diff options
Diffstat (limited to 'drivers/gpu/drm/amd/amdgpu')
123 files changed, 3257 insertions, 1588 deletions
diff --git a/drivers/gpu/drm/amd/amdgpu/Kconfig b/drivers/gpu/drm/amd/amdgpu/Kconfig index 7777d55275de..5fcd510f1abb 100644 --- a/drivers/gpu/drm/amd/amdgpu/Kconfig +++ b/drivers/gpu/drm/amd/amdgpu/Kconfig @@ -1,4 +1,33 @@ # SPDX-License-Identifier: MIT + +config DRM_AMDGPU + tristate "AMD GPU" + depends on DRM && PCI && MMU + select FW_LOADER + select DRM_DISPLAY_DP_HELPER + select DRM_DISPLAY_HDMI_HELPER + select DRM_DISPLAY_HELPER + select DRM_KMS_HELPER + select DRM_SCHED + select DRM_TTM + select DRM_TTM_HELPER + select POWER_SUPPLY + select HWMON + select BACKLIGHT_CLASS_DEVICE + select INTERVAL_TREE + select DRM_BUDDY + # amdgpu depends on ACPI_VIDEO when ACPI is enabled, for select to work + # ACPI_VIDEO's dependencies must also be selected. + select INPUT if ACPI + select ACPI_VIDEO if ACPI + # On x86 ACPI_VIDEO also needs ACPI_WMI + select X86_PLATFORM_DEVICES if ACPI && X86 + select ACPI_WMI if ACPI && X86 + help + Choose this option if you have a recent AMD Radeon graphics card. + + If M is selected, the module will be called amdgpu. + config DRM_AMDGPU_SI bool "Enable amdgpu support for SI parts" depends on DRM_AMDGPU diff --git a/drivers/gpu/drm/amd/amdgpu/Makefile b/drivers/gpu/drm/amd/amdgpu/Makefile index 6ad39cf71bdd..798d0e9a60b7 100644 --- a/drivers/gpu/drm/amd/amdgpu/Makefile +++ b/drivers/gpu/drm/amd/amdgpu/Makefile @@ -58,7 +58,8 @@ amdgpu-y += amdgpu_device.o amdgpu_kms.o \ amdgpu_vm_sdma.o amdgpu_discovery.o amdgpu_ras_eeprom.o amdgpu_nbio.o \ amdgpu_umc.o smu_v11_0_i2c.o amdgpu_fru_eeprom.o amdgpu_rap.o \ amdgpu_fw_attestation.o amdgpu_securedisplay.o \ - amdgpu_eeprom.o amdgpu_mca.o amdgpu_psp_ta.o amdgpu_lsdma.o + amdgpu_eeprom.o amdgpu_mca.o amdgpu_psp_ta.o amdgpu_lsdma.o \ + amdgpu_ring_mux.o amdgpu-$(CONFIG_PROC_FS) += amdgpu_fdinfo.o @@ -250,7 +251,7 @@ endif amdgpu-$(CONFIG_COMPAT) += amdgpu_ioc32.o amdgpu-$(CONFIG_VGA_SWITCHEROO) += amdgpu_atpx_handler.o amdgpu-$(CONFIG_ACPI) += amdgpu_acpi.o -amdgpu-$(CONFIG_HMM_MIRROR) += amdgpu_mn.o +amdgpu-$(CONFIG_HMM_MIRROR) += amdgpu_hmm.o include $(FULL_AMD_PATH)/pm/Makefile diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu.h b/drivers/gpu/drm/amd/amdgpu/amdgpu.h index 2eca58220550..e3e2e6e3b485 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu.h +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu.h @@ -82,7 +82,6 @@ #include "amdgpu_vce.h" #include "amdgpu_vcn.h" #include "amdgpu_jpeg.h" -#include "amdgpu_mn.h" #include "amdgpu_gmc.h" #include "amdgpu_gfx.h" #include "amdgpu_sdma.h" @@ -196,6 +195,7 @@ extern int amdgpu_emu_mode; extern uint amdgpu_smu_memory_pool_size; extern int amdgpu_smu_pptable_id; extern uint amdgpu_dc_feature_mask; +extern uint amdgpu_freesync_vid_mode; extern uint amdgpu_dc_debug_mask; extern uint amdgpu_dc_visual_confirm; extern uint amdgpu_dm_abm_level; @@ -219,10 +219,12 @@ extern int amdgpu_use_xgmi_p2p; extern int sched_policy; extern bool debug_evictions; extern bool no_system_mem_limit; +extern int halt_if_hws_hang; #else static const int __maybe_unused sched_policy = KFD_SCHED_POLICY_HWS; static const bool __maybe_unused debug_evictions; /* = false */ static const bool __maybe_unused no_system_mem_limit; +static const int __maybe_unused halt_if_hws_hang; #endif #ifdef CONFIG_HSA_AMD_P2P extern bool pcie_p2p; @@ -675,7 +677,7 @@ enum amd_hw_ip_block_type { MAX_HWIP }; -#define HWIP_MAX_INSTANCE 11 +#define HWIP_MAX_INSTANCE 28 #define HW_ID_MAX 300 #define IP_VERSION(mj, mn, rv) (((mj) << 16) | ((mn) << 8) | (rv)) @@ -1063,6 +1065,7 @@ struct amdgpu_device { struct work_struct reset_work; bool job_hang; + bool dc_enabled; }; static inline struct amdgpu_device *drm_to_adev(struct drm_device *ddev) @@ -1120,6 +1123,8 @@ void amdgpu_device_indirect_wreg64(struct amdgpu_device *adev, bool amdgpu_device_asic_has_dc_support(enum amd_asic_type asic_type); bool amdgpu_device_has_dc_support(struct amdgpu_device *adev); +void amdgpu_device_set_sriov_virtual_display(struct amdgpu_device *adev); + int amdgpu_device_pre_asic_reset(struct amdgpu_device *adev, struct amdgpu_reset_context *reset_context); diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_acpi.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_acpi.c index b14800ac179e..57b5e11446c6 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_acpi.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_acpi.c @@ -847,7 +847,7 @@ int amdgpu_acpi_init(struct amdgpu_device *adev) struct amdgpu_atif *atif = &amdgpu_acpi_priv.atif; if (atif->notifications.brightness_change) { - if (amdgpu_device_has_dc_support(adev)) { + if (adev->dc_enabled) { #if defined(CONFIG_DRM_AMD_DC) struct amdgpu_display_manager *dm = &adev->dm; diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd.c index 5d9a34601a1a..f99d4873bf22 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd.c @@ -195,7 +195,7 @@ void amdgpu_amdkfd_device_init(struct amdgpu_device *adev) } adev->kfd.init_complete = kgd2kfd_device_init(adev->kfd.dev, - adev_to_drm(adev), &gpu_resources); + &gpu_resources); amdgpu_amdkfd_total_mem_size += adev->gmc.real_vram_size; @@ -673,7 +673,7 @@ int amdgpu_amdkfd_submit_ib(struct amdgpu_device *adev, goto err; } - ret = amdgpu_job_alloc(adev, 1, &job, NULL); + ret = amdgpu_job_alloc(adev, NULL, NULL, NULL, 1, &job); if (ret) goto err; @@ -760,9 +760,7 @@ bool amdgpu_amdkfd_have_atomics_support(struct amdgpu_device *adev) void amdgpu_amdkfd_ras_poison_consumption_handler(struct amdgpu_device *adev, bool reset) { - struct ras_err_data err_data = {0, 0, 0, NULL}; - - amdgpu_umc_poison_handler(adev, &err_data, reset); + amdgpu_umc_poison_handler(adev, reset); } bool amdgpu_amdkfd_ras_query_utcl2_poison_status(struct amdgpu_device *adev) diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd.h b/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd.h index 647220a8762d..0040deaf8a83 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd.h +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd.h @@ -29,6 +29,7 @@ #include <linux/mm.h> #include <linux/kthread.h> #include <linux/workqueue.h> +#include <linux/mmu_notifier.h> #include <kgd_kfd_interface.h> #include <drm/ttm/ttm_execbuf_util.h> #include "amdgpu_sync.h" @@ -65,6 +66,7 @@ struct kgd_mem { struct mutex lock; struct amdgpu_bo *bo; struct dma_buf *dmabuf; + struct hmm_range *range; struct list_head attachments; /* protected by amdkfd_process_info.lock */ struct ttm_validate_buffer validate_list; @@ -75,7 +77,7 @@ struct kgd_mem { uint32_t alloc_flags; - atomic_t invalid; + uint32_t invalid; struct amdkfd_process_info *process_info; struct amdgpu_sync sync; @@ -131,7 +133,8 @@ struct amdkfd_process_info { struct amdgpu_amdkfd_fence *eviction_fence; /* MMU-notifier related fields */ - atomic_t evicted_bos; + struct mutex notifier_lock; + uint32_t evicted_bos; struct delayed_work restore_userptr_work; struct pid *pid; bool block_mmu_notifications; @@ -180,7 +183,8 @@ int kfd_debugfs_kfd_mem_limits(struct seq_file *m, void *data); bool amdkfd_fence_check_mm(struct dma_fence *f, struct mm_struct *mm); struct amdgpu_amdkfd_fence *to_amdgpu_amdkfd_fence(struct dma_fence *f); int amdgpu_amdkfd_remove_fence_on_pt_pd_bos(struct amdgpu_bo *bo); -int amdgpu_amdkfd_evict_userptr(struct kgd_mem *mem, struct mm_struct *mm); +int amdgpu_amdkfd_evict_userptr(struct mmu_interval_notifier *mni, + unsigned long cur_seq, struct kgd_mem *mem); #else static inline bool amdkfd_fence_check_mm(struct dma_fence *f, struct mm_struct *mm) @@ -201,7 +205,8 @@ int amdgpu_amdkfd_remove_fence_on_pt_pd_bos(struct amdgpu_bo *bo) } static inline -int amdgpu_amdkfd_evict_userptr(struct kgd_mem *mem, struct mm_struct *mm) +int amdgpu_amdkfd_evict_userptr(struct mmu_interval_notifier *mni, + unsigned long cur_seq, struct kgd_mem *mem) { return 0; } @@ -265,8 +270,10 @@ int amdgpu_amdkfd_get_pcie_bandwidth_mbytes(struct amdgpu_device *adev, bool is_ (&((struct amdgpu_fpriv *) \ ((struct drm_file *)(drm_priv))->driver_priv)->vm) +int amdgpu_amdkfd_gpuvm_set_vm_pasid(struct amdgpu_device *adev, + struct file *filp, u32 pasid); int amdgpu_amdkfd_gpuvm_acquire_process_vm(struct amdgpu_device *adev, - struct file *filp, u32 pasid, + struct file *filp, void **process_info, struct dma_fence **ef); void amdgpu_amdkfd_gpuvm_release_process_vm(struct amdgpu_device *adev, @@ -353,7 +360,6 @@ int kgd2kfd_init(void); void kgd2kfd_exit(void); struct kfd_dev *kgd2kfd_probe(struct amdgpu_device *adev, bool vf); bool kgd2kfd_device_init(struct kfd_dev *kfd, - struct drm_device *ddev, const struct kgd2kfd_shared_resources *gpu_resources); void kgd2kfd_device_exit(struct kfd_dev *kfd); void kgd2kfd_suspend(struct kfd_dev *kfd, bool run_pm); @@ -381,7 +387,7 @@ struct kfd_dev *kgd2kfd_probe(struct amdgpu_device *adev, bool vf) } static inline -bool kgd2kfd_device_init(struct kfd_dev *kfd, struct drm_device *ddev, +bool kgd2kfd_device_init(struct kfd_dev *kfd, const struct kgd2kfd_shared_resources *gpu_resources) { return false; diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gfx_v9.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gfx_v9.c index 81e3b528bbc9..e92b93557c13 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gfx_v9.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gfx_v9.c @@ -787,7 +787,7 @@ void kgd_gfx_v9_get_cu_occupancy(struct amdgpu_device *adev, int pasid, for (se_idx = 0; se_idx < se_cnt; se_idx++) { for (sh_idx = 0; sh_idx < sh_cnt; sh_idx++) { - gfx_v9_0_select_se_sh(adev, se_idx, sh_idx, 0xffffffff); + amdgpu_gfx_select_se_sh(adev, se_idx, sh_idx, 0xffffffff); queue_map = RREG32_SOC15(GC, 0, mmSPI_CSQ_WF_ACTIVE_STATUS); /* @@ -820,7 +820,7 @@ void kgd_gfx_v9_get_cu_occupancy(struct amdgpu_device *adev, int pasid, } } - gfx_v9_0_select_se_sh(adev, 0xffffffff, 0xffffffff, 0xffffffff); + amdgpu_gfx_select_se_sh(adev, 0xffffffff, 0xffffffff, 0xffffffff); soc15_grbm_select(adev, 0, 0, 0, 0); unlock_spi_csq_mutexes(adev); diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gpuvm.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gpuvm.c index 1f76e27f1a35..3b5c53712d31 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gpuvm.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gpuvm.c @@ -29,6 +29,7 @@ #include "amdgpu_object.h" #include "amdgpu_gem.h" #include "amdgpu_vm.h" +#include "amdgpu_hmm.h" #include "amdgpu_amdkfd.h" #include "amdgpu_dma_buf.h" #include <uapi/linux/kfd_ioctl.h> @@ -403,63 +404,15 @@ static int vm_update_pds(struct amdgpu_vm *vm, struct amdgpu_sync *sync) static uint64_t get_pte_flags(struct amdgpu_device *adev, struct kgd_mem *mem) { - struct amdgpu_device *bo_adev = amdgpu_ttm_adev(mem->bo->tbo.bdev); - bool coherent = mem->alloc_flags & KFD_IOC_ALLOC_MEM_FLAGS_COHERENT; - bool uncached = mem->alloc_flags & KFD_IOC_ALLOC_MEM_FLAGS_UNCACHED; - uint32_t mapping_flags; - uint64_t pte_flags; - bool snoop = false; + uint32_t mapping_flags = AMDGPU_VM_PAGE_READABLE | + AMDGPU_VM_MTYPE_DEFAULT; - mapping_flags = AMDGPU_VM_PAGE_READABLE; if (mem->alloc_flags & KFD_IOC_ALLOC_MEM_FLAGS_WRITABLE) mapping_flags |= AMDGPU_VM_PAGE_WRITEABLE; if (mem->alloc_flags & KFD_IOC_ALLOC_MEM_FLAGS_EXECUTABLE) mapping_flags |= AMDGPU_VM_PAGE_EXECUTABLE; - switch (adev->asic_type) { - case CHIP_ARCTURUS: - case CHIP_ALDEBARAN: - if (mem->alloc_flags & KFD_IOC_ALLOC_MEM_FLAGS_VRAM) { - if (bo_adev == adev) { - if (uncached) - mapping_flags |= AMDGPU_VM_MTYPE_UC; - else if (coherent) - mapping_flags |= AMDGPU_VM_MTYPE_CC; - else - mapping_flags |= AMDGPU_VM_MTYPE_RW; - if (adev->asic_type == CHIP_ALDEBARAN && - adev->gmc.xgmi.connected_to_cpu) - snoop = true; - } else { - if (uncached || coherent) - mapping_flags |= AMDGPU_VM_MTYPE_UC; - else - mapping_flags |= AMDGPU_VM_MTYPE_NC; - if (amdgpu_xgmi_same_hive(adev, bo_adev)) - snoop = true; - } - } else { - if (uncached || coherent) - mapping_flags |= AMDGPU_VM_MTYPE_UC; - else - mapping_flags |= AMDGPU_VM_MTYPE_NC; - snoop = true; - } - break; - default: - if (uncached || coherent) - mapping_flags |= AMDGPU_VM_MTYPE_UC; - else - mapping_flags |= AMDGPU_VM_MTYPE_NC; - - if (!(mem->alloc_flags & KFD_IOC_ALLOC_MEM_FLAGS_VRAM)) - snoop = true; - } - - pte_flags = amdgpu_gem_va_map_flags(adev, mapping_flags); - pte_flags |= snoop ? AMDGPU_PTE_SNOOPED : 0; - - return pte_flags; + return amdgpu_gem_va_map_flags(adev, mapping_flags); } /** @@ -997,7 +950,7 @@ static int init_user_pages(struct kgd_mem *mem, uint64_t user_addr, goto out; } - ret = amdgpu_mn_register(bo, user_addr); + ret = amdgpu_hmm_register(bo, user_addr); if (ret) { pr_err("%s: Failed to register MMU notifier: %d\n", __func__, ret); @@ -1011,7 +964,9 @@ static int init_user_pages(struct kgd_mem *mem, uint64_t user_addr, * later stage when it is scheduled by another ioctl called by * CRIU master process for the target pid for restore. */ - atomic_inc(&mem->invalid); + mutex_lock(&process_info->notifier_lock); + mem->invalid++; + mutex_unlock(&process_info->notifier_lock); mutex_unlock(&process_info->lock); return 0; } @@ -1037,7 +992,7 @@ release_out: amdgpu_ttm_tt_get_user_pages_done(bo->tbo.ttm, range); unregister_out: if (ret) - amdgpu_mn_unregister(bo); + amdgpu_hmm_unregister(bo); out: mutex_unlock(&process_info->lock); return ret; @@ -1348,6 +1303,7 @@ static int init_kfd_vm(struct amdgpu_vm *vm, void **process_info, return -ENOMEM; mutex_init(&info->lock); + mutex_init(&info->notifier_lock); INIT_LIST_HEAD(&info->vm_list_head); INIT_LIST_HEAD(&info->kfd_bo_list); INIT_LIST_HEAD(&info->userptr_valid_list); @@ -1364,7 +1320,6 @@ static int init_kfd_vm(struct amdgpu_vm *vm, void **process_info, } info->pid = get_task_pid(current->group_leader, PIDTYPE_PID); - atomic_set(&info->evicted_bos, 0); INIT_DELAYED_WORK(&info->restore_userptr_work, amdgpu_amdkfd_restore_userptr_worker); @@ -1419,6 +1374,7 @@ reserve_pd_fail: put_pid(info->pid); create_evict_fence_fail: mutex_destroy(&info->lock); + mutex_destroy(&info->notifier_lock); kfree(info); } return ret; @@ -1473,10 +1429,9 @@ static void amdgpu_amdkfd_gpuvm_unpin_bo(struct amdgpu_bo *bo) amdgpu_bo_unreserve(bo); } -int amdgpu_amdkfd_gpuvm_acquire_process_vm(struct amdgpu_device *adev, - struct file *filp, u32 pasid, - void **process_info, - struct dma_fence **ef) +int amdgpu_amdkfd_gpuvm_set_vm_pasid(struct amdgpu_device *adev, + struct file *filp, u32 pasid) + { struct amdgpu_fpriv *drv_priv; struct amdgpu_vm *avm; @@ -1487,10 +1442,6 @@ int amdgpu_amdkfd_gpuvm_acquire_process_vm(struct amdgpu_device *adev, return ret; avm = &drv_priv->vm; - /* Already a compute VM? */ - if (avm->process_info) - return -EINVAL; - /* Free the original amdgpu allocated pasid, * will be replaced with kfd allocated pasid. */ @@ -1499,14 +1450,36 @@ int amdgpu_amdkfd_gpuvm_acquire_process_vm(struct amdgpu_device *adev, amdgpu_vm_set_pasid(adev, avm, 0); } - /* Convert VM into a compute VM */ - ret = amdgpu_vm_make_compute(adev, avm); + ret = amdgpu_vm_set_pasid(adev, avm, pasid); if (ret) return ret; - ret = amdgpu_vm_set_pasid(adev, avm, pasid); + return 0; +} + +int amdgpu_amdkfd_gpuvm_acquire_process_vm(struct amdgpu_device *adev, + struct file *filp, + void **process_info, + struct dma_fence **ef) +{ + struct amdgpu_fpriv *drv_priv; + struct amdgpu_vm *avm; + int ret; + + ret = amdgpu_file_to_fpriv(filp, &drv_priv); + if (ret) + return ret; + avm = &drv_priv->vm; + + /* Already a compute VM? */ + if (avm->process_info) + return -EINVAL; + + /* Convert VM into a compute VM */ + ret = amdgpu_vm_make_compute(adev, avm); if (ret) return ret; + /* Initialize KFD part of the VM and process info */ ret = init_kfd_vm(avm, process_info, ef); if (ret) @@ -1543,6 +1516,7 @@ void amdgpu_amdkfd_gpuvm_destroy_cb(struct amdgpu_device *adev, cancel_delayed_work_sync(&process_info->restore_userptr_work); put_pid(process_info->pid); mutex_destroy(&process_info->lock); + mutex_destroy(&process_info->notifier_lock); kfree(process_info); } } @@ -1595,7 +1569,9 @@ int amdgpu_amdkfd_criu_resume(void *p) mutex_lock(&pinfo->lock); pr_debug("scheduling work\n"); - atomic_inc(&pinfo->evicted_bos); + mutex_lock(&pinfo->notifier_lock); + pinfo->evicted_bos++; + mutex_unlock(&pinfo->notifier_lock); if (!READ_ONCE(pinfo->block_mmu_notifications)) { ret = -EINVAL; goto out_unlock; @@ -1672,6 +1648,11 @@ int amdgpu_amdkfd_gpuvm_alloc_memory_of_gpu( } } + if (flags & KFD_IOC_ALLOC_MEM_FLAGS_COHERENT) + alloc_flags |= AMDGPU_GEM_CREATE_COHERENT; + if (flags & KFD_IOC_ALLOC_MEM_FLAGS_UNCACHED) + alloc_flags |= AMDGPU_GEM_CREATE_UNCACHED; + *mem = kzalloc(sizeof(struct kgd_mem), GFP_KERNEL); if (!*mem) { ret = -ENOMEM; @@ -1815,8 +1796,13 @@ int amdgpu_amdkfd_gpuvm_free_memory_of_gpu( list_del(&bo_list_entry->head); mutex_unlock(&process_info->lock); - /* No more MMU notifiers */ - amdgpu_mn_unregister(mem->bo); + /* Cleanup user pages and MMU notifiers */ + if (amdgpu_ttm_tt_get_usermm(mem->bo->tbo.ttm)) { + amdgpu_hmm_unregister(mem->bo); + mutex_lock(&process_info->notifier_lock); + amdgpu_ttm_tt_discard_user_pages(mem->bo->tbo.ttm, mem->range); + mutex_unlock(&process_info->notifier_lock); + } ret = reserve_bo_and_cond_vms(mem, NULL, BO_VM_ALL, &ctx); if (unlikely(ret)) @@ -1906,14 +1892,14 @@ int amdgpu_amdkfd_gpuvm_map_memory_to_gpu( */ mutex_lock(&mem->process_info->lock); - /* Lock mmap-sem. If we find an invalid userptr BO, we can be + /* Lock notifier lock. If we find an invalid userptr BO, we can be * sure that the MMU notifier is no longer running * concurrently and the queues are actually stopped */ if (amdgpu_ttm_tt_get_usermm(bo->tbo.ttm)) { - mmap_write_lock(current->mm); - is_invalid_userptr = atomic_read(&mem->invalid); - mmap_write_unlock(current->mm); + mutex_lock(&mem->process_info->notifier_lock); + is_invalid_userptr = !!mem->invalid; + mutex_unlock(&mem->process_info->notifier_lock); } mutex_lock(&mem->lock); @@ -2113,7 +2099,7 @@ int amdgpu_amdkfd_map_gtt_bo_to_gart(struct amdgpu_device *adev, struct amdgpu_b } amdgpu_amdkfd_remove_eviction_fence( - bo, bo->kfd_bo->process_info->eviction_fence); + bo, bo->vm_bo->vm->process_info->eviction_fence); amdgpu_bo_unreserve(bo); @@ -2256,7 +2242,7 @@ int amdgpu_amdkfd_gpuvm_import_dmabuf(struct amdgpu_device *adev, ret = drm_vma_node_allow(&obj->vma_node, drm_priv); if (ret) { - kfree(mem); + kfree(*mem); return ret; } @@ -2293,34 +2279,38 @@ int amdgpu_amdkfd_gpuvm_import_dmabuf(struct amdgpu_device *adev, * * Runs in MMU notifier, may be in RECLAIM_FS context. This means it * cannot do any memory allocations, and cannot take any locks that - * are held elsewhere while allocating memory. Therefore this is as - * simple as possible, using atomic counters. + * are held elsewhere while allocating memory. * * It doesn't do anything to the BO itself. The real work happens in * restore, where we get updated page addresses. This function only * ensures that GPU access to the BO is stopped. */ -int amdgpu_amdkfd_evict_userptr(struct kgd_mem *mem, - struct mm_struct *mm) +int amdgpu_amdkfd_evict_userptr(struct mmu_interval_notifier *mni, + unsigned long cur_seq, struct kgd_mem *mem) { struct amdkfd_process_info *process_info = mem->process_info; - int evicted_bos; int r = 0; - /* Do not process MMU notifications until stage-4 IOCTL is received */ + /* Do not process MMU notifications during CRIU restore until + * KFD_CRIU_OP_RESUME IOCTL is received + */ if (READ_ONCE(process_info->block_mmu_notifications)) return 0; - atomic_inc(&mem->invalid); - evicted_bos = atomic_inc_return(&process_info->evicted_bos); - if (evicted_bos == 1) { + mutex_lock(&process_info->notifier_lock); + mmu_interval_set_seq(mni, cur_seq); + + mem->invalid++; + if (++process_info->evicted_bos == 1) { /* First eviction, stop the queues */ - r = kgd2kfd_quiesce_mm(mm, KFD_QUEUE_EVICTION_TRIGGER_USERPTR); + r = kgd2kfd_quiesce_mm(mni->mm, + KFD_QUEUE_EVICTION_TRIGGER_USERPTR); if (r) pr_err("Failed to quiesce KFD\n"); schedule_delayed_work(&process_info->restore_userptr_work, msecs_to_jiffies(AMDGPU_USERPTR_RESTORE_DELAY_MS)); } + mutex_unlock(&process_info->notifier_lock); return r; } @@ -2337,54 +2327,58 @@ static int update_invalid_user_pages(struct amdkfd_process_info *process_info, struct kgd_mem *mem, *tmp_mem; struct amdgpu_bo *bo; struct ttm_operation_ctx ctx = { false, false }; - int invalid, ret; + uint32_t invalid; + int ret = 0; - /* Move all invalidated BOs to the userptr_inval_list and - * release their user pages by migration to the CPU domain - */ + mutex_lock(&process_info->notifier_lock); + + /* Move all invalidated BOs to the userptr_inval_list */ list_for_each_entry_safe(mem, tmp_mem, &process_info->userptr_valid_list, - validate_list.head) { - if (!atomic_read(&mem->invalid)) - continue; /* BO is still valid */ - - bo = mem->bo; - - if (amdgpu_bo_reserve(bo, true)) - return -EAGAIN; - amdgpu_bo_placement_from_domain(bo, AMDGPU_GEM_DOMAIN_CPU); - ret = ttm_bo_validate(&bo->tbo, &bo->placement, &ctx); - amdgpu_bo_unreserve(bo); - if (ret) { - pr_err("%s: Failed to invalidate userptr BO\n", - __func__); - return -EAGAIN; - } - - list_move_tail(&mem->validate_list.head, - &process_info->userptr_inval_list); - } - - if (list_empty(&process_info->userptr_inval_list)) - return 0; /* All evicted userptr BOs were freed */ + validate_list.head) + if (mem->invalid) + list_move_tail(&mem->validate_list.head, + &process_info->userptr_inval_list); /* Go through userptr_inval_list and update any invalid user_pages */ list_for_each_entry(mem, &process_info->userptr_inval_list, validate_list.head) { - struct hmm_range *range; - - invalid = atomic_read(&mem->invalid); + invalid = mem->invalid; if (!invalid) /* BO hasn't been invalidated since the last - * revalidation attempt. Keep its BO list. + * revalidation attempt. Keep its page list. */ continue; bo = mem->bo; + amdgpu_ttm_tt_discard_user_pages(bo->tbo.ttm, mem->range); + mem->range = NULL; + + /* BO reservations and getting user pages (hmm_range_fault) + * must happen outside the notifier lock + */ + mutex_unlock(&process_info->notifier_lock); + + /* Move the BO to system (CPU) domain if necessary to unmap + * and free the SG table + */ + if (bo->tbo.resource->mem_type != TTM_PL_SYSTEM) { + if (amdgpu_bo_reserve(bo, true)) + return -EAGAIN; + amdgpu_bo_placement_from_domain(bo, AMDGPU_GEM_DOMAIN_CPU); + ret = ttm_bo_validate(&bo->tbo, &bo->placement, &ctx); + amdgpu_bo_unreserve(bo); + if (ret) { + pr_err("%s: Failed to invalidate userptr BO\n", + __func__); + return -EAGAIN; + } + } + /* Get updated user pages */ ret = amdgpu_ttm_tt_get_user_pages(bo, bo->tbo.ttm->pages, - &range); + &mem->range); if (ret) { pr_debug("Failed %d to get user pages\n", ret); @@ -2397,30 +2391,32 @@ static int update_invalid_user_pages(struct amdkfd_process_info *process_info, */ if (ret != -EFAULT) return ret; - } else { - /* - * FIXME: Cannot ignore the return code, must hold - * notifier_lock - */ - amdgpu_ttm_tt_get_user_pages_done(bo->tbo.ttm, range); + ret = 0; } + mutex_lock(&process_info->notifier_lock); + /* Mark the BO as valid unless it was invalidated * again concurrently. */ - if (atomic_cmpxchg(&mem->invalid, invalid, 0) != invalid) - return -EAGAIN; + if (mem->invalid != invalid) { + ret = -EAGAIN; + goto unlock_out; + } + mem->invalid = 0; } - return 0; +unlock_out: + mutex_unlock(&process_info->notifier_lock); + + return ret; } /* Validate invalid userptr BOs * - * Validates BOs on the userptr_inval_list, and moves them back to the - * userptr_valid_list. Also updates GPUVM page tables with new page - * addresses and waits for the page table updates to complete. + * Validates BOs on the userptr_inval_list. Also updates GPUVM page tables + * with new page addresses and waits for the page table updates to complete. */ static int validate_invalid_user_pages(struct amdkfd_process_info *process_info) { @@ -2491,9 +2487,6 @@ static int validate_invalid_user_pages(struct amdkfd_process_info *process_info) } } - list_move_tail(&mem->validate_list.head, - &process_info->userptr_valid_list); - /* Update mapping. If the BO was not validated * (because we couldn't get user pages), this will * clear the page table entries, which will result in @@ -2509,7 +2502,9 @@ static int validate_invalid_user_pages(struct amdkfd_process_info *process_info) if (ret) { pr_err("%s: update PTE failed\n", __func__); /* make sure this gets validated again */ - atomic_inc(&mem->invalid); + mutex_lock(&process_info->notifier_lock); + mem->invalid++; + mutex_unlock(&process_info->notifier_lock); goto unreserve_out; } } @@ -2529,6 +2524,36 @@ out_no_mem: return ret; } +/* Confirm that all user pages are valid while holding the notifier lock + * + * Moves valid BOs from the userptr_inval_list back to userptr_val_list. + */ +static int confirm_valid_user_pages_locked(struct amdkfd_process_info *process_info) +{ + struct kgd_mem *mem, *tmp_mem; + int ret = 0; + + list_for_each_entry_safe(mem, tmp_mem, + &process_info->userptr_inval_list, + validate_list.head) { + bool valid = amdgpu_ttm_tt_get_user_pages_done( + mem->bo->tbo.ttm, mem->range); + + mem->range = NULL; + if (!valid) { + WARN(!mem->invalid, "Invalid BO not marked invalid"); + ret = -EAGAIN; + continue; + } + WARN(mem->invalid, "Valid BO is marked invalid"); + + list_move_tail(&mem->validate_list.head, + &process_info->userptr_valid_list); + } + + return ret; +} + /* Worker callback to restore evicted userptr BOs * * Tries to update and validate all userptr BOs. If successful and no @@ -2543,9 +2568,11 @@ static void amdgpu_amdkfd_restore_userptr_worker(struct work_struct *work) restore_userptr_work); struct task_struct *usertask; struct mm_struct *mm; - int evicted_bos; + uint32_t evicted_bos; - evicted_bos = atomic_read(&process_info->evicted_bos); + mutex_lock(&process_info->notifier_lock); + evicted_bos = process_info->evicted_bos; + mutex_unlock(&process_info->notifier_lock); if (!evicted_bos) return; @@ -2568,9 +2595,6 @@ static void amdgpu_amdkfd_restore_userptr_worker(struct work_struct *work) * and we can just restart the queues. */ if (!list_empty(&process_info->userptr_inval_list)) { - if (atomic_read(&process_info->evicted_bos) != evicted_bos) - goto unlock_out; /* Concurrent eviction, try again */ - if (validate_invalid_user_pages(process_info)) goto unlock_out; } @@ -2579,10 +2603,17 @@ static void amdgpu_amdkfd_restore_userptr_worker(struct work_struct *work) * be a first eviction that calls quiesce_mm. The eviction * reference counting inside KFD will handle this case. */ - if (atomic_cmpxchg(&process_info->evicted_bos, evicted_bos, 0) != - evicted_bos) - goto unlock_out; - evicted_bos = 0; + mutex_lock(&process_info->notifier_lock); + if (process_info->evicted_bos != evicted_bos) + goto unlock_notifier_out; + + if (confirm_valid_user_pages_locked(process_info)) { + WARN(1, "User pages unexpectedly invalid"); + goto unlock_notifier_out; + } + + process_info->evicted_bos = evicted_bos = 0; + if (kgd2kfd_resume_mm(mm)) { pr_err("%s: Failed to resume KFD\n", __func__); /* No recovery from this failure. Probably the CP is @@ -2590,6 +2621,8 @@ static void amdgpu_amdkfd_restore_userptr_worker(struct work_struct *work) */ } +unlock_notifier_out: + mutex_unlock(&process_info->notifier_lock); unlock_out: mutex_unlock(&process_info->lock); diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_atomfirmware.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_atomfirmware.c index b81b77a9efa6..ac6fe0ae4609 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_atomfirmware.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_atomfirmware.c @@ -101,39 +101,101 @@ void amdgpu_atomfirmware_scratch_regs_init(struct amdgpu_device *adev) } } +static int amdgpu_atomfirmware_allocate_fb_v2_1(struct amdgpu_device *adev, + struct vram_usagebyfirmware_v2_1 *fw_usage, int *usage_bytes) +{ + u32 start_addr, fw_size, drv_size; + + start_addr = le32_to_cpu(fw_usage->start_address_in_kb); + fw_size = le16_to_cpu(fw_usage->used_by_firmware_in_kb); + drv_size = le16_to_cpu(fw_usage->used_by_driver_in_kb); + + DRM_DEBUG("atom firmware v2_1 requested %08x %dkb fw %dkb drv\n", + start_addr, + fw_size, + drv_size); + + if ((start_addr & ATOM_VRAM_OPERATION_FLAGS_MASK) == + (u32)(ATOM_VRAM_BLOCK_SRIOV_MSG_SHARE_RESERVATION << + ATOM_VRAM_OPERATION_FLAGS_SHIFT)) { + /* Firmware request VRAM reservation for SR-IOV */ + adev->mman.fw_vram_usage_start_offset = (start_addr & + (~ATOM_VRAM_OPERATION_FLAGS_MASK)) << 10; + adev->mman.fw_vram_usage_size = fw_size << 10; + /* Use the default scratch size */ + *usage_bytes = 0; + } else { + *usage_bytes = drv_size << 10; + } + return 0; +} + +static int amdgpu_atomfirmware_allocate_fb_v2_2(struct amdgpu_device *adev, + struct vram_usagebyfirmware_v2_2 *fw_usage, int *usage_bytes) +{ + u32 fw_start_addr, fw_size, drv_start_addr, drv_size; + + fw_start_addr = le32_to_cpu(fw_usage->fw_region_start_address_in_kb); + fw_size = le16_to_cpu(fw_usage->used_by_firmware_in_kb); + + drv_start_addr = le32_to_cpu(fw_usage->driver_region0_start_address_in_kb); + drv_size = le32_to_cpu(fw_usage->used_by_driver_region0_in_kb); + + DRM_DEBUG("atom requested fw start at %08x %dkb and drv start at %08x %dkb\n", + fw_start_addr, + fw_size, + drv_start_addr, + drv_size); + + if (amdgpu_sriov_vf(adev) && + ((fw_start_addr & (ATOM_VRAM_BLOCK_NEEDS_NO_RESERVATION << + ATOM_VRAM_OPERATION_FLAGS_SHIFT)) == 0)) { + /* Firmware request VRAM reservation for SR-IOV */ + adev->mman.fw_vram_usage_start_offset = (fw_start_addr & + (~ATOM_VRAM_OPERATION_FLAGS_MASK)) << 10; + adev->mman.fw_vram_usage_size = fw_size << 10; + } + + if (amdgpu_sriov_vf(adev) && + ((drv_start_addr & (ATOM_VRAM_BLOCK_NEEDS_NO_RESERVATION << + ATOM_VRAM_OPERATION_FLAGS_SHIFT)) == 0)) { + /* driver request VRAM reservation for SR-IOV */ + adev->mman.drv_vram_usage_start_offset = (drv_start_addr & + (~ATOM_VRAM_OPERATION_FLAGS_MASK)) << 10; + adev->mman.drv_vram_usage_size = drv_size << 10; + } + + *usage_bytes = 0; + return 0; +} + int amdgpu_atomfirmware_allocate_fb_scratch(struct amdgpu_device *adev) { struct atom_context *ctx = adev->mode_info.atom_context; int index = get_index_into_master_table(atom_master_list_of_data_tables_v2_1, vram_usagebyfirmware); - struct vram_usagebyfirmware_v2_1 *firmware_usage; - uint32_t start_addr, size; - uint16_t data_offset; + struct vram_usagebyfirmware_v2_1 *fw_usage_v2_1; + struct vram_usagebyfirmware_v2_2 *fw_usage_v2_2; + u16 data_offset; + u8 frev, crev; int usage_bytes = 0; - if (amdgpu_atom_parse_data_header(ctx, index, NULL, NULL, NULL, &data_offset)) { - firmware_usage = (struct vram_usagebyfirmware_v2_1 *)(ctx->bios + data_offset); - DRM_DEBUG("atom firmware requested %08x %dkb fw %dkb drv\n", - le32_to_cpu(firmware_usage->start_address_in_kb), - le16_to_cpu(firmware_usage->used_by_firmware_in_kb), - le16_to_cpu(firmware_usage->used_by_driver_in_kb)); - - start_addr = le32_to_cpu(firmware_usage->start_address_in_kb); - size = le16_to_cpu(firmware_usage->used_by_firmware_in_kb); - - if ((uint32_t)(start_addr & ATOM_VRAM_OPERATION_FLAGS_MASK) == - (uint32_t)(ATOM_VRAM_BLOCK_SRIOV_MSG_SHARE_RESERVATION << - ATOM_VRAM_OPERATION_FLAGS_SHIFT)) { - /* Firmware request VRAM reservation for SR-IOV */ - adev->mman.fw_vram_usage_start_offset = (start_addr & - (~ATOM_VRAM_OPERATION_FLAGS_MASK)) << 10; - adev->mman.fw_vram_usage_size = size << 10; - /* Use the default scratch size */ - usage_bytes = 0; - } else { - usage_bytes = le16_to_cpu(firmware_usage->used_by_driver_in_kb) << 10; + if (amdgpu_atom_parse_data_header(ctx, index, NULL, &frev, &crev, &data_offset)) { + if (frev == 2 && crev == 1) { + fw_usage_v2_1 = + (struct vram_usagebyfirmware_v2_1 *)(ctx->bios + data_offset); + amdgpu_atomfirmware_allocate_fb_v2_1(adev, + fw_usage_v2_1, + &usage_bytes); + } else if (frev >= 2 && crev >= 2) { + fw_usage_v2_2 = + (struct vram_usagebyfirmware_v2_2 *)(ctx->bios + data_offset); + amdgpu_atomfirmware_allocate_fb_v2_2(adev, + fw_usage_v2_2, + &usage_bytes); } } + ctx->scratch_size_bytes = 0; if (usage_bytes == 0) usage_bytes = 20 * 1024; diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_bios.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_bios.c index e363f56c72af..30c28a69e847 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_bios.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_bios.c @@ -317,6 +317,7 @@ static bool amdgpu_atrm_get_bios(struct amdgpu_device *adev) if (!found) return false; + pci_dev_put(pdev); adev->bios = kmalloc(size, GFP_KERNEL); if (!adev->bios) { diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_connectors.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_connectors.c index cfb262911bfc..2ebbc6382a06 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_connectors.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_connectors.c @@ -26,7 +26,6 @@ #include <drm/display/drm_dp_helper.h> #include <drm/drm_edid.h> -#include <drm/drm_fb_helper.h> #include <drm/drm_probe_helper.h> #include <drm/amdgpu_drm.h> #include "amdgpu.h" diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_cs.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_cs.c index 365e3fb6a9e5..7b5ce00f0602 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_cs.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_cs.c @@ -61,6 +61,8 @@ static int amdgpu_cs_parser_init(struct amdgpu_cs_parser *p, amdgpu_ctx_put(p->ctx); return -ECANCELED; } + + amdgpu_sync_create(&p->sync); return 0; } @@ -294,12 +296,8 @@ static int amdgpu_cs_pass1(struct amdgpu_cs_parser *p, } for (i = 0; i < p->gang_size; ++i) { - ret = amdgpu_job_alloc(p->adev, num_ibs[i], &p->jobs[i], vm); - if (ret) - goto free_all_kdata; - - ret = drm_sched_job_init(&p->jobs[i]->base, p->entities[i], - &fpriv->vm); + ret = amdgpu_job_alloc(p->adev, vm, p->entities[i], vm, + num_ibs[i], &p->jobs[i]); if (ret) goto free_all_kdata; } @@ -433,7 +431,7 @@ static int amdgpu_cs_p2_dependencies(struct amdgpu_cs_parser *p, dma_fence_put(old); } - r = amdgpu_sync_fence(&p->gang_leader->sync, fence); + r = amdgpu_sync_fence(&p->sync, fence); dma_fence_put(fence); if (r) return r; @@ -455,9 +453,8 @@ static int amdgpu_syncobj_lookup_and_add(struct amdgpu_cs_parser *p, return r; } - r = amdgpu_sync_fence(&p->gang_leader->sync, fence); + r = amdgpu_sync_fence(&p->sync, fence); dma_fence_put(fence); - return r; } @@ -1106,7 +1103,7 @@ static int amdgpu_cs_vm_handling(struct amdgpu_cs_parser *p) if (r) return r; - r = amdgpu_sync_fence(&job->sync, fpriv->prt_va->last_pt_update); + r = amdgpu_sync_fence(&p->sync, fpriv->prt_va->last_pt_update); if (r) return r; @@ -1117,7 +1114,7 @@ static int amdgpu_cs_vm_handling(struct amdgpu_cs_parser *p) if (r) return r; - r = amdgpu_sync_fence(&job->sync, bo_va->last_pt_update); + r = amdgpu_sync_fence(&p->sync, bo_va->last_pt_update); if (r) return r; } @@ -1136,7 +1133,7 @@ static int amdgpu_cs_vm_handling(struct amdgpu_cs_parser *p) if (r) return r; - r = amdgpu_sync_fence(&job->sync, bo_va->last_pt_update); + r = amdgpu_sync_fence(&p->sync, bo_va->last_pt_update); if (r) return r; } @@ -1149,7 +1146,7 @@ static int amdgpu_cs_vm_handling(struct amdgpu_cs_parser *p) if (r) return r; - r = amdgpu_sync_fence(&job->sync, vm->last_update); + r = amdgpu_sync_fence(&p->sync, vm->last_update); if (r) return r; @@ -1181,11 +1178,19 @@ static int amdgpu_cs_vm_handling(struct amdgpu_cs_parser *p) static int amdgpu_cs_sync_rings(struct amdgpu_cs_parser *p) { struct amdgpu_fpriv *fpriv = p->filp->driver_priv; - struct amdgpu_job *leader = p->gang_leader; + struct drm_gpu_scheduler *sched; struct amdgpu_bo_list_entry *e; + struct dma_fence *fence; unsigned int i; int r; + r = amdgpu_ctx_wait_prev_fence(p->ctx, p->entities[p->gang_leader_idx]); + if (r) { + if (r != -ERESTARTSYS) + DRM_ERROR("amdgpu_ctx_wait_prev_fence failed.\n"); + return r; + } + list_for_each_entry(e, &p->validated, tv.head) { struct amdgpu_bo *bo = ttm_to_amdgpu_bo(e->tv.bo); struct dma_resv *resv = bo->tbo.base.resv; @@ -1193,25 +1198,36 @@ static int amdgpu_cs_sync_rings(struct amdgpu_cs_parser *p) sync_mode = amdgpu_bo_explicit_sync(bo) ? AMDGPU_SYNC_EXPLICIT : AMDGPU_SYNC_NE_OWNER; - r = amdgpu_sync_resv(p->adev, &leader->sync, resv, sync_mode, + r = amdgpu_sync_resv(p->adev, &p->sync, resv, sync_mode, &fpriv->vm); if (r) return r; } for (i = 0; i < p->gang_size; ++i) { - if (p->jobs[i] == leader) + r = amdgpu_sync_push_to_job(&p->sync, p->jobs[i]); + if (r) + return r; + } + + sched = p->gang_leader->base.entity->rq->sched; + while ((fence = amdgpu_sync_get_fence(&p->sync))) { + struct drm_sched_fence *s_fence = to_drm_sched_fence(fence); + + /* + * When we have an dependency it might be necessary to insert a + * pipeline sync to make sure that all caches etc are flushed and the + * next job actually sees the results from the previous one + * before we start executing on the same scheduler ring. + */ + if (!s_fence || s_fence->sched != sched) continue; - r = amdgpu_sync_clone(&leader->sync, &p->jobs[i]->sync); + r = amdgpu_sync_fence(&p->gang_leader->explicit_sync, fence); if (r) return r; } - - r = amdgpu_ctx_wait_prev_fence(p->ctx, p->entities[p->gang_leader_idx]); - if (r && r != -ERESTARTSYS) - DRM_ERROR("amdgpu_ctx_wait_prev_fence failed.\n"); - return r; + return 0; } static void amdgpu_cs_post_dependencies(struct amdgpu_cs_parser *p) @@ -1251,9 +1267,12 @@ static int amdgpu_cs_submit(struct amdgpu_cs_parser *p, continue; fence = &p->jobs[i]->base.s_fence->scheduled; - r = amdgpu_sync_fence(&leader->sync, fence); - if (r) + dma_fence_get(fence); + r = drm_sched_job_add_dependency(&leader->base, fence); + if (r) { + dma_fence_put(fence); goto error_cleanup; + } } if (p->gang_size > 1) { @@ -1341,6 +1360,7 @@ static void amdgpu_cs_parser_fini(struct amdgpu_cs_parser *parser) { unsigned i; + amdgpu_sync_free(&parser->sync); for (i = 0; i < parser->num_post_deps; i++) { drm_syncobj_put(parser->post_deps[i].syncobj); kfree(parser->post_deps[i].chain); diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_cs.h b/drivers/gpu/drm/amd/amdgpu/amdgpu_cs.h index f80adf9069ec..113f39510a72 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_cs.h +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_cs.h @@ -76,6 +76,8 @@ struct amdgpu_cs_parser { unsigned num_post_deps; struct amdgpu_cs_post_dep *post_deps; + + struct amdgpu_sync sync; }; int amdgpu_cs_find_mapping(struct amdgpu_cs_parser *parser, diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_debugfs.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_debugfs.c index de61a85c4b02..0f16d3c09309 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_debugfs.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_debugfs.c @@ -1969,7 +1969,7 @@ int amdgpu_debugfs_init(struct amdgpu_device *adev) amdgpu_ta_if_debugfs_init(adev); #if defined(CONFIG_DRM_AMD_DC) - if (amdgpu_device_has_dc_support(adev)) + if (adev->dc_enabled) dtn_debugfs_init(adev); #endif diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_device.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_device.c index f1e9663b4051..2f28a8c02f64 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_device.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_device.c @@ -36,7 +36,9 @@ #include <generated/utsrelease.h> #include <linux/pci-p2pdma.h> +#include <drm/drm_aperture.h> #include <drm/drm_atomic_helper.h> +#include <drm/drm_fb_helper.h> #include <drm/drm_probe_helper.h> #include <drm/amdgpu_drm.h> #include <linux/vgaarb.h> @@ -89,6 +91,8 @@ MODULE_FIRMWARE("amdgpu/navi12_gpu_info.bin"); #define AMDGPU_MAX_RETRY_LIMIT 2 #define AMDGPU_RETRY_SRIOV_RESET(r) ((r) == -EBUSY || (r) == -ETIMEDOUT || (r) == -EINVAL) +static const struct drm_driver amdgpu_kms_driver; + const char *amdgpu_asic_name[] = { "TAHITI", "PITCAIRN", @@ -1568,7 +1572,7 @@ static int amdgpu_device_check_arguments(struct amdgpu_device *adev) * @pdev: pci dev pointer * @state: vga_switcheroo state * - * Callback for the switcheroo driver. Suspends or resumes the + * Callback for the switcheroo driver. Suspends or resumes * the asics before or after it is powered up using ACPI methods. */ static void amdgpu_switcheroo_set_state(struct pci_dev *pdev, @@ -1915,6 +1919,16 @@ static void amdgpu_device_enable_virtual_display(struct amdgpu_device *adev) } } +void amdgpu_device_set_sriov_virtual_display(struct amdgpu_device *adev) +{ + if (amdgpu_sriov_vf(adev) && !adev->enable_virtual_display) { + adev->mode_info.num_crtc = 1; + adev->enable_virtual_display = true; + DRM_INFO("virtual_display:%d, num_crtc:%d\n", + adev->enable_virtual_display, adev->mode_info.num_crtc); + } +} + /** * amdgpu_device_parse_gpu_info_fw - parse gpu info firmware * @@ -2397,7 +2411,7 @@ static int amdgpu_device_ip_init(struct amdgpu_device *adev) adev->ip_blocks[i].status.hw = true; /* right after GMC hw init, we create CSA */ - if (amdgpu_mcbp || amdgpu_sriov_vf(adev)) { + if (amdgpu_mcbp) { r = amdgpu_allocate_static_csa(adev, &adev->virt.csa_obj, AMDGPU_GEM_DOMAIN_VRAM, AMDGPU_CSA_SIZE); @@ -2462,6 +2476,11 @@ static int amdgpu_device_ip_init(struct amdgpu_device *adev) if (!amdgpu_sriov_vf(adev)) { struct amdgpu_hive_info *hive = amdgpu_get_xgmi_hive(adev); + if (WARN_ON(!hive)) { + r = -ENOENT; + goto init_failed; + } + if (!hive->reset_domain || !amdgpu_reset_get_reset_domain(hive->reset_domain)) { r = -ENOENT; @@ -3000,14 +3019,15 @@ static int amdgpu_device_ip_suspend_phase2(struct amdgpu_device *adev) continue; } - /* skip suspend of gfx and psp for S0ix + /* skip suspend of gfx/mes and psp for S0ix * gfx is in gfxoff state, so on resume it will exit gfxoff just * like at runtime. PSP is also part of the always on hardware * so no need to suspend it. */ if (adev->in_s0ix && (adev->ip_blocks[i].version->type == AMD_IP_BLOCK_TYPE_PSP || - adev->ip_blocks[i].version->type == AMD_IP_BLOCK_TYPE_GFX)) + adev->ip_blocks[i].version->type == AMD_IP_BLOCK_TYPE_GFX || + adev->ip_blocks[i].version->type == AMD_IP_BLOCK_TYPE_MES)) continue; /* XXX handle errors */ @@ -3347,8 +3367,7 @@ bool amdgpu_device_asic_has_dc_support(enum amd_asic_type asic_type) */ bool amdgpu_device_has_dc_support(struct amdgpu_device *adev) { - if (amdgpu_sriov_vf(adev) || - adev->enable_virtual_display || + if (adev->enable_virtual_display || (adev->harvest_ip_mask & AMD_HARVEST_IP_DMU_MASK)) return false; @@ -3671,6 +3690,11 @@ int amdgpu_device_init(struct amdgpu_device *adev, if (r) return r; + /* Get rid of things like offb */ + r = drm_aperture_remove_conflicting_pci_framebuffers(adev->pdev, &amdgpu_kms_driver); + if (r) + return r; + /* Enable TMZ based on IP_VERSION */ amdgpu_gmc_tmz_set(adev); @@ -4097,6 +4121,11 @@ int amdgpu_device_suspend(struct drm_device *dev, bool fbcon) adev->in_suspend = true; + /* Evict the majority of BOs before grabbing the full access */ + r = amdgpu_device_evict_resources(adev); + if (r) + return r; + if (amdgpu_sriov_vf(adev)) { amdgpu_virt_fini_data_exchange(adev); r = amdgpu_virt_request_full_gpu(adev, false); @@ -4171,21 +4200,15 @@ int amdgpu_device_resume(struct drm_device *dev, bool fbcon) r = amdgpu_device_ip_resume(adev); - /* no matter what r is, always need to properly release full GPU */ - if (amdgpu_sriov_vf(adev)) { - amdgpu_virt_init_data_exchange(adev); - amdgpu_virt_release_full_gpu(adev, true); - } - if (r) { dev_err(adev->dev, "amdgpu_device_ip_resume failed (%d).\n", r); - return r; + goto exit; } amdgpu_fence_driver_hw_init(adev); r = amdgpu_device_ip_late_init(adev); if (r) - return r; + goto exit; queue_delayed_work(system_wq, &adev->delayed_init_work, msecs_to_jiffies(AMDGPU_RESUME_MS)); @@ -4193,9 +4216,18 @@ int amdgpu_device_resume(struct drm_device *dev, bool fbcon) if (!adev->in_s0ix) { r = amdgpu_amdkfd_resume(adev, adev->in_runpm); if (r) - return r; + goto exit; } +exit: + if (amdgpu_sriov_vf(adev)) { + amdgpu_virt_init_data_exchange(adev); + amdgpu_virt_release_full_gpu(adev, true); + } + + if (r) + return r; + /* Make sure IB tests flushed */ flush_delayed_work(&adev->delayed_init_work); @@ -4213,25 +4245,27 @@ int amdgpu_device_resume(struct drm_device *dev, bool fbcon) amdgpu_ras_resume(adev); - /* - * Most of the connector probing functions try to acquire runtime pm - * refs to ensure that the GPU is powered on when connector polling is - * performed. Since we're calling this from a runtime PM callback, - * trying to acquire rpm refs will cause us to deadlock. - * - * Since we're guaranteed to be holding the rpm lock, it's safe to - * temporarily disable the rpm helpers so this doesn't deadlock us. - */ + if (adev->mode_info.num_crtc) { + /* + * Most of the connector probing functions try to acquire runtime pm + * refs to ensure that the GPU is powered on when connector polling is + * performed. Since we're calling this from a runtime PM callback, + * trying to acquire rpm refs will cause us to deadlock. + * + * Since we're guaranteed to be holding the rpm lock, it's safe to + * temporarily disable the rpm helpers so this doesn't deadlock us. + */ #ifdef CONFIG_PM - dev->dev->power.disable_depth++; + dev->dev->power.disable_depth++; #endif - if (!amdgpu_device_has_dc_support(adev)) - drm_helper_hpd_irq_event(dev); - else - drm_kms_helper_hotplug_event(dev); + if (!adev->dc_enabled) + drm_helper_hpd_irq_event(dev); + else + drm_kms_helper_hotplug_event(dev); #ifdef CONFIG_PM - dev->dev->power.disable_depth--; + dev->dev->power.disable_depth--; #endif + } adev->in_suspend = false; if (amdgpu_acpi_smart_shift_update(dev, AMDGPU_SS_DEV_D0)) @@ -4580,6 +4614,10 @@ bool amdgpu_device_should_recover_gpu(struct amdgpu_device *adev) if (amdgpu_gpu_recovery == 0) goto disabled; + /* Skip soft reset check in fatal error mode */ + if (!amdgpu_ras_is_poison_mode_supported(adev)) + return true; + if (!amdgpu_device_ip_check_soft_reset(adev)) { dev_info(adev->dev,"Timeout, but no hardware hang detected.\n"); return false; @@ -5027,6 +5065,8 @@ static void amdgpu_device_resume_display_audio(struct amdgpu_device *adev) pm_runtime_enable(&(p->dev)); pm_runtime_resume(&(p->dev)); } + + pci_dev_put(p); } static int amdgpu_device_suspend_display_audio(struct amdgpu_device *adev) @@ -5065,6 +5105,7 @@ static int amdgpu_device_suspend_display_audio(struct amdgpu_device *adev) if (expires < ktime_get_mono_fast_ns()) { dev_warn(adev->dev, "failed to suspend display audio\n"); + pci_dev_put(p); /* TODO: abort the succeeding gpu reset? */ return -ETIMEDOUT; } @@ -5072,97 +5113,10 @@ static int amdgpu_device_suspend_display_audio(struct amdgpu_device *adev) pm_runtime_disable(&(p->dev)); + pci_dev_put(p); return 0; } -static void amdgpu_device_recheck_guilty_jobs( - struct amdgpu_device *adev, struct list_head *device_list_handle, - struct amdgpu_reset_context *reset_context) -{ - int i, r = 0; - - for (i = 0; i < AMDGPU_MAX_RINGS; ++i) { - struct amdgpu_ring *ring = adev->rings[i]; - int ret = 0; - struct drm_sched_job *s_job; - - if (!ring || !ring->sched.thread) - continue; - - s_job = list_first_entry_or_null(&ring->sched.pending_list, - struct drm_sched_job, list); - if (s_job == NULL) - continue; - - /* clear job's guilty and depend the folowing step to decide the real one */ - drm_sched_reset_karma(s_job); - drm_sched_resubmit_jobs_ext(&ring->sched, 1); - - if (!s_job->s_fence->parent) { - DRM_WARN("Failed to get a HW fence for job!"); - continue; - } - - ret = dma_fence_wait_timeout(s_job->s_fence->parent, false, ring->sched.timeout); - if (ret == 0) { /* timeout */ - DRM_ERROR("Found the real bad job! ring:%s, job_id:%llx\n", - ring->sched.name, s_job->id); - - - amdgpu_fence_driver_isr_toggle(adev, true); - - /* Clear this failed job from fence array */ - amdgpu_fence_driver_clear_job_fences(ring); - - amdgpu_fence_driver_isr_toggle(adev, false); - - /* Since the job won't signal and we go for - * another resubmit drop this parent pointer - */ - dma_fence_put(s_job->s_fence->parent); - s_job->s_fence->parent = NULL; - - /* set guilty */ - drm_sched_increase_karma(s_job); - amdgpu_reset_prepare_hwcontext(adev, reset_context); -retry: - /* do hw reset */ - if (amdgpu_sriov_vf(adev)) { - amdgpu_virt_fini_data_exchange(adev); - r = amdgpu_device_reset_sriov(adev, false); - if (r) - adev->asic_reset_res = r; - } else { - clear_bit(AMDGPU_SKIP_HW_RESET, - &reset_context->flags); - r = amdgpu_do_asic_reset(device_list_handle, - reset_context); - if (r && r == -EAGAIN) - goto retry; - } - - /* - * add reset counter so that the following - * resubmitted job could flush vmid - */ - atomic_inc(&adev->gpu_reset_counter); - continue; - } - - /* got the hw fence, signal finished fence */ - atomic_dec(ring->sched.score); - dma_fence_get(&s_job->s_fence->finished); - dma_fence_signal(&s_job->s_fence->finished); - dma_fence_put(&s_job->s_fence->finished); - - /* remove node from list and free the job */ - spin_lock(&ring->sched.job_list_lock); - list_del_init(&s_job->list); - spin_unlock(&ring->sched.job_list_lock); - ring->sched.ops->free_job(s_job); - } -} - static inline void amdgpu_device_stop_pending_resets(struct amdgpu_device *adev) { struct amdgpu_ras *con = amdgpu_ras_get_context(adev); @@ -5183,7 +5137,6 @@ static inline void amdgpu_device_stop_pending_resets(struct amdgpu_device *adev) } - /** * amdgpu_device_gpu_recover - reset the asic and recover scheduler * @@ -5206,7 +5159,6 @@ int amdgpu_device_gpu_recover(struct amdgpu_device *adev, int i, r = 0; bool need_emergency_restart = false; bool audio_suspended = false; - int tmp_vram_lost_counter; bool gpu_reset_for_dev_remove = false; gpu_reset_for_dev_remove = @@ -5352,7 +5304,6 @@ retry: /* Rest of adevs pre asic reset from XGMI hive. */ amdgpu_device_stop_pending_resets(tmp_adev); } - tmp_vram_lost_counter = atomic_read(&((adev)->vram_lost_counter)); /* Actual ASIC resets if needed.*/ /* Host driver will handle XGMI hive reset for SRIOV */ if (amdgpu_sriov_vf(adev)) { @@ -5377,29 +5328,13 @@ skip_hw_reset: /* Post ASIC reset for all devs .*/ list_for_each_entry(tmp_adev, device_list_handle, reset_list) { - /* - * Sometimes a later bad compute job can block a good gfx job as gfx - * and compute ring share internal GC HW mutually. We add an additional - * guilty jobs recheck step to find the real guilty job, it synchronously - * submits and pends for the first job being signaled. If it gets timeout, - * we identify it as a real guilty job. - */ - if (amdgpu_gpu_recovery == 2 && - !(tmp_vram_lost_counter < atomic_read(&adev->vram_lost_counter))) - amdgpu_device_recheck_guilty_jobs( - tmp_adev, device_list_handle, reset_context); - for (i = 0; i < AMDGPU_MAX_RINGS; ++i) { struct amdgpu_ring *ring = tmp_adev->rings[i]; if (!ring || !ring->sched.thread) continue; - /* No point to resubmit jobs if we didn't HW reset*/ - if (!tmp_adev->asic_reset_res && !job_signaled) - drm_sched_resubmit_jobs(&ring->sched); - - drm_sched_start(&ring->sched, !tmp_adev->asic_reset_res); + drm_sched_start(&ring->sched, true); } if (adev->enable_mes && adev->ip_versions[GC_HWIP][0] != IP_VERSION(11, 0, 3)) @@ -5441,6 +5376,8 @@ skip_sched_resume: amdgpu_device_resume_display_audio(tmp_adev); amdgpu_device_unset_mp1_state(tmp_adev); + + amdgpu_ras_set_error_query_ready(tmp_adev, true); } recover_end: @@ -5852,8 +5789,6 @@ void amdgpu_pci_resume(struct pci_dev *pdev) if (!ring || !ring->sched.thread) continue; - - drm_sched_resubmit_jobs(&ring->sched); drm_sched_start(&ring->sched, true); } diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_discovery.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_discovery.c index 3993e6134914..1bbd56029a4f 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_discovery.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_discovery.c @@ -305,8 +305,13 @@ static int amdgpu_discovery_init(struct amdgpu_device *adev) goto out; } - if (!amdgpu_discovery_verify_binary_signature(adev->mman.discovery_bin)) { - dev_warn(adev->dev, "get invalid ip discovery binary signature from vram\n"); + if (!amdgpu_discovery_verify_binary_signature(adev->mman.discovery_bin) || amdgpu_discovery == 2) { + /* ignore the discovery binary from vram if discovery=2 in kernel module parameter */ + if (amdgpu_discovery == 2) + dev_info(adev->dev,"force read ip discovery binary from file"); + else + dev_warn(adev->dev, "get invalid ip discovery binary signature from vram\n"); + /* retry read ip discovery binary from file */ r = amdgpu_discovery_read_binary_from_file(adev, adev->mman.discovery_bin); if (r) { @@ -1507,6 +1512,7 @@ static int amdgpu_discovery_set_common_ip_blocks(struct amdgpu_device *adev) case IP_VERSION(11, 0, 1): case IP_VERSION(11, 0, 2): case IP_VERSION(11, 0, 3): + case IP_VERSION(11, 0, 4): amdgpu_device_ip_block_add(adev, &soc21_common_ip_block); break; default: @@ -1551,6 +1557,7 @@ static int amdgpu_discovery_set_gmc_ip_blocks(struct amdgpu_device *adev) case IP_VERSION(11, 0, 1): case IP_VERSION(11, 0, 2): case IP_VERSION(11, 0, 3): + case IP_VERSION(11, 0, 4): amdgpu_device_ip_block_add(adev, &gmc_v11_0_ip_block); break; default: @@ -1636,6 +1643,7 @@ static int amdgpu_discovery_set_psp_ip_blocks(struct amdgpu_device *adev) case IP_VERSION(13, 0, 7): case IP_VERSION(13, 0, 8): case IP_VERSION(13, 0, 10): + case IP_VERSION(13, 0, 11): amdgpu_device_ip_block_add(adev, &psp_v13_0_ip_block); break; case IP_VERSION(13, 0, 4): @@ -1686,6 +1694,7 @@ static int amdgpu_discovery_set_smu_ip_blocks(struct amdgpu_device *adev) case IP_VERSION(13, 0, 7): case IP_VERSION(13, 0, 8): case IP_VERSION(13, 0, 10): + case IP_VERSION(13, 0, 11): amdgpu_device_ip_block_add(adev, &smu_v13_0_ip_block); break; default: @@ -1697,9 +1706,17 @@ static int amdgpu_discovery_set_smu_ip_blocks(struct amdgpu_device *adev) return 0; } +#if defined(CONFIG_DRM_AMD_DC) +static void amdgpu_discovery_set_sriov_display(struct amdgpu_device *adev) +{ + amdgpu_device_set_sriov_virtual_display(adev); + amdgpu_device_ip_block_add(adev, &amdgpu_vkms_ip_block); +} +#endif + static int amdgpu_discovery_set_display_ip_blocks(struct amdgpu_device *adev) { - if (adev->enable_virtual_display || amdgpu_sriov_vf(adev)) { + if (adev->enable_virtual_display) { amdgpu_device_ip_block_add(adev, &amdgpu_vkms_ip_block); return 0; } @@ -1727,7 +1744,10 @@ static int amdgpu_discovery_set_display_ip_blocks(struct amdgpu_device *adev) case IP_VERSION(3, 1, 6): case IP_VERSION(3, 2, 0): case IP_VERSION(3, 2, 1): - amdgpu_device_ip_block_add(adev, &dm_ip_block); + if (amdgpu_sriov_vf(adev)) + amdgpu_discovery_set_sriov_display(adev); + else + amdgpu_device_ip_block_add(adev, &dm_ip_block); break; default: dev_err(adev->dev, @@ -1740,7 +1760,10 @@ static int amdgpu_discovery_set_display_ip_blocks(struct amdgpu_device *adev) case IP_VERSION(12, 0, 0): case IP_VERSION(12, 0, 1): case IP_VERSION(12, 1, 0): - amdgpu_device_ip_block_add(adev, &dm_ip_block); + if (amdgpu_sriov_vf(adev)) + amdgpu_discovery_set_sriov_display(adev); + else + amdgpu_device_ip_block_add(adev, &dm_ip_block); break; default: dev_err(adev->dev, @@ -1785,6 +1808,7 @@ static int amdgpu_discovery_set_gc_ip_blocks(struct amdgpu_device *adev) case IP_VERSION(11, 0, 1): case IP_VERSION(11, 0, 2): case IP_VERSION(11, 0, 3): + case IP_VERSION(11, 0, 4): amdgpu_device_ip_block_add(adev, &gfx_v11_0_ip_block); break; default: @@ -1948,6 +1972,7 @@ static int amdgpu_discovery_set_mes_ip_blocks(struct amdgpu_device *adev) case IP_VERSION(11, 0, 1): case IP_VERSION(11, 0, 2): case IP_VERSION(11, 0, 3): + case IP_VERSION(11, 0, 4): amdgpu_device_ip_block_add(adev, &mes_v11_0_ip_block); adev->enable_mes = true; adev->enable_mes_kiq = true; @@ -2161,6 +2186,7 @@ int amdgpu_discovery_set_ip_blocks(struct amdgpu_device *adev) break; case IP_VERSION(10, 3, 1): adev->family = AMDGPU_FAMILY_VGH; + adev->apu_flags |= AMD_APU_IS_VANGOGH; break; case IP_VERSION(10, 3, 3): adev->family = AMDGPU_FAMILY_YC; @@ -2177,6 +2203,7 @@ int amdgpu_discovery_set_ip_blocks(struct amdgpu_device *adev) adev->family = AMDGPU_FAMILY_GC_11_0_0; break; case IP_VERSION(11, 0, 1): + case IP_VERSION(11, 0, 4): adev->family = AMDGPU_FAMILY_GC_11_0_1; break; default: @@ -2194,6 +2221,7 @@ int amdgpu_discovery_set_ip_blocks(struct amdgpu_device *adev) case IP_VERSION(10, 3, 6): case IP_VERSION(10, 3, 7): case IP_VERSION(11, 0, 1): + case IP_VERSION(11, 0, 4): adev->flags |= AMD_IS_APU; break; default: @@ -2250,6 +2278,7 @@ int amdgpu_discovery_set_ip_blocks(struct amdgpu_device *adev) adev->nbio.hdp_flush_reg = &nbio_v4_3_hdp_flush_reg; break; case IP_VERSION(7, 7, 0): + case IP_VERSION(7, 7, 1): adev->nbio.funcs = &nbio_v7_7_funcs; adev->nbio.hdp_flush_reg = &nbio_v7_7_hdp_flush_reg; break; diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_display.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_display.c index 1a06b8d724f3..b22471b3bd63 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_display.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_display.c @@ -39,11 +39,46 @@ #include <linux/pm_runtime.h> #include <drm/drm_crtc_helper.h> #include <drm/drm_edid.h> -#include <drm/drm_gem_framebuffer_helper.h> #include <drm/drm_fb_helper.h> +#include <drm/drm_gem_framebuffer_helper.h> #include <drm/drm_fourcc.h> #include <drm/drm_vblank.h> +/** + * amdgpu_display_hotplug_work_func - work handler for display hotplug event + * + * @work: work struct pointer + * + * This is the hotplug event work handler (all ASICs). + * The work gets scheduled from the IRQ handler if there + * was a hotplug interrupt. It walks through the connector table + * and calls hotplug handler for each connector. After this, it sends + * a DRM hotplug event to alert userspace. + * + * This design approach is required in order to defer hotplug event handling + * from the IRQ handler to a work handler because hotplug handler has to use + * mutexes which cannot be locked in an IRQ handler (since &mutex_lock may + * sleep). + */ +void amdgpu_display_hotplug_work_func(struct work_struct *work) +{ + struct amdgpu_device *adev = container_of(work, struct amdgpu_device, + hotplug_work); + struct drm_device *dev = adev_to_drm(adev); + struct drm_mode_config *mode_config = &dev->mode_config; + struct drm_connector *connector; + struct drm_connector_list_iter iter; + + mutex_lock(&mode_config->mutex); + drm_connector_list_iter_begin(dev, &iter); + drm_for_each_connector_iter(connector, &iter) + amdgpu_connector_hotplug(connector); + drm_connector_list_iter_end(&iter); + mutex_unlock(&mode_config->mutex); + /* Just fire off a uevent and let userspace tell us what to do */ + drm_helper_hpd_irq_event(dev); +} + static int amdgpu_display_framebuffer_init(struct drm_device *dev, struct amdgpu_framebuffer *rfb, const struct drm_mode_fb_cmd2 *mode_cmd, @@ -514,7 +549,7 @@ uint32_t amdgpu_display_supported_domains(struct amdgpu_device *adev, */ if ((bo_flags & AMDGPU_GEM_CREATE_CPU_GTT_USWC) && amdgpu_bo_support_uswc(bo_flags) && - amdgpu_device_asic_has_dc_support(adev->asic_type) && + adev->dc_enabled && adev->mode_info.gpu_vm_support) domain |= AMDGPU_GEM_DOMAIN_GTT; #endif @@ -1214,7 +1249,6 @@ amdgpu_display_user_framebuffer_create(struct drm_device *dev, const struct drm_mode_config_funcs amdgpu_mode_funcs = { .fb_create = amdgpu_display_user_framebuffer_create, - .output_poll_changed = drm_fb_helper_output_poll_changed, }; static const struct drm_prop_enum_list amdgpu_underscan_enum_list[] = @@ -1281,7 +1315,7 @@ int amdgpu_display_modeset_create_props(struct amdgpu_device *adev) "dither", amdgpu_dither_enum_list, sz); - if (amdgpu_device_has_dc_support(adev)) { + if (adev->dc_enabled) { adev->mode_info.abm_level_property = drm_property_create_range(adev_to_drm(adev), 0, "abm level", 0, 4); diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_display.h b/drivers/gpu/drm/amd/amdgpu/amdgpu_display.h index 560352f7c317..9d19940f73c8 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_display.h +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_display.h @@ -35,6 +35,7 @@ #define amdgpu_display_add_encoder(adev, e, s, c) (adev)->mode_info.funcs->add_encoder((adev), (e), (s), (c)) #define amdgpu_display_add_connector(adev, ci, sd, ct, ib, coi, h, r) (adev)->mode_info.funcs->add_connector((adev), (ci), (sd), (ct), (ib), (coi), (h), (r)) +void amdgpu_display_hotplug_work_func(struct work_struct *work); void amdgpu_display_update_priority(struct amdgpu_device *adev); uint32_t amdgpu_display_supported_domains(struct amdgpu_device *adev, uint64_t bo_flags); diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_dma_buf.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_dma_buf.c index 7bd8e33b14be..271e30e34d93 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_dma_buf.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_dma_buf.c @@ -328,7 +328,9 @@ amdgpu_dma_buf_create_obj(struct drm_device *dev, struct dma_buf *dma_buf) if (dma_buf->ops == &amdgpu_dmabuf_ops) { struct amdgpu_bo *other = gem_to_amdgpu_bo(dma_buf->priv); - flags |= other->flags & AMDGPU_GEM_CREATE_CPU_GTT_USWC; + flags |= other->flags & (AMDGPU_GEM_CREATE_CPU_GTT_USWC | + AMDGPU_GEM_CREATE_COHERENT | + AMDGPU_GEM_CREATE_UNCACHED); } ret = amdgpu_gem_object_create(adev, dma_buf->size, PAGE_SIZE, diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_drv.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_drv.c index bf2d50c8c92a..cd4caaa29528 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_drv.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_drv.c @@ -23,8 +23,8 @@ */ #include <drm/amdgpu_drm.h> -#include <drm/drm_aperture.h> #include <drm/drm_drv.h> +#include <drm/drm_fbdev_generic.h> #include <drm/drm_gem.h> #include <drm/drm_vblank.h> #include <drm/drm_managed.h> @@ -180,6 +180,7 @@ int amdgpu_mes_kiq; int amdgpu_noretry = -1; int amdgpu_force_asic_type = -1; int amdgpu_tmz = -1; /* auto */ +uint amdgpu_freesync_vid_mode; int amdgpu_reset_method = -1; /* auto */ int amdgpu_num_kcq = -1; int amdgpu_smartshift_bias; @@ -230,17 +231,18 @@ module_param_named(vis_vramlimit, amdgpu_vis_vram_limit, int, 0444); /** * DOC: gartsize (uint) - * Restrict the size of GART in Mib (32, 64, etc.) for testing. The default is -1 (The size depends on asic). + * Restrict the size of GART (for kernel use) in Mib (32, 64, etc.) for testing. + * The default is -1 (The size depends on asic). */ -MODULE_PARM_DESC(gartsize, "Size of GART to setup in megabytes (32, 64, etc., -1=auto)"); +MODULE_PARM_DESC(gartsize, "Size of kernel GART to setup in megabytes (32, 64, etc., -1=auto)"); module_param_named(gartsize, amdgpu_gart_size, uint, 0600); /** * DOC: gttsize (int) - * Restrict the size of GTT domain in MiB for testing. The default is -1 (It's VRAM size if 3GB < VRAM < 3/4 RAM, - * otherwise 3/4 RAM size). + * Restrict the size of GTT domain (for userspace use) in MiB for testing. + * The default is -1 (Use 1/2 RAM, minimum value is 3GB). */ -MODULE_PARM_DESC(gttsize, "Size of the GTT domain in megabytes (-1 = auto)"); +MODULE_PARM_DESC(gttsize, "Size of the GTT userspace domain in megabytes (-1 = auto)"); module_param_named(gttsize, amdgpu_gtt_size, int, 0600); /** @@ -533,7 +535,7 @@ module_param_named(compute_multipipe, amdgpu_compute_multipipe, int, 0444); * DOC: gpu_recovery (int) * Set to enable GPU recovery mechanism (1 = enable, 0 = disable). The default is -1 (auto, disabled except SRIOV). */ -MODULE_PARM_DESC(gpu_recovery, "Enable GPU recovery mechanism, (2 = advanced tdr mode, 1 = enable, 0 = disable, -1 = auto)"); +MODULE_PARM_DESC(gpu_recovery, "Enable GPU recovery mechanism, (1 = enable, 0 = disable, -1 = auto)"); module_param_named(gpu_recovery, amdgpu_gpu_recovery, int, 0444); /** @@ -878,6 +880,32 @@ MODULE_PARM_DESC(tmz, "Enable TMZ feature (-1 = auto (default), 0 = off, 1 = on) module_param_named(tmz, amdgpu_tmz, int, 0444); /** + * DOC: freesync_video (uint) + * Enable the optimization to adjust front porch timing to achieve seamless + * mode change experience when setting a freesync supported mode for which full + * modeset is not needed. + * + * The Display Core will add a set of modes derived from the base FreeSync + * video mode into the corresponding connector's mode list based on commonly + * used refresh rates and VRR range of the connected display, when users enable + * this feature. From the userspace perspective, they can see a seamless mode + * change experience when the change between different refresh rates under the + * same resolution. Additionally, userspace applications such as Video playback + * can read this modeset list and change the refresh rate based on the video + * frame rate. Finally, the userspace can also derive an appropriate mode for a + * particular refresh rate based on the FreeSync Mode and add it to the + * connector's mode list. + * + * Note: This is an experimental feature. + * + * The default value: 0 (off). + */ +MODULE_PARM_DESC( + freesync_video, + "Enable freesync modesetting optimization feature (0 = off (default), 1 = on)"); +module_param_named(freesync_video, amdgpu_freesync_vid_mode, uint, 0444); + +/** * DOC: reset_method (int) * GPU reset method (-1 = auto (default), 0 = legacy, 1 = mode0, 2 = mode1, 3 = mode2, 4 = baco) */ @@ -1924,9 +1952,6 @@ static const struct pci_device_id pciidlist[] = { {0x1002, 0x73AF, PCI_ANY_ID, PCI_ANY_ID, 0, 0, CHIP_SIENNA_CICHLID}, {0x1002, 0x73BF, PCI_ANY_ID, PCI_ANY_ID, 0, 0, CHIP_SIENNA_CICHLID}, - /* Van Gogh */ - {0x1002, 0x163F, PCI_ANY_ID, PCI_ANY_ID, 0, 0, CHIP_VANGOGH|AMD_IS_APU}, - /* Yellow Carp */ {0x1002, 0x164D, PCI_ANY_ID, PCI_ANY_ID, 0, 0, CHIP_YELLOW_CARP|AMD_IS_APU}, {0x1002, 0x1681, PCI_ANY_ID, PCI_ANY_ID, 0, 0, CHIP_YELLOW_CARP|AMD_IS_APU}, @@ -2040,6 +2065,15 @@ static int amdgpu_pci_probe(struct pci_dev *pdev, "See modparam exp_hw_support\n"); return -ENODEV; } + /* differentiate between P10 and P11 asics with the same DID */ + if (pdev->device == 0x67FF && + (pdev->revision == 0xE3 || + pdev->revision == 0xE7 || + pdev->revision == 0xF3 || + pdev->revision == 0xF7)) { + flags &= ~AMD_ASIC_MASK; + flags |= CHIP_POLARIS10; + } /* Due to hardware bugs, S/G Display on raven requires a 1:1 IOMMU mapping, * however, SME requires an indirect IOMMU mapping because the encryption @@ -2087,11 +2121,6 @@ static int amdgpu_pci_probe(struct pci_dev *pdev, } #endif - /* Get rid of things like offb */ - ret = drm_aperture_remove_conflicting_pci_framebuffers(pdev, &amdgpu_kms_driver); - if (ret) - return ret; - adev = devm_drm_dev_alloc(&pdev->dev, &amdgpu_kms_driver, typeof(*adev), ddev); if (IS_ERR(adev)) return PTR_ERR(adev); @@ -2109,12 +2138,12 @@ static int amdgpu_pci_probe(struct pci_dev *pdev, pci_set_drvdata(pdev, ddev); - ret = amdgpu_driver_load_kms(adev, ent->driver_data); + ret = amdgpu_driver_load_kms(adev, flags); if (ret) goto err_pci; retry_init: - ret = drm_dev_register(ddev, ent->driver_data); + ret = drm_dev_register(ddev, flags); if (ret == -EAGAIN && ++retry <= 3) { DRM_INFO("retry init %d\n", retry); /* Don't request EX mode too frequently which is attacking */ @@ -2471,7 +2500,7 @@ static int amdgpu_runtime_idle_check_display(struct device *dev) if (ret) return ret; - if (amdgpu_device_has_dc_support(adev)) { + if (adev->dc_enabled) { struct drm_crtc *crtc; drm_for_each_crtc(crtc, drm_dev) { @@ -2572,6 +2601,8 @@ static int amdgpu_pmops_runtime_suspend(struct device *dev) amdgpu_device_baco_enter(drm_dev); } + dev_dbg(&pdev->dev, "asic/device is runtime suspended\n"); + return 0; } diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_eeprom.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_eeprom.c index 4d9eb0137f8c..7d2a908438e9 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_eeprom.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_eeprom.c @@ -79,13 +79,15 @@ * That is, for an I2C EEPROM driver everything is controlled by * the "eeprom_addr". * + * See also top of amdgpu_ras_eeprom.c. + * * P.S. If you need to write, lock and read the Identification Page, * (M24M02-DR device only, which we do not use), change the "7" to * "0xF" in the macro below, and let the client set bit 20 to 1 in * "eeprom_addr", and set A10 to 0 to write into it, and A10 and A1 to * 1 to lock it permanently. */ -#define MAKE_I2C_ADDR(_aa) ((0xA << 3) | (((_aa) >> 16) & 7)) +#define MAKE_I2C_ADDR(_aa) ((0xA << 3) | (((_aa) >> 16) & 0xF)) static int __amdgpu_eeprom_xfer(struct i2c_adapter *i2c_adap, u32 eeprom_addr, u8 *eeprom_buf, u16 buf_size, bool read) diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_fence.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_fence.c index d0d99ed607dd..00444203220d 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_fence.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_fence.c @@ -55,6 +55,7 @@ struct amdgpu_fence { /* RB, DMA, etc. */ struct amdgpu_ring *ring; + ktime_t start_timestamp; }; static struct kmem_cache *amdgpu_fence_slab; @@ -199,6 +200,8 @@ int amdgpu_fence_emit(struct amdgpu_ring *ring, struct dma_fence **f, struct amd } } + to_amdgpu_fence(fence)->start_timestamp = ktime_get(); + /* This function can't be called concurrently anyway, otherwise * emitting the fence would mess up the hardware ring buffer. */ @@ -407,6 +410,57 @@ unsigned amdgpu_fence_count_emitted(struct amdgpu_ring *ring) } /** + * amdgpu_fence_last_unsignaled_time_us - the time fence emitted until now + * @ring: ring the fence is associated with + * + * Find the earliest fence unsignaled until now, calculate the time delta + * between the time fence emitted and now. + */ +u64 amdgpu_fence_last_unsignaled_time_us(struct amdgpu_ring *ring) +{ + struct amdgpu_fence_driver *drv = &ring->fence_drv; + struct dma_fence *fence; + uint32_t last_seq, sync_seq; + + last_seq = atomic_read(&ring->fence_drv.last_seq); + sync_seq = READ_ONCE(ring->fence_drv.sync_seq); + if (last_seq == sync_seq) + return 0; + + ++last_seq; + last_seq &= drv->num_fences_mask; + fence = drv->fences[last_seq]; + if (!fence) + return 0; + + return ktime_us_delta(ktime_get(), + to_amdgpu_fence(fence)->start_timestamp); +} + +/** + * amdgpu_fence_update_start_timestamp - update the timestamp of the fence + * @ring: ring the fence is associated with + * @seq: the fence seq number to update. + * @timestamp: the start timestamp to update. + * + * The function called at the time the fence and related ib is about to + * resubmit to gpu in MCBP scenario. Thus we do not consider race condition + * with amdgpu_fence_process to modify the same fence. + */ +void amdgpu_fence_update_start_timestamp(struct amdgpu_ring *ring, uint32_t seq, ktime_t timestamp) +{ + struct amdgpu_fence_driver *drv = &ring->fence_drv; + struct dma_fence *fence; + + seq &= drv->num_fences_mask; + fence = drv->fences[seq]; + if (!fence) + return; + + to_amdgpu_fence(fence)->start_timestamp = timestamp; +} + +/** * amdgpu_fence_driver_start_ring - make the fence driver * ready for use on the requested ring. * diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_fru_eeprom.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_fru_eeprom.c index e325150879df..4620c4712ce3 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_fru_eeprom.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_fru_eeprom.c @@ -29,9 +29,10 @@ #include "amdgpu_fru_eeprom.h" #include "amdgpu_eeprom.h" -#define FRU_EEPROM_MADDR 0x60000 +#define FRU_EEPROM_MADDR_6 0x60000 +#define FRU_EEPROM_MADDR_8 0x80000 -static bool is_fru_eeprom_supported(struct amdgpu_device *adev) +static bool is_fru_eeprom_supported(struct amdgpu_device *adev, u32 *fru_addr) { /* Only server cards have the FRU EEPROM * TODO: See if we can figure this out dynamically instead of @@ -45,6 +46,11 @@ static bool is_fru_eeprom_supported(struct amdgpu_device *adev) if (amdgpu_sriov_vf(adev)) return false; + /* The default I2C EEPROM address of the FRU. + */ + if (fru_addr) + *fru_addr = FRU_EEPROM_MADDR_8; + /* VBIOS is of the format ###-DXXXYYYY-##. For SKU identification, * we can use just the "DXXX" portion. If there were more models, we * could convert the 3 characters to a hex integer and use a switch @@ -57,21 +63,31 @@ static bool is_fru_eeprom_supported(struct amdgpu_device *adev) if (strnstr(atom_ctx->vbios_version, "D161", sizeof(atom_ctx->vbios_version)) || strnstr(atom_ctx->vbios_version, "D163", - sizeof(atom_ctx->vbios_version))) + sizeof(atom_ctx->vbios_version))) { + if (fru_addr) + *fru_addr = FRU_EEPROM_MADDR_6; return true; - else + } else { return false; + } case CHIP_ALDEBARAN: - /* All Aldebaran SKUs have the FRU */ + /* All Aldebaran SKUs have an FRU */ + if (!strnstr(atom_ctx->vbios_version, "D673", + sizeof(atom_ctx->vbios_version))) + if (fru_addr) + *fru_addr = FRU_EEPROM_MADDR_6; return true; case CHIP_SIENNA_CICHLID: if (strnstr(atom_ctx->vbios_version, "D603", - sizeof(atom_ctx->vbios_version))) { + sizeof(atom_ctx->vbios_version))) { if (strnstr(atom_ctx->vbios_version, "D603GLXE", - sizeof(atom_ctx->vbios_version))) + sizeof(atom_ctx->vbios_version))) { return false; - else + } else { + if (fru_addr) + *fru_addr = FRU_EEPROM_MADDR_6; return true; + } } else { return false; } @@ -80,41 +96,14 @@ static bool is_fru_eeprom_supported(struct amdgpu_device *adev) } } -static int amdgpu_fru_read_eeprom(struct amdgpu_device *adev, uint32_t addrptr, - unsigned char *buf, size_t buf_size) -{ - int ret; - u8 size; - - ret = amdgpu_eeprom_read(adev->pm.fru_eeprom_i2c_bus, addrptr, buf, 1); - if (ret < 1) { - DRM_WARN("FRU: Failed to get size field"); - return ret; - } - - /* The size returned by the i2c requires subtraction of 0xC0 since the - * size apparently always reports as 0xC0+actual size. - */ - size = buf[0] & 0x3F; - size = min_t(size_t, size, buf_size); - - ret = amdgpu_eeprom_read(adev->pm.fru_eeprom_i2c_bus, addrptr + 1, - buf, size); - if (ret < 1) { - DRM_WARN("FRU: Failed to get data field"); - return ret; - } - - return size; -} - int amdgpu_fru_get_product_info(struct amdgpu_device *adev) { - unsigned char buf[AMDGPU_PRODUCT_NAME_LEN]; - u32 addrptr; + unsigned char buf[8], *pia; + u32 addr, fru_addr; int size, len; + u8 csum; - if (!is_fru_eeprom_supported(adev)) + if (!is_fru_eeprom_supported(adev, &fru_addr)) return 0; /* If algo exists, it means that the i2c_adapter's initialized */ @@ -123,88 +112,102 @@ int amdgpu_fru_get_product_info(struct amdgpu_device *adev) return -ENODEV; } - /* There's a lot of repetition here. This is due to the FRU having - * variable-length fields. To get the information, we have to find the - * size of each field, and then keep reading along and reading along - * until we get all of the data that we want. We use addrptr to track - * the address as we go - */ - - /* The first fields are all of size 1-byte, from 0-7 are offsets that - * contain information that isn't useful to us. - * Bytes 8-a are all 1-byte and refer to the size of the entire struct, - * and the language field, so just start from 0xb, manufacturer size - */ - addrptr = FRU_EEPROM_MADDR + 0xb; - size = amdgpu_fru_read_eeprom(adev, addrptr, buf, sizeof(buf)); - if (size < 1) { - DRM_ERROR("Failed to read FRU Manufacturer, ret:%d", size); - return -EINVAL; + /* Read the IPMI Common header */ + len = amdgpu_eeprom_read(adev->pm.fru_eeprom_i2c_bus, fru_addr, buf, + sizeof(buf)); + if (len != 8) { + DRM_ERROR("Couldn't read the IPMI Common Header: %d", len); + return len < 0 ? len : -EIO; } - /* Increment the addrptr by the size of the field, and 1 due to the - * size field being 1 byte. This pattern continues below. - */ - addrptr += size + 1; - size = amdgpu_fru_read_eeprom(adev, addrptr, buf, sizeof(buf)); - if (size < 1) { - DRM_ERROR("Failed to read FRU product name, ret:%d", size); - return -EINVAL; + if (buf[0] != 1) { + DRM_ERROR("Bad IPMI Common Header version: 0x%02x", buf[0]); + return -EIO; } - len = size; - if (len >= AMDGPU_PRODUCT_NAME_LEN) { - DRM_WARN("FRU Product Name is larger than %d characters. This is likely a mistake", - AMDGPU_PRODUCT_NAME_LEN); - len = AMDGPU_PRODUCT_NAME_LEN - 1; - } - memcpy(adev->product_name, buf, len); - adev->product_name[len] = '\0'; - - addrptr += size + 1; - size = amdgpu_fru_read_eeprom(adev, addrptr, buf, sizeof(buf)); - if (size < 1) { - DRM_ERROR("Failed to read FRU product number, ret:%d", size); - return -EINVAL; + for (csum = 0; len > 0; len--) + csum += buf[len - 1]; + if (csum) { + DRM_ERROR("Bad IPMI Common Header checksum: 0x%02x", csum); + return -EIO; } - len = size; - /* Product number should only be 16 characters. Any more, - * and something could be wrong. Cap it at 16 to be safe - */ - if (len >= sizeof(adev->product_number)) { - DRM_WARN("FRU Product Number is larger than 16 characters. This is likely a mistake"); - len = sizeof(adev->product_number) - 1; - } - memcpy(adev->product_number, buf, len); - adev->product_number[len] = '\0'; + /* Get the offset to the Product Info Area (PIA). */ + addr = buf[4] * 8; + if (!addr) + return 0; - addrptr += size + 1; - size = amdgpu_fru_read_eeprom(adev, addrptr, buf, sizeof(buf)); + /* Get the absolute address to the PIA. */ + addr += fru_addr; - if (size < 1) { - DRM_ERROR("Failed to read FRU product version, ret:%d", size); - return -EINVAL; + /* Read the header of the PIA. */ + len = amdgpu_eeprom_read(adev->pm.fru_eeprom_i2c_bus, addr, buf, 3); + if (len != 3) { + DRM_ERROR("Couldn't read the Product Info Area header: %d", len); + return len < 0 ? len : -EIO; } - addrptr += size + 1; - size = amdgpu_fru_read_eeprom(adev, addrptr, buf, sizeof(buf)); + if (buf[0] != 1) { + DRM_ERROR("Bad IPMI Product Info Area version: 0x%02x", buf[0]); + return -EIO; + } - if (size < 1) { - DRM_ERROR("Failed to read FRU serial number, ret:%d", size); - return -EINVAL; + size = buf[1] * 8; + pia = kzalloc(size, GFP_KERNEL); + if (!pia) + return -ENOMEM; + + /* Read the whole PIA. */ + len = amdgpu_eeprom_read(adev->pm.fru_eeprom_i2c_bus, addr, pia, size); + if (len != size) { + kfree(pia); + DRM_ERROR("Couldn't read the Product Info Area: %d", len); + return len < 0 ? len : -EIO; } - len = size; - /* Serial number should only be 16 characters. Any more, - * and something could be wrong. Cap it at 16 to be safe - */ - if (len >= sizeof(adev->serial)) { - DRM_WARN("FRU Serial Number is larger than 16 characters. This is likely a mistake"); - len = sizeof(adev->serial) - 1; + for (csum = 0; size > 0; size--) + csum += pia[size - 1]; + if (csum) { + DRM_ERROR("Bad Product Info Area checksum: 0x%02x", csum); + return -EIO; } - memcpy(adev->serial, buf, len); - adev->serial[len] = '\0'; + /* Now extract useful information from the PIA. + * + * Skip the Manufacturer Name at [3] and go directly to + * the Product Name field. + */ + addr = 3 + 1 + (pia[3] & 0x3F); + if (addr + 1 >= len) + goto Out; + memcpy(adev->product_name, pia + addr + 1, + min_t(size_t, + sizeof(adev->product_name), + pia[addr] & 0x3F)); + adev->product_name[sizeof(adev->product_name) - 1] = '\0'; + + /* Go to the Product Part/Model Number field. */ + addr += 1 + (pia[addr] & 0x3F); + if (addr + 1 >= len) + goto Out; + memcpy(adev->product_number, pia + addr + 1, + min_t(size_t, + sizeof(adev->product_number), + pia[addr] & 0x3F)); + adev->product_number[sizeof(adev->product_number) - 1] = '\0'; + + /* Go to the Product Version field. */ + addr += 1 + (pia[addr] & 0x3F); + + /* Go to the Product Serial Number field. */ + addr += 1 + (pia[addr] & 0x3F); + if (addr + 1 >= len) + goto Out; + memcpy(adev->serial, pia + addr + 1, min_t(size_t, + sizeof(adev->serial), + pia[addr] & 0x3F)); + adev->serial[sizeof(adev->serial) - 1] = '\0'; +Out: + kfree(pia); return 0; } diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_gem.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_gem.c index 91571b1324f2..bb7350ea1d75 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_gem.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_gem.c @@ -38,6 +38,7 @@ #include "amdgpu.h" #include "amdgpu_display.h" #include "amdgpu_dma_buf.h" +#include "amdgpu_hmm.h" #include "amdgpu_xgmi.h" static const struct drm_gem_object_funcs amdgpu_gem_object_funcs; @@ -87,7 +88,7 @@ static void amdgpu_gem_object_free(struct drm_gem_object *gobj) struct amdgpu_bo *robj = gem_to_amdgpu_bo(gobj); if (robj) { - amdgpu_mn_unregister(robj); + amdgpu_hmm_unregister(robj); amdgpu_bo_unref(&robj); } } @@ -255,7 +256,7 @@ static int amdgpu_gem_object_mmap(struct drm_gem_object *obj, struct vm_area_str * becoming writable and makes is_cow_mapping(vm_flags) false. */ if (is_cow_mapping(vma->vm_flags) && - !(vma->vm_flags & (VM_READ | VM_WRITE | VM_EXEC))) + !(vma->vm_flags & VM_ACCESS_FLAGS)) vma->vm_flags &= ~VM_MAYWRITE; return drm_gem_ttm_mmap(obj, vma); @@ -414,7 +415,7 @@ int amdgpu_gem_userptr_ioctl(struct drm_device *dev, void *data, if (r) goto release_object; - r = amdgpu_mn_register(bo, args->addr); + r = amdgpu_hmm_register(bo, args->addr); if (r) goto release_object; diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_gfx.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_gfx.c index 9546adc8a76f..23692e5d4d13 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_gfx.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_gfx.c @@ -583,10 +583,14 @@ void amdgpu_gfx_off_ctrl(struct amdgpu_device *adev, bool enable) if (adev->gfx.gfx_off_req_count == 0 && !adev->gfx.gfx_off_state) { /* If going to s2idle, no need to wait */ - if (adev->in_s0ix) - delay = GFX_OFF_NO_DELAY; - schedule_delayed_work(&adev->gfx.gfx_off_delay_work, + if (adev->in_s0ix) { + if (!amdgpu_dpm_set_powergating_by_smu(adev, + AMD_IP_BLOCK_TYPE_GFX, true)) + adev->gfx.gfx_off_state = true; + } else { + schedule_delayed_work(&adev->gfx.gfx_off_delay_work, delay); + } } } else { if (adev->gfx.gfx_off_req_count == 0) { diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_gfx.h b/drivers/gpu/drm/amd/amdgpu/amdgpu_gfx.h index 832b3807f1d6..b3df4787877e 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_gfx.h +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_gfx.h @@ -33,6 +33,7 @@ #include "amdgpu_imu.h" #include "soc15.h" #include "amdgpu_ras.h" +#include "amdgpu_ring_mux.h" /* GFX current status */ #define AMDGPU_GFX_NORMAL_MODE 0x00000000L @@ -352,6 +353,9 @@ struct amdgpu_gfx { struct amdgpu_gfx_ras *ras; bool is_poweron; + + struct amdgpu_ring sw_gfx_ring[AMDGPU_MAX_SW_GFX_RINGS]; + struct amdgpu_ring_mux muxer; }; #define amdgpu_gfx_get_gpu_clock_counter(adev) (adev)->gfx.funcs->get_gpu_clock_counter((adev)) diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_gmc.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_gmc.c index 28612e56d0d4..02a4c93673ce 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_gmc.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_gmc.c @@ -548,6 +548,8 @@ void amdgpu_gmc_tmz_set(struct amdgpu_device *adev) case IP_VERSION(10, 3, 1): /* YELLOW_CARP*/ case IP_VERSION(10, 3, 3): + case IP_VERSION(11, 0, 1): + case IP_VERSION(11, 0, 4): /* Don't enable it by default yet. */ if (amdgpu_tmz < 1) { diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_gtt_mgr.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_gtt_mgr.c index 1f3302aebeff..44367f03316f 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_gtt_mgr.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_gtt_mgr.c @@ -144,7 +144,7 @@ static int amdgpu_gtt_mgr_new(struct ttm_resource_manager *man, node->base.start = node->mm_nodes[0].start; } else { node->mm_nodes[0].start = 0; - node->mm_nodes[0].size = node->base.num_pages; + node->mm_nodes[0].size = PFN_UP(node->base.size); node->base.start = AMDGPU_BO_INVALID_OFFSET; } diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_mn.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_hmm.c index b86c0b8252a5..2dadcfe43d03 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_mn.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_hmm.c @@ -49,9 +49,12 @@ #include "amdgpu.h" #include "amdgpu_amdkfd.h" +#include "amdgpu_hmm.h" + +#define MAX_WALK_BYTE (2UL << 30) /** - * amdgpu_mn_invalidate_gfx - callback to notify about mm change + * amdgpu_hmm_invalidate_gfx - callback to notify about mm change * * @mni: the range (mm) is about to update * @range: details on the invalidation @@ -60,9 +63,9 @@ * Block for operations on BOs to finish and mark pages as accessed and * potentially dirty. */ -static bool amdgpu_mn_invalidate_gfx(struct mmu_interval_notifier *mni, - const struct mmu_notifier_range *range, - unsigned long cur_seq) +static bool amdgpu_hmm_invalidate_gfx(struct mmu_interval_notifier *mni, + const struct mmu_notifier_range *range, + unsigned long cur_seq) { struct amdgpu_bo *bo = container_of(mni, struct amdgpu_bo, notifier); struct amdgpu_device *adev = amdgpu_ttm_adev(bo->tbo.bdev); @@ -83,12 +86,12 @@ static bool amdgpu_mn_invalidate_gfx(struct mmu_interval_notifier *mni, return true; } -static const struct mmu_interval_notifier_ops amdgpu_mn_gfx_ops = { - .invalidate = amdgpu_mn_invalidate_gfx, +static const struct mmu_interval_notifier_ops amdgpu_hmm_gfx_ops = { + .invalidate = amdgpu_hmm_invalidate_gfx, }; /** - * amdgpu_mn_invalidate_hsa - callback to notify about mm change + * amdgpu_hmm_invalidate_hsa - callback to notify about mm change * * @mni: the range (mm) is about to update * @range: details on the invalidation @@ -97,32 +100,26 @@ static const struct mmu_interval_notifier_ops amdgpu_mn_gfx_ops = { * We temporarily evict the BO attached to this range. This necessitates * evicting all user-mode queues of the process. */ -static bool amdgpu_mn_invalidate_hsa(struct mmu_interval_notifier *mni, - const struct mmu_notifier_range *range, - unsigned long cur_seq) +static bool amdgpu_hmm_invalidate_hsa(struct mmu_interval_notifier *mni, + const struct mmu_notifier_range *range, + unsigned long cur_seq) { struct amdgpu_bo *bo = container_of(mni, struct amdgpu_bo, notifier); - struct amdgpu_device *adev = amdgpu_ttm_adev(bo->tbo.bdev); if (!mmu_notifier_range_blockable(range)) return false; - mutex_lock(&adev->notifier_lock); - - mmu_interval_set_seq(mni, cur_seq); - - amdgpu_amdkfd_evict_userptr(bo->kfd_bo, bo->notifier.mm); - mutex_unlock(&adev->notifier_lock); + amdgpu_amdkfd_evict_userptr(mni, cur_seq, bo->kfd_bo); return true; } -static const struct mmu_interval_notifier_ops amdgpu_mn_hsa_ops = { - .invalidate = amdgpu_mn_invalidate_hsa, +static const struct mmu_interval_notifier_ops amdgpu_hmm_hsa_ops = { + .invalidate = amdgpu_hmm_invalidate_hsa, }; /** - * amdgpu_mn_register - register a BO for notifier updates + * amdgpu_hmm_register - register a BO for notifier updates * * @bo: amdgpu buffer object * @addr: userptr addr we should monitor @@ -130,25 +127,25 @@ static const struct mmu_interval_notifier_ops amdgpu_mn_hsa_ops = { * Registers a mmu_notifier for the given BO at the specified address. * Returns 0 on success, -ERRNO if anything goes wrong. */ -int amdgpu_mn_register(struct amdgpu_bo *bo, unsigned long addr) +int amdgpu_hmm_register(struct amdgpu_bo *bo, unsigned long addr) { if (bo->kfd_bo) return mmu_interval_notifier_insert(&bo->notifier, current->mm, addr, amdgpu_bo_size(bo), - &amdgpu_mn_hsa_ops); + &amdgpu_hmm_hsa_ops); return mmu_interval_notifier_insert(&bo->notifier, current->mm, addr, amdgpu_bo_size(bo), - &amdgpu_mn_gfx_ops); + &amdgpu_hmm_gfx_ops); } /** - * amdgpu_mn_unregister - unregister a BO for notifier updates + * amdgpu_hmm_unregister - unregister a BO for notifier updates * * @bo: amdgpu buffer object * * Remove any registration of mmu notifier updates from the buffer object. */ -void amdgpu_mn_unregister(struct amdgpu_bo *bo) +void amdgpu_hmm_unregister(struct amdgpu_bo *bo) { if (!bo->notifier.mm) return; @@ -157,12 +154,12 @@ void amdgpu_mn_unregister(struct amdgpu_bo *bo) } int amdgpu_hmm_range_get_pages(struct mmu_interval_notifier *notifier, - struct mm_struct *mm, struct page **pages, - uint64_t start, uint64_t npages, - struct hmm_range **phmm_range, bool readonly, - bool mmap_locked, void *owner) + uint64_t start, uint64_t npages, bool readonly, + void *owner, struct page **pages, + struct hmm_range **phmm_range) { struct hmm_range *hmm_range; + unsigned long end; unsigned long timeout; unsigned long i; unsigned long *pfns; @@ -184,32 +181,42 @@ int amdgpu_hmm_range_get_pages(struct mmu_interval_notifier *notifier, hmm_range->default_flags |= HMM_PFN_REQ_WRITE; hmm_range->hmm_pfns = pfns; hmm_range->start = start; - hmm_range->end = start + npages * PAGE_SIZE; + end = start + npages * PAGE_SIZE; hmm_range->dev_private_owner = owner; - /* Assuming 512MB takes maxmium 1 second to fault page address */ - timeout = max(npages >> 17, 1ULL) * HMM_RANGE_DEFAULT_TIMEOUT; - timeout = jiffies + msecs_to_jiffies(timeout); + do { + hmm_range->end = min(hmm_range->start + MAX_WALK_BYTE, end); + + pr_debug("hmm range: start = 0x%lx, end = 0x%lx", + hmm_range->start, hmm_range->end); + + /* Assuming 512MB takes maxmium 1 second to fault page address */ + timeout = max((hmm_range->end - hmm_range->start) >> 29, 1UL); + timeout *= HMM_RANGE_DEFAULT_TIMEOUT; + timeout = jiffies + msecs_to_jiffies(timeout); retry: - hmm_range->notifier_seq = mmu_interval_read_begin(notifier); - - if (likely(!mmap_locked)) - mmap_read_lock(mm); - - r = hmm_range_fault(hmm_range); - - if (likely(!mmap_locked)) - mmap_read_unlock(mm); - if (unlikely(r)) { - /* - * FIXME: This timeout should encompass the retry from - * mmu_interval_read_retry() as well. - */ - if (r == -EBUSY && !time_after(jiffies, timeout)) - goto retry; - goto out_free_pfns; - } + hmm_range->notifier_seq = mmu_interval_read_begin(notifier); + r = hmm_range_fault(hmm_range); + if (unlikely(r)) { + /* + * FIXME: This timeout should encompass the retry from + * mmu_interval_read_retry() as well. + */ + if (r == -EBUSY && !time_after(jiffies, timeout)) + goto retry; + goto out_free_pfns; + } + + if (hmm_range->end == end) + break; + hmm_range->hmm_pfns += MAX_WALK_BYTE >> PAGE_SHIFT; + hmm_range->start = hmm_range->end; + schedule(); + } while (hmm_range->end < end); + + hmm_range->start = start; + hmm_range->hmm_pfns = pfns; /* * Due to default_flags, all pages are HMM_PFN_VALID or @@ -231,9 +238,9 @@ out_free_range: return r; } -int amdgpu_hmm_range_get_pages_done(struct hmm_range *hmm_range) +bool amdgpu_hmm_range_get_pages_done(struct hmm_range *hmm_range) { - int r; + bool r; r = mmu_interval_read_retry(hmm_range->notifier, hmm_range->notifier_seq); diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_mn.h b/drivers/gpu/drm/amd/amdgpu/amdgpu_hmm.h index 14a3c1864085..e2edcd010ccc 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_mn.h +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_hmm.h @@ -29,25 +29,25 @@ #include <linux/rwsem.h> #include <linux/workqueue.h> #include <linux/interval_tree.h> +#include <linux/mmu_notifier.h> int amdgpu_hmm_range_get_pages(struct mmu_interval_notifier *notifier, - struct mm_struct *mm, struct page **pages, - uint64_t start, uint64_t npages, - struct hmm_range **phmm_range, bool readonly, - bool mmap_locked, void *owner); -int amdgpu_hmm_range_get_pages_done(struct hmm_range *hmm_range); + uint64_t start, uint64_t npages, bool readonly, + void *owner, struct page **pages, + struct hmm_range **phmm_range); +bool amdgpu_hmm_range_get_pages_done(struct hmm_range *hmm_range); #if defined(CONFIG_HMM_MIRROR) -int amdgpu_mn_register(struct amdgpu_bo *bo, unsigned long addr); -void amdgpu_mn_unregister(struct amdgpu_bo *bo); +int amdgpu_hmm_register(struct amdgpu_bo *bo, unsigned long addr); +void amdgpu_hmm_unregister(struct amdgpu_bo *bo); #else -static inline int amdgpu_mn_register(struct amdgpu_bo *bo, unsigned long addr) +static inline int amdgpu_hmm_register(struct amdgpu_bo *bo, unsigned long addr) { DRM_WARN_ONCE("HMM_MIRROR kernel config option is not enabled, " "add CONFIG_ZONE_DEVICE=y in config file to fix this\n"); return -ENODEV; } -static inline void amdgpu_mn_unregister(struct amdgpu_bo *bo) {} +static inline void amdgpu_hmm_unregister(struct amdgpu_bo *bo) {} #endif #endif diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_ib.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_ib.c index 258cffe3c06a..bcccc348dbe2 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_ib.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_ib.c @@ -182,7 +182,7 @@ int amdgpu_ib_schedule(struct amdgpu_ring *ring, unsigned num_ibs, need_ctx_switch = ring->current_ctx != fence_ctx; if (ring->funcs->emit_pipeline_sync && job && - ((tmp = amdgpu_sync_get_fence(&job->sched_sync)) || + ((tmp = amdgpu_sync_get_fence(&job->explicit_sync)) || (amdgpu_sriov_vf(adev) && need_ctx_switch) || amdgpu_vm_need_pipeline_sync(ring, job))) { need_pipe_sync = true; @@ -211,6 +211,7 @@ int amdgpu_ib_schedule(struct amdgpu_ring *ring, unsigned num_ibs, } } + amdgpu_ring_ib_begin(ring); if (job && ring->funcs->init_cond_exec) patch_offset = amdgpu_ring_init_cond_exec(ring); @@ -285,6 +286,7 @@ int amdgpu_ib_schedule(struct amdgpu_ring *ring, unsigned num_ibs, ring->hw_prio == AMDGPU_GFX_PIPE_PRIO_HIGH) ring->funcs->emit_wave_limit(ring, false); + amdgpu_ring_ib_end(ring); amdgpu_ring_commit(ring); return 0; } diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_ids.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_ids.c index 03d115d2b5ed..fcb711a11a5b 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_ids.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_ids.c @@ -165,31 +165,52 @@ bool amdgpu_vmid_had_gpu_reset(struct amdgpu_device *adev, atomic_read(&adev->gpu_reset_counter); } +/* Check if we need to switch to another set of resources */ +static bool amdgpu_vmid_gds_switch_needed(struct amdgpu_vmid *id, + struct amdgpu_job *job) +{ + return id->gds_base != job->gds_base || + id->gds_size != job->gds_size || + id->gws_base != job->gws_base || + id->gws_size != job->gws_size || + id->oa_base != job->oa_base || + id->oa_size != job->oa_size; +} + +/* Check if the id is compatible with the job */ +static bool amdgpu_vmid_compatible(struct amdgpu_vmid *id, + struct amdgpu_job *job) +{ + return id->pd_gpu_addr == job->vm_pd_addr && + !amdgpu_vmid_gds_switch_needed(id, job); +} + /** * amdgpu_vmid_grab_idle - grab idle VMID * * @vm: vm to allocate id for * @ring: ring we want to submit job to - * @sync: sync object where we add dependencies * @idle: resulting idle VMID + * @fence: fence to wait for if no id could be grabbed * * Try to find an idle VMID, if none is idle add a fence to wait to the sync * object. Returns -ENOMEM when we are out of memory. */ static int amdgpu_vmid_grab_idle(struct amdgpu_vm *vm, struct amdgpu_ring *ring, - struct amdgpu_sync *sync, - struct amdgpu_vmid **idle) + struct amdgpu_vmid **idle, + struct dma_fence **fence) { struct amdgpu_device *adev = ring->adev; unsigned vmhub = ring->funcs->vmhub; struct amdgpu_vmid_mgr *id_mgr = &adev->vm_manager.id_mgr[vmhub]; struct dma_fence **fences; unsigned i; - int r; - if (!dma_fence_is_signaled(ring->vmid_wait)) - return amdgpu_sync_fence(sync, ring->vmid_wait); + if (!dma_fence_is_signaled(ring->vmid_wait)) { + *fence = dma_fence_get(ring->vmid_wait); + return 0; + } fences = kmalloc_array(id_mgr->num_ids, sizeof(void *), GFP_KERNEL); if (!fences) @@ -228,10 +249,10 @@ static int amdgpu_vmid_grab_idle(struct amdgpu_vm *vm, return -ENOMEM; } - r = amdgpu_sync_fence(sync, &array->base); + *fence = dma_fence_get(&array->base); dma_fence_put(ring->vmid_wait); ring->vmid_wait = &array->base; - return r; + return 0; } kfree(fences); @@ -243,30 +264,29 @@ static int amdgpu_vmid_grab_idle(struct amdgpu_vm *vm, * * @vm: vm to allocate id for * @ring: ring we want to submit job to - * @sync: sync object where we add dependencies - * @fence: fence protecting ID from reuse * @job: job who wants to use the VMID * @id: resulting VMID + * @fence: fence to wait for if no id could be grabbed * * Try to assign a reserved VMID. */ static int amdgpu_vmid_grab_reserved(struct amdgpu_vm *vm, struct amdgpu_ring *ring, - struct amdgpu_sync *sync, - struct dma_fence *fence, struct amdgpu_job *job, - struct amdgpu_vmid **id) + struct amdgpu_vmid **id, + struct dma_fence **fence) { struct amdgpu_device *adev = ring->adev; unsigned vmhub = ring->funcs->vmhub; + struct amdgpu_vmid_mgr *id_mgr = &adev->vm_manager.id_mgr[vmhub]; uint64_t fence_context = adev->fence_context + ring->idx; bool needs_flush = vm->use_cpu_for_update; uint64_t updates = amdgpu_vm_tlb_seq(vm); int r; - *id = vm->reserved_vmid[vmhub]; + *id = id_mgr->reserved; if ((*id)->owner != vm->immediate.fence_context || - (*id)->pd_gpu_addr != job->vm_pd_addr || + !amdgpu_vmid_compatible(*id, job) || (*id)->flushed_updates < updates || !(*id)->last_flush || ((*id)->last_flush->context != fence_context && @@ -282,7 +302,8 @@ static int amdgpu_vmid_grab_reserved(struct amdgpu_vm *vm, tmp = amdgpu_sync_peek_fence(&(*id)->active, ring); if (tmp) { *id = NULL; - return amdgpu_sync_fence(sync, tmp); + *fence = dma_fence_get(tmp); + return 0; } needs_flush = true; } @@ -290,12 +311,12 @@ static int amdgpu_vmid_grab_reserved(struct amdgpu_vm *vm, /* Good we can use this VMID. Remember this submission as * user of the VMID. */ - r = amdgpu_sync_fence(&(*id)->active, fence); + r = amdgpu_sync_fence(&(*id)->active, &job->base.s_fence->finished); if (r) return r; - (*id)->flushed_updates = updates; job->vm_needs_flush = needs_flush; + job->spm_update_needed = true; return 0; } @@ -304,19 +325,17 @@ static int amdgpu_vmid_grab_reserved(struct amdgpu_vm *vm, * * @vm: vm to allocate id for * @ring: ring we want to submit job to - * @sync: sync object where we add dependencies - * @fence: fence protecting ID from reuse * @job: job who wants to use the VMID * @id: resulting VMID + * @fence: fence to wait for if no id could be grabbed * * Try to reuse a VMID for this submission. */ static int amdgpu_vmid_grab_used(struct amdgpu_vm *vm, struct amdgpu_ring *ring, - struct amdgpu_sync *sync, - struct dma_fence *fence, struct amdgpu_job *job, - struct amdgpu_vmid **id) + struct amdgpu_vmid **id, + struct dma_fence **fence) { struct amdgpu_device *adev = ring->adev; unsigned vmhub = ring->funcs->vmhub; @@ -335,7 +354,7 @@ static int amdgpu_vmid_grab_used(struct amdgpu_vm *vm, if ((*id)->owner != vm->immediate.fence_context) continue; - if ((*id)->pd_gpu_addr != job->vm_pd_addr) + if (!amdgpu_vmid_compatible(*id, job)) continue; if (!(*id)->last_flush || @@ -352,11 +371,11 @@ static int amdgpu_vmid_grab_used(struct amdgpu_vm *vm, /* Good, we can use this VMID. Remember this submission as * user of the VMID. */ - r = amdgpu_sync_fence(&(*id)->active, fence); + r = amdgpu_sync_fence(&(*id)->active, + &job->base.s_fence->finished); if (r) return r; - (*id)->flushed_updates = updates; job->vm_needs_flush |= needs_flush; return 0; } @@ -370,15 +389,13 @@ static int amdgpu_vmid_grab_used(struct amdgpu_vm *vm, * * @vm: vm to allocate id for * @ring: ring we want to submit job to - * @sync: sync object where we add dependencies - * @fence: fence protecting ID from reuse * @job: job who wants to use the VMID + * @fence: fence to wait for if no id could be grabbed * * Allocate an id for the vm, adding fences to the sync obj as necessary. */ int amdgpu_vmid_grab(struct amdgpu_vm *vm, struct amdgpu_ring *ring, - struct amdgpu_sync *sync, struct dma_fence *fence, - struct amdgpu_job *job) + struct amdgpu_job *job, struct dma_fence **fence) { struct amdgpu_device *adev = ring->adev; unsigned vmhub = ring->funcs->vmhub; @@ -388,16 +405,16 @@ int amdgpu_vmid_grab(struct amdgpu_vm *vm, struct amdgpu_ring *ring, int r = 0; mutex_lock(&id_mgr->lock); - r = amdgpu_vmid_grab_idle(vm, ring, sync, &idle); + r = amdgpu_vmid_grab_idle(vm, ring, &idle, fence); if (r || !idle) goto error; if (vm->reserved_vmid[vmhub]) { - r = amdgpu_vmid_grab_reserved(vm, ring, sync, fence, job, &id); + r = amdgpu_vmid_grab_reserved(vm, ring, job, &id, fence); if (r || !id) goto error; } else { - r = amdgpu_vmid_grab_used(vm, ring, sync, fence, job, &id); + r = amdgpu_vmid_grab_used(vm, ring, job, &id, fence); if (r) goto error; @@ -406,26 +423,35 @@ int amdgpu_vmid_grab(struct amdgpu_vm *vm, struct amdgpu_ring *ring, id = idle; /* Remember this submission as user of the VMID */ - r = amdgpu_sync_fence(&id->active, fence); + r = amdgpu_sync_fence(&id->active, + &job->base.s_fence->finished); if (r) goto error; - id->flushed_updates = amdgpu_vm_tlb_seq(vm); job->vm_needs_flush = true; } list_move_tail(&id->list, &id_mgr->ids_lru); } - id->pd_gpu_addr = job->vm_pd_addr; - id->owner = vm->immediate.fence_context; - + job->gds_switch_needed = amdgpu_vmid_gds_switch_needed(id, job); if (job->vm_needs_flush) { + id->flushed_updates = amdgpu_vm_tlb_seq(vm); dma_fence_put(id->last_flush); id->last_flush = NULL; } job->vmid = id - id_mgr->ids; job->pasid = vm->pasid; + + id->gds_base = job->gds_base; + id->gds_size = job->gds_size; + id->gws_base = job->gws_base; + id->gws_size = job->gws_size; + id->oa_base = job->oa_base; + id->oa_size = job->oa_size; + id->pd_gpu_addr = job->vm_pd_addr; + id->owner = vm->immediate.fence_context; + trace_amdgpu_vm_grab_id(vm, ring, job); error: @@ -437,31 +463,27 @@ int amdgpu_vmid_alloc_reserved(struct amdgpu_device *adev, struct amdgpu_vm *vm, unsigned vmhub) { - struct amdgpu_vmid_mgr *id_mgr; - struct amdgpu_vmid *idle; - int r = 0; + struct amdgpu_vmid_mgr *id_mgr = &adev->vm_manager.id_mgr[vmhub]; - id_mgr = &adev->vm_manager.id_mgr[vmhub]; mutex_lock(&id_mgr->lock); if (vm->reserved_vmid[vmhub]) goto unlock; - if (atomic_inc_return(&id_mgr->reserved_vmid_num) > - AMDGPU_VM_MAX_RESERVED_VMID) { - DRM_ERROR("Over limitation of reserved vmid\n"); - atomic_dec(&id_mgr->reserved_vmid_num); - r = -EINVAL; - goto unlock; + + ++id_mgr->reserved_use_count; + if (!id_mgr->reserved) { + struct amdgpu_vmid *id; + + id = list_first_entry(&id_mgr->ids_lru, struct amdgpu_vmid, + list); + /* Remove from normal round robin handling */ + list_del_init(&id->list); + id_mgr->reserved = id; } - /* Select the first entry VMID */ - idle = list_first_entry(&id_mgr->ids_lru, struct amdgpu_vmid, list); - list_del_init(&idle->list); - vm->reserved_vmid[vmhub] = idle; - mutex_unlock(&id_mgr->lock); + vm->reserved_vmid[vmhub] = true; - return 0; unlock: mutex_unlock(&id_mgr->lock); - return r; + return 0; } void amdgpu_vmid_free_reserved(struct amdgpu_device *adev, @@ -471,12 +493,12 @@ void amdgpu_vmid_free_reserved(struct amdgpu_device *adev, struct amdgpu_vmid_mgr *id_mgr = &adev->vm_manager.id_mgr[vmhub]; mutex_lock(&id_mgr->lock); - if (vm->reserved_vmid[vmhub]) { - list_add(&vm->reserved_vmid[vmhub]->list, - &id_mgr->ids_lru); - vm->reserved_vmid[vmhub] = NULL; - atomic_dec(&id_mgr->reserved_vmid_num); + if (vm->reserved_vmid[vmhub] && + !--id_mgr->reserved_use_count) { + /* give the reserved ID back to normal round robin */ + list_add(&id_mgr->reserved->list, &id_mgr->ids_lru); } + vm->reserved_vmid[vmhub] = false; mutex_unlock(&id_mgr->lock); } @@ -543,7 +565,7 @@ void amdgpu_vmid_mgr_init(struct amdgpu_device *adev) mutex_init(&id_mgr->lock); INIT_LIST_HEAD(&id_mgr->ids_lru); - atomic_set(&id_mgr->reserved_vmid_num, 0); + id_mgr->reserved_use_count = 0; /* manage only VMIDs not used by KFD */ id_mgr->num_ids = adev->vm_manager.first_kfd_vmid; diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_ids.h b/drivers/gpu/drm/amd/amdgpu/amdgpu_ids.h index 06c8a0034fa5..d1cc09b45da4 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_ids.h +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_ids.h @@ -67,7 +67,8 @@ struct amdgpu_vmid_mgr { unsigned num_ids; struct list_head ids_lru; struct amdgpu_vmid ids[AMDGPU_NUM_VMID]; - atomic_t reserved_vmid_num; + struct amdgpu_vmid *reserved; + unsigned int reserved_use_count; }; int amdgpu_pasid_alloc(unsigned int bits); @@ -84,8 +85,7 @@ void amdgpu_vmid_free_reserved(struct amdgpu_device *adev, struct amdgpu_vm *vm, unsigned vmhub); int amdgpu_vmid_grab(struct amdgpu_vm *vm, struct amdgpu_ring *ring, - struct amdgpu_sync *sync, struct dma_fence *fence, - struct amdgpu_job *job); + struct amdgpu_job *job, struct dma_fence **fence); void amdgpu_vmid_reset(struct amdgpu_device *adev, unsigned vmhub, unsigned vmid); void amdgpu_vmid_reset_all(struct amdgpu_device *adev); diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_irq.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_irq.c index 89011bae7588..a6aef488a822 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_irq.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_irq.c @@ -101,41 +101,6 @@ const char *soc15_ih_clientid_name[] = { }; /** - * amdgpu_hotplug_work_func - work handler for display hotplug event - * - * @work: work struct pointer - * - * This is the hotplug event work handler (all ASICs). - * The work gets scheduled from the IRQ handler if there - * was a hotplug interrupt. It walks through the connector table - * and calls hotplug handler for each connector. After this, it sends - * a DRM hotplug event to alert userspace. - * - * This design approach is required in order to defer hotplug event handling - * from the IRQ handler to a work handler because hotplug handler has to use - * mutexes which cannot be locked in an IRQ handler (since &mutex_lock may - * sleep). - */ -static void amdgpu_hotplug_work_func(struct work_struct *work) -{ - struct amdgpu_device *adev = container_of(work, struct amdgpu_device, - hotplug_work); - struct drm_device *dev = adev_to_drm(adev); - struct drm_mode_config *mode_config = &dev->mode_config; - struct drm_connector *connector; - struct drm_connector_list_iter iter; - - mutex_lock(&mode_config->mutex); - drm_connector_list_iter_begin(dev, &iter); - drm_for_each_connector_iter(connector, &iter) - amdgpu_connector_hotplug(connector); - drm_connector_list_iter_end(&iter); - mutex_unlock(&mode_config->mutex); - /* Just fire off a uevent and let userspace tell us what to do */ - drm_helper_hpd_irq_event(dev); -} - -/** * amdgpu_irq_disable_all - disable *all* interrupts * * @adev: amdgpu device pointer @@ -317,21 +282,6 @@ int amdgpu_irq_init(struct amdgpu_device *adev) } } - if (!amdgpu_device_has_dc_support(adev)) { - if (!adev->enable_virtual_display) - /* Disable vblank IRQs aggressively for power-saving */ - /* XXX: can this be enabled for DC? */ - adev_to_drm(adev)->vblank_disable_immediate = true; - - r = drm_vblank_init(adev_to_drm(adev), adev->mode_info.num_crtc); - if (r) - return r; - - /* Pre-DCE11 */ - INIT_WORK(&adev->hotplug_work, - amdgpu_hotplug_work_func); - } - INIT_WORK(&adev->irq.ih1_work, amdgpu_irq_handle_ih1); INIT_WORK(&adev->irq.ih2_work, amdgpu_irq_handle_ih2); INIT_WORK(&adev->irq.ih_soft_work, amdgpu_irq_handle_ih_soft); @@ -345,11 +295,8 @@ int amdgpu_irq_init(struct amdgpu_device *adev) /* PCI devices require shared interrupts. */ r = request_irq(irq, amdgpu_irq_handler, IRQF_SHARED, adev_to_drm(adev)->driver->name, adev_to_drm(adev)); - if (r) { - if (!amdgpu_device_has_dc_support(adev)) - flush_work(&adev->hotplug_work); + if (r) return r; - } adev->irq.installed = true; adev->irq.irq = irq; adev_to_drm(adev)->max_vblank_count = 0x00ffffff; @@ -366,9 +313,6 @@ void amdgpu_irq_fini_hw(struct amdgpu_device *adev) adev->irq.installed = false; if (adev->irq.msi_enabled) pci_free_irq_vectors(adev->pdev); - - if (!amdgpu_device_has_dc_support(adev)) - flush_work(&adev->hotplug_work); } amdgpu_ih_ring_fini(adev, &adev->irq.ih_soft); diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_job.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_job.c index adac650cf544..9e549923622b 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_job.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_job.c @@ -88,8 +88,9 @@ exit: return DRM_GPU_SCHED_STAT_NOMINAL; } -int amdgpu_job_alloc(struct amdgpu_device *adev, unsigned num_ibs, - struct amdgpu_job **job, struct amdgpu_vm *vm) +int amdgpu_job_alloc(struct amdgpu_device *adev, struct amdgpu_vm *vm, + struct drm_sched_entity *entity, void *owner, + unsigned int num_ibs, struct amdgpu_job **job) { if (num_ibs == 0) return -EINVAL; @@ -105,28 +106,34 @@ int amdgpu_job_alloc(struct amdgpu_device *adev, unsigned num_ibs, (*job)->base.sched = &adev->rings[0]->sched; (*job)->vm = vm; - amdgpu_sync_create(&(*job)->sync); - amdgpu_sync_create(&(*job)->sched_sync); + amdgpu_sync_create(&(*job)->explicit_sync); (*job)->vram_lost_counter = atomic_read(&adev->vram_lost_counter); (*job)->vm_pd_addr = AMDGPU_BO_INVALID_OFFSET; - return 0; + if (!entity) + return 0; + + return drm_sched_job_init(&(*job)->base, entity, owner); } -int amdgpu_job_alloc_with_ib(struct amdgpu_device *adev, unsigned size, - enum amdgpu_ib_pool_type pool_type, - struct amdgpu_job **job) +int amdgpu_job_alloc_with_ib(struct amdgpu_device *adev, + struct drm_sched_entity *entity, void *owner, + size_t size, enum amdgpu_ib_pool_type pool_type, + struct amdgpu_job **job) { int r; - r = amdgpu_job_alloc(adev, 1, job, NULL); + r = amdgpu_job_alloc(adev, NULL, entity, owner, 1, job); if (r) return r; (*job)->num_ibs = 1; r = amdgpu_ib_get(adev, NULL, size, pool_type, &(*job)->ibs[0]); - if (r) + if (r) { + if (entity) + drm_sched_job_cleanup(&(*job)->base); kfree(*job); + } return r; } @@ -166,8 +173,7 @@ static void amdgpu_job_free_cb(struct drm_sched_job *s_job) drm_sched_job_cleanup(s_job); - amdgpu_sync_free(&job->sync); - amdgpu_sync_free(&job->sched_sync); + amdgpu_sync_free(&job->explicit_sync); /* only put the hw fence if has embedded fence */ if (!job->hw_fence.ops) @@ -194,9 +200,11 @@ void amdgpu_job_set_gang_leader(struct amdgpu_job *job, void amdgpu_job_free(struct amdgpu_job *job) { + if (job->base.entity) + drm_sched_job_cleanup(&job->base); + amdgpu_job_free_resources(job); - amdgpu_sync_free(&job->sync); - amdgpu_sync_free(&job->sched_sync); + amdgpu_sync_free(&job->explicit_sync); if (job->gang_submit != &job->base.s_fence->scheduled) dma_fence_put(job->gang_submit); @@ -206,25 +214,16 @@ void amdgpu_job_free(struct amdgpu_job *job) dma_fence_put(&job->hw_fence); } -int amdgpu_job_submit(struct amdgpu_job *job, struct drm_sched_entity *entity, - void *owner, struct dma_fence **f) +struct dma_fence *amdgpu_job_submit(struct amdgpu_job *job) { - int r; - - if (!f) - return -EINVAL; - - r = drm_sched_job_init(&job->base, entity, owner); - if (r) - return r; + struct dma_fence *f; drm_sched_job_arm(&job->base); - - *f = dma_fence_get(&job->base.s_fence->finished); + f = dma_fence_get(&job->base.s_fence->finished); amdgpu_job_free_resources(job); drm_sched_entity_push_job(&job->base); - return 0; + return f; } int amdgpu_job_submit_direct(struct amdgpu_job *job, struct amdgpu_ring *ring, @@ -242,33 +241,22 @@ int amdgpu_job_submit_direct(struct amdgpu_job *job, struct amdgpu_ring *ring, return 0; } -static struct dma_fence *amdgpu_job_dependency(struct drm_sched_job *sched_job, - struct drm_sched_entity *s_entity) +static struct dma_fence * +amdgpu_job_prepare_job(struct drm_sched_job *sched_job, + struct drm_sched_entity *s_entity) { struct amdgpu_ring *ring = to_amdgpu_ring(s_entity->rq->sched); struct amdgpu_job *job = to_amdgpu_job(sched_job); - struct amdgpu_vm *vm = job->vm; - struct dma_fence *fence; + struct dma_fence *fence = NULL; int r; - fence = amdgpu_sync_get_fence(&job->sync); - if (fence && drm_sched_dependency_optimized(fence, s_entity)) { - r = amdgpu_sync_fence(&job->sched_sync, fence); - if (r) - DRM_ERROR("Error adding fence (%d)\n", r); - } - if (!fence && job->gang_submit) fence = amdgpu_device_switch_gang(ring->adev, job->gang_submit); - while (fence == NULL && vm && !job->vmid) { - r = amdgpu_vmid_grab(vm, ring, &job->sync, - &job->base.s_fence->finished, - job); + while (!fence && job->vm && !job->vmid) { + r = amdgpu_vmid_grab(job->vm, ring, job, &fence); if (r) DRM_ERROR("Error getting VM ID (%d)\n", r); - - fence = amdgpu_sync_get_fence(&job->sync); } return fence; @@ -285,8 +273,6 @@ static struct dma_fence *amdgpu_job_run(struct drm_sched_job *sched_job) job = to_amdgpu_job(sched_job); finished = &job->base.s_fence->finished; - BUG_ON(amdgpu_sync_peek_fence(&job->sync, NULL)); - trace_amdgpu_sched_run_job(job); /* Skip job if VRAM is lost and never resubmit gangs */ @@ -345,7 +331,7 @@ void amdgpu_job_stop_all_jobs_on_sched(struct drm_gpu_scheduler *sched) } const struct drm_sched_backend_ops amdgpu_sched_ops = { - .dependency = amdgpu_job_dependency, + .prepare_job = amdgpu_job_prepare_job, .run_job = amdgpu_job_run, .timedout_job = amdgpu_job_timedout, .free_job = amdgpu_job_free_cb diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_job.h b/drivers/gpu/drm/amd/amdgpu/amdgpu_job.h index ab7b150e5d50..52f2e313ea17 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_job.h +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_job.h @@ -47,13 +47,14 @@ enum amdgpu_ib_pool_type; struct amdgpu_job { struct drm_sched_job base; struct amdgpu_vm *vm; - struct amdgpu_sync sync; - struct amdgpu_sync sched_sync; + struct amdgpu_sync explicit_sync; struct dma_fence hw_fence; struct dma_fence *gang_submit; uint32_t preamble_status; uint32_t preemption_status; bool vm_needs_flush; + bool gds_switch_needed; + bool spm_update_needed; uint64_t vm_pd_addr; unsigned vmid; unsigned pasid; @@ -78,18 +79,20 @@ static inline struct amdgpu_ring *amdgpu_job_ring(struct amdgpu_job *job) return to_amdgpu_ring(job->base.entity->rq->sched); } -int amdgpu_job_alloc(struct amdgpu_device *adev, unsigned num_ibs, - struct amdgpu_job **job, struct amdgpu_vm *vm); -int amdgpu_job_alloc_with_ib(struct amdgpu_device *adev, unsigned size, - enum amdgpu_ib_pool_type pool, struct amdgpu_job **job); +int amdgpu_job_alloc(struct amdgpu_device *adev, struct amdgpu_vm *vm, + struct drm_sched_entity *entity, void *owner, + unsigned int num_ibs, struct amdgpu_job **job); +int amdgpu_job_alloc_with_ib(struct amdgpu_device *adev, + struct drm_sched_entity *entity, void *owner, + size_t size, enum amdgpu_ib_pool_type pool_type, + struct amdgpu_job **job); void amdgpu_job_set_resources(struct amdgpu_job *job, struct amdgpu_bo *gds, struct amdgpu_bo *gws, struct amdgpu_bo *oa); void amdgpu_job_free_resources(struct amdgpu_job *job); void amdgpu_job_set_gang_leader(struct amdgpu_job *job, struct amdgpu_job *leader); void amdgpu_job_free(struct amdgpu_job *job); -int amdgpu_job_submit(struct amdgpu_job *job, struct drm_sched_entity *entity, - void *owner, struct dma_fence **f); +struct dma_fence *amdgpu_job_submit(struct amdgpu_job *job); int amdgpu_job_submit_direct(struct amdgpu_job *job, struct amdgpu_ring *ring, struct dma_fence **fence); diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_jpeg.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_jpeg.c index 518eb0e40d32..6f81ed4fb0d9 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_jpeg.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_jpeg.c @@ -150,14 +150,15 @@ static int amdgpu_jpeg_dec_set_reg(struct amdgpu_ring *ring, uint32_t handle, const unsigned ib_size_dw = 16; int i, r; - r = amdgpu_job_alloc_with_ib(ring->adev, ib_size_dw * 4, - AMDGPU_IB_POOL_DIRECT, &job); + r = amdgpu_job_alloc_with_ib(ring->adev, NULL, NULL, ib_size_dw * 4, + AMDGPU_IB_POOL_DIRECT, &job); if (r) return r; ib = &job->ibs[0]; - ib->ptr[0] = PACKETJ(adev->jpeg.internal.jpeg_pitch, 0, 0, PACKETJ_TYPE0); + ib->ptr[0] = PACKETJ(adev->jpeg.internal.jpeg_pitch, 0, 0, + PACKETJ_TYPE0); ib->ptr[1] = 0xDEADBEEF; for (i = 2; i < 16; i += 2) { ib->ptr[i] = PACKETJ(0, 0, 0, PACKETJ_TYPE6); @@ -234,3 +235,20 @@ int amdgpu_jpeg_process_poison_irq(struct amdgpu_device *adev, return 0; } + +void jpeg_set_ras_funcs(struct amdgpu_device *adev) +{ + if (!adev->jpeg.ras) + return; + + amdgpu_ras_register_ras_block(adev, &adev->jpeg.ras->ras_block); + + strcpy(adev->jpeg.ras->ras_block.ras_comm.name, "jpeg"); + adev->jpeg.ras->ras_block.ras_comm.block = AMDGPU_RAS_BLOCK__JPEG; + adev->jpeg.ras->ras_block.ras_comm.type = AMDGPU_RAS_ERROR__POISON; + adev->jpeg.ras_if = &adev->jpeg.ras->ras_block.ras_comm; + + /* If don't define special ras_late_init function, use default ras_late_init */ + if (!adev->jpeg.ras->ras_block.ras_late_init) + adev->jpeg.ras->ras_block.ras_late_init = amdgpu_ras_block_late_init; +} diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_jpeg.h b/drivers/gpu/drm/amd/amdgpu/amdgpu_jpeg.h index 635dca59a70a..e8ca3e32ad52 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_jpeg.h +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_jpeg.h @@ -72,5 +72,6 @@ int amdgpu_jpeg_dec_ring_test_ib(struct amdgpu_ring *ring, long timeout); int amdgpu_jpeg_process_poison_irq(struct amdgpu_device *adev, struct amdgpu_irq_src *source, struct amdgpu_iv_entry *entry); +void jpeg_set_ras_funcs(struct amdgpu_device *adev); #endif /*__AMDGPU_JPEG_H__*/ diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_kms.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_kms.c index 4e42dcb1950f..7aa7e52ca784 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_kms.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_kms.c @@ -29,6 +29,7 @@ #include "amdgpu.h" #include <drm/amdgpu_drm.h> #include <drm/drm_drv.h> +#include <drm/drm_fb_helper.h> #include "amdgpu_uvd.h" #include "amdgpu_vce.h" #include "atom.h" @@ -430,7 +431,7 @@ static int amdgpu_hw_ip_info(struct amdgpu_device *adev, case AMDGPU_HW_IP_VCN_DEC: type = AMD_IP_BLOCK_TYPE_VCN; for (i = 0; i < adev->vcn.num_vcn_inst; i++) { - if (adev->uvd.harvest_config & (1 << i)) + if (adev->vcn.harvest_config & (1 << i)) continue; if (adev->vcn.inst[i].ring_dec.sched.ready) @@ -442,7 +443,7 @@ static int amdgpu_hw_ip_info(struct amdgpu_device *adev, case AMDGPU_HW_IP_VCN_ENC: type = AMD_IP_BLOCK_TYPE_VCN; for (i = 0; i < adev->vcn.num_vcn_inst; i++) { - if (adev->uvd.harvest_config & (1 << i)) + if (adev->vcn.harvest_config & (1 << i)) continue; for (j = 0; j < adev->vcn.num_enc_rings; j++) @@ -796,7 +797,7 @@ int amdgpu_info_ioctl(struct drm_device *dev, void *data, struct drm_file *filp) dev_info->ids_flags = 0; if (adev->flags & AMD_IS_APU) dev_info->ids_flags |= AMDGPU_IDS_FLAGS_FUSION; - if (amdgpu_mcbp || amdgpu_sriov_vf(adev)) + if (amdgpu_mcbp) dev_info->ids_flags |= AMDGPU_IDS_FLAGS_PREEMPTION; if (amdgpu_is_tmz(adev)) dev_info->ids_flags |= AMDGPU_IDS_FLAGS_TMZ; @@ -1172,7 +1173,7 @@ int amdgpu_driver_open_kms(struct drm_device *dev, struct drm_file *file_priv) goto error_vm; } - if (amdgpu_mcbp || amdgpu_sriov_vf(adev)) { + if (amdgpu_mcbp) { uint64_t csa_addr = amdgpu_csa_vaddr(adev) & AMDGPU_GMC_HOLE_MASK; r = amdgpu_map_static_csa(adev, &fpriv->vm, adev->virt.csa_obj, @@ -1236,7 +1237,7 @@ void amdgpu_driver_postclose_kms(struct drm_device *dev, if (amdgpu_device_ip_get_ip_block(adev, AMD_IP_BLOCK_TYPE_VCE) != NULL) amdgpu_vce_free_handles(adev, file_priv); - if (amdgpu_mcbp || amdgpu_sriov_vf(adev)) { + if (amdgpu_mcbp) { /* TODO: how to handle reserve failure */ BUG_ON(amdgpu_bo_reserve(adev->virt.csa_obj, true)); amdgpu_vm_bo_del(adev, fpriv->csa_va); diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_mes.h b/drivers/gpu/drm/amd/amdgpu/amdgpu_mes.h index ad980f4b66e1..97c05d08a551 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_mes.h +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_mes.h @@ -91,14 +91,12 @@ struct amdgpu_mes { struct amdgpu_bo *ucode_fw_obj[AMDGPU_MAX_MES_PIPES]; uint64_t ucode_fw_gpu_addr[AMDGPU_MAX_MES_PIPES]; uint32_t *ucode_fw_ptr[AMDGPU_MAX_MES_PIPES]; - uint32_t ucode_fw_version[AMDGPU_MAX_MES_PIPES]; uint64_t uc_start_addr[AMDGPU_MAX_MES_PIPES]; /* mes ucode data */ struct amdgpu_bo *data_fw_obj[AMDGPU_MAX_MES_PIPES]; uint64_t data_fw_gpu_addr[AMDGPU_MAX_MES_PIPES]; uint32_t *data_fw_ptr[AMDGPU_MAX_MES_PIPES]; - uint32_t data_fw_version[AMDGPU_MAX_MES_PIPES]; uint64_t data_start_addr[AMDGPU_MAX_MES_PIPES]; /* eop gpu obj */ diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_mode.h b/drivers/gpu/drm/amd/amdgpu/amdgpu_mode.h index 37322550d750..8a39300b1a84 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_mode.h +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_mode.h @@ -36,7 +36,6 @@ #include <drm/drm_encoder.h> #include <drm/drm_fixed.h> #include <drm/drm_crtc_helper.h> -#include <drm/drm_fb_helper.h> #include <drm/drm_framebuffer.h> #include <drm/drm_probe_helper.h> #include <linux/i2c.h> diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_object.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_object.c index 2e8f6cd7a729..25a68d8888e0 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_object.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_object.c @@ -346,17 +346,16 @@ int amdgpu_bo_create_kernel(struct amdgpu_device *adev, * @adev: amdgpu device object * @offset: offset of the BO * @size: size of the BO - * @domain: where to place it * @bo_ptr: used to initialize BOs in structures * @cpu_addr: optional CPU address mapping * - * Creates a kernel BO at a specific offset in the address space of the domain. + * Creates a kernel BO at a specific offset in VRAM. * * Returns: * 0 on success, negative error code otherwise. */ int amdgpu_bo_create_kernel_at(struct amdgpu_device *adev, - uint64_t offset, uint64_t size, uint32_t domain, + uint64_t offset, uint64_t size, struct amdgpu_bo **bo_ptr, void **cpu_addr) { struct ttm_operation_ctx ctx = { false, false }; @@ -366,8 +365,9 @@ int amdgpu_bo_create_kernel_at(struct amdgpu_device *adev, offset &= PAGE_MASK; size = ALIGN(size, PAGE_SIZE); - r = amdgpu_bo_create_reserved(adev, size, PAGE_SIZE, domain, bo_ptr, - NULL, cpu_addr); + r = amdgpu_bo_create_reserved(adev, size, PAGE_SIZE, + AMDGPU_GEM_DOMAIN_VRAM, bo_ptr, NULL, + cpu_addr); if (r) return r; @@ -422,6 +422,8 @@ void amdgpu_bo_free_kernel(struct amdgpu_bo **bo, u64 *gpu_addr, if (*bo == NULL) return; + WARN_ON(amdgpu_ttm_adev((*bo)->tbo.bdev)->in_suspend); + if (likely(amdgpu_bo_reserve(*bo, true) == 0)) { if (cpu_addr) amdgpu_bo_kunmap(*bo); @@ -446,33 +448,31 @@ static bool amdgpu_bo_validate_size(struct amdgpu_device *adev, /* * If GTT is part of requested domains the check must succeed to - * allow fall back to GTT + * allow fall back to GTT. */ if (domain & AMDGPU_GEM_DOMAIN_GTT) { man = ttm_manager_type(&adev->mman.bdev, TTM_PL_TT); - if (size < man->size) + if (man && size < man->size) return true; - else - goto fail; - } - - if (domain & AMDGPU_GEM_DOMAIN_VRAM) { + else if (!man) + WARN_ON_ONCE("GTT domain requested but GTT mem manager uninitialized"); + goto fail; + } else if (domain & AMDGPU_GEM_DOMAIN_VRAM) { man = ttm_manager_type(&adev->mman.bdev, TTM_PL_VRAM); - if (size < man->size) + if (man && size < man->size) return true; - else - goto fail; + goto fail; } - /* TODO add more domains checks, such as AMDGPU_GEM_DOMAIN_CPU */ return true; fail: - DRM_DEBUG("BO size %lu > total memory in domain: %llu\n", size, - man->size); + if (man) + DRM_DEBUG("BO size %lu > total memory in domain: %llu\n", size, + man->size); return false; } @@ -542,6 +542,7 @@ int amdgpu_bo_create(struct amdgpu_device *adev, /* GWS and OA don't need any alignment. */ page_align = bp->byte_align; size <<= PAGE_SHIFT; + } else if (bp->domain & AMDGPU_GEM_DOMAIN_GDS) { /* Both size and alignment must be a multiple of 4. */ page_align = ALIGN(bp->byte_align, 4); @@ -776,7 +777,7 @@ int amdgpu_bo_kmap(struct amdgpu_bo *bo, void **ptr) return 0; } - r = ttm_bo_kmap(&bo->tbo, 0, bo->tbo.resource->num_pages, &bo->kmap); + r = ttm_bo_kmap(&bo->tbo, 0, PFN_UP(bo->tbo.base.size), &bo->kmap); if (r) return r; @@ -1509,7 +1510,8 @@ u64 amdgpu_bo_gpu_offset_no_check(struct amdgpu_bo *bo) uint32_t amdgpu_bo_get_preferred_domain(struct amdgpu_device *adev, uint32_t domain) { - if (domain == (AMDGPU_GEM_DOMAIN_VRAM | AMDGPU_GEM_DOMAIN_GTT)) { + if ((domain == (AMDGPU_GEM_DOMAIN_VRAM | AMDGPU_GEM_DOMAIN_GTT)) && + ((adev->asic_type == CHIP_CARRIZO) || (adev->asic_type == CHIP_STONEY))) { domain = AMDGPU_GEM_DOMAIN_VRAM; if (adev->gmc.real_vram_size <= AMDGPU_SG_THRESHOLD) domain = AMDGPU_GEM_DOMAIN_GTT; diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_object.h b/drivers/gpu/drm/amd/amdgpu/amdgpu_object.h index 147b79c10cbb..93207badf83f 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_object.h +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_object.h @@ -284,7 +284,7 @@ int amdgpu_bo_create_kernel(struct amdgpu_device *adev, u32 domain, struct amdgpu_bo **bo_ptr, u64 *gpu_addr, void **cpu_addr); int amdgpu_bo_create_kernel_at(struct amdgpu_device *adev, - uint64_t offset, uint64_t size, uint32_t domain, + uint64_t offset, uint64_t size, struct amdgpu_bo **bo_ptr, void **cpu_addr); int amdgpu_bo_create_user(struct amdgpu_device *adev, struct amdgpu_bo_param *bp, diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_psp.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_psp.c index 7978307e1d6d..7a2fc920739b 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_psp.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_psp.c @@ -52,6 +52,32 @@ static int psp_load_smu_fw(struct psp_context *psp); static int psp_rap_terminate(struct psp_context *psp); static int psp_securedisplay_terminate(struct psp_context *psp); +static int psp_ring_init(struct psp_context *psp, + enum psp_ring_type ring_type) +{ + int ret = 0; + struct psp_ring *ring; + struct amdgpu_device *adev = psp->adev; + + ring = &psp->km_ring; + + ring->ring_type = ring_type; + + /* allocate 4k Page of Local Frame Buffer memory for ring */ + ring->ring_size = 0x1000; + ret = amdgpu_bo_create_kernel(adev, ring->ring_size, PAGE_SIZE, + AMDGPU_GEM_DOMAIN_VRAM, + &adev->firmware.rbuf, + &ring->ring_mem_mc_addr, + (void **)&ring->ring_mem); + if (ret) { + ring->ring_size = 0; + return ret; + } + + return 0; +} + /* * Due to DF Cstate management centralized to PMFW, the firmware * loading sequence will be updated as below: @@ -139,6 +165,7 @@ static int psp_early_init(void *handle) case IP_VERSION(13, 0, 5): case IP_VERSION(13, 0, 8): case IP_VERSION(13, 0, 10): + case IP_VERSION(13, 0, 11): psp_v13_0_set_psp_funcs(psp); psp->autoload_supported = true; break; @@ -486,26 +513,22 @@ static int psp_sw_fini(void *handle) struct psp_gfx_cmd_resp *cmd = psp->cmd; psp_memory_training_fini(psp); - if (psp->sos_fw) { - release_firmware(psp->sos_fw); - psp->sos_fw = NULL; - } - if (psp->asd_fw) { - release_firmware(psp->asd_fw); - psp->asd_fw = NULL; - } - if (psp->ta_fw) { - release_firmware(psp->ta_fw); - psp->ta_fw = NULL; - } - if (psp->cap_fw) { - release_firmware(psp->cap_fw); - psp->cap_fw = NULL; - } - if (psp->toc_fw) { - release_firmware(psp->toc_fw); - psp->toc_fw = NULL; - } + + release_firmware(psp->sos_fw); + psp->sos_fw = NULL; + + release_firmware(psp->asd_fw); + psp->asd_fw = NULL; + + release_firmware(psp->ta_fw); + psp->ta_fw = NULL; + + release_firmware(psp->cap_fw); + psp->cap_fw = NULL; + + release_firmware(psp->toc_fw); + psp->toc_fw = NULL; + if (adev->ip_versions[MP0_HWIP][0] == IP_VERSION(11, 0, 0) || adev->ip_versions[MP0_HWIP][0] == IP_VERSION(11, 0, 7)) psp_sysfs_fini(adev); @@ -835,7 +858,7 @@ static int psp_tmr_unload(struct psp_context *psp) struct psp_gfx_cmd_resp *cmd = acquire_psp_cmd_buf(psp); psp_prep_tmr_unload_cmd_buf(psp, cmd); - dev_info(psp->adev->dev, "free PSP TMR buffer\n"); + dev_dbg(psp->adev->dev, "free PSP TMR buffer\n"); ret = psp_cmd_submit_buf(psp, NULL, cmd, psp->fence_buf_mc_addr); @@ -992,6 +1015,8 @@ int psp_ta_unload(struct psp_context *psp, struct ta_context *context) ret = psp_cmd_submit_buf(psp, NULL, cmd, psp->fence_buf_mc_addr); + context->resp_status = cmd->resp.status; + release_psp_cmd_buf(psp); return ret; @@ -1073,42 +1098,6 @@ int psp_ta_init_shared_buf(struct psp_context *psp, &mem_ctx->shared_buf); } -static void psp_prep_ta_invoke_indirect_cmd_buf(struct psp_gfx_cmd_resp *cmd, - uint32_t ta_cmd_id, - struct ta_context *context) -{ - cmd->cmd_id = GFX_CMD_ID_INVOKE_CMD; - cmd->cmd.cmd_invoke_cmd.session_id = context->session_id; - cmd->cmd.cmd_invoke_cmd.ta_cmd_id = ta_cmd_id; - - cmd->cmd.cmd_invoke_cmd.buf.num_desc = 1; - cmd->cmd.cmd_invoke_cmd.buf.total_size = context->mem_context.shared_mem_size; - cmd->cmd.cmd_invoke_cmd.buf.buf_desc[0].buf_size = context->mem_context.shared_mem_size; - cmd->cmd.cmd_invoke_cmd.buf.buf_desc[0].buf_phy_addr_lo = - lower_32_bits(context->mem_context.shared_mc_addr); - cmd->cmd.cmd_invoke_cmd.buf.buf_desc[0].buf_phy_addr_hi = - upper_32_bits(context->mem_context.shared_mc_addr); -} - -int psp_ta_invoke_indirect(struct psp_context *psp, - uint32_t ta_cmd_id, - struct ta_context *context) -{ - int ret; - struct psp_gfx_cmd_resp *cmd = acquire_psp_cmd_buf(psp); - - psp_prep_ta_invoke_indirect_cmd_buf(cmd, ta_cmd_id, context); - - ret = psp_cmd_submit_buf(psp, NULL, cmd, - psp->fence_buf_mc_addr); - - context->resp_status = cmd->resp.status; - - release_psp_cmd_buf(psp); - - return ret; -} - static void psp_prep_ta_invoke_cmd_buf(struct psp_gfx_cmd_resp *cmd, uint32_t ta_cmd_id, uint32_t session_id) @@ -1551,7 +1540,7 @@ int psp_ras_terminate(struct psp_context *psp) return ret; } -static int psp_ras_initialize(struct psp_context *psp) +int psp_ras_initialize(struct psp_context *psp) { int ret; uint32_t boot_cfg = 0xFF; @@ -1614,7 +1603,7 @@ static int psp_ras_initialize(struct psp_context *psp) psp->ras_context.context.mem_context.shared_mem_size = PSP_RAS_SHARED_MEM_SIZE; psp->ras_context.context.ta_load_type = GFX_CMD_ID_LOAD_TA; - if (!psp->ras_context.context.initialized) { + if (!psp->ras_context.context.mem_context.shared_buf) { ret = psp_ta_init_shared_buf(psp, &psp->ras_context.context.mem_context); if (ret) return ret; @@ -1635,7 +1624,9 @@ static int psp_ras_initialize(struct psp_context *psp) else { if (ras_cmd->ras_status) dev_warn(psp->adev->dev, "RAS Init Status: 0x%X\n", ras_cmd->ras_status); - amdgpu_ras_fini(psp->adev); + + /* fail to load RAS TA */ + psp->ras_context.context.initialized = false; } return ret; @@ -1942,10 +1933,15 @@ static int psp_securedisplay_initialize(struct psp_context *psp) } else return ret; + mutex_lock(&psp->securedisplay_context.mutex); + psp_prep_securedisplay_cmd_buf(psp, &securedisplay_cmd, TA_SECUREDISPLAY_COMMAND__QUERY_TA); ret = psp_securedisplay_invoke(psp, TA_SECUREDISPLAY_COMMAND__QUERY_TA); + + mutex_unlock(&psp->securedisplay_context.mutex); + if (ret) { psp_securedisplay_terminate(psp); /* free securedisplay shared memory */ @@ -1994,12 +1990,8 @@ int psp_securedisplay_invoke(struct psp_context *psp, uint32_t ta_cmd_id) ta_cmd_id != TA_SECUREDISPLAY_COMMAND__SEND_ROI_CRC) return -EINVAL; - mutex_lock(&psp->securedisplay_context.mutex); - ret = psp_ta_invoke(psp, ta_cmd_id, &psp->securedisplay_context.context); - mutex_unlock(&psp->securedisplay_context.mutex); - return ret; } /* SECUREDISPLAY end */ diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_psp.h b/drivers/gpu/drm/amd/amdgpu/amdgpu_psp.h index 58ce3ebb446c..cf4f60c66122 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_psp.h +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_psp.h @@ -118,7 +118,6 @@ struct psp_funcs int (*bootloader_load_dbg_drv)(struct psp_context *psp); int (*bootloader_load_ras_drv)(struct psp_context *psp); int (*bootloader_load_sos)(struct psp_context *psp); - int (*ring_init)(struct psp_context *psp, enum psp_ring_type ring_type); int (*ring_create)(struct psp_context *psp, enum psp_ring_type ring_type); int (*ring_stop)(struct psp_context *psp, @@ -136,6 +135,12 @@ struct psp_funcs int (*vbflash_stat)(struct psp_context *psp); }; +struct ta_funcs { + int (*fn_ta_initialize)(struct psp_context *psp); + int (*fn_ta_invoke)(struct psp_context *psp, uint32_t ta_cmd_id); + int (*fn_ta_terminate)(struct psp_context *psp); +}; + #define AMDGPU_XGMI_MAX_CONNECTED_NODES 64 struct psp_xgmi_node_info { uint64_t node_id; @@ -309,6 +314,7 @@ struct psp_context struct psp_gfx_cmd_resp *cmd; const struct psp_funcs *funcs; + const struct ta_funcs *ta_funcs; /* firmware buffer */ struct amdgpu_bo *fw_pri_bo; @@ -389,7 +395,6 @@ struct amdgpu_psp_funcs { }; -#define psp_ring_init(psp, type) (psp)->funcs->ring_init((psp), (type)) #define psp_ring_create(psp, type) (psp)->funcs->ring_create((psp), (type)) #define psp_ring_stop(psp, type) (psp)->funcs->ring_stop((psp), (type)) #define psp_ring_destroy(psp, type) ((psp)->funcs->ring_destroy((psp), (type))) @@ -463,9 +468,6 @@ int psp_ta_load(struct psp_context *psp, struct ta_context *context); int psp_ta_invoke(struct psp_context *psp, uint32_t ta_cmd_id, struct ta_context *context); -int psp_ta_invoke_indirect(struct psp_context *psp, - uint32_t ta_cmd_id, - struct ta_context *context); int psp_xgmi_initialize(struct psp_context *psp, bool set_extended_data, bool load_ta); int psp_xgmi_terminate(struct psp_context *psp); @@ -479,7 +481,7 @@ int psp_xgmi_get_topology_info(struct psp_context *psp, int psp_xgmi_set_topology_info(struct psp_context *psp, int number_devices, struct psp_xgmi_topology_info *topology); - +int psp_ras_initialize(struct psp_context *psp); int psp_ras_invoke(struct psp_context *psp, uint32_t ta_cmd_id); int psp_ras_enable_features(struct psp_context *psp, union ta_ras_cmd_input *info, bool enable); diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_psp_ta.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_psp_ta.c index 0988e00612e5..468a67b302d4 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_psp_ta.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_psp_ta.c @@ -41,30 +41,46 @@ static uint32_t get_bin_version(const uint8_t *bin) return hdr->ucode_version; } -static void prep_ta_mem_context(struct psp_context *psp, - struct ta_context *context, +static int prep_ta_mem_context(struct ta_mem_context *mem_context, uint8_t *shared_buf, uint32_t shared_buf_len) { - context->mem_context.shared_mem_size = PAGE_ALIGN(shared_buf_len); - psp_ta_init_shared_buf(psp, &context->mem_context); + if (mem_context->shared_mem_size < shared_buf_len) + return -EINVAL; + memset(mem_context->shared_buf, 0, mem_context->shared_mem_size); + memcpy((void *)mem_context->shared_buf, shared_buf, shared_buf_len); - memcpy((void *)context->mem_context.shared_buf, shared_buf, shared_buf_len); + return 0; } static bool is_ta_type_valid(enum ta_type_id ta_type) { - bool ret = false; + switch (ta_type) { + case TA_TYPE_RAS: + return true; + default: + return false; + } +} + +static const struct ta_funcs ras_ta_funcs = { + .fn_ta_initialize = psp_ras_initialize, + .fn_ta_invoke = psp_ras_invoke, + .fn_ta_terminate = psp_ras_terminate +}; +static void set_ta_context_funcs(struct psp_context *psp, + enum ta_type_id ta_type, + struct ta_context **pcontext) +{ switch (ta_type) { case TA_TYPE_RAS: - ret = true; + *pcontext = &psp->ras_context.context; + psp->ta_funcs = &ras_ta_funcs; break; default: break; } - - return ret; } static const struct file_operations ta_load_debugfs_fops = { @@ -85,8 +101,7 @@ static const struct file_operations ta_invoke_debugfs_fops = { .owner = THIS_MODULE }; - -/** +/* * DOC: AMDGPU TA debugfs interfaces * * Three debugfs interfaces can be opened by a program to @@ -111,15 +126,18 @@ static const struct file_operations ta_invoke_debugfs_fops = { * * - For TA invoke debugfs interface: * Transmit buffer: + * - TA type (4bytes) * - TA ID (4bytes) * - TA CMD ID (4bytes) - * - TA shard buf length (4bytes) + * - TA shard buf length + * (4bytes, value not beyond TA shared memory size) * - TA shared buf * Receive buffer: * - TA shared buf * * - For TA unload debugfs interface: * Transmit buffer: + * - TA type (4bytes) * - TA ID (4bytes) */ @@ -131,59 +149,92 @@ static ssize_t ta_if_load_debugfs_write(struct file *fp, const char *buf, size_t uint32_t copy_pos = 0; int ret = 0; - struct amdgpu_device *adev = (struct amdgpu_device *)file_inode(fp)->i_private; - struct psp_context *psp = &adev->psp; - struct ta_context context = {0}; + struct amdgpu_device *adev = (struct amdgpu_device *)file_inode(fp)->i_private; + struct psp_context *psp = &adev->psp; + struct ta_context *context = NULL; if (!buf) return -EINVAL; ret = copy_from_user((void *)&ta_type, &buf[copy_pos], sizeof(uint32_t)); if (ret || (!is_ta_type_valid(ta_type))) - return -EINVAL; + return -EFAULT; copy_pos += sizeof(uint32_t); ret = copy_from_user((void *)&ta_bin_len, &buf[copy_pos], sizeof(uint32_t)); if (ret) - return -EINVAL; + return -EFAULT; copy_pos += sizeof(uint32_t); ta_bin = kzalloc(ta_bin_len, GFP_KERNEL); if (!ta_bin) - ret = -ENOMEM; + return -ENOMEM; if (copy_from_user((void *)ta_bin, &buf[copy_pos], ta_bin_len)) { ret = -EFAULT; goto err_free_bin; } - ret = psp_ras_terminate(psp); - if (ret) { - dev_err(adev->dev, "Failed to unload embedded RAS TA\n"); + /* Set TA context and functions */ + set_ta_context_funcs(psp, ta_type, &context); + + if (!psp->ta_funcs || !psp->ta_funcs->fn_ta_terminate) { + dev_err(adev->dev, "Unsupported function to terminate TA\n"); + ret = -EOPNOTSUPP; goto err_free_bin; } - context.ta_type = ta_type; - context.ta_load_type = GFX_CMD_ID_LOAD_TA; - context.bin_desc.fw_version = get_bin_version(ta_bin); - context.bin_desc.size_bytes = ta_bin_len; - context.bin_desc.start_addr = ta_bin; + /* + * Allocate TA shared buf in case shared buf was freed + * due to loading TA failed before. + */ + if (!context->mem_context.shared_buf) { + ret = psp_ta_init_shared_buf(psp, &context->mem_context); + if (ret) { + ret = -ENOMEM; + goto err_free_bin; + } + } + + ret = psp_fn_ta_terminate(psp); + if (ret || context->resp_status) { + dev_err(adev->dev, + "Failed to unload embedded TA (%d) and status (0x%X)\n", + ret, context->resp_status); + if (!ret) + ret = -EINVAL; + goto err_free_ta_shared_buf; + } + + /* Prepare TA context for TA initialization */ + context->ta_type = ta_type; + context->bin_desc.fw_version = get_bin_version(ta_bin); + context->bin_desc.size_bytes = ta_bin_len; + context->bin_desc.start_addr = ta_bin; - ret = psp_ta_load(psp, &context); + if (!psp->ta_funcs->fn_ta_initialize) { + dev_err(adev->dev, "Unsupported function to initialize TA\n"); + ret = -EOPNOTSUPP; + goto err_free_ta_shared_buf; + } - if (ret || context.resp_status) { - dev_err(adev->dev, "TA load via debugfs failed (%d) status %d\n", - ret, context.resp_status); + ret = psp_fn_ta_initialize(psp); + if (ret || context->resp_status) { + dev_err(adev->dev, "Failed to load TA via debugfs (%d) and status (0x%X)\n", + ret, context->resp_status); if (!ret) ret = -EINVAL; - goto err_free_bin; + goto err_free_ta_shared_buf; } - context.initialized = true; - if (copy_to_user((char *)buf, (void *)&context.session_id, sizeof(uint32_t))) + if (copy_to_user((char *)buf, (void *)&context->session_id, sizeof(uint32_t))) ret = -EFAULT; +err_free_ta_shared_buf: + /* Only free TA shared buf when returns error code */ + if (ret && context->mem_context.shared_buf) + psp_ta_free_shared_buf(&context->mem_context); err_free_bin: kfree(ta_bin); @@ -192,58 +243,85 @@ err_free_bin: static ssize_t ta_if_unload_debugfs_write(struct file *fp, const char *buf, size_t len, loff_t *off) { - uint32_t ta_id = 0; - int ret = 0; + uint32_t ta_type = 0; + uint32_t ta_id = 0; + uint32_t copy_pos = 0; + int ret = 0; - struct amdgpu_device *adev = (struct amdgpu_device *)file_inode(fp)->i_private; - struct psp_context *psp = &adev->psp; - struct ta_context context = {0}; + struct amdgpu_device *adev = (struct amdgpu_device *)file_inode(fp)->i_private; + struct psp_context *psp = &adev->psp; + struct ta_context *context = NULL; if (!buf) return -EINVAL; - ret = copy_from_user((void *)&ta_id, buf, sizeof(uint32_t)); + ret = copy_from_user((void *)&ta_type, &buf[copy_pos], sizeof(uint32_t)); + if (ret || (!is_ta_type_valid(ta_type))) + return -EFAULT; + + copy_pos += sizeof(uint32_t); + + ret = copy_from_user((void *)&ta_id, &buf[copy_pos], sizeof(uint32_t)); if (ret) - return -EINVAL; + return -EFAULT; - context.session_id = ta_id; + set_ta_context_funcs(psp, ta_type, &context); + context->session_id = ta_id; - ret = psp_ta_unload(psp, &context); - if (!ret) - context.initialized = false; + if (!psp->ta_funcs || !psp->ta_funcs->fn_ta_terminate) { + dev_err(adev->dev, "Unsupported function to terminate TA\n"); + return -EOPNOTSUPP; + } + + ret = psp_fn_ta_terminate(psp); + if (ret || context->resp_status) { + dev_err(adev->dev, "Failed to unload TA via debugfs (%d) and status (0x%X)\n", + ret, context->resp_status); + if (!ret) + ret = -EINVAL; + } + + if (context->mem_context.shared_buf) + psp_ta_free_shared_buf(&context->mem_context); return ret; } static ssize_t ta_if_invoke_debugfs_write(struct file *fp, const char *buf, size_t len, loff_t *off) { + uint32_t ta_type = 0; uint32_t ta_id = 0; uint32_t cmd_id = 0; uint32_t shared_buf_len = 0; - uint8_t *shared_buf = NULL; + uint8_t *shared_buf = NULL; uint32_t copy_pos = 0; int ret = 0; - struct amdgpu_device *adev = (struct amdgpu_device *)file_inode(fp)->i_private; - struct psp_context *psp = &adev->psp; - struct ta_context context = {0}; + struct amdgpu_device *adev = (struct amdgpu_device *)file_inode(fp)->i_private; + struct psp_context *psp = &adev->psp; + struct ta_context *context = NULL; if (!buf) return -EINVAL; + ret = copy_from_user((void *)&ta_type, &buf[copy_pos], sizeof(uint32_t)); + if (ret) + return -EFAULT; + copy_pos += sizeof(uint32_t); + ret = copy_from_user((void *)&ta_id, &buf[copy_pos], sizeof(uint32_t)); if (ret) - return -EINVAL; + return -EFAULT; copy_pos += sizeof(uint32_t); ret = copy_from_user((void *)&cmd_id, &buf[copy_pos], sizeof(uint32_t)); if (ret) - return -EINVAL; + return -EFAULT; copy_pos += sizeof(uint32_t); ret = copy_from_user((void *)&shared_buf_len, &buf[copy_pos], sizeof(uint32_t)); if (ret) - return -EINVAL; + return -EFAULT; copy_pos += sizeof(uint32_t); shared_buf = kzalloc(shared_buf_len, GFP_KERNEL); @@ -254,26 +332,39 @@ static ssize_t ta_if_invoke_debugfs_write(struct file *fp, const char *buf, size goto err_free_shared_buf; } - context.session_id = ta_id; + set_ta_context_funcs(psp, ta_type, &context); + + if (!context->initialized) { + dev_err(adev->dev, "TA is not initialized\n"); + ret = -EINVAL; + goto err_free_shared_buf; + } + + if (!psp->ta_funcs || !psp->ta_funcs->fn_ta_invoke) { + dev_err(adev->dev, "Unsupported function to invoke TA\n"); + ret = -EOPNOTSUPP; + goto err_free_shared_buf; + } - prep_ta_mem_context(psp, &context, shared_buf, shared_buf_len); + context->session_id = ta_id; - ret = psp_ta_invoke_indirect(psp, cmd_id, &context); + ret = prep_ta_mem_context(&context->mem_context, shared_buf, shared_buf_len); + if (ret) + goto err_free_shared_buf; - if (ret || context.resp_status) { - dev_err(adev->dev, "TA invoke via debugfs failed (%d) status %d\n", - ret, context.resp_status); - if (!ret) + ret = psp_fn_ta_invoke(psp, cmd_id); + if (ret || context->resp_status) { + dev_err(adev->dev, "Failed to invoke TA via debugfs (%d) and status (0x%X)\n", + ret, context->resp_status); + if (!ret) { ret = -EINVAL; - goto err_free_ta_shared_buf; + goto err_free_shared_buf; + } } - if (copy_to_user((char *)buf, context.mem_context.shared_buf, shared_buf_len)) + if (copy_to_user((char *)buf, context->mem_context.shared_buf, shared_buf_len)) ret = -EFAULT; -err_free_ta_shared_buf: - psp_ta_free_shared_buf(&context.mem_context); - err_free_shared_buf: kfree(shared_buf); diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_psp_ta.h b/drivers/gpu/drm/amd/amdgpu/amdgpu_psp_ta.h index cfc1542f63ef..14cd1c81c3e6 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_psp_ta.h +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_psp_ta.h @@ -24,6 +24,11 @@ #ifndef __AMDGPU_PSP_TA_H__ #define __AMDGPU_PSP_TA_H__ +/* Calling set_ta_context_funcs is required before using the following macros */ +#define psp_fn_ta_initialize(psp) ((psp)->ta_funcs->fn_ta_initialize((psp))) +#define psp_fn_ta_invoke(psp, ta_cmd_id) ((psp)->ta_funcs->fn_ta_invoke((psp), (ta_cmd_id))) +#define psp_fn_ta_terminate(psp) ((psp)->ta_funcs->fn_ta_terminate((psp))) + void amdgpu_ta_if_debugfs_init(struct amdgpu_device *adev); #endif diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_ras.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_ras.c index a4b47e1bd111..ad490c1e2f57 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_ras.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_ras.c @@ -1267,7 +1267,7 @@ static ssize_t amdgpu_ras_sysfs_features_read(struct device *dev, struct amdgpu_ras *con = container_of(attr, struct amdgpu_ras, features_attr); - return scnprintf(buf, PAGE_SIZE, "feature mask: 0x%x\n", con->features); + return sysfs_emit(buf, "feature mask: 0x%x\n", con->features); } static void amdgpu_ras_sysfs_remove_bad_page_node(struct amdgpu_device *adev) @@ -1561,7 +1561,6 @@ static void amdgpu_ras_interrupt_poison_consumption_handler(struct ras_manager * { bool poison_stat = false; struct amdgpu_device *adev = obj->adev; - struct ras_err_data err_data = {0, 0, 0, NULL}; struct amdgpu_ras_block_object *block_obj = amdgpu_ras_get_ras_block(adev, obj->head.block, 0); @@ -1584,7 +1583,7 @@ static void amdgpu_ras_interrupt_poison_consumption_handler(struct ras_manager * } if (!adev->gmc.xgmi.connected_to_cpu) - amdgpu_umc_poison_handler(adev, &err_data, false); + amdgpu_umc_poison_handler(adev, false); if (block_obj->hw_ops->handle_poison_consumption) poison_stat = block_obj->hw_ops->handle_poison_consumption(adev); @@ -1949,7 +1948,12 @@ static void amdgpu_ras_do_recovery(struct work_struct *work) reset_context.method = AMD_RESET_METHOD_NONE; reset_context.reset_req_dev = adev; - clear_bit(AMDGPU_NEED_FULL_RESET, &reset_context.flags); + + /* Perform full reset in fatal error mode */ + if (!amdgpu_ras_is_poison_mode_supported(ras->adev)) + set_bit(AMDGPU_NEED_FULL_RESET, &reset_context.flags); + else + clear_bit(AMDGPU_NEED_FULL_RESET, &reset_context.flags); amdgpu_device_gpu_recover(ras->adev, NULL, &reset_context); } @@ -2344,7 +2348,8 @@ static void amdgpu_ras_check_supported(struct amdgpu_device *adev) adev->ras_hw_enabled |= ~(1 << AMDGPU_RAS_BLOCK__UMC | 1 << AMDGPU_RAS_BLOCK__DF); - if (adev->ip_versions[VCN_HWIP][0] == IP_VERSION(2, 6, 0)) + if (adev->ip_versions[VCN_HWIP][0] == IP_VERSION(2, 6, 0) || + adev->ip_versions[VCN_HWIP][0] == IP_VERSION(4, 0, 0)) adev->ras_hw_enabled |= (1 << AMDGPU_RAS_BLOCK__VCN | 1 << AMDGPU_RAS_BLOCK__JPEG); else @@ -2848,7 +2853,6 @@ static int amdgpu_bad_page_notifier(struct notifier_block *nb, struct amdgpu_device *adev = NULL; uint32_t gpu_id = 0; uint32_t umc_inst = 0, ch_inst = 0; - struct ras_err_data err_data = {0, 0, 0, NULL}; /* * If the error was generated in UMC_V2, which belongs to GPU UMCs, @@ -2887,31 +2891,10 @@ static int amdgpu_bad_page_notifier(struct notifier_block *nb, dev_info(adev->dev, "Uncorrectable error detected in UMC inst: %d, chan_idx: %d", umc_inst, ch_inst); - err_data.err_addr = - kcalloc(adev->umc.max_ras_err_cnt_per_query, - sizeof(struct eeprom_table_record), GFP_KERNEL); - if (!err_data.err_addr) { - dev_warn(adev->dev, - "Failed to alloc memory for umc error record in mca notifier!\n"); + if (!amdgpu_umc_page_retirement_mca(adev, m->addr, ch_inst, umc_inst)) + return NOTIFY_OK; + else return NOTIFY_DONE; - } - - /* - * Translate UMC channel address to Physical address - */ - if (adev->umc.ras && - adev->umc.ras->convert_ras_error_address) - adev->umc.ras->convert_ras_error_address(adev, - &err_data, m->addr, ch_inst, umc_inst); - - if (amdgpu_bad_page_threshold != 0) { - amdgpu_ras_add_bad_pages(adev, err_data.err_addr, - err_data.err_addr_cnt); - amdgpu_ras_save_bad_pages(adev); - } - - kfree(err_data.err_addr); - return NOTIFY_OK; } static struct notifier_block amdgpu_bad_page_nb = { diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_ras_eeprom.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_ras_eeprom.c index 84c241b9a2a1..2d9f3f4cd79e 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_ras_eeprom.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_ras_eeprom.c @@ -33,12 +33,29 @@ #include "amdgpu_reset.h" -#define EEPROM_I2C_MADDR_VEGA20 0x0 -#define EEPROM_I2C_MADDR_ARCTURUS 0x40000 -#define EEPROM_I2C_MADDR_ARCTURUS_D342 0x0 -#define EEPROM_I2C_MADDR_SIENNA_CICHLID 0x0 -#define EEPROM_I2C_MADDR_ALDEBARAN 0x0 -#define EEPROM_I2C_MADDR_SMU_13_0_0 (0x54UL << 16) +/* These are memory addresses as would be seen by one or more EEPROM + * chips strung on the I2C bus, usually by manipulating pins 1-3 of a + * set of EEPROM devices. They form a continuous memory space. + * + * The I2C device address includes the device type identifier, 1010b, + * which is a reserved value and indicates that this is an I2C EEPROM + * device. It also includes the top 3 bits of the 19 bit EEPROM memory + * address, namely bits 18, 17, and 16. This makes up the 7 bit + * address sent on the I2C bus with bit 0 being the direction bit, + * which is not represented here, and sent by the hardware directly. + * + * For instance, + * 50h = 1010000b => device type identifier 1010b, bits 18:16 = 000b, address 0. + * 54h = 1010100b => --"--, bits 18:16 = 100b, address 40000h. + * 56h = 1010110b => --"--, bits 18:16 = 110b, address 60000h. + * Depending on the size of the I2C EEPROM device(s), bits 18:16 may + * address memory in a device or a device on the I2C bus, depending on + * the status of pins 1-3. See top of amdgpu_eeprom.c. + * + * The RAS table lives either at address 0 or address 40000h of EEPROM. + */ +#define EEPROM_I2C_MADDR_0 0x0 +#define EEPROM_I2C_MADDR_4 0x40000 /* * The 2 macros bellow represent the actual size in bytes that @@ -90,6 +107,16 @@ static bool __is_ras_eeprom_supported(struct amdgpu_device *adev) { + if (adev->asic_type == CHIP_IP_DISCOVERY) { + switch (adev->ip_versions[MP1_HWIP][0]) { + case IP_VERSION(13, 0, 0): + case IP_VERSION(13, 0, 10): + return true; + default: + return false; + } + } + return adev->asic_type == CHIP_VEGA20 || adev->asic_type == CHIP_ARCTURUS || adev->asic_type == CHIP_SIENNA_CICHLID || @@ -107,16 +134,30 @@ static bool __get_eeprom_i2c_addr_arct(struct amdgpu_device *adev, if (strnstr(atom_ctx->vbios_version, "D342", sizeof(atom_ctx->vbios_version))) - control->i2c_address = EEPROM_I2C_MADDR_ARCTURUS_D342; + control->i2c_address = EEPROM_I2C_MADDR_0; else - control->i2c_address = EEPROM_I2C_MADDR_ARCTURUS; + control->i2c_address = EEPROM_I2C_MADDR_4; return true; } +static bool __get_eeprom_i2c_addr_ip_discovery(struct amdgpu_device *adev, + struct amdgpu_ras_eeprom_control *control) +{ + switch (adev->ip_versions[MP1_HWIP][0]) { + case IP_VERSION(13, 0, 0): + case IP_VERSION(13, 0, 10): + control->i2c_address = EEPROM_I2C_MADDR_4; + return true; + default: + return false; + } +} + static bool __get_eeprom_i2c_addr(struct amdgpu_device *adev, struct amdgpu_ras_eeprom_control *control) { + struct atom_context *atom_ctx = adev->mode_info.atom_context; u8 i2c_addr; if (!control) @@ -139,27 +180,34 @@ static bool __get_eeprom_i2c_addr(struct amdgpu_device *adev, switch (adev->asic_type) { case CHIP_VEGA20: - control->i2c_address = EEPROM_I2C_MADDR_VEGA20; + control->i2c_address = EEPROM_I2C_MADDR_0; break; case CHIP_ARCTURUS: return __get_eeprom_i2c_addr_arct(adev, control); case CHIP_SIENNA_CICHLID: - control->i2c_address = EEPROM_I2C_MADDR_SIENNA_CICHLID; + control->i2c_address = EEPROM_I2C_MADDR_0; break; case CHIP_ALDEBARAN: - control->i2c_address = EEPROM_I2C_MADDR_ALDEBARAN; + if (strnstr(atom_ctx->vbios_version, "D673", + sizeof(atom_ctx->vbios_version))) + control->i2c_address = EEPROM_I2C_MADDR_4; + else + control->i2c_address = EEPROM_I2C_MADDR_0; break; + case CHIP_IP_DISCOVERY: + return __get_eeprom_i2c_addr_ip_discovery(adev, control); + default: return false; } switch (adev->ip_versions[MP1_HWIP][0]) { case IP_VERSION(13, 0, 0): - control->i2c_address = EEPROM_I2C_MADDR_SMU_13_0_0; + control->i2c_address = EEPROM_I2C_MADDR_4; break; default: diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_res_cursor.h b/drivers/gpu/drm/amd/amdgpu/amdgpu_res_cursor.h index 6546552e596c..5c4f93ee0c57 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_res_cursor.h +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_res_cursor.h @@ -62,7 +62,7 @@ static inline void amdgpu_res_first(struct ttm_resource *res, if (!res) goto fallback; - BUG_ON(start + size > res->num_pages << PAGE_SHIFT); + BUG_ON(start + size > res->size); cur->mem_type = res->mem_type; @@ -110,7 +110,7 @@ fallback: cur->size = size; cur->remaining = size; cur->node = NULL; - WARN_ON(res && start + size > res->num_pages << PAGE_SHIFT); + WARN_ON(res && start + size > res->size); return; } diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_ring.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_ring.c index d3558c34d406..dc474b809604 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_ring.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_ring.c @@ -569,3 +569,15 @@ int amdgpu_ring_init_mqd(struct amdgpu_ring *ring) return mqd_mgr->init_mqd(adev, ring->mqd_ptr, &prop); } + +void amdgpu_ring_ib_begin(struct amdgpu_ring *ring) +{ + if (ring->is_sw_ring) + amdgpu_sw_ring_ib_begin(ring); +} + +void amdgpu_ring_ib_end(struct amdgpu_ring *ring) +{ + if (ring->is_sw_ring) + amdgpu_sw_ring_ib_end(ring); +} diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_ring.h b/drivers/gpu/drm/amd/amdgpu/amdgpu_ring.h index 82c178a9033a..f752c7ae7f60 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_ring.h +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_ring.h @@ -39,6 +39,7 @@ struct amdgpu_vm; #define AMDGPU_MAX_RINGS 28 #define AMDGPU_MAX_HWIP_RINGS 8 #define AMDGPU_MAX_GFX_RINGS 2 +#define AMDGPU_MAX_SW_GFX_RINGS 2 #define AMDGPU_MAX_COMPUTE_RINGS 8 #define AMDGPU_MAX_VCE_RINGS 3 #define AMDGPU_MAX_UVD_ENC_RINGS 2 @@ -59,6 +60,7 @@ enum amdgpu_ring_priority_level { #define AMDGPU_FENCE_FLAG_64BIT (1 << 0) #define AMDGPU_FENCE_FLAG_INT (1 << 1) #define AMDGPU_FENCE_FLAG_TC_WB_ONLY (1 << 2) +#define AMDGPU_FENCE_FLAG_EXEC (1 << 3) #define to_amdgpu_ring(s) container_of((s), struct amdgpu_ring, sched) @@ -143,8 +145,13 @@ signed long amdgpu_fence_wait_polling(struct amdgpu_ring *ring, uint32_t wait_seq, signed long timeout); unsigned amdgpu_fence_count_emitted(struct amdgpu_ring *ring); + void amdgpu_fence_driver_isr_toggle(struct amdgpu_device *adev, bool stop); +u64 amdgpu_fence_last_unsignaled_time_us(struct amdgpu_ring *ring); +void amdgpu_fence_update_start_timestamp(struct amdgpu_ring *ring, uint32_t seq, + ktime_t timestamp); + /* * Rings. */ @@ -279,6 +286,10 @@ struct amdgpu_ring { bool is_mes_queue; uint32_t hw_queue_id; struct amdgpu_mes_ctx_data *mes_ctx; + + bool is_sw_ring; + unsigned int entry_index; + }; #define amdgpu_ring_parse_cs(r, p, job, ib) ((r)->funcs->parse_cs((p), (job), (ib))) @@ -307,6 +318,9 @@ struct amdgpu_ring { #define amdgpu_ring_preempt_ib(r) (r)->funcs->preempt_ib(r) int amdgpu_ring_alloc(struct amdgpu_ring *ring, unsigned ndw); +void amdgpu_ring_ib_begin(struct amdgpu_ring *ring); +void amdgpu_ring_ib_end(struct amdgpu_ring *ring); + void amdgpu_ring_insert_nop(struct amdgpu_ring *ring, uint32_t count); void amdgpu_ring_generic_pad_ib(struct amdgpu_ring *ring, struct amdgpu_ib *ib); void amdgpu_ring_commit(struct amdgpu_ring *ring); diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_ring_mux.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_ring_mux.c new file mode 100644 index 000000000000..62079f0e3ee8 --- /dev/null +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_ring_mux.c @@ -0,0 +1,514 @@ +/* + * Copyright 2022 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. + * + */ +#include <linux/slab.h> +#include <drm/drm_print.h> + +#include "amdgpu_ring_mux.h" +#include "amdgpu_ring.h" +#include "amdgpu.h" + +#define AMDGPU_MUX_RESUBMIT_JIFFIES_TIMEOUT (HZ / 2) +#define AMDGPU_MAX_LAST_UNSIGNALED_THRESHOLD_US 10000 + +static const struct ring_info { + unsigned int hw_pio; + const char *ring_name; +} sw_ring_info[] = { + { AMDGPU_RING_PRIO_DEFAULT, "gfx_low"}, + { AMDGPU_RING_PRIO_2, "gfx_high"}, +}; + +static struct kmem_cache *amdgpu_mux_chunk_slab; + +static inline struct amdgpu_mux_entry *amdgpu_ring_mux_sw_entry(struct amdgpu_ring_mux *mux, + struct amdgpu_ring *ring) +{ + return ring->entry_index < mux->ring_entry_size ? + &mux->ring_entry[ring->entry_index] : NULL; +} + +/* copy packages on sw ring range[begin, end) */ +static void amdgpu_ring_mux_copy_pkt_from_sw_ring(struct amdgpu_ring_mux *mux, + struct amdgpu_ring *ring, + u64 s_start, u64 s_end) +{ + u64 start, end; + struct amdgpu_ring *real_ring = mux->real_ring; + + start = s_start & ring->buf_mask; + end = s_end & ring->buf_mask; + + if (start == end) { + DRM_ERROR("no more data copied from sw ring\n"); + return; + } + if (start > end) { + amdgpu_ring_alloc(real_ring, (ring->ring_size >> 2) + end - start); + amdgpu_ring_write_multiple(real_ring, (void *)&ring->ring[start], + (ring->ring_size >> 2) - start); + amdgpu_ring_write_multiple(real_ring, (void *)&ring->ring[0], end); + } else { + amdgpu_ring_alloc(real_ring, end - start); + amdgpu_ring_write_multiple(real_ring, (void *)&ring->ring[start], end - start); + } +} + +static void amdgpu_mux_resubmit_chunks(struct amdgpu_ring_mux *mux) +{ + struct amdgpu_mux_entry *e = NULL; + struct amdgpu_mux_chunk *chunk; + uint32_t seq, last_seq; + int i; + + /*find low priority entries:*/ + if (!mux->s_resubmit) + return; + + for (i = 0; i < mux->num_ring_entries; i++) { + if (mux->ring_entry[i].ring->hw_prio <= AMDGPU_RING_PRIO_DEFAULT) { + e = &mux->ring_entry[i]; + break; + } + } + + if (!e) { + DRM_ERROR("%s no low priority ring found\n", __func__); + return; + } + + last_seq = atomic_read(&e->ring->fence_drv.last_seq); + seq = mux->seqno_to_resubmit; + if (last_seq < seq) { + /*resubmit all the fences between (last_seq, seq]*/ + list_for_each_entry(chunk, &e->list, entry) { + if (chunk->sync_seq > last_seq && chunk->sync_seq <= seq) { + amdgpu_fence_update_start_timestamp(e->ring, + chunk->sync_seq, + ktime_get()); + amdgpu_ring_mux_copy_pkt_from_sw_ring(mux, e->ring, + chunk->start, + chunk->end); + mux->wptr_resubmit = chunk->end; + amdgpu_ring_commit(mux->real_ring); + } + } + } + + del_timer(&mux->resubmit_timer); + mux->s_resubmit = false; +} + +static void amdgpu_ring_mux_schedule_resubmit(struct amdgpu_ring_mux *mux) +{ + mod_timer(&mux->resubmit_timer, jiffies + AMDGPU_MUX_RESUBMIT_JIFFIES_TIMEOUT); +} + +static void amdgpu_mux_resubmit_fallback(struct timer_list *t) +{ + struct amdgpu_ring_mux *mux = from_timer(mux, t, resubmit_timer); + + if (!spin_trylock(&mux->lock)) { + amdgpu_ring_mux_schedule_resubmit(mux); + DRM_ERROR("reschedule resubmit\n"); + return; + } + amdgpu_mux_resubmit_chunks(mux); + spin_unlock(&mux->lock); +} + +int amdgpu_ring_mux_init(struct amdgpu_ring_mux *mux, struct amdgpu_ring *ring, + unsigned int entry_size) +{ + mux->real_ring = ring; + mux->num_ring_entries = 0; + + mux->ring_entry = kcalloc(entry_size, sizeof(struct amdgpu_mux_entry), GFP_KERNEL); + if (!mux->ring_entry) + return -ENOMEM; + + mux->ring_entry_size = entry_size; + mux->s_resubmit = false; + + amdgpu_mux_chunk_slab = kmem_cache_create("amdgpu_mux_chunk", + sizeof(struct amdgpu_mux_chunk), 0, + SLAB_HWCACHE_ALIGN, NULL); + if (!amdgpu_mux_chunk_slab) { + DRM_ERROR("create amdgpu_mux_chunk cache failed\n"); + return -ENOMEM; + } + + spin_lock_init(&mux->lock); + timer_setup(&mux->resubmit_timer, amdgpu_mux_resubmit_fallback, 0); + + return 0; +} + +void amdgpu_ring_mux_fini(struct amdgpu_ring_mux *mux) +{ + struct amdgpu_mux_entry *e; + struct amdgpu_mux_chunk *chunk, *chunk2; + int i; + + for (i = 0; i < mux->num_ring_entries; i++) { + e = &mux->ring_entry[i]; + list_for_each_entry_safe(chunk, chunk2, &e->list, entry) { + list_del(&chunk->entry); + kmem_cache_free(amdgpu_mux_chunk_slab, chunk); + } + } + kmem_cache_destroy(amdgpu_mux_chunk_slab); + kfree(mux->ring_entry); + mux->ring_entry = NULL; + mux->num_ring_entries = 0; + mux->ring_entry_size = 0; +} + +int amdgpu_ring_mux_add_sw_ring(struct amdgpu_ring_mux *mux, struct amdgpu_ring *ring) +{ + struct amdgpu_mux_entry *e; + + if (mux->num_ring_entries >= mux->ring_entry_size) { + DRM_ERROR("add sw ring exceeding max entry size\n"); + return -ENOENT; + } + + e = &mux->ring_entry[mux->num_ring_entries]; + ring->entry_index = mux->num_ring_entries; + e->ring = ring; + + INIT_LIST_HEAD(&e->list); + mux->num_ring_entries += 1; + return 0; +} + +void amdgpu_ring_mux_set_wptr(struct amdgpu_ring_mux *mux, struct amdgpu_ring *ring, u64 wptr) +{ + struct amdgpu_mux_entry *e; + + spin_lock(&mux->lock); + + if (ring->hw_prio <= AMDGPU_RING_PRIO_DEFAULT) + amdgpu_mux_resubmit_chunks(mux); + + e = amdgpu_ring_mux_sw_entry(mux, ring); + if (!e) { + DRM_ERROR("cannot find entry for sw ring\n"); + spin_unlock(&mux->lock); + return; + } + + /* We could skip this set wptr as preemption in process. */ + if (ring->hw_prio <= AMDGPU_RING_PRIO_DEFAULT && mux->pending_trailing_fence_signaled) { + spin_unlock(&mux->lock); + return; + } + + e->sw_cptr = e->sw_wptr; + /* Update cptr if the package already copied in resubmit functions */ + if (ring->hw_prio <= AMDGPU_RING_PRIO_DEFAULT && e->sw_cptr < mux->wptr_resubmit) + e->sw_cptr = mux->wptr_resubmit; + e->sw_wptr = wptr; + e->start_ptr_in_hw_ring = mux->real_ring->wptr; + + /* Skip copying for the packages already resubmitted.*/ + if (ring->hw_prio > AMDGPU_RING_PRIO_DEFAULT || mux->wptr_resubmit < wptr) { + amdgpu_ring_mux_copy_pkt_from_sw_ring(mux, ring, e->sw_cptr, wptr); + e->end_ptr_in_hw_ring = mux->real_ring->wptr; + amdgpu_ring_commit(mux->real_ring); + } else { + e->end_ptr_in_hw_ring = mux->real_ring->wptr; + } + spin_unlock(&mux->lock); +} + +u64 amdgpu_ring_mux_get_wptr(struct amdgpu_ring_mux *mux, struct amdgpu_ring *ring) +{ + struct amdgpu_mux_entry *e; + + e = amdgpu_ring_mux_sw_entry(mux, ring); + if (!e) { + DRM_ERROR("cannot find entry for sw ring\n"); + return 0; + } + + return e->sw_wptr; +} + +/** + * amdgpu_ring_mux_get_rptr - get the readptr of the software ring + * @mux: the multiplexer the software rings attach to + * @ring: the software ring of which we calculate the readptr + * + * The return value of the readptr is not precise while the other rings could + * write data onto the real ring buffer.After overwriting on the real ring, we + * can not decide if our packages have been excuted or not read yet. However, + * this function is only called by the tools such as umr to collect the latest + * packages for the hang analysis. We assume the hang happens near our latest + * submit. Thus we could use the following logic to give the clue: + * If the readptr is between start and end, then we return the copy pointer + * plus the distance from start to readptr. If the readptr is before start, we + * return the copy pointer. Lastly, if the readptr is past end, we return the + * write pointer. + */ +u64 amdgpu_ring_mux_get_rptr(struct amdgpu_ring_mux *mux, struct amdgpu_ring *ring) +{ + struct amdgpu_mux_entry *e; + u64 readp, offset, start, end; + + e = amdgpu_ring_mux_sw_entry(mux, ring); + if (!e) { + DRM_ERROR("no sw entry found!\n"); + return 0; + } + + readp = amdgpu_ring_get_rptr(mux->real_ring); + + start = e->start_ptr_in_hw_ring & mux->real_ring->buf_mask; + end = e->end_ptr_in_hw_ring & mux->real_ring->buf_mask; + if (start > end) { + if (readp <= end) + readp += mux->real_ring->ring_size >> 2; + end += mux->real_ring->ring_size >> 2; + } + + if (start <= readp && readp <= end) { + offset = readp - start; + e->sw_rptr = (e->sw_cptr + offset) & ring->buf_mask; + } else if (readp < start) { + e->sw_rptr = e->sw_cptr; + } else { + /* end < readptr */ + e->sw_rptr = e->sw_wptr; + } + + return e->sw_rptr; +} + +u64 amdgpu_sw_ring_get_rptr_gfx(struct amdgpu_ring *ring) +{ + struct amdgpu_device *adev = ring->adev; + struct amdgpu_ring_mux *mux = &adev->gfx.muxer; + + WARN_ON(!ring->is_sw_ring); + return amdgpu_ring_mux_get_rptr(mux, ring); +} + +u64 amdgpu_sw_ring_get_wptr_gfx(struct amdgpu_ring *ring) +{ + struct amdgpu_device *adev = ring->adev; + struct amdgpu_ring_mux *mux = &adev->gfx.muxer; + + WARN_ON(!ring->is_sw_ring); + return amdgpu_ring_mux_get_wptr(mux, ring); +} + +void amdgpu_sw_ring_set_wptr_gfx(struct amdgpu_ring *ring) +{ + struct amdgpu_device *adev = ring->adev; + struct amdgpu_ring_mux *mux = &adev->gfx.muxer; + + WARN_ON(!ring->is_sw_ring); + amdgpu_ring_mux_set_wptr(mux, ring, ring->wptr); +} + +/* Override insert_nop to prevent emitting nops to the software rings */ +void amdgpu_sw_ring_insert_nop(struct amdgpu_ring *ring, uint32_t count) +{ + WARN_ON(!ring->is_sw_ring); +} + +const char *amdgpu_sw_ring_name(int idx) +{ + return idx < ARRAY_SIZE(sw_ring_info) ? + sw_ring_info[idx].ring_name : NULL; +} + +unsigned int amdgpu_sw_ring_priority(int idx) +{ + return idx < ARRAY_SIZE(sw_ring_info) ? + sw_ring_info[idx].hw_pio : AMDGPU_RING_PRIO_DEFAULT; +} + +/*Scan on low prio rings to have unsignaled fence and high ring has no fence.*/ +static int amdgpu_mcbp_scan(struct amdgpu_ring_mux *mux) +{ + struct amdgpu_ring *ring; + int i, need_preempt; + + need_preempt = 0; + for (i = 0; i < mux->num_ring_entries; i++) { + ring = mux->ring_entry[i].ring; + if (ring->hw_prio > AMDGPU_RING_PRIO_DEFAULT && + amdgpu_fence_count_emitted(ring) > 0) + return 0; + if (ring->hw_prio <= AMDGPU_RING_PRIO_DEFAULT && + amdgpu_fence_last_unsignaled_time_us(ring) > + AMDGPU_MAX_LAST_UNSIGNALED_THRESHOLD_US) + need_preempt = 1; + } + return need_preempt && !mux->s_resubmit; +} + +/* Trigger Mid-Command Buffer Preemption (MCBP) and find if we need to resubmit. */ +static int amdgpu_mcbp_trigger_preempt(struct amdgpu_ring_mux *mux) +{ + int r; + + spin_lock(&mux->lock); + mux->pending_trailing_fence_signaled = true; + r = amdgpu_ring_preempt_ib(mux->real_ring); + spin_unlock(&mux->lock); + return r; +} + +void amdgpu_sw_ring_ib_begin(struct amdgpu_ring *ring) +{ + struct amdgpu_device *adev = ring->adev; + struct amdgpu_ring_mux *mux = &adev->gfx.muxer; + + WARN_ON(!ring->is_sw_ring); + if (ring->hw_prio > AMDGPU_RING_PRIO_DEFAULT) { + if (amdgpu_mcbp_scan(mux) > 0) + amdgpu_mcbp_trigger_preempt(mux); + return; + } + + amdgpu_ring_mux_start_ib(mux, ring); +} + +void amdgpu_sw_ring_ib_end(struct amdgpu_ring *ring) +{ + struct amdgpu_device *adev = ring->adev; + struct amdgpu_ring_mux *mux = &adev->gfx.muxer; + + WARN_ON(!ring->is_sw_ring); + if (ring->hw_prio > AMDGPU_RING_PRIO_DEFAULT) + return; + amdgpu_ring_mux_end_ib(mux, ring); +} + +void amdgpu_ring_mux_start_ib(struct amdgpu_ring_mux *mux, struct amdgpu_ring *ring) +{ + struct amdgpu_mux_entry *e; + struct amdgpu_mux_chunk *chunk; + + spin_lock(&mux->lock); + amdgpu_mux_resubmit_chunks(mux); + spin_unlock(&mux->lock); + + e = amdgpu_ring_mux_sw_entry(mux, ring); + if (!e) { + DRM_ERROR("cannot find entry!\n"); + return; + } + + chunk = kmem_cache_alloc(amdgpu_mux_chunk_slab, GFP_KERNEL); + if (!chunk) { + DRM_ERROR("alloc amdgpu_mux_chunk_slab failed\n"); + return; + } + + chunk->start = ring->wptr; + list_add_tail(&chunk->entry, &e->list); +} + +static void scan_and_remove_signaled_chunk(struct amdgpu_ring_mux *mux, struct amdgpu_ring *ring) +{ + uint32_t last_seq = 0; + struct amdgpu_mux_entry *e; + struct amdgpu_mux_chunk *chunk, *tmp; + + e = amdgpu_ring_mux_sw_entry(mux, ring); + if (!e) { + DRM_ERROR("cannot find entry!\n"); + return; + } + + last_seq = atomic_read(&ring->fence_drv.last_seq); + + list_for_each_entry_safe(chunk, tmp, &e->list, entry) { + if (chunk->sync_seq <= last_seq) { + list_del(&chunk->entry); + kmem_cache_free(amdgpu_mux_chunk_slab, chunk); + } + } +} + +void amdgpu_ring_mux_end_ib(struct amdgpu_ring_mux *mux, struct amdgpu_ring *ring) +{ + struct amdgpu_mux_entry *e; + struct amdgpu_mux_chunk *chunk; + + e = amdgpu_ring_mux_sw_entry(mux, ring); + if (!e) { + DRM_ERROR("cannot find entry!\n"); + return; + } + + chunk = list_last_entry(&e->list, struct amdgpu_mux_chunk, entry); + if (!chunk) { + DRM_ERROR("cannot find chunk!\n"); + return; + } + + chunk->end = ring->wptr; + chunk->sync_seq = READ_ONCE(ring->fence_drv.sync_seq); + + scan_and_remove_signaled_chunk(mux, ring); +} + +bool amdgpu_mcbp_handle_trailing_fence_irq(struct amdgpu_ring_mux *mux) +{ + struct amdgpu_mux_entry *e; + struct amdgpu_ring *ring = NULL; + int i; + + if (!mux->pending_trailing_fence_signaled) + return false; + + if (mux->real_ring->trail_seq != le32_to_cpu(*mux->real_ring->trail_fence_cpu_addr)) + return false; + + for (i = 0; i < mux->num_ring_entries; i++) { + e = &mux->ring_entry[i]; + if (e->ring->hw_prio <= AMDGPU_RING_PRIO_DEFAULT) { + ring = e->ring; + break; + } + } + + if (!ring) { + DRM_ERROR("cannot find low priority ring\n"); + return false; + } + + amdgpu_fence_process(ring); + if (amdgpu_fence_count_emitted(ring) > 0) { + mux->s_resubmit = true; + mux->seqno_to_resubmit = ring->fence_drv.sync_seq; + amdgpu_ring_mux_schedule_resubmit(mux); + } + + mux->pending_trailing_fence_signaled = false; + return true; +} diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_ring_mux.h b/drivers/gpu/drm/amd/amdgpu/amdgpu_ring_mux.h new file mode 100644 index 000000000000..4be45fc14954 --- /dev/null +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_ring_mux.h @@ -0,0 +1,103 @@ +/* + * Copyright 2022 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 __AMDGPU_RING_MUX__ +#define __AMDGPU_RING_MUX__ + +#include <linux/timer.h> +#include <linux/spinlock.h> +#include "amdgpu_ring.h" + +struct amdgpu_ring; + +/** + * struct amdgpu_mux_entry - the entry recording software rings copying information. + * @ring: the pointer to the software ring. + * @start_ptr_in_hw_ring: last start location copied to in the hardware ring. + * @end_ptr_in_hw_ring: last end location copied to in the hardware ring. + * @sw_cptr: the position of the copy pointer in the sw ring. + * @sw_rptr: the read pointer in software ring. + * @sw_wptr: the write pointer in software ring. + * @list: list head for amdgpu_mux_chunk + */ +struct amdgpu_mux_entry { + struct amdgpu_ring *ring; + u64 start_ptr_in_hw_ring; + u64 end_ptr_in_hw_ring; + u64 sw_cptr; + u64 sw_rptr; + u64 sw_wptr; + struct list_head list; +}; + +struct amdgpu_ring_mux { + struct amdgpu_ring *real_ring; + + struct amdgpu_mux_entry *ring_entry; + unsigned int num_ring_entries; + unsigned int ring_entry_size; + /*the lock for copy data from different software rings*/ + spinlock_t lock; + bool s_resubmit; + uint32_t seqno_to_resubmit; + u64 wptr_resubmit; + struct timer_list resubmit_timer; + + bool pending_trailing_fence_signaled; +}; + +/** + * struct amdgpu_mux_chunk - save the location of indirect buffer's package on softare rings. + * @entry: the list entry. + * @sync_seq: the fence seqno related with the saved IB. + * @start:- start location on the software ring. + * @end:- end location on the software ring. + */ +struct amdgpu_mux_chunk { + struct list_head entry; + uint32_t sync_seq; + u64 start; + u64 end; +}; + +int amdgpu_ring_mux_init(struct amdgpu_ring_mux *mux, struct amdgpu_ring *ring, + unsigned int entry_size); +void amdgpu_ring_mux_fini(struct amdgpu_ring_mux *mux); +int amdgpu_ring_mux_add_sw_ring(struct amdgpu_ring_mux *mux, struct amdgpu_ring *ring); +void amdgpu_ring_mux_set_wptr(struct amdgpu_ring_mux *mux, struct amdgpu_ring *ring, u64 wptr); +u64 amdgpu_ring_mux_get_wptr(struct amdgpu_ring_mux *mux, struct amdgpu_ring *ring); +u64 amdgpu_ring_mux_get_rptr(struct amdgpu_ring_mux *mux, struct amdgpu_ring *ring); +void amdgpu_ring_mux_start_ib(struct amdgpu_ring_mux *mux, struct amdgpu_ring *ring); +void amdgpu_ring_mux_end_ib(struct amdgpu_ring_mux *mux, struct amdgpu_ring *ring); +bool amdgpu_mcbp_handle_trailing_fence_irq(struct amdgpu_ring_mux *mux); + +u64 amdgpu_sw_ring_get_rptr_gfx(struct amdgpu_ring *ring); +u64 amdgpu_sw_ring_get_wptr_gfx(struct amdgpu_ring *ring); +void amdgpu_sw_ring_set_wptr_gfx(struct amdgpu_ring *ring); +void amdgpu_sw_ring_insert_nop(struct amdgpu_ring *ring, uint32_t count); +void amdgpu_sw_ring_ib_begin(struct amdgpu_ring *ring); +void amdgpu_sw_ring_ib_end(struct amdgpu_ring *ring); +const char *amdgpu_sw_ring_name(int idx); +unsigned int amdgpu_sw_ring_priority(int idx); + +#endif diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_securedisplay.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_securedisplay.c index cc7597a15fe9..2c1d82fc4c34 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_securedisplay.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_securedisplay.c @@ -121,6 +121,7 @@ static ssize_t amdgpu_securedisplay_debugfs_write(struct file *f, const char __u switch (op) { case 1: + mutex_lock(&psp->securedisplay_context.mutex); psp_prep_securedisplay_cmd_buf(psp, &securedisplay_cmd, TA_SECUREDISPLAY_COMMAND__QUERY_TA); ret = psp_securedisplay_invoke(psp, TA_SECUREDISPLAY_COMMAND__QUERY_TA); @@ -131,8 +132,10 @@ static ssize_t amdgpu_securedisplay_debugfs_write(struct file *f, const char __u else psp_securedisplay_parse_resp_status(psp, securedisplay_cmd->status); } + mutex_unlock(&psp->securedisplay_context.mutex); break; case 2: + mutex_lock(&psp->securedisplay_context.mutex); psp_prep_securedisplay_cmd_buf(psp, &securedisplay_cmd, TA_SECUREDISPLAY_COMMAND__SEND_ROI_CRC); securedisplay_cmd->securedisplay_in_message.send_roi_crc.phy_id = phy_id; @@ -146,6 +149,7 @@ static ssize_t amdgpu_securedisplay_debugfs_write(struct file *f, const char __u psp_securedisplay_parse_resp_status(psp, securedisplay_cmd->status); } } + mutex_unlock(&psp->securedisplay_context.mutex); break; default: dev_err(adev->dev, "Invalid input: %s\n", str); diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_sync.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_sync.c index 090e66a1b284..dcd8c066bc1f 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_sync.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_sync.c @@ -259,6 +259,14 @@ int amdgpu_sync_resv(struct amdgpu_device *adev, struct amdgpu_sync *sync, return 0; } +/* Free the entry back to the slab */ +static void amdgpu_sync_entry_free(struct amdgpu_sync_entry *e) +{ + hash_del(&e->node); + dma_fence_put(e->fence); + kmem_cache_free(amdgpu_sync_slab, e); +} + /** * amdgpu_sync_peek_fence - get the next fence not signaled yet * @@ -280,9 +288,7 @@ struct dma_fence *amdgpu_sync_peek_fence(struct amdgpu_sync *sync, struct drm_sched_fence *s_fence = to_drm_sched_fence(f); if (dma_fence_is_signaled(f)) { - hash_del(&e->node); - dma_fence_put(f); - kmem_cache_free(amdgpu_sync_slab, e); + amdgpu_sync_entry_free(e); continue; } if (ring && s_fence) { @@ -355,15 +361,44 @@ int amdgpu_sync_clone(struct amdgpu_sync *source, struct amdgpu_sync *clone) if (r) return r; } else { - hash_del(&e->node); - dma_fence_put(f); - kmem_cache_free(amdgpu_sync_slab, e); + amdgpu_sync_entry_free(e); } } return 0; } +/** + * amdgpu_sync_push_to_job - push fences into job + * @sync: sync object to get the fences from + * @job: job to push the fences into + * + * Add all unsignaled fences from sync to job. + */ +int amdgpu_sync_push_to_job(struct amdgpu_sync *sync, struct amdgpu_job *job) +{ + struct amdgpu_sync_entry *e; + struct hlist_node *tmp; + struct dma_fence *f; + int i, r; + + hash_for_each_safe(sync->fences, i, tmp, e, node) { + f = e->fence; + if (dma_fence_is_signaled(f)) { + amdgpu_sync_entry_free(e); + continue; + } + + dma_fence_get(f); + r = drm_sched_job_add_dependency(&job->base, f); + if (r) { + dma_fence_put(f); + return r; + } + } + return 0; +} + int amdgpu_sync_wait(struct amdgpu_sync *sync, bool intr) { struct amdgpu_sync_entry *e; @@ -375,9 +410,7 @@ int amdgpu_sync_wait(struct amdgpu_sync *sync, bool intr) if (r) return r; - hash_del(&e->node); - dma_fence_put(e->fence); - kmem_cache_free(amdgpu_sync_slab, e); + amdgpu_sync_entry_free(e); } return 0; @@ -396,11 +429,8 @@ void amdgpu_sync_free(struct amdgpu_sync *sync) struct hlist_node *tmp; unsigned int i; - hash_for_each_safe(sync->fences, i, tmp, e, node) { - hash_del(&e->node); - dma_fence_put(e->fence); - kmem_cache_free(amdgpu_sync_slab, e); - } + hash_for_each_safe(sync->fences, i, tmp, e, node) + amdgpu_sync_entry_free(e); } /** diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_sync.h b/drivers/gpu/drm/amd/amdgpu/amdgpu_sync.h index 2d5c613cda10..cf1e9e858efd 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_sync.h +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_sync.h @@ -30,6 +30,7 @@ struct dma_fence; struct dma_resv; struct amdgpu_device; struct amdgpu_ring; +struct amdgpu_job; enum amdgpu_sync_mode { AMDGPU_SYNC_ALWAYS, @@ -54,6 +55,7 @@ struct dma_fence *amdgpu_sync_peek_fence(struct amdgpu_sync *sync, struct amdgpu_ring *ring); struct dma_fence *amdgpu_sync_get_fence(struct amdgpu_sync *sync); int amdgpu_sync_clone(struct amdgpu_sync *source, struct amdgpu_sync *clone); +int amdgpu_sync_push_to_job(struct amdgpu_sync *sync, struct amdgpu_job *job); int amdgpu_sync_wait(struct amdgpu_sync *sync, bool intr); void amdgpu_sync_free(struct amdgpu_sync *sync); int amdgpu_sync_init(void); diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_trace.h b/drivers/gpu/drm/amd/amdgpu/amdgpu_trace.h index 5e6ddc7e101c..677ad2016976 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_trace.h +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_trace.h @@ -127,7 +127,7 @@ TRACE_EVENT(amdgpu_bo_create, TP_fast_assign( __entry->bo = bo; - __entry->pages = bo->tbo.resource->num_pages; + __entry->pages = PFN_UP(bo->tbo.resource->size); __entry->type = bo->tbo.resource->mem_type; __entry->prefer = bo->preferred_domains; __entry->allow = bo->allowed_domains; diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_ttm.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_ttm.c index b64938ed8cb6..55e0284b2bdd 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_ttm.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_ttm.c @@ -58,6 +58,7 @@ #include "amdgpu_amdkfd.h" #include "amdgpu_sdma.h" #include "amdgpu_ras.h" +#include "amdgpu_hmm.h" #include "amdgpu_atomfirmware.h" #include "amdgpu_res_cursor.h" #include "bif/bif_4_1_d.h" @@ -189,7 +190,6 @@ static int amdgpu_ttm_map_buffer(struct ttm_buffer_object *bo, struct amdgpu_device *adev = ring->adev; unsigned offset, num_pages, num_dw, num_bytes; uint64_t src_addr, dst_addr; - struct dma_fence *fence; struct amdgpu_job *job; void *cpu_addr; uint64_t flags; @@ -229,7 +229,9 @@ static int amdgpu_ttm_map_buffer(struct ttm_buffer_object *bo, num_dw = ALIGN(adev->mman.buffer_funcs->copy_num_dw, 8); num_bytes = num_pages * 8 * AMDGPU_GPU_PAGES_IN_CPU_PAGE; - r = amdgpu_job_alloc_with_ib(adev, num_dw * 4 + num_bytes, + r = amdgpu_job_alloc_with_ib(adev, &adev->mman.entity, + AMDGPU_FENCE_OWNER_UNDEFINED, + num_dw * 4 + num_bytes, AMDGPU_IB_POOL_DELAYED, &job); if (r) return r; @@ -269,18 +271,8 @@ static int amdgpu_ttm_map_buffer(struct ttm_buffer_object *bo, } } - r = amdgpu_job_submit(job, &adev->mman.entity, - AMDGPU_FENCE_OWNER_UNDEFINED, &fence); - if (r) - goto error_free; - - dma_fence_put(fence); - - return r; - -error_free: - amdgpu_job_free(job); - return r; + dma_fence_put(amdgpu_job_submit(job)); + return 0; } /** @@ -381,7 +373,7 @@ static int amdgpu_move_blit(struct ttm_buffer_object *bo, dst.offset = 0; r = amdgpu_ttm_copy_mem_to_mem(adev, &src, &dst, - new_mem->num_pages << PAGE_SHIFT, + new_mem->size, amdgpu_bo_encrypted(abo), bo->base.resv, &fence); if (r) @@ -424,7 +416,7 @@ error: static bool amdgpu_mem_visible(struct amdgpu_device *adev, struct ttm_resource *mem) { - u64 mem_size = (u64)mem->num_pages << PAGE_SHIFT; + u64 mem_size = (u64)mem->size; struct amdgpu_res_cursor cursor; u64 end; @@ -571,7 +563,7 @@ static int amdgpu_ttm_io_mem_reserve(struct ttm_device *bdev, struct ttm_resource *mem) { struct amdgpu_device *adev = amdgpu_ttm_adev(bdev); - size_t bus_size = (size_t)mem->num_pages << PAGE_SHIFT; + size_t bus_size = (size_t)mem->size; switch (mem->mem_type) { case TTM_PL_SYSTEM: @@ -691,9 +683,8 @@ int amdgpu_ttm_tt_get_user_pages(struct amdgpu_bo *bo, struct page **pages, } readonly = amdgpu_ttm_tt_is_readonly(ttm); - r = amdgpu_hmm_range_get_pages(&bo->notifier, mm, pages, start, - ttm->num_pages, range, readonly, - true, NULL); + r = amdgpu_hmm_range_get_pages(&bo->notifier, start, ttm->num_pages, + readonly, NULL, pages, range); out_unlock: mmap_read_unlock(mm); if (r) @@ -704,8 +695,19 @@ out_unlock: return r; } +/* amdgpu_ttm_tt_discard_user_pages - Discard range and pfn array allocations + */ +void amdgpu_ttm_tt_discard_user_pages(struct ttm_tt *ttm, + struct hmm_range *range) +{ + struct amdgpu_ttm_tt *gtt = (void *)ttm; + + if (gtt && gtt->userptr && range) + amdgpu_hmm_range_get_pages_done(range); +} + /* - * amdgpu_ttm_tt_userptr_range_done - stop HMM track the CPU page table change + * amdgpu_ttm_tt_get_user_pages_done - stop HMM track the CPU page table change * Check if the pages backing this ttm range have been invalidated * * Returns: true if pages are still valid @@ -723,10 +725,6 @@ bool amdgpu_ttm_tt_get_user_pages_done(struct ttm_tt *ttm, WARN_ONCE(!range->hmm_pfns, "No user pages to check\n"); - /* - * FIXME: Must always hold notifier_lock for this, and must - * not ignore the return code. - */ return !amdgpu_hmm_range_get_pages_done(range); } #endif @@ -1154,8 +1152,9 @@ int amdgpu_ttm_tt_get_userptr(const struct ttm_buffer_object *tbo, * @addr: The address in the current tasks VM space to use * @flags: Requirements of userptr object. * - * Called by amdgpu_gem_userptr_ioctl() to bind userptr pages - * to current task + * Called by amdgpu_gem_userptr_ioctl() and kfd_ioctl_alloc_memory_of_gpu() to + * bind userptr pages to current task and by kfd_ioctl_acquire_vm() to + * initialize GPU VM for a KFD process. */ int amdgpu_ttm_tt_set_userptr(struct ttm_buffer_object *bo, uint64_t addr, uint32_t flags) @@ -1394,7 +1393,8 @@ static void amdgpu_ttm_vram_mm_access(struct amdgpu_device *adev, loff_t pos, } static int amdgpu_ttm_access_memory_sdma(struct ttm_buffer_object *bo, - unsigned long offset, void *buf, int len, int write) + unsigned long offset, void *buf, + int len, int write) { struct amdgpu_bo *abo = ttm_to_amdgpu_bo(bo); struct amdgpu_device *adev = amdgpu_ttm_adev(abo->tbo.bdev); @@ -1418,26 +1418,27 @@ static int amdgpu_ttm_access_memory_sdma(struct ttm_buffer_object *bo, memcpy(adev->mman.sdma_access_ptr, buf, len); num_dw = ALIGN(adev->mman.buffer_funcs->copy_num_dw, 8); - r = amdgpu_job_alloc_with_ib(adev, num_dw * 4, AMDGPU_IB_POOL_DELAYED, &job); + r = amdgpu_job_alloc_with_ib(adev, &adev->mman.entity, + AMDGPU_FENCE_OWNER_UNDEFINED, + num_dw * 4, AMDGPU_IB_POOL_DELAYED, + &job); if (r) goto out; amdgpu_res_first(abo->tbo.resource, offset, len, &src_mm); - src_addr = amdgpu_ttm_domain_start(adev, bo->resource->mem_type) + src_mm.start; + src_addr = amdgpu_ttm_domain_start(adev, bo->resource->mem_type) + + src_mm.start; dst_addr = amdgpu_bo_gpu_offset(adev->mman.sdma_access_bo); if (write) swap(src_addr, dst_addr); - amdgpu_emit_copy_buffer(adev, &job->ibs[0], src_addr, dst_addr, PAGE_SIZE, false); + amdgpu_emit_copy_buffer(adev, &job->ibs[0], src_addr, dst_addr, + PAGE_SIZE, false); amdgpu_ring_pad_ib(adev->mman.buffer_funcs_ring, &job->ibs[0]); WARN_ON(job->ibs[0].length_dw > num_dw); - r = amdgpu_job_submit(job, &adev->mman.entity, AMDGPU_FENCE_OWNER_UNDEFINED, &fence); - if (r) { - amdgpu_job_free(job); - goto out; - } + fence = amdgpu_job_submit(job); if (!dma_fence_wait_timeout(fence, false, adev->sdma_timeout)) r = -ETIMEDOUT; @@ -1537,6 +1538,23 @@ static void amdgpu_ttm_fw_reserve_vram_fini(struct amdgpu_device *adev) NULL, &adev->mman.fw_vram_usage_va); } +/* + * Driver Reservation functions + */ +/** + * amdgpu_ttm_drv_reserve_vram_fini - free drv reserved vram + * + * @adev: amdgpu_device pointer + * + * free drv reserved vram if it has been reserved. + */ +static void amdgpu_ttm_drv_reserve_vram_fini(struct amdgpu_device *adev) +{ + amdgpu_bo_free_kernel(&adev->mman.drv_vram_usage_reserved_bo, + NULL, + &adev->mman.drv_vram_usage_va); +} + /** * amdgpu_ttm_fw_reserve_vram_init - create bo vram reservation from fw * @@ -1558,11 +1576,35 @@ static int amdgpu_ttm_fw_reserve_vram_init(struct amdgpu_device *adev) return amdgpu_bo_create_kernel_at(adev, adev->mman.fw_vram_usage_start_offset, adev->mman.fw_vram_usage_size, - AMDGPU_GEM_DOMAIN_VRAM, &adev->mman.fw_vram_usage_reserved_bo, &adev->mman.fw_vram_usage_va); } +/** + * amdgpu_ttm_drv_reserve_vram_init - create bo vram reservation from driver + * + * @adev: amdgpu_device pointer + * + * create bo vram reservation from drv. + */ +static int amdgpu_ttm_drv_reserve_vram_init(struct amdgpu_device *adev) +{ + u64 vram_size = adev->gmc.visible_vram_size; + + adev->mman.drv_vram_usage_va = NULL; + adev->mman.drv_vram_usage_reserved_bo = NULL; + + if (adev->mman.drv_vram_usage_size == 0 || + adev->mman.drv_vram_usage_size > vram_size) + return 0; + + return amdgpu_bo_create_kernel_at(adev, + adev->mman.drv_vram_usage_start_offset, + adev->mman.drv_vram_usage_size, + &adev->mman.drv_vram_usage_reserved_bo, + &adev->mman.drv_vram_usage_va); +} + /* * Memoy training reservation functions */ @@ -1639,7 +1681,6 @@ static int amdgpu_ttm_reserve_tmr(struct amdgpu_device *adev) ret = amdgpu_bo_create_kernel_at(adev, ctx->c2p_train_data_offset, ctx->train_data_size, - AMDGPU_GEM_DOMAIN_VRAM, &ctx->c2p_bo, NULL); if (ret) { @@ -1653,7 +1694,6 @@ static int amdgpu_ttm_reserve_tmr(struct amdgpu_device *adev) ret = amdgpu_bo_create_kernel_at(adev, adev->gmc.real_vram_size - adev->mman.discovery_tmr_size, adev->mman.discovery_tmr_size, - AMDGPU_GEM_DOMAIN_VRAM, &adev->mman.discovery_memory, NULL); if (ret) { @@ -1731,6 +1771,14 @@ int amdgpu_ttm_init(struct amdgpu_device *adev) } /* + *The reserved vram for driver must be pinned to the specified + *place on the VRAM, so reserve it early. + */ + r = amdgpu_ttm_drv_reserve_vram_init(adev); + if (r) + return r; + + /* * only NAVI10 and onwards ASIC support for IP discovery. * If IP discovery enabled, a block of memory should be * reserved for IP discovey. @@ -1746,21 +1794,18 @@ int amdgpu_ttm_init(struct amdgpu_device *adev) * avoid display artifacts while transitioning between pre-OS * and driver. */ r = amdgpu_bo_create_kernel_at(adev, 0, adev->mman.stolen_vga_size, - AMDGPU_GEM_DOMAIN_VRAM, &adev->mman.stolen_vga_memory, NULL); if (r) return r; r = amdgpu_bo_create_kernel_at(adev, adev->mman.stolen_vga_size, adev->mman.stolen_extended_size, - AMDGPU_GEM_DOMAIN_VRAM, &adev->mman.stolen_extended_memory, NULL); if (r) return r; r = amdgpu_bo_create_kernel_at(adev, adev->mman.stolen_reserved_offset, adev->mman.stolen_reserved_size, - AMDGPU_GEM_DOMAIN_VRAM, &adev->mman.stolen_reserved_memory, NULL); if (r) @@ -1855,6 +1900,7 @@ void amdgpu_ttm_fini(struct amdgpu_device *adev) amdgpu_bo_free_kernel(&adev->mman.sdma_access_bo, NULL, &adev->mman.sdma_access_ptr); amdgpu_ttm_fw_reserve_vram_fini(adev); + amdgpu_ttm_drv_reserve_vram_fini(adev); if (drm_dev_enter(adev_to_drm(adev), &idx)) { @@ -1936,7 +1982,9 @@ static int amdgpu_ttm_prepare_job(struct amdgpu_device *adev, AMDGPU_IB_POOL_DELAYED; int r; - r = amdgpu_job_alloc_with_ib(adev, num_dw * 4, pool, job); + r = amdgpu_job_alloc_with_ib(adev, &adev->mman.entity, + AMDGPU_FENCE_OWNER_UNDEFINED, + num_dw * 4, pool, job); if (r) return r; @@ -1946,17 +1994,11 @@ static int amdgpu_ttm_prepare_job(struct amdgpu_device *adev, adev->gart.bo); (*job)->vm_needs_flush = true; } - if (resv) { - r = amdgpu_sync_resv(adev, &(*job)->sync, resv, - AMDGPU_SYNC_ALWAYS, - AMDGPU_FENCE_OWNER_UNDEFINED); - if (r) { - DRM_ERROR("sync failed (%d).\n", r); - amdgpu_job_free(*job); - return r; - } - } - return 0; + if (!resv) + return 0; + + return drm_sched_job_add_resv_dependencies(&(*job)->base, resv, + DMA_RESV_USAGE_BOOKKEEP); } int amdgpu_copy_buffer(struct amdgpu_ring *ring, uint64_t src_offset, @@ -2001,8 +2043,7 @@ int amdgpu_copy_buffer(struct amdgpu_ring *ring, uint64_t src_offset, if (direct_submit) r = amdgpu_job_submit_direct(job, ring, fence); else - r = amdgpu_job_submit(job, &adev->mman.entity, - AMDGPU_FENCE_OWNER_UNDEFINED, fence); + *fence = amdgpu_job_submit(job); if (r) goto error_free; @@ -2047,16 +2088,8 @@ static int amdgpu_ttm_fill_mem(struct amdgpu_ring *ring, uint32_t src_data, amdgpu_ring_pad_ib(ring, &job->ibs[0]); WARN_ON(job->ibs[0].length_dw > num_dw); - r = amdgpu_job_submit(job, &adev->mman.entity, - AMDGPU_FENCE_OWNER_UNDEFINED, fence); - if (r) - goto error_free; - + *fence = amdgpu_job_submit(job); return 0; - -error_free: - amdgpu_job_free(job); - return r; } int amdgpu_fill_buffer(struct amdgpu_bo *bo, @@ -2272,9 +2305,9 @@ static ssize_t amdgpu_iomem_read(struct file *f, char __user *buf, if (p->mapping != adev->mman.bdev.dev_mapping) return -EPERM; - ptr = kmap(p); + ptr = kmap_local_page(p); r = copy_to_user(buf, ptr + off, bytes); - kunmap(p); + kunmap_local(ptr); if (r) return -EFAULT; @@ -2323,9 +2356,9 @@ static ssize_t amdgpu_iomem_write(struct file *f, const char __user *buf, if (p->mapping != adev->mman.bdev.dev_mapping) return -EPERM; - ptr = kmap(p); + ptr = kmap_local_page(p); r = copy_from_user(ptr + off, buf, bytes); - kunmap(p); + kunmap_local(ptr); if (r) return -EFAULT; diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_ttm.h b/drivers/gpu/drm/amd/amdgpu/amdgpu_ttm.h index a37207011a69..e2cd5894afc9 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_ttm.h +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_ttm.h @@ -86,6 +86,12 @@ struct amdgpu_mman { struct amdgpu_bo *fw_vram_usage_reserved_bo; void *fw_vram_usage_va; + /* driver VRAM reservation */ + u64 drv_vram_usage_start_offset; + u64 drv_vram_usage_size; + struct amdgpu_bo *drv_vram_usage_reserved_bo; + void *drv_vram_usage_va; + /* PAGE_SIZE'd BO for process memory r/w over SDMA. */ struct amdgpu_bo *sdma_access_bo; void *sdma_access_ptr; @@ -153,6 +159,8 @@ uint64_t amdgpu_ttm_domain_start(struct amdgpu_device *adev, uint32_t type); #if IS_ENABLED(CONFIG_DRM_AMDGPU_USERPTR) int amdgpu_ttm_tt_get_user_pages(struct amdgpu_bo *bo, struct page **pages, struct hmm_range **range); +void amdgpu_ttm_tt_discard_user_pages(struct ttm_tt *ttm, + struct hmm_range *range); bool amdgpu_ttm_tt_get_user_pages_done(struct ttm_tt *ttm, struct hmm_range *range); #else @@ -162,6 +170,10 @@ static inline int amdgpu_ttm_tt_get_user_pages(struct amdgpu_bo *bo, { return -EPERM; } +static inline void amdgpu_ttm_tt_discard_user_pages(struct ttm_tt *ttm, + struct hmm_range *range) +{ +} static inline bool amdgpu_ttm_tt_get_user_pages_done(struct ttm_tt *ttm, struct hmm_range *range) { diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_ucode.h b/drivers/gpu/drm/amd/amdgpu/amdgpu_ucode.h index 1c36235b4539..552e06929229 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_ucode.h +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_ucode.h @@ -125,6 +125,7 @@ enum psp_fw_type { PSP_FW_TYPE_PSP_INTF_DRV, PSP_FW_TYPE_PSP_DBG_DRV, PSP_FW_TYPE_PSP_RAS_DRV, + PSP_FW_TYPE_MAX_INDEX, }; /* version_major=2, version_minor=0 */ diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_umc.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_umc.c index aad3c8b4c810..f76c19fc0392 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_umc.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_umc.c @@ -22,6 +22,59 @@ */ #include "amdgpu.h" +#include "umc_v6_7.h" + +static int amdgpu_umc_convert_error_address(struct amdgpu_device *adev, + struct ras_err_data *err_data, uint64_t err_addr, + uint32_t ch_inst, uint32_t umc_inst) +{ + switch (adev->ip_versions[UMC_HWIP][0]) { + case IP_VERSION(6, 7, 0): + umc_v6_7_convert_error_address(adev, + err_data, err_addr, ch_inst, umc_inst); + break; + default: + dev_warn(adev->dev, + "UMC address to Physical address translation is not supported\n"); + return AMDGPU_RAS_FAIL; + } + + return AMDGPU_RAS_SUCCESS; +} + +int amdgpu_umc_page_retirement_mca(struct amdgpu_device *adev, + uint64_t err_addr, uint32_t ch_inst, uint32_t umc_inst) +{ + struct ras_err_data err_data = {0, 0, 0, NULL}; + int ret = AMDGPU_RAS_FAIL; + + err_data.err_addr = + kcalloc(adev->umc.max_ras_err_cnt_per_query, + sizeof(struct eeprom_table_record), GFP_KERNEL); + if (!err_data.err_addr) { + dev_warn(adev->dev, + "Failed to alloc memory for umc error record in MCA notifier!\n"); + return AMDGPU_RAS_FAIL; + } + + /* + * Translate UMC channel address to Physical address + */ + ret = amdgpu_umc_convert_error_address(adev, &err_data, err_addr, + ch_inst, umc_inst); + if (ret) + goto out; + + if (amdgpu_bad_page_threshold != 0) { + amdgpu_ras_add_bad_pages(adev, err_data.err_addr, + err_data.err_addr_cnt); + amdgpu_ras_save_bad_pages(adev); + } + +out: + kfree(err_data.err_addr); + return ret; +} static int amdgpu_umc_do_page_retirement(struct amdgpu_device *adev, void *ras_error_status, @@ -112,23 +165,29 @@ static int amdgpu_umc_do_page_retirement(struct amdgpu_device *adev, return AMDGPU_RAS_SUCCESS; } -int amdgpu_umc_poison_handler(struct amdgpu_device *adev, - void *ras_error_status, - bool reset) +int amdgpu_umc_poison_handler(struct amdgpu_device *adev, bool reset) { - int ret; - struct ras_err_data *err_data = (struct ras_err_data *)ras_error_status; - struct ras_common_if head = { - .block = AMDGPU_RAS_BLOCK__UMC, - }; - struct ras_manager *obj = amdgpu_ras_find_obj(adev, &head); + int ret = AMDGPU_RAS_SUCCESS; - ret = - amdgpu_umc_do_page_retirement(adev, ras_error_status, NULL, reset); + if (!adev->gmc.xgmi.connected_to_cpu) { + struct ras_err_data err_data = {0, 0, 0, NULL}; + struct ras_common_if head = { + .block = AMDGPU_RAS_BLOCK__UMC, + }; + struct ras_manager *obj = amdgpu_ras_find_obj(adev, &head); - if (ret == AMDGPU_RAS_SUCCESS && obj) { - obj->err_data.ue_count += err_data->ue_count; - obj->err_data.ce_count += err_data->ce_count; + ret = amdgpu_umc_do_page_retirement(adev, &err_data, NULL, reset); + + if (ret == AMDGPU_RAS_SUCCESS && obj) { + obj->err_data.ue_count += err_data.ue_count; + obj->err_data.ce_count += err_data.ce_count; + } + } else if (reset) { + /* MCA poison handler is only responsible for GPU reset, + * let MCA notifier do page retirement. + */ + kgd2kfd_set_sram_ecc_flag(adev->kfd.dev); + amdgpu_ras_reset_gpu(adev); } return ret; diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_umc.h b/drivers/gpu/drm/amd/amdgpu/amdgpu_umc.h index e46439274f3a..a6951160f13a 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_umc.h +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_umc.h @@ -51,9 +51,6 @@ struct amdgpu_umc_ras { struct amdgpu_ras_block_object ras_block; void (*err_cnt_init)(struct amdgpu_device *adev); bool (*query_ras_poison_mode)(struct amdgpu_device *adev); - void (*convert_ras_error_address)(struct amdgpu_device *adev, - struct ras_err_data *err_data, uint64_t err_addr, - uint32_t ch_inst, uint32_t umc_inst); void (*ecc_info_query_ras_error_count)(struct amdgpu_device *adev, void *ras_error_status); void (*ecc_info_query_ras_error_address)(struct amdgpu_device *adev, @@ -86,9 +83,7 @@ struct amdgpu_umc { }; int amdgpu_umc_ras_late_init(struct amdgpu_device *adev, struct ras_common_if *ras_block); -int amdgpu_umc_poison_handler(struct amdgpu_device *adev, - void *ras_error_status, - bool reset); +int amdgpu_umc_poison_handler(struct amdgpu_device *adev, bool reset); int amdgpu_umc_process_ecc_irq(struct amdgpu_device *adev, struct amdgpu_irq_src *source, struct amdgpu_iv_entry *entry); @@ -101,4 +96,6 @@ void amdgpu_umc_fill_error_record(struct ras_err_data *err_data, int amdgpu_umc_process_ras_data_cb(struct amdgpu_device *adev, void *ras_error_status, struct amdgpu_iv_entry *entry); +int amdgpu_umc_page_retirement_mca(struct amdgpu_device *adev, + uint64_t err_addr, uint32_t ch_inst, uint32_t umc_inst); #endif diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_uvd.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_uvd.c index 6eac649499d3..e00bb654e24b 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_uvd.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_uvd.c @@ -1132,7 +1132,9 @@ static int amdgpu_uvd_send_msg(struct amdgpu_ring *ring, struct amdgpu_bo *bo, unsigned offset_idx = 0; unsigned offset[3] = { UVD_BASE_SI, 0, 0 }; - r = amdgpu_job_alloc_with_ib(adev, 64, direct ? AMDGPU_IB_POOL_DIRECT : + r = amdgpu_job_alloc_with_ib(ring->adev, &adev->uvd.entity, + AMDGPU_FENCE_OWNER_UNDEFINED, + 64, direct ? AMDGPU_IB_POOL_DIRECT : AMDGPU_IB_POOL_DELAYED, &job); if (r) return r; @@ -1175,16 +1177,13 @@ static int amdgpu_uvd_send_msg(struct amdgpu_ring *ring, struct amdgpu_bo *bo, if (r) goto err_free; } else { - r = amdgpu_sync_resv(adev, &job->sync, bo->tbo.base.resv, - AMDGPU_SYNC_ALWAYS, - AMDGPU_FENCE_OWNER_UNDEFINED); + r = drm_sched_job_add_resv_dependencies(&job->base, + bo->tbo.base.resv, + DMA_RESV_USAGE_KERNEL); if (r) goto err_free; - r = amdgpu_job_submit(job, &adev->uvd.entity, - AMDGPU_FENCE_OWNER_UNDEFINED, &f); - if (r) - goto err_free; + f = amdgpu_job_submit(job); } amdgpu_bo_reserve(bo, true); diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_vce.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_vce.c index 02cb3a12dd76..b239e874f2d5 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_vce.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_vce.c @@ -450,8 +450,10 @@ static int amdgpu_vce_get_create_msg(struct amdgpu_ring *ring, uint32_t handle, uint64_t addr; int i, r; - r = amdgpu_job_alloc_with_ib(ring->adev, ib_size_dw * 4, - AMDGPU_IB_POOL_DIRECT, &job); + r = amdgpu_job_alloc_with_ib(ring->adev, &ring->adev->vce.entity, + AMDGPU_FENCE_OWNER_UNDEFINED, + ib_size_dw * 4, AMDGPU_IB_POOL_DIRECT, + &job); if (r) return r; @@ -538,7 +540,9 @@ static int amdgpu_vce_get_destroy_msg(struct amdgpu_ring *ring, uint32_t handle, struct dma_fence *f = NULL; int i, r; - r = amdgpu_job_alloc_with_ib(ring->adev, ib_size_dw * 4, + r = amdgpu_job_alloc_with_ib(ring->adev, &ring->adev->vce.entity, + AMDGPU_FENCE_OWNER_UNDEFINED, + ib_size_dw * 4, direct ? AMDGPU_IB_POOL_DIRECT : AMDGPU_IB_POOL_DELAYED, &job); if (r) @@ -570,8 +574,7 @@ static int amdgpu_vce_get_destroy_msg(struct amdgpu_ring *ring, uint32_t handle, if (direct) r = amdgpu_job_submit_direct(job, ring, &f); else - r = amdgpu_job_submit(job, &ring->adev->vce.entity, - AMDGPU_FENCE_OWNER_UNDEFINED, &f); + f = amdgpu_job_submit(job); if (r) goto err; diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_vcn.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_vcn.c index ce64ca1c6e66..b1622ac9949f 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_vcn.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_vcn.c @@ -603,15 +603,16 @@ static int amdgpu_vcn_dec_send_msg(struct amdgpu_ring *ring, struct amdgpu_ib *ib_msg, struct dma_fence **fence) { + u64 addr = AMDGPU_GPU_PAGE_ALIGN(ib_msg->gpu_addr); struct amdgpu_device *adev = ring->adev; struct dma_fence *f = NULL; struct amdgpu_job *job; struct amdgpu_ib *ib; - uint64_t addr = AMDGPU_GPU_PAGE_ALIGN(ib_msg->gpu_addr); int i, r; - r = amdgpu_job_alloc_with_ib(adev, 64, - AMDGPU_IB_POOL_DIRECT, &job); + r = amdgpu_job_alloc_with_ib(ring->adev, NULL, NULL, + 64, AMDGPU_IB_POOL_DIRECT, + &job); if (r) goto err; @@ -790,8 +791,9 @@ static int amdgpu_vcn_dec_sw_send_msg(struct amdgpu_ring *ring, if (sq) ib_size_dw += 8; - r = amdgpu_job_alloc_with_ib(adev, ib_size_dw * 4, - AMDGPU_IB_POOL_DIRECT, &job); + r = amdgpu_job_alloc_with_ib(ring->adev, NULL, NULL, + ib_size_dw * 4, AMDGPU_IB_POOL_DIRECT, + &job); if (r) goto err; @@ -919,8 +921,9 @@ static int amdgpu_vcn_enc_get_create_msg(struct amdgpu_ring *ring, uint32_t hand if (sq) ib_size_dw += 8; - r = amdgpu_job_alloc_with_ib(ring->adev, ib_size_dw * 4, - AMDGPU_IB_POOL_DIRECT, &job); + r = amdgpu_job_alloc_with_ib(ring->adev, NULL, NULL, + ib_size_dw * 4, AMDGPU_IB_POOL_DIRECT, + &job); if (r) return r; @@ -985,8 +988,9 @@ static int amdgpu_vcn_enc_get_destroy_msg(struct amdgpu_ring *ring, uint32_t han if (sq) ib_size_dw += 8; - r = amdgpu_job_alloc_with_ib(ring->adev, ib_size_dw * 4, - AMDGPU_IB_POOL_DIRECT, &job); + r = amdgpu_job_alloc_with_ib(ring->adev, NULL, NULL, + ib_size_dw * 4, AMDGPU_IB_POOL_DIRECT, + &job); if (r) return r; @@ -1251,3 +1255,20 @@ int amdgpu_vcn_process_poison_irq(struct amdgpu_device *adev, return 0; } + +void amdgpu_vcn_set_ras_funcs(struct amdgpu_device *adev) +{ + if (!adev->vcn.ras) + return; + + amdgpu_ras_register_ras_block(adev, &adev->vcn.ras->ras_block); + + strcpy(adev->vcn.ras->ras_block.ras_comm.name, "vcn"); + adev->vcn.ras->ras_block.ras_comm.block = AMDGPU_RAS_BLOCK__VCN; + adev->vcn.ras->ras_block.ras_comm.type = AMDGPU_RAS_ERROR__POISON; + adev->vcn.ras_if = &adev->vcn.ras->ras_block.ras_comm; + + /* If don't define special ras_late_init function, use default ras_late_init */ + if (!adev->vcn.ras->ras_block.ras_late_init) + adev->vcn.ras->ras_block.ras_late_init = amdgpu_ras_block_late_init; +} diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_vcn.h b/drivers/gpu/drm/amd/amdgpu/amdgpu_vcn.h index 253ea6b159df..dbb8d68a30c6 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_vcn.h +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_vcn.h @@ -399,5 +399,6 @@ void amdgpu_debugfs_vcn_fwlog_init(struct amdgpu_device *adev, int amdgpu_vcn_process_poison_irq(struct amdgpu_device *adev, struct amdgpu_irq_src *source, struct amdgpu_iv_entry *entry); +void amdgpu_vcn_set_ras_funcs(struct amdgpu_device *adev); #endif diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_virt.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_virt.c index c73abe54d974..2994b9db196f 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_virt.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_virt.c @@ -64,6 +64,10 @@ void amdgpu_virt_init_setting(struct amdgpu_device *adev) ddev->driver_features &= ~DRIVER_ATOMIC; adev->cg_flags = 0; adev->pg_flags = 0; + + /* enable mcbp for sriov asic_type before soc21 */ + amdgpu_mcbp = (adev->asic_type < CHIP_IP_DISCOVERY) ? 1 : 0; + } void amdgpu_virt_kiq_reg_write_reg_wait(struct amdgpu_device *adev, @@ -391,7 +395,6 @@ static void amdgpu_virt_ras_reserve_bps(struct amdgpu_device *adev) */ if (amdgpu_bo_create_kernel_at(adev, bp << AMDGPU_GPU_PAGE_SHIFT, AMDGPU_GPU_PAGE_SIZE, - AMDGPU_GEM_DOMAIN_VRAM, &bo, NULL)) DRM_DEBUG("RAS WARN: reserve vram for retired page %llx fail\n", bp); @@ -424,11 +427,17 @@ static void amdgpu_virt_add_bad_page(struct amdgpu_device *adev, struct eeprom_table_record bp; uint64_t retired_page; uint32_t bp_idx, bp_cnt; + void *vram_usage_va = NULL; + + if (adev->mman.fw_vram_usage_va) + vram_usage_va = adev->mman.fw_vram_usage_va; + else + vram_usage_va = adev->mman.drv_vram_usage_va; if (bp_block_size) { bp_cnt = bp_block_size / sizeof(uint64_t); for (bp_idx = 0; bp_idx < bp_cnt; bp_idx++) { - retired_page = *(uint64_t *)(adev->mman.fw_vram_usage_va + + retired_page = *(uint64_t *)(vram_usage_va + bp_block_offset + bp_idx * sizeof(uint64_t)); bp.retired_page = retired_page; @@ -639,7 +648,9 @@ void amdgpu_virt_init_data_exchange(struct amdgpu_device *adev) adev->virt.fw_reserve.p_vf2pf = NULL; adev->virt.vf2pf_update_interval_ms = 0; - if (adev->mman.fw_vram_usage_va != NULL) { + if (adev->mman.fw_vram_usage_va && adev->mman.drv_vram_usage_va) { + DRM_WARN("Currently fw_vram and drv_vram should not have values at the same time!"); + } else if (adev->mman.fw_vram_usage_va || adev->mman.drv_vram_usage_va) { /* go through this logic in ip_init and reset to init workqueue*/ amdgpu_virt_exchange_data(adev); @@ -662,32 +673,40 @@ void amdgpu_virt_exchange_data(struct amdgpu_device *adev) uint32_t bp_block_size = 0; struct amd_sriov_msg_pf2vf_info *pf2vf_v2 = NULL; - if (adev->mman.fw_vram_usage_va != NULL) { - - adev->virt.fw_reserve.p_pf2vf = - (struct amd_sriov_msg_pf2vf_info_header *) - (adev->mman.fw_vram_usage_va + (AMD_SRIOV_MSG_PF2VF_OFFSET_KB << 10)); - adev->virt.fw_reserve.p_vf2pf = - (struct amd_sriov_msg_vf2pf_info_header *) - (adev->mman.fw_vram_usage_va + (AMD_SRIOV_MSG_VF2PF_OFFSET_KB << 10)); + if (adev->mman.fw_vram_usage_va || adev->mman.drv_vram_usage_va) { + if (adev->mman.fw_vram_usage_va) { + adev->virt.fw_reserve.p_pf2vf = + (struct amd_sriov_msg_pf2vf_info_header *) + (adev->mman.fw_vram_usage_va + (AMD_SRIOV_MSG_PF2VF_OFFSET_KB << 10)); + adev->virt.fw_reserve.p_vf2pf = + (struct amd_sriov_msg_vf2pf_info_header *) + (adev->mman.fw_vram_usage_va + (AMD_SRIOV_MSG_VF2PF_OFFSET_KB << 10)); + } else if (adev->mman.drv_vram_usage_va) { + adev->virt.fw_reserve.p_pf2vf = + (struct amd_sriov_msg_pf2vf_info_header *) + (adev->mman.drv_vram_usage_va + (AMD_SRIOV_MSG_PF2VF_OFFSET_KB << 10)); + adev->virt.fw_reserve.p_vf2pf = + (struct amd_sriov_msg_vf2pf_info_header *) + (adev->mman.drv_vram_usage_va + (AMD_SRIOV_MSG_VF2PF_OFFSET_KB << 10)); + } amdgpu_virt_read_pf2vf_data(adev); amdgpu_virt_write_vf2pf_data(adev); /* bad page handling for version 2 */ if (adev->virt.fw_reserve.p_pf2vf->version == 2) { - pf2vf_v2 = (struct amd_sriov_msg_pf2vf_info *)adev->virt.fw_reserve.p_pf2vf; + pf2vf_v2 = (struct amd_sriov_msg_pf2vf_info *)adev->virt.fw_reserve.p_pf2vf; - bp_block_offset = ((uint64_t)pf2vf_v2->bp_block_offset_low & 0xFFFFFFFF) | - ((((uint64_t)pf2vf_v2->bp_block_offset_high) << 32) & 0xFFFFFFFF00000000); - bp_block_size = pf2vf_v2->bp_block_size; + bp_block_offset = ((uint64_t)pf2vf_v2->bp_block_offset_low & 0xFFFFFFFF) | + ((((uint64_t)pf2vf_v2->bp_block_offset_high) << 32) & 0xFFFFFFFF00000000); + bp_block_size = pf2vf_v2->bp_block_size; - if (bp_block_size && !adev->virt.ras_init_done) - amdgpu_virt_init_ras_err_handler_data(adev); + if (bp_block_size && !adev->virt.ras_init_done) + amdgpu_virt_init_ras_err_handler_data(adev); - if (adev->virt.ras_init_done) - amdgpu_virt_add_bad_page(adev, bp_block_offset, bp_block_size); - } + if (adev->virt.ras_init_done) + amdgpu_virt_add_bad_page(adev, bp_block_offset, bp_block_size); + } } } diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_virt.h b/drivers/gpu/drm/amd/amdgpu/amdgpu_virt.h index 49c4347d154c..2b9d806e23af 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_virt.h +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_virt.h @@ -75,6 +75,8 @@ struct amdgpu_vf_error_buffer { uint64_t data[AMDGPU_VF_ERROR_ENTRY_SIZE]; }; +enum idh_request; + /** * struct amdgpu_virt_ops - amdgpu device virt operations */ @@ -84,7 +86,8 @@ struct amdgpu_virt_ops { int (*req_init_data)(struct amdgpu_device *adev); int (*reset_gpu)(struct amdgpu_device *adev); int (*wait_reset)(struct amdgpu_device *adev); - void (*trans_msg)(struct amdgpu_device *adev, u32 req, u32 data1, u32 data2, u32 data3); + void (*trans_msg)(struct amdgpu_device *adev, enum idh_request req, + u32 data1, u32 data2, u32 data3); }; /* diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_vkms.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_vkms.c index 500a1dc4fe02..53ff91fc6cf6 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_vkms.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_vkms.c @@ -498,7 +498,7 @@ static int amdgpu_vkms_sw_init(void *handle) adev_to_drm(adev)->mode_config.preferred_depth = 24; adev_to_drm(adev)->mode_config.prefer_shadow = 1; - adev_to_drm(adev)->mode_config.fb_base = adev->gmc.aper_base; + adev_to_drm(adev)->mode_config.fb_modifiers_not_supported = true; adev_to_drm(adev)->mode_config.fb_modifiers_not_supported = true; @@ -513,6 +513,10 @@ static int amdgpu_vkms_sw_init(void *handle) return r; } + r = drm_vblank_init(adev_to_drm(adev), adev->mode_info.num_crtc); + if (r) + return r; + drm_kms_helper_poll_init(adev_to_drm(adev)); adev->mode_info.mode_config_initialized = true; diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_vm.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_vm.c index 003aa9e47085..dc379dc22c77 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_vm.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_vm.c @@ -45,22 +45,43 @@ /** * DOC: GPUVM * - * GPUVM is similar to the legacy gart on older asics, however - * rather than there being a single global gart table - * for the entire GPU, there are multiple VM page tables active - * at any given time. The VM page tables can contain a mix - * vram pages and system memory pages and system memory pages + * GPUVM is the MMU functionality provided on the GPU. + * GPUVM is similar to the legacy GART on older asics, however + * rather than there being a single global GART table + * for the entire GPU, there can be multiple GPUVM page tables active + * at any given time. The GPUVM page tables can contain a mix + * VRAM pages and system pages (both memory and MMIO) and system pages * can be mapped as snooped (cached system pages) or unsnooped * (uncached system pages). - * Each VM has an ID associated with it and there is a page table - * associated with each VMID. When executing a command buffer, - * the kernel tells the ring what VMID to use for that command + * + * Each active GPUVM has an ID associated with it and there is a page table + * linked with each VMID. When executing a command buffer, + * the kernel tells the engine what VMID to use for that command * buffer. VMIDs are allocated dynamically as commands are submitted. * The userspace drivers maintain their own address space and the kernel * sets up their pages tables accordingly when they submit their * command buffers and a VMID is assigned. - * Cayman/Trinity support up to 8 active VMs at any given time; - * SI supports 16. + * The hardware supports up to 16 active GPUVMs at any given time. + * + * Each GPUVM is represented by a 1-2 or 1-5 level page table, depending + * on the ASIC family. GPUVM supports RWX attributes on each page as well + * as other features such as encryption and caching attributes. + * + * VMID 0 is special. It is the GPUVM used for the kernel driver. In + * addition to an aperture managed by a page table, VMID 0 also has + * several other apertures. There is an aperture for direct access to VRAM + * and there is a legacy AGP aperture which just forwards accesses directly + * to the matching system physical addresses (or IOVAs when an IOMMU is + * present). These apertures provide direct access to these memories without + * incurring the overhead of a page table. VMID 0 is used by the kernel + * driver for tasks like memory management. + * + * GPU clients (i.e., engines on the GPU) use GPUVM VMIDs to access memory. + * For user applications, each application can have their own unique GPUVM + * address space. The application manages the address space and the kernel + * driver manages the GPUVM page tables for each process. If an GPU client + * accesses an invalid page, it will generate a GPU page fault, similar to + * accessing an invalid page on a CPU. */ #define START(node) ((node)->start) @@ -463,25 +484,20 @@ bool amdgpu_vm_need_pipeline_sync(struct amdgpu_ring *ring, struct amdgpu_device *adev = ring->adev; unsigned vmhub = ring->funcs->vmhub; struct amdgpu_vmid_mgr *id_mgr = &adev->vm_manager.id_mgr[vmhub]; - struct amdgpu_vmid *id; - bool gds_switch_needed; - bool vm_flush_needed = job->vm_needs_flush || ring->has_compute_vm_bug; if (job->vmid == 0) return false; - id = &id_mgr->ids[job->vmid]; - gds_switch_needed = ring->funcs->emit_gds_switch && ( - id->gds_base != job->gds_base || - id->gds_size != job->gds_size || - id->gws_base != job->gws_base || - id->gws_size != job->gws_size || - id->oa_base != job->oa_base || - id->oa_size != job->oa_size); - - if (amdgpu_vmid_had_gpu_reset(adev, id)) + + if (job->vm_needs_flush || ring->has_compute_vm_bug) return true; - return vm_flush_needed || gds_switch_needed; + if (ring->funcs->emit_gds_switch && job->gds_switch_needed) + return true; + + if (amdgpu_vmid_had_gpu_reset(adev, &id_mgr->ids[job->vmid])) + return true; + + return false; } /** @@ -503,27 +519,20 @@ int amdgpu_vm_flush(struct amdgpu_ring *ring, struct amdgpu_job *job, unsigned vmhub = ring->funcs->vmhub; struct amdgpu_vmid_mgr *id_mgr = &adev->vm_manager.id_mgr[vmhub]; struct amdgpu_vmid *id = &id_mgr->ids[job->vmid]; - bool gds_switch_needed = ring->funcs->emit_gds_switch && ( - id->gds_base != job->gds_base || - id->gds_size != job->gds_size || - id->gws_base != job->gws_base || - id->gws_size != job->gws_size || - id->oa_base != job->oa_base || - id->oa_size != job->oa_size); + bool spm_update_needed = job->spm_update_needed; + bool gds_switch_needed = ring->funcs->emit_gds_switch && + job->gds_switch_needed; bool vm_flush_needed = job->vm_needs_flush; struct dma_fence *fence = NULL; bool pasid_mapping_needed = false; unsigned patch_offset = 0; - bool update_spm_vmid_needed = (job->vm && (job->vm->reserved_vmid[vmhub] != NULL)); int r; - if (update_spm_vmid_needed && adev->gfx.rlc.funcs->update_spm_vmid) - adev->gfx.rlc.funcs->update_spm_vmid(adev, job->vmid); - if (amdgpu_vmid_had_gpu_reset(adev, id)) { gds_switch_needed = true; vm_flush_needed = true; pasid_mapping_needed = true; + spm_update_needed = true; } mutex_lock(&id_mgr->lock); @@ -541,6 +550,7 @@ int amdgpu_vm_flush(struct amdgpu_ring *ring, struct amdgpu_job *job, if (!vm_flush_needed && !gds_switch_needed && !need_pipe_sync) return 0; + amdgpu_ring_ib_begin(ring); if (ring->funcs->init_cond_exec) patch_offset = amdgpu_ring_init_cond_exec(ring); @@ -555,6 +565,17 @@ int amdgpu_vm_flush(struct amdgpu_ring *ring, struct amdgpu_job *job, if (pasid_mapping_needed) amdgpu_gmc_emit_pasid_mapping(ring, job->vmid, job->pasid); + if (spm_update_needed && adev->gfx.rlc.funcs->update_spm_vmid) + adev->gfx.rlc.funcs->update_spm_vmid(adev, job->vmid); + + if (!ring->is_mes_queue && ring->funcs->emit_gds_switch && + gds_switch_needed) { + amdgpu_ring_emit_gds_switch(ring, job->vmid, job->gds_base, + job->gds_size, job->gws_base, + job->gws_size, job->oa_base, + job->oa_size); + } + if (vm_flush_needed || pasid_mapping_needed) { r = amdgpu_fence_emit(ring, &fence, NULL, 0); if (r) @@ -579,20 +600,6 @@ int amdgpu_vm_flush(struct amdgpu_ring *ring, struct amdgpu_job *job, } dma_fence_put(fence); - if (!ring->is_mes_queue && ring->funcs->emit_gds_switch && - gds_switch_needed) { - id->gds_base = job->gds_base; - id->gds_size = job->gds_size; - id->gws_base = job->gws_base; - id->gws_size = job->gws_size; - id->oa_base = job->oa_base; - id->oa_size = job->oa_size; - amdgpu_ring_emit_gds_switch(ring, job->vmid, job->gds_base, - job->gds_size, job->gws_base, - job->gws_size, job->oa_base, - job->oa_size); - } - if (ring->funcs->patch_cond_exec) amdgpu_ring_patch_cond_exec(ring, patch_offset); @@ -601,6 +608,7 @@ int amdgpu_vm_flush(struct amdgpu_ring *ring, struct amdgpu_job *job, amdgpu_ring_emit_switch_buffer(ring); amdgpu_ring_emit_switch_buffer(ring); } + amdgpu_ring_ib_end(ring); return 0; } @@ -2360,7 +2368,6 @@ int amdgpu_vm_ioctl(struct drm_device *dev, void *data, struct drm_file *filp) union drm_amdgpu_vm *args = data; struct amdgpu_device *adev = drm_to_adev(dev); struct amdgpu_fpriv *fpriv = filp->driver_priv; - long timeout = msecs_to_jiffies(2000); int r; switch (args->in.op) { @@ -2372,21 +2379,6 @@ int amdgpu_vm_ioctl(struct drm_device *dev, void *data, struct drm_file *filp) return r; break; case AMDGPU_VM_OP_UNRESERVE_VMID: - if (amdgpu_sriov_runtime(adev)) - timeout = 8 * timeout; - - /* Wait vm idle to make sure the vmid set in SPM_VMID is - * not referenced anymore. - */ - r = amdgpu_bo_reserve(fpriv->vm.root.bo, true); - if (r) - return r; - - r = amdgpu_vm_wait_idle(&fpriv->vm, timeout); - if (r < 0) - return r; - - amdgpu_bo_unreserve(fpriv->vm.root.bo); amdgpu_vmid_free_reserved(adev, &fpriv->vm, AMDGPU_GFXHUB_0); break; default: diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_vm.h b/drivers/gpu/drm/amd/amdgpu/amdgpu_vm.h index 6546e786bf00..094bb4807303 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_vm.h +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_vm.h @@ -119,9 +119,6 @@ struct amdgpu_bo_vm; /* Reserve 2MB at top/bottom of address space for kernel use */ #define AMDGPU_VA_RESERVED_SIZE (2ULL << 20) -/* max vmids dedicated for process */ -#define AMDGPU_VM_MAX_RESERVED_VMID 1 - /* See vm_update_mode */ #define AMDGPU_VM_USE_CPU_FOR_GFX (1 << 0) #define AMDGPU_VM_USE_CPU_FOR_COMPUTE (1 << 1) @@ -298,8 +295,7 @@ struct amdgpu_vm { struct dma_fence *last_unlocked; unsigned int pasid; - /* dedicated to vm */ - struct amdgpu_vmid *reserved_vmid[AMDGPU_MAX_VMHUBS]; + bool reserved_vmid[AMDGPU_MAX_VMHUBS]; /* Flag to indicate if VM tables are updated by CPU or GPU (SDMA) */ bool use_cpu_for_update; diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_vm_sdma.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_vm_sdma.c index 69e105fa41f6..535cd6569bcc 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_vm_sdma.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_vm_sdma.c @@ -47,6 +47,32 @@ static int amdgpu_vm_sdma_map_table(struct amdgpu_bo_vm *table) return r; } +/* Allocate a new job for @count PTE updates */ +static int amdgpu_vm_sdma_alloc_job(struct amdgpu_vm_update_params *p, + unsigned int count) +{ + enum amdgpu_ib_pool_type pool = p->immediate ? AMDGPU_IB_POOL_IMMEDIATE + : AMDGPU_IB_POOL_DELAYED; + struct drm_sched_entity *entity = p->immediate ? &p->vm->immediate + : &p->vm->delayed; + unsigned int ndw; + int r; + + /* estimate how many dw we need */ + ndw = AMDGPU_VM_SDMA_MIN_NUM_DW; + if (p->pages_addr) + ndw += count * 2; + ndw = min(ndw, AMDGPU_VM_SDMA_MAX_NUM_DW); + + r = amdgpu_job_alloc_with_ib(p->adev, entity, AMDGPU_FENCE_OWNER_VM, + ndw * 4, pool, &p->job); + if (r) + return r; + + p->num_dw_left = ndw; + return 0; +} + /** * amdgpu_vm_sdma_prepare - prepare SDMA command submission * @@ -61,21 +87,22 @@ static int amdgpu_vm_sdma_prepare(struct amdgpu_vm_update_params *p, struct dma_resv *resv, enum amdgpu_sync_mode sync_mode) { - enum amdgpu_ib_pool_type pool = p->immediate ? AMDGPU_IB_POOL_IMMEDIATE - : AMDGPU_IB_POOL_DELAYED; - unsigned int ndw = AMDGPU_VM_SDMA_MIN_NUM_DW; + struct amdgpu_sync sync; int r; - r = amdgpu_job_alloc_with_ib(p->adev, ndw * 4, pool, &p->job); + r = amdgpu_vm_sdma_alloc_job(p, 0); if (r) return r; - p->num_dw_left = ndw; - if (!resv) return 0; - return amdgpu_sync_resv(p->adev, &p->job->sync, resv, sync_mode, p->vm); + amdgpu_sync_create(&sync); + r = amdgpu_sync_resv(p->adev, &sync, resv, sync_mode, p->vm); + if (!r) + r = amdgpu_sync_push_to_job(&sync, p->job); + amdgpu_sync_free(&sync); + return r; } /** @@ -91,20 +118,16 @@ static int amdgpu_vm_sdma_commit(struct amdgpu_vm_update_params *p, struct dma_fence **fence) { struct amdgpu_ib *ib = p->job->ibs; - struct drm_sched_entity *entity; struct amdgpu_ring *ring; struct dma_fence *f; - int r; - entity = p->immediate ? &p->vm->immediate : &p->vm->delayed; - ring = container_of(entity->rq->sched, struct amdgpu_ring, sched); + ring = container_of(p->vm->delayed.rq->sched, struct amdgpu_ring, + sched); WARN_ON(ib->length_dw == 0); amdgpu_ring_pad_ib(ring, ib); WARN_ON(ib->length_dw > p->num_dw_left); - r = amdgpu_job_submit(p->job, entity, AMDGPU_FENCE_OWNER_VM, &f); - if (r) - goto error; + f = amdgpu_job_submit(p->job); if (p->unlocked) { struct dma_fence *tmp = dma_fence_get(f); @@ -127,10 +150,6 @@ static int amdgpu_vm_sdma_commit(struct amdgpu_vm_update_params *p, } dma_fence_put(f); return 0; - -error: - amdgpu_job_free(p->job); - return r; } /** @@ -210,8 +229,6 @@ static int amdgpu_vm_sdma_update(struct amdgpu_vm_update_params *p, uint64_t flags) { struct amdgpu_bo *bo = &vmbo->bo; - enum amdgpu_ib_pool_type pool = p->immediate ? AMDGPU_IB_POOL_IMMEDIATE - : AMDGPU_IB_POOL_DELAYED; struct dma_resv_iter cursor; unsigned int i, ndw, nptes; struct dma_fence *fence; @@ -221,8 +238,10 @@ static int amdgpu_vm_sdma_update(struct amdgpu_vm_update_params *p, /* Wait for PD/PT moves to be completed */ dma_resv_iter_begin(&cursor, bo->tbo.base.resv, DMA_RESV_USAGE_KERNEL); dma_resv_for_each_fence_unlocked(&cursor, fence) { - r = amdgpu_sync_fence(&p->job->sync, fence); + dma_fence_get(fence); + r = drm_sched_job_add_dependency(&p->job->base, fence); if (r) { + dma_fence_put(fence); dma_resv_iter_end(&cursor); return r; } @@ -238,19 +257,9 @@ static int amdgpu_vm_sdma_update(struct amdgpu_vm_update_params *p, if (r) return r; - /* estimate how many dw we need */ - ndw = 32; - if (p->pages_addr) - ndw += count * 2; - ndw = max(ndw, AMDGPU_VM_SDMA_MIN_NUM_DW); - ndw = min(ndw, AMDGPU_VM_SDMA_MAX_NUM_DW); - - r = amdgpu_job_alloc_with_ib(p->adev, ndw * 4, pool, - &p->job); + r = amdgpu_vm_sdma_alloc_job(p, count); if (r) return r; - - p->num_dw_left = ndw; } if (!p->pages_addr) { diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_vram_mgr.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_vram_mgr.c index 80dd1343594c..9fa1d814508a 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_vram_mgr.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_vram_mgr.c @@ -439,7 +439,7 @@ static int amdgpu_vram_mgr_new(struct ttm_resource_manager *man, /* Allocate blocks in desired range */ vres->flags |= DRM_BUDDY_RANGE_ALLOCATION; - remaining_size = (u64)vres->base.num_pages << PAGE_SHIFT; + remaining_size = (u64)vres->base.size; mutex_lock(&mgr->lock); while (remaining_size) { @@ -498,7 +498,7 @@ static int amdgpu_vram_mgr_new(struct ttm_resource_manager *man, LIST_HEAD(temp); trim_list = &vres->blocks; - original_size = (u64)vres->base.num_pages << PAGE_SHIFT; + original_size = (u64)vres->base.size; /* * If size value is rounded up to min_block_size, trim the last @@ -533,8 +533,8 @@ static int amdgpu_vram_mgr_new(struct ttm_resource_manager *man, amdgpu_vram_mgr_block_size(block); start >>= PAGE_SHIFT; - if (start > vres->base.num_pages) - start -= vres->base.num_pages; + if (start > PFN_UP(vres->base.size)) + start -= PFN_UP(vres->base.size); else start = 0; vres->base.start = max(vres->base.start, start); @@ -882,7 +882,7 @@ void amdgpu_vram_mgr_fini(struct amdgpu_device *adev) kfree(rsv); list_for_each_entry_safe(rsv, temp, &mgr->reserved_pages, blocks) { - drm_buddy_free_list(&mgr->mm, &rsv->blocks); + drm_buddy_free_list(&mgr->mm, &rsv->allocated); kfree(rsv); } drm_buddy_fini(&mgr->mm); diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_xgmi.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_xgmi.c index 47159e9a0884..4b9e7b050ccd 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_xgmi.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_xgmi.c @@ -386,7 +386,6 @@ struct amdgpu_hive_info *amdgpu_get_xgmi_hive(struct amdgpu_device *adev) if (ret) { dev_err(adev->dev, "XGMI: failed initializing kobject for xgmi hive\n"); kobject_put(&hive->kobj); - kfree(hive); hive = NULL; goto pro_end; } @@ -410,7 +409,6 @@ struct amdgpu_hive_info *amdgpu_get_xgmi_hive(struct amdgpu_device *adev) dev_err(adev->dev, "XGMI: failed initializing reset domain for xgmi hive\n"); ret = -ENOMEM; kobject_put(&hive->kobj); - kfree(hive); hive = NULL; goto pro_end; } diff --git a/drivers/gpu/drm/amd/amdgpu/atombios_encoders.c b/drivers/gpu/drm/amd/amdgpu/atombios_encoders.c index 6be9ac2b9c5b..18ae9433e463 100644 --- a/drivers/gpu/drm/amd/amdgpu/atombios_encoders.c +++ b/drivers/gpu/drm/amd/amdgpu/atombios_encoders.c @@ -2081,8 +2081,11 @@ amdgpu_atombios_encoder_get_lcd_info(struct amdgpu_encoder *encoder) } } record += fake_edid_record->ucFakeEDIDLength ? - fake_edid_record->ucFakeEDIDLength + 2 : - sizeof(ATOM_FAKE_EDID_PATCH_RECORD); + struct_size(fake_edid_record, + ucFakeEDIDString, + fake_edid_record->ucFakeEDIDLength) : + /* empty fake edid record must be 3 bytes long */ + sizeof(ATOM_FAKE_EDID_PATCH_RECORD) + 1; break; case LCD_PANEL_RESOLUTION_RECORD_TYPE: panel_res_record = (ATOM_PANEL_RESOLUTION_PATCH_RECORD *)record; diff --git a/drivers/gpu/drm/amd/amdgpu/dce_v10_0.c b/drivers/gpu/drm/amd/amdgpu/dce_v10_0.c index 288fce7dc0ed..248f1a4e915f 100644 --- a/drivers/gpu/drm/amd/amdgpu/dce_v10_0.c +++ b/drivers/gpu/drm/amd/amdgpu/dce_v10_0.c @@ -21,6 +21,7 @@ * */ +#include <drm/drm_fb_helper.h> #include <drm/drm_fourcc.h> #include <drm/drm_vblank.h> @@ -2800,8 +2801,6 @@ static int dce_v10_0_sw_init(void *handle) adev_to_drm(adev)->mode_config.fb_modifiers_not_supported = true; - adev_to_drm(adev)->mode_config.fb_base = adev->gmc.aper_base; - r = amdgpu_display_modeset_create_props(adev); if (r) return r; @@ -2830,6 +2829,17 @@ static int dce_v10_0_sw_init(void *handle) if (r) return r; + /* Disable vblank IRQs aggressively for power-saving */ + /* XXX: can this be enabled for DC? */ + adev_to_drm(adev)->vblank_disable_immediate = true; + + r = drm_vblank_init(adev_to_drm(adev), adev->mode_info.num_crtc); + if (r) + return r; + + INIT_WORK(&adev->hotplug_work, + amdgpu_display_hotplug_work_func); + drm_kms_helper_poll_init(adev_to_drm(adev)); adev->mode_info.mode_config_initialized = true; @@ -2892,6 +2902,8 @@ static int dce_v10_0_hw_fini(void *handle) dce_v10_0_pageflip_interrupt_fini(adev); + flush_work(&adev->hotplug_work); + return 0; } diff --git a/drivers/gpu/drm/amd/amdgpu/dce_v11_0.c b/drivers/gpu/drm/amd/amdgpu/dce_v11_0.c index cbe5250b31cb..cd9c19060d89 100644 --- a/drivers/gpu/drm/amd/amdgpu/dce_v11_0.c +++ b/drivers/gpu/drm/amd/amdgpu/dce_v11_0.c @@ -21,6 +21,7 @@ * */ +#include <drm/drm_fb_helper.h> #include <drm/drm_fourcc.h> #include <drm/drm_vblank.h> @@ -2918,8 +2919,6 @@ static int dce_v11_0_sw_init(void *handle) adev_to_drm(adev)->mode_config.fb_modifiers_not_supported = true; - adev_to_drm(adev)->mode_config.fb_base = adev->gmc.aper_base; - r = amdgpu_display_modeset_create_props(adev); if (r) return r; @@ -2949,6 +2948,17 @@ static int dce_v11_0_sw_init(void *handle) if (r) return r; + /* Disable vblank IRQs aggressively for power-saving */ + /* XXX: can this be enabled for DC? */ + adev_to_drm(adev)->vblank_disable_immediate = true; + + r = drm_vblank_init(adev_to_drm(adev), adev->mode_info.num_crtc); + if (r) + return r; + + INIT_WORK(&adev->hotplug_work, + amdgpu_display_hotplug_work_func); + drm_kms_helper_poll_init(adev_to_drm(adev)); adev->mode_info.mode_config_initialized = true; @@ -3022,6 +3032,8 @@ static int dce_v11_0_hw_fini(void *handle) dce_v11_0_pageflip_interrupt_fini(adev); + flush_work(&adev->hotplug_work); + return 0; } diff --git a/drivers/gpu/drm/amd/amdgpu/dce_v6_0.c b/drivers/gpu/drm/amd/amdgpu/dce_v6_0.c index b1c44fab074f..76323deecc58 100644 --- a/drivers/gpu/drm/amd/amdgpu/dce_v6_0.c +++ b/drivers/gpu/drm/amd/amdgpu/dce_v6_0.c @@ -23,6 +23,7 @@ #include <linux/pci.h> +#include <drm/drm_fb_helper.h> #include <drm/drm_fourcc.h> #include <drm/drm_vblank.h> @@ -2675,7 +2676,6 @@ static int dce_v6_0_sw_init(void *handle) adev_to_drm(adev)->mode_config.preferred_depth = 24; adev_to_drm(adev)->mode_config.prefer_shadow = 1; adev_to_drm(adev)->mode_config.fb_modifiers_not_supported = true; - adev_to_drm(adev)->mode_config.fb_base = adev->gmc.aper_base; r = amdgpu_display_modeset_create_props(adev); if (r) @@ -2706,6 +2706,18 @@ static int dce_v6_0_sw_init(void *handle) if (r) return r; + /* Disable vblank IRQs aggressively for power-saving */ + /* XXX: can this be enabled for DC? */ + adev_to_drm(adev)->vblank_disable_immediate = true; + + r = drm_vblank_init(adev_to_drm(adev), adev->mode_info.num_crtc); + if (r) + return r; + + /* Pre-DCE11 */ + INIT_WORK(&adev->hotplug_work, + amdgpu_display_hotplug_work_func); + drm_kms_helper_poll_init(adev_to_drm(adev)); return r; @@ -2764,6 +2776,8 @@ static int dce_v6_0_hw_fini(void *handle) dce_v6_0_pageflip_interrupt_fini(adev); + flush_work(&adev->hotplug_work); + return 0; } diff --git a/drivers/gpu/drm/amd/amdgpu/dce_v8_0.c b/drivers/gpu/drm/amd/amdgpu/dce_v8_0.c index a22b45c92792..01cf3ab111cb 100644 --- a/drivers/gpu/drm/amd/amdgpu/dce_v8_0.c +++ b/drivers/gpu/drm/amd/amdgpu/dce_v8_0.c @@ -21,6 +21,7 @@ * */ +#include <drm/drm_fb_helper.h> #include <drm/drm_fourcc.h> #include <drm/drm_vblank.h> @@ -2701,8 +2702,6 @@ static int dce_v8_0_sw_init(void *handle) adev_to_drm(adev)->mode_config.fb_modifiers_not_supported = true; - adev_to_drm(adev)->mode_config.fb_base = adev->gmc.aper_base; - r = amdgpu_display_modeset_create_props(adev); if (r) return r; @@ -2731,6 +2730,18 @@ static int dce_v8_0_sw_init(void *handle) if (r) return r; + /* Disable vblank IRQs aggressively for power-saving */ + /* XXX: can this be enabled for DC? */ + adev_to_drm(adev)->vblank_disable_immediate = true; + + r = drm_vblank_init(adev_to_drm(adev), adev->mode_info.num_crtc); + if (r) + return r; + + /* Pre-DCE11 */ + INIT_WORK(&adev->hotplug_work, + amdgpu_display_hotplug_work_func); + drm_kms_helper_poll_init(adev_to_drm(adev)); adev->mode_info.mode_config_initialized = true; @@ -2791,6 +2802,8 @@ static int dce_v8_0_hw_fini(void *handle) dce_v8_0_pageflip_interrupt_fini(adev); + flush_work(&adev->hotplug_work); + return 0; } diff --git a/drivers/gpu/drm/amd/amdgpu/gfx_v10_0.c b/drivers/gpu/drm/amd/amdgpu/gfx_v10_0.c index af94ac580d3e..49d34c7bbf20 100644 --- a/drivers/gpu/drm/amd/amdgpu/gfx_v10_0.c +++ b/drivers/gpu/drm/amd/amdgpu/gfx_v10_0.c @@ -4453,8 +4453,6 @@ static void gfx_v10_0_gpu_early_init(struct amdgpu_device *adev) { u32 gb_addr_config; - adev->gfx.funcs = &gfx_v10_0_gfx_funcs; - switch (adev->ip_versions[GC_HWIP][0]) { case IP_VERSION(10, 1, 10): case IP_VERSION(10, 1, 1): @@ -6911,6 +6909,8 @@ static int gfx_v10_0_kiq_init_queue(struct amdgpu_ring *ring) mutex_unlock(&adev->srbm_mutex); } else { memset((void *)mqd, 0, sizeof(*mqd)); + if (amdgpu_sriov_vf(adev) && adev->in_suspend) + amdgpu_ring_clear_ring(ring); mutex_lock(&adev->srbm_mutex); nv_grbm_select(adev, ring->me, ring->pipe, ring->queue, 0); amdgpu_ring_init_mqd(ring); @@ -7593,6 +7593,8 @@ static int gfx_v10_0_early_init(void *handle) { struct amdgpu_device *adev = (struct amdgpu_device *)handle; + adev->gfx.funcs = &gfx_v10_0_gfx_funcs; + switch (adev->ip_versions[GC_HWIP][0]) { case IP_VERSION(10, 1, 10): case IP_VERSION(10, 1, 1): @@ -8489,7 +8491,7 @@ static void gfx_v10_0_ring_emit_ib_gfx(struct amdgpu_ring *ring, control |= ib->length_dw | (vmid << 24); - if ((amdgpu_sriov_vf(ring->adev) || amdgpu_mcbp) && (ib->flags & AMDGPU_IB_FLAG_PREEMPT)) { + if (amdgpu_mcbp && (ib->flags & AMDGPU_IB_FLAG_PREEMPT)) { control |= INDIRECT_BUFFER_PRE_ENB(1); if (flags & AMDGPU_IB_PREEMPTED) @@ -8664,7 +8666,7 @@ static void gfx_v10_0_ring_emit_cntxcntl(struct amdgpu_ring *ring, { uint32_t dw2 = 0; - if (amdgpu_mcbp || amdgpu_sriov_vf(ring->adev)) + if (amdgpu_mcbp) gfx_v10_0_ring_emit_ce_meta(ring, (!amdgpu_sriov_vf(ring->adev) && flags & AMDGPU_IB_PREEMPTED) ? true : false); diff --git a/drivers/gpu/drm/amd/amdgpu/gfx_v11_0.c b/drivers/gpu/drm/amd/amdgpu/gfx_v11_0.c index 0fecc5bf45bc..a56c6e106d00 100644 --- a/drivers/gpu/drm/amd/amdgpu/gfx_v11_0.c +++ b/drivers/gpu/drm/amd/amdgpu/gfx_v11_0.c @@ -77,6 +77,10 @@ MODULE_FIRMWARE("amdgpu/gc_11_0_3_pfp.bin"); MODULE_FIRMWARE("amdgpu/gc_11_0_3_me.bin"); MODULE_FIRMWARE("amdgpu/gc_11_0_3_mec.bin"); MODULE_FIRMWARE("amdgpu/gc_11_0_3_rlc.bin"); +MODULE_FIRMWARE("amdgpu/gc_11_0_4_pfp.bin"); +MODULE_FIRMWARE("amdgpu/gc_11_0_4_me.bin"); +MODULE_FIRMWARE("amdgpu/gc_11_0_4_mec.bin"); +MODULE_FIRMWARE("amdgpu/gc_11_0_4_rlc.bin"); static const struct soc15_reg_golden golden_settings_gc_11_0_1[] = { @@ -262,6 +266,7 @@ static void gfx_v11_0_init_golden_registers(struct amdgpu_device *adev) { switch (adev->ip_versions[GC_HWIP][0]) { case IP_VERSION(11, 0, 1): + case IP_VERSION(11, 0, 4): soc15_program_register_sequence(adev, golden_settings_gc_11_0_1, (const u32)ARRAY_SIZE(golden_settings_gc_11_0_1)); @@ -843,7 +848,6 @@ static const struct amdgpu_gfx_funcs gfx_v11_0_gfx_funcs = { static int gfx_v11_0_gpu_early_init(struct amdgpu_device *adev) { - adev->gfx.funcs = &gfx_v11_0_gfx_funcs; switch (adev->ip_versions[GC_HWIP][0]) { case IP_VERSION(11, 0, 0): @@ -856,6 +860,7 @@ static int gfx_v11_0_gpu_early_init(struct amdgpu_device *adev) adev->gfx.config.sc_earlyz_tile_fifo_size = 0x4C0; break; case IP_VERSION(11, 0, 1): + case IP_VERSION(11, 0, 4): adev->gfx.config.max_hw_contexts = 8; adev->gfx.config.sc_prim_fifo_size_frontend = 0x20; adev->gfx.config.sc_prim_fifo_size_backend = 0x100; @@ -1285,6 +1290,7 @@ static int gfx_v11_0_sw_init(void *handle) case IP_VERSION(11, 0, 1): case IP_VERSION(11, 0, 2): case IP_VERSION(11, 0, 3): + case IP_VERSION(11, 0, 4): adev->gfx.me.num_me = 1; adev->gfx.me.num_pipe_per_me = 1; adev->gfx.me.num_queue_per_pipe = 1; @@ -1626,7 +1632,8 @@ static void gfx_v11_0_constants_init(struct amdgpu_device *adev) u32 tmp; int i; - WREG32_FIELD15_PREREG(GC, 0, GRBM_CNTL, READ_TIMEOUT, 0xff); + if (!amdgpu_sriov_vf(adev)) + WREG32_FIELD15_PREREG(GC, 0, GRBM_CNTL, READ_TIMEOUT, 0xff); gfx_v11_0_setup_rb(adev); gfx_v11_0_get_cu_info(adev, &adev->gfx.cu_info); @@ -2486,7 +2493,8 @@ static int gfx_v11_0_wait_for_rlc_autoload_complete(struct amdgpu_device *adev) for (i = 0; i < adev->usec_timeout; i++) { cp_status = RREG32_SOC15(GC, 0, regCP_STAT); - if (adev->ip_versions[GC_HWIP][0] == IP_VERSION(11, 0, 1)) + if (adev->ip_versions[GC_HWIP][0] == IP_VERSION(11, 0, 1) || + adev->ip_versions[GC_HWIP][0] == IP_VERSION(11, 0, 4)) bootload_status = RREG32_SOC15(GC, 0, regRLC_RLCS_BOOTLOAD_STATUS_gc_11_0_1); else @@ -4004,6 +4012,8 @@ static int gfx_v11_0_kiq_init_queue(struct amdgpu_ring *ring) mutex_unlock(&adev->srbm_mutex); } else { memset((void *)mqd, 0, sizeof(*mqd)); + if (amdgpu_sriov_vf(adev) && adev->in_suspend) + amdgpu_ring_clear_ring(ring); mutex_lock(&adev->srbm_mutex); soc21_grbm_select(adev, ring->me, ring->pipe, ring->queue, 0); amdgpu_ring_init_mqd(ring); @@ -4390,7 +4400,6 @@ static int gfx_v11_0_hw_fini(void *handle) { struct amdgpu_device *adev = (struct amdgpu_device *)handle; int r; - uint32_t tmp; amdgpu_irq_put(adev, &adev->gfx.priv_reg_irq, 0); amdgpu_irq_put(adev, &adev->gfx.priv_inst_irq, 0); @@ -4409,15 +4418,14 @@ static int gfx_v11_0_hw_fini(void *handle) amdgpu_mes_kiq_hw_fini(adev); } - if (amdgpu_sriov_vf(adev)) { - gfx_v11_0_cp_gfx_enable(adev, false); - /* Program KIQ position of RLC_CP_SCHEDULERS during destroy */ - tmp = RREG32_SOC15(GC, 0, regRLC_CP_SCHEDULERS); - tmp &= 0xffffff00; - WREG32_SOC15(GC, 0, regRLC_CP_SCHEDULERS, tmp); - + if (amdgpu_sriov_vf(adev)) + /* Remove the steps disabling CPG and clearing KIQ position, + * so that CP could perform IDLE-SAVE during switch. Those + * steps are necessary to avoid a DMAR error in gfx9 but it is + * not reproduced on gfx11. + */ return 0; - } + gfx_v11_0_cp_enable(adev, false); gfx_v11_0_enable_gui_idle_interrupt(adev, false); @@ -4656,6 +4664,8 @@ static int gfx_v11_0_early_init(void *handle) { struct amdgpu_device *adev = (struct amdgpu_device *)handle; + adev->gfx.funcs = &gfx_v11_0_gfx_funcs; + adev->gfx.num_gfx_rings = GFX11_NUM_GFX_RINGS; adev->gfx.num_compute_rings = min(amdgpu_gfx_get_num_kcq(adev), AMDGPU_MAX_COMPUTE_RINGS); @@ -4673,6 +4683,26 @@ static int gfx_v11_0_early_init(void *handle) return 0; } +static int gfx_v11_0_ras_late_init(void *handle) +{ + struct amdgpu_device *adev = (struct amdgpu_device *)handle; + struct ras_common_if *gfx_common_if; + int ret; + + gfx_common_if = kzalloc(sizeof(struct ras_common_if), GFP_KERNEL); + if (!gfx_common_if) + return -ENOMEM; + + gfx_common_if->block = AMDGPU_RAS_BLOCK__GFX; + + ret = amdgpu_ras_feature_enable(adev, gfx_common_if, true); + if (ret) + dev_warn(adev->dev, "Failed to enable gfx11 ras feature\n"); + + kfree(gfx_common_if); + return 0; +} + static int gfx_v11_0_late_init(void *handle) { struct amdgpu_device *adev = (struct amdgpu_device *)handle; @@ -4686,6 +4716,12 @@ static int gfx_v11_0_late_init(void *handle) if (r) return r; + if (adev->ip_versions[GC_HWIP][0] == IP_VERSION(11, 0, 3)) { + r = gfx_v11_0_ras_late_init(handle); + if (r) + return r; + } + return 0; } @@ -5022,6 +5058,7 @@ static void gfx_v11_cntl_power_gating(struct amdgpu_device *adev, bool enable) if (enable && (adev->pg_flags & AMD_PG_SUPPORT_GFX_PG)) { switch (adev->ip_versions[GC_HWIP][0]) { case IP_VERSION(11, 0, 1): + case IP_VERSION(11, 0, 4): WREG32_SOC15(GC, 0, regRLC_PG_DELAY_3, RLC_PG_DELAY_3_DEFAULT_GC_11_0_1); break; default: @@ -5055,6 +5092,7 @@ static int gfx_v11_0_set_powergating_state(void *handle, amdgpu_gfx_off_ctrl(adev, enable); break; case IP_VERSION(11, 0, 1): + case IP_VERSION(11, 0, 4): gfx_v11_cntl_pg(adev, enable); amdgpu_gfx_off_ctrl(adev, enable); break; @@ -5078,6 +5116,7 @@ static int gfx_v11_0_set_clockgating_state(void *handle, case IP_VERSION(11, 0, 1): case IP_VERSION(11, 0, 2): case IP_VERSION(11, 0, 3): + case IP_VERSION(11, 0, 4): gfx_v11_0_update_gfx_clock_gating(adev, state == AMD_CG_STATE_GATE); break; @@ -5299,7 +5338,7 @@ static void gfx_v11_0_ring_emit_ib_gfx(struct amdgpu_ring *ring, control |= ib->length_dw | (vmid << 24); - if ((amdgpu_sriov_vf(ring->adev) || amdgpu_mcbp) && (ib->flags & AMDGPU_IB_FLAG_PREEMPT)) { + if (amdgpu_mcbp && (ib->flags & AMDGPU_IB_FLAG_PREEMPT)) { control |= INDIRECT_BUFFER_PRE_ENB(1); if (flags & AMDGPU_IB_PREEMPTED) @@ -6060,6 +6099,7 @@ static const struct amdgpu_ring_funcs gfx_v11_0_ring_funcs_gfx = { .align_mask = 0xff, .nop = PACKET3(PACKET3_NOP, 0x3FFF), .support_64bit_ptrs = true, + .secure_submission_supported = true, .vmhub = AMDGPU_GFXHUB_0, .get_rptr = gfx_v11_0_ring_get_rptr_gfx, .get_wptr = gfx_v11_0_ring_get_wptr_gfx, diff --git a/drivers/gpu/drm/amd/amdgpu/gfx_v8_0.c b/drivers/gpu/drm/amd/amdgpu/gfx_v8_0.c index 7f0b18b0d4c4..d47135606e3e 100644 --- a/drivers/gpu/drm/amd/amdgpu/gfx_v8_0.c +++ b/drivers/gpu/drm/amd/amdgpu/gfx_v8_0.c @@ -4643,6 +4643,8 @@ static int gfx_v8_0_kiq_init_queue(struct amdgpu_ring *ring) memset((void *)mqd, 0, sizeof(struct vi_mqd_allocation)); ((struct vi_mqd_allocation *)mqd)->dynamic_cu_mask = 0xFFFFFFFF; ((struct vi_mqd_allocation *)mqd)->dynamic_rb_mask = 0xFFFFFFFF; + if (amdgpu_sriov_vf(adev) && adev->in_suspend) + amdgpu_ring_clear_ring(ring); mutex_lock(&adev->srbm_mutex); vi_srbm_select(adev, ring->me, ring->pipe, ring->queue, 0); gfx_v8_0_mqd_init(ring); diff --git a/drivers/gpu/drm/amd/amdgpu/gfx_v9_0.c b/drivers/gpu/drm/amd/amdgpu/gfx_v9_0.c index 0320be4a5fc6..f202b45c413c 100644 --- a/drivers/gpu/drm/amd/amdgpu/gfx_v9_0.c +++ b/drivers/gpu/drm/amd/amdgpu/gfx_v9_0.c @@ -47,6 +47,7 @@ #include "amdgpu_ras.h" +#include "amdgpu_ring_mux.h" #include "gfx_v9_4.h" #include "gfx_v9_0.h" #include "gfx_v9_4_2.h" @@ -56,6 +57,7 @@ #include "asic_reg/gc/gc_9_0_default.h" #define GFX9_NUM_GFX_RINGS 1 +#define GFX9_NUM_SW_GFX_RINGS 2 #define GFX9_MEC_HPD_SIZE 4096 #define RLCG_UCODE_LOADING_START_ADDRESS 0x00002000L #define RLC_SAVE_RESTORE_ADDR_STARTING_OFFSET 0x00000000L @@ -753,7 +755,7 @@ static void gfx_v9_0_set_rlc_funcs(struct amdgpu_device *adev); static int gfx_v9_0_get_cu_info(struct amdgpu_device *adev, struct amdgpu_cu_info *cu_info); static uint64_t gfx_v9_0_get_gpu_clock_counter(struct amdgpu_device *adev); -static void gfx_v9_0_ring_emit_de_meta(struct amdgpu_ring *ring); +static void gfx_v9_0_ring_emit_de_meta(struct amdgpu_ring *ring, bool resume); static u64 gfx_v9_0_ring_get_rptr_compute(struct amdgpu_ring *ring); static void gfx_v9_0_query_ras_error_count(struct amdgpu_device *adev, void *ras_error_status); @@ -826,9 +828,10 @@ static void gfx_v9_0_kiq_unmap_queues(struct amdgpu_ring *kiq_ring, PACKET3_UNMAP_QUEUES_DOORBELL_OFFSET0(ring->doorbell_index)); if (action == PREEMPT_QUEUES_NO_UNMAP) { - amdgpu_ring_write(kiq_ring, lower_32_bits(gpu_addr)); - amdgpu_ring_write(kiq_ring, upper_32_bits(gpu_addr)); - amdgpu_ring_write(kiq_ring, seq); + amdgpu_ring_write(kiq_ring, lower_32_bits(ring->wptr & ring->buf_mask)); + amdgpu_ring_write(kiq_ring, 0); + amdgpu_ring_write(kiq_ring, 0); + } else { amdgpu_ring_write(kiq_ring, 0); amdgpu_ring_write(kiq_ring, 0); @@ -1564,7 +1567,7 @@ static void gfx_v9_0_init_always_on_cu_mask(struct amdgpu_device *adev) mask = 1; cu_bitmap = 0; counter = 0; - gfx_v9_0_select_se_sh(adev, i, j, 0xffffffff); + amdgpu_gfx_select_se_sh(adev, i, j, 0xffffffff); for (k = 0; k < adev->gfx.config.max_cu_per_sh; k ++) { if (cu_info->bitmap[i][j] & mask) { @@ -1583,7 +1586,7 @@ static void gfx_v9_0_init_always_on_cu_mask(struct amdgpu_device *adev) cu_info->ao_cu_bitmap[i][j] = cu_bitmap; } } - gfx_v9_0_select_se_sh(adev, 0xffffffff, 0xffffffff, 0xffffffff); + amdgpu_gfx_select_se_sh(adev, 0xffffffff, 0xffffffff, 0xffffffff); mutex_unlock(&adev->grbm_idx_mutex); } @@ -1605,7 +1608,7 @@ static void gfx_v9_0_init_lbpw(struct amdgpu_device *adev) mutex_lock(&adev->grbm_idx_mutex); /* set mmRLC_LB_INIT_CU_MASK thru broadcast mode to enable all SE/SH*/ - gfx_v9_0_select_se_sh(adev, 0xffffffff, 0xffffffff, 0xffffffff); + amdgpu_gfx_select_se_sh(adev, 0xffffffff, 0xffffffff, 0xffffffff); WREG32_SOC15(GC, 0, mmRLC_LB_INIT_CU_MASK, 0xffffffff); /* set mmRLC_LB_PARAMS = 0x003F_1006 */ @@ -1654,7 +1657,7 @@ static void gfx_v9_4_init_lbpw(struct amdgpu_device *adev) mutex_lock(&adev->grbm_idx_mutex); /* set mmRLC_LB_INIT_CU_MASK thru broadcast mode to enable all SE/SH*/ - gfx_v9_0_select_se_sh(adev, 0xffffffff, 0xffffffff, 0xffffffff); + amdgpu_gfx_select_se_sh(adev, 0xffffffff, 0xffffffff, 0xffffffff); WREG32_SOC15(GC, 0, mmRLC_LB_INIT_CU_MASK, 0xffffffff); /* set mmRLC_LB_PARAMS = 0x003F_1006 */ @@ -1919,8 +1922,6 @@ static int gfx_v9_0_gpu_early_init(struct amdgpu_device *adev) u32 gb_addr_config; int err; - adev->gfx.funcs = &gfx_v9_0_gfx_funcs; - switch (adev->ip_versions[GC_HWIP][0]) { case IP_VERSION(9, 0, 1): adev->gfx.config.max_hw_contexts = 8; @@ -2105,6 +2106,7 @@ static int gfx_v9_0_sw_init(void *handle) struct amdgpu_ring *ring; struct amdgpu_kiq *kiq; struct amdgpu_device *adev = (struct amdgpu_device *)handle; + unsigned int hw_prio; switch (adev->ip_versions[GC_HWIP][0]) { case IP_VERSION(9, 0, 1): @@ -2188,6 +2190,9 @@ static int gfx_v9_0_sw_init(void *handle) sprintf(ring->name, "gfx_%d", i); ring->use_doorbell = true; ring->doorbell_index = adev->doorbell_index.gfx_ring0 << 1; + + /* disable scheduler on the real ring */ + ring->no_scheduler = true; r = amdgpu_ring_init(adev, ring, 1024, &adev->gfx.eop_irq, AMDGPU_CP_IRQ_GFX_ME0_PIPE0_EOP, AMDGPU_RING_PRIO_DEFAULT, NULL); @@ -2195,6 +2200,41 @@ static int gfx_v9_0_sw_init(void *handle) return r; } + /* set up the software rings */ + if (adev->gfx.num_gfx_rings) { + for (i = 0; i < GFX9_NUM_SW_GFX_RINGS; i++) { + ring = &adev->gfx.sw_gfx_ring[i]; + ring->ring_obj = NULL; + sprintf(ring->name, amdgpu_sw_ring_name(i)); + ring->use_doorbell = true; + ring->doorbell_index = adev->doorbell_index.gfx_ring0 << 1; + ring->is_sw_ring = true; + hw_prio = amdgpu_sw_ring_priority(i); + r = amdgpu_ring_init(adev, ring, 1024, &adev->gfx.eop_irq, + AMDGPU_CP_IRQ_GFX_ME0_PIPE0_EOP, hw_prio, + NULL); + if (r) + return r; + ring->wptr = 0; + } + + /* init the muxer and add software rings */ + r = amdgpu_ring_mux_init(&adev->gfx.muxer, &adev->gfx.gfx_ring[0], + GFX9_NUM_SW_GFX_RINGS); + if (r) { + DRM_ERROR("amdgpu_ring_mux_init failed(%d)\n", r); + return r; + } + for (i = 0; i < GFX9_NUM_SW_GFX_RINGS; i++) { + r = amdgpu_ring_mux_add_sw_ring(&adev->gfx.muxer, + &adev->gfx.sw_gfx_ring[i]); + if (r) { + DRM_ERROR("amdgpu_ring_mux_add_sw_ring failed(%d)\n", r); + return r; + } + } + } + /* set up the compute queues - allocate horizontally across pipes */ ring_id = 0; for (i = 0; i < adev->gfx.mec.num_mec; ++i) { @@ -2245,6 +2285,12 @@ static int gfx_v9_0_sw_fini(void *handle) int i; struct amdgpu_device *adev = (struct amdgpu_device *)handle; + if (adev->gfx.num_gfx_rings) { + for (i = 0; i < GFX9_NUM_SW_GFX_RINGS; i++) + amdgpu_ring_fini(&adev->gfx.sw_gfx_ring[i]); + amdgpu_ring_mux_fini(&adev->gfx.muxer); + } + for (i = 0; i < adev->gfx.num_gfx_rings; i++) amdgpu_ring_fini(&adev->gfx.gfx_ring[i]); for (i = 0; i < adev->gfx.num_compute_rings; i++) @@ -2324,13 +2370,13 @@ static void gfx_v9_0_setup_rb(struct amdgpu_device *adev) mutex_lock(&adev->grbm_idx_mutex); for (i = 0; i < adev->gfx.config.max_shader_engines; i++) { for (j = 0; j < adev->gfx.config.max_sh_per_se; j++) { - gfx_v9_0_select_se_sh(adev, i, j, 0xffffffff); + amdgpu_gfx_select_se_sh(adev, i, j, 0xffffffff); data = gfx_v9_0_get_rb_active_bitmap(adev); active_rbs |= data << ((i * adev->gfx.config.max_sh_per_se + j) * rb_bitmap_width_per_sh); } } - gfx_v9_0_select_se_sh(adev, 0xffffffff, 0xffffffff, 0xffffffff); + amdgpu_gfx_select_se_sh(adev, 0xffffffff, 0xffffffff, 0xffffffff); mutex_unlock(&adev->grbm_idx_mutex); adev->gfx.config.backend_enable_mask = active_rbs; @@ -2467,14 +2513,14 @@ static void gfx_v9_0_wait_for_rlc_serdes(struct amdgpu_device *adev) mutex_lock(&adev->grbm_idx_mutex); for (i = 0; i < adev->gfx.config.max_shader_engines; i++) { for (j = 0; j < adev->gfx.config.max_sh_per_se; j++) { - gfx_v9_0_select_se_sh(adev, i, j, 0xffffffff); + amdgpu_gfx_select_se_sh(adev, i, j, 0xffffffff); for (k = 0; k < adev->usec_timeout; k++) { if (RREG32_SOC15(GC, 0, mmRLC_SERDES_CU_MASTER_BUSY) == 0) break; udelay(1); } if (k == adev->usec_timeout) { - gfx_v9_0_select_se_sh(adev, 0xffffffff, + amdgpu_gfx_select_se_sh(adev, 0xffffffff, 0xffffffff, 0xffffffff); mutex_unlock(&adev->grbm_idx_mutex); DRM_INFO("Timeout wait for RLC serdes %u,%u\n", @@ -2483,7 +2529,7 @@ static void gfx_v9_0_wait_for_rlc_serdes(struct amdgpu_device *adev) } } } - gfx_v9_0_select_se_sh(adev, 0xffffffff, 0xffffffff, 0xffffffff); + amdgpu_gfx_select_se_sh(adev, 0xffffffff, 0xffffffff, 0xffffffff); mutex_unlock(&adev->grbm_idx_mutex); mask = RLC_SERDES_NONCU_MASTER_BUSY__SE_MASTER_BUSY_MASK | @@ -3583,6 +3629,8 @@ static int gfx_v9_0_kiq_init_queue(struct amdgpu_ring *ring) memset((void *)mqd, 0, sizeof(struct v9_mqd_allocation)); ((struct v9_mqd_allocation *)mqd)->dynamic_cu_mask = 0xFFFFFFFF; ((struct v9_mqd_allocation *)mqd)->dynamic_rb_mask = 0xFFFFFFFF; + if (amdgpu_sriov_vf(adev) && adev->in_suspend) + amdgpu_ring_clear_ring(ring); mutex_lock(&adev->srbm_mutex); soc15_grbm_select(adev, ring->me, ring->pipe, ring->queue, 0); gfx_v9_0_mqd_init(ring); @@ -4539,6 +4587,8 @@ static int gfx_v9_0_early_init(void *handle) { struct amdgpu_device *adev = (struct amdgpu_device *)handle; + adev->gfx.funcs = &gfx_v9_0_gfx_funcs; + if (adev->ip_versions[GC_HWIP][0] == IP_VERSION(9, 4, 1) || adev->ip_versions[GC_HWIP][0] == IP_VERSION(9, 4, 2)) adev->gfx.num_gfx_rings = 0; @@ -5155,11 +5205,17 @@ static void gfx_v9_0_ring_emit_ib_gfx(struct amdgpu_ring *ring, control |= ib->length_dw | (vmid << 24); - if (amdgpu_sriov_vf(ring->adev) && (ib->flags & AMDGPU_IB_FLAG_PREEMPT)) { + if (ib->flags & AMDGPU_IB_FLAG_PREEMPT) { control |= INDIRECT_BUFFER_PRE_ENB(1); + if (flags & AMDGPU_IB_PREEMPTED) + control |= INDIRECT_BUFFER_PRE_RESUME(1); + if (!(ib->flags & AMDGPU_IB_FLAG_CE) && vmid) - gfx_v9_0_ring_emit_de_meta(ring); + gfx_v9_0_ring_emit_de_meta(ring, + (!amdgpu_sriov_vf(ring->adev) && + flags & AMDGPU_IB_PREEMPTED) ? + true : false); } amdgpu_ring_write(ring, header); @@ -5214,17 +5270,24 @@ static void gfx_v9_0_ring_emit_fence(struct amdgpu_ring *ring, u64 addr, bool write64bit = flags & AMDGPU_FENCE_FLAG_64BIT; bool int_sel = flags & AMDGPU_FENCE_FLAG_INT; bool writeback = flags & AMDGPU_FENCE_FLAG_TC_WB_ONLY; + bool exec = flags & AMDGPU_FENCE_FLAG_EXEC; + uint32_t dw2 = 0; /* RELEASE_MEM - flush caches, send int */ amdgpu_ring_write(ring, PACKET3(PACKET3_RELEASE_MEM, 6)); - amdgpu_ring_write(ring, ((writeback ? (EOP_TC_WB_ACTION_EN | - EOP_TC_NC_ACTION_EN) : - (EOP_TCL1_ACTION_EN | - EOP_TC_ACTION_EN | - EOP_TC_WB_ACTION_EN | - EOP_TC_MD_ACTION_EN)) | - EVENT_TYPE(CACHE_FLUSH_AND_INV_TS_EVENT) | - EVENT_INDEX(5))); + + if (writeback) { + dw2 = EOP_TC_NC_ACTION_EN; + } else { + dw2 = EOP_TCL1_ACTION_EN | EOP_TC_ACTION_EN | + EOP_TC_MD_ACTION_EN; + } + dw2 |= EOP_TC_WB_ACTION_EN | EVENT_TYPE(CACHE_FLUSH_AND_INV_TS_EVENT) | + EVENT_INDEX(5); + if (exec) + dw2 |= EOP_EXEC; + + amdgpu_ring_write(ring, dw2); amdgpu_ring_write(ring, DATA_SEL(write64bit ? 2 : 1) | INT_SEL(int_sel ? 2 : 0)); /* @@ -5329,33 +5392,135 @@ static void gfx_v9_ring_emit_sb(struct amdgpu_ring *ring) amdgpu_ring_write(ring, 0); } -static void gfx_v9_0_ring_emit_ce_meta(struct amdgpu_ring *ring) +static void gfx_v9_0_ring_emit_ce_meta(struct amdgpu_ring *ring, bool resume) { + struct amdgpu_device *adev = ring->adev; struct v9_ce_ib_state ce_payload = {0}; - uint64_t csa_addr; + uint64_t offset, ce_payload_gpu_addr; + void *ce_payload_cpu_addr; int cnt; cnt = (sizeof(ce_payload) >> 2) + 4 - 2; - csa_addr = amdgpu_csa_vaddr(ring->adev); + + if (ring->is_mes_queue) { + offset = offsetof(struct amdgpu_mes_ctx_meta_data, + gfx[0].gfx_meta_data) + + offsetof(struct v9_gfx_meta_data, ce_payload); + ce_payload_gpu_addr = + amdgpu_mes_ctx_get_offs_gpu_addr(ring, offset); + ce_payload_cpu_addr = + amdgpu_mes_ctx_get_offs_cpu_addr(ring, offset); + } else { + offset = offsetof(struct v9_gfx_meta_data, ce_payload); + ce_payload_gpu_addr = amdgpu_csa_vaddr(ring->adev) + offset; + ce_payload_cpu_addr = adev->virt.csa_cpu_addr + offset; + } amdgpu_ring_write(ring, PACKET3(PACKET3_WRITE_DATA, cnt)); amdgpu_ring_write(ring, (WRITE_DATA_ENGINE_SEL(2) | WRITE_DATA_DST_SEL(8) | WR_CONFIRM) | WRITE_DATA_CACHE_POLICY(0)); - amdgpu_ring_write(ring, lower_32_bits(csa_addr + offsetof(struct v9_gfx_meta_data, ce_payload))); - amdgpu_ring_write(ring, upper_32_bits(csa_addr + offsetof(struct v9_gfx_meta_data, ce_payload))); - amdgpu_ring_write_multiple(ring, (void *)&ce_payload, sizeof(ce_payload) >> 2); + amdgpu_ring_write(ring, lower_32_bits(ce_payload_gpu_addr)); + amdgpu_ring_write(ring, upper_32_bits(ce_payload_gpu_addr)); + + if (resume) + amdgpu_ring_write_multiple(ring, ce_payload_cpu_addr, + sizeof(ce_payload) >> 2); + else + amdgpu_ring_write_multiple(ring, (void *)&ce_payload, + sizeof(ce_payload) >> 2); +} + +static int gfx_v9_0_ring_preempt_ib(struct amdgpu_ring *ring) +{ + int i, r = 0; + struct amdgpu_device *adev = ring->adev; + struct amdgpu_kiq *kiq = &adev->gfx.kiq; + struct amdgpu_ring *kiq_ring = &kiq->ring; + unsigned long flags; + + if (!kiq->pmf || !kiq->pmf->kiq_unmap_queues) + return -EINVAL; + + spin_lock_irqsave(&kiq->ring_lock, flags); + + if (amdgpu_ring_alloc(kiq_ring, kiq->pmf->unmap_queues_size)) { + spin_unlock_irqrestore(&kiq->ring_lock, flags); + return -ENOMEM; + } + + /* assert preemption condition */ + amdgpu_ring_set_preempt_cond_exec(ring, false); + + ring->trail_seq += 1; + amdgpu_ring_alloc(ring, 13); + gfx_v9_0_ring_emit_fence(ring, ring->trail_fence_gpu_addr, + ring->trail_seq, AMDGPU_FENCE_FLAG_EXEC | AMDGPU_FENCE_FLAG_INT); + /*reset the CP_VMID_PREEMPT after trailing fence*/ + amdgpu_ring_emit_wreg(ring, + SOC15_REG_OFFSET(GC, 0, mmCP_VMID_PREEMPT), + 0x0); + + /* assert IB preemption, emit the trailing fence */ + kiq->pmf->kiq_unmap_queues(kiq_ring, ring, PREEMPT_QUEUES_NO_UNMAP, + ring->trail_fence_gpu_addr, + ring->trail_seq); + + amdgpu_ring_commit(kiq_ring); + spin_unlock_irqrestore(&kiq->ring_lock, flags); + + /* poll the trailing fence */ + for (i = 0; i < adev->usec_timeout; i++) { + if (ring->trail_seq == + le32_to_cpu(*ring->trail_fence_cpu_addr)) + break; + udelay(1); + } + + if (i >= adev->usec_timeout) { + r = -EINVAL; + DRM_WARN("ring %d timeout to preempt ib\n", ring->idx); + } + + amdgpu_ring_commit(ring); + + /* deassert preemption condition */ + amdgpu_ring_set_preempt_cond_exec(ring, true); + return r; } -static void gfx_v9_0_ring_emit_de_meta(struct amdgpu_ring *ring) +static void gfx_v9_0_ring_emit_de_meta(struct amdgpu_ring *ring, bool resume) { + struct amdgpu_device *adev = ring->adev; struct v9_de_ib_state de_payload = {0}; - uint64_t csa_addr, gds_addr; + uint64_t offset, gds_addr, de_payload_gpu_addr; + void *de_payload_cpu_addr; int cnt; - csa_addr = amdgpu_csa_vaddr(ring->adev); - gds_addr = csa_addr + 4096; + if (ring->is_mes_queue) { + offset = offsetof(struct amdgpu_mes_ctx_meta_data, + gfx[0].gfx_meta_data) + + offsetof(struct v9_gfx_meta_data, de_payload); + de_payload_gpu_addr = + amdgpu_mes_ctx_get_offs_gpu_addr(ring, offset); + de_payload_cpu_addr = + amdgpu_mes_ctx_get_offs_cpu_addr(ring, offset); + + offset = offsetof(struct amdgpu_mes_ctx_meta_data, + gfx[0].gds_backup) + + offsetof(struct v9_gfx_meta_data, de_payload); + gds_addr = amdgpu_mes_ctx_get_offs_gpu_addr(ring, offset); + } else { + offset = offsetof(struct v9_gfx_meta_data, de_payload); + de_payload_gpu_addr = amdgpu_csa_vaddr(ring->adev) + offset; + de_payload_cpu_addr = adev->virt.csa_cpu_addr + offset; + + gds_addr = ALIGN(amdgpu_csa_vaddr(ring->adev) + + AMDGPU_CSA_SIZE - adev->gds.gds_size, + PAGE_SIZE); + } + de_payload.gds_backup_addrlo = lower_32_bits(gds_addr); de_payload.gds_backup_addrhi = upper_32_bits(gds_addr); @@ -5365,9 +5530,15 @@ static void gfx_v9_0_ring_emit_de_meta(struct amdgpu_ring *ring) WRITE_DATA_DST_SEL(8) | WR_CONFIRM) | WRITE_DATA_CACHE_POLICY(0)); - amdgpu_ring_write(ring, lower_32_bits(csa_addr + offsetof(struct v9_gfx_meta_data, de_payload))); - amdgpu_ring_write(ring, upper_32_bits(csa_addr + offsetof(struct v9_gfx_meta_data, de_payload))); - amdgpu_ring_write_multiple(ring, (void *)&de_payload, sizeof(de_payload) >> 2); + amdgpu_ring_write(ring, lower_32_bits(de_payload_gpu_addr)); + amdgpu_ring_write(ring, upper_32_bits(de_payload_gpu_addr)); + + if (resume) + amdgpu_ring_write_multiple(ring, de_payload_cpu_addr, + sizeof(de_payload) >> 2); + else + amdgpu_ring_write_multiple(ring, (void *)&de_payload, + sizeof(de_payload) >> 2); } static void gfx_v9_0_ring_emit_frame_cntl(struct amdgpu_ring *ring, bool start, @@ -5383,8 +5554,9 @@ static void gfx_v9_ring_emit_cntxcntl(struct amdgpu_ring *ring, uint32_t flags) { uint32_t dw2 = 0; - if (amdgpu_sriov_vf(ring->adev)) - gfx_v9_0_ring_emit_ce_meta(ring); + gfx_v9_0_ring_emit_ce_meta(ring, + (!amdgpu_sriov_vf(ring->adev) && + flags & AMDGPU_IB_PREEMPTED) ? true : false); dw2 |= 0x80000000; /* set load_enable otherwise this package is just NOPs */ if (flags & AMDGPU_HAVE_CTX_SWITCH) { @@ -5710,7 +5882,12 @@ static int gfx_v9_0_eop_irq(struct amdgpu_device *adev, switch (me_id) { case 0: - amdgpu_fence_process(&adev->gfx.gfx_ring[0]); + if (adev->gfx.num_gfx_rings && + !amdgpu_mcbp_handle_trailing_fence_irq(&adev->gfx.muxer)) { + /* Fence signals are handled on the software rings*/ + for (i = 0; i < GFX9_NUM_SW_GFX_RINGS; i++) + amdgpu_fence_process(&adev->gfx.sw_gfx_ring[i]); + } break; case 1: case 2: @@ -6482,7 +6659,7 @@ static void gfx_v9_0_reset_ras_error_count(struct amdgpu_device *adev) for (i = 0; i < ARRAY_SIZE(gfx_v9_0_edc_counter_regs); i++) { for (j = 0; j < gfx_v9_0_edc_counter_regs[i].se_num; j++) { for (k = 0; k < gfx_v9_0_edc_counter_regs[i].instance; k++) { - gfx_v9_0_select_se_sh(adev, j, 0x0, k); + amdgpu_gfx_select_se_sh(adev, j, 0x0, k); RREG32(SOC15_REG_ENTRY_OFFSET(gfx_v9_0_edc_counter_regs[i])); } } @@ -6544,7 +6721,7 @@ static void gfx_v9_0_query_ras_error_count(struct amdgpu_device *adev, for (i = 0; i < ARRAY_SIZE(gfx_v9_0_edc_counter_regs); i++) { for (j = 0; j < gfx_v9_0_edc_counter_regs[i].se_num; j++) { for (k = 0; k < gfx_v9_0_edc_counter_regs[i].instance; k++) { - gfx_v9_0_select_se_sh(adev, j, 0, k); + amdgpu_gfx_select_se_sh(adev, j, 0, k); reg_value = RREG32(SOC15_REG_ENTRY_OFFSET(gfx_v9_0_edc_counter_regs[i])); if (reg_value) @@ -6559,7 +6736,7 @@ static void gfx_v9_0_query_ras_error_count(struct amdgpu_device *adev, err_data->ce_count += sec_count; err_data->ue_count += ded_count; - gfx_v9_0_select_se_sh(adev, 0xffffffff, 0xffffffff, 0xffffffff); + amdgpu_gfx_select_se_sh(adev, 0xffffffff, 0xffffffff, 0xffffffff); mutex_unlock(&adev->grbm_idx_mutex); gfx_v9_0_query_utc_edc_status(adev, err_data); @@ -6707,6 +6884,62 @@ static const struct amdgpu_ring_funcs gfx_v9_0_ring_funcs_gfx = { .emit_cntxcntl = gfx_v9_ring_emit_cntxcntl, .init_cond_exec = gfx_v9_0_ring_emit_init_cond_exec, .patch_cond_exec = gfx_v9_0_ring_emit_patch_cond_exec, + .preempt_ib = gfx_v9_0_ring_preempt_ib, + .emit_frame_cntl = gfx_v9_0_ring_emit_frame_cntl, + .emit_wreg = gfx_v9_0_ring_emit_wreg, + .emit_reg_wait = gfx_v9_0_ring_emit_reg_wait, + .emit_reg_write_reg_wait = gfx_v9_0_ring_emit_reg_write_reg_wait, + .soft_recovery = gfx_v9_0_ring_soft_recovery, + .emit_mem_sync = gfx_v9_0_emit_mem_sync, +}; + +static const struct amdgpu_ring_funcs gfx_v9_0_sw_ring_funcs_gfx = { + .type = AMDGPU_RING_TYPE_GFX, + .align_mask = 0xff, + .nop = PACKET3(PACKET3_NOP, 0x3FFF), + .support_64bit_ptrs = true, + .secure_submission_supported = true, + .vmhub = AMDGPU_GFXHUB_0, + .get_rptr = amdgpu_sw_ring_get_rptr_gfx, + .get_wptr = amdgpu_sw_ring_get_wptr_gfx, + .set_wptr = amdgpu_sw_ring_set_wptr_gfx, + .emit_frame_size = /* totally 242 maximum if 16 IBs */ + 5 + /* COND_EXEC */ + 7 + /* PIPELINE_SYNC */ + SOC15_FLUSH_GPU_TLB_NUM_WREG * 5 + + SOC15_FLUSH_GPU_TLB_NUM_REG_WAIT * 7 + + 2 + /* VM_FLUSH */ + 8 + /* FENCE for VM_FLUSH */ + 20 + /* GDS switch */ + 4 + /* double SWITCH_BUFFER, + * the first COND_EXEC jump to the place just + * prior to this double SWITCH_BUFFER + */ + 5 + /* COND_EXEC */ + 7 + /* HDP_flush */ + 4 + /* VGT_flush */ + 14 + /* CE_META */ + 31 + /* DE_META */ + 3 + /* CNTX_CTRL */ + 5 + /* HDP_INVL */ + 8 + 8 + /* FENCE x2 */ + 2 + /* SWITCH_BUFFER */ + 7, /* gfx_v9_0_emit_mem_sync */ + .emit_ib_size = 4, /* gfx_v9_0_ring_emit_ib_gfx */ + .emit_ib = gfx_v9_0_ring_emit_ib_gfx, + .emit_fence = gfx_v9_0_ring_emit_fence, + .emit_pipeline_sync = gfx_v9_0_ring_emit_pipeline_sync, + .emit_vm_flush = gfx_v9_0_ring_emit_vm_flush, + .emit_gds_switch = gfx_v9_0_ring_emit_gds_switch, + .emit_hdp_flush = gfx_v9_0_ring_emit_hdp_flush, + .test_ring = gfx_v9_0_ring_test_ring, + .test_ib = gfx_v9_0_ring_test_ib, + .insert_nop = amdgpu_sw_ring_insert_nop, + .pad_ib = amdgpu_ring_generic_pad_ib, + .emit_switch_buffer = gfx_v9_ring_emit_sb, + .emit_cntxcntl = gfx_v9_ring_emit_cntxcntl, + .init_cond_exec = gfx_v9_0_ring_emit_init_cond_exec, + .patch_cond_exec = gfx_v9_0_ring_emit_patch_cond_exec, .emit_frame_cntl = gfx_v9_0_ring_emit_frame_cntl, .emit_wreg = gfx_v9_0_ring_emit_wreg, .emit_reg_wait = gfx_v9_0_ring_emit_reg_wait, @@ -6792,6 +7025,11 @@ static void gfx_v9_0_set_ring_funcs(struct amdgpu_device *adev) for (i = 0; i < adev->gfx.num_gfx_rings; i++) adev->gfx.gfx_ring[i].funcs = &gfx_v9_0_ring_funcs_gfx; + if (adev->gfx.num_gfx_rings) { + for (i = 0; i < GFX9_NUM_SW_GFX_RINGS; i++) + adev->gfx.sw_gfx_ring[i].funcs = &gfx_v9_0_sw_ring_funcs_gfx; + } + for (i = 0; i < adev->gfx.num_compute_rings; i++) adev->gfx.compute_ring[i].funcs = &gfx_v9_0_ring_funcs_compute; } @@ -6963,7 +7201,7 @@ static int gfx_v9_0_get_cu_info(struct amdgpu_device *adev, mask = 1; ao_bitmap = 0; counter = 0; - gfx_v9_0_select_se_sh(adev, i, j, 0xffffffff); + amdgpu_gfx_select_se_sh(adev, i, j, 0xffffffff); gfx_v9_0_set_user_cu_inactive_bitmap( adev, disable_masks[i * adev->gfx.config.max_sh_per_se + j]); bitmap = gfx_v9_0_get_cu_active_bitmap(adev); @@ -6996,7 +7234,7 @@ static int gfx_v9_0_get_cu_info(struct amdgpu_device *adev, cu_info->ao_cu_bitmap[i % 4][j + i / 4] = ao_bitmap; } } - gfx_v9_0_select_se_sh(adev, 0xffffffff, 0xffffffff, 0xffffffff); + amdgpu_gfx_select_se_sh(adev, 0xffffffff, 0xffffffff, 0xffffffff); mutex_unlock(&adev->grbm_idx_mutex); cu_info->number = active_cu_number; diff --git a/drivers/gpu/drm/amd/amdgpu/gfxhub_v2_1.c b/drivers/gpu/drm/amd/amdgpu/gfxhub_v2_1.c index 8cf53e039c11..3f8676d23a5e 100644 --- a/drivers/gpu/drm/amd/amdgpu/gfxhub_v2_1.c +++ b/drivers/gpu/drm/amd/amdgpu/gfxhub_v2_1.c @@ -397,6 +397,9 @@ static void gfxhub_v2_1_gart_disable(struct amdgpu_device *adev) ENABLE_ADVANCED_DRIVER_MODEL, 0); WREG32_SOC15(GC, 0, mmGCMC_VM_MX_L1_TLB_CNTL, tmp); + if (amdgpu_sriov_vf(adev)) + return; + /* Setup L2 cache */ WREG32_FIELD15(GC, 0, GCVM_L2_CNTL, ENABLE_L2_CACHE, 0); WREG32_SOC15(GC, 0, mmGCVM_L2_CNTL3, 0); diff --git a/drivers/gpu/drm/amd/amdgpu/gfxhub_v3_0_3.c b/drivers/gpu/drm/amd/amdgpu/gfxhub_v3_0_3.c index 5d3fffd4929f..080ff11ca305 100644 --- a/drivers/gpu/drm/amd/amdgpu/gfxhub_v3_0_3.c +++ b/drivers/gpu/drm/amd/amdgpu/gfxhub_v3_0_3.c @@ -154,6 +154,9 @@ static void gfxhub_v3_0_3_init_system_aperture_regs(struct amdgpu_device *adev) { uint64_t value; + if (amdgpu_sriov_vf(adev)) + return; + /* Disable AGP. */ WREG32_SOC15(GC, 0, regGCMC_VM_AGP_BASE, 0); WREG32_SOC15(GC, 0, regGCMC_VM_AGP_TOP, 0); @@ -354,18 +357,6 @@ static void gfxhub_v3_0_3_program_invalidation(struct amdgpu_device *adev) static int gfxhub_v3_0_3_gart_enable(struct amdgpu_device *adev) { - if (amdgpu_sriov_vf(adev)) { - /* - * GCMC_VM_FB_LOCATION_BASE/TOP is NULL for VF, becuase they are - * VF copy registers so vbios post doesn't program them, for - * SRIOV driver need to program them - */ - WREG32_SOC15(GC, 0, regGCMC_VM_FB_LOCATION_BASE, - adev->gmc.vram_start >> 24); - WREG32_SOC15(GC, 0, regGCMC_VM_FB_LOCATION_TOP, - adev->gmc.vram_end >> 24); - } - /* GART Enable. */ gfxhub_v3_0_3_init_gart_aperture_regs(adev); gfxhub_v3_0_3_init_system_aperture_regs(adev); diff --git a/drivers/gpu/drm/amd/amdgpu/gmc_v10_0.c b/drivers/gpu/drm/amd/amdgpu/gmc_v10_0.c index f513e2c2e964..21e46817d82d 100644 --- a/drivers/gpu/drm/amd/amdgpu/gmc_v10_0.c +++ b/drivers/gpu/drm/amd/amdgpu/gmc_v10_0.c @@ -371,7 +371,9 @@ static void gmc_v10_0_flush_gpu_tlb(struct amdgpu_device *adev, uint32_t vmid, * translation. Avoid this by doing the invalidation from the SDMA * itself. */ - r = amdgpu_job_alloc_with_ib(adev, 16 * 4, AMDGPU_IB_POOL_IMMEDIATE, + r = amdgpu_job_alloc_with_ib(ring->adev, &adev->mman.entity, + AMDGPU_FENCE_OWNER_UNDEFINED, + 16 * 4, AMDGPU_IB_POOL_IMMEDIATE, &job); if (r) goto error_alloc; @@ -380,10 +382,7 @@ static void gmc_v10_0_flush_gpu_tlb(struct amdgpu_device *adev, uint32_t vmid, job->vm_needs_flush = true; job->ibs->ptr[job->ibs->length_dw++] = ring->funcs->nop; amdgpu_ring_pad_ib(ring, &job->ibs[0]); - r = amdgpu_job_submit(job, &adev->mman.entity, - AMDGPU_FENCE_OWNER_UNDEFINED, &fence); - if (r) - goto error_submit; + fence = amdgpu_job_submit(job); mutex_unlock(&adev->mman.gtt_window_lock); @@ -392,9 +391,6 @@ static void gmc_v10_0_flush_gpu_tlb(struct amdgpu_device *adev, uint32_t vmid, return; -error_submit: - amdgpu_job_free(job); - error_alloc: mutex_unlock(&adev->mman.gtt_window_lock); DRM_ERROR("Error flushing GPU TLB using the SDMA (%d)!\n", r); @@ -612,6 +608,8 @@ static void gmc_v10_0_get_vm_pte(struct amdgpu_device *adev, struct amdgpu_bo_va_mapping *mapping, uint64_t *flags) { + struct amdgpu_bo *bo = mapping->bo_va->base.bo; + *flags &= ~AMDGPU_PTE_EXECUTABLE; *flags |= mapping->flags & AMDGPU_PTE_EXECUTABLE; @@ -628,6 +626,11 @@ static void gmc_v10_0_get_vm_pte(struct amdgpu_device *adev, *flags |= AMDGPU_PTE_SYSTEM; *flags &= ~AMDGPU_PTE_VALID; } + + if (bo && bo->flags & (AMDGPU_GEM_CREATE_COHERENT | + AMDGPU_GEM_CREATE_UNCACHED)) + *flags = (*flags & ~AMDGPU_PTE_MTYPE_NV10_MASK) | + AMDGPU_PTE_MTYPE_NV10(MTYPE_UC); } static unsigned gmc_v10_0_get_vbios_fb_size(struct amdgpu_device *adev) diff --git a/drivers/gpu/drm/amd/amdgpu/gmc_v11_0.c b/drivers/gpu/drm/amd/amdgpu/gmc_v11_0.c index 66dfb574cc7d..4326078689cd 100644 --- a/drivers/gpu/drm/amd/amdgpu/gmc_v11_0.c +++ b/drivers/gpu/drm/amd/amdgpu/gmc_v11_0.c @@ -503,6 +503,8 @@ static void gmc_v11_0_get_vm_pte(struct amdgpu_device *adev, struct amdgpu_bo_va_mapping *mapping, uint64_t *flags) { + struct amdgpu_bo *bo = mapping->bo_va->base.bo; + *flags &= ~AMDGPU_PTE_EXECUTABLE; *flags |= mapping->flags & AMDGPU_PTE_EXECUTABLE; @@ -519,6 +521,11 @@ static void gmc_v11_0_get_vm_pte(struct amdgpu_device *adev, *flags |= AMDGPU_PTE_SYSTEM; *flags &= ~AMDGPU_PTE_VALID; } + + if (bo && bo->flags & (AMDGPU_GEM_CREATE_COHERENT | + AMDGPU_GEM_CREATE_UNCACHED)) + *flags = (*flags & ~AMDGPU_PTE_MTYPE_NV10_MASK) | + AMDGPU_PTE_MTYPE_NV10(MTYPE_UC); } static unsigned gmc_v11_0_get_vbios_fb_size(struct amdgpu_device *adev) @@ -551,7 +558,10 @@ static void gmc_v11_0_set_umc_funcs(struct amdgpu_device *adev) adev->umc.node_inst_num = adev->gmc.num_umc; adev->umc.max_ras_err_cnt_per_query = UMC_V8_10_TOTAL_CHANNEL_NUM(adev); adev->umc.channel_offs = UMC_V8_10_PER_CHANNEL_OFFSET; - adev->umc.channel_idx_tbl = &umc_v8_10_channel_idx_tbl[0][0][0]; + if (adev->umc.node_inst_num == 4) + adev->umc.channel_idx_tbl = &umc_v8_10_channel_idx_tbl_ext0[0][0][0]; + else + adev->umc.channel_idx_tbl = &umc_v8_10_channel_idx_tbl[0][0][0]; adev->umc.ras = &umc_v8_10_ras; break; case IP_VERSION(8, 11, 0): @@ -749,6 +759,7 @@ static int gmc_v11_0_sw_init(void *handle) case IP_VERSION(11, 0, 1): case IP_VERSION(11, 0, 2): case IP_VERSION(11, 0, 3): + case IP_VERSION(11, 0, 4): adev->num_vmhubs = 2; /* * To fulfill 4-level page support, diff --git a/drivers/gpu/drm/amd/amdgpu/gmc_v9_0.c b/drivers/gpu/drm/amd/amdgpu/gmc_v9_0.c index 67ca16a8027c..08d6cf79fb15 100644 --- a/drivers/gpu/drm/amd/amdgpu/gmc_v9_0.c +++ b/drivers/gpu/drm/amd/amdgpu/gmc_v9_0.c @@ -1113,10 +1113,80 @@ static void gmc_v9_0_get_vm_pde(struct amdgpu_device *adev, int level, } } +static void gmc_v9_0_get_coherence_flags(struct amdgpu_device *adev, + struct amdgpu_bo *bo, + struct amdgpu_bo_va_mapping *mapping, + uint64_t *flags) +{ + struct amdgpu_device *bo_adev = amdgpu_ttm_adev(bo->tbo.bdev); + bool is_vram = bo->tbo.resource->mem_type == TTM_PL_VRAM; + bool coherent = bo->flags & AMDGPU_GEM_CREATE_COHERENT; + bool uncached = bo->flags & AMDGPU_GEM_CREATE_UNCACHED; + unsigned int mtype; + bool snoop = false; + + switch (adev->ip_versions[GC_HWIP][0]) { + case IP_VERSION(9, 4, 1): + case IP_VERSION(9, 4, 2): + if (is_vram) { + if (bo_adev == adev) { + if (uncached) + mtype = MTYPE_UC; + else if (coherent) + mtype = MTYPE_CC; + else + mtype = MTYPE_RW; + /* FIXME: is this still needed? Or does + * amdgpu_ttm_tt_pde_flags already handle this? + */ + if (adev->ip_versions[GC_HWIP][0] == + IP_VERSION(9, 4, 2) && + adev->gmc.xgmi.connected_to_cpu) + snoop = true; + } else { + if (uncached || coherent) + mtype = MTYPE_UC; + else + mtype = MTYPE_NC; + if (mapping->bo_va->is_xgmi) + snoop = true; + } + } else { + if (uncached || coherent) + mtype = MTYPE_UC; + else + mtype = MTYPE_NC; + /* FIXME: is this still needed? Or does + * amdgpu_ttm_tt_pde_flags already handle this? + */ + snoop = true; + } + break; + default: + if (uncached || coherent) + mtype = MTYPE_UC; + else + mtype = MTYPE_NC; + + /* FIXME: is this still needed? Or does + * amdgpu_ttm_tt_pde_flags already handle this? + */ + if (!is_vram) + snoop = true; + } + + if (mtype != MTYPE_NC) + *flags = (*flags & ~AMDGPU_PTE_MTYPE_VG10_MASK) | + AMDGPU_PTE_MTYPE_VG10(mtype); + *flags |= snoop ? AMDGPU_PTE_SNOOPED : 0; +} + static void gmc_v9_0_get_vm_pte(struct amdgpu_device *adev, struct amdgpu_bo_va_mapping *mapping, uint64_t *flags) { + struct amdgpu_bo *bo = mapping->bo_va->base.bo; + *flags &= ~AMDGPU_PTE_EXECUTABLE; *flags |= mapping->flags & AMDGPU_PTE_EXECUTABLE; @@ -1128,14 +1198,9 @@ static void gmc_v9_0_get_vm_pte(struct amdgpu_device *adev, *flags &= ~AMDGPU_PTE_VALID; } - if ((adev->ip_versions[GC_HWIP][0] == IP_VERSION(9, 4, 1) || - adev->ip_versions[GC_HWIP][0] == IP_VERSION(9, 4, 2)) && - !(*flags & AMDGPU_PTE_SYSTEM) && - mapping->bo_va->is_xgmi) - *flags |= AMDGPU_PTE_SNOOPED; - - if (adev->ip_versions[GC_HWIP][0] == IP_VERSION(9, 4, 2)) - *flags |= mapping->flags & AMDGPU_PTE_SNOOPED; + if (bo && bo->tbo.resource) + gmc_v9_0_get_coherence_flags(adev, mapping->bo_va->base.bo, + mapping, flags); } static unsigned gmc_v9_0_get_vbios_fb_size(struct amdgpu_device *adev) diff --git a/drivers/gpu/drm/amd/amdgpu/jpeg_v2_5.c b/drivers/gpu/drm/amd/amdgpu/jpeg_v2_5.c index f87d0f6ffc93..f2b743a93915 100644 --- a/drivers/gpu/drm/amd/amdgpu/jpeg_v2_5.c +++ b/drivers/gpu/drm/amd/amdgpu/jpeg_v2_5.c @@ -807,16 +807,5 @@ static void jpeg_v2_5_set_ras_funcs(struct amdgpu_device *adev) break; } - if (adev->jpeg.ras) { - amdgpu_ras_register_ras_block(adev, &adev->jpeg.ras->ras_block); - - strcpy(adev->jpeg.ras->ras_block.ras_comm.name, "jpeg"); - adev->jpeg.ras->ras_block.ras_comm.block = AMDGPU_RAS_BLOCK__JPEG; - adev->jpeg.ras->ras_block.ras_comm.type = AMDGPU_RAS_ERROR__POISON; - adev->jpeg.ras_if = &adev->jpeg.ras->ras_block.ras_comm; - - /* If don't define special ras_late_init function, use default ras_late_init */ - if (!adev->jpeg.ras->ras_block.ras_late_init) - adev->jpeg.ras->ras_block.ras_late_init = amdgpu_ras_block_late_init; - } + jpeg_set_ras_funcs(adev); } diff --git a/drivers/gpu/drm/amd/amdgpu/jpeg_v4_0.c b/drivers/gpu/drm/amd/amdgpu/jpeg_v4_0.c index 63b0d0b810ec..3beb731b2ce5 100644 --- a/drivers/gpu/drm/amd/amdgpu/jpeg_v4_0.c +++ b/drivers/gpu/drm/amd/amdgpu/jpeg_v4_0.c @@ -27,6 +27,7 @@ #include "soc15.h" #include "soc15d.h" #include "jpeg_v2_0.h" +#include "jpeg_v4_0.h" #include "vcn/vcn_4_0_0_offset.h" #include "vcn/vcn_4_0_0_sh_mask.h" @@ -38,6 +39,7 @@ static void jpeg_v4_0_set_dec_ring_funcs(struct amdgpu_device *adev); static void jpeg_v4_0_set_irq_funcs(struct amdgpu_device *adev); static int jpeg_v4_0_set_powergating_state(void *handle, enum amd_powergating_state state); +static void jpeg_v4_0_set_ras_funcs(struct amdgpu_device *adev); /** * jpeg_v4_0_early_init - set function pointers @@ -55,6 +57,7 @@ static int jpeg_v4_0_early_init(void *handle) jpeg_v4_0_set_dec_ring_funcs(adev); jpeg_v4_0_set_irq_funcs(adev); + jpeg_v4_0_set_ras_funcs(adev); return 0; } @@ -78,6 +81,18 @@ static int jpeg_v4_0_sw_init(void *handle) if (r) 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->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->irq); + if (r) + return r; + r = amdgpu_jpeg_sw_init(adev); if (r) return r; @@ -167,6 +182,8 @@ static int jpeg_v4_0_hw_fini(void *handle) RREG32_SOC15(JPEG, 0, regUVD_JRBC_STATUS)) jpeg_v4_0_set_powergating_state(adev, AMD_PG_STATE_GATE); + amdgpu_irq_put(adev, &adev->jpeg.inst->irq, 0); + return 0; } @@ -524,6 +541,10 @@ static int jpeg_v4_0_process_interrupt(struct amdgpu_device *adev, case VCN_4_0__SRCID__JPEG_DECODE: amdgpu_fence_process(&adev->jpeg.inst->ring_dec); break; + case VCN_4_0__SRCID_DJPEG0_POISON: + case VCN_4_0__SRCID_EJPEG0_POISON: + amdgpu_jpeg_process_poison_irq(adev, source, entry); + break; default: DRM_DEV_ERROR(adev->dev, "Unhandled interrupt: %d %d\n", entry->src_id, entry->src_data[0]); @@ -607,3 +628,63 @@ const struct amdgpu_ip_block_version jpeg_v4_0_ip_block = { .rev = 0, .funcs = &jpeg_v4_0_ip_funcs, }; + +static uint32_t jpeg_v4_0_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_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_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_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_MAX_SUB_BLOCK; sub++) + poison_stat += + jpeg_v4_0_query_poison_by_instance(adev, inst, sub); + + return !!poison_stat; +} + +const struct amdgpu_ras_block_hw_ops jpeg_v4_0_ras_hw_ops = { + .query_poison_status = jpeg_v4_0_query_ras_poison_status, +}; + +static struct amdgpu_jpeg_ras jpeg_v4_0_ras = { + .ras_block = { + .hw_ops = &jpeg_v4_0_ras_hw_ops, + }, +}; + +static void jpeg_v4_0_set_ras_funcs(struct amdgpu_device *adev) +{ + switch (adev->ip_versions[JPEG_HWIP][0]) { + case IP_VERSION(4, 0, 0): + adev->jpeg.ras = &jpeg_v4_0_ras; + break; + default: + break; + } + + jpeg_set_ras_funcs(adev); +} diff --git a/drivers/gpu/drm/amd/amdgpu/jpeg_v4_0.h b/drivers/gpu/drm/amd/amdgpu/jpeg_v4_0.h index f1ed6ccfedca..07d36c2abd6b 100644 --- a/drivers/gpu/drm/amd/amdgpu/jpeg_v4_0.h +++ b/drivers/gpu/drm/amd/amdgpu/jpeg_v4_0.h @@ -24,6 +24,13 @@ #ifndef __JPEG_V4_0_H__ #define __JPEG_V4_0_H__ +enum amdgpu_jpeg_v4_0_sub_block { + AMDGPU_JPEG_V4_0_JPEG0 = 0, + AMDGPU_JPEG_V4_0_JPEG1, + + AMDGPU_JPEG_V4_0_MAX_SUB_BLOCK, +}; + extern const struct amdgpu_ip_block_version jpeg_v4_0_ip_block; #endif /* __JPEG_V4_0_H__ */ diff --git a/drivers/gpu/drm/amd/amdgpu/mes_v10_1.c b/drivers/gpu/drm/amd/amdgpu/mes_v10_1.c index 067d10073a56..614394118a53 100644 --- a/drivers/gpu/drm/amd/amdgpu/mes_v10_1.c +++ b/drivers/gpu/drm/amd/amdgpu/mes_v10_1.c @@ -121,6 +121,10 @@ static int mes_v10_1_submit_pkt_and_poll_completion(struct amdgpu_mes *mes, if (r < 1) { DRM_ERROR("MES failed to response msg=%d\n", x_pkt->header.opcode); + + while (halt_if_hws_hang) + schedule(); + return -ETIMEDOUT; } @@ -415,10 +419,6 @@ static int mes_v10_1_init_microcode(struct amdgpu_device *adev, mes_hdr = (const struct mes_firmware_header_v1_0 *) adev->mes.fw[pipe]->data; - adev->mes.ucode_fw_version[pipe] = - le32_to_cpu(mes_hdr->mes_ucode_version); - adev->mes.ucode_fw_version[pipe] = - le32_to_cpu(mes_hdr->mes_ucode_data_version); adev->mes.uc_start_addr[pipe] = le32_to_cpu(mes_hdr->mes_uc_start_addr_lo) | ((uint64_t)(le32_to_cpu(mes_hdr->mes_uc_start_addr_hi)) << 32); diff --git a/drivers/gpu/drm/amd/amdgpu/mes_v11_0.c b/drivers/gpu/drm/amd/amdgpu/mes_v11_0.c index f141fadd2d86..970b066b37bb 100644 --- a/drivers/gpu/drm/amd/amdgpu/mes_v11_0.c +++ b/drivers/gpu/drm/amd/amdgpu/mes_v11_0.c @@ -129,6 +129,10 @@ static int mes_v11_0_submit_pkt_and_poll_completion(struct amdgpu_mes *mes, if (r < 1) { DRM_ERROR("MES failed to response msg=%d\n", x_pkt->header.opcode); + + while (halt_if_hws_hang) + schedule(); + return -ETIMEDOUT; } @@ -384,6 +388,7 @@ static int mes_v11_0_set_hw_resources(struct amdgpu_mes *mes) mes_set_hw_res_pkt.disable_reset = 1; mes_set_hw_res_pkt.disable_mes_log = 1; mes_set_hw_res_pkt.use_different_vmid_compute = 1; + mes_set_hw_res_pkt.enable_reg_active_poll = 1; mes_set_hw_res_pkt.oversubscription_timer = 50; return mes_v11_0_submit_pkt_and_poll_completion(mes, @@ -485,10 +490,6 @@ static int mes_v11_0_init_microcode(struct amdgpu_device *adev, mes_hdr = (const struct mes_firmware_header_v1_0 *) adev->mes.fw[pipe]->data; - adev->mes.ucode_fw_version[pipe] = - le32_to_cpu(mes_hdr->mes_ucode_version); - adev->mes.ucode_fw_version[pipe] = - le32_to_cpu(mes_hdr->mes_ucode_data_version); adev->mes.uc_start_addr[pipe] = le32_to_cpu(mes_hdr->mes_uc_start_addr_lo) | ((uint64_t)(le32_to_cpu(mes_hdr->mes_uc_start_addr_hi)) << 32); @@ -1253,7 +1254,9 @@ static int mes_v11_0_kiq_hw_fini(struct amdgpu_device *adev) if (adev->mes.ring.sched.ready) mes_v11_0_kiq_dequeue_sched(adev); - mes_v11_0_enable(adev, false); + if (!amdgpu_sriov_vf(adev)) + mes_v11_0_enable(adev, false); + return 0; } @@ -1339,7 +1342,8 @@ static int mes_v11_0_late_init(void *handle) { struct amdgpu_device *adev = (struct amdgpu_device *)handle; - if (!amdgpu_in_reset(adev) && + /* it's only intended for use in mes_self_test case, not for s0ix and reset */ + if (!amdgpu_in_reset(adev) && !adev->in_s0ix && (adev->ip_versions[GC_HWIP][0] != IP_VERSION(11, 0, 3))) amdgpu_mes_self_test(adev); diff --git a/drivers/gpu/drm/amd/amdgpu/mmhub_v2_0.c b/drivers/gpu/drm/amd/amdgpu/mmhub_v2_0.c index 998b5d17b271..0e664d0cc8d5 100644 --- a/drivers/gpu/drm/amd/amdgpu/mmhub_v2_0.c +++ b/drivers/gpu/drm/amd/amdgpu/mmhub_v2_0.c @@ -319,7 +319,7 @@ static void mmhub_v2_0_init_cache_regs(struct amdgpu_device *adev) tmp = mmMMVM_L2_CNTL5_DEFAULT; tmp = REG_SET_FIELD(tmp, MMVM_L2_CNTL5, L2_CACHE_SMALLK_FRAGMENT_SIZE, 0); - WREG32_SOC15(GC, 0, mmMMVM_L2_CNTL5, tmp); + WREG32_SOC15(MMHUB, 0, mmMMVM_L2_CNTL5, tmp); } static void mmhub_v2_0_enable_system_domain(struct amdgpu_device *adev) diff --git a/drivers/gpu/drm/amd/amdgpu/mmhub_v2_3.c b/drivers/gpu/drm/amd/amdgpu/mmhub_v2_3.c index 1b027d069ab4..4638ea7c2eec 100644 --- a/drivers/gpu/drm/amd/amdgpu/mmhub_v2_3.c +++ b/drivers/gpu/drm/amd/amdgpu/mmhub_v2_3.c @@ -243,7 +243,7 @@ static void mmhub_v2_3_init_cache_regs(struct amdgpu_device *adev) tmp = mmMMVM_L2_CNTL5_DEFAULT; tmp = REG_SET_FIELD(tmp, MMVM_L2_CNTL5, L2_CACHE_SMALLK_FRAGMENT_SIZE, 0); - WREG32_SOC15(GC, 0, mmMMVM_L2_CNTL5, tmp); + WREG32_SOC15(MMHUB, 0, mmMMVM_L2_CNTL5, tmp); } static void mmhub_v2_3_enable_system_domain(struct amdgpu_device *adev) diff --git a/drivers/gpu/drm/amd/amdgpu/mmhub_v3_0.c b/drivers/gpu/drm/amd/amdgpu/mmhub_v3_0.c index a1d26c4d80b8..16cc82215e2e 100644 --- a/drivers/gpu/drm/amd/amdgpu/mmhub_v3_0.c +++ b/drivers/gpu/drm/amd/amdgpu/mmhub_v3_0.c @@ -275,7 +275,7 @@ static void mmhub_v3_0_init_cache_regs(struct amdgpu_device *adev) tmp = regMMVM_L2_CNTL5_DEFAULT; tmp = REG_SET_FIELD(tmp, MMVM_L2_CNTL5, L2_CACHE_SMALLK_FRAGMENT_SIZE, 0); - WREG32_SOC15(GC, 0, regMMVM_L2_CNTL5, tmp); + WREG32_SOC15(MMHUB, 0, regMMVM_L2_CNTL5, tmp); } static void mmhub_v3_0_enable_system_domain(struct amdgpu_device *adev) diff --git a/drivers/gpu/drm/amd/amdgpu/mmhub_v3_0_1.c b/drivers/gpu/drm/amd/amdgpu/mmhub_v3_0_1.c index e8058edc1d10..6bdf2ef0298d 100644 --- a/drivers/gpu/drm/amd/amdgpu/mmhub_v3_0_1.c +++ b/drivers/gpu/drm/amd/amdgpu/mmhub_v3_0_1.c @@ -269,7 +269,7 @@ static void mmhub_v3_0_1_init_cache_regs(struct amdgpu_device *adev) tmp = regMMVM_L2_CNTL5_DEFAULT; tmp = REG_SET_FIELD(tmp, MMVM_L2_CNTL5, L2_CACHE_SMALLK_FRAGMENT_SIZE, 0); - WREG32_SOC15(GC, 0, regMMVM_L2_CNTL5, tmp); + WREG32_SOC15(MMHUB, 0, regMMVM_L2_CNTL5, tmp); } static void mmhub_v3_0_1_enable_system_domain(struct amdgpu_device *adev) diff --git a/drivers/gpu/drm/amd/amdgpu/mmhub_v3_0_2.c b/drivers/gpu/drm/amd/amdgpu/mmhub_v3_0_2.c index 770be0a8f7ce..45465acaa943 100644 --- a/drivers/gpu/drm/amd/amdgpu/mmhub_v3_0_2.c +++ b/drivers/gpu/drm/amd/amdgpu/mmhub_v3_0_2.c @@ -268,7 +268,7 @@ static void mmhub_v3_0_2_init_cache_regs(struct amdgpu_device *adev) tmp = regMMVM_L2_CNTL5_DEFAULT; tmp = REG_SET_FIELD(tmp, MMVM_L2_CNTL5, L2_CACHE_SMALLK_FRAGMENT_SIZE, 0); - WREG32_SOC15(GC, 0, regMMVM_L2_CNTL5, tmp); + WREG32_SOC15(MMHUB, 0, regMMVM_L2_CNTL5, tmp); } static void mmhub_v3_0_2_enable_system_domain(struct amdgpu_device *adev) diff --git a/drivers/gpu/drm/amd/amdgpu/nv.c b/drivers/gpu/drm/amd/amdgpu/nv.c index b3fba8dea63c..6853b93ac82e 100644 --- a/drivers/gpu/drm/amd/amdgpu/nv.c +++ b/drivers/gpu/drm/amd/amdgpu/nv.c @@ -82,10 +82,10 @@ static const struct amdgpu_video_codecs nv_video_codecs_encode = /* Navi1x */ static const struct amdgpu_video_codec_info nv_video_codecs_decode_array[] = { - {codec_info_build(AMDGPU_INFO_VIDEO_CAPS_CODEC_IDX_MPEG2, 4096, 4906, 3)}, - {codec_info_build(AMDGPU_INFO_VIDEO_CAPS_CODEC_IDX_MPEG4, 4096, 4906, 5)}, - {codec_info_build(AMDGPU_INFO_VIDEO_CAPS_CODEC_IDX_MPEG4_AVC, 4096, 4906, 52)}, - {codec_info_build(AMDGPU_INFO_VIDEO_CAPS_CODEC_IDX_VC1, 4096, 4906, 4)}, + {codec_info_build(AMDGPU_INFO_VIDEO_CAPS_CODEC_IDX_MPEG2, 4096, 4096, 3)}, + {codec_info_build(AMDGPU_INFO_VIDEO_CAPS_CODEC_IDX_MPEG4, 4096, 4096, 5)}, + {codec_info_build(AMDGPU_INFO_VIDEO_CAPS_CODEC_IDX_MPEG4_AVC, 4096, 4096, 52)}, + {codec_info_build(AMDGPU_INFO_VIDEO_CAPS_CODEC_IDX_VC1, 4096, 4096, 4)}, {codec_info_build(AMDGPU_INFO_VIDEO_CAPS_CODEC_IDX_HEVC, 8192, 4352, 186)}, {codec_info_build(AMDGPU_INFO_VIDEO_CAPS_CODEC_IDX_JPEG, 4096, 4096, 0)}, {codec_info_build(AMDGPU_INFO_VIDEO_CAPS_CODEC_IDX_VP9, 8192, 4352, 0)}, @@ -100,10 +100,10 @@ static const struct amdgpu_video_codecs nv_video_codecs_decode = /* Sienna Cichlid */ static const struct amdgpu_video_codec_info sc_video_codecs_decode_array[] = { - {codec_info_build(AMDGPU_INFO_VIDEO_CAPS_CODEC_IDX_MPEG2, 4096, 4906, 3)}, - {codec_info_build(AMDGPU_INFO_VIDEO_CAPS_CODEC_IDX_MPEG4, 4096, 4906, 5)}, - {codec_info_build(AMDGPU_INFO_VIDEO_CAPS_CODEC_IDX_MPEG4_AVC, 4096, 4906, 52)}, - {codec_info_build(AMDGPU_INFO_VIDEO_CAPS_CODEC_IDX_VC1, 4096, 4906, 4)}, + {codec_info_build(AMDGPU_INFO_VIDEO_CAPS_CODEC_IDX_MPEG2, 4096, 4096, 3)}, + {codec_info_build(AMDGPU_INFO_VIDEO_CAPS_CODEC_IDX_MPEG4, 4096, 4096, 5)}, + {codec_info_build(AMDGPU_INFO_VIDEO_CAPS_CODEC_IDX_MPEG4_AVC, 4096, 4096, 52)}, + {codec_info_build(AMDGPU_INFO_VIDEO_CAPS_CODEC_IDX_VC1, 4096, 4096, 4)}, {codec_info_build(AMDGPU_INFO_VIDEO_CAPS_CODEC_IDX_HEVC, 8192, 4352, 186)}, {codec_info_build(AMDGPU_INFO_VIDEO_CAPS_CODEC_IDX_JPEG, 4096, 4096, 0)}, {codec_info_build(AMDGPU_INFO_VIDEO_CAPS_CODEC_IDX_VP9, 8192, 4352, 0)}, @@ -125,10 +125,10 @@ static struct amdgpu_video_codec_info sriov_sc_video_codecs_encode_array[] = static struct amdgpu_video_codec_info sriov_sc_video_codecs_decode_array[] = { - {codec_info_build(AMDGPU_INFO_VIDEO_CAPS_CODEC_IDX_MPEG2, 4096, 4906, 3)}, - {codec_info_build(AMDGPU_INFO_VIDEO_CAPS_CODEC_IDX_MPEG4, 4096, 4906, 5)}, - {codec_info_build(AMDGPU_INFO_VIDEO_CAPS_CODEC_IDX_MPEG4_AVC, 4096, 4906, 52)}, - {codec_info_build(AMDGPU_INFO_VIDEO_CAPS_CODEC_IDX_VC1, 4096, 4906, 4)}, + {codec_info_build(AMDGPU_INFO_VIDEO_CAPS_CODEC_IDX_MPEG2, 4096, 4096, 3)}, + {codec_info_build(AMDGPU_INFO_VIDEO_CAPS_CODEC_IDX_MPEG4, 4096, 4096, 5)}, + {codec_info_build(AMDGPU_INFO_VIDEO_CAPS_CODEC_IDX_MPEG4_AVC, 4096, 4096, 52)}, + {codec_info_build(AMDGPU_INFO_VIDEO_CAPS_CODEC_IDX_VC1, 4096, 4096, 4)}, {codec_info_build(AMDGPU_INFO_VIDEO_CAPS_CODEC_IDX_HEVC, 8192, 4352, 186)}, {codec_info_build(AMDGPU_INFO_VIDEO_CAPS_CODEC_IDX_JPEG, 4096, 4096, 0)}, {codec_info_build(AMDGPU_INFO_VIDEO_CAPS_CODEC_IDX_VP9, 8192, 4352, 0)}, @@ -149,7 +149,7 @@ static struct amdgpu_video_codecs sriov_sc_video_codecs_decode = /* Beige Goby*/ static const struct amdgpu_video_codec_info bg_video_codecs_decode_array[] = { - {codec_info_build(AMDGPU_INFO_VIDEO_CAPS_CODEC_IDX_MPEG4_AVC, 4096, 4906, 52)}, + {codec_info_build(AMDGPU_INFO_VIDEO_CAPS_CODEC_IDX_MPEG4_AVC, 4096, 4096, 52)}, {codec_info_build(AMDGPU_INFO_VIDEO_CAPS_CODEC_IDX_HEVC, 8192, 4352, 186)}, {codec_info_build(AMDGPU_INFO_VIDEO_CAPS_CODEC_IDX_VP9, 8192, 4352, 0)}, }; @@ -166,7 +166,7 @@ static const struct amdgpu_video_codecs bg_video_codecs_encode = { /* Yellow Carp*/ static const struct amdgpu_video_codec_info yc_video_codecs_decode_array[] = { - {codec_info_build(AMDGPU_INFO_VIDEO_CAPS_CODEC_IDX_MPEG4_AVC, 4096, 4906, 52)}, + {codec_info_build(AMDGPU_INFO_VIDEO_CAPS_CODEC_IDX_MPEG4_AVC, 4096, 4096, 52)}, {codec_info_build(AMDGPU_INFO_VIDEO_CAPS_CODEC_IDX_HEVC, 8192, 4352, 186)}, {codec_info_build(AMDGPU_INFO_VIDEO_CAPS_CODEC_IDX_VP9, 8192, 4352, 0)}, {codec_info_build(AMDGPU_INFO_VIDEO_CAPS_CODEC_IDX_JPEG, 4096, 4096, 0)}, diff --git a/drivers/gpu/drm/amd/amdgpu/psp_v10_0.c b/drivers/gpu/drm/amd/amdgpu/psp_v10_0.c index ed2293686f0d..9de46fa8f46c 100644 --- a/drivers/gpu/drm/amd/amdgpu/psp_v10_0.c +++ b/drivers/gpu/drm/amd/amdgpu/psp_v10_0.c @@ -126,32 +126,6 @@ out: return err; } -static int psp_v10_0_ring_init(struct psp_context *psp, - enum psp_ring_type ring_type) -{ - int ret = 0; - struct psp_ring *ring; - struct amdgpu_device *adev = psp->adev; - - ring = &psp->km_ring; - - ring->ring_type = ring_type; - - /* allocate 4k Page of Local Frame Buffer memory for ring */ - ring->ring_size = 0x1000; - ret = amdgpu_bo_create_kernel(adev, ring->ring_size, PAGE_SIZE, - AMDGPU_GEM_DOMAIN_VRAM, - &adev->firmware.rbuf, - &ring->ring_mem_mc_addr, - (void **)&ring->ring_mem); - if (ret) { - ring->ring_size = 0; - return ret; - } - - return 0; -} - static int psp_v10_0_ring_create(struct psp_context *psp, enum psp_ring_type ring_type) { @@ -245,7 +219,6 @@ static void psp_v10_0_ring_set_wptr(struct psp_context *psp, uint32_t value) static const struct psp_funcs psp_v10_0_funcs = { .init_microcode = psp_v10_0_init_microcode, - .ring_init = psp_v10_0_ring_init, .ring_create = psp_v10_0_ring_create, .ring_stop = psp_v10_0_ring_stop, .ring_destroy = psp_v10_0_ring_destroy, diff --git a/drivers/gpu/drm/amd/amdgpu/psp_v11_0.c b/drivers/gpu/drm/amd/amdgpu/psp_v11_0.c index 9518b4394a6e..bd3e3e23a939 100644 --- a/drivers/gpu/drm/amd/amdgpu/psp_v11_0.c +++ b/drivers/gpu/drm/amd/amdgpu/psp_v11_0.c @@ -360,32 +360,6 @@ static int psp_v11_0_bootloader_load_sos(struct psp_context *psp) return ret; } -static int psp_v11_0_ring_init(struct psp_context *psp, - enum psp_ring_type ring_type) -{ - int ret = 0; - struct psp_ring *ring; - struct amdgpu_device *adev = psp->adev; - - ring = &psp->km_ring; - - ring->ring_type = ring_type; - - /* allocate 4k Page of Local Frame Buffer memory for ring */ - ring->ring_size = 0x1000; - ret = amdgpu_bo_create_kernel(adev, ring->ring_size, PAGE_SIZE, - AMDGPU_GEM_DOMAIN_VRAM, - &adev->firmware.rbuf, - &ring->ring_mem_mc_addr, - (void **)&ring->ring_mem); - if (ret) { - ring->ring_size = 0; - return ret; - } - - return 0; -} - static int psp_v11_0_ring_stop(struct psp_context *psp, enum psp_ring_type ring_type) { @@ -779,7 +753,6 @@ static const struct psp_funcs psp_v11_0_funcs = { .bootloader_load_spl = psp_v11_0_bootloader_load_spl, .bootloader_load_sysdrv = psp_v11_0_bootloader_load_sysdrv, .bootloader_load_sos = psp_v11_0_bootloader_load_sos, - .ring_init = psp_v11_0_ring_init, .ring_create = psp_v11_0_ring_create, .ring_stop = psp_v11_0_ring_stop, .ring_destroy = psp_v11_0_ring_destroy, diff --git a/drivers/gpu/drm/amd/amdgpu/psp_v11_0_8.c b/drivers/gpu/drm/amd/amdgpu/psp_v11_0_8.c index ff13e1beb49b..5697760a819b 100644 --- a/drivers/gpu/drm/amd/amdgpu/psp_v11_0_8.c +++ b/drivers/gpu/drm/amd/amdgpu/psp_v11_0_8.c @@ -28,32 +28,6 @@ #include "mp/mp_11_0_8_offset.h" -static int psp_v11_0_8_ring_init(struct psp_context *psp, - enum psp_ring_type ring_type) -{ - int ret = 0; - struct psp_ring *ring; - struct amdgpu_device *adev = psp->adev; - - ring = &psp->km_ring; - - ring->ring_type = ring_type; - - /* allocate 4k Page of Local Frame Buffer memory for ring */ - ring->ring_size = 0x1000; - ret = amdgpu_bo_create_kernel(adev, ring->ring_size, PAGE_SIZE, - AMDGPU_GEM_DOMAIN_VRAM, - &adev->firmware.rbuf, - &ring->ring_mem_mc_addr, - (void **)&ring->ring_mem); - if (ret) { - ring->ring_size = 0; - return ret; - } - - return 0; -} - static int psp_v11_0_8_ring_stop(struct psp_context *psp, enum psp_ring_type ring_type) { @@ -194,7 +168,6 @@ static void psp_v11_0_8_ring_set_wptr(struct psp_context *psp, uint32_t value) } static const struct psp_funcs psp_v11_0_8_funcs = { - .ring_init = psp_v11_0_8_ring_init, .ring_create = psp_v11_0_8_ring_create, .ring_stop = psp_v11_0_8_ring_stop, .ring_destroy = psp_v11_0_8_ring_destroy, diff --git a/drivers/gpu/drm/amd/amdgpu/psp_v12_0.c b/drivers/gpu/drm/amd/amdgpu/psp_v12_0.c index 0b2ac418e4ac..8ed2281b6557 100644 --- a/drivers/gpu/drm/amd/amdgpu/psp_v12_0.c +++ b/drivers/gpu/drm/amd/amdgpu/psp_v12_0.c @@ -236,34 +236,6 @@ static void psp_v12_0_reroute_ih(struct psp_context *psp) 0x80000000, 0x8000FFFF, false); } -static int psp_v12_0_ring_init(struct psp_context *psp, - enum psp_ring_type ring_type) -{ - int ret = 0; - struct psp_ring *ring; - struct amdgpu_device *adev = psp->adev; - - psp_v12_0_reroute_ih(psp); - - ring = &psp->km_ring; - - ring->ring_type = ring_type; - - /* allocate 4k Page of Local Frame Buffer memory for ring */ - ring->ring_size = 0x1000; - ret = amdgpu_bo_create_kernel(adev, ring->ring_size, PAGE_SIZE, - AMDGPU_GEM_DOMAIN_VRAM, - &adev->firmware.rbuf, - &ring->ring_mem_mc_addr, - (void **)&ring->ring_mem); - if (ret) { - ring->ring_size = 0; - return ret; - } - - return 0; -} - static int psp_v12_0_ring_create(struct psp_context *psp, enum psp_ring_type ring_type) { @@ -272,6 +244,8 @@ static int psp_v12_0_ring_create(struct psp_context *psp, struct psp_ring *ring = &psp->km_ring; struct amdgpu_device *adev = psp->adev; + psp_v12_0_reroute_ih(psp); + if (amdgpu_sriov_vf(psp->adev)) { /* Write low address of the ring to C2PMSG_102 */ psp_ring_reg = lower_32_bits(ring->ring_mem_mc_addr); @@ -425,7 +399,6 @@ static const struct psp_funcs psp_v12_0_funcs = { .init_microcode = psp_v12_0_init_microcode, .bootloader_load_sysdrv = psp_v12_0_bootloader_load_sysdrv, .bootloader_load_sos = psp_v12_0_bootloader_load_sos, - .ring_init = psp_v12_0_ring_init, .ring_create = psp_v12_0_ring_create, .ring_stop = psp_v12_0_ring_stop, .ring_destroy = psp_v12_0_ring_destroy, diff --git a/drivers/gpu/drm/amd/amdgpu/psp_v13_0.c b/drivers/gpu/drm/amd/amdgpu/psp_v13_0.c index 88f9b327183a..e6a26a7e5e5e 100644 --- a/drivers/gpu/drm/amd/amdgpu/psp_v13_0.c +++ b/drivers/gpu/drm/amd/amdgpu/psp_v13_0.c @@ -46,6 +46,8 @@ MODULE_FIRMWARE("amdgpu/psp_13_0_7_sos.bin"); MODULE_FIRMWARE("amdgpu/psp_13_0_7_ta.bin"); MODULE_FIRMWARE("amdgpu/psp_13_0_10_sos.bin"); MODULE_FIRMWARE("amdgpu/psp_13_0_10_ta.bin"); +MODULE_FIRMWARE("amdgpu/psp_13_0_11_toc.bin"); +MODULE_FIRMWARE("amdgpu/psp_13_0_11_ta.bin"); /* For large FW files the time to complete can be very long */ #define USBC_PD_POLLING_LIMIT_S 240 @@ -102,6 +104,7 @@ static int psp_v13_0_init_microcode(struct psp_context *psp) case IP_VERSION(13, 0, 3): case IP_VERSION(13, 0, 5): case IP_VERSION(13, 0, 8): + case IP_VERSION(13, 0, 11): err = psp_init_toc_microcode(psp, chip_name); if (err) return err; @@ -268,32 +271,6 @@ static int psp_v13_0_bootloader_load_sos(struct psp_context *psp) return ret; } -static int psp_v13_0_ring_init(struct psp_context *psp, - enum psp_ring_type ring_type) -{ - int ret = 0; - struct psp_ring *ring; - struct amdgpu_device *adev = psp->adev; - - ring = &psp->km_ring; - - ring->ring_type = ring_type; - - /* allocate 4k Page of Local Frame Buffer memory for ring */ - ring->ring_size = 0x1000; - ret = amdgpu_bo_create_kernel(adev, ring->ring_size, PAGE_SIZE, - AMDGPU_GEM_DOMAIN_VRAM, - &adev->firmware.rbuf, - &ring->ring_mem_mc_addr, - (void **)&ring->ring_mem); - if (ret) { - ring->ring_size = 0; - return ret; - } - - return 0; -} - static int psp_v13_0_ring_stop(struct psp_context *psp, enum psp_ring_type ring_type) { @@ -729,7 +706,6 @@ static const struct psp_funcs psp_v13_0_funcs = { .bootloader_load_dbg_drv = psp_v13_0_bootloader_load_dbg_drv, .bootloader_load_ras_drv = psp_v13_0_bootloader_load_ras_drv, .bootloader_load_sos = psp_v13_0_bootloader_load_sos, - .ring_init = psp_v13_0_ring_init, .ring_create = psp_v13_0_ring_create, .ring_stop = psp_v13_0_ring_stop, .ring_destroy = psp_v13_0_ring_destroy, diff --git a/drivers/gpu/drm/amd/amdgpu/psp_v13_0_4.c b/drivers/gpu/drm/amd/amdgpu/psp_v13_0_4.c index 321089dfa7db..9d4e24e518e8 100644 --- a/drivers/gpu/drm/amd/amdgpu/psp_v13_0_4.c +++ b/drivers/gpu/drm/amd/amdgpu/psp_v13_0_4.c @@ -199,32 +199,6 @@ static int psp_v13_0_4_bootloader_load_sos(struct psp_context *psp) return ret; } -static int psp_v13_0_4_ring_init(struct psp_context *psp, - enum psp_ring_type ring_type) -{ - int ret = 0; - struct psp_ring *ring; - struct amdgpu_device *adev = psp->adev; - - ring = &psp->km_ring; - - ring->ring_type = ring_type; - - /* allocate 4k Page of Local Frame Buffer memory for ring */ - ring->ring_size = 0x1000; - ret = amdgpu_bo_create_kernel(adev, ring->ring_size, PAGE_SIZE, - AMDGPU_GEM_DOMAIN_VRAM, - &adev->firmware.rbuf, - &ring->ring_mem_mc_addr, - (void **)&ring->ring_mem); - if (ret) { - ring->ring_size = 0; - return ret; - } - - return 0; -} - static int psp_v13_0_4_ring_stop(struct psp_context *psp, enum psp_ring_type ring_type) { @@ -373,7 +347,6 @@ static const struct psp_funcs psp_v13_0_4_funcs = { .bootloader_load_intf_drv = psp_v13_0_4_bootloader_load_intf_drv, .bootloader_load_dbg_drv = psp_v13_0_4_bootloader_load_dbg_drv, .bootloader_load_sos = psp_v13_0_4_bootloader_load_sos, - .ring_init = psp_v13_0_4_ring_init, .ring_create = psp_v13_0_4_ring_create, .ring_stop = psp_v13_0_4_ring_stop, .ring_destroy = psp_v13_0_4_ring_destroy, diff --git a/drivers/gpu/drm/amd/amdgpu/psp_v3_1.c b/drivers/gpu/drm/amd/amdgpu/psp_v3_1.c index 01f3bcc62a6c..157147c6c94e 100644 --- a/drivers/gpu/drm/amd/amdgpu/psp_v3_1.c +++ b/drivers/gpu/drm/amd/amdgpu/psp_v3_1.c @@ -160,32 +160,6 @@ static int psp_v3_1_bootloader_load_sos(struct psp_context *psp) return ret; } -static int psp_v3_1_ring_init(struct psp_context *psp, - enum psp_ring_type ring_type) -{ - int ret = 0; - struct psp_ring *ring; - struct amdgpu_device *adev = psp->adev; - - ring = &psp->km_ring; - - ring->ring_type = ring_type; - - /* allocate 4k Page of Local Frame Buffer memory for ring */ - ring->ring_size = 0x1000; - ret = amdgpu_bo_create_kernel(adev, ring->ring_size, PAGE_SIZE, - AMDGPU_GEM_DOMAIN_VRAM, - &adev->firmware.rbuf, - &ring->ring_mem_mc_addr, - (void **)&ring->ring_mem); - if (ret) { - ring->ring_size = 0; - return ret; - } - - return 0; -} - static void psp_v3_1_reroute_ih(struct psp_context *psp) { struct amdgpu_device *adev = psp->adev; @@ -401,7 +375,6 @@ static const struct psp_funcs psp_v3_1_funcs = { .init_microcode = psp_v3_1_init_microcode, .bootloader_load_sysdrv = psp_v3_1_bootloader_load_sysdrv, .bootloader_load_sos = psp_v3_1_bootloader_load_sos, - .ring_init = psp_v3_1_ring_init, .ring_create = psp_v3_1_ring_create, .ring_stop = psp_v3_1_ring_stop, .ring_destroy = psp_v3_1_ring_destroy, diff --git a/drivers/gpu/drm/amd/amdgpu/sdma_v6_0.c b/drivers/gpu/drm/amd/amdgpu/sdma_v6_0.c index da3beb0bf2fa..049c26a45d85 100644 --- a/drivers/gpu/drm/amd/amdgpu/sdma_v6_0.c +++ b/drivers/gpu/drm/amd/amdgpu/sdma_v6_0.c @@ -455,6 +455,9 @@ static void sdma_v6_0_enable(struct amdgpu_device *adev, bool enable) sdma_v6_0_rlc_stop(adev); } + if (amdgpu_sriov_vf(adev)) + return; + for (i = 0; i < adev->sdma.num_instances; i++) { f32_cntl = RREG32_SOC15_IP(GC, sdma_v6_0_get_reg_offset(adev, i, regSDMA0_F32_CNTL)); f32_cntl = REG_SET_FIELD(f32_cntl, SDMA0_F32_CNTL, HALT, enable ? 0 : 1); @@ -1523,6 +1526,7 @@ static const struct amdgpu_ring_funcs sdma_v6_0_ring_funcs = { .align_mask = 0xf, .nop = SDMA_PKT_NOP_HEADER_OP(SDMA_OP_NOP), .support_64bit_ptrs = true, + .secure_submission_supported = true, .vmhub = AMDGPU_GFXHUB_0, .get_rptr = sdma_v6_0_ring_get_rptr, .get_wptr = sdma_v6_0_ring_get_wptr, diff --git a/drivers/gpu/drm/amd/amdgpu/si_dma.c b/drivers/gpu/drm/amd/amdgpu/si_dma.c index 4d5e718540aa..abca8b529721 100644 --- a/drivers/gpu/drm/amd/amdgpu/si_dma.c +++ b/drivers/gpu/drm/amd/amdgpu/si_dma.c @@ -112,14 +112,12 @@ static void si_dma_ring_emit_fence(struct amdgpu_ring *ring, u64 addr, u64 seq, static void si_dma_stop(struct amdgpu_device *adev) { - struct amdgpu_ring *ring; u32 rb_cntl; unsigned i; amdgpu_sdma_unset_buffer_funcs_helper(adev); for (i = 0; i < adev->sdma.num_instances; i++) { - ring = &adev->sdma.instance[i].ring; /* dma0 */ rb_cntl = RREG32(DMA_RB_CNTL + sdma_offsets[i]); rb_cntl &= ~DMA_RB_ENABLE; diff --git a/drivers/gpu/drm/amd/amdgpu/soc15.c b/drivers/gpu/drm/amd/amdgpu/soc15.c index e3b2b6b4f1a6..7cd17dda32ce 100644 --- a/drivers/gpu/drm/amd/amdgpu/soc15.c +++ b/drivers/gpu/drm/amd/amdgpu/soc15.c @@ -103,10 +103,10 @@ static const struct amdgpu_video_codecs vega_video_codecs_encode = /* Vega */ static const struct amdgpu_video_codec_info vega_video_codecs_decode_array[] = { - {codec_info_build(AMDGPU_INFO_VIDEO_CAPS_CODEC_IDX_MPEG2, 4096, 4906, 3)}, - {codec_info_build(AMDGPU_INFO_VIDEO_CAPS_CODEC_IDX_MPEG4, 4096, 4906, 5)}, - {codec_info_build(AMDGPU_INFO_VIDEO_CAPS_CODEC_IDX_MPEG4_AVC, 4096, 4906, 52)}, - {codec_info_build(AMDGPU_INFO_VIDEO_CAPS_CODEC_IDX_VC1, 4096, 4906, 4)}, + {codec_info_build(AMDGPU_INFO_VIDEO_CAPS_CODEC_IDX_MPEG2, 4096, 4096, 3)}, + {codec_info_build(AMDGPU_INFO_VIDEO_CAPS_CODEC_IDX_MPEG4, 4096, 4096, 5)}, + {codec_info_build(AMDGPU_INFO_VIDEO_CAPS_CODEC_IDX_MPEG4_AVC, 4096, 4096, 52)}, + {codec_info_build(AMDGPU_INFO_VIDEO_CAPS_CODEC_IDX_VC1, 4096, 4096, 4)}, {codec_info_build(AMDGPU_INFO_VIDEO_CAPS_CODEC_IDX_HEVC, 4096, 4096, 186)}, {codec_info_build(AMDGPU_INFO_VIDEO_CAPS_CODEC_IDX_JPEG, 4096, 4096, 0)}, }; @@ -120,10 +120,10 @@ static const struct amdgpu_video_codecs vega_video_codecs_decode = /* Raven */ static const struct amdgpu_video_codec_info rv_video_codecs_decode_array[] = { - {codec_info_build(AMDGPU_INFO_VIDEO_CAPS_CODEC_IDX_MPEG2, 4096, 4906, 3)}, - {codec_info_build(AMDGPU_INFO_VIDEO_CAPS_CODEC_IDX_MPEG4, 4096, 4906, 5)}, - {codec_info_build(AMDGPU_INFO_VIDEO_CAPS_CODEC_IDX_MPEG4_AVC, 4096, 4906, 52)}, - {codec_info_build(AMDGPU_INFO_VIDEO_CAPS_CODEC_IDX_VC1, 4096, 4906, 4)}, + {codec_info_build(AMDGPU_INFO_VIDEO_CAPS_CODEC_IDX_MPEG2, 4096, 4096, 3)}, + {codec_info_build(AMDGPU_INFO_VIDEO_CAPS_CODEC_IDX_MPEG4, 4096, 4096, 5)}, + {codec_info_build(AMDGPU_INFO_VIDEO_CAPS_CODEC_IDX_MPEG4_AVC, 4096, 4096, 52)}, + {codec_info_build(AMDGPU_INFO_VIDEO_CAPS_CODEC_IDX_VC1, 4096, 4096, 4)}, {codec_info_build(AMDGPU_INFO_VIDEO_CAPS_CODEC_IDX_HEVC, 4096, 4096, 186)}, {codec_info_build(AMDGPU_INFO_VIDEO_CAPS_CODEC_IDX_JPEG, 4096, 4096, 0)}, {codec_info_build(AMDGPU_INFO_VIDEO_CAPS_CODEC_IDX_VP9, 4096, 4096, 0)}, @@ -138,10 +138,10 @@ static const struct amdgpu_video_codecs rv_video_codecs_decode = /* Renoir, Arcturus */ static const struct amdgpu_video_codec_info rn_video_codecs_decode_array[] = { - {codec_info_build(AMDGPU_INFO_VIDEO_CAPS_CODEC_IDX_MPEG2, 4096, 4906, 3)}, - {codec_info_build(AMDGPU_INFO_VIDEO_CAPS_CODEC_IDX_MPEG4, 4096, 4906, 5)}, - {codec_info_build(AMDGPU_INFO_VIDEO_CAPS_CODEC_IDX_MPEG4_AVC, 4096, 4906, 52)}, - {codec_info_build(AMDGPU_INFO_VIDEO_CAPS_CODEC_IDX_VC1, 4096, 4906, 4)}, + {codec_info_build(AMDGPU_INFO_VIDEO_CAPS_CODEC_IDX_MPEG2, 4096, 4096, 3)}, + {codec_info_build(AMDGPU_INFO_VIDEO_CAPS_CODEC_IDX_MPEG4, 4096, 4096, 5)}, + {codec_info_build(AMDGPU_INFO_VIDEO_CAPS_CODEC_IDX_MPEG4_AVC, 4096, 4096, 52)}, + {codec_info_build(AMDGPU_INFO_VIDEO_CAPS_CODEC_IDX_VC1, 4096, 4096, 4)}, {codec_info_build(AMDGPU_INFO_VIDEO_CAPS_CODEC_IDX_HEVC, 8192, 4352, 186)}, {codec_info_build(AMDGPU_INFO_VIDEO_CAPS_CODEC_IDX_JPEG, 4096, 4096, 0)}, {codec_info_build(AMDGPU_INFO_VIDEO_CAPS_CODEC_IDX_VP9, 8192, 4352, 0)}, diff --git a/drivers/gpu/drm/amd/amdgpu/soc15d.h b/drivers/gpu/drm/amd/amdgpu/soc15d.h index 799925d22fc8..2357ff39323f 100644 --- a/drivers/gpu/drm/amd/amdgpu/soc15d.h +++ b/drivers/gpu/drm/amd/amdgpu/soc15d.h @@ -162,6 +162,7 @@ * 2 - Bypass */ #define INDIRECT_BUFFER_PRE_ENB(x) ((x) << 21) +#define INDIRECT_BUFFER_PRE_RESUME(x) ((x) << 30) #define PACKET3_COPY_DATA 0x40 #define PACKET3_PFP_SYNC_ME 0x42 #define PACKET3_COND_WRITE 0x45 @@ -184,6 +185,7 @@ #define EOP_TC_ACTION_EN (1 << 17) /* L2 */ #define EOP_TC_NC_ACTION_EN (1 << 19) #define EOP_TC_MD_ACTION_EN (1 << 21) /* L2 metadata */ +#define EOP_EXEC (1 << 28) /* For Trailing Fence */ #define DATA_SEL(x) ((x) << 29) /* 0 - discard diff --git a/drivers/gpu/drm/amd/amdgpu/soc21.c b/drivers/gpu/drm/amd/amdgpu/soc21.c index e08044008186..5562670b7b52 100644 --- a/drivers/gpu/drm/amd/amdgpu/soc21.c +++ b/drivers/gpu/drm/amd/amdgpu/soc21.c @@ -43,6 +43,7 @@ #include "soc15.h" #include "soc15_common.h" #include "soc21.h" +#include "mxgpu_nv.h" static const struct amd_ip_funcs soc21_common_ip_funcs; @@ -61,7 +62,7 @@ static const struct amdgpu_video_codecs vcn_4_0_0_video_codecs_encode = static const struct amdgpu_video_codec_info vcn_4_0_0_video_codecs_decode_array[] = { - {codec_info_build(AMDGPU_INFO_VIDEO_CAPS_CODEC_IDX_MPEG4_AVC, 4096, 4906, 52)}, + {codec_info_build(AMDGPU_INFO_VIDEO_CAPS_CODEC_IDX_MPEG4_AVC, 4096, 4096, 52)}, {codec_info_build(AMDGPU_INFO_VIDEO_CAPS_CODEC_IDX_HEVC, 8192, 4352, 186)}, {codec_info_build(AMDGPU_INFO_VIDEO_CAPS_CODEC_IDX_JPEG, 4096, 4096, 0)}, {codec_info_build(AMDGPU_INFO_VIDEO_CAPS_CODEC_IDX_VP9, 8192, 4352, 0)}, @@ -322,8 +323,10 @@ soc21_asic_reset_method(struct amdgpu_device *adev) switch (adev->ip_versions[MP1_HWIP][0]) { case IP_VERSION(13, 0, 0): case IP_VERSION(13, 0, 7): + case IP_VERSION(13, 0, 10): return AMD_RESET_METHOD_MODE1; case IP_VERSION(13, 0, 4): + case IP_VERSION(13, 0, 11): return AMD_RESET_METHOD_MODE2; default: if (amdgpu_dpm_is_baco_supported(adev)) @@ -584,10 +587,6 @@ static int soc21_common_early_init(void *handle) AMD_PG_SUPPORT_JPEG | AMD_PG_SUPPORT_ATHUB | AMD_PG_SUPPORT_MMHUB; - if (amdgpu_sriov_vf(adev)) { - adev->cg_flags = 0; - adev->pg_flags = 0; - } adev->external_rev_id = adev->rev_id + 0x1; // TODO: need update break; case IP_VERSION(11, 0, 2): @@ -645,28 +644,64 @@ static int soc21_common_early_init(void *handle) adev->pg_flags = AMD_PG_SUPPORT_VCN | AMD_PG_SUPPORT_VCN_DPG | AMD_PG_SUPPORT_JPEG; - if (amdgpu_sriov_vf(adev)) { - /* hypervisor control CG and PG enablement */ - adev->cg_flags = 0; - adev->pg_flags = 0; - } adev->external_rev_id = adev->rev_id + 0x20; break; + case IP_VERSION(11, 0, 4): + adev->cg_flags = + AMD_CG_SUPPORT_GFX_CGCG | + AMD_CG_SUPPORT_GFX_CGLS | + AMD_CG_SUPPORT_GFX_MGCG | + AMD_CG_SUPPORT_GFX_FGCG | + AMD_CG_SUPPORT_REPEATER_FGCG | + AMD_CG_SUPPORT_GFX_PERF_CLK | + AMD_CG_SUPPORT_MC_MGCG | + AMD_CG_SUPPORT_MC_LS | + AMD_CG_SUPPORT_HDP_MGCG | + AMD_CG_SUPPORT_HDP_LS | + AMD_CG_SUPPORT_ATHUB_MGCG | + AMD_CG_SUPPORT_ATHUB_LS | + AMD_CG_SUPPORT_IH_CG | + AMD_CG_SUPPORT_BIF_MGCG | + AMD_CG_SUPPORT_BIF_LS | + AMD_CG_SUPPORT_VCN_MGCG | + AMD_CG_SUPPORT_JPEG_MGCG; + adev->pg_flags = AMD_PG_SUPPORT_VCN | + AMD_PG_SUPPORT_VCN_DPG | + AMD_PG_SUPPORT_GFX_PG | + AMD_PG_SUPPORT_JPEG; + adev->external_rev_id = adev->rev_id + 0x1; + break; + default: /* FIXME: not supported yet */ return -EINVAL; } + if (amdgpu_sriov_vf(adev)) { + amdgpu_virt_init_setting(adev); + xgpu_nv_mailbox_set_irq_funcs(adev); + } + return 0; } static int soc21_common_late_init(void *handle) { + struct amdgpu_device *adev = (struct amdgpu_device *)handle; + + if (amdgpu_sriov_vf(adev)) + xgpu_nv_mailbox_get_irq(adev); + return 0; } static int soc21_common_sw_init(void *handle) { + struct amdgpu_device *adev = (struct amdgpu_device *)handle; + + if (amdgpu_sriov_vf(adev)) + xgpu_nv_mailbox_add_irq_id(adev); + return 0; } @@ -704,6 +739,9 @@ static int soc21_common_hw_fini(void *handle) /* disable the doorbell aperture */ soc21_enable_doorbell_aperture(adev, false); + if (amdgpu_sriov_vf(adev)) + xgpu_nv_mailbox_put_irq(adev); + return 0; } diff --git a/drivers/gpu/drm/amd/amdgpu/umc_v6_7.c b/drivers/gpu/drm/amd/amdgpu/umc_v6_7.c index 5d5d031c9e7d..72fd963f178b 100644 --- a/drivers/gpu/drm/amd/amdgpu/umc_v6_7.c +++ b/drivers/gpu/drm/amd/amdgpu/umc_v6_7.c @@ -187,9 +187,9 @@ static void umc_v6_7_ecc_info_query_ras_error_count(struct amdgpu_device *adev, } } -static void umc_v6_7_convert_error_address(struct amdgpu_device *adev, - struct ras_err_data *err_data, uint64_t err_addr, - uint32_t ch_inst, uint32_t umc_inst) +void umc_v6_7_convert_error_address(struct amdgpu_device *adev, + struct ras_err_data *err_data, uint64_t err_addr, + uint32_t ch_inst, uint32_t umc_inst) { uint32_t channel_index; uint64_t soc_pa, retired_page, column; @@ -553,5 +553,4 @@ struct amdgpu_umc_ras umc_v6_7_ras = { .query_ras_poison_mode = umc_v6_7_query_ras_poison_mode, .ecc_info_query_ras_error_count = umc_v6_7_ecc_info_query_ras_error_count, .ecc_info_query_ras_error_address = umc_v6_7_ecc_info_query_ras_error_address, - .convert_ras_error_address = umc_v6_7_convert_error_address, }; diff --git a/drivers/gpu/drm/amd/amdgpu/umc_v6_7.h b/drivers/gpu/drm/amd/amdgpu/umc_v6_7.h index fe41ed2f5945..105245d5b6e5 100644 --- a/drivers/gpu/drm/amd/amdgpu/umc_v6_7.h +++ b/drivers/gpu/drm/amd/amdgpu/umc_v6_7.h @@ -71,5 +71,7 @@ extern const uint32_t umc_v6_7_channel_idx_tbl_second[UMC_V6_7_UMC_INSTANCE_NUM][UMC_V6_7_CHANNEL_INSTANCE_NUM]; extern const uint32_t umc_v6_7_channel_idx_tbl_first[UMC_V6_7_UMC_INSTANCE_NUM][UMC_V6_7_CHANNEL_INSTANCE_NUM]; - +void umc_v6_7_convert_error_address(struct amdgpu_device *adev, + struct ras_err_data *err_data, uint64_t err_addr, + uint32_t ch_inst, uint32_t umc_inst); #endif diff --git a/drivers/gpu/drm/amd/amdgpu/umc_v8_10.c b/drivers/gpu/drm/amd/amdgpu/umc_v8_10.c index 91235df54e22..b7da4528cf0a 100644 --- a/drivers/gpu/drm/amd/amdgpu/umc_v8_10.c +++ b/drivers/gpu/drm/amd/amdgpu/umc_v8_10.c @@ -46,6 +46,16 @@ const struct channelnum_map_colbit umc_v8_10_channelnum_map_colbit_table[] = { }; const uint32_t + umc_v8_10_channel_idx_tbl_ext0[] + [UMC_V8_10_UMC_INSTANCE_NUM] + [UMC_V8_10_CHANNEL_INSTANCE_NUM] = { + {{1, 5}, {7, 3}}, + {{14, 15}, {13, 12}}, + {{10, 11}, {9, 8}}, + {{6, 2}, {0, 4}} + }; + +const uint32_t umc_v8_10_channel_idx_tbl[] [UMC_V8_10_UMC_INSTANCE_NUM] [UMC_V8_10_CHANNEL_INSTANCE_NUM] = { diff --git a/drivers/gpu/drm/amd/amdgpu/umc_v8_10.h b/drivers/gpu/drm/amd/amdgpu/umc_v8_10.h index 849ede88e111..25eaf4af5fcf 100644 --- a/drivers/gpu/drm/amd/amdgpu/umc_v8_10.h +++ b/drivers/gpu/drm/amd/amdgpu/umc_v8_10.h @@ -66,5 +66,9 @@ extern const uint32_t [UMC_V8_10_UMC_INSTANCE_NUM] [UMC_V8_10_CHANNEL_INSTANCE_NUM]; +extern const uint32_t + umc_v8_10_channel_idx_tbl_ext0[] + [UMC_V8_10_UMC_INSTANCE_NUM] + [UMC_V8_10_CHANNEL_INSTANCE_NUM]; #endif diff --git a/drivers/gpu/drm/amd/amdgpu/uvd_v6_0.c b/drivers/gpu/drm/amd/amdgpu/uvd_v6_0.c index 375c440957dc..5fe872f4bea7 100644 --- a/drivers/gpu/drm/amd/amdgpu/uvd_v6_0.c +++ b/drivers/gpu/drm/amd/amdgpu/uvd_v6_0.c @@ -216,8 +216,8 @@ static int uvd_v6_0_enc_get_create_msg(struct amdgpu_ring *ring, uint32_t handle uint64_t addr; int i, r; - r = amdgpu_job_alloc_with_ib(ring->adev, ib_size_dw * 4, - AMDGPU_IB_POOL_DIRECT, &job); + r = amdgpu_job_alloc_with_ib(ring->adev, NULL, NULL, ib_size_dw * 4, + AMDGPU_IB_POOL_DIRECT, &job); if (r) return r; @@ -280,8 +280,8 @@ static int uvd_v6_0_enc_get_destroy_msg(struct amdgpu_ring *ring, uint64_t addr; int i, r; - r = amdgpu_job_alloc_with_ib(ring->adev, ib_size_dw * 4, - AMDGPU_IB_POOL_DIRECT, &job); + r = amdgpu_job_alloc_with_ib(ring->adev, NULL, NULL, ib_size_dw * 4, + AMDGPU_IB_POOL_DIRECT, &job); if (r) return r; diff --git a/drivers/gpu/drm/amd/amdgpu/uvd_v7_0.c b/drivers/gpu/drm/amd/amdgpu/uvd_v7_0.c index e668b3baa8c6..e407be6cb63c 100644 --- a/drivers/gpu/drm/amd/amdgpu/uvd_v7_0.c +++ b/drivers/gpu/drm/amd/amdgpu/uvd_v7_0.c @@ -213,7 +213,7 @@ static int uvd_v7_0_enc_ring_test_ring(struct amdgpu_ring *ring) * * Open up a stream for HW test */ -static int uvd_v7_0_enc_get_create_msg(struct amdgpu_ring *ring, uint32_t handle, +static int uvd_v7_0_enc_get_create_msg(struct amdgpu_ring *ring, u32 handle, struct amdgpu_bo *bo, struct dma_fence **fence) { @@ -224,8 +224,8 @@ static int uvd_v7_0_enc_get_create_msg(struct amdgpu_ring *ring, uint32_t handle uint64_t addr; int i, r; - r = amdgpu_job_alloc_with_ib(ring->adev, ib_size_dw * 4, - AMDGPU_IB_POOL_DIRECT, &job); + r = amdgpu_job_alloc_with_ib(ring->adev, NULL, NULL, ib_size_dw * 4, + AMDGPU_IB_POOL_DIRECT, &job); if (r) return r; @@ -276,7 +276,7 @@ err: * * Close up a stream for HW test or if userspace failed to do so */ -static int uvd_v7_0_enc_get_destroy_msg(struct amdgpu_ring *ring, uint32_t handle, +static int uvd_v7_0_enc_get_destroy_msg(struct amdgpu_ring *ring, u32 handle, struct amdgpu_bo *bo, struct dma_fence **fence) { @@ -287,8 +287,8 @@ static int uvd_v7_0_enc_get_destroy_msg(struct amdgpu_ring *ring, uint32_t handl uint64_t addr; int i, r; - r = amdgpu_job_alloc_with_ib(ring->adev, ib_size_dw * 4, - AMDGPU_IB_POOL_DIRECT, &job); + r = amdgpu_job_alloc_with_ib(ring->adev, NULL, NULL, ib_size_dw * 4, + AMDGPU_IB_POOL_DIRECT, &job); if (r) return r; diff --git a/drivers/gpu/drm/amd/amdgpu/vcn_v2_5.c b/drivers/gpu/drm/amd/amdgpu/vcn_v2_5.c index 8a7006d62a87..ec87b00f2e05 100644 --- a/drivers/gpu/drm/amd/amdgpu/vcn_v2_5.c +++ b/drivers/gpu/drm/amd/amdgpu/vcn_v2_5.c @@ -770,6 +770,33 @@ static void vcn_v2_5_enable_clock_gating(struct amdgpu_device *adev) } } +static void vcn_v2_6_enable_ras(struct amdgpu_device *adev, int inst_idx, + bool indirect) +{ + uint32_t tmp; + + if (adev->ip_versions[UVD_HWIP][0] != IP_VERSION(2, 6, 0)) + return; + + tmp = VCN_RAS_CNTL__VCPU_VCODEC_REARM_MASK | + VCN_RAS_CNTL__VCPU_VCODEC_IH_EN_MASK | + VCN_RAS_CNTL__VCPU_VCODEC_PMI_EN_MASK | + VCN_RAS_CNTL__VCPU_VCODEC_STALL_EN_MASK; + WREG32_SOC15_DPG_MODE(inst_idx, + SOC15_DPG_MODE_OFFSET(VCN, 0, mmVCN_RAS_CNTL), + tmp, 0, indirect); + + tmp = UVD_VCPU_INT_EN__RASCNTL_VCPU_VCODEC_EN_MASK; + WREG32_SOC15_DPG_MODE(inst_idx, + SOC15_DPG_MODE_OFFSET(VCN, 0, mmUVD_VCPU_INT_EN), + tmp, 0, indirect); + + tmp = UVD_SYS_INT_EN__RASCNTL_VCPU_VCODEC_EN_MASK; + WREG32_SOC15_DPG_MODE(inst_idx, + SOC15_DPG_MODE_OFFSET(VCN, 0, mmUVD_SYS_INT_EN), + tmp, 0, indirect); +} + static int vcn_v2_5_start_dpg_mode(struct amdgpu_device *adev, int inst_idx, bool indirect) { volatile struct amdgpu_fw_shared *fw_shared = adev->vcn.inst[inst_idx].fw_shared.cpu_addr; @@ -849,6 +876,8 @@ static int vcn_v2_5_start_dpg_mode(struct amdgpu_device *adev, int inst_idx, boo WREG32_SOC15_DPG_MODE(inst_idx, SOC15_DPG_MODE_OFFSET( VCN, 0, mmUVD_LMI_CTRL2), 0, 0, indirect); + vcn_v2_6_enable_ras(adev, inst_idx, indirect); + /* unblock VCPU register access */ WREG32_SOC15_DPG_MODE(inst_idx, SOC15_DPG_MODE_OFFSET( VCN, 0, mmUVD_RB_ARB_CTRL), 0, 0, indirect); @@ -2002,16 +2031,5 @@ static void vcn_v2_5_set_ras_funcs(struct amdgpu_device *adev) break; } - if (adev->vcn.ras) { - amdgpu_ras_register_ras_block(adev, &adev->vcn.ras->ras_block); - - strcpy(adev->vcn.ras->ras_block.ras_comm.name, "vcn"); - adev->vcn.ras->ras_block.ras_comm.block = AMDGPU_RAS_BLOCK__VCN; - adev->vcn.ras->ras_block.ras_comm.type = AMDGPU_RAS_ERROR__POISON; - adev->vcn.ras_if = &adev->vcn.ras->ras_block.ras_comm; - - /* If don't define special ras_late_init function, use default ras_late_init */ - if (!adev->vcn.ras->ras_block.ras_late_init) - adev->vcn.ras->ras_block.ras_late_init = amdgpu_ras_block_late_init; - } + amdgpu_vcn_set_ras_funcs(adev); } diff --git a/drivers/gpu/drm/amd/amdgpu/vcn_v4_0.c b/drivers/gpu/drm/amd/amdgpu/vcn_v4_0.c index dcc49b01bd59..1e2b22299975 100644 --- a/drivers/gpu/drm/amd/amdgpu/vcn_v4_0.c +++ b/drivers/gpu/drm/amd/amdgpu/vcn_v4_0.c @@ -31,6 +31,7 @@ #include "soc15_hw_ip.h" #include "vcn_v2_0.h" #include "mmsch_v4_0.h" +#include "vcn_v4_0.h" #include "vcn/vcn_4_0_0_offset.h" #include "vcn/vcn_4_0_0_sh_mask.h" @@ -64,6 +65,7 @@ static int vcn_v4_0_set_powergating_state(void *handle, static int vcn_v4_0_pause_dpg_mode(struct amdgpu_device *adev, int inst_idx, struct dpg_pause_state *new_state); static void vcn_v4_0_unified_ring_set_wptr(struct amdgpu_ring *ring); +static void vcn_v4_0_set_ras_funcs(struct amdgpu_device *adev); /** * vcn_v4_0_early_init - set function pointers @@ -84,6 +86,7 @@ static int vcn_v4_0_early_init(void *handle) vcn_v4_0_set_unified_ring_funcs(adev); vcn_v4_0_set_irq_funcs(adev); + vcn_v4_0_set_ras_funcs(adev); return 0; } @@ -125,6 +128,12 @@ static int vcn_v4_0_sw_init(void *handle) if (r) return r; + /* VCN POISON TRAP */ + r = amdgpu_irq_add_id(adev, amdgpu_ih_clientid_vcns[i], + VCN_4_0__SRCID_UVD_POISON, &adev->vcn.inst[i].irq); + if (r) + return r; + ring = &adev->vcn.inst[i].ring_enc[0]; ring->use_doorbell = true; if (amdgpu_sriov_vf(adev)) @@ -289,6 +298,7 @@ static int vcn_v4_0_hw_fini(void *handle) } } + amdgpu_irq_put(adev, &adev->vcn.inst[i].irq, 0); } return 0; @@ -852,6 +862,28 @@ static void vcn_v4_0_enable_clock_gating(struct amdgpu_device *adev, int inst) return; } +static void vcn_v4_0_enable_ras(struct amdgpu_device *adev, int inst_idx, + bool indirect) +{ + uint32_t tmp; + + if (!amdgpu_ras_is_supported(adev, AMDGPU_RAS_BLOCK__VCN)) + return; + + tmp = VCN_RAS_CNTL__VCPU_VCODEC_REARM_MASK | + VCN_RAS_CNTL__VCPU_VCODEC_IH_EN_MASK | + VCN_RAS_CNTL__VCPU_VCODEC_PMI_EN_MASK | + VCN_RAS_CNTL__VCPU_VCODEC_STALL_EN_MASK; + WREG32_SOC15_DPG_MODE(inst_idx, + SOC15_DPG_MODE_OFFSET(VCN, 0, regVCN_RAS_CNTL), + tmp, 0, indirect); + + tmp = UVD_SYS_INT_EN__RASCNTL_VCPU_VCODEC_EN_MASK; + WREG32_SOC15_DPG_MODE(inst_idx, + SOC15_DPG_MODE_OFFSET(VCN, 0, regUVD_SYS_INT_EN), + tmp, 0, indirect); +} + /** * vcn_v4_0_start_dpg_mode - VCN start with dpg mode * @@ -940,6 +972,8 @@ static int vcn_v4_0_start_dpg_mode(struct amdgpu_device *adev, int inst_idx, boo WREG32_SOC15_DPG_MODE(inst_idx, SOC15_DPG_MODE_OFFSET( VCN, inst_idx, regUVD_LMI_CTRL2), tmp, 0, indirect); + vcn_v4_0_enable_ras(adev, inst_idx, indirect); + /* enable master interrupt */ WREG32_SOC15_DPG_MODE(inst_idx, SOC15_DPG_MODE_OFFSET( VCN, inst_idx, regUVD_MASTINT_EN), @@ -1932,6 +1966,9 @@ static int vcn_v4_0_process_interrupt(struct amdgpu_device *adev, struct amdgpu_ case VCN_4_0__SRCID__UVD_ENC_GENERAL_PURPOSE: amdgpu_fence_process(&adev->vcn.inst[ip_instance].ring_enc[0]); break; + case VCN_4_0__SRCID_UVD_POISON: + amdgpu_vcn_process_poison_irq(adev, source, entry); + break; default: DRM_ERROR("Unhandled interrupt: %d %d\n", entry->src_id, entry->src_data[0]); @@ -1994,3 +2031,60 @@ const struct amdgpu_ip_block_version vcn_v4_0_ip_block = .rev = 0, .funcs = &vcn_v4_0_ip_funcs, }; + +static uint32_t vcn_v4_0_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_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_query_ras_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_MAX_SUB_BLOCK; sub++) + poison_stat += + vcn_v4_0_query_poison_by_instance(adev, inst, sub); + + return !!poison_stat; +} + +const struct amdgpu_ras_block_hw_ops vcn_v4_0_ras_hw_ops = { + .query_poison_status = vcn_v4_0_query_ras_poison_status, +}; + +static struct amdgpu_vcn_ras vcn_v4_0_ras = { + .ras_block = { + .hw_ops = &vcn_v4_0_ras_hw_ops, + }, +}; + +static void vcn_v4_0_set_ras_funcs(struct amdgpu_device *adev) +{ + switch (adev->ip_versions[VCN_HWIP][0]) { + case IP_VERSION(4, 0, 0): + adev->vcn.ras = &vcn_v4_0_ras; + break; + default: + break; + } + + amdgpu_vcn_set_ras_funcs(adev); +} diff --git a/drivers/gpu/drm/amd/amdgpu/vcn_v4_0.h b/drivers/gpu/drm/amd/amdgpu/vcn_v4_0.h index 7c5c9d91bb52..7d3d11f40f27 100644 --- a/drivers/gpu/drm/amd/amdgpu/vcn_v4_0.h +++ b/drivers/gpu/drm/amd/amdgpu/vcn_v4_0.h @@ -24,6 +24,12 @@ #ifndef __VCN_V4_0_H__ #define __VCN_V4_0_H__ +enum amdgpu_vcn_v4_0_sub_block { + AMDGPU_VCN_V4_0_VCPU_VCODEC = 0, + + AMDGPU_VCN_V4_0_MAX_SUB_BLOCK, +}; + extern const struct amdgpu_ip_block_version vcn_v4_0_ip_block; #endif /* __VCN_V4_0_H__ */ diff --git a/drivers/gpu/drm/amd/amdgpu/vega20_ih.c b/drivers/gpu/drm/amd/amdgpu/vega20_ih.c index 59dfca093155..1706081d054d 100644 --- a/drivers/gpu/drm/amd/amdgpu/vega20_ih.c +++ b/drivers/gpu/drm/amd/amdgpu/vega20_ih.c @@ -267,7 +267,7 @@ static void vega20_ih_reroute_ih(struct amdgpu_device *adev) /* vega20 ih reroute will go through psp this * function is used for newer asics starting arcturus */ - if (adev->asic_type >= CHIP_ARCTURUS) { + if (adev->ip_versions[OSSSYS_HWIP][0] >= IP_VERSION(4, 2, 1)) { /* Reroute to IH ring 1 for VMC */ WREG32_SOC15(OSSSYS, 0, mmIH_CLIENT_CFG_INDEX, 0x12); tmp = RREG32_SOC15(OSSSYS, 0, mmIH_CLIENT_CFG_DATA); @@ -308,7 +308,7 @@ static int vega20_ih_irq_init(struct amdgpu_device *adev) adev->nbio.funcs->ih_control(adev); - if (adev->asic_type == CHIP_ARCTURUS && + if ((adev->ip_versions[OSSSYS_HWIP][0] == IP_VERSION(4, 2, 1)) && adev->firmware.load_type == AMDGPU_FW_LOAD_DIRECT) { ih_chicken = RREG32_SOC15(OSSSYS, 0, mmIH_CHICKEN); if (adev->irq.ih.use_bus_addr) { @@ -321,7 +321,7 @@ static int vega20_ih_irq_init(struct amdgpu_device *adev) /* psp firmware won't program IH_CHICKEN for aldebaran * driver needs to program it properly according to * MC_SPACE type in IH_RB_CNTL */ - if (adev->asic_type == CHIP_ALDEBARAN) { + if (adev->ip_versions[OSSSYS_HWIP][0] == IP_VERSION(4, 4, 0)) { ih_chicken = RREG32_SOC15(OSSSYS, 0, mmIH_CHICKEN_ALDEBARAN); if (adev->irq.ih.use_bus_addr) { ih_chicken = REG_SET_FIELD(ih_chicken, IH_CHICKEN, diff --git a/drivers/gpu/drm/amd/amdgpu/vi.c b/drivers/gpu/drm/amd/amdgpu/vi.c index f6ffd7c96ff9..12ef782eb478 100644 --- a/drivers/gpu/drm/amd/amdgpu/vi.c +++ b/drivers/gpu/drm/amd/amdgpu/vi.c @@ -2111,6 +2111,8 @@ void vi_set_virt_ops(struct amdgpu_device *adev) int vi_set_ip_blocks(struct amdgpu_device *adev) { + amdgpu_device_set_sriov_virtual_display(adev); + switch (adev->asic_type) { case CHIP_TOPAZ: /* topaz has no DCE, UVD, VCE */ @@ -2130,7 +2132,7 @@ int vi_set_ip_blocks(struct amdgpu_device *adev) amdgpu_device_ip_block_add(adev, &gfx_v8_0_ip_block); amdgpu_device_ip_block_add(adev, &sdma_v3_0_ip_block); amdgpu_device_ip_block_add(adev, &pp_smu_ip_block); - if (adev->enable_virtual_display || amdgpu_sriov_vf(adev)) + if (adev->enable_virtual_display) amdgpu_device_ip_block_add(adev, &amdgpu_vkms_ip_block); #if defined(CONFIG_DRM_AMD_DC) else if (amdgpu_device_has_dc_support(adev)) @@ -2150,7 +2152,7 @@ int vi_set_ip_blocks(struct amdgpu_device *adev) amdgpu_device_ip_block_add(adev, &gfx_v8_0_ip_block); amdgpu_device_ip_block_add(adev, &sdma_v3_0_ip_block); amdgpu_device_ip_block_add(adev, &pp_smu_ip_block); - if (adev->enable_virtual_display || amdgpu_sriov_vf(adev)) + if (adev->enable_virtual_display) amdgpu_device_ip_block_add(adev, &amdgpu_vkms_ip_block); #if defined(CONFIG_DRM_AMD_DC) else if (amdgpu_device_has_dc_support(adev)) |