable_static_power_gating(vinst); + /* Keeping one read-back to ensure all register writes are done, + * otherwise it may introduce race conditions. + */ + RREG32_SOC15(VCN, i, regUVD_STATUS); + done: if (adev->pm.dpm_enabled) amdgpu_dpm_enable_vcn(adev, false, i); -- cgit v1.2.3 From a8bce9b7a2ccb5cbbe9967a37e617a9ef963e540 Mon Sep 17 00:00:00 2001 From: "David (Ming Qiang) Wu" Date: Wed, 14 May 2025 18:58:37 -0400 Subject: drm/amdgpu/vcn5: read back register after written The addition of register read-back in VCN v5.0.0 is intended to prevent potential race conditions. Reviewed-by: Ruijing Dong Signed-off-by: David (Ming Qiang) Wu Signed-off-by: Alex Deucher --- drivers/gpu/drm/amd/amdgpu/vcn_v5_0_0.c | 20 ++++++++++++++++++++ 1 file changed, 20 insertions(+) (limited to 'drivers/gpu') diff --git a/drivers/gpu/drm/amd/amdgpu/vcn_v5_0_0.c b/drivers/gpu/drm/amd/amdgpu/vcn_v5_0_0.c index 27dcc6f37a73..f8e3f0b882da 100644 --- a/drivers/gpu/drm/amd/amdgpu/vcn_v5_0_0.c +++ b/drivers/gpu/drm/amd/amdgpu/vcn_v5_0_0.c @@ -794,6 +794,11 @@ static int vcn_v5_0_0_start_dpg_mode(struct amdgpu_vcn_inst *vinst, ring->doorbell_index << VCN_RB1_DB_CTRL__OFFSET__SHIFT | VCN_RB1_DB_CTRL__EN_MASK); + /* Keeping one read-back to ensure all register writes are done, + * otherwise it may introduce race conditions. + */ + RREG32_SOC15(VCN, inst_idx, regUVD_STATUS); + return 0; } @@ -946,6 +951,11 @@ static int vcn_v5_0_0_start(struct amdgpu_vcn_inst *vinst) WREG32_SOC15(VCN, i, regVCN_RB_ENABLE, tmp); fw_shared->sq.queue_mode &= ~(FW_QUEUE_RING_RESET | FW_QUEUE_DPG_HOLD_OFF); + /* Keeping one read-back to ensure all register writes are done, + * otherwise it may introduce race conditions. + */ + RREG32_SOC15(VCN, i, regUVD_STATUS); + return 0; } @@ -977,6 +987,11 @@ static void vcn_v5_0_0_stop_dpg_mode(struct amdgpu_vcn_inst *vinst) WREG32_P(SOC15_REG_OFFSET(VCN, inst_idx, regUVD_POWER_STATUS), 0, ~UVD_POWER_STATUS__UVD_PG_MODE_MASK); + /* Keeping one read-back to ensure all register writes are done, + * otherwise it may introduce race conditions. + */ + RREG32_SOC15(VCN, inst_idx, regUVD_STATUS); + return; } @@ -1058,6 +1073,11 @@ static int vcn_v5_0_0_stop(struct amdgpu_vcn_inst *vinst) /* enable VCN power gating */ vcn_v5_0_0_enable_static_power_gating(vinst); + /* Keeping one read-back to ensure all register writes are done, + * otherwise it may introduce race conditions. + */ + RREG32_SOC15(VCN, i, regUVD_STATUS); + done: if (adev->pm.dpm_enabled) amdgpu_dpm_enable_vcn(adev, false, i); -- cgit v1.2.3 From bf394d28548c3c0a01e113fdef20ddb6cd2df106 Mon Sep 17 00:00:00 2001 From: "David (Ming Qiang) Wu" Date: Wed, 14 May 2025 18:59:11 -0400 Subject: drm/amdgpu/vcn5.0.1: read back register after written The addition of register read-back in VCN v5.0.1 is intended to prevent potential race conditions. Reviewed-by: Ruijing Dong Signed-off-by: David (Ming Qiang) Wu Signed-off-by: Alex Deucher --- drivers/gpu/drm/amd/amdgpu/vcn_v5_0_1.c | 15 +++++++++++++++ 1 file changed, 15 insertions(+) (limited to 'drivers/gpu') diff --git a/drivers/gpu/drm/amd/amdgpu/vcn_v5_0_1.c b/drivers/gpu/drm/amd/amdgpu/vcn_v5_0_1.c index 1e9d2aedf279..338cf43c45fe 100644 --- a/drivers/gpu/drm/amd/amdgpu/vcn_v5_0_1.c +++ b/drivers/gpu/drm/amd/amdgpu/vcn_v5_0_1.c @@ -1038,6 +1038,11 @@ static int vcn_v5_0_1_start(struct amdgpu_vcn_inst *vinst) WREG32_SOC15(VCN, vcn_inst, regVCN_RB_ENABLE, tmp); fw_shared->sq.queue_mode &= ~(FW_QUEUE_RING_RESET | FW_QUEUE_DPG_HOLD_OFF); + /* Keeping one read-back to ensure all register writes are done, + * otherwise it may introduce race conditions. + */ + RREG32_SOC15(VCN, vcn_inst, regUVD_STATUS); + return 0; } @@ -1072,6 +1077,11 @@ static void vcn_v5_0_1_stop_dpg_mode(struct amdgpu_vcn_inst *vinst) /* disable dynamic power gating mode */ WREG32_P(SOC15_REG_OFFSET(VCN, vcn_inst, regUVD_POWER_STATUS), 0, ~UVD_POWER_STATUS__UVD_PG_MODE_MASK); + + /* Keeping one read-back to ensure all register writes are done, + * otherwise it may introduce race conditions. + */ + RREG32_SOC15(VCN, vcn_inst, regUVD_STATUS); } /** @@ -1147,6 +1157,11 @@ static int vcn_v5_0_1_stop(struct amdgpu_vcn_inst *vinst) /* clear status */ WREG32_SOC15(VCN, vcn_inst, regUVD_STATUS, 0); + /* Keeping one read-back to ensure all register writes are done, + * otherwise it may introduce race conditions. + */ + RREG32_SOC15(VCN, vcn_inst, regUVD_STATUS); + return 0; } -- cgit v1.2.3 From 4d2f6b4e4c7ed32e7fa39fcea37344a9eab99094 Mon Sep 17 00:00:00 2001 From: John Olender Date: Tue, 29 Apr 2025 07:24:28 -0400 Subject: drm/amdgpu: amdgpu_vram_mgr_new(): Clamp lpfn to total vram The drm_mm allocator tolerated being passed end > mm->size, but the drm_buddy allocator does not. Restore the pre-buddy-allocator behavior of allowing such placements. Closes: https://gitlab.freedesktop.org/drm/amd/-/issues/3448 Signed-off-by: John Olender Reviewed-by: Alex Deucher Reviewed-by: Arunpravin Paneer Selvam Signed-off-by: Alex Deucher Cc: stable@vger.kernel.org --- drivers/gpu/drm/amd/amdgpu/amdgpu_vram_mgr.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) (limited to 'drivers/gpu') diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_vram_mgr.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_vram_mgr.c index 2d7f82e98df9..abdc52b0895a 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_vram_mgr.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_vram_mgr.c @@ -463,7 +463,7 @@ static int amdgpu_vram_mgr_new(struct ttm_resource_manager *man, int r; lpfn = (u64)place->lpfn << PAGE_SHIFT; - if (!lpfn) + if (!lpfn || lpfn > man->size) lpfn = man->size; fpfn = (u64)place->fpfn << PAGE_SHIFT; -- cgit v1.2.3 From 978592136cfed3491330acfff12a90812d0c6a58 Mon Sep 17 00:00:00 2001 From: Asad Kamal Date: Wed, 21 May 2025 19:17:50 +0800 Subject: drm/amd/pm: Optimize get gpu metrics data function Optimize get gpu metrics data function for smu_v13_0_12 to allocate metrics structure only once v2: Free and alloc moved to same function(Kevin) Signed-off-by: Asad Kamal Reviewed-by: Yang Wang Reviewed-by: Lijo Lazar Signed-off-by: Alex Deucher --- drivers/gpu/drm/amd/pm/swsmu/smu13/smu_v13_0_12_ppt.c | 6 ++---- drivers/gpu/drm/amd/pm/swsmu/smu13/smu_v13_0_6_ppt.c | 8 ++++++-- drivers/gpu/drm/amd/pm/swsmu/smu13/smu_v13_0_6_ppt.h | 2 +- 3 files changed, 9 insertions(+), 7 deletions(-) (limited to 'drivers/gpu') 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 69f92bd35bf2..e0d356f93ab0 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 @@ -378,7 +378,7 @@ ssize_t smu_v13_0_12_get_xcp_metrics(struct smu_context *smu, struct amdgpu_xcp return sizeof(*xcp_metrics); } -ssize_t smu_v13_0_12_get_gpu_metrics(struct smu_context *smu, void **table) +ssize_t smu_v13_0_12_get_gpu_metrics(struct smu_context *smu, void **table, void *smu_metrics) { struct smu_table_context *smu_table = &smu->smu_table; struct gpu_metrics_v1_8 *gpu_metrics = @@ -390,8 +390,7 @@ ssize_t smu_v13_0_12_get_gpu_metrics(struct smu_context *smu, void **table) struct amdgpu_xcp *xcp; u32 inst_mask; - metrics = kzalloc(sizeof(MetricsTable_t), GFP_KERNEL); - memcpy(metrics, smu_table->metrics_table, sizeof(MetricsTable_t)); + metrics = (MetricsTable_t *)smu_metrics; smu_cmn_init_soft_gpu_metrics(gpu_metrics, 1, 8); @@ -533,7 +532,6 @@ ssize_t smu_v13_0_12_get_gpu_metrics(struct smu_context *smu, void **table) gpu_metrics->firmware_timestamp = metrics->Timestamp; *table = (void *)gpu_metrics; - kfree(metrics); return sizeof(*gpu_metrics); } diff --git a/drivers/gpu/drm/amd/pm/swsmu/smu13/smu_v13_0_6_ppt.c b/drivers/gpu/drm/amd/pm/swsmu/smu13/smu_v13_0_6_ppt.c index 0a9488576a4e..271458b74f2d 100644 --- a/drivers/gpu/drm/amd/pm/swsmu/smu13/smu_v13_0_6_ppt.c +++ b/drivers/gpu/drm/amd/pm/swsmu/smu13/smu_v13_0_6_ppt.c @@ -2669,6 +2669,7 @@ static ssize_t smu_v13_0_6_get_gpu_metrics(struct smu_context *smu, void **table MetricsTableV2_t *metrics_v2; struct amdgpu_xcp *xcp; u16 link_width_level; + ssize_t num_bytes; u8 num_jpeg_rings; u32 inst_mask; bool per_inst; @@ -2681,8 +2682,11 @@ static ssize_t smu_v13_0_6_get_gpu_metrics(struct smu_context *smu, void **table } if (amdgpu_ip_version(smu->adev, MP1_HWIP, 0) == IP_VERSION(13, 0, 12) && - smu_v13_0_6_cap_supported(smu, SMU_CAP(STATIC_METRICS))) - return smu_v13_0_12_get_gpu_metrics(smu, table); + smu_v13_0_6_cap_supported(smu, SMU_CAP(STATIC_METRICS))) { + num_bytes = smu_v13_0_12_get_gpu_metrics(smu, table, metrics_v0); + kfree(metrics_v0); + return num_bytes; + } metrics_v1 = (MetricsTableV1_t *)metrics_v0; metrics_v2 = (MetricsTableV2_t *)metrics_v0; diff --git a/drivers/gpu/drm/amd/pm/swsmu/smu13/smu_v13_0_6_ppt.h b/drivers/gpu/drm/amd/pm/swsmu/smu13/smu_v13_0_6_ppt.h index 1a54675c576f..d38d6d76b1e7 100644 --- a/drivers/gpu/drm/amd/pm/swsmu/smu13/smu_v13_0_6_ppt.h +++ b/drivers/gpu/drm/amd/pm/swsmu/smu13/smu_v13_0_6_ppt.h @@ -80,7 +80,7 @@ int smu_v13_0_12_get_max_metrics_size(void); int smu_v13_0_12_setup_driver_pptable(struct smu_context *smu); int smu_v13_0_12_get_smu_metrics_data(struct smu_context *smu, MetricsMember_t member, uint32_t *value); -ssize_t smu_v13_0_12_get_gpu_metrics(struct smu_context *smu, void **table); +ssize_t smu_v13_0_12_get_gpu_metrics(struct smu_context *smu, void **table, void *smu_metrics); ssize_t smu_v13_0_12_get_xcp_metrics(struct smu_context *smu, struct amdgpu_xcp *xcp, void *table, void *smu_metrics); -- cgit v1.2.3 From 1091fba163834f51a02d5d149bd657804e6ab749 Mon Sep 17 00:00:00 2001 From: Sunday Clement Date: Fri, 23 May 2025 17:49:50 -0400 Subject: drm/amdkfd: Identical code for different branches This patch removes the if/else statement in the cik_event_interrupt_wq function because it is redundant with both branches resulting in identical outcomes, this improves code readibility. Signed-off-by: Sunday Clement Reviewed-by: Harish Kasiviswanathan Signed-off-by: Alex Deucher --- drivers/gpu/drm/amd/amdkfd/cik_event_interrupt.c | 7 +------ 1 file changed, 1 insertion(+), 6 deletions(-) (limited to 'drivers/gpu') diff --git a/drivers/gpu/drm/amd/amdkfd/cik_event_interrupt.c b/drivers/gpu/drm/amd/amdkfd/cik_event_interrupt.c index 981d9adcc5e1..73acbe0b7c21 100644 --- a/drivers/gpu/drm/amd/amdkfd/cik_event_interrupt.c +++ b/drivers/gpu/drm/amd/amdkfd/cik_event_interrupt.c @@ -91,7 +91,6 @@ static void cik_event_interrupt_wq(struct kfd_node *dev, const struct cik_ih_ring_entry *ihre = (const struct cik_ih_ring_entry *)ih_ring_entry; uint32_t context_id = ihre->data & 0xfffffff; - unsigned int vmid = (ihre->ring_id & 0x0000ff00) >> 8; u32 pasid = (ihre->ring_id & 0xffff0000) >> 16; if (pasid == 0) @@ -125,11 +124,7 @@ static void cik_event_interrupt_wq(struct kfd_node *dev, return; } - if (info.vmid == vmid) - kfd_signal_vm_fault_event(pdd, &info, NULL); - else - kfd_signal_vm_fault_event(pdd, &info, NULL); - + kfd_signal_vm_fault_event(pdd, &info, NULL); kfd_unref_process(p); } } -- cgit v1.2.3 From 5ae9de5867dbf23e53d244dfd62216bec95234a8 Mon Sep 17 00:00:00 2001 From: Arunpravin Paneer Selvam Date: Mon, 30 Dec 2024 22:24:53 +0530 Subject: drm/amdgpu: Add userq fence support to SDMAv6.0 MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit Add userq fence support to SDMAv6.0 Signed-off-by: Arunpravin Paneer Selvam Acked-by: Christian König Signed-off-by: Alex Deucher --- drivers/gpu/drm/amd/amdgpu/amdgpu_sdma.h | 1 + drivers/gpu/drm/amd/amdgpu/sdma_v6_0.c | 55 +++++++++++++++------- .../amd/include/ivsrcid/gfx/irqsrcs_gfx_11_0_0.h | 1 + 3 files changed, 41 insertions(+), 16 deletions(-) (limited to 'drivers/gpu') diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_sdma.h b/drivers/gpu/drm/amd/amdgpu/amdgpu_sdma.h index 5605921212f0..e5f8951bbb6f 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_sdma.h +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_sdma.h @@ -113,6 +113,7 @@ struct amdgpu_sdma { struct amdgpu_sdma_instance instance[AMDGPU_MAX_SDMA_INSTANCES]; struct amdgpu_irq_src trap_irq; struct amdgpu_irq_src illegal_inst_irq; + struct amdgpu_irq_src fence_irq; struct amdgpu_irq_src ecc_irq; struct amdgpu_irq_src vm_hole_irq; struct amdgpu_irq_src doorbell_invalid_irq; diff --git a/drivers/gpu/drm/amd/amdgpu/sdma_v6_0.c b/drivers/gpu/drm/amd/amdgpu/sdma_v6_0.c index da5b5d64f137..5a70ae17be04 100644 --- a/drivers/gpu/drm/amd/amdgpu/sdma_v6_0.c +++ b/drivers/gpu/drm/amd/amdgpu/sdma_v6_0.c @@ -44,6 +44,7 @@ #include "sdma_v6_0.h" #include "v11_structs.h" #include "mes_userqueue.h" +#include "amdgpu_userq_fence.h" MODULE_FIRMWARE("amdgpu/sdma_6_0_0.bin"); MODULE_FIRMWARE("amdgpu/sdma_6_0_1.bin"); @@ -893,6 +894,9 @@ static int sdma_v6_0_mqd_init(struct amdgpu_device *adev, void *mqd, m->sdmax_rlcx_csa_addr_lo = lower_32_bits(prop->csa_addr); m->sdmax_rlcx_csa_addr_hi = upper_32_bits(prop->csa_addr); + m->sdmax_rlcx_f32_dbg0 = lower_32_bits(prop->fence_address); + m->sdmax_rlcx_f32_dbg1 = upper_32_bits(prop->fence_address); + return 0; } @@ -1315,6 +1319,13 @@ static int sdma_v6_0_sw_init(struct amdgpu_ip_block *ip_block) if (r) return r; + /* SDMA user fence event */ + r = amdgpu_irq_add_id(adev, SOC21_IH_CLIENTID_GFX, + GFX_11_0_0__SRCID__SDMA_FENCE, + &adev->sdma.fence_irq); + if (r) + return r; + for (i = 0; i < adev->sdma.num_instances; i++) { ring = &adev->sdma.instance[i].ring; ring->ring_obj = NULL; @@ -1575,25 +1586,9 @@ static int sdma_v6_0_process_trap_irq(struct amdgpu_device *adev, struct amdgpu_iv_entry *entry) { int instances, queue; - uint32_t mes_queue_id = entry->src_data[0]; DRM_DEBUG("IH: SDMA trap\n"); - if (adev->enable_mes && (mes_queue_id & AMDGPU_FENCE_MES_QUEUE_FLAG)) { - struct amdgpu_mes_queue *queue; - - mes_queue_id &= AMDGPU_FENCE_MES_QUEUE_ID_MASK; - - spin_lock(&adev->mes.queue_id_lock); - queue = idr_find(&adev->mes.queue_id_idr, mes_queue_id); - if (queue) { - DRM_DEBUG("process smda queue id = %d\n", mes_queue_id); - amdgpu_fence_process(queue->ring); - } - spin_unlock(&adev->mes.queue_id_lock); - return 0; - } - queue = entry->ring_id & 0xf; instances = (entry->ring_id & 0xf0) >> 4; if (instances > 1) { @@ -1615,6 +1610,29 @@ static int sdma_v6_0_process_trap_irq(struct amdgpu_device *adev, return 0; } +static int sdma_v6_0_process_fence_irq(struct amdgpu_device *adev, + struct amdgpu_irq_src *source, + struct amdgpu_iv_entry *entry) +{ + u32 doorbell_offset = entry->src_data[0]; + + if (adev->enable_mes && doorbell_offset) { + struct amdgpu_userq_fence_driver *fence_drv = NULL; + struct xarray *xa = &adev->userq_xa; + unsigned long flags; + + doorbell_offset >>= SDMA0_QUEUE0_DOORBELL_OFFSET__OFFSET__SHIFT; + + xa_lock_irqsave(xa, flags); + fence_drv = xa_load(xa, doorbell_offset); + if (fence_drv) + amdgpu_userq_fence_driver_process(fence_drv); + xa_unlock_irqrestore(xa, flags); + } + + return 0; +} + static int sdma_v6_0_process_illegal_inst_irq(struct amdgpu_device *adev, struct amdgpu_irq_src *source, struct amdgpu_iv_entry *entry) @@ -1751,6 +1769,10 @@ static const struct amdgpu_irq_src_funcs sdma_v6_0_trap_irq_funcs = { .process = sdma_v6_0_process_trap_irq, }; +static const struct amdgpu_irq_src_funcs sdma_v6_0_fence_irq_funcs = { + .process = sdma_v6_0_process_fence_irq, +}; + static const struct amdgpu_irq_src_funcs sdma_v6_0_illegal_inst_irq_funcs = { .process = sdma_v6_0_process_illegal_inst_irq, }; @@ -1760,6 +1782,7 @@ static void sdma_v6_0_set_irq_funcs(struct amdgpu_device *adev) adev->sdma.trap_irq.num_types = AMDGPU_SDMA_IRQ_INSTANCE0 + adev->sdma.num_instances; adev->sdma.trap_irq.funcs = &sdma_v6_0_trap_irq_funcs; + adev->sdma.fence_irq.funcs = &sdma_v6_0_fence_irq_funcs; adev->sdma.illegal_inst_irq.funcs = &sdma_v6_0_illegal_inst_irq_funcs; } diff --git a/drivers/gpu/drm/amd/include/ivsrcid/gfx/irqsrcs_gfx_11_0_0.h b/drivers/gpu/drm/amd/include/ivsrcid/gfx/irqsrcs_gfx_11_0_0.h index 3a4670bc4449..b98b7ae551b5 100644 --- a/drivers/gpu/drm/amd/include/ivsrcid/gfx/irqsrcs_gfx_11_0_0.h +++ b/drivers/gpu/drm/amd/include/ivsrcid/gfx/irqsrcs_gfx_11_0_0.h @@ -48,6 +48,7 @@ #define GFX_11_0_0__SRCID__SDMA_SRAM_ECC 64 // 0x40 SRAM ECC Error #define GFX_11_0_0__SRCID__SDMA_SEM_INCOMPLETE_TIMEOUT 65 // 0x41 GPF(Sem incomplete timeout) #define GFX_11_0_0__SRCID__SDMA_SEM_WAIT_FAIL_TIMEOUT 66 // 0x42 Semaphore wait fail timeout +#define GFX_11_0_0__SRCID__SDMA_FENCE 67 // 0x43 User fence #define GFX_11_0_0__SRCID__RLC_GC_FED_INTERRUPT 128 // 0x80 FED Interrupt (for data poisoning) -- cgit v1.2.3 From 90237b16ec1d7afa16e2173cc9a664377214cdd9 Mon Sep 17 00:00:00 2001 From: Yifan Zhang Date: Wed, 21 May 2025 18:06:28 +0800 Subject: amd/amdkfd: fix a kfd_process ref leak This patch is to fix a kfd_prcess ref leak. Signed-off-by: Yifan Zhang Reviewed-by: Philip Yang Signed-off-by: Alex Deucher --- drivers/gpu/drm/amd/amdkfd/kfd_events.c | 1 + 1 file changed, 1 insertion(+) (limited to 'drivers/gpu') diff --git a/drivers/gpu/drm/amd/amdkfd/kfd_events.c b/drivers/gpu/drm/amd/amdkfd/kfd_events.c index e54e708ed82d..2b294ada3ec0 100644 --- a/drivers/gpu/drm/amd/amdkfd/kfd_events.c +++ b/drivers/gpu/drm/amd/amdkfd/kfd_events.c @@ -1350,6 +1350,7 @@ void kfd_signal_poison_consumed_event(struct kfd_node *dev, u32 pasid) user_gpu_id = kfd_process_get_user_gpu_id(p, dev->id); if (unlikely(user_gpu_id == -EINVAL)) { WARN_ONCE(1, "Could not get user_gpu_id from dev->id:%x\n", dev->id); + kfd_unref_process(p); return; } -- cgit v1.2.3 From 31e837d242cbb58afed8e0d2a722745bc50ad154 Mon Sep 17 00:00:00 2001 From: ganglxie Date: Thu, 22 May 2025 14:28:52 +0800 Subject: drm/amdgpu: handle old RAS eeprom data in non-nps1 mode Get MCA address from PA in nps1, then convert MCA address to PA in specific nps mode. Signed-off-by: ganglxie Reviewed-by: Tao Zhou Signed-off-by: Alex Deucher --- drivers/gpu/drm/amd/amdgpu/amdgpu_ras.c | 16 ++++++++++++++-- drivers/gpu/drm/amd/amdgpu/amdgpu_umc.c | 23 +++++++++++++++++++++++ drivers/gpu/drm/amd/amdgpu/amdgpu_umc.h | 2 ++ 3 files changed, 39 insertions(+), 2 deletions(-) (limited to 'drivers/gpu') diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_ras.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_ras.c index 8a13a0c7b9cd..b0b1dbd4ae35 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_ras.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_ras.c @@ -2886,8 +2886,20 @@ static int __amdgpu_ras_convert_rec_from_rom(struct amdgpu_device *adev, bps->retired_page << AMDGPU_GPU_PAGE_SHIFT)) return -EINVAL; } else { - if (amdgpu_ras_mca2pa_by_idx(adev, bps, err_data)) - return -EINVAL; + if (bps->address) { + if (amdgpu_ras_mca2pa_by_idx(adev, bps, err_data)) + return -EINVAL; + } else { + /* for specific old eeprom data, mca address is not stored, + * calc it from pa + */ + if (amdgpu_umc_pa2mca(adev, bps->retired_page << AMDGPU_GPU_PAGE_SHIFT, + &(bps->address), AMDGPU_NPS1_PARTITION_MODE)) + return -EINVAL; + + if (amdgpu_ras_mca2pa(adev, bps, err_data)) + return -EOPNOTSUPP; + } } return __amdgpu_ras_restore_bad_pages(adev, err_data->err_addr, diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_umc.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_umc.c index 8c6e55b5b967..c92b8794aa73 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_umc.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_umc.c @@ -562,3 +562,26 @@ int amdgpu_umc_mca_to_addr(struct amdgpu_device *adev, return 0; } + +int amdgpu_umc_pa2mca(struct amdgpu_device *adev, + uint64_t pa, uint64_t *mca, enum amdgpu_memory_partition nps) +{ + struct ta_ras_query_address_input addr_in; + struct ta_ras_query_address_output addr_out; + int ret; + + /* nps: the pa belongs to */ + addr_in.pa.pa = pa | ((uint64_t)nps << 58); + addr_in.addr_type = TA_RAS_PA_TO_MCA; + ret = psp_ras_query_address(&adev->psp, &addr_in, &addr_out); + if (ret) { + dev_warn(adev->dev, "Failed to query RAS MCA address for 0x%llx", + pa); + + return ret; + } + + *mca = addr_out.ma.err_addr; + + return 0; +} diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_umc.h b/drivers/gpu/drm/amd/amdgpu/amdgpu_umc.h index 29ce6b1d214a..ec203f9e5ffa 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_umc.h +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_umc.h @@ -189,4 +189,6 @@ int amdgpu_umc_mca_to_addr(struct amdgpu_device *adev, uint64_t err_addr, uint32_t ch, uint32_t umc, uint32_t node, uint32_t socket, struct ta_ras_query_address_output *addr_out, bool dump_addr); +int amdgpu_umc_pa2mca(struct amdgpu_device *adev, + uint64_t pa, uint64_t *mca, enum amdgpu_memory_partition nps); #endif -- cgit v1.2.3 From fce0afca3562ceb5386241159280590beeee2f51 Mon Sep 17 00:00:00 2001 From: ganglxie Date: Thu, 22 May 2025 15:14:28 +0800 Subject: drm/amdgpu: Get mca address for old eeprom records after getting mca address for old eeprom records with 'address==0', it can be correctly parsed under none-nps1, or it will be dropped. Signed-off-by: ganglxie Reviewed-by: Tao Zhou Signed-off-by: Alex Deucher --- drivers/gpu/drm/amd/amdgpu/amdgpu_ras.c | 9 +++++++++ 1 file changed, 9 insertions(+) (limited to 'drivers/gpu') diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_ras.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_ras.c index b0b1dbd4ae35..de0944947eaf 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_ras.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_ras.c @@ -2859,6 +2859,15 @@ static int __amdgpu_ras_convert_rec_array_from_rom(struct amdgpu_device *adev, return -EINVAL; } } else { + if (bps[0].address == 0) { + /* for specific old eeprom data, mca address is not stored, + * calc it from pa + */ + if (amdgpu_umc_pa2mca(adev, bps[0].retired_page << AMDGPU_GPU_PAGE_SHIFT, + &(bps[0].address), AMDGPU_NPS1_PARTITION_MODE)) + return -EINVAL; + } + if (amdgpu_ras_mca2pa(adev, &bps[0], err_data)) { if (nps == AMDGPU_NPS1_PARTITION_MODE) memcpy(err_data->err_addr, bps, -- cgit v1.2.3 From c3e9826a22027a21d998d3e64882fa377b613006 Mon Sep 17 00:00:00 2001 From: Wentao Liang Date: Mon, 26 May 2025 10:37:31 +0800 Subject: drm/amd/display: Add null pointer check for get_first_active_display() The function mod_hdcp_hdcp1_enable_encryption() calls the function get_first_active_display(), but does not check its return value. The return value is a null pointer if the display list is empty. This will lead to a null pointer dereference in mod_hdcp_hdcp2_enable_encryption(). Add a null pointer check for get_first_active_display() and return MOD_HDCP_STATUS_DISPLAY_NOT_FOUND if the function return null. Fixes: 2deade5ede56 ("drm/amd/display: Remove hdcp display state with mst fix") Signed-off-by: Wentao Liang Reviewed-by: Alex Hung Signed-off-by: Alex Deucher Cc: stable@vger.kernel.org # v5.8 --- drivers/gpu/drm/amd/display/modules/hdcp/hdcp_psp.c | 3 +++ 1 file changed, 3 insertions(+) (limited to 'drivers/gpu') diff --git a/drivers/gpu/drm/amd/display/modules/hdcp/hdcp_psp.c b/drivers/gpu/drm/amd/display/modules/hdcp/hdcp_psp.c index 8c137d7c032e..e58e7b93810b 100644 --- a/drivers/gpu/drm/amd/display/modules/hdcp/hdcp_psp.c +++ b/drivers/gpu/drm/amd/display/modules/hdcp/hdcp_psp.c @@ -368,6 +368,9 @@ enum mod_hdcp_status mod_hdcp_hdcp1_enable_encryption(struct mod_hdcp *hdcp) struct mod_hdcp_display *display = get_first_active_display(hdcp); enum mod_hdcp_status status = MOD_HDCP_STATUS_SUCCESS; + if (!display) + return MOD_HDCP_STATUS_DISPLAY_NOT_FOUND; + mutex_lock(&psp->hdcp_context.mutex); hdcp_cmd = (struct ta_hdcp_shared_memory *)psp->hdcp_context.context.mem_context.shared_buf; memset(hdcp_cmd, 0, sizeof(struct ta_hdcp_shared_memory)); -- cgit v1.2.3 From 347efe5b396c61bc129832910665d899ee86448e Mon Sep 17 00:00:00 2001 From: Christophe JAILLET Date: Sat, 24 May 2025 18:51:25 +0200 Subject: drm/amd/display: Constify struct timing_generator_funcs 'struct timing_generator_funcs' are not modified in these drivers. Constifying these structures moves some data to a read-only section, so increases overall security, especially when the structure holds some function pointers. Signed-off-by: Christophe JAILLET Reviewed-by: Alex Hung Signed-off-by: Alex Deucher --- drivers/gpu/drm/amd/display/dc/optc/dcn20/dcn20_optc.c | 2 +- drivers/gpu/drm/amd/display/dc/optc/dcn201/dcn201_optc.c | 2 +- drivers/gpu/drm/amd/display/dc/optc/dcn30/dcn30_optc.c | 2 +- drivers/gpu/drm/amd/display/dc/optc/dcn301/dcn301_optc.c | 2 +- drivers/gpu/drm/amd/display/dc/optc/dcn31/dcn31_optc.c | 2 +- drivers/gpu/drm/amd/display/dc/optc/dcn314/dcn314_optc.c | 2 +- drivers/gpu/drm/amd/display/dc/optc/dcn32/dcn32_optc.c | 2 +- drivers/gpu/drm/amd/display/dc/optc/dcn35/dcn35_optc.c | 2 +- drivers/gpu/drm/amd/display/dc/optc/dcn401/dcn401_optc.c | 2 +- 9 files changed, 9 insertions(+), 9 deletions(-) (limited to 'drivers/gpu') diff --git a/drivers/gpu/drm/amd/display/dc/optc/dcn20/dcn20_optc.c b/drivers/gpu/drm/amd/display/dc/optc/dcn20/dcn20_optc.c index 81857ce6d68d..e7a90a437fff 100644 --- a/drivers/gpu/drm/amd/display/dc/optc/dcn20/dcn20_optc.c +++ b/drivers/gpu/drm/amd/display/dc/optc/dcn20/dcn20_optc.c @@ -502,7 +502,7 @@ void optc2_get_last_used_drr_vtotal(struct timing_generator *optc, uint32_t *ref REG_GET(OTG_DRR_CONTROL, OTG_V_TOTAL_LAST_USED_BY_DRR, refresh_rate); } -static struct timing_generator_funcs dcn20_tg_funcs = { +static const struct timing_generator_funcs dcn20_tg_funcs = { .validate_timing = optc1_validate_timing, .program_timing = optc1_program_timing, .setup_vertical_interrupt0 = optc1_setup_vertical_interrupt0, diff --git a/drivers/gpu/drm/amd/display/dc/optc/dcn201/dcn201_optc.c b/drivers/gpu/drm/amd/display/dc/optc/dcn201/dcn201_optc.c index f2415eebdc09..772a8bfb949c 100644 --- a/drivers/gpu/drm/amd/display/dc/optc/dcn201/dcn201_optc.c +++ b/drivers/gpu/drm/amd/display/dc/optc/dcn201/dcn201_optc.c @@ -129,7 +129,7 @@ static void optc201_get_optc_source(struct timing_generator *optc, *num_of_src_opp = 1; } -static struct timing_generator_funcs dcn201_tg_funcs = { +static const struct timing_generator_funcs dcn201_tg_funcs = { .validate_timing = optc201_validate_timing, .program_timing = optc1_program_timing, .setup_vertical_interrupt0 = optc1_setup_vertical_interrupt0, diff --git a/drivers/gpu/drm/amd/display/dc/optc/dcn30/dcn30_optc.c b/drivers/gpu/drm/amd/display/dc/optc/dcn30/dcn30_optc.c index 78b58a449fa4..ee4665aa49e9 100644 --- a/drivers/gpu/drm/amd/display/dc/optc/dcn30/dcn30_optc.c +++ b/drivers/gpu/drm/amd/display/dc/optc/dcn30/dcn30_optc.c @@ -357,7 +357,7 @@ void optc3_tg_init(struct timing_generator *optc) optc1_clear_optc_underflow(optc); } -static struct timing_generator_funcs dcn30_tg_funcs = { +static const struct timing_generator_funcs dcn30_tg_funcs = { .validate_timing = optc1_validate_timing, .program_timing = optc1_program_timing, .setup_vertical_interrupt0 = optc1_setup_vertical_interrupt0, diff --git a/drivers/gpu/drm/amd/display/dc/optc/dcn301/dcn301_optc.c b/drivers/gpu/drm/amd/display/dc/optc/dcn301/dcn301_optc.c index 65e9089b7f31..38f85bc2681a 100644 --- a/drivers/gpu/drm/amd/display/dc/optc/dcn301/dcn301_optc.c +++ b/drivers/gpu/drm/amd/display/dc/optc/dcn301/dcn301_optc.c @@ -109,7 +109,7 @@ void optc301_setup_manual_trigger(struct timing_generator *optc) OTG_TRIGA_CLEAR, 1); } -static struct timing_generator_funcs dcn30_tg_funcs = { +static const struct timing_generator_funcs dcn30_tg_funcs = { .validate_timing = optc1_validate_timing, .program_timing = optc1_program_timing, .setup_vertical_interrupt0 = optc1_setup_vertical_interrupt0, diff --git a/drivers/gpu/drm/amd/display/dc/optc/dcn31/dcn31_optc.c b/drivers/gpu/drm/amd/display/dc/optc/dcn31/dcn31_optc.c index ef536f37b4ed..4f1830ba619f 100644 --- a/drivers/gpu/drm/amd/display/dc/optc/dcn31/dcn31_optc.c +++ b/drivers/gpu/drm/amd/display/dc/optc/dcn31/dcn31_optc.c @@ -315,7 +315,7 @@ void optc31_read_otg_state(struct timing_generator *optc, s->otg_double_buffer_control = REG_READ(OTG_DOUBLE_BUFFER_CONTROL); } -static struct timing_generator_funcs dcn31_tg_funcs = { +static const struct timing_generator_funcs dcn31_tg_funcs = { .validate_timing = optc1_validate_timing, .program_timing = optc1_program_timing, .setup_vertical_interrupt0 = optc1_setup_vertical_interrupt0, diff --git a/drivers/gpu/drm/amd/display/dc/optc/dcn314/dcn314_optc.c b/drivers/gpu/drm/amd/display/dc/optc/dcn314/dcn314_optc.c index 0e603bad0d12..4a2caca37255 100644 --- a/drivers/gpu/drm/amd/display/dc/optc/dcn314/dcn314_optc.c +++ b/drivers/gpu/drm/amd/display/dc/optc/dcn314/dcn314_optc.c @@ -192,7 +192,7 @@ static void optc314_set_h_timing_div_manual_mode(struct timing_generator *optc, } -static struct timing_generator_funcs dcn314_tg_funcs = { +static const struct timing_generator_funcs dcn314_tg_funcs = { .validate_timing = optc1_validate_timing, .program_timing = optc1_program_timing, .setup_vertical_interrupt0 = optc1_setup_vertical_interrupt0, diff --git a/drivers/gpu/drm/amd/display/dc/optc/dcn32/dcn32_optc.c b/drivers/gpu/drm/amd/display/dc/optc/dcn32/dcn32_optc.c index 2cdd19ba634b..b2b226bcd871 100644 --- a/drivers/gpu/drm/amd/display/dc/optc/dcn32/dcn32_optc.c +++ b/drivers/gpu/drm/amd/display/dc/optc/dcn32/dcn32_optc.c @@ -297,7 +297,7 @@ static void optc32_set_drr( optc32_setup_manual_trigger(optc); } -static struct timing_generator_funcs dcn32_tg_funcs = { +static const struct timing_generator_funcs dcn32_tg_funcs = { .validate_timing = optc1_validate_timing, .program_timing = optc1_program_timing, .setup_vertical_interrupt0 = optc1_setup_vertical_interrupt0, diff --git a/drivers/gpu/drm/amd/display/dc/optc/dcn35/dcn35_optc.c b/drivers/gpu/drm/amd/display/dc/optc/dcn35/dcn35_optc.c index 4cfc6c0fa147..72bff94cb57d 100644 --- a/drivers/gpu/drm/amd/display/dc/optc/dcn35/dcn35_optc.c +++ b/drivers/gpu/drm/amd/display/dc/optc/dcn35/dcn35_optc.c @@ -428,7 +428,7 @@ static void optc35_set_long_vtotal( } } -static struct timing_generator_funcs dcn35_tg_funcs = { +static const struct timing_generator_funcs dcn35_tg_funcs = { .validate_timing = optc1_validate_timing, .program_timing = optc1_program_timing, .setup_vertical_interrupt0 = optc1_setup_vertical_interrupt0, diff --git a/drivers/gpu/drm/amd/display/dc/optc/dcn401/dcn401_optc.c b/drivers/gpu/drm/amd/display/dc/optc/dcn401/dcn401_optc.c index 382ac18e7854..ff79c38287df 100644 --- a/drivers/gpu/drm/amd/display/dc/optc/dcn401/dcn401_optc.c +++ b/drivers/gpu/drm/amd/display/dc/optc/dcn401/dcn401_optc.c @@ -459,7 +459,7 @@ bool optc401_wait_update_lock_status(struct timing_generator *tg, bool locked) return true; } -static struct timing_generator_funcs dcn401_tg_funcs = { +static const struct timing_generator_funcs dcn401_tg_funcs = { .validate_timing = optc1_validate_timing, .program_timing = optc1_program_timing, .setup_vertical_interrupt0 = optc1_setup_vertical_interrupt0, -- cgit v1.2.3 From c7d43a148593e71563a4d3b83527d2917149526d Mon Sep 17 00:00:00 2001 From: Asad Kamal Date: Mon, 26 May 2025 15:23:41 +0800 Subject: drm/amd/pm: Enable static metrics table support Enable static metrics support to fetch board voltage and pldm version for other smu_v13_0_6 program Signed-off-by: Asad Kamal Reviewed-by: Lijo Lazar Reviewed-by: Hawking Zhang Signed-off-by: Alex Deucher --- drivers/gpu/drm/amd/pm/swsmu/smu13/smu_v13_0_6_ppt.c | 6 ++++-- 1 file changed, 4 insertions(+), 2 deletions(-) (limited to 'drivers/gpu') diff --git a/drivers/gpu/drm/amd/pm/swsmu/smu13/smu_v13_0_6_ppt.c b/drivers/gpu/drm/amd/pm/swsmu/smu13/smu_v13_0_6_ppt.c index 271458b74f2d..2651c8c41ad7 100644 --- a/drivers/gpu/drm/amd/pm/swsmu/smu13/smu_v13_0_6_ppt.c +++ b/drivers/gpu/drm/amd/pm/swsmu/smu13/smu_v13_0_6_ppt.c @@ -392,11 +392,13 @@ static void smu_v13_0_6_init_caps(struct smu_context *smu) if ((pgm == 7 && fw_ver >= 0x7550E00) || (pgm == 0 && fw_ver >= 0x00557E00)) smu_v13_0_6_cap_set(smu, SMU_CAP(HST_LIMIT_METRICS)); - if (fw_ver >= 0x00557F01) { + if ((pgm == 0 && fw_ver >= 0x00557F01) || + (pgm == 7 && fw_ver >= 0x7551000)) { smu_v13_0_6_cap_set(smu, SMU_CAP(STATIC_METRICS)); smu_v13_0_6_cap_set(smu, SMU_CAP(BOARD_VOLTAGE)); } - if (fw_ver >= 0x00558000) + if ((pgm == 0 && fw_ver >= 0x00558000) || + (pgm == 7 && fw_ver >= 0x7551000)) smu_v13_0_6_cap_set(smu, SMU_CAP(PLDM_VERSION)); } if (((pgm == 7) && (fw_ver >= 0x7550700)) || -- cgit v1.2.3 From 324194b528b490d431af24b88c7596f480ed7c81 Mon Sep 17 00:00:00 2001 From: Asad Kamal Date: Mon, 26 May 2025 15:27:08 +0800 Subject: drm/amd/pm: Enable static metrics table support Enable static metrics support to fetch board voltage and pldm version for smu_v13_0_14 Signed-off-by: Asad Kamal Reviewed-by: Lijo Lazar Reviewed-by: Hawking Zhang Signed-off-by: Alex Deucher --- drivers/gpu/drm/amd/pm/swsmu/smu13/smu_v13_0_6_ppt.c | 5 +++++ 1 file changed, 5 insertions(+) (limited to 'drivers/gpu') diff --git a/drivers/gpu/drm/amd/pm/swsmu/smu13/smu_v13_0_6_ppt.c b/drivers/gpu/drm/amd/pm/swsmu/smu13/smu_v13_0_6_ppt.c index 2651c8c41ad7..f00ef7f3f355 100644 --- a/drivers/gpu/drm/amd/pm/swsmu/smu13/smu_v13_0_6_ppt.c +++ b/drivers/gpu/drm/amd/pm/swsmu/smu13/smu_v13_0_6_ppt.c @@ -312,6 +312,11 @@ static void smu_v13_0_14_init_caps(struct smu_context *smu) smu_v13_0_6_cap_set(smu, SMU_CAP(PER_INST_METRICS)); if (fw_ver >= 0x5551200) smu_v13_0_6_cap_set(smu, SMU_CAP(SDMA_RESET)); + if (fw_ver >= 0x5551600) { + smu_v13_0_6_cap_set(smu, SMU_CAP(STATIC_METRICS)); + smu_v13_0_6_cap_set(smu, SMU_CAP(BOARD_VOLTAGE)); + smu_v13_0_6_cap_set(smu, SMU_CAP(PLDM_VERSION)); + } } static void smu_v13_0_12_init_caps(struct smu_context *smu) -- cgit v1.2.3 From 82a277d529f3fa03d2374f961143afd97469e44e Mon Sep 17 00:00:00 2001 From: Mario Limonciello Date: Tue, 27 May 2025 10:59:42 -0500 Subject: drm/amd: Export DMCUB version to sysfs For supported ASICs DMCU version is exported, but ASICs that support DMCUB there is no information exported to sysfs. Add an attribute for DMCUB. Reviewed-by: Lijo Lazar Link: https://lore.kernel.org/r/20250527155942.476354-1-mario.limonciello@amd.com Signed-off-by: Mario Limonciello Signed-off-by: Alex Deucher --- drivers/gpu/drm/amd/amdgpu/amdgpu_ucode.c | 8 +++++--- 1 file changed, 5 insertions(+), 3 deletions(-) (limited to 'drivers/gpu') diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_ucode.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_ucode.c index 4a72c2bbd49e..2505c46a9c3d 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_ucode.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_ucode.c @@ -765,6 +765,7 @@ FW_VERSION_ATTR(sdma_fw_version, 0444, sdma.instance[0].fw_version); FW_VERSION_ATTR(sdma2_fw_version, 0444, sdma.instance[1].fw_version); FW_VERSION_ATTR(vcn_fw_version, 0444, vcn.fw_version); FW_VERSION_ATTR(dmcu_fw_version, 0444, dm.dmcu_fw_version); +FW_VERSION_ATTR(dmcub_fw_version, 0444, dm.dmcub_fw_version); FW_VERSION_ATTR(mes_fw_version, 0444, mes.sched_version & AMDGPU_MES_VERSION_MASK); FW_VERSION_ATTR(mes_kiq_fw_version, 0444, mes.kiq_version & AMDGPU_MES_VERSION_MASK); FW_VERSION_ATTR(pldm_fw_version, 0444, firmware.pldm_version); @@ -780,9 +781,10 @@ static struct attribute *fw_attrs[] = { &dev_attr_ta_ras_fw_version.attr, &dev_attr_ta_xgmi_fw_version.attr, &dev_attr_smc_fw_version.attr, &dev_attr_sdma_fw_version.attr, &dev_attr_sdma2_fw_version.attr, &dev_attr_vcn_fw_version.attr, - &dev_attr_dmcu_fw_version.attr, &dev_attr_imu_fw_version.attr, - &dev_attr_mes_fw_version.attr, &dev_attr_mes_kiq_fw_version.attr, - &dev_attr_pldm_fw_version.attr, NULL + &dev_attr_dmcu_fw_version.attr, &dev_attr_dmcub_fw_version.attr, + &dev_attr_imu_fw_version.attr, &dev_attr_mes_fw_version.attr, + &dev_attr_mes_kiq_fw_version.attr, &dev_attr_pldm_fw_version.attr, + NULL }; #define to_dev_attr(x) container_of(x, struct device_attribute, attr) -- cgit v1.2.3 From 40f970ba7a4ab77be2ffe6d50a70416c8876496a Mon Sep 17 00:00:00 2001 From: Alex Deucher Date: Mon, 19 May 2025 15:46:25 -0400 Subject: drm/amdgpu/mes: add missing locking in helper functions We need to take the MES lock. Reviewed-by: Michael Chen Signed-off-by: Alex Deucher Cc: stable@vger.kernel.org --- drivers/gpu/drm/amd/amdgpu/amdgpu_mes.c | 16 ++++++++++++++++ 1 file changed, 16 insertions(+) (limited to 'drivers/gpu') diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_mes.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_mes.c index 2febb63ab232..fe772c380120 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_mes.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_mes.c @@ -300,7 +300,9 @@ int amdgpu_mes_map_legacy_queue(struct amdgpu_device *adev, queue_input.mqd_addr = amdgpu_bo_gpu_offset(ring->mqd_obj); queue_input.wptr_addr = ring->wptr_gpu_addr; + amdgpu_mes_lock(&adev->mes); r = adev->mes.funcs->map_legacy_queue(&adev->mes, &queue_input); + amdgpu_mes_unlock(&adev->mes); if (r) DRM_ERROR("failed to map legacy queue\n"); @@ -323,7 +325,9 @@ int amdgpu_mes_unmap_legacy_queue(struct amdgpu_device *adev, queue_input.trail_fence_addr = gpu_addr; queue_input.trail_fence_data = seq; + amdgpu_mes_lock(&adev->mes); r = adev->mes.funcs->unmap_legacy_queue(&adev->mes, &queue_input); + amdgpu_mes_unlock(&adev->mes); if (r) DRM_ERROR("failed to unmap legacy queue\n"); @@ -353,7 +357,9 @@ int amdgpu_mes_reset_legacy_queue(struct amdgpu_device *adev, if (ring->funcs->type == AMDGPU_RING_TYPE_GFX) queue_input.legacy_gfx = true; + amdgpu_mes_lock(&adev->mes); r = adev->mes.funcs->reset_hw_queue(&adev->mes, &queue_input); + amdgpu_mes_unlock(&adev->mes); if (r) DRM_ERROR("failed to reset legacy queue\n"); @@ -383,7 +389,9 @@ uint32_t amdgpu_mes_rreg(struct amdgpu_device *adev, uint32_t reg) goto error; } + amdgpu_mes_lock(&adev->mes); r = adev->mes.funcs->misc_op(&adev->mes, &op_input); + amdgpu_mes_unlock(&adev->mes); if (r) dev_err(adev->dev, "failed to read reg (0x%x)\n", reg); else @@ -411,7 +419,9 @@ int amdgpu_mes_wreg(struct amdgpu_device *adev, goto error; } + amdgpu_mes_lock(&adev->mes); r = adev->mes.funcs->misc_op(&adev->mes, &op_input); + amdgpu_mes_unlock(&adev->mes); if (r) dev_err(adev->dev, "failed to write reg (0x%x)\n", reg); @@ -438,7 +448,9 @@ int amdgpu_mes_reg_write_reg_wait(struct amdgpu_device *adev, goto error; } + amdgpu_mes_lock(&adev->mes); r = adev->mes.funcs->misc_op(&adev->mes, &op_input); + amdgpu_mes_unlock(&adev->mes); if (r) dev_err(adev->dev, "failed to reg_write_reg_wait\n"); @@ -463,7 +475,9 @@ int amdgpu_mes_reg_wait(struct amdgpu_device *adev, uint32_t reg, goto error; } + amdgpu_mes_lock(&adev->mes); r = adev->mes.funcs->misc_op(&adev->mes, &op_input); + amdgpu_mes_unlock(&adev->mes); if (r) dev_err(adev->dev, "failed to reg_write_reg_wait\n"); @@ -694,7 +708,9 @@ static int amdgpu_mes_set_enforce_isolation(struct amdgpu_device *adev, goto error; } + amdgpu_mes_lock(&adev->mes); r = adev->mes.funcs->misc_op(&adev->mes, &op_input); + amdgpu_mes_unlock(&adev->mes); if (r) dev_err(adev->dev, "failed to change_config.\n"); -- cgit v1.2.3 From 684530526f07d488986bc34244034e54e00759a4 Mon Sep 17 00:00:00 2001 From: Alex Deucher Date: Mon, 19 May 2025 15:51:47 -0400 Subject: drm/amdgpu/mes: remove some unused functions Nothing uses them so remove them. Leftover from MES bring up. Reviewed-by: Michael Chen Signed-off-by: Alex Deucher --- drivers/gpu/drm/amd/amdgpu/amdgpu_mes.c | 63 --------------------------------- drivers/gpu/drm/amd/amdgpu/amdgpu_mes.h | 4 --- 2 files changed, 67 deletions(-) (limited to 'drivers/gpu') diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_mes.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_mes.c index fe772c380120..6fa9fa11c8f3 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_mes.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_mes.c @@ -458,33 +458,6 @@ error: return r; } -int amdgpu_mes_reg_wait(struct amdgpu_device *adev, uint32_t reg, - uint32_t val, uint32_t mask) -{ - struct mes_misc_op_input op_input; - int r; - - op_input.op = MES_MISC_OP_WRM_REG_WAIT; - op_input.wrm_reg.reg0 = reg; - op_input.wrm_reg.ref = val; - op_input.wrm_reg.mask = mask; - - if (!adev->mes.funcs->misc_op) { - dev_err(adev->dev, "mes reg wait is not supported!\n"); - r = -EINVAL; - goto error; - } - - amdgpu_mes_lock(&adev->mes); - r = adev->mes.funcs->misc_op(&adev->mes, &op_input); - amdgpu_mes_unlock(&adev->mes); - if (r) - dev_err(adev->dev, "failed to reg_write_reg_wait\n"); - -error: - return r; -} - int amdgpu_mes_set_shader_debugger(struct amdgpu_device *adev, uint64_t process_context_addr, uint32_t spi_gdbg_per_vmid_cntl, @@ -553,42 +526,6 @@ int amdgpu_mes_flush_shader_debugger(struct amdgpu_device *adev, return r; } -#define DEFINE_AMDGPU_MES_CTX_GET_OFFS_ENG(_eng) \ -do { \ - if (id_offs < AMDGPU_MES_CTX_MAX_OFFS) \ - return offsetof(struct amdgpu_mes_ctx_meta_data, \ - _eng[ring->idx].slots[id_offs]); \ - else if (id_offs == AMDGPU_MES_CTX_RING_OFFS) \ - return offsetof(struct amdgpu_mes_ctx_meta_data, \ - _eng[ring->idx].ring); \ - else if (id_offs == AMDGPU_MES_CTX_IB_OFFS) \ - return offsetof(struct amdgpu_mes_ctx_meta_data, \ - _eng[ring->idx].ib); \ - else if (id_offs == AMDGPU_MES_CTX_PADDING_OFFS) \ - return offsetof(struct amdgpu_mes_ctx_meta_data, \ - _eng[ring->idx].padding); \ -} while(0) - -int amdgpu_mes_ctx_get_offs(struct amdgpu_ring *ring, unsigned int id_offs) -{ - switch (ring->funcs->type) { - case AMDGPU_RING_TYPE_GFX: - DEFINE_AMDGPU_MES_CTX_GET_OFFS_ENG(gfx); - break; - case AMDGPU_RING_TYPE_COMPUTE: - DEFINE_AMDGPU_MES_CTX_GET_OFFS_ENG(compute); - break; - case AMDGPU_RING_TYPE_SDMA: - DEFINE_AMDGPU_MES_CTX_GET_OFFS_ENG(sdma); - break; - default: - break; - } - - WARN_ON(1); - return -EINVAL; -} - uint32_t amdgpu_mes_get_aggregated_doorbell_index(struct amdgpu_device *adev, enum amdgpu_mes_priority_level prio) { diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_mes.h b/drivers/gpu/drm/amd/amdgpu/amdgpu_mes.h index a41f65b4f733..c0d2c195fe2e 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_mes.h +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_mes.h @@ -372,8 +372,6 @@ struct amdgpu_mes_funcs { #define amdgpu_mes_kiq_hw_init(adev) (adev)->mes.kiq_hw_init((adev)) #define amdgpu_mes_kiq_hw_fini(adev) (adev)->mes.kiq_hw_fini((adev)) -int amdgpu_mes_ctx_get_offs(struct amdgpu_ring *ring, unsigned int id_offs); - int amdgpu_mes_init_microcode(struct amdgpu_device *adev, int pipe); int amdgpu_mes_init(struct amdgpu_device *adev); void amdgpu_mes_fini(struct amdgpu_device *adev); @@ -395,8 +393,6 @@ int amdgpu_mes_reset_legacy_queue(struct amdgpu_device *adev, uint32_t amdgpu_mes_rreg(struct amdgpu_device *adev, uint32_t reg); int amdgpu_mes_wreg(struct amdgpu_device *adev, uint32_t reg, uint32_t val); -int amdgpu_mes_reg_wait(struct amdgpu_device *adev, uint32_t reg, - uint32_t val, uint32_t mask); int amdgpu_mes_reg_write_reg_wait(struct amdgpu_device *adev, uint32_t reg0, uint32_t reg1, uint32_t ref, uint32_t mask); -- cgit v1.2.3 From 30837a49bd0aba0f311d4056cd48753955f60d40 Mon Sep 17 00:00:00 2001 From: Lang Yu Date: Fri, 23 May 2025 10:04:54 +0800 Subject: drm/amdkfd: Map wptr BO to GART unconditionally For simulation C models that don't run CP FW where adev->mes.sched_version is not populated correctly. This causes NULL dereference in amdgpu_amdkfd_free_gtt_mem(dev->adev, (void **)&pqn->q->wptr_bo_gart) and warning on unpinned BO in amdgpu_bo_gpu_offset(q->properties.wptr_bo). Compared with adding version check here and there, always map wptr BO to GART simplifies things. v2: Add NULL check in amdgpu_amdkfd_free_gtt_mem.(Philip) Signed-off-by: Lang Yu Reviewed-by: Felix Kuehling Signed-off-by: Alex Deucher --- drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd.c | 3 +++ .../gpu/drm/amd/amdkfd/kfd_process_queue_manager.c | 23 ++++++++++------------ 2 files changed, 13 insertions(+), 13 deletions(-) (limited to 'drivers/gpu') diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd.c index 4cec3a873995..d8ac4b1051a8 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd.c @@ -368,6 +368,9 @@ void amdgpu_amdkfd_free_gtt_mem(struct amdgpu_device *adev, void **mem_obj) { struct amdgpu_bo **bo = (struct amdgpu_bo **) mem_obj; + if (!bo || !*bo) + return; + (void)amdgpu_bo_reserve(*bo, true); amdgpu_bo_kunmap(*bo); amdgpu_bo_unpin(*bo); diff --git a/drivers/gpu/drm/amd/amdkfd/kfd_process_queue_manager.c b/drivers/gpu/drm/amd/amdkfd/kfd_process_queue_manager.c index 6d5fa57d4a23..c643e0ccec52 100644 --- a/drivers/gpu/drm/amd/amdkfd/kfd_process_queue_manager.c +++ b/drivers/gpu/drm/amd/amdkfd/kfd_process_queue_manager.c @@ -279,20 +279,17 @@ static int init_user_queue(struct process_queue_manager *pqm, /* Starting with GFX11, wptr BOs must be mapped to GART for MES to determine work * on unmapped queues for usermode queue oversubscription (no aggregated doorbell) */ - if (((dev->adev->mes.sched_version & AMDGPU_MES_API_VERSION_MASK) - >> AMDGPU_MES_API_VERSION_SHIFT) >= 2) { - if (dev->adev != amdgpu_ttm_adev(q_properties->wptr_bo->tbo.bdev)) { - pr_err("Queue memory allocated to wrong device\n"); - retval = -EINVAL; - goto free_gang_ctx_bo; - } + if (dev->adev != amdgpu_ttm_adev(q_properties->wptr_bo->tbo.bdev)) { + pr_err("Queue memory allocated to wrong device\n"); + retval = -EINVAL; + goto free_gang_ctx_bo; + } - retval = amdgpu_amdkfd_map_gtt_bo_to_gart(q_properties->wptr_bo, - &(*q)->wptr_bo_gart); - if (retval) { - pr_err("Failed to map wptr bo to GART\n"); - goto free_gang_ctx_bo; - } + retval = amdgpu_amdkfd_map_gtt_bo_to_gart(q_properties->wptr_bo, + &(*q)->wptr_bo_gart); + if (retval) { + pr_err("Failed to map wptr bo to GART\n"); + goto free_gang_ctx_bo; } } -- cgit v1.2.3 From c557fd1050f6691dde36818dfc1a4c415c42901b Mon Sep 17 00:00:00 2001 From: Jesus Narvaez Date: Wed, 14 May 2025 15:52:24 -0700 Subject: drm/i915/guc: Check if expecting reply before decrementing outstanding_submission_g2h When sending a H2G message where a reply is expected in guc_submission_send_busy_loop(), outstanding_submission_g2h is incremented before the send. However, if there is an error sending the message, outstanding_submission_g2h is decremented without checking if a reply is expected. Therefore, check if reply is expected when there is a failure before decrementing outstanding_submission_g2h. Fixes: 2f2cc53b5fe7 ("drm/i915/guc: Close deregister-context race against CT-loss") Signed-off-by: Jesus Narvaez Cc: Daniele Ceraolo Spurio Cc: Alan Previn Cc: Anshuman Gupta Cc: Mousumi Jana Cc: Rodrigo Vivi Cc: Matt Roper Reviewed-by: Daniele Ceraolo Spurio Signed-off-by: John Harrison Link: https://lore.kernel.org/r/20250514225224.4142684-1-jesus.narvaez@intel.com (cherry picked from commit a6a26786f22a4ab0227bcf610510c4c9c2df0808) Signed-off-by: Joonas Lahtinen --- drivers/gpu/drm/i915/gt/uc/intel_guc_submission.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) (limited to 'drivers/gpu') diff --git a/drivers/gpu/drm/i915/gt/uc/intel_guc_submission.c b/drivers/gpu/drm/i915/gt/uc/intel_guc_submission.c index f8cb7c630d5b..108331a69995 100644 --- a/drivers/gpu/drm/i915/gt/uc/intel_guc_submission.c +++ b/drivers/gpu/drm/i915/gt/uc/intel_guc_submission.c @@ -633,7 +633,7 @@ static int guc_submission_send_busy_loop(struct intel_guc *guc, atomic_inc(&guc->outstanding_submission_g2h); ret = intel_guc_send_busy_loop(guc, action, len, g2h_len_dw, loop); - if (ret) + if (ret && g2h_len_dw) atomic_dec(&guc->outstanding_submission_g2h); return ret; -- cgit v1.2.3 From 57d63c6cd0851d3af612a556ec61b0f2a9bd522f Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Jouni=20H=C3=B6gander?= Date: Mon, 26 May 2025 15:05:11 +0300 Subject: drm/i915/psr: Fix using wrong mask in REG_FIELD_PREP MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit Wrong mask is used in PORT_ALPM_LFPS_CTL_FIRST_LFPS_HALF_CYCLE_DURATION and PORT_ALPM_LFPS_CTL_LAST_LFPS_HALF_CYCLE_DURATION. Fixes: 295099580f04 ("drm/i915/psr: Add missing ALPM AUX-Less register definitions") Signed-off-by: Jouni Högander Reviewed-by: Ankit Nautiyal Link: https://lore.kernel.org/r/20250526120512.1702815-12-jouni.hogander@intel.com (cherry picked from commit 8097128a40ff378761034ec72cdbf6f46e466dc0) Signed-off-by: Joonas Lahtinen --- drivers/gpu/drm/i915/display/intel_psr_regs.h | 4 ++-- 1 file changed, 2 insertions(+), 2 deletions(-) (limited to 'drivers/gpu') diff --git a/drivers/gpu/drm/i915/display/intel_psr_regs.h b/drivers/gpu/drm/i915/display/intel_psr_regs.h index 795e6b9cc575..248136456048 100644 --- a/drivers/gpu/drm/i915/display/intel_psr_regs.h +++ b/drivers/gpu/drm/i915/display/intel_psr_regs.h @@ -325,8 +325,8 @@ #define PORT_ALPM_LFPS_CTL_LFPS_HALF_CYCLE_DURATION_MASK REG_GENMASK(20, 16) #define PORT_ALPM_LFPS_CTL_LFPS_HALF_CYCLE_DURATION(val) REG_FIELD_PREP(PORT_ALPM_LFPS_CTL_LFPS_HALF_CYCLE_DURATION_MASK, val) #define PORT_ALPM_LFPS_CTL_FIRST_LFPS_HALF_CYCLE_DURATION_MASK REG_GENMASK(12, 8) -#define PORT_ALPM_LFPS_CTL_FIRST_LFPS_HALF_CYCLE_DURATION(val) REG_FIELD_PREP(PORT_ALPM_LFPS_CTL_LFPS_HALF_CYCLE_DURATION_MASK, val) +#define PORT_ALPM_LFPS_CTL_FIRST_LFPS_HALF_CYCLE_DURATION(val) REG_FIELD_PREP(PORT_ALPM_LFPS_CTL_FIRST_LFPS_HALF_CYCLE_DURATION_MASK, val) #define PORT_ALPM_LFPS_CTL_LAST_LFPS_HALF_CYCLE_DURATION_MASK REG_GENMASK(4, 0) -#define PORT_ALPM_LFPS_CTL_LAST_LFPS_HALF_CYCLE_DURATION(val) REG_FIELD_PREP(PORT_ALPM_LFPS_CTL_LFPS_HALF_CYCLE_DURATION_MASK, val) +#define PORT_ALPM_LFPS_CTL_LAST_LFPS_HALF_CYCLE_DURATION(val) REG_FIELD_PREP(PORT_ALPM_LFPS_CTL_LAST_LFPS_HALF_CYCLE_DURATION_MASK, val) #endif /* __INTEL_PSR_REGS_H__ */ -- cgit v1.2.3 From 0323a5127e7c534cfc88efe0f850a0cb777e938b Mon Sep 17 00:00:00 2001 From: Jesus Narvaez Date: Wed, 28 May 2025 16:05:51 -0700 Subject: drm/i915/guc: Handle race condition where wakeref count drops below 0 There is a rare race condition when preparing for a reset where guc_lrc_desc_unpin() could be in the process of deregistering a context while a different thread is scrubbing outstanding contexts and it alters the context state and does a wakeref put. Then, if there is a failure with deregister_context(), a second wakeref put could occur. As a result the wakeref count could drop below 0 and fail an INTEL_WAKEREF_BUG_ON() check. Therefore if there is a failure with deregister_context(), undo the context state changes and do a wakeref put only if the context was set to be destroyed earlier. v2: Expand comment to better explain change. (Daniele) v3: Removed addition to the original comment. (Daniele) Fixes: 2f2cc53b5fe7 ("drm/i915/guc: Close deregister-context race against CT-loss") Signed-off-by: Jesus Narvaez Cc: Daniele Ceraolo Spurio Cc: Alan Previn Cc: Anshuman Gupta Cc: Mousumi Jana Cc: Rodrigo Vivi Cc: Matt Roper Reviewed-by: Daniele Ceraolo Spurio Signed-off-by: John Harrison Link: https://lore.kernel.org/r/20250528230551.1855177-1-jesus.narvaez@intel.com (cherry picked from commit f36a75aba1c3176d177964bca76f86a075d2943a) Signed-off-by: Joonas Lahtinen --- drivers/gpu/drm/i915/gt/uc/intel_guc_submission.c | 17 ++++++++++++++--- 1 file changed, 14 insertions(+), 3 deletions(-) (limited to 'drivers/gpu') diff --git a/drivers/gpu/drm/i915/gt/uc/intel_guc_submission.c b/drivers/gpu/drm/i915/gt/uc/intel_guc_submission.c index 108331a69995..127316d2c8aa 100644 --- a/drivers/gpu/drm/i915/gt/uc/intel_guc_submission.c +++ b/drivers/gpu/drm/i915/gt/uc/intel_guc_submission.c @@ -3443,18 +3443,29 @@ static inline int guc_lrc_desc_unpin(struct intel_context *ce) * GuC is active, lets destroy this context, but at this point we can still be racing * with suspend, so we undo everything if the H2G fails in deregister_context so * that GuC reset will find this context during clean up. + * + * There is a race condition where the reset code could have altered + * this context's state and done a wakeref put before we try to + * deregister it here. So check if the context is still set to be + * destroyed before undoing earlier changes, to avoid two wakeref puts + * on the same context. */ ret = deregister_context(ce, ce->guc_id.id); if (ret) { + bool pending_destroyed; spin_lock_irqsave(&ce->guc_state.lock, flags); - set_context_registered(ce); - clr_context_destroyed(ce); + pending_destroyed = context_destroyed(ce); + if (pending_destroyed) { + set_context_registered(ce); + clr_context_destroyed(ce); + } spin_unlock_irqrestore(&ce->guc_state.lock, flags); /* * As gt-pm is awake at function entry, intel_wakeref_put_async merely decrements * the wakeref immediately but per function spec usage call this after unlock. */ - intel_wakeref_put_async(>->wakeref); + if (pending_destroyed) + intel_wakeref_put_async(>->wakeref); } return ret; -- cgit v1.2.3 From 4e696906e9a82d4cab75f3083fabd65433c77e20 Mon Sep 17 00:00:00 2001 From: Xuemei Liu Date: Thu, 29 May 2025 10:25:11 +0800 Subject: drm/amdkfd: enable kfd on RISCV systems KFD has been confirmed that can run on RISCV systems. It's necessary to support CONFIG_HSA_AMD on RISCV. Signed-off-by: Xuemei Liu Signed-off-by: Felix Kuehling Reviewed-by: Felix Kuehling Signed-off-by: Alex Deucher --- drivers/gpu/drm/amd/amdkfd/Kconfig | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) (limited to 'drivers/gpu') diff --git a/drivers/gpu/drm/amd/amdkfd/Kconfig b/drivers/gpu/drm/amd/amdkfd/Kconfig index d3c3d3ab7225..62e88e5362e9 100644 --- a/drivers/gpu/drm/amd/amdkfd/Kconfig +++ b/drivers/gpu/drm/amd/amdkfd/Kconfig @@ -5,7 +5,7 @@ config HSA_AMD bool "HSA kernel driver for AMD GPU devices" - depends on DRM_AMDGPU && (X86_64 || ARM64 || PPC64) + depends on DRM_AMDGPU && (X86_64 || ARM64 || PPC64 || (RISCV && 64BIT)) select HMM_MIRROR select MMU_NOTIFIER select DRM_AMDGPU_USERPTR -- cgit v1.2.3 From 719d84f8a812608fc0f7be18a96d7dee96eaf3ba Mon Sep 17 00:00:00 2001 From: Lijo Lazar Date: Thu, 29 May 2025 15:57:44 +0530 Subject: drm/amdgpu: Add more checks to discovery fetch Add more checks for valid vram size and log error, if any. Signed-off-by: Lijo Lazar Reviewed-by: Alex Deucher Signed-off-by: Alex Deucher --- drivers/gpu/drm/amd/amdgpu/amdgpu_discovery.c | 16 +++++++++++++--- 1 file changed, 13 insertions(+), 3 deletions(-) (limited to 'drivers/gpu') diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_discovery.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_discovery.c index 9e738fae2b74..a0e9bf9b2710 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_discovery.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_discovery.c @@ -270,9 +270,10 @@ static int amdgpu_discovery_read_binary_from_sysmem(struct amdgpu_device *adev, static int amdgpu_discovery_read_binary_from_mem(struct amdgpu_device *adev, uint8_t *binary) { + bool sz_valid = true; uint64_t vram_size; - u32 msg; int i, ret = 0; + u32 msg; if (!amdgpu_sriov_vf(adev)) { /* It can take up to a second for IFWI init to complete on some dGPUs, @@ -291,9 +292,13 @@ static int amdgpu_discovery_read_binary_from_mem(struct amdgpu_device *adev, } } - vram_size = (uint64_t)RREG32(mmRCC_CONFIG_MEMSIZE) << 20; + vram_size = RREG32(mmRCC_CONFIG_MEMSIZE); + if (!vram_size || vram_size == U32_MAX) + sz_valid = false; + else + vram_size <<= 20; - if (vram_size) { + if (sz_valid) { uint64_t pos = vram_size - DISCOVERY_TMR_OFFSET; amdgpu_device_vram_access(adev, pos, (uint32_t *)binary, adev->mman.discovery_tmr_size, false); @@ -301,6 +306,11 @@ static int amdgpu_discovery_read_binary_from_mem(struct amdgpu_device *adev, ret = amdgpu_discovery_read_binary_from_sysmem(adev, binary); } + if (ret) + dev_err(adev->dev, + "failed to read discovery info from memory, vram size read: %llx", + vram_size); + return ret; } -- cgit v1.2.3 From d26625d034fb8d596f0488472969493fa02d03f3 Mon Sep 17 00:00:00 2001 From: Vitaly Prosyak Date: Tue, 6 May 2025 16:45:33 -0400 Subject: drm/amdgpu/gfx10: Refine Cleaner Shader for GFX10.1.10 MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit This patch updates the cleaner shader, which is responsible for initializing GPU resources such as Local Data Share (LDS), Vector General Purpose Registers (VGPRs), and Scalar General Purpose Registers (SGPRs). Changes include adjustments to register clearing and shader configuration. - Updated GPU resource initialization addresses in the cleaner shader from `be803080` to `be803000`. - Simplified the logic in the SGPR clearing section, ensuring all SGPRs are set to zero. Fixes: 25961bad9212 ("drm/amdgpu/gfx10: Add cleaner shader for GFX10.1.10") Cc: Christian König Cc: Alex Deucher Signed-off-by: Manu Rastogi Signed-off-by: Vitaly Prosyak Signed-off-by: Srinivasan Shanmugam Acked-by: Alex Deucher Signed-off-by: Alex Deucher --- drivers/gpu/drm/amd/amdgpu/gfx_v10_0_cleaner_shader.h | 6 +++--- drivers/gpu/drm/amd/amdgpu/gfx_v10_1_10_cleaner_shader.asm | 13 ++++++------- 2 files changed, 9 insertions(+), 10 deletions(-) (limited to 'drivers/gpu') diff --git a/drivers/gpu/drm/amd/amdgpu/gfx_v10_0_cleaner_shader.h b/drivers/gpu/drm/amd/amdgpu/gfx_v10_0_cleaner_shader.h index 5255378af53c..f67569ccf9f6 100644 --- a/drivers/gpu/drm/amd/amdgpu/gfx_v10_0_cleaner_shader.h +++ b/drivers/gpu/drm/amd/amdgpu/gfx_v10_0_cleaner_shader.h @@ -43,9 +43,9 @@ static const u32 gfx_10_1_10_cleaner_shader_hex[] = { 0xd70f6a01, 0x000202ff, 0x00000400, 0x80828102, 0xbf84fff7, 0xbefc03ff, - 0x00000068, 0xbe803080, - 0xbe813080, 0xbe823080, - 0xbe833080, 0x80fc847c, + 0x00000068, 0xbe803000, + 0xbe813000, 0xbe823000, + 0xbe833000, 0x80fc847c, 0xbf84fffa, 0xbeea0480, 0xbeec0480, 0xbeee0480, 0xbef00480, 0xbef20480, diff --git a/drivers/gpu/drm/amd/amdgpu/gfx_v10_1_10_cleaner_shader.asm b/drivers/gpu/drm/amd/amdgpu/gfx_v10_1_10_cleaner_shader.asm index 9ba3359253c9..54f7ed9e2801 100644 --- a/drivers/gpu/drm/amd/amdgpu/gfx_v10_1_10_cleaner_shader.asm +++ b/drivers/gpu/drm/amd/amdgpu/gfx_v10_1_10_cleaner_shader.asm @@ -40,7 +40,6 @@ shader main type(CS) wave_size(32) // Note: original source code from SQ team - // // Create 32 waves in a threadgroup (CS waves) // Each allocates 64 VGPRs @@ -71,8 +70,8 @@ label_0005: s_sub_u32 s2, s2, 8 s_cbranch_scc0 label_0005 // - s_mov_b32 s2, 0x80000000 // Bit31 is first_wave - s_and_b32 s2, s2, s0 // sgpr0 has tg_size (first_wave) term as in ucode only COMPUTE_PGM_RSRC2.tg_size_en is set + s_mov_b32 s2, 0x80000000 // Bit31 is first_wave + s_and_b32 s2, s2, s1 // sgpr0 has tg_size (first_wave) term as in ucode only COMPUTE_PGM_RSRC2.tg_size_en is set s_cbranch_scc0 label_0023 // Clean LDS if its first wave of ThreadGroup/WorkGroup // CLEAR LDS // @@ -99,10 +98,10 @@ label_001F: label_0023: s_mov_b32 m0, 0x00000068 // Loop 108/4=27 times (loop unrolled for performance) label_sgpr_loop: - s_movreld_b32 s0, 0 - s_movreld_b32 s1, 0 - s_movreld_b32 s2, 0 - s_movreld_b32 s3, 0 + s_movreld_b32 s0, s0 + s_movreld_b32 s1, s0 + s_movreld_b32 s2, s0 + s_movreld_b32 s3, s0 s_sub_u32 m0, m0, 4 s_cbranch_scc0 label_sgpr_loop -- cgit v1.2.3 From 5cccf10f652122a17b40df9d672ccf2ed69cd82f Mon Sep 17 00:00:00 2001 From: Alex Deucher Date: Tue, 27 May 2025 10:13:31 -0400 Subject: drm/amdgpu: disable workload profile switching when OD is enabled Users have reported that they have to reduce the level of undervolting to acheive stability when dynamic workload profiles are enabled on GC 10.3.x. Disable dynamic workload profiles if the user has enabled OD. Fixes: b9467983b774 ("drm/amdgpu: add dynamic workload profile switching for gfx10") Closes: https://gitlab.freedesktop.org/drm/amd/-/issues/4262 Reviewed-by: Kenneth Feng Signed-off-by: Alex Deucher Cc: stable@vger.kernel.org # 6.15.x --- drivers/gpu/drm/amd/amdgpu/amdgpu_gfx.c | 8 ++++++++ drivers/gpu/drm/amd/pm/amdgpu_dpm.c | 22 ++++++++++++++++++++++ drivers/gpu/drm/amd/pm/inc/amdgpu_dpm.h | 1 + 3 files changed, 31 insertions(+) (limited to 'drivers/gpu') diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_gfx.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_gfx.c index 1db1e6ec0184..c5646af055ab 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_gfx.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_gfx.c @@ -2228,6 +2228,9 @@ void amdgpu_gfx_profile_ring_begin_use(struct amdgpu_ring *ring) enum PP_SMC_POWER_PROFILE profile; int r; + if (amdgpu_dpm_is_overdrive_enabled(adev)) + return; + if (adev->gfx.num_gfx_rings) profile = PP_SMC_POWER_PROFILE_FULLSCREEN3D; else @@ -2258,6 +2261,11 @@ void amdgpu_gfx_profile_ring_begin_use(struct amdgpu_ring *ring) void amdgpu_gfx_profile_ring_end_use(struct amdgpu_ring *ring) { + struct amdgpu_device *adev = ring->adev; + + if (amdgpu_dpm_is_overdrive_enabled(adev)) + return; + atomic_dec(&ring->adev->gfx.total_submission_cnt); schedule_delayed_work(&ring->adev->gfx.idle_work, GFX_PROFILE_IDLE_TIMEOUT); diff --git a/drivers/gpu/drm/amd/pm/amdgpu_dpm.c b/drivers/gpu/drm/amd/pm/amdgpu_dpm.c index d98c95d1ed83..5c1cbdc122d2 100644 --- a/drivers/gpu/drm/amd/pm/amdgpu_dpm.c +++ b/drivers/gpu/drm/amd/pm/amdgpu_dpm.c @@ -1697,6 +1697,28 @@ int amdgpu_dpm_is_overdrive_supported(struct amdgpu_device *adev) } } +int amdgpu_dpm_is_overdrive_enabled(struct amdgpu_device *adev) +{ + if (is_support_sw_smu(adev)) { + struct smu_context *smu = adev->powerplay.pp_handle; + + return smu->od_enabled; + } else { + struct pp_hwmgr *hwmgr; + + /* + * dpm on some legacy asics don't carry od_enabled member + * as its pp_handle is casted directly from adev. + */ + if (amdgpu_dpm_is_legacy_dpm(adev)) + return false; + + hwmgr = (struct pp_hwmgr *)adev->powerplay.pp_handle; + + return hwmgr->od_enabled; + } +} + int amdgpu_dpm_set_pp_table(struct amdgpu_device *adev, const char *buf, size_t size) diff --git a/drivers/gpu/drm/amd/pm/inc/amdgpu_dpm.h b/drivers/gpu/drm/amd/pm/inc/amdgpu_dpm.h index c0f9ecb97fcc..768317ee1486 100644 --- a/drivers/gpu/drm/amd/pm/inc/amdgpu_dpm.h +++ b/drivers/gpu/drm/amd/pm/inc/amdgpu_dpm.h @@ -563,6 +563,7 @@ int amdgpu_dpm_get_smu_prv_buf_details(struct amdgpu_device *adev, void **addr, size_t *size); int amdgpu_dpm_is_overdrive_supported(struct amdgpu_device *adev); +int amdgpu_dpm_is_overdrive_enabled(struct amdgpu_device *adev); int amdgpu_dpm_set_pp_table(struct amdgpu_device *adev, const char *buf, size_t size); -- cgit v1.2.3 From 98a46a408998102af5c45adce0871acd7967bb59 Mon Sep 17 00:00:00 2001 From: Dan Carpenter Date: Fri, 23 May 2025 19:05:58 +0300 Subject: drm/amdgpu: Fix integer overflow issues in amdgpu_userq_fence.c This patch only affects 32bit systems. There are several integer overflows bugs here but only the "sizeof(u32) * num_syncobj" multiplication is a problem at runtime. (The last lines of this patch). These variables are u32 variables that come from the user. The issue is the multiplications can overflow leading to us allocating a smaller buffer than intended. For the first couple integer overflows, the syncobj_handles = memdup_user() allocation is immediately followed by a kmalloc_array(): syncobj = kmalloc_array(num_syncobj_handles, sizeof(*syncobj), GFP_KERNEL); In that situation the kmalloc_array() works as a bounds check and we haven't accessed the syncobj_handlesp[] array yet so the integer overflow is harmless. But the "num_syncobj" multiplication doesn't have that and the integer overflow could lead to an out of bounds access. Fixes: a292fdecd728 ("drm/amdgpu: Implement userqueue signal/wait IOCTL") Signed-off-by: Dan Carpenter Signed-off-by: Alex Deucher --- drivers/gpu/drm/amd/amdgpu/amdgpu_userq_fence.c | 8 ++++---- 1 file changed, 4 insertions(+), 4 deletions(-) (limited to 'drivers/gpu') diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_userq_fence.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_userq_fence.c index fc4d0d42e223..a86616c6deef 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_userq_fence.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_userq_fence.c @@ -430,7 +430,7 @@ int amdgpu_userq_signal_ioctl(struct drm_device *dev, void *data, num_syncobj_handles = args->num_syncobj_handles; syncobj_handles = memdup_user(u64_to_user_ptr(args->syncobj_handles), - sizeof(u32) * num_syncobj_handles); + size_mul(sizeof(u32), num_syncobj_handles)); if (IS_ERR(syncobj_handles)) return PTR_ERR(syncobj_handles); @@ -612,13 +612,13 @@ int amdgpu_userq_wait_ioctl(struct drm_device *dev, void *data, num_read_bo_handles = wait_info->num_bo_read_handles; bo_handles_read = memdup_user(u64_to_user_ptr(wait_info->bo_read_handles), - sizeof(u32) * num_read_bo_handles); + size_mul(sizeof(u32), num_read_bo_handles)); if (IS_ERR(bo_handles_read)) return PTR_ERR(bo_handles_read); num_write_bo_handles = wait_info->num_bo_write_handles; bo_handles_write = memdup_user(u64_to_user_ptr(wait_info->bo_write_handles), - sizeof(u32) * num_write_bo_handles); + size_mul(sizeof(u32), num_write_bo_handles)); if (IS_ERR(bo_handles_write)) { r = PTR_ERR(bo_handles_write); goto free_bo_handles_read; @@ -626,7 +626,7 @@ int amdgpu_userq_wait_ioctl(struct drm_device *dev, void *data, num_syncobj = wait_info->num_syncobj_handles; syncobj_handles = memdup_user(u64_to_user_ptr(wait_info->syncobj_handles), - sizeof(u32) * num_syncobj); + size_mul(sizeof(u32), num_syncobj)); if (IS_ERR(syncobj_handles)) { r = PTR_ERR(syncobj_handles); goto free_bo_handles_write; -- cgit v1.2.3 From 335f1e797c32cbe6f313805125526b35d29280b0 Mon Sep 17 00:00:00 2001 From: Dan Carpenter Date: Fri, 23 May 2025 19:09:52 +0300 Subject: drm/amdgpu: Fix integer overflow in amdgpu_gem_add_input_fence() The "num_syncobj_handles" is a u32 value that comes from the user via the ioctl. On 32bit systems the "sizeof(uint32_t) * num_syncobj_handles" multiplication can have an integer overflow. Use size_mul() to fix that. Fixes: 38c67ec9aa4b ("drm/amdgpu: Add input fence to sync bo map/unmap") Signed-off-by: Dan Carpenter Signed-off-by: Alex Deucher --- drivers/gpu/drm/amd/amdgpu/amdgpu_gem.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) (limited to 'drivers/gpu') diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_gem.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_gem.c index 2c68118fe9fd..0ecc88df7208 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_gem.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_gem.c @@ -58,7 +58,7 @@ amdgpu_gem_add_input_fence(struct drm_file *filp, return 0; syncobj_handles = memdup_user(u64_to_user_ptr(syncobj_handles_array), - sizeof(uint32_t) * num_syncobj_handles); + size_mul(sizeof(uint32_t), num_syncobj_handles)); if (IS_ERR(syncobj_handles)) return PTR_ERR(syncobj_handles); -- cgit v1.2.3 From e34bcf1594b59f9f63c084bf0646b19edf581adc Mon Sep 17 00:00:00 2001 From: Arunpravin Paneer Selvam Date: Tue, 27 May 2025 19:13:20 +0530 Subject: drm/amdgpu: Add userq fence support to SDMAv7.0 MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit - Add userq fence support to SDMAv7.0. - GFX12's user fence irq src id differs from GFX11's, hence we need create a new irq srcid header file for GFX12. User fence irq src id information- GFX11 and SDMA6.0 - 0x43 GFX12 and SDMA7.0 - 0x46 Signed-off-by: Arunpravin Paneer Selvam Acked-by: Christian König Signed-off-by: Alex Deucher --- drivers/gpu/drm/amd/amdgpu/gfx_v12_0.c | 10 +-- drivers/gpu/drm/amd/amdgpu/sdma_v7_0.c | 59 +++++++++++------ .../amd/include/ivsrcid/gfx/irqsrcs_gfx_12_0_0.h | 74 ++++++++++++++++++++++ 3 files changed, 120 insertions(+), 23 deletions(-) create mode 100644 drivers/gpu/drm/amd/include/ivsrcid/gfx/irqsrcs_gfx_12_0_0.h (limited to 'drivers/gpu') diff --git a/drivers/gpu/drm/amd/amdgpu/gfx_v12_0.c b/drivers/gpu/drm/amd/amdgpu/gfx_v12_0.c index f09d96bfee16..1234c8d64e20 100644 --- a/drivers/gpu/drm/amd/amdgpu/gfx_v12_0.c +++ b/drivers/gpu/drm/amd/amdgpu/gfx_v12_0.c @@ -36,7 +36,7 @@ #include "gc/gc_12_0_0_offset.h" #include "gc/gc_12_0_0_sh_mask.h" #include "soc24_enum.h" -#include "ivsrcid/gfx/irqsrcs_gfx_11_0_0.h" +#include "ivsrcid/gfx/irqsrcs_gfx_12_0_0.h" #include "soc15.h" #include "clearstate_gfx12.h" @@ -1453,28 +1453,28 @@ static int gfx_v12_0_sw_init(struct amdgpu_ip_block *ip_block) /* EOP Event */ r = amdgpu_irq_add_id(adev, SOC21_IH_CLIENTID_GRBM_CP, - GFX_11_0_0__SRCID__CP_EOP_INTERRUPT, + GFX_12_0_0__SRCID__CP_EOP_INTERRUPT, &adev->gfx.eop_irq); if (r) return r; /* Bad opcode Event */ r = amdgpu_irq_add_id(adev, SOC21_IH_CLIENTID_GRBM_CP, - GFX_11_0_0__SRCID__CP_BAD_OPCODE_ERROR, + GFX_12_0_0__SRCID__CP_BAD_OPCODE_ERROR, &adev->gfx.bad_op_irq); if (r) return r; /* Privileged reg */ r = amdgpu_irq_add_id(adev, SOC21_IH_CLIENTID_GRBM_CP, - GFX_11_0_0__SRCID__CP_PRIV_REG_FAULT, + GFX_12_0_0__SRCID__CP_PRIV_REG_FAULT, &adev->gfx.priv_reg_irq); if (r) return r; /* Privileged inst */ r = amdgpu_irq_add_id(adev, SOC21_IH_CLIENTID_GRBM_CP, - GFX_11_0_0__SRCID__CP_PRIV_INSTR_FAULT, + GFX_12_0_0__SRCID__CP_PRIV_INSTR_FAULT, &adev->gfx.priv_inst_irq); if (r) return r; diff --git a/drivers/gpu/drm/amd/amdgpu/sdma_v7_0.c b/drivers/gpu/drm/amd/amdgpu/sdma_v7_0.c index befe013b11a7..ad47d0bdf777 100644 --- a/drivers/gpu/drm/amd/amdgpu/sdma_v7_0.c +++ b/drivers/gpu/drm/amd/amdgpu/sdma_v7_0.c @@ -33,7 +33,7 @@ #include "gc/gc_12_0_0_offset.h" #include "gc/gc_12_0_0_sh_mask.h" #include "hdp/hdp_6_0_0_offset.h" -#include "ivsrcid/gfx/irqsrcs_gfx_11_0_0.h" +#include "ivsrcid/gfx/irqsrcs_gfx_12_0_0.h" #include "soc15_common.h" #include "soc15.h" @@ -43,6 +43,7 @@ #include "sdma_v7_0.h" #include "v12_structs.h" #include "mes_userqueue.h" +#include "amdgpu_userq_fence.h" MODULE_FIRMWARE("amdgpu/sdma_7_0_0.bin"); MODULE_FIRMWARE("amdgpu/sdma_7_0_1.bin"); @@ -910,6 +911,9 @@ static int sdma_v7_0_mqd_init(struct amdgpu_device *adev, void *mqd, m->sdmax_rlcx_csa_addr_lo = lower_32_bits(prop->csa_addr); m->sdmax_rlcx_csa_addr_hi = upper_32_bits(prop->csa_addr); + m->sdmax_rlcx_mcu_dbg0 = lower_32_bits(prop->fence_address); + m->sdmax_rlcx_mcu_dbg1 = upper_32_bits(prop->fence_address); + return 0; } @@ -1296,11 +1300,18 @@ static int sdma_v7_0_sw_init(struct amdgpu_ip_block *ip_block) /* SDMA trap event */ r = amdgpu_irq_add_id(adev, SOC21_IH_CLIENTID_GFX, - GFX_11_0_0__SRCID__SDMA_TRAP, + GFX_12_0_0__SRCID__SDMA_TRAP, &adev->sdma.trap_irq); if (r) return r; + /* SDMA user fence event */ + r = amdgpu_irq_add_id(adev, SOC21_IH_CLIENTID_GFX, + GFX_12_0_0__SRCID__SDMA_FENCE, + &adev->sdma.fence_irq); + if (r) + return r; + for (i = 0; i < adev->sdma.num_instances; i++) { ring = &adev->sdma.instance[i].ring; ring->ring_obj = NULL; @@ -1526,25 +1537,9 @@ static int sdma_v7_0_process_trap_irq(struct amdgpu_device *adev, struct amdgpu_iv_entry *entry) { int instances, queue; - uint32_t mes_queue_id = entry->src_data[0]; DRM_DEBUG("IH: SDMA trap\n"); - if (adev->enable_mes && (mes_queue_id & AMDGPU_FENCE_MES_QUEUE_FLAG)) { - struct amdgpu_mes_queue *queue; - - mes_queue_id &= AMDGPU_FENCE_MES_QUEUE_ID_MASK; - - spin_lock(&adev->mes.queue_id_lock); - queue = idr_find(&adev->mes.queue_id_idr, mes_queue_id); - if (queue) { - DRM_DEBUG("process smda queue id = %d\n", mes_queue_id); - amdgpu_fence_process(queue->ring); - } - spin_unlock(&adev->mes.queue_id_lock); - return 0; - } - queue = entry->ring_id & 0xf; instances = (entry->ring_id & 0xf0) >> 4; if (instances > 1) { @@ -1566,6 +1561,29 @@ static int sdma_v7_0_process_trap_irq(struct amdgpu_device *adev, return 0; } +static int sdma_v7_0_process_fence_irq(struct amdgpu_device *adev, + struct amdgpu_irq_src *source, + struct amdgpu_iv_entry *entry) +{ + u32 doorbell_offset = entry->src_data[0]; + + if (adev->enable_mes && doorbell_offset) { + struct amdgpu_userq_fence_driver *fence_drv = NULL; + struct xarray *xa = &adev->userq_xa; + unsigned long flags; + + doorbell_offset >>= SDMA0_QUEUE0_DOORBELL_OFFSET__OFFSET__SHIFT; + + xa_lock_irqsave(xa, flags); + fence_drv = xa_load(xa, doorbell_offset); + if (fence_drv) + amdgpu_userq_fence_driver_process(fence_drv); + xa_unlock_irqrestore(xa, flags); + } + + return 0; +} + static int sdma_v7_0_process_illegal_inst_irq(struct amdgpu_device *adev, struct amdgpu_irq_src *source, struct amdgpu_iv_entry *entry) @@ -1703,6 +1721,10 @@ static const struct amdgpu_irq_src_funcs sdma_v7_0_trap_irq_funcs = { .process = sdma_v7_0_process_trap_irq, }; +static const struct amdgpu_irq_src_funcs sdma_v7_0_fence_irq_funcs = { + .process = sdma_v7_0_process_fence_irq, +}; + static const struct amdgpu_irq_src_funcs sdma_v7_0_illegal_inst_irq_funcs = { .process = sdma_v7_0_process_illegal_inst_irq, }; @@ -1712,6 +1734,7 @@ static void sdma_v7_0_set_irq_funcs(struct amdgpu_device *adev) adev->sdma.trap_irq.num_types = AMDGPU_SDMA_IRQ_INSTANCE0 + adev->sdma.num_instances; adev->sdma.trap_irq.funcs = &sdma_v7_0_trap_irq_funcs; + adev->sdma.fence_irq.funcs = &sdma_v7_0_fence_irq_funcs; adev->sdma.illegal_inst_irq.funcs = &sdma_v7_0_illegal_inst_irq_funcs; } diff --git a/drivers/gpu/drm/amd/include/ivsrcid/gfx/irqsrcs_gfx_12_0_0.h b/drivers/gpu/drm/amd/include/ivsrcid/gfx/irqsrcs_gfx_12_0_0.h new file mode 100644 index 000000000000..467897ec2e65 --- /dev/null +++ b/drivers/gpu/drm/amd/include/ivsrcid/gfx/irqsrcs_gfx_12_0_0.h @@ -0,0 +1,74 @@ +/* SPDX-License-Identifier: MIT */ +/* + * Copyright 2024 Advanced Micro Devices, Inc. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * the rights to use, copy, modify, merge, publish, distribute, sublicense, + * and/or sell copies of the Software, and to permit persons to whom the + * Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice shall be included in + * all copies or substantial portions of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL + * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR + * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, + * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR + * OTHER DEALINGS IN THE SOFTWARE. + * + */ +#ifndef __IRQSRCS_GFX_12_0_0_H__ +#define __IRQSRCS_GFX_12_0_0_H__ + +#define GFX_12_0_0__SRCID__UTCL2_FAULT 0 // UTCL2 has encountered a fault or retry scenario +#define GFX_12_0_0__SRCID__UTCL2_DATA_POISONING 1 // UTCL2 for data poisoning +#define GFX_12_0_0__SRCID__MEM_ACCES_MON 10 // 0x0A EA memory access monitor interrupt +#define GFX_12_0_0__SRCID__SDMA_ATOMIC_RTN_DONE 48 // 0x30 SDMA atomic*_rtn ops complete +#define GFX_12_0_0__SRCID__SDMA_TRAP 49 // 0x31 Trap +#define GFX_12_0_0__SRCID__SDMA_SRBMWRITE 50 // 0x32 SRBM write Protection +#define GFX_12_0_0__SRCID__SDMA_CTXEMPTY 51 // 0x33 Context Empty +#define GFX_12_0_0__SRCID__SDMA_PREEMPT 52 // 0x34 SDMA New Run List +#define GFX_12_0_0__SRCID__SDMA_IB_PREEMPT 53 // 0x35 sdma mid - command buffer preempt interrupt +#define GFX_12_0_0__SRCID__SDMA_DOORBELL_INVALID 54 // 0x36 Doorbell BE invalid +#define GFX_12_0_0__SRCID__SDMA_QUEUE_HANG 55 // 0x37 Queue hang or Command timeout +#define GFX_12_0_0__SRCID__SDMA_ATOMIC_TIMEOUT 56 // 0x38 SDMA atomic CMPSWAP loop timeout +#define GFX_12_0_0__SRCID__SDMA_POLL_TIMEOUT 57 // 0x39 SRBM read poll timeout +#define GFX_12_0_0__SRCID__SDMA_PAGE_TIMEOUT 58 // 0x3A Page retry timeout after UTCL2 return nack = 1 +#define GFX_12_0_0__SRCID__SDMA_PAGE_NULL 59 // 0x3B Page Null from UTCL2 when nack = 2 +#define GFX_12_0_0__SRCID__SDMA_PAGE_FAULT 60 // 0x3C Page Fault Error from UTCL2 when nack = 3 +#define GFX_12_0_0__SRCID__SDMA_VM_HOLE 61 // 0x3D MC or SEM address in VM hole +#define GFX_12_0_0__SRCID__SDMA_ECC 62 // 0x3E ECC Error +#define GFX_12_0_0__SRCID__SDMA_FROZEN 63 // 0x3F SDMA Frozen +#define GFX_12_0_0__SRCID__SDMA_SRAM_ECC 64 // 0x40 SRAM ECC Error +#define GFX_12_0_0__SRCID__SDMA_SEM_INCOMPLETE_TIMEOUT 65 // 0x41 GPF(Sem incomplete timeout) +#define GFX_12_0_0__SRCID__SDMA_SEM_WAIT_FAIL_TIMEOUT 66 // 0x42 Semaphore wait fail timeout +#define GFX_12_0_0__SRCID__SDMA_FENCE 70 // 0x46 User fence +#define GFX_12_0_0__SRCID__RLC_GC_FED_INTERRUPT 128 // 0x80 FED Interrupt (for data poisoning) +#define GFX_12_0_0__SRCID__CP_GENERIC_INT 177 // 0xB1 CP_GENERIC int +#define GFX_12_0_0__SRCID__CP_PM4_PKT_RSVD_BIT_ERROR 180 // 0xB4 PM4 Pkt Rsvd Bits Error +#define GFX_12_0_0__SRCID__CP_EOP_INTERRUPT 181 // 0xB5 End-of-Pipe Interrupt +#define GFX_12_0_0__SRCID__CP_BAD_OPCODE_ERROR 183 // 0xB7 Bad Opcode Error +#define GFX_12_0_0__SRCID__CP_PRIV_REG_FAULT 184 // 0xB8 Privileged Register Fault +#define GFX_12_0_0__SRCID__CP_PRIV_INSTR_FAULT 185 // 0xB9 Privileged Instr Fault +#define GFX_12_0_0__SRCID__CP_WAIT_MEM_SEM_FAULT 186 // 0xBA Wait Memory Semaphore Fault (Sync Object Fault) +#define GFX_12_0_0__SRCID__CP_CTX_EMPTY_INTERRUPT 187 // 0xBB Context Empty Interrupt +#define GFX_12_0_0__SRCID__CP_CTX_BUSY_INTERRUPT 188 // 0xBC Context Busy Interrupt +#define GFX_12_0_0__SRCID__CP_ME_WAIT_REG_MEM_POLL_TIMEOUT 192 // 0xC0 CP.ME Wait_Reg_Mem Poll Timeout +#define GFX_12_0_0__SRCID__CP_SIG_INCOMPLETE 193 // 0xC1 "Surface Probe Fault Signal Incomplete" +#define GFX_12_0_0__SRCID__CP_PREEMPT_ACK 194 // 0xC2 Preemption Ack-wledge +#define GFX_12_0_0__SRCID__CP_GPF 195 // 0xC3 General Protection Fault (GPF) +#define GFX_12_0_0__SRCID__CP_GDS_ALLOC_ERROR 196 // 0xC4 GDS Alloc Error +#define GFX_12_0_0__SRCID__CP_ECC_ERROR 197 // 0xC5 ECC Error +#define GFX_12_0_0__SRCID__CP_COMPUTE_QUERY_STATUS 199 // 0xC7 Compute query status +#define GFX_12_0_0__SRCID__CP_VM_DOORBELL 200 // 0xC8 Unattached VM Doorbell Received +#define GFX_12_0_0__SRCID__CP_FUE_ERROR 201 // 0xC9 ECC FUE Error +#define GFX_12_0_0__SRCID__RLC_STRM_PERF_MONITOR_INTERRUPT 202 // 0xCA Streaming Perf Monitor Interrupt +#define GFX_12_0_0__SRCID__GRBM_RD_TIMEOUT_ERROR 232 // 0xE8 CRead timeout error +#define GFX_12_0_0__SRCID__GRBM_REG_GUI_IDLE 233 // 0xE9 Register GUI Idle +#define GFX_12_0_0__SRCID__SQ_INTERRUPT_ID 239 // 0xEF SQ Interrupt (ttrace wrap, errors) + +#endif -- cgit v1.2.3 From 893f07452bca56ff146a6be02b3294a9ea23d18a Mon Sep 17 00:00:00 2001 From: Zhongwei Zhang Date: Tue, 13 May 2025 16:45:59 +0800 Subject: drm/amd/display: Correct non-OLED pre_T11_delay. [Why] Only OLED panels require non-zero pre_T11_delay defaultly. Others should be controlled by power sequence. [How] For non OLED, pre_T11_delay delay in code should be zero. Also post_T7_delay. Cc: Mario Limonciello Cc: Alex Deucher Cc: stable@vger.kernel.org Reviewed-by: Charlene Liu Signed-off-by: Zhongwei Zhang Signed-off-by: Wayne Lin Tested-by: Daniel Wheeler Signed-off-by: Alex Deucher --- drivers/gpu/drm/amd/display/dc/hwss/dce110/dce110_hwseq.c | 7 ++++--- 1 file changed, 4 insertions(+), 3 deletions(-) (limited to 'drivers/gpu') diff --git a/drivers/gpu/drm/amd/display/dc/hwss/dce110/dce110_hwseq.c b/drivers/gpu/drm/amd/display/dc/hwss/dce110/dce110_hwseq.c index 23bec5d25ed6..b88b2d6b4e81 100644 --- a/drivers/gpu/drm/amd/display/dc/hwss/dce110/dce110_hwseq.c +++ b/drivers/gpu/drm/amd/display/dc/hwss/dce110/dce110_hwseq.c @@ -952,8 +952,8 @@ void dce110_edp_backlight_control( struct dc_context *ctx = link->ctx; struct bp_transmitter_control cntl = { 0 }; uint8_t pwrseq_instance = 0; - unsigned int pre_T11_delay = OLED_PRE_T11_DELAY; - unsigned int post_T7_delay = OLED_POST_T7_DELAY; + unsigned int pre_T11_delay = (link->dpcd_sink_ext_caps.bits.oled ? OLED_PRE_T11_DELAY : 0); + unsigned int post_T7_delay = (link->dpcd_sink_ext_caps.bits.oled ? OLED_POST_T7_DELAY : 0); if (dal_graphics_object_id_get_connector_id(link->link_enc->connector) != CONNECTOR_ID_EDP) { @@ -1069,7 +1069,8 @@ void dce110_edp_backlight_control( if (!enable) { /*follow oem panel config's requirement*/ pre_T11_delay += link->panel_config.pps.extra_pre_t11_ms; - msleep(pre_T11_delay); + if (pre_T11_delay) + msleep(pre_T11_delay); } } -- cgit v1.2.3 From 747bfca45e07b201cf80f3ba7338006f4525aeed Mon Sep 17 00:00:00 2001 From: Zhongwei Zhang Date: Fri, 16 May 2025 14:44:21 +0800 Subject: drm/amd/display: Avoid calling blank_stream() twice [Why] We've made fix for garbage in dcn31_reset_back_end_for_pipe(), adding blank_stream() before disable_crtc(). And set_dpms_off() will call blank_stream() again. [How] Add flag to avoid calling blank_stream() twice. Reviewed-by: Nicholas Kazlauskas Signed-off-by: Zhongwei Zhang Signed-off-by: Wayne Lin Tested-by: Daniel Wheeler Signed-off-by: Alex Deucher --- drivers/gpu/drm/amd/display/dc/hwss/dce110/dce110_hwseq.c | 3 +++ drivers/gpu/drm/amd/display/dc/hwss/dcn31/dcn31_hwseq.c | 11 +++++++++-- drivers/gpu/drm/amd/display/dc/hwss/hw_sequencer_private.h | 1 + 3 files changed, 13 insertions(+), 2 deletions(-) (limited to 'drivers/gpu') diff --git a/drivers/gpu/drm/amd/display/dc/hwss/dce110/dce110_hwseq.c b/drivers/gpu/drm/amd/display/dc/hwss/dce110/dce110_hwseq.c index b88b2d6b4e81..e8730cc40edb 100644 --- a/drivers/gpu/drm/amd/display/dc/hwss/dce110/dce110_hwseq.c +++ b/drivers/gpu/drm/amd/display/dc/hwss/dce110/dce110_hwseq.c @@ -1221,6 +1221,9 @@ void dce110_blank_stream(struct pipe_ctx *pipe_ctx) struct dc_link *link = stream->link; struct dce_hwseq *hws = link->dc->hwseq; + if (hws && hws->wa_state.skip_blank_stream) + return; + if (link->local_sink && link->local_sink->sink_signal == SIGNAL_TYPE_EDP) { if (!link->skip_implict_edp_power_control) hws->funcs.edp_backlight_control(link, false); diff --git a/drivers/gpu/drm/amd/display/dc/hwss/dcn31/dcn31_hwseq.c b/drivers/gpu/drm/amd/display/dc/hwss/dcn31/dcn31_hwseq.c index f38340aa3f15..5ba3999991b0 100644 --- a/drivers/gpu/drm/amd/display/dc/hwss/dcn31/dcn31_hwseq.c +++ b/drivers/gpu/drm/amd/display/dc/hwss/dcn31/dcn31_hwseq.c @@ -526,9 +526,15 @@ static void dcn31_reset_back_end_for_pipe( link = pipe_ctx->stream->link; + if (dc->hwseq) + dc->hwseq->wa_state.skip_blank_stream = false; + if ((!pipe_ctx->stream->dpms_off || link->link_status.link_active) && - (link->connector_signal == SIGNAL_TYPE_EDP)) + (link->connector_signal == SIGNAL_TYPE_EDP)) { dc->hwss.blank_stream(pipe_ctx); + if (dc->hwseq) + dc->hwseq->wa_state.skip_blank_stream = true; + } pipe_ctx->stream_res.tg->funcs->set_dsc_config( pipe_ctx->stream_res.tg, @@ -570,7 +576,8 @@ static void dcn31_reset_back_end_for_pipe( pipe_ctx->stream_res.audio = NULL; } } - + if (dc->hwseq) + dc->hwseq->wa_state.skip_blank_stream = false; pipe_ctx->stream = NULL; DC_LOG_DEBUG("Reset back end for pipe %d, tg:%d\n", pipe_ctx->pipe_idx, pipe_ctx->stream_res.tg->inst); diff --git a/drivers/gpu/drm/amd/display/dc/hwss/hw_sequencer_private.h b/drivers/gpu/drm/amd/display/dc/hwss/hw_sequencer_private.h index 09bc65c2fa23..1e2d247fbbac 100644 --- a/drivers/gpu/drm/amd/display/dc/hwss/hw_sequencer_private.h +++ b/drivers/gpu/drm/amd/display/dc/hwss/hw_sequencer_private.h @@ -49,6 +49,7 @@ struct hwseq_wa_state { bool DEGVIDCN10_253_applied; bool disallow_self_refresh_during_multi_plane_transition_applied; unsigned int disallow_self_refresh_during_multi_plane_transition_applied_on_frame; + bool skip_blank_stream; }; struct pipe_ctx; -- cgit v1.2.3 From c73375d918452e58f8903685d30ae21603040709 Mon Sep 17 00:00:00 2001 From: Cruise Hung Date: Thu, 22 May 2025 18:02:14 +0800 Subject: drm/amd/display: Use DC log instead of using DM error msg [Why & How] It sent an error msg when it failed to read the DP tunneling DPCD field. This should just be a warning msg. Use a DC log instead of a DM error msg. Reviewed-by: Wenjing Liu Signed-off-by: Cruise Hung Signed-off-by: Wayne Lin Tested-by: Daniel Wheeler Signed-off-by: Alex Deucher --- drivers/gpu/drm/amd/display/dc/link/protocols/link_dp_capability.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) (limited to 'drivers/gpu') diff --git a/drivers/gpu/drm/amd/display/dc/link/protocols/link_dp_capability.c b/drivers/gpu/drm/amd/display/dc/link/protocols/link_dp_capability.c index 8f79881ad9f1..a5127c2d47ef 100644 --- a/drivers/gpu/drm/amd/display/dc/link/protocols/link_dp_capability.c +++ b/drivers/gpu/drm/amd/display/dc/link/protocols/link_dp_capability.c @@ -2023,7 +2023,7 @@ static bool retrieve_link_cap(struct dc_link *link) /* Read DP tunneling information. */ status = dpcd_get_tunneling_device_data(link); if (status != DC_OK) - dm_error("%s: Read DP tunneling device data failed.\n", __func__); + DC_LOG_DP2("%s: Read DP tunneling device data failed.\n", __func__); retrieve_cable_id(link); dpcd_write_cable_id_to_dprx(link); -- cgit v1.2.3 From 4b61b8a390511a1864f26cc42bab72881e93468d Mon Sep 17 00:00:00 2001 From: Mario Limonciello Date: Wed, 14 May 2025 16:00:43 -0500 Subject: drm/amd/display: Add debugging message for brightness caps [Why] Default BIOS brightness caps are buried in ACPI. [How] Add extra dynamic debug that can show default brightness caps. Reviewed-by: Alex Hung Signed-off-by: Mario Limonciello Signed-off-by: Wayne Lin Tested-by: Daniel Wheeler Signed-off-by: Alex Deucher --- drivers/gpu/drm/amd/display/amdgpu_dm/amdgpu_dm.c | 7 ++++++- 1 file changed, 6 insertions(+), 1 deletion(-) (limited to 'drivers/gpu') 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 1797fa85fac6..588f6afb1447 100644 --- a/drivers/gpu/drm/amd/display/amdgpu_dm/amdgpu_dm.c +++ b/drivers/gpu/drm/amd/display/amdgpu_dm/amdgpu_dm.c @@ -4910,6 +4910,7 @@ amdgpu_dm_register_backlight_device(struct amdgpu_dm_connector *aconnector) struct backlight_properties props = { 0 }; struct amdgpu_dm_backlight_caps caps = { 0 }; char bl_name[16]; + int min, max; if (aconnector->bl_idx == -1) return; @@ -4922,11 +4923,15 @@ amdgpu_dm_register_backlight_device(struct amdgpu_dm_connector *aconnector) } amdgpu_acpi_get_backlight_caps(&caps); - if (caps.caps_valid) { + if (caps.caps_valid && get_brightness_range(&caps, &min, &max)) { if (power_supply_is_system_supplied() > 0) props.brightness = caps.ac_level; else props.brightness = caps.dc_level; + /* min is zero, so max needs to be adjusted */ + props.max_brightness = max - min; + drm_dbg(drm, "Backlight caps: min: %d, max: %d, ac %d, dc %d\n", min, max, + caps.ac_level, caps.dc_level); } else props.brightness = AMDGPU_MAX_BL_LEVEL; -- cgit v1.2.3 From 8b5f3a229a70d242322b78c8e13744ca00212def Mon Sep 17 00:00:00 2001 From: Mario Limonciello Date: Wed, 14 May 2025 16:06:40 -0500 Subject: drm/amd/display: Fix default DC and AC levels [Why] DC and AC levels are advertised in a percentage, not a luminance. [How] Scale DC and AC levels to supported values. Closes: https://gitlab.freedesktop.org/drm/amd/-/issues/4221 Reviewed-by: Alex Hung Signed-off-by: Mario Limonciello Signed-off-by: Wayne Lin Tested-by: Daniel Wheeler Signed-off-by: Alex Deucher --- drivers/gpu/drm/amd/display/amdgpu_dm/amdgpu_dm.c | 4 ++-- 1 file changed, 2 insertions(+), 2 deletions(-) (limited to 'drivers/gpu') 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 588f6afb1447..d3100f641ac6 100644 --- a/drivers/gpu/drm/amd/display/amdgpu_dm/amdgpu_dm.c +++ b/drivers/gpu/drm/amd/display/amdgpu_dm/amdgpu_dm.c @@ -4925,9 +4925,9 @@ amdgpu_dm_register_backlight_device(struct amdgpu_dm_connector *aconnector) amdgpu_acpi_get_backlight_caps(&caps); if (caps.caps_valid && get_brightness_range(&caps, &min, &max)) { if (power_supply_is_system_supplied() > 0) - props.brightness = caps.ac_level; + props.brightness = (max - min) * DIV_ROUND_CLOSEST(caps.ac_level, 100); else - props.brightness = caps.dc_level; + props.brightness = (max - min) * DIV_ROUND_CLOSEST(caps.dc_level, 100); /* min is zero, so max needs to be adjusted */ props.max_brightness = max - min; drm_dbg(drm, "Backlight caps: min: %d, max: %d, ac %d, dc %d\n", min, max, -- cgit v1.2.3 From 791d76005de0ab556b590473eb4cbfede727fce0 Mon Sep 17 00:00:00 2001 From: Dibin Moolakadan Subrahmanian Date: Wed, 28 May 2025 12:15:56 +0530 Subject: drm/i915/display: Fix u32 overflow in SNPS PHY HDMI PLL setup When configuring the HDMI PLL, calculations use DIV_ROUND_UP_ULL and DIV_ROUND_DOWN_ULL macros, which internally rely on do_div. However, do_div expects a 32-bit (u32) divisor, and at higher data rates, the divisor can exceed this limit. This leads to incorrect division results and ultimately misconfigured PLL values. This fix replaces do_div calls with div64_base64 calls where diviser can exceed u32 limit. Fixes: 5947642004bf ("drm/i915/display: Add support for SNPS PHY HDMI PLL algorithm for DG2") Cc: Ankit Nautiyal Cc: Suraj Kandpal Cc: Jani Nikula Signed-off-by: Dibin Moolakadan Subrahmanian Reviewed-by: Ankit Nautiyal Signed-off-by: Ankit Nautiyal Link: https://lore.kernel.org/r/20250528064557.4172149-1-dibin.moolakadan.subrahmanian@intel.com (cherry picked from commit ce924116e43ffbfa544d82976c4b9d11bcde9334) Signed-off-by: Joonas Lahtinen --- drivers/gpu/drm/i915/display/intel_snps_hdmi_pll.c | 16 ++++++++-------- 1 file changed, 8 insertions(+), 8 deletions(-) (limited to 'drivers/gpu') diff --git a/drivers/gpu/drm/i915/display/intel_snps_hdmi_pll.c b/drivers/gpu/drm/i915/display/intel_snps_hdmi_pll.c index c6321dafef4f..74bb3bedf30f 100644 --- a/drivers/gpu/drm/i915/display/intel_snps_hdmi_pll.c +++ b/drivers/gpu/drm/i915/display/intel_snps_hdmi_pll.c @@ -41,12 +41,12 @@ static s64 interp(s64 x, s64 x1, s64 x2, s64 y1, s64 y2) { s64 dydx; - dydx = DIV_ROUND_UP_ULL((y2 - y1) * 100000, (x2 - x1)); + dydx = DIV64_U64_ROUND_UP((y2 - y1) * 100000, (x2 - x1)); - return (y1 + DIV_ROUND_UP_ULL(dydx * (x - x1), 100000)); + return (y1 + DIV64_U64_ROUND_UP(dydx * (x - x1), 100000)); } -static void get_ana_cp_int_prop(u32 vco_clk, +static void get_ana_cp_int_prop(u64 vco_clk, u32 refclk_postscalar, int mpll_ana_v2i, int c, int a, @@ -115,16 +115,16 @@ static void get_ana_cp_int_prop(u32 vco_clk, CURVE0_MULTIPLIER)); scaled_interpolated_sqrt = - int_sqrt(DIV_ROUND_UP_ULL(interpolated_product, vco_div_refclk_float) * + int_sqrt(DIV64_U64_ROUND_UP(interpolated_product, vco_div_refclk_float) * DIV_ROUND_DOWN_ULL(1000000000000ULL, 55)); /* Scale vco_div_refclk for ana_cp_int */ scaled_vco_div_refclk2 = DIV_ROUND_UP_ULL(vco_div_refclk_float, 1000000); - adjusted_vco_clk2 = 1460281 * DIV_ROUND_UP_ULL(scaled_interpolated_sqrt * + adjusted_vco_clk2 = 1460281 * DIV64_U64_ROUND_UP(scaled_interpolated_sqrt * scaled_vco_div_refclk2, curve_1_interpolated); - *ana_cp_prop = DIV_ROUND_UP_ULL(adjusted_vco_clk2, curve_2_scaled2); + *ana_cp_prop = DIV64_U64_ROUND_UP(adjusted_vco_clk2, curve_2_scaled2); *ana_cp_prop = max(1, min(*ana_cp_prop, 127)); } @@ -165,10 +165,10 @@ static void compute_hdmi_tmds_pll(u64 pixel_clock, u32 refclk, /* Select appropriate v2i point */ if (datarate <= INTEL_SNPS_PHY_HDMI_9999MHZ) { mpll_ana_v2i = 2; - tx_clk_div = ilog2(DIV_ROUND_DOWN_ULL(INTEL_SNPS_PHY_HDMI_9999MHZ, datarate)); + tx_clk_div = ilog2(div64_u64(INTEL_SNPS_PHY_HDMI_9999MHZ, datarate)); } else { mpll_ana_v2i = 3; - tx_clk_div = ilog2(DIV_ROUND_DOWN_ULL(INTEL_SNPS_PHY_HDMI_16GHZ, datarate)); + tx_clk_div = ilog2(div64_u64(INTEL_SNPS_PHY_HDMI_16GHZ, datarate)); } vco_clk = (datarate << tx_clk_div) >> 1; -- cgit v1.2.3 From a63e99b4d6d3a0353ef47146dd5bd562f08e1786 Mon Sep 17 00:00:00 2001 From: Matthew Auld Date: Wed, 14 May 2025 16:24:25 +0100 Subject: drm/xe/vm: move rebind_work init earlier MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit In xe_vm_close_and_put() we need to be able to call flush_work(rebind_work), however during vm creation we can call this on the error path, before having actually set up the worker, leading to a splat from flush_work(). It looks like we can simply move the worker init step earlier to fix this. Fixes: dd08ebf6c352 ("drm/xe: Introduce a new DRM driver for Intel GPUs") Signed-off-by: Matthew Auld Cc: Matthew Brost Cc: # v6.8+ Reviewed-by: Matthew Brost Link: https://lore.kernel.org/r/20250514152424.149591-3-matthew.auld@intel.com (cherry picked from commit 96af397aa1a2d1032a6e28ff3f4bc0ab4be40e1d) Signed-off-by: Thomas Hellström --- drivers/gpu/drm/xe/xe_vm.c | 8 ++++---- 1 file changed, 4 insertions(+), 4 deletions(-) (limited to 'drivers/gpu') diff --git a/drivers/gpu/drm/xe/xe_vm.c b/drivers/gpu/drm/xe/xe_vm.c index 79323c78130f..a68fd99ddfdb 100644 --- a/drivers/gpu/drm/xe/xe_vm.c +++ b/drivers/gpu/drm/xe/xe_vm.c @@ -1678,8 +1678,10 @@ struct xe_vm *xe_vm_create(struct xe_device *xe, u32 flags) * scheduler drops all the references of it, hence protecting the VM * for this case is necessary. */ - if (flags & XE_VM_FLAG_LR_MODE) + if (flags & XE_VM_FLAG_LR_MODE) { + INIT_WORK(&vm->preempt.rebind_work, preempt_rebind_work_func); xe_pm_runtime_get_noresume(xe); + } vm_resv_obj = drm_gpuvm_resv_object_alloc(&xe->drm); if (!vm_resv_obj) { @@ -1724,10 +1726,8 @@ struct xe_vm *xe_vm_create(struct xe_device *xe, u32 flags) vm->batch_invalidate_tlb = true; } - if (vm->flags & XE_VM_FLAG_LR_MODE) { - INIT_WORK(&vm->preempt.rebind_work, preempt_rebind_work_func); + if (vm->flags & XE_VM_FLAG_LR_MODE) vm->batch_invalidate_tlb = false; - } /* Fill pt_root after allocating scratch tables */ for_each_tile(tile, xe, id) { -- cgit v1.2.3 From 8cf8cde41ad01150afbd1327ad1942387787f7fd Mon Sep 17 00:00:00 2001 From: Matthew Auld Date: Wed, 14 May 2025 16:24:26 +0100 Subject: drm/xe/vm: move xe_svm_init() earlier MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit In xe_vm_close_and_put() we need to be able to call xe_svm_fini(), however during vm creation we can call this on the error path, before having actually initialised the svm state, leading to various splats followed by a fatal NPD. Fixes: 6fd979c2f331 ("drm/xe: Add SVM init / close / fini to faulting VMs") Link: https://gitlab.freedesktop.org/drm/xe/kernel/-/issues/4967 Signed-off-by: Matthew Auld Cc: Matthew Brost Reviewed-by: Matthew Brost Link: https://lore.kernel.org/r/20250514152424.149591-4-matthew.auld@intel.com (cherry picked from commit 4f296d77cf49fcb5f90b4674123ad7f3a0676165) Signed-off-by: Thomas Hellström --- drivers/gpu/drm/xe/xe_vm.c | 19 ++++++++++++------- 1 file changed, 12 insertions(+), 7 deletions(-) (limited to 'drivers/gpu') diff --git a/drivers/gpu/drm/xe/xe_vm.c b/drivers/gpu/drm/xe/xe_vm.c index a68fd99ddfdb..861577746929 100644 --- a/drivers/gpu/drm/xe/xe_vm.c +++ b/drivers/gpu/drm/xe/xe_vm.c @@ -1683,10 +1683,16 @@ struct xe_vm *xe_vm_create(struct xe_device *xe, u32 flags) xe_pm_runtime_get_noresume(xe); } + if (flags & XE_VM_FLAG_FAULT_MODE) { + err = xe_svm_init(vm); + if (err) + goto err_no_resv; + } + vm_resv_obj = drm_gpuvm_resv_object_alloc(&xe->drm); if (!vm_resv_obj) { err = -ENOMEM; - goto err_no_resv; + goto err_svm_fini; } drm_gpuvm_init(&vm->gpuvm, "Xe VM", DRM_GPUVM_RESV_PROTECTED, &xe->drm, @@ -1757,12 +1763,6 @@ struct xe_vm *xe_vm_create(struct xe_device *xe, u32 flags) } } - if (flags & XE_VM_FLAG_FAULT_MODE) { - err = xe_svm_init(vm); - if (err) - goto err_close; - } - if (number_tiles > 1) vm->composite_fence_ctx = dma_fence_context_alloc(1); @@ -1776,6 +1776,11 @@ err_close: xe_vm_close_and_put(vm); return ERR_PTR(err); +err_svm_fini: + if (flags & XE_VM_FLAG_FAULT_MODE) { + vm->size = 0; /* close the vm */ + xe_svm_fini(vm); + } err_no_resv: mutex_destroy(&vm->snap_mutex); for_each_tile(tile, xe, id) -- cgit v1.2.3 From 25a2aa779fc39c4559a5bde0f841d2cd4cbc4d66 Mon Sep 17 00:00:00 2001 From: Karthik Poosa Date: Thu, 29 May 2025 22:04:53 +0530 Subject: drm/xe/hwmon: Add support to manage power limits though mailbox MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit Add support to manage power limits using pcode mailbox commands for supported platforms. v2: - Address review comments. (Badal) - Use mailbox commands instead of registers to manage power limits for BMG. - Clamp the maximum power limit to GPU firmware default value. v3: - Clamp power limit in write also for platforms with mailbox support. v4: - Remove unnecessary debug prints. (Badal) v5: - Update description of variable pl1_on_boot to fix kernel-doc error. v6: - Improve commit message, refer to BIOS as GPU firmware. - Change macro READ_PL_FROM_BIOS to READ_PL_FROM_FW. - Rectify drm_warn to drm_info. Signed-off-by: Karthik Poosa Fixes: e90f7a58e659 ("drm/xe/hwmon: Add HWMON support for BMG") Reviewed-by: Badal Nilawar Link: https://lore.kernel.org/r/20250529163458.2354509-2-karthik.poosa@intel.com Signed-off-by: Rodrigo Vivi (cherry picked from commit 7596d839f6228757fe17a810da2d1c5f3305078c) Signed-off-by: Thomas Hellström --- drivers/gpu/drm/xe/regs/xe_mchbar_regs.h | 10 +- drivers/gpu/drm/xe/regs/xe_pcode_regs.h | 4 - drivers/gpu/drm/xe/xe_device_types.h | 4 + drivers/gpu/drm/xe/xe_hwmon.c | 380 +++++++++++++++++++++++-------- drivers/gpu/drm/xe/xe_pci.c | 5 + drivers/gpu/drm/xe/xe_pcode.c | 11 + drivers/gpu/drm/xe/xe_pcode.h | 3 + drivers/gpu/drm/xe/xe_pcode_api.h | 7 + 8 files changed, 318 insertions(+), 106 deletions(-) (limited to 'drivers/gpu') diff --git a/drivers/gpu/drm/xe/regs/xe_mchbar_regs.h b/drivers/gpu/drm/xe/regs/xe_mchbar_regs.h index f5e5234857c1..5394a1373a6b 100644 --- a/drivers/gpu/drm/xe/regs/xe_mchbar_regs.h +++ b/drivers/gpu/drm/xe/regs/xe_mchbar_regs.h @@ -38,10 +38,10 @@ #define TEMP_MASK REG_GENMASK(7, 0) #define PCU_CR_PACKAGE_RAPL_LIMIT XE_REG(MCHBAR_MIRROR_BASE_SNB + 0x59a0) -#define PKG_PWR_LIM_1 REG_GENMASK(14, 0) -#define PKG_PWR_LIM_1_EN REG_BIT(15) -#define PKG_PWR_LIM_1_TIME REG_GENMASK(23, 17) -#define PKG_PWR_LIM_1_TIME_X REG_GENMASK(23, 22) -#define PKG_PWR_LIM_1_TIME_Y REG_GENMASK(21, 17) +#define PWR_LIM_VAL REG_GENMASK(14, 0) +#define PWR_LIM_EN REG_BIT(15) +#define PWR_LIM_TIME REG_GENMASK(23, 17) +#define PWR_LIM_TIME_X REG_GENMASK(23, 22) +#define PWR_LIM_TIME_Y REG_GENMASK(21, 17) #endif /* _XE_MCHBAR_REGS_H_ */ diff --git a/drivers/gpu/drm/xe/regs/xe_pcode_regs.h b/drivers/gpu/drm/xe/regs/xe_pcode_regs.h index c7d5d782e3f9..c556a04670ee 100644 --- a/drivers/gpu/drm/xe/regs/xe_pcode_regs.h +++ b/drivers/gpu/drm/xe/regs/xe_pcode_regs.h @@ -18,16 +18,12 @@ #define PVC_GT0_PLATFORM_ENERGY_STATUS XE_REG(0x28106c) #define PVC_GT0_PACKAGE_POWER_SKU XE_REG(0x281080) -#define BMG_PACKAGE_POWER_SKU XE_REG(0x138098) -#define BMG_PACKAGE_POWER_SKU_UNIT XE_REG(0x1380dc) #define BMG_PACKAGE_ENERGY_STATUS XE_REG(0x138120) #define BMG_FAN_1_SPEED XE_REG(0x138140) #define BMG_FAN_2_SPEED XE_REG(0x138170) #define BMG_FAN_3_SPEED XE_REG(0x1381a0) #define BMG_VRAM_TEMPERATURE XE_REG(0x1382c0) #define BMG_PACKAGE_TEMPERATURE XE_REG(0x138434) -#define BMG_PACKAGE_RAPL_LIMIT XE_REG(0x138440) #define BMG_PLATFORM_ENERGY_STATUS XE_REG(0x138458) -#define BMG_PLATFORM_POWER_LIMIT XE_REG(0x138460) #endif /* _XE_PCODE_REGS_H_ */ diff --git a/drivers/gpu/drm/xe/xe_device_types.h b/drivers/gpu/drm/xe/xe_device_types.h index 06c65dace026..b591c99f6f8a 100644 --- a/drivers/gpu/drm/xe/xe_device_types.h +++ b/drivers/gpu/drm/xe/xe_device_types.h @@ -325,6 +325,10 @@ struct xe_device { u8 has_heci_gscfi:1; /** @info.has_llc: Device has a shared CPU+GPU last level cache */ u8 has_llc:1; + /** @info.has_mbx_power_limits: Device has support to manage power limits using + * pcode mailbox commands. + */ + u8 has_mbx_power_limits:1; /** @info.has_pxp: Device has PXP support */ u8 has_pxp:1; /** @info.has_range_tlb_invalidation: Has range based TLB invalidations */ diff --git a/drivers/gpu/drm/xe/xe_hwmon.c b/drivers/gpu/drm/xe/xe_hwmon.c index eb293aec36a0..e272128f5145 100644 --- a/drivers/gpu/drm/xe/xe_hwmon.c +++ b/drivers/gpu/drm/xe/xe_hwmon.c @@ -51,6 +51,14 @@ enum xe_fan_channel { FAN_MAX, }; +/* + * For platforms that support mailbox commands for power limits, REG_PKG_POWER_SKU_UNIT is + * not supported and below are SKU units to be used. + */ +#define PWR_UNIT 0x3 +#define ENERGY_UNIT 0xe +#define TIME_UNIT 0xa + /* * SF_* - scale factors for particular quantities according to hwmon spec. */ @@ -60,6 +68,18 @@ enum xe_fan_channel { #define SF_ENERGY 1000000 /* microjoules */ #define SF_TIME 1000 /* milliseconds */ +/* + * PL*_HWMON_ATTR - mapping of hardware power limits to corresponding hwmon power attribute. + */ +#define PL1_HWMON_ATTR hwmon_power_max + +#define PWR_ATTR_TO_STR(attr) (((attr) == hwmon_power_max) ? "PL1" : "Invalid") + +/* + * Timeout for power limit write mailbox command. + */ +#define PL_WRITE_MBX_TIMEOUT_MS (1) + /** * struct xe_hwmon_energy_info - to accumulate energy */ @@ -100,8 +120,80 @@ struct xe_hwmon { struct xe_hwmon_energy_info ei[CHANNEL_MAX]; /** @fi: Fan info for fanN_input */ struct xe_hwmon_fan_info fi[FAN_MAX]; + /** @boot_power_limit_read: is boot power limits read */ + bool boot_power_limit_read; + /** @pl1_on_boot: power limit PL1 on boot */ + u32 pl1_on_boot[CHANNEL_MAX]; }; +static int xe_hwmon_pcode_read_power_limit(const struct xe_hwmon *hwmon, u32 attr, int channel, + u32 *uval) +{ + struct xe_tile *root_tile = xe_device_get_root_tile(hwmon->xe); + u32 val0 = 0, val1 = 0; + int ret = 0; + + ret = xe_pcode_read(root_tile, PCODE_MBOX(PCODE_POWER_SETUP, + (channel == CHANNEL_CARD) ? + READ_PSYSGPU_POWER_LIMIT : + READ_PACKAGE_POWER_LIMIT, + hwmon->boot_power_limit_read ? + READ_PL_FROM_PCODE : READ_PL_FROM_FW), + &val0, &val1); + + if (ret) { + drm_dbg(&hwmon->xe->drm, "read failed ch %d val0 0x%08x, val1 0x%08x, ret %d\n", + channel, val0, val1, ret); + *uval = 0; + return ret; + } + + /* return the value only if limit is enabled */ + if (attr == PL1_HWMON_ATTR) + *uval = (val0 & PWR_LIM_EN) ? val0 : 0; + else if (attr == hwmon_power_label) + *uval = (val0 & PWR_LIM_EN) ? 1 : 0; + else + *uval = 0; + + return ret; +} + +static int xe_hwmon_pcode_write_power_limit(const struct xe_hwmon *hwmon, u32 attr, u8 channel, + u32 uval) +{ + struct xe_tile *root_tile = xe_device_get_root_tile(hwmon->xe); + u32 val0, val1; + int ret = 0; + + ret = xe_pcode_read(root_tile, PCODE_MBOX(PCODE_POWER_SETUP, + (channel == CHANNEL_CARD) ? + READ_PSYSGPU_POWER_LIMIT : + READ_PACKAGE_POWER_LIMIT, + hwmon->boot_power_limit_read ? + READ_PL_FROM_PCODE : READ_PL_FROM_FW), + &val0, &val1); + + if (ret) + drm_dbg(&hwmon->xe->drm, "read failed ch %d val0 0x%08x, val1 0x%08x, ret %d\n", + channel, val0, val1, ret); + + if (attr == PL1_HWMON_ATTR) + val0 = uval; + else + return -EIO; + + ret = xe_pcode_write64_timeout(root_tile, PCODE_MBOX(PCODE_POWER_SETUP, + (channel == CHANNEL_CARD) ? + WRITE_PSYSGPU_POWER_LIMIT : + WRITE_PACKAGE_POWER_LIMIT, 0), + val0, val1, PL_WRITE_MBX_TIMEOUT_MS); + if (ret) + drm_dbg(&hwmon->xe->drm, "write failed ch %d val0 0x%08x, val1 0x%08x, ret %d\n", + channel, val0, val1, ret); + return ret; +} + static struct xe_reg xe_hwmon_get_reg(struct xe_hwmon *hwmon, enum xe_hwmon_reg hwmon_reg, int channel) { @@ -122,29 +214,19 @@ static struct xe_reg xe_hwmon_get_reg(struct xe_hwmon *hwmon, enum xe_hwmon_reg } break; case REG_PKG_RAPL_LIMIT: - if (xe->info.platform == XE_BATTLEMAGE) { - if (channel == CHANNEL_PKG) - return BMG_PACKAGE_RAPL_LIMIT; - else - return BMG_PLATFORM_POWER_LIMIT; - } else if (xe->info.platform == XE_PVC && channel == CHANNEL_PKG) { + if (xe->info.platform == XE_PVC && channel == CHANNEL_PKG) return PVC_GT0_PACKAGE_RAPL_LIMIT; - } else if ((xe->info.platform == XE_DG2) && (channel == CHANNEL_PKG)) { + else if ((xe->info.platform == XE_DG2) && (channel == CHANNEL_PKG)) return PCU_CR_PACKAGE_RAPL_LIMIT; - } break; case REG_PKG_POWER_SKU: - if (xe->info.platform == XE_BATTLEMAGE) - return BMG_PACKAGE_POWER_SKU; - else if (xe->info.platform == XE_PVC && channel == CHANNEL_PKG) + if (xe->info.platform == XE_PVC && channel == CHANNEL_PKG) return PVC_GT0_PACKAGE_POWER_SKU; else if ((xe->info.platform == XE_DG2) && (channel == CHANNEL_PKG)) return PCU_CR_PACKAGE_POWER_SKU; break; case REG_PKG_POWER_SKU_UNIT: - if (xe->info.platform == XE_BATTLEMAGE) - return BMG_PACKAGE_POWER_SKU_UNIT; - else if (xe->info.platform == XE_PVC) + if (xe->info.platform == XE_PVC) return PVC_GT0_PACKAGE_POWER_SKU_UNIT; else if (xe->info.platform == XE_DG2) return PCU_CR_PACKAGE_POWER_SKU_UNIT; @@ -181,7 +263,7 @@ static struct xe_reg xe_hwmon_get_reg(struct xe_hwmon *hwmon, enum xe_hwmon_reg return XE_REG(0); } -#define PL1_DISABLE 0 +#define PL_DISABLE 0 /* * HW allows arbitrary PL1 limits to be set but silently clamps these values to @@ -189,67 +271,83 @@ static struct xe_reg xe_hwmon_get_reg(struct xe_hwmon *hwmon, enum xe_hwmon_reg * same pattern for sysfs, allow arbitrary PL1 limits to be set but display * clamped values when read. */ -static void xe_hwmon_power_max_read(struct xe_hwmon *hwmon, int channel, long *value) +static void xe_hwmon_power_max_read(struct xe_hwmon *hwmon, u32 attr, int channel, long *value) { u64 reg_val, min, max; struct xe_device *xe = hwmon->xe; struct xe_reg rapl_limit, pkg_power_sku; struct xe_mmio *mmio = xe_root_tile_mmio(xe); - rapl_limit = xe_hwmon_get_reg(hwmon, REG_PKG_RAPL_LIMIT, channel); - pkg_power_sku = xe_hwmon_get_reg(hwmon, REG_PKG_POWER_SKU, channel); + mutex_lock(&hwmon->hwmon_lock); - /* - * Valid check of REG_PKG_RAPL_LIMIT is already done in xe_hwmon_power_is_visible. - * So not checking it again here. - */ - if (!xe_reg_is_valid(pkg_power_sku)) { - drm_warn(&xe->drm, "pkg_power_sku invalid\n"); - *value = 0; - return; + if (hwmon->xe->info.has_mbx_power_limits) { + xe_hwmon_pcode_read_power_limit(hwmon, attr, channel, (u32 *)®_val); + } else { + rapl_limit = xe_hwmon_get_reg(hwmon, REG_PKG_RAPL_LIMIT, channel); + pkg_power_sku = xe_hwmon_get_reg(hwmon, REG_PKG_POWER_SKU, channel); + + /* + * Valid check of REG_PKG_RAPL_LIMIT is already done in xe_hwmon_power_is_visible. + * So not checking it again here. + */ + if (!xe_reg_is_valid(pkg_power_sku)) { + drm_warn(&xe->drm, "pkg_power_sku invalid\n"); + *value = 0; + goto unlock; + } + reg_val = xe_mmio_read32(mmio, rapl_limit); } - mutex_lock(&hwmon->hwmon_lock); - - reg_val = xe_mmio_read32(mmio, rapl_limit); - /* Check if PL1 limit is disabled */ - if (!(reg_val & PKG_PWR_LIM_1_EN)) { - *value = PL1_DISABLE; + /* Check if PL limits are disabled. */ + if (!(reg_val & PWR_LIM_EN)) { + *value = PL_DISABLE; + drm_info(&hwmon->xe->drm, "%s disabled for channel %d, val 0x%016llx\n", + PWR_ATTR_TO_STR(attr), channel, reg_val); goto unlock; } - reg_val = REG_FIELD_GET(PKG_PWR_LIM_1, reg_val); + reg_val = REG_FIELD_GET(PWR_LIM_VAL, reg_val); *value = mul_u64_u32_shr(reg_val, SF_POWER, hwmon->scl_shift_power); - reg_val = xe_mmio_read64_2x32(mmio, pkg_power_sku); - min = REG_FIELD_GET(PKG_MIN_PWR, reg_val); - min = mul_u64_u32_shr(min, SF_POWER, hwmon->scl_shift_power); - max = REG_FIELD_GET(PKG_MAX_PWR, reg_val); - max = mul_u64_u32_shr(max, SF_POWER, hwmon->scl_shift_power); - - if (min && max) - *value = clamp_t(u64, *value, min, max); + /* For platforms with mailbox power limit support clamping would be done by pcode. */ + if (!hwmon->xe->info.has_mbx_power_limits) { + reg_val = xe_mmio_read64_2x32(mmio, pkg_power_sku); + min = REG_FIELD_GET(PKG_MIN_PWR, reg_val); + max = REG_FIELD_GET(PKG_MAX_PWR, reg_val); + min = mul_u64_u32_shr(min, SF_POWER, hwmon->scl_shift_power); + max = mul_u64_u32_shr(max, SF_POWER, hwmon->scl_shift_power); + if (min && max) + *value = clamp_t(u64, *value, min, max); + } unlock: mutex_unlock(&hwmon->hwmon_lock); } -static int xe_hwmon_power_max_write(struct xe_hwmon *hwmon, int channel, long value) +static int xe_hwmon_power_max_write(struct xe_hwmon *hwmon, u32 attr, int channel, long value) { struct xe_mmio *mmio = xe_root_tile_mmio(hwmon->xe); int ret = 0; - u64 reg_val; + u32 reg_val; struct xe_reg rapl_limit; + mutex_lock(&hwmon->hwmon_lock); + rapl_limit = xe_hwmon_get_reg(hwmon, REG_PKG_RAPL_LIMIT, channel); - mutex_lock(&hwmon->hwmon_lock); + /* Disable Power Limit and verify, as limit cannot be disabled on all platforms. */ + if (value == PL_DISABLE) { + if (hwmon->xe->info.has_mbx_power_limits) { + drm_dbg(&hwmon->xe->drm, "disabling %s on channel %d\n", + PWR_ATTR_TO_STR(attr), channel); + xe_hwmon_pcode_write_power_limit(hwmon, attr, channel, 0); + xe_hwmon_pcode_read_power_limit(hwmon, attr, channel, ®_val); + } else { + reg_val = xe_mmio_rmw32(mmio, rapl_limit, PWR_LIM_EN, 0); + reg_val = xe_mmio_read32(mmio, rapl_limit); + } - /* Disable PL1 limit and verify, as limit cannot be disabled on all platforms */ - if (value == PL1_DISABLE) { - reg_val = xe_mmio_rmw32(mmio, rapl_limit, PKG_PWR_LIM_1_EN, 0); - reg_val = xe_mmio_read32(mmio, rapl_limit); - if (reg_val & PKG_PWR_LIM_1_EN) { - drm_warn(&hwmon->xe->drm, "PL1 disable is not supported!\n"); + if (reg_val & PWR_LIM_EN) { + drm_warn(&hwmon->xe->drm, "Power limit disable is not supported!\n"); ret = -EOPNOTSUPP; } goto unlock; @@ -257,26 +355,50 @@ static int xe_hwmon_power_max_write(struct xe_hwmon *hwmon, int channel, long va /* Computation in 64-bits to avoid overflow. Round to nearest. */ reg_val = DIV_ROUND_CLOSEST_ULL((u64)value << hwmon->scl_shift_power, SF_POWER); - reg_val = PKG_PWR_LIM_1_EN | REG_FIELD_PREP(PKG_PWR_LIM_1, reg_val); - reg_val = xe_mmio_rmw32(mmio, rapl_limit, PKG_PWR_LIM_1_EN | PKG_PWR_LIM_1, reg_val); + reg_val = PWR_LIM_EN | REG_FIELD_PREP(PWR_LIM_VAL, reg_val); + /* + * Clamp power limit to card-firmware default as maximum, as an additional protection to + * pcode clamp. + */ + if (hwmon->xe->info.has_mbx_power_limits) { + if (reg_val > REG_FIELD_GET(PWR_LIM_VAL, hwmon->pl1_on_boot[channel])) { + reg_val = REG_FIELD_GET(PWR_LIM_VAL, hwmon->pl1_on_boot[channel]); + drm_dbg(&hwmon->xe->drm, "Clamping power limit to firmware default 0x%x\n", + reg_val); + } + } + + if (hwmon->xe->info.has_mbx_power_limits) + ret = xe_hwmon_pcode_write_power_limit(hwmon, attr, channel, reg_val); + else + reg_val = xe_mmio_rmw32(mmio, rapl_limit, PWR_LIM_EN | PWR_LIM_VAL, + reg_val); unlock: mutex_unlock(&hwmon->hwmon_lock); return ret; } -static void xe_hwmon_power_rated_max_read(struct xe_hwmon *hwmon, int channel, long *value) +static void xe_hwmon_power_rated_max_read(struct xe_hwmon *hwmon, u32 attr, int channel, + long *value) { struct xe_mmio *mmio = xe_root_tile_mmio(hwmon->xe); - struct xe_reg reg = xe_hwmon_get_reg(hwmon, REG_PKG_POWER_SKU, channel); - u64 reg_val; + u32 reg_val; + + if (hwmon->xe->info.has_mbx_power_limits) { + /* PL1 is rated max if supported. */ + xe_hwmon_pcode_read_power_limit(hwmon, PL1_HWMON_ATTR, channel, ®_val); + } else { + /* + * This sysfs file won't be visible if REG_PKG_POWER_SKU is invalid, so valid check + * for this register can be skipped. + * See xe_hwmon_power_is_visible. + */ + struct xe_reg reg = xe_hwmon_get_reg(hwmon, REG_PKG_POWER_SKU, channel); + + reg_val = xe_mmio_read32(mmio, reg); + } - /* - * This sysfs file won't be visible if REG_PKG_POWER_SKU is invalid, so valid check - * for this register can be skipped. - * See xe_hwmon_power_is_visible. - */ - reg_val = xe_mmio_read32(mmio, reg); reg_val = REG_FIELD_GET(PKG_TDP, reg_val); *value = mul_u64_u32_shr(reg_val, SF_POWER, hwmon->scl_shift_power); } @@ -330,23 +452,35 @@ xe_hwmon_power_max_interval_show(struct device *dev, struct device_attribute *at struct xe_mmio *mmio = xe_root_tile_mmio(hwmon->xe); u32 x, y, x_w = 2; /* 2 bits */ u64 r, tau4, out; - int sensor_index = to_sensor_dev_attr(attr)->index; + int channel = to_sensor_dev_attr(attr)->index; + u32 power_attr = PL1_HWMON_ATTR; + int ret = 0; xe_pm_runtime_get(hwmon->xe); mutex_lock(&hwmon->hwmon_lock); - r = xe_mmio_read32(mmio, xe_hwmon_get_reg(hwmon, REG_PKG_RAPL_LIMIT, sensor_index)); + if (hwmon->xe->info.has_mbx_power_limits) { + ret = xe_hwmon_pcode_read_power_limit(hwmon, power_attr, channel, (u32 *)&r); + if (ret) { + drm_err(&hwmon->xe->drm, + "power interval read fail, ch %d, attr %d, r 0%llx, ret %d\n", + channel, power_attr, r, ret); + r = 0; + } + } else { + r = xe_mmio_read32(mmio, xe_hwmon_get_reg(hwmon, REG_PKG_RAPL_LIMIT, channel)); + } mutex_unlock(&hwmon->hwmon_lock); xe_pm_runtime_put(hwmon->xe); - x = REG_FIELD_GET(PKG_PWR_LIM_1_TIME_X, r); - y = REG_FIELD_GET(PKG_PWR_LIM_1_TIME_Y, r); + x = REG_FIELD_GET(PWR_LIM_TIME_X, r); + y = REG_FIELD_GET(PWR_LIM_TIME_Y, r); /* - * tau = 1.x * power(2,y), x = bits(23:22), y = bits(21:17) + * tau = (1 + (x / 4)) * power(2,y), x = bits(23:22), y = bits(21:17) * = (4 | x) << (y - 2) * * Here (y - 2) ensures a 1.x fixed point representation of 1.x @@ -373,14 +507,15 @@ xe_hwmon_power_max_interval_store(struct device *dev, struct device_attribute *a u64 tau4, r, max_win; unsigned long val; int ret; - int sensor_index = to_sensor_dev_attr(attr)->index; + int channel = to_sensor_dev_attr(attr)->index; + u32 power_attr = PL1_HWMON_ATTR; ret = kstrtoul(buf, 0, &val); if (ret) return ret; /* - * Max HW supported tau in '1.x * power(2,y)' format, x = 0, y = 0x12. + * Max HW supported tau in '(1 + (x / 4)) * power(2,y)' format, x = 0, y = 0x12. * The hwmon->scl_shift_time default of 0xa results in a max tau of 256 seconds. * * The ideal scenario is for PKG_MAX_WIN to be read from the PKG_PWR_SKU register. @@ -400,11 +535,13 @@ xe_hwmon_power_max_interval_store(struct device *dev, struct device_attribute *a tau4 = (u64)((1 << x_w) | x) << y; max_win = mul_u64_u32_shr(tau4, SF_TIME, hwmon->scl_shift_time + x_w); - if (val > max_win) + if (val > max_win) { + drm_warn(&hwmon->xe->drm, "power_interval invalid val 0x%lx\n", val); return -EINVAL; + } /* val in hw units */ - val = DIV_ROUND_CLOSEST_ULL((u64)val << hwmon->scl_shift_time, SF_TIME); + val = DIV_ROUND_CLOSEST_ULL((u64)val << hwmon->scl_shift_time, SF_TIME) + 1; /* * Convert val to 1.x * power(2,y) @@ -419,14 +556,21 @@ xe_hwmon_power_max_interval_store(struct device *dev, struct device_attribute *a x = (val - (1ul << y)) << x_w >> y; } - rxy = REG_FIELD_PREP(PKG_PWR_LIM_1_TIME_X, x) | REG_FIELD_PREP(PKG_PWR_LIM_1_TIME_Y, y); + rxy = REG_FIELD_PREP(PWR_LIM_TIME_X, x) | + REG_FIELD_PREP(PWR_LIM_TIME_Y, y); xe_pm_runtime_get(hwmon->xe); mutex_lock(&hwmon->hwmon_lock); - r = xe_mmio_rmw32(mmio, xe_hwmon_get_reg(hwmon, REG_PKG_RAPL_LIMIT, sensor_index), - PKG_PWR_LIM_1_TIME, rxy); + if (hwmon->xe->info.has_mbx_power_limits) { + ret = xe_hwmon_pcode_read_power_limit(hwmon, power_attr, channel, (u32 *)&r); + r = (r & ~PWR_LIM_TIME) | rxy; + xe_hwmon_pcode_write_power_limit(hwmon, power_attr, channel, r); + } else { + r = xe_mmio_rmw32(mmio, xe_hwmon_get_reg(hwmon, REG_PKG_RAPL_LIMIT, channel), + PWR_LIM_TIME, rxy); + } mutex_unlock(&hwmon->hwmon_lock); @@ -435,6 +579,7 @@ xe_hwmon_power_max_interval_store(struct device *dev, struct device_attribute *a return count; } +/* PSYS PL1 */ static SENSOR_DEVICE_ATTR(power1_max_interval, 0664, xe_hwmon_power_max_interval_show, xe_hwmon_power_max_interval_store, CHANNEL_CARD); @@ -455,10 +600,19 @@ static umode_t xe_hwmon_attributes_visible(struct kobject *kobj, struct device *dev = kobj_to_dev(kobj); struct xe_hwmon *hwmon = dev_get_drvdata(dev); int ret = 0; + int channel = index ? CHANNEL_PKG : CHANNEL_CARD; + u32 power_attr = PL1_HWMON_ATTR; + u32 uval; xe_pm_runtime_get(hwmon->xe); - ret = xe_reg_is_valid(xe_hwmon_get_reg(hwmon, REG_PKG_RAPL_LIMIT, index)) ? attr->mode : 0; + if (hwmon->xe->info.has_mbx_power_limits) { + xe_hwmon_pcode_read_power_limit(hwmon, power_attr, channel, &uval); + ret = (uval & PWR_LIM_EN) ? attr->mode : 0; + } else { + ret = xe_reg_is_valid(xe_hwmon_get_reg(hwmon, REG_PKG_RAPL_LIMIT, + channel)) ? attr->mode : 0; + } xe_pm_runtime_put(hwmon->xe); @@ -604,19 +758,27 @@ xe_hwmon_power_is_visible(struct xe_hwmon *hwmon, u32 attr, int channel) switch (attr) { case hwmon_power_max: - return xe_reg_is_valid(xe_hwmon_get_reg(hwmon, REG_PKG_RAPL_LIMIT, + if (hwmon->xe->info.has_mbx_power_limits) { + xe_hwmon_pcode_read_power_limit(hwmon, attr, channel, &uval); + return (uval) ? 0664 : 0; + } else { + return xe_reg_is_valid(xe_hwmon_get_reg(hwmon, REG_PKG_RAPL_LIMIT, channel)) ? 0664 : 0; + } case hwmon_power_rated_max: - return xe_reg_is_valid(xe_hwmon_get_reg(hwmon, REG_PKG_POWER_SKU, - channel)) ? 0444 : 0; + if (hwmon->xe->info.has_mbx_power_limits) + return 0; + else + return xe_reg_is_valid(xe_hwmon_get_reg(hwmon, REG_PKG_POWER_SKU, + channel)) ? 0444 : 0; case hwmon_power_crit: - if (channel == CHANNEL_PKG) - return (xe_hwmon_pcode_read_i1(hwmon, &uval) || - !(uval & POWER_SETUP_I1_WATTS)) ? 0 : 0644; - break; case hwmon_power_label: - return xe_reg_is_valid(xe_hwmon_get_reg(hwmon, REG_PKG_POWER_SKU_UNIT, - channel)) ? 0444 : 0; + if (channel == CHANNEL_PKG) { + xe_hwmon_pcode_read_i1(hwmon, &uval); + return (uval & POWER_SETUP_I1_WATTS) ? (attr == hwmon_power_label) ? + 0444 : 0644 : 0; + } + break; default: return 0; } @@ -628,10 +790,10 @@ xe_hwmon_power_read(struct xe_hwmon *hwmon, u32 attr, int channel, long *val) { switch (attr) { case hwmon_power_max: - xe_hwmon_power_max_read(hwmon, channel, val); + xe_hwmon_power_max_read(hwmon, attr, channel, val); return 0; case hwmon_power_rated_max: - xe_hwmon_power_rated_max_read(hwmon, channel, val); + xe_hwmon_power_rated_max_read(hwmon, attr, channel, val); return 0; case hwmon_power_crit: return xe_hwmon_power_curr_crit_read(hwmon, channel, val, SF_POWER); @@ -645,7 +807,7 @@ xe_hwmon_power_write(struct xe_hwmon *hwmon, u32 attr, int channel, long val) { switch (attr) { case hwmon_power_max: - return xe_hwmon_power_max_write(hwmon, channel, val); + return xe_hwmon_power_max_write(hwmon, attr, channel, val); case hwmon_power_crit: return xe_hwmon_power_curr_crit_write(hwmon, channel, val, SF_POWER); default: @@ -965,18 +1127,42 @@ xe_hwmon_get_preregistration_info(struct xe_hwmon *hwmon) int channel; struct xe_reg pkg_power_sku_unit; - /* - * The contents of register PKG_POWER_SKU_UNIT do not change, - * so read it once and store the shift values. - */ - pkg_power_sku_unit = xe_hwmon_get_reg(hwmon, REG_PKG_POWER_SKU_UNIT, 0); - if (xe_reg_is_valid(pkg_power_sku_unit)) { - val_sku_unit = xe_mmio_read32(mmio, pkg_power_sku_unit); - hwmon->scl_shift_power = REG_FIELD_GET(PKG_PWR_UNIT, val_sku_unit); - hwmon->scl_shift_energy = REG_FIELD_GET(PKG_ENERGY_UNIT, val_sku_unit); - hwmon->scl_shift_time = REG_FIELD_GET(PKG_TIME_UNIT, val_sku_unit); + if (hwmon->xe->info.has_mbx_power_limits) { + /* Check if card firmware support mailbox power limits commands. */ + if (xe_hwmon_pcode_read_power_limit(hwmon, PL1_HWMON_ATTR, CHANNEL_CARD, + &hwmon->pl1_on_boot[CHANNEL_CARD]) | + xe_hwmon_pcode_read_power_limit(hwmon, PL1_HWMON_ATTR, CHANNEL_PKG, + &hwmon->pl1_on_boot[CHANNEL_PKG])) { + drm_warn(&hwmon->xe->drm, + "Failed to read power limits, check card firmware !\n"); + } else { + drm_info(&hwmon->xe->drm, "Using mailbox commands for power limits\n"); + /* Write default limits to read from pcode from now on. */ + xe_hwmon_pcode_write_power_limit(hwmon, PL1_HWMON_ATTR, + CHANNEL_CARD, + hwmon->pl1_on_boot[CHANNEL_CARD]); + xe_hwmon_pcode_write_power_limit(hwmon, PL1_HWMON_ATTR, + CHANNEL_PKG, + hwmon->pl1_on_boot[CHANNEL_PKG]); + hwmon->scl_shift_power = PWR_UNIT; + hwmon->scl_shift_energy = ENERGY_UNIT; + hwmon->scl_shift_time = TIME_UNIT; + hwmon->boot_power_limit_read = true; + } + } else { + drm_info(&hwmon->xe->drm, "Using register for power limits\n"); + /* + * The contents of register PKG_POWER_SKU_UNIT do not change, + * so read it once and store the shift values. + */ + pkg_power_sku_unit = xe_hwmon_get_reg(hwmon, REG_PKG_POWER_SKU_UNIT, 0); + if (xe_reg_is_valid(pkg_power_sku_unit)) { + val_sku_unit = xe_mmio_read32(mmio, pkg_power_sku_unit); + hwmon->scl_shift_power = REG_FIELD_GET(PKG_PWR_UNIT, val_sku_unit); + hwmon->scl_shift_energy = REG_FIELD_GET(PKG_ENERGY_UNIT, val_sku_unit); + hwmon->scl_shift_time = REG_FIELD_GET(PKG_TIME_UNIT, val_sku_unit); + } } - /* * Initialize 'struct xe_hwmon_energy_info', i.e. set fields to the * first value of the energy register read diff --git a/drivers/gpu/drm/xe/xe_pci.c b/drivers/gpu/drm/xe/xe_pci.c index 882398e09b7e..95a2a458e8f7 100644 --- a/drivers/gpu/drm/xe/xe_pci.c +++ b/drivers/gpu/drm/xe/xe_pci.c @@ -66,6 +66,7 @@ struct xe_device_desc { u8 has_heci_gscfi:1; u8 has_heci_cscfi:1; u8 has_llc:1; + u8 has_mbx_power_limits:1; u8 has_pxp:1; u8 has_sriov:1; u8 needs_scratch:1; @@ -305,6 +306,7 @@ static const struct xe_device_desc dg2_desc = { DG2_FEATURES, .has_display = true, .has_fan_control = true, + .has_mbx_power_limits = false, }; static const __maybe_unused struct xe_device_desc pvc_desc = { @@ -316,6 +318,7 @@ static const __maybe_unused struct xe_device_desc pvc_desc = { .has_heci_gscfi = 1, .max_remote_tiles = 1, .require_force_probe = true, + .has_mbx_power_limits = false, }; static const struct xe_device_desc mtl_desc = { @@ -341,6 +344,7 @@ static const struct xe_device_desc bmg_desc = { .dma_mask_size = 46, .has_display = true, .has_fan_control = true, + .has_mbx_power_limits = true, .has_heci_cscfi = 1, .needs_scratch = true, }; @@ -583,6 +587,7 @@ static int xe_info_init_early(struct xe_device *xe, xe->info.dma_mask_size = desc->dma_mask_size; xe->info.is_dgfx = desc->is_dgfx; xe->info.has_fan_control = desc->has_fan_control; + xe->info.has_mbx_power_limits = desc->has_mbx_power_limits; xe->info.has_heci_gscfi = desc->has_heci_gscfi; xe->info.has_heci_cscfi = desc->has_heci_cscfi; xe->info.has_llc = desc->has_llc; diff --git a/drivers/gpu/drm/xe/xe_pcode.c b/drivers/gpu/drm/xe/xe_pcode.c index cf955b3ed52c..9189117fe825 100644 --- a/drivers/gpu/drm/xe/xe_pcode.c +++ b/drivers/gpu/drm/xe/xe_pcode.c @@ -109,6 +109,17 @@ int xe_pcode_write_timeout(struct xe_tile *tile, u32 mbox, u32 data, int timeout return err; } +int xe_pcode_write64_timeout(struct xe_tile *tile, u32 mbox, u32 data0, u32 data1, int timeout) +{ + int err; + + mutex_lock(&tile->pcode.lock); + err = pcode_mailbox_rw(tile, mbox, &data0, &data1, timeout, false, false); + mutex_unlock(&tile->pcode.lock); + + return err; +} + int xe_pcode_read(struct xe_tile *tile, u32 mbox, u32 *val, u32 *val1) { int err; diff --git a/drivers/gpu/drm/xe/xe_pcode.h b/drivers/gpu/drm/xe/xe_pcode.h index ba33991d72a7..de38f44f3201 100644 --- a/drivers/gpu/drm/xe/xe_pcode.h +++ b/drivers/gpu/drm/xe/xe_pcode.h @@ -18,6 +18,9 @@ int xe_pcode_init_min_freq_table(struct xe_tile *tile, u32 min_gt_freq, int xe_pcode_read(struct xe_tile *tile, u32 mbox, u32 *val, u32 *val1); int xe_pcode_write_timeout(struct xe_tile *tile, u32 mbox, u32 val, int timeout_ms); +int xe_pcode_write64_timeout(struct xe_tile *tile, u32 mbox, u32 data0, + u32 data1, int timeout); + #define xe_pcode_write(tile, mbox, val) \ xe_pcode_write_timeout(tile, mbox, val, 1) diff --git a/drivers/gpu/drm/xe/xe_pcode_api.h b/drivers/gpu/drm/xe/xe_pcode_api.h index 127d4d26c4cf..0befdea77db1 100644 --- a/drivers/gpu/drm/xe/xe_pcode_api.h +++ b/drivers/gpu/drm/xe/xe_pcode_api.h @@ -43,6 +43,13 @@ #define POWER_SETUP_I1_SHIFT 6 /* 10.6 fixed point format */ #define POWER_SETUP_I1_DATA_MASK REG_GENMASK(15, 0) +#define READ_PSYSGPU_POWER_LIMIT 0x6 +#define WRITE_PSYSGPU_POWER_LIMIT 0x7 +#define READ_PACKAGE_POWER_LIMIT 0x8 +#define WRITE_PACKAGE_POWER_LIMIT 0x9 +#define READ_PL_FROM_FW 0x1 +#define READ_PL_FROM_PCODE 0x0 + #define PCODE_FREQUENCY_CONFIG 0x6e /* Frequency Config Sub Commands (param1) */ #define PCODE_MBOX_FC_SC_READ_FUSED_P0 0x0 -- cgit v1.2.3 From b885ae2e9db3dba8e9b3bc4df36744f22455d889 Mon Sep 17 00:00:00 2001 From: Karthik Poosa Date: Thu, 29 May 2025 22:04:54 +0530 Subject: drm/xe/hwmon: Move card reactive critical power under channel card MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit Move power2/curr2_crit to channel 1 i.e power1/curr1_crit as this represents the entire card critical power/current. v2: Update the date of curr1_crit also in hwmon documentation. Signed-off-by: Karthik Poosa Fixes: 345dadc4f68b ("drm/xe/hwmon: Add infra to support card power and energy attributes") Reviewed-by: Badal Nilawar Link: https://lore.kernel.org/r/20250529163458.2354509-3-karthik.poosa@intel.com Signed-off-by: Rodrigo Vivi (cherry picked from commit 25e963a09e059ffdb15c09cc79cfded855b43668) Signed-off-by: Thomas Hellström --- drivers/gpu/drm/xe/xe_hwmon.c | 6 +++--- 1 file changed, 3 insertions(+), 3 deletions(-) (limited to 'drivers/gpu') diff --git a/drivers/gpu/drm/xe/xe_hwmon.c b/drivers/gpu/drm/xe/xe_hwmon.c index e272128f5145..74f31639b37f 100644 --- a/drivers/gpu/drm/xe/xe_hwmon.c +++ b/drivers/gpu/drm/xe/xe_hwmon.c @@ -632,8 +632,8 @@ static const struct attribute_group *hwmon_groups[] = { static const struct hwmon_channel_info * const hwmon_info[] = { HWMON_CHANNEL_INFO(temp, HWMON_T_LABEL, HWMON_T_INPUT | HWMON_T_LABEL, HWMON_T_INPUT | HWMON_T_LABEL), - HWMON_CHANNEL_INFO(power, HWMON_P_MAX | HWMON_P_RATED_MAX | HWMON_P_LABEL, - HWMON_P_MAX | HWMON_P_RATED_MAX | HWMON_P_CRIT | HWMON_P_LABEL), + HWMON_CHANNEL_INFO(power, HWMON_P_MAX | HWMON_P_RATED_MAX | HWMON_P_LABEL | HWMON_P_CRIT, + HWMON_P_MAX | HWMON_P_RATED_MAX | HWMON_P_LABEL), HWMON_CHANNEL_INFO(curr, HWMON_C_LABEL, HWMON_C_CRIT | HWMON_C_LABEL), HWMON_CHANNEL_INFO(in, HWMON_I_INPUT | HWMON_I_LABEL, HWMON_I_INPUT | HWMON_I_LABEL), HWMON_CHANNEL_INFO(energy, HWMON_E_INPUT | HWMON_E_LABEL, HWMON_E_INPUT | HWMON_E_LABEL), @@ -773,7 +773,7 @@ xe_hwmon_power_is_visible(struct xe_hwmon *hwmon, u32 attr, int channel) channel)) ? 0444 : 0; case hwmon_power_crit: case hwmon_power_label: - if (channel == CHANNEL_PKG) { + if (channel == CHANNEL_CARD) { xe_hwmon_pcode_read_i1(hwmon, &uval); return (uval & POWER_SETUP_I1_WATTS) ? (attr == hwmon_power_label) ? 0444 : 0644 : 0; -- cgit v1.2.3 From 94110827925a2512e480176b3002e08105f98d66 Mon Sep 17 00:00:00 2001 From: Raag Jadav Date: Thu, 29 May 2025 21:39:37 +0530 Subject: drm/xe: drop redundant conversion to bool MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit The result of integer comparison already evaluates to bool. No need for explicit conversion. No functional impact. Fixes: 0e414bf7ad01 ("drm/xe: Expose PCIe link downgrade attributes") Reported-by: kernel test robot Closes: https://lore.kernel.org/oe-kbuild-all/202505292205.MoljmkjQ-lkp@intel.com/ Signed-off-by: Raag Jadav Reviewed-by: Rodrigo Vivi Link: https://lore.kernel.org/r/20250529160937.490147-1-raag.jadav@intel.com Signed-off-by: Rodrigo Vivi (cherry picked from commit 61761a6b57f2818983466d24aab60baab471ba21) Signed-off-by: Thomas Hellström --- drivers/gpu/drm/xe/xe_device_sysfs.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) (limited to 'drivers/gpu') diff --git a/drivers/gpu/drm/xe/xe_device_sysfs.c b/drivers/gpu/drm/xe/xe_device_sysfs.c index 2e657692e5b5..b9440f8c781e 100644 --- a/drivers/gpu/drm/xe/xe_device_sysfs.c +++ b/drivers/gpu/drm/xe/xe_device_sysfs.c @@ -115,7 +115,7 @@ auto_link_downgrade_capable_show(struct device *dev, struct device_attribute *at xe_pm_runtime_put(xe); cap = REG_FIELD_GET(LINK_DOWNGRADE, val); - return sysfs_emit(buf, "%u\n", cap == DOWNGRADE_CAPABLE ? true : false); + return sysfs_emit(buf, "%u\n", cap == DOWNGRADE_CAPABLE); } static DEVICE_ATTR_ADMIN_RO(auto_link_downgrade_capable); -- cgit v1.2.3 From 2182f358fb138f81a586ffdddd510f2a4fc61702 Mon Sep 17 00:00:00 2001 From: Arnd Bergmann Date: Thu, 29 May 2025 10:23:56 -0700 Subject: drm/xe/vsec: fix CONFIG_INTEL_VSEC dependency MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit The XE driver can be built with or without VSEC support, but fails to link as built-in if vsec is in a loadable module: x86_64-linux-ld: vmlinux.o: in function `xe_vsec_init': (.text+0x1e83e16): undefined reference to `intel_vsec_register' The normal fix for this is to add a 'depends on INTEL_VSEC || !INTEL_VSEC', forcing XE to be a loadable module as well, but that causes a circular dependency: symbol DRM_XE depends on INTEL_VSEC symbol INTEL_VSEC depends on X86_PLATFORM_DEVICES symbol X86_PLATFORM_DEVICES is selected by DRM_XE The problem here is selecting a symbol from another subsystem, so change that as well and rephrase the 'select' into the corresponding dependency. Since X86_PLATFORM_DEVICES is 'default y', there is no change to defconfig builds here. Fixes: 0c45e76fcc62 ("drm/xe/vsec: Support BMG devices") Signed-off-by: Arnd Bergmann Reviewed-by: Lucas De Marchi Link: https://lore.kernel.org/r/20250529172355.2395634-2-lucas.demarchi@intel.com Signed-off-by: Lucas De Marchi (cherry picked from commit e4931f8be347ec5f19df4d6d33aea37145378c42) Signed-off-by: Thomas Hellström --- drivers/gpu/drm/xe/Kconfig | 3 ++- 1 file changed, 2 insertions(+), 1 deletion(-) (limited to 'drivers/gpu') diff --git a/drivers/gpu/drm/xe/Kconfig b/drivers/gpu/drm/xe/Kconfig index 9bce047901b2..98b46c534278 100644 --- a/drivers/gpu/drm/xe/Kconfig +++ b/drivers/gpu/drm/xe/Kconfig @@ -2,6 +2,8 @@ config DRM_XE tristate "Intel Xe Graphics" depends on DRM && PCI && MMU && (m || (y && KUNIT=y)) + depends on INTEL_VSEC || !INTEL_VSEC + depends on X86_PLATFORM_DEVICES || !(X86 && ACPI) select INTERVAL_TREE # we need shmfs for the swappable backing store, and in particular # the shmem_readpage() which depends upon tmpfs @@ -27,7 +29,6 @@ config DRM_XE select BACKLIGHT_CLASS_DEVICE if ACPI select INPUT if ACPI select ACPI_VIDEO if X86 && ACPI - select X86_PLATFORM_DEVICES if X86 && ACPI select ACPI_WMI if X86 && ACPI select SYNC_FILE select IOSF_MBI -- cgit v1.2.3 From 5cc3325584c425069c1c3355c775314d64bf8770 Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Thomas=20Hellstr=C3=B6m?= Date: Wed, 28 May 2025 18:41:05 +0200 Subject: drm/xe: Rework eviction rejection of bound external bos MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit For preempt_fence mode VM's we're rejecting eviction of shared bos during VM_BIND. However, since we do this in the move() callback, we're getting an eviction failure warning from TTM. The TTM callback intended for these things is eviction_valuable(). However, the latter doesn't pass in the struct ttm_operation_ctx needed to determine whether the caller needs this. Instead, attach the needed information to the vm under the vm->resv, until we've been able to update TTM to provide the needed information. And add sufficient lockdep checks to prevent misuse and races. v2: - Fix a copy-paste error in xe_vm_clear_validating() v3: - Fix kerneldoc errors. Signed-off-by: Thomas Hellström Fixes: 0af944f0e308 ("drm/xe: Reject BO eviction if BO is bound to current VM") Reviewed-by: Matthew Brost Link: https://lore.kernel.org/r/20250528164105.234718-1-thomas.hellstrom@linux.intel.com (cherry picked from commit 9d5558649f68e2e84a87a909631b30e15ca0f8ec) Signed-off-by: Thomas Hellström --- drivers/gpu/drm/xe/xe_bo.c | 46 ++++++++++++++++----------- drivers/gpu/drm/xe/xe_vm.h | 69 ++++++++++++++++++++++++++++++++++++++++ drivers/gpu/drm/xe/xe_vm_types.h | 8 +++++ 3 files changed, 105 insertions(+), 18 deletions(-) (limited to 'drivers/gpu') diff --git a/drivers/gpu/drm/xe/xe_bo.c b/drivers/gpu/drm/xe/xe_bo.c index d99d91fe8aa9..3c48a8c5f439 100644 --- a/drivers/gpu/drm/xe/xe_bo.c +++ b/drivers/gpu/drm/xe/xe_bo.c @@ -841,21 +841,6 @@ static int xe_bo_move(struct ttm_buffer_object *ttm_bo, bool evict, goto out; } - /* Reject BO eviction if BO is bound to current VM. */ - if (evict && ctx->resv) { - struct drm_gpuvm_bo *vm_bo; - - drm_gem_for_each_gpuvm_bo(vm_bo, &bo->ttm.base) { - struct xe_vm *vm = gpuvm_to_vm(vm_bo->vm); - - if (xe_vm_resv(vm) == ctx->resv && - xe_vm_in_preempt_fence_mode(vm)) { - ret = -EBUSY; - goto out; - } - } - } - /* * Failed multi-hop where the old_mem is still marked as * TTM_PL_FLAG_TEMPORARY, should just be a dummy move. @@ -1013,6 +998,25 @@ static long xe_bo_shrink_purge(struct ttm_operation_ctx *ctx, return lret; } +static bool +xe_bo_eviction_valuable(struct ttm_buffer_object *bo, const struct ttm_place *place) +{ + struct drm_gpuvm_bo *vm_bo; + + if (!ttm_bo_eviction_valuable(bo, place)) + return false; + + if (!xe_bo_is_xe_bo(bo)) + return true; + + drm_gem_for_each_gpuvm_bo(vm_bo, &bo->base) { + if (xe_vm_is_validating(gpuvm_to_vm(vm_bo->vm))) + return false; + } + + return true; +} + /** * xe_bo_shrink() - Try to shrink an xe bo. * @ctx: The struct ttm_operation_ctx used for shrinking. @@ -1047,7 +1051,7 @@ long xe_bo_shrink(struct ttm_operation_ctx *ctx, struct ttm_buffer_object *bo, (flags.purge && !xe_tt->purgeable)) return -EBUSY; - if (!ttm_bo_eviction_valuable(bo, &place)) + if (!xe_bo_eviction_valuable(bo, &place)) return -EBUSY; if (!xe_bo_is_xe_bo(bo) || !xe_bo_get_unless_zero(xe_bo)) @@ -1588,7 +1592,7 @@ const struct ttm_device_funcs xe_ttm_funcs = { .io_mem_pfn = xe_ttm_io_mem_pfn, .access_memory = xe_ttm_access_memory, .release_notify = xe_ttm_bo_release_notify, - .eviction_valuable = ttm_bo_eviction_valuable, + .eviction_valuable = xe_bo_eviction_valuable, .delete_mem_notify = xe_ttm_bo_delete_mem_notify, .swap_notify = xe_ttm_bo_swap_notify, }; @@ -2431,6 +2435,8 @@ int xe_bo_validate(struct xe_bo *bo, struct xe_vm *vm, bool allow_res_evict) .no_wait_gpu = false, .gfp_retry_mayfail = true, }; + struct pin_cookie cookie; + int ret; if (vm) { lockdep_assert_held(&vm->lock); @@ -2440,8 +2446,12 @@ int xe_bo_validate(struct xe_bo *bo, struct xe_vm *vm, bool allow_res_evict) ctx.resv = xe_vm_resv(vm); } + cookie = xe_vm_set_validating(vm, allow_res_evict); trace_xe_bo_validate(bo); - return ttm_bo_validate(&bo->ttm, &bo->placement, &ctx); + ret = ttm_bo_validate(&bo->ttm, &bo->placement, &ctx); + xe_vm_clear_validating(vm, allow_res_evict, cookie); + + return ret; } bool xe_bo_is_xe_bo(struct ttm_buffer_object *bo) diff --git a/drivers/gpu/drm/xe/xe_vm.h b/drivers/gpu/drm/xe/xe_vm.h index 0ef811fc2bde..494af6bdc646 100644 --- a/drivers/gpu/drm/xe/xe_vm.h +++ b/drivers/gpu/drm/xe/xe_vm.h @@ -301,6 +301,75 @@ void xe_vm_snapshot_capture_delayed(struct xe_vm_snapshot *snap); void xe_vm_snapshot_print(struct xe_vm_snapshot *snap, struct drm_printer *p); void xe_vm_snapshot_free(struct xe_vm_snapshot *snap); +/** + * xe_vm_set_validating() - Register this task as currently making bos resident + * @allow_res_evict: Allow eviction of buffer objects bound to @vm when + * validating. + * @vm: Pointer to the vm or NULL. + * + * Register this task as currently making bos resident for the vm. Intended + * to avoid eviction by the same task of shared bos bound to the vm. + * Call with the vm's resv lock held. + * + * Return: A pin cookie that should be used for xe_vm_clear_validating(). + */ +static inline struct pin_cookie xe_vm_set_validating(struct xe_vm *vm, + bool allow_res_evict) +{ + struct pin_cookie cookie = {}; + + if (vm && !allow_res_evict) { + xe_vm_assert_held(vm); + cookie = lockdep_pin_lock(&xe_vm_resv(vm)->lock.base); + /* Pairs with READ_ONCE in xe_vm_is_validating() */ + WRITE_ONCE(vm->validating, current); + } + + return cookie; +} + +/** + * xe_vm_clear_validating() - Unregister this task as currently making bos resident + * @vm: Pointer to the vm or NULL + * @allow_res_evict: Eviction from @vm was allowed. Must be set to the same + * value as for xe_vm_set_validation(). + * @cookie: Cookie obtained from xe_vm_set_validating(). + * + * Register this task as currently making bos resident for the vm. Intended + * to avoid eviction by the same task of shared bos bound to the vm. + * Call with the vm's resv lock held. + */ +static inline void xe_vm_clear_validating(struct xe_vm *vm, bool allow_res_evict, + struct pin_cookie cookie) +{ + if (vm && !allow_res_evict) { + lockdep_unpin_lock(&xe_vm_resv(vm)->lock.base, cookie); + /* Pairs with READ_ONCE in xe_vm_is_validating() */ + WRITE_ONCE(vm->validating, NULL); + } +} + +/** + * xe_vm_is_validating() - Whether bos bound to the vm are currently being made resident + * by the current task. + * @vm: Pointer to the vm. + * + * If this function returns %true, we should be in a vm resv locked region, since + * the current process is the same task that called xe_vm_set_validating(). + * The function asserts that that's indeed the case. + * + * Return: %true if the task is currently making bos resident, %false otherwise. + */ +static inline bool xe_vm_is_validating(struct xe_vm *vm) +{ + /* Pairs with WRITE_ONCE in xe_vm_is_validating() */ + if (READ_ONCE(vm->validating) == current) { + xe_vm_assert_held(vm); + return true; + } + return false; +} + #if IS_ENABLED(CONFIG_DRM_XE_USERPTR_INVAL_INJECT) void xe_vma_userptr_force_invalidate(struct xe_userptr_vma *uvma); #else diff --git a/drivers/gpu/drm/xe/xe_vm_types.h b/drivers/gpu/drm/xe/xe_vm_types.h index 1662604c4486..1979e9bdbdf3 100644 --- a/drivers/gpu/drm/xe/xe_vm_types.h +++ b/drivers/gpu/drm/xe/xe_vm_types.h @@ -310,6 +310,14 @@ struct xe_vm { * protected by the vm resv. */ u64 tlb_flush_seqno; + /** + * @validating: The task that is currently making bos resident for this vm. + * Protected by the VM's resv for writing. Opportunistic reading can be done + * using READ_ONCE. Note: This is a workaround for the + * TTM eviction_valuable() callback not being passed a struct + * ttm_operation_context(). Future work might want to address this. + */ + struct task_struct *validating; /** @batch_invalidate_tlb: Always invalidate TLB before batch start */ bool batch_invalidate_tlb; /** @xef: XE file handle for tracking this VM's drm client */ -- cgit v1.2.3 From 0ee54d5cacc0276ec631ac149825a24b59c51c38 Mon Sep 17 00:00:00 2001 From: Matthew Auld Date: Wed, 28 May 2025 12:33:29 +0100 Subject: drm/xe/sched: stop re-submitting signalled jobs MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit Customer is reporting a really subtle issue where we get random DMAR faults, hangs and other nasties for kernel migration jobs when stressing stuff like s2idle/s3/s4. The explosions seems to happen somewhere after resuming the system with splats looking something like: PM: suspend exit rfkill: input handler disabled xe 0000:00:02.0: [drm] GT0: Engine reset: engine_class=bcs, logical_mask: 0x2, guc_id=0 xe 0000:00:02.0: [drm] GT0: Timedout job: seqno=24496, lrc_seqno=24496, guc_id=0, flags=0x13 in no process [-1] xe 0000:00:02.0: [drm] GT0: Kernel-submitted job timed out The likely cause appears to be a race between suspend cancelling the worker that processes the free_job()'s, such that we still have pending jobs to be freed after the cancel. Following from this, on resume the pending_list will now contain at least one already complete job, but it looks like we call drm_sched_resubmit_jobs(), which will then call run_job() on everything still on the pending_list. But if the job was already complete, then all the resources tied to the job, like the bb itself, any memory that is being accessed, the iommu mappings etc. might be long gone since those are usually tied to the fence signalling. This scenario can be seen in ftrace when running a slightly modified xe_pm IGT (kernel was only modified to inject artificial latency into free_job to make the race easier to hit): xe_sched_job_run: dev=0000:00:02.0, fence=0xffff888276cc8540, seqno=0, lrc_seqno=0, gt=0, guc_id=0, batch_addr=0x000000146910 ... xe_exec_queue_stop: dev=0000:00:02.0, 3:0x2, gt=0, width=1, guc_id=0, guc_state=0x0, flags=0x13 xe_exec_queue_stop: dev=0000:00:02.0, 3:0x2, gt=0, width=1, guc_id=1, guc_state=0x0, flags=0x4 xe_exec_queue_stop: dev=0000:00:02.0, 4:0x1, gt=1, width=1, guc_id=0, guc_state=0x0, flags=0x3 xe_exec_queue_stop: dev=0000:00:02.0, 1:0x1, gt=1, width=1, guc_id=1, guc_state=0x0, flags=0x3 xe_exec_queue_stop: dev=0000:00:02.0, 4:0x1, gt=1, width=1, guc_id=2, guc_state=0x0, flags=0x3 xe_exec_queue_resubmit: dev=0000:00:02.0, 3:0x2, gt=0, width=1, guc_id=0, guc_state=0x0, flags=0x13 xe_sched_job_run: dev=0000:00:02.0, fence=0xffff888276cc8540, seqno=0, lrc_seqno=0, gt=0, guc_id=0, batch_addr=0x000000146910 ... ..... xe_exec_queue_memory_cat_error: dev=0000:00:02.0, 3:0x2, gt=0, width=1, guc_id=0, guc_state=0x3, flags=0x13 So the job_run() is clearly triggered twice for the same job, even though the first must have already signalled to completion during suspend. We can also see a CAT error after the re-submit. To prevent this only resubmit jobs on the pending_list that have not yet signalled. v2: - Make sure to re-arm the fence callbacks with sched_start(). v3 (Matt B): - Stop using drm_sched_resubmit_jobs(), which appears to be deprecated and just open-code a simple loop such that we skip calling run_job() on anything already signalled. Link: https://gitlab.freedesktop.org/drm/xe/kernel/-/issues/4856 Fixes: dd08ebf6c352 ("drm/xe: Introduce a new DRM driver for Intel GPUs") Signed-off-by: Matthew Auld Cc: Thomas Hellström Cc: Matthew Brost Cc: William Tseng Cc: # v6.8+ Reviewed-by: Matthew Brost Reviewed-by: Tejas Upadhyay Link: https://lore.kernel.org/r/20250528113328.289392-2-matthew.auld@intel.com (cherry picked from commit 38fafa9f392f3110d2de431432d43f4eef99cd1b) Signed-off-by: Thomas Hellström --- drivers/gpu/drm/xe/xe_gpu_scheduler.h | 10 +++++++++- 1 file changed, 9 insertions(+), 1 deletion(-) (limited to 'drivers/gpu') diff --git a/drivers/gpu/drm/xe/xe_gpu_scheduler.h b/drivers/gpu/drm/xe/xe_gpu_scheduler.h index c250ea773491..308061f0cf37 100644 --- a/drivers/gpu/drm/xe/xe_gpu_scheduler.h +++ b/drivers/gpu/drm/xe/xe_gpu_scheduler.h @@ -51,7 +51,15 @@ static inline void xe_sched_tdr_queue_imm(struct xe_gpu_scheduler *sched) static inline void xe_sched_resubmit_jobs(struct xe_gpu_scheduler *sched) { - drm_sched_resubmit_jobs(&sched->base); + struct drm_sched_job *s_job; + + list_for_each_entry(s_job, &sched->base.pending_list, list) { + struct drm_sched_fence *s_fence = s_job->s_fence; + struct dma_fence *hw_fence = s_fence->parent; + + if (hw_fence && !dma_fence_is_signaled(hw_fence)) + sched->base.ops->run_job(s_job); + } } static inline bool -- cgit v1.2.3 From 6bf4d5649230ca65725ec4793333fb5eba18d646 Mon Sep 17 00:00:00 2001 From: Daniele Ceraolo Spurio Date: Thu, 22 May 2025 15:54:03 -0700 Subject: drm/xe/pxp: Use the correct define in the set_property_funcs array MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit The define of the extension type was accidentally used instead of the one of the property itself. They're both zero, so no functional issue, but we should use the correct define for code correctness. Fixes: 41a97c4a1294 ("drm/xe/pxp/uapi: Add API to mark a BO as using PXP") Signed-off-by: Daniele Ceraolo Spurio Cc: John Harrison Reviewed-by: John Harrison Link: https://lore.kernel.org/r/20250522225401.3953243-6-daniele.ceraolospurio@intel.com (cherry picked from commit 1d891ee820fd0fbb4101eacb0d922b5050a24933) Signed-off-by: Thomas Hellström --- drivers/gpu/drm/xe/xe_bo.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) (limited to 'drivers/gpu') diff --git a/drivers/gpu/drm/xe/xe_bo.c b/drivers/gpu/drm/xe/xe_bo.c index 3c48a8c5f439..7aa2c17825da 100644 --- a/drivers/gpu/drm/xe/xe_bo.c +++ b/drivers/gpu/drm/xe/xe_bo.c @@ -2567,7 +2567,7 @@ typedef int (*xe_gem_create_set_property_fn)(struct xe_device *xe, u64 value); static const xe_gem_create_set_property_fn gem_create_set_property_funcs[] = { - [DRM_XE_GEM_CREATE_EXTENSION_SET_PROPERTY] = gem_create_set_pxp_type, + [DRM_XE_GEM_CREATE_SET_PROPERTY_PXP_TYPE] = gem_create_set_pxp_type, }; static int gem_create_user_ext_set_property(struct xe_device *xe, -- cgit v1.2.3 From 69a58ef4fa77759b0e0c2f79834fa51b00a50c0b Mon Sep 17 00:00:00 2001 From: Daniele Ceraolo Spurio Date: Thu, 22 May 2025 15:54:04 -0700 Subject: drm/xe/pxp: Clarify PXP queue creation behavior if PXP is not ready MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit The expected flow of operations when using PXP is to query the PXP status and wait for it to transition to "ready" before attempting to create an exec_queue. This flow is followed by the Mesa driver, but there is no guarantee that an incorrectly coded (or malicious) app will not attempt to create the queue first without querying the status. Therefore, we need to clarify what the expected behavior of the queue creation ioctl is in this scenario. Currently, the ioctl always fails with an -EBUSY code no matter the error, but for consistency it is better to distinguish between "failed to init" (-EIO) and "not ready" (-EBUSY), the same way the query ioctl does. Note that, while this is a change in the return code of an ioctl, the behavior of the ioctl in this particular corner case was not clearly spec'd, so no one should have been relying on it (and we know that Mesa, which is the only known userspace for this, didn't). v2: Minor rework of the doc (Rodrigo) Fixes: 72d479601d67 ("drm/xe/pxp/uapi: Add userspace and LRC support for PXP-using queues") Signed-off-by: Daniele Ceraolo Spurio Cc: John Harrison Cc: José Roberto de Souza Reviewed-by: José Roberto de Souza Reviewed-by: John Harrison Acked-by: Rodrigo Vivi Link: https://lore.kernel.org/r/20250522225401.3953243-7-daniele.ceraolospurio@intel.com (cherry picked from commit 21784ca96025b62d95b670b7639ad70ddafa69b8) Signed-off-by: Thomas Hellström --- drivers/gpu/drm/xe/xe_pxp.c | 8 ++++++-- 1 file changed, 6 insertions(+), 2 deletions(-) (limited to 'drivers/gpu') diff --git a/drivers/gpu/drm/xe/xe_pxp.c b/drivers/gpu/drm/xe/xe_pxp.c index 454ea7dc08ac..b5bc15f436fa 100644 --- a/drivers/gpu/drm/xe/xe_pxp.c +++ b/drivers/gpu/drm/xe/xe_pxp.c @@ -541,10 +541,14 @@ int xe_pxp_exec_queue_add(struct xe_pxp *pxp, struct xe_exec_queue *q) */ xe_pm_runtime_get(pxp->xe); - if (!pxp_prerequisites_done(pxp)) { - ret = -EBUSY; + /* get_readiness_status() returns 0 for in-progress and 1 for done */ + ret = xe_pxp_get_readiness_status(pxp); + if (ret <= 0) { + if (!ret) + ret = -EBUSY; goto out; } + ret = 0; wait_for_idle: /* -- cgit v1.2.3 From 2e824747cfbdf1fba88df5e5800d284b2602ae8f Mon Sep 17 00:00:00 2001 From: Matthew Auld Date: Tue, 3 Jun 2025 18:42:14 +0100 Subject: drm/xe/guc_submit: add back fix MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit Daniele noticed that the fix in commit 2d2be279f1ca ("drm/xe: fix UAF around queue destruction") looks to have been unintentionally removed as part of handling a conflict in some past merge commit. Add it back. Fixes: ac44ff7cec33 ("Merge tag 'drm-xe-fixes-2024-10-10' of https://gitlab.freedesktop.org/drm/xe/kernel into drm-fixes") Reported-by: Daniele Ceraolo Spurio Signed-off-by: Matthew Auld Cc: Matthew Brost Cc: # v6.12+ Reviewed-by: Matthew Brost Link: https://lore.kernel.org/r/20250603174213.1543579-2-matthew.auld@intel.com (cherry picked from commit 9d9fca62dc49d96f97045b6d8e7402a95f8cf92a) Signed-off-by: Thomas Hellström --- drivers/gpu/drm/xe/xe_guc_submit.c | 11 +++++++++++ 1 file changed, 11 insertions(+) (limited to 'drivers/gpu') diff --git a/drivers/gpu/drm/xe/xe_guc_submit.c b/drivers/gpu/drm/xe/xe_guc_submit.c index 369be36f7dc5..66a5a75898ac 100644 --- a/drivers/gpu/drm/xe/xe_guc_submit.c +++ b/drivers/gpu/drm/xe/xe_guc_submit.c @@ -229,6 +229,17 @@ static bool exec_queue_killed_or_banned_or_wedged(struct xe_exec_queue *q) static void guc_submit_fini(struct drm_device *drm, void *arg) { struct xe_guc *guc = arg; + struct xe_device *xe = guc_to_xe(guc); + struct xe_gt *gt = guc_to_gt(guc); + int ret; + + ret = wait_event_timeout(guc->submission_state.fini_wq, + xa_empty(&guc->submission_state.exec_queue_lookup), + HZ * 5); + + drain_workqueue(xe->destroy_wq); + + xe_gt_assert(gt, ret); xa_destroy(&guc->submission_state.exec_queue_lookup); } -- cgit v1.2.3 From 2b0a0ce0c20bbedf83f78ba5926f6cae7470cd38 Mon Sep 17 00:00:00 2001 From: Niranjana Vishwanathapura Date: Wed, 28 May 2025 22:20:32 -0700 Subject: drm/xe: Create LRC BO without VM MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit Specifying VM during lrc->bo creation requires VM's reference to be held for the lifetime of lrc->bo as it will use VM's dma reservation object. Using VM's dma reservation object for lrc->bo doesn't provide any advantage. Hence do not pass VM while creating lrc->bo. v2: Use xe_bo_unpin_map_no_vm (Matthew Brost) Fixes: 264eecdba211 ("drm/xe: Decouple xe_exec_queue and xe_lrc") Signed-off-by: Niranjana Vishwanathapura Reviewed-by: Matthew Brost Reviewed-by: Maarten Lankhorst Signed-off-by: Matthew Brost Link: https://lore.kernel.org/r/20250529052031.2429120-2-niranjana.vishwanathapura@intel.com (cherry picked from commit fbeaad071a98fef87deccee81d564de1c8e8e16d) Signed-off-by: Thomas Hellström --- drivers/gpu/drm/xe/xe_exec_queue.c | 9 --------- drivers/gpu/drm/xe/xe_lrc.c | 23 ++++------------------- 2 files changed, 4 insertions(+), 28 deletions(-) (limited to 'drivers/gpu') diff --git a/drivers/gpu/drm/xe/xe_exec_queue.c b/drivers/gpu/drm/xe/xe_exec_queue.c index 21d4ced31dd9..338487dc74c0 100644 --- a/drivers/gpu/drm/xe/xe_exec_queue.c +++ b/drivers/gpu/drm/xe/xe_exec_queue.c @@ -132,12 +132,6 @@ static int __xe_exec_queue_init(struct xe_exec_queue *q) flags |= XE_LRC_CREATE_RUNALONE; } - if (vm) { - err = xe_vm_lock(vm, true); - if (err) - return err; - } - for (i = 0; i < q->width; ++i) { q->lrc[i] = xe_lrc_create(q->hwe, q->vm, SZ_16K, q->msix_vec, flags); if (IS_ERR(q->lrc[i])) { @@ -146,9 +140,6 @@ static int __xe_exec_queue_init(struct xe_exec_queue *q) } } - if (vm) - xe_vm_unlock(vm); - err = q->ops->init(q); if (err) goto err_lrc; diff --git a/drivers/gpu/drm/xe/xe_lrc.c b/drivers/gpu/drm/xe/xe_lrc.c index 855c8acaf3f1..e1db6f2a1ad0 100644 --- a/drivers/gpu/drm/xe/xe_lrc.c +++ b/drivers/gpu/drm/xe/xe_lrc.c @@ -876,10 +876,7 @@ static void xe_lrc_set_ppgtt(struct xe_lrc *lrc, struct xe_vm *vm) static void xe_lrc_finish(struct xe_lrc *lrc) { xe_hw_fence_ctx_finish(&lrc->fence_ctx); - xe_bo_lock(lrc->bo, false); - xe_bo_unpin(lrc->bo); - xe_bo_unlock(lrc->bo); - xe_bo_put(lrc->bo); + xe_bo_unpin_map_no_vm(lrc->bo); } #define PVC_CTX_ASID (0x2e + 1) @@ -914,7 +911,7 @@ static int xe_lrc_init(struct xe_lrc *lrc, struct xe_hw_engine *hwe, * FIXME: Perma-pinning LRC as we don't yet support moving GGTT address * via VM bind calls. */ - lrc->bo = xe_bo_create_pin_map(xe, tile, vm, lrc_size, + lrc->bo = xe_bo_create_pin_map(xe, tile, NULL, lrc_size, ttm_bo_type_kernel, bo_flags); if (IS_ERR(lrc->bo)) @@ -1676,9 +1673,6 @@ struct xe_lrc_snapshot *xe_lrc_snapshot_capture(struct xe_lrc *lrc) if (!snapshot) return NULL; - if (lrc->bo->vm) - xe_vm_get(lrc->bo->vm); - snapshot->context_desc = xe_lrc_ggtt_addr(lrc); snapshot->ring_addr = __xe_lrc_ring_ggtt_addr(lrc); snapshot->indirect_context_desc = xe_lrc_indirect_ring_ggtt_addr(lrc); @@ -1700,14 +1694,12 @@ struct xe_lrc_snapshot *xe_lrc_snapshot_capture(struct xe_lrc *lrc) void xe_lrc_snapshot_capture_delayed(struct xe_lrc_snapshot *snapshot) { struct xe_bo *bo; - struct xe_vm *vm; struct iosys_map src; if (!snapshot) return; bo = snapshot->lrc_bo; - vm = bo->vm; snapshot->lrc_bo = NULL; snapshot->lrc_snapshot = kvmalloc(snapshot->lrc_size, GFP_KERNEL); @@ -1727,8 +1719,6 @@ void xe_lrc_snapshot_capture_delayed(struct xe_lrc_snapshot *snapshot) xe_bo_unlock(bo); put_bo: xe_bo_put(bo); - if (vm) - xe_vm_put(vm); } void xe_lrc_snapshot_print(struct xe_lrc_snapshot *snapshot, struct drm_printer *p) @@ -1781,14 +1771,9 @@ void xe_lrc_snapshot_free(struct xe_lrc_snapshot *snapshot) return; kvfree(snapshot->lrc_snapshot); - if (snapshot->lrc_bo) { - struct xe_vm *vm; - - vm = snapshot->lrc_bo->vm; + if (snapshot->lrc_bo) xe_bo_put(snapshot->lrc_bo); - if (vm) - xe_vm_put(vm); - } + kfree(snapshot); } -- cgit v1.2.3 From 7c7c5cb5b5bf9d8ccc6a51b28687c9e7ff7f1890 Mon Sep 17 00:00:00 2001 From: Maciej Patelczyk Date: Fri, 30 May 2025 15:56:27 +0200 Subject: drm/xe: remove unmatched xe_vm_unlock() from __xe_exec_queue_init() MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit There is unmatched xe_vm_unlock() in the __xe_exec_queue_init(). Leftover from commit fbeaad071a98 ("drm/xe: Create LRC BO without VM") Fixes: 2b0a0ce0c20b ("drm/xe: Create LRC BO without VM") Signed-off-by: Maciej Patelczyk Reviewed-by: Jonathan Cavitt Reviewed-by: Matthew Brost Signed-off-by: Niranjana Vishwanathapura Link: https://lore.kernel.org/r/20250530135627.2821612-1-maciej.patelczyk@intel.com (cherry picked from commit 28b996ce73982a44fa86736ca0e3684cb1ae8b24) Signed-off-by: Thomas Hellström --- drivers/gpu/drm/xe/xe_exec_queue.c | 6 +----- 1 file changed, 1 insertion(+), 5 deletions(-) (limited to 'drivers/gpu') diff --git a/drivers/gpu/drm/xe/xe_exec_queue.c b/drivers/gpu/drm/xe/xe_exec_queue.c index 338487dc74c0..0161a80e92bc 100644 --- a/drivers/gpu/drm/xe/xe_exec_queue.c +++ b/drivers/gpu/drm/xe/xe_exec_queue.c @@ -114,7 +114,6 @@ static struct xe_exec_queue *__xe_exec_queue_alloc(struct xe_device *xe, static int __xe_exec_queue_init(struct xe_exec_queue *q) { - struct xe_vm *vm = q->vm; int i, err; u32 flags = 0; @@ -136,7 +135,7 @@ static int __xe_exec_queue_init(struct xe_exec_queue *q) q->lrc[i] = xe_lrc_create(q->hwe, q->vm, SZ_16K, q->msix_vec, flags); if (IS_ERR(q->lrc[i])) { err = PTR_ERR(q->lrc[i]); - goto err_unlock; + goto err_lrc; } } @@ -146,9 +145,6 @@ static int __xe_exec_queue_init(struct xe_exec_queue *q) return 0; -err_unlock: - if (vm) - xe_vm_unlock(vm); err_lrc: for (i = i - 1; i >= 0; --i) xe_lrc_put(q->lrc[i]); -- cgit v1.2.3 From 04c8970771b4f1f39bb8453a2eeb188c4d5edbd6 Mon Sep 17 00:00:00 2001 From: Chen Ni Date: Tue, 3 Jun 2025 14:10:27 +0800 Subject: drm/nouveau/vfn/r535: Convert comma to semicolon Replace comma between expressions with semicolons. Using a ',' in place of a ';' can have unintended side effects. Although that is not the case here, it is seems best to use ';' unless ',' is intended. Found by inspection. No functional change intended. Compile tested only. Signed-off-by: Chen Ni Fixes: cd3c62282b61 ("drm/nouveau/gsp: add usermode class id to gpu hal") Signed-off-by: Dave Airlie Link: https://lore.kernel.org/r/20250603061027.1310267-1-nichen@iscas.ac.cn --- drivers/gpu/drm/nouveau/nvkm/subdev/vfn/r535.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) (limited to 'drivers/gpu') diff --git a/drivers/gpu/drm/nouveau/nvkm/subdev/vfn/r535.c b/drivers/gpu/drm/nouveau/nvkm/subdev/vfn/r535.c index 9446049642e1..d294844d9eae 100644 --- a/drivers/gpu/drm/nouveau/nvkm/subdev/vfn/r535.c +++ b/drivers/gpu/drm/nouveau/nvkm/subdev/vfn/r535.c @@ -42,7 +42,7 @@ r535_vfn_new(const struct nvkm_vfn_func *hw, return -ENOMEM; rm->dtor = r535_vfn_dtor; - rm->intr = &tu102_vfn_intr, + rm->intr = &tu102_vfn_intr; rm->user.addr = 0x030000; rm->user.size = 0x010000; rm->user.base.minver = -1; -- cgit v1.2.3