diff options
Diffstat (limited to 'drivers/gpu/drm/amd/amdgpu')
42 files changed, 997 insertions, 368 deletions
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu.h b/drivers/gpu/drm/amd/amdgpu/amdgpu.h index ebfc267467ee..a23b8af95319 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu.h +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu.h @@ -121,6 +121,7 @@ extern int amdgpu_cntl_sb_buf_per_se; extern int amdgpu_param_buf_per_se; extern int amdgpu_job_hang_limit; extern int amdgpu_lbpw; +extern int amdgpu_compute_multipipe; #ifdef CONFIG_DRM_AMDGPU_SI extern int amdgpu_si_support; @@ -1310,6 +1311,8 @@ int amdgpu_gem_va_ioctl(struct drm_device *dev, void *data, int amdgpu_gem_op_ioctl(struct drm_device *dev, void *data, struct drm_file *filp); int amdgpu_cs_ioctl(struct drm_device *dev, void *data, struct drm_file *filp); +int amdgpu_cs_fence_to_handle_ioctl(struct drm_device *dev, void *data, + struct drm_file *filp); int amdgpu_cs_wait_ioctl(struct drm_device *dev, void *data, struct drm_file *filp); int amdgpu_cs_wait_fences_ioctl(struct drm_device *dev, void *data, struct drm_file *filp); @@ -1524,7 +1527,6 @@ struct amdgpu_device { /* powerplay */ struct amd_powerplay powerplay; - bool pp_enabled; bool pp_force_state_enabled; /* dpm */ diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gfx_v7.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gfx_v7.c index dc7e25cce741..47d1c132ac40 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gfx_v7.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gfx_v7.c @@ -338,6 +338,7 @@ static int kgd_hqd_load(struct kgd_dev *kgd, void *mqd, uint32_t pipe_id, struct cik_mqd *m; uint32_t *mqd_hqd; uint32_t reg, wptr_val, data; + bool valid_wptr = false; m = get_mqd(mqd); @@ -356,7 +357,14 @@ static int kgd_hqd_load(struct kgd_dev *kgd, void *mqd, uint32_t pipe_id, CP_HQD_PQ_DOORBELL_CONTROL, DOORBELL_EN, 1); WREG32(mmCP_HQD_PQ_DOORBELL_CONTROL, data); - if (read_user_wptr(mm, wptr, wptr_val)) + /* read_user_ptr may take the mm->mmap_sem. + * release srbm_mutex to avoid circular dependency between + * srbm_mutex->mm_sem->reservation_ww_class_mutex->srbm_mutex. + */ + release_queue(kgd); + valid_wptr = read_user_wptr(mm, wptr, wptr_val); + acquire_queue(kgd, pipe_id, queue_id); + if (valid_wptr) WREG32(mmCP_HQD_PQ_WPTR, (wptr_val << wptr_shift) & wptr_mask); data = REG_SET_FIELD(m->cp_hqd_active, CP_HQD_ACTIVE, ACTIVE, 1); diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gfx_v8.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gfx_v8.c index c678c69936a0..056929b8ccd0 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gfx_v8.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gfx_v8.c @@ -292,6 +292,7 @@ static int kgd_hqd_load(struct kgd_dev *kgd, void *mqd, uint32_t pipe_id, struct vi_mqd *m; uint32_t *mqd_hqd; uint32_t reg, wptr_val, data; + bool valid_wptr = false; m = get_mqd(mqd); @@ -339,7 +340,14 @@ static int kgd_hqd_load(struct kgd_dev *kgd, void *mqd, uint32_t pipe_id, CP_HQD_PQ_DOORBELL_CONTROL, DOORBELL_EN, 1); WREG32(mmCP_HQD_PQ_DOORBELL_CONTROL, data); - if (read_user_wptr(mm, wptr, wptr_val)) + /* read_user_ptr may take the mm->mmap_sem. + * release srbm_mutex to avoid circular dependency between + * srbm_mutex->mm_sem->reservation_ww_class_mutex->srbm_mutex. + */ + release_queue(kgd); + valid_wptr = read_user_wptr(mm, wptr, wptr_val); + acquire_queue(kgd, pipe_id, queue_id); + if (valid_wptr) WREG32(mmCP_HQD_PQ_WPTR, (wptr_val << wptr_shift) & wptr_mask); data = REG_SET_FIELD(m->cp_hqd_active, CP_HQD_ACTIVE, ACTIVE, 1); diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_cgs.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_cgs.c index 383204e911a4..a7afe553e0a1 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_cgs.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_cgs.c @@ -42,6 +42,28 @@ struct amdgpu_cgs_device { struct amdgpu_device *adev = \ ((struct amdgpu_cgs_device *)cgs_device)->adev +static void *amdgpu_cgs_register_pp_handle(struct cgs_device *cgs_device, + int (*call_back_func)(struct amd_pp_init *, void **)) +{ + CGS_FUNC_ADEV; + struct amd_pp_init pp_init; + struct amd_powerplay *amd_pp; + + if (call_back_func == NULL) + return NULL; + + amd_pp = &(adev->powerplay); + pp_init.chip_family = adev->family; + pp_init.chip_id = adev->asic_type; + pp_init.pm_en = (amdgpu_dpm != 0 && !amdgpu_sriov_vf(adev)) ? true : false; + pp_init.feature_mask = amdgpu_pp_feature_mask; + pp_init.device = cgs_device; + if (call_back_func(&pp_init, &(amd_pp->pp_handle))) + return NULL; + + return adev->powerplay.pp_handle; +} + static int amdgpu_cgs_alloc_gpu_mem(struct cgs_device *cgs_device, enum cgs_gpu_mem_type type, uint64_t size, uint64_t align, @@ -1179,6 +1201,7 @@ static const struct cgs_ops amdgpu_cgs_ops = { .is_virtualization_enabled = amdgpu_cgs_is_virtualization_enabled, .enter_safe_mode = amdgpu_cgs_enter_safe_mode, .lock_grbm_idx = amdgpu_cgs_lock_grbm_idx, + .register_pp_handle = amdgpu_cgs_register_pp_handle, }; static const struct cgs_os_ops amdgpu_cgs_os_ops = { diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_cs.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_cs.c index c6a214f1e991..ab83dfcabb41 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_cs.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_cs.c @@ -25,6 +25,7 @@ * Jerome Glisse <glisse@freedesktop.org> */ #include <linux/pagemap.h> +#include <linux/sync_file.h> #include <drm/drmP.h> #include <drm/amdgpu_drm.h> #include <drm/drm_syncobj.h> @@ -1330,6 +1331,66 @@ static struct dma_fence *amdgpu_cs_get_fence(struct amdgpu_device *adev, return fence; } +int amdgpu_cs_fence_to_handle_ioctl(struct drm_device *dev, void *data, + struct drm_file *filp) +{ + struct amdgpu_device *adev = dev->dev_private; + struct amdgpu_fpriv *fpriv = filp->driver_priv; + union drm_amdgpu_fence_to_handle *info = data; + struct dma_fence *fence; + struct drm_syncobj *syncobj; + struct sync_file *sync_file; + int fd, r; + + if (amdgpu_kms_vram_lost(adev, fpriv)) + return -ENODEV; + + fence = amdgpu_cs_get_fence(adev, filp, &info->in.fence); + if (IS_ERR(fence)) + return PTR_ERR(fence); + + switch (info->in.what) { + case AMDGPU_FENCE_TO_HANDLE_GET_SYNCOBJ: + r = drm_syncobj_create(&syncobj, 0, fence); + dma_fence_put(fence); + if (r) + return r; + r = drm_syncobj_get_handle(filp, syncobj, &info->out.handle); + drm_syncobj_put(syncobj); + return r; + + case AMDGPU_FENCE_TO_HANDLE_GET_SYNCOBJ_FD: + r = drm_syncobj_create(&syncobj, 0, fence); + dma_fence_put(fence); + if (r) + return r; + r = drm_syncobj_get_fd(syncobj, (int*)&info->out.handle); + drm_syncobj_put(syncobj); + return r; + + case AMDGPU_FENCE_TO_HANDLE_GET_SYNC_FILE_FD: + fd = get_unused_fd_flags(O_CLOEXEC); + if (fd < 0) { + dma_fence_put(fence); + return fd; + } + + sync_file = sync_file_create(fence); + dma_fence_put(fence); + if (!sync_file) { + put_unused_fd(fd); + return -ENOMEM; + } + + fd_install(fd, sync_file->file); + info->out.handle = fd; + return 0; + + default: + return -EINVAL; + } +} + /** * amdgpu_cs_wait_all_fence - wait on all fences to signal * diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_device.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_device.c index 3e84ddf9e3b5..1949d8aedf49 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_device.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_device.c @@ -56,6 +56,7 @@ #include "amdgpu_vf_error.h" #include "amdgpu_amdkfd.h" +#include "amdgpu_pm.h" MODULE_FIRMWARE("amdgpu/vega10_gpu_info.bin"); MODULE_FIRMWARE("amdgpu/raven_gpu_info.bin"); @@ -1603,6 +1604,7 @@ static int amdgpu_init(struct amdgpu_device *adev) return r; } adev->ip_blocks[i].status.sw = true; + /* need to do gmc hw init early so we can allocate gpu mem */ if (adev->ip_blocks[i].version->type == AMD_IP_BLOCK_TYPE_GMC) { r = amdgpu_vram_scratch_init(adev); @@ -1633,6 +1635,11 @@ static int amdgpu_init(struct amdgpu_device *adev) } } + mutex_lock(&adev->firmware.mutex); + if (amdgpu_ucode_init_bo(adev)) + adev->firmware.load_type = AMDGPU_FW_LOAD_DIRECT; + mutex_unlock(&adev->firmware.mutex); + for (i = 0; i < adev->num_ip_blocks; i++) { if (!adev->ip_blocks[i].status.sw) continue; @@ -1768,6 +1775,8 @@ static int amdgpu_fini(struct amdgpu_device *adev) adev->ip_blocks[i].status.hw = false; } + if (adev->firmware.load_type != AMDGPU_FW_LOAD_DIRECT) + amdgpu_ucode_fini_bo(adev); for (i = adev->num_ip_blocks - 1; i >= 0; i--) { if (!adev->ip_blocks[i].status.sw) @@ -2040,6 +2049,7 @@ int amdgpu_device_init(struct amdgpu_device *adev, mutex_init(&adev->srbm_mutex); mutex_init(&adev->grbm_idx_mutex); mutex_init(&adev->mn_lock); + mutex_init(&adev->virt.vf_errors.lock); hash_init(adev->mn_hash); amdgpu_check_arguments(adev); @@ -2125,7 +2135,7 @@ int amdgpu_device_init(struct amdgpu_device *adev, r = amdgpu_atombios_init(adev); if (r) { dev_err(adev->dev, "amdgpu_atombios_init failed\n"); - amdgpu_vf_error_put(AMDGIM_ERROR_VF_ATOMBIOS_INIT_FAIL, 0, 0); + amdgpu_vf_error_put(adev, AMDGIM_ERROR_VF_ATOMBIOS_INIT_FAIL, 0, 0); goto failed; } @@ -2136,7 +2146,7 @@ int amdgpu_device_init(struct amdgpu_device *adev, if (amdgpu_vpost_needed(adev)) { if (!adev->bios) { dev_err(adev->dev, "no vBIOS found\n"); - amdgpu_vf_error_put(AMDGIM_ERROR_VF_NO_VBIOS, 0, 0); + amdgpu_vf_error_put(adev, AMDGIM_ERROR_VF_NO_VBIOS, 0, 0); r = -EINVAL; goto failed; } @@ -2144,7 +2154,7 @@ int amdgpu_device_init(struct amdgpu_device *adev, r = amdgpu_atom_asic_init(adev->mode_info.atom_context); if (r) { dev_err(adev->dev, "gpu post error!\n"); - amdgpu_vf_error_put(AMDGIM_ERROR_VF_GPU_POST_ERROR, 0, 0); + amdgpu_vf_error_put(adev, AMDGIM_ERROR_VF_GPU_POST_ERROR, 0, 0); goto failed; } } else { @@ -2156,7 +2166,7 @@ int amdgpu_device_init(struct amdgpu_device *adev, r = amdgpu_atomfirmware_get_clock_info(adev); if (r) { dev_err(adev->dev, "amdgpu_atomfirmware_get_clock_info failed\n"); - amdgpu_vf_error_put(AMDGIM_ERROR_VF_ATOMBIOS_GET_CLOCK_FAIL, 0, 0); + amdgpu_vf_error_put(adev, AMDGIM_ERROR_VF_ATOMBIOS_GET_CLOCK_FAIL, 0, 0); goto failed; } } else { @@ -2164,7 +2174,7 @@ int amdgpu_device_init(struct amdgpu_device *adev, r = amdgpu_atombios_get_clock_info(adev); if (r) { dev_err(adev->dev, "amdgpu_atombios_get_clock_info failed\n"); - amdgpu_vf_error_put(AMDGIM_ERROR_VF_ATOMBIOS_GET_CLOCK_FAIL, 0, 0); + amdgpu_vf_error_put(adev, AMDGIM_ERROR_VF_ATOMBIOS_GET_CLOCK_FAIL, 0, 0); goto failed; } /* init i2c buses */ @@ -2175,7 +2185,7 @@ int amdgpu_device_init(struct amdgpu_device *adev, r = amdgpu_fence_driver_init(adev); if (r) { dev_err(adev->dev, "amdgpu_fence_driver_init failed\n"); - amdgpu_vf_error_put(AMDGIM_ERROR_VF_FENCE_INIT_FAIL, 0, 0); + amdgpu_vf_error_put(adev, AMDGIM_ERROR_VF_FENCE_INIT_FAIL, 0, 0); goto failed; } @@ -2185,7 +2195,7 @@ int amdgpu_device_init(struct amdgpu_device *adev, r = amdgpu_init(adev); if (r) { dev_err(adev->dev, "amdgpu_init failed\n"); - amdgpu_vf_error_put(AMDGIM_ERROR_VF_AMDGPU_INIT_FAIL, 0, 0); + amdgpu_vf_error_put(adev, AMDGIM_ERROR_VF_AMDGPU_INIT_FAIL, 0, 0); amdgpu_fini(adev); goto failed; } @@ -2205,7 +2215,7 @@ int amdgpu_device_init(struct amdgpu_device *adev, r = amdgpu_ib_pool_init(adev); if (r) { dev_err(adev->dev, "IB initialization failed (%d).\n", r); - amdgpu_vf_error_put(AMDGIM_ERROR_VF_IB_INIT_FAIL, 0, r); + amdgpu_vf_error_put(adev, AMDGIM_ERROR_VF_IB_INIT_FAIL, 0, r); goto failed; } @@ -2215,6 +2225,10 @@ int amdgpu_device_init(struct amdgpu_device *adev, amdgpu_fbdev_init(adev); + r = amdgpu_pm_sysfs_init(adev); + if (r) + DRM_ERROR("registering pm debugfs failed (%d).\n", r); + r = amdgpu_gem_debugfs_init(adev); if (r) DRM_ERROR("registering gem debugfs failed (%d).\n", r); @@ -2254,7 +2268,7 @@ int amdgpu_device_init(struct amdgpu_device *adev, r = amdgpu_late_init(adev); if (r) { dev_err(adev->dev, "amdgpu_late_init failed\n"); - amdgpu_vf_error_put(AMDGIM_ERROR_VF_AMDGPU_LATE_INIT_FAIL, 0, r); + amdgpu_vf_error_put(adev, AMDGIM_ERROR_VF_AMDGPU_LATE_INIT_FAIL, 0, r); goto failed; } @@ -2311,6 +2325,7 @@ void amdgpu_device_fini(struct amdgpu_device *adev) iounmap(adev->rmmio); adev->rmmio = NULL; amdgpu_doorbell_fini(adev); + amdgpu_pm_sysfs_fini(adev); amdgpu_debugfs_regs_cleanup(adev); } @@ -2936,7 +2951,7 @@ out: } } else { dev_err(adev->dev, "asic resume failed (%d).\n", r); - amdgpu_vf_error_put(AMDGIM_ERROR_VF_ASIC_RESUME_FAIL, 0, r); + amdgpu_vf_error_put(adev, AMDGIM_ERROR_VF_ASIC_RESUME_FAIL, 0, r); for (i = 0; i < AMDGPU_MAX_RINGS; ++i) { if (adev->rings[i] && adev->rings[i]->sched.thread) { kthread_unpark(adev->rings[i]->sched.thread); @@ -2950,7 +2965,7 @@ out: if (r) { /* bad news, how to tell it to userspace ? */ dev_info(adev->dev, "GPU reset failed\n"); - amdgpu_vf_error_put(AMDGIM_ERROR_VF_GPU_RESET_FAIL, 0, r); + amdgpu_vf_error_put(adev, AMDGIM_ERROR_VF_GPU_RESET_FAIL, 0, r); } else { dev_info(adev->dev, "GPU reset successed!\n"); diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_dpm.h b/drivers/gpu/drm/amd/amdgpu/amdgpu_dpm.h index f79f9ea58b17..7279fb5c3abc 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_dpm.h +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_dpm.h @@ -356,6 +356,10 @@ enum amdgpu_pcie_gen { ((adev)->powerplay.pp_funcs->switch_power_profile(\ (adev)->powerplay.pp_handle, type)) +#define amdgpu_dpm_set_clockgating_by_smu(adev, msg_id) \ + ((adev)->powerplay.pp_funcs->set_clockgating_by_smu(\ + (adev)->powerplay.pp_handle, msg_id)) + struct amdgpu_dpm { struct amdgpu_ps *ps; /* number of valid power states */ diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_drv.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_drv.c index 4f98960e47f9..ad02d3fbb44c 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_drv.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_drv.c @@ -70,9 +70,10 @@ * - 3.18.0 - Export gpu always on cu bitmap * - 3.19.0 - Add support for UVD MJPEG decode * - 3.20.0 - Add support for local BOs + * - 3.21.0 - Add DRM_AMDGPU_FENCE_TO_HANDLE ioctl */ #define KMS_DRIVER_MAJOR 3 -#define KMS_DRIVER_MINOR 20 +#define KMS_DRIVER_MINOR 21 #define KMS_DRIVER_PATCHLEVEL 0 int amdgpu_vram_limit = 0; @@ -122,6 +123,7 @@ int amdgpu_cntl_sb_buf_per_se = 0; int amdgpu_param_buf_per_se = 0; int amdgpu_job_hang_limit = 0; int amdgpu_lbpw = -1; +int amdgpu_compute_multipipe = -1; MODULE_PARM_DESC(vramlimit, "Restrict VRAM for testing, in megabytes"); module_param_named(vramlimit, amdgpu_vram_limit, int, 0600); @@ -265,6 +267,9 @@ module_param_named(job_hang_limit, amdgpu_job_hang_limit, int ,0444); MODULE_PARM_DESC(lbpw, "Load Balancing Per Watt (LBPW) support (1 = enable, 0 = disable, -1 = auto)"); module_param_named(lbpw, amdgpu_lbpw, int, 0444); +MODULE_PARM_DESC(compute_multipipe, "Force compute queues to be spread across pipes (1 = enable, 0 = disable, -1 = auto)"); +module_param_named(compute_multipipe, amdgpu_compute_multipipe, int, 0444); + #ifdef CONFIG_DRM_AMDGPU_SI #if defined(CONFIG_DRM_RADEON) || defined(CONFIG_DRM_RADEON_MODULE) diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_gfx.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_gfx.c index 4fcd98e65998..83435ccbad44 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_gfx.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_gfx.c @@ -109,9 +109,26 @@ void amdgpu_gfx_parse_disable_cu(unsigned *mask, unsigned max_se, unsigned max_s } } +static bool amdgpu_gfx_is_multipipe_capable(struct amdgpu_device *adev) +{ + if (amdgpu_compute_multipipe != -1) { + DRM_INFO("amdgpu: forcing compute pipe policy %d\n", + amdgpu_compute_multipipe); + return amdgpu_compute_multipipe == 1; + } + + /* FIXME: spreading the queues across pipes causes perf regressions + * on POLARIS11 compute workloads */ + if (adev->asic_type == CHIP_POLARIS11) + return false; + + return adev->gfx.mec.num_mec > 1; +} + void amdgpu_gfx_compute_queue_acquire(struct amdgpu_device *adev) { int i, queue, pipe, mec; + bool multipipe_policy = amdgpu_gfx_is_multipipe_capable(adev); /* policy for amdgpu compute queue ownership */ for (i = 0; i < AMDGPU_MAX_COMPUTE_QUEUES; ++i) { @@ -125,8 +142,7 @@ void amdgpu_gfx_compute_queue_acquire(struct amdgpu_device *adev) if (mec >= adev->gfx.mec.num_mec) break; - /* FIXME: spreading the queues across pipes causes perf regressions */ - if (0) { + if (multipipe_policy) { /* policy: amdgpu owns the first two queues of the first MEC */ if (mec == 0 && queue < 2) set_bit(i, adev->gfx.mec.queue_bitmap); diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_kms.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_kms.c index 4fd06f8d9768..51841259e23f 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_kms.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_kms.c @@ -1024,6 +1024,7 @@ const struct drm_ioctl_desc amdgpu_ioctls_kms[] = { DRM_IOCTL_DEF_DRV(AMDGPU_CTX, amdgpu_ctx_ioctl, DRM_AUTH|DRM_RENDER_ALLOW), DRM_IOCTL_DEF_DRV(AMDGPU_VM, amdgpu_vm_ioctl, DRM_AUTH|DRM_RENDER_ALLOW), DRM_IOCTL_DEF_DRV(AMDGPU_BO_LIST, amdgpu_bo_list_ioctl, DRM_AUTH|DRM_RENDER_ALLOW), + DRM_IOCTL_DEF_DRV(AMDGPU_FENCE_TO_HANDLE, amdgpu_cs_fence_to_handle_ioctl, DRM_AUTH|DRM_RENDER_ALLOW), /* KMS */ DRM_IOCTL_DEF_DRV(AMDGPU_GEM_MMAP, amdgpu_gem_mmap_ioctl, DRM_AUTH|DRM_RENDER_ALLOW), DRM_IOCTL_DEF_DRV(AMDGPU_GEM_WAIT_IDLE, amdgpu_gem_wait_idle_ioctl, DRM_AUTH|DRM_RENDER_ALLOW), diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_pm.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_pm.c index f6ce52956e6d..a59e04f3eeba 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_pm.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_pm.c @@ -64,10 +64,6 @@ static const struct cg_flag_name clocks[] = { void amdgpu_pm_acpi_event_handler(struct amdgpu_device *adev) { - if (adev->pp_enabled) - /* TODO */ - return; - if (adev->pm.dpm_enabled) { mutex_lock(&adev->pm.mutex); if (power_supply_is_system_supplied() > 0) @@ -118,7 +114,7 @@ static ssize_t amdgpu_set_dpm_state(struct device *dev, goto fail; } - if (adev->pp_enabled) { + if (adev->powerplay.pp_funcs->dispatch_tasks) { amdgpu_dpm_dispatch_task(adev, AMD_PP_TASK_ENABLE_USER_STATE, &state, NULL); } else { mutex_lock(&adev->pm.mutex); @@ -303,7 +299,8 @@ static ssize_t amdgpu_set_pp_force_state(struct device *dev, if (strlen(buf) == 1) adev->pp_force_state_enabled = false; - else if (adev->pp_enabled) { + else if (adev->powerplay.pp_funcs->dispatch_tasks && + adev->powerplay.pp_funcs->get_pp_num_states) { struct pp_states_info data; ret = kstrtoul(buf, 0, &idx); @@ -531,7 +528,7 @@ static ssize_t amdgpu_set_pp_sclk_od(struct device *dev, if (adev->powerplay.pp_funcs->set_sclk_od) amdgpu_dpm_set_sclk_od(adev, (uint32_t)value); - if (adev->pp_enabled) { + if (adev->powerplay.pp_funcs->dispatch_tasks) { amdgpu_dpm_dispatch_task(adev, AMD_PP_TASK_READJUST_POWER_STATE, NULL, NULL); } else { adev->pm.dpm.current_ps = adev->pm.dpm.boot_ps; @@ -575,7 +572,7 @@ static ssize_t amdgpu_set_pp_mclk_od(struct device *dev, if (adev->powerplay.pp_funcs->set_mclk_od) amdgpu_dpm_set_mclk_od(adev, (uint32_t)value); - if (adev->pp_enabled) { + if (adev->powerplay.pp_funcs->dispatch_tasks) { amdgpu_dpm_dispatch_task(adev, AMD_PP_TASK_READJUST_POWER_STATE, NULL, NULL); } else { adev->pm.dpm.current_ps = adev->pm.dpm.boot_ps; @@ -959,9 +956,6 @@ static umode_t hwmon_attributes_visible(struct kobject *kobj, attr == &sensor_dev_attr_pwm1_min.dev_attr.attr)) return 0; - if (adev->pp_enabled) - return effective_mode; - /* Skip fan attributes if fan is not present */ if (adev->pm.no_fan && (attr == &sensor_dev_attr_pwm1.dev_attr.attr || @@ -1317,6 +1311,9 @@ int amdgpu_pm_sysfs_init(struct amdgpu_device *adev) if (adev->pm.sysfs_initialized) return 0; + if (adev->pm.dpm_enabled == 0) + return 0; + if (adev->powerplay.pp_funcs->get_temperature == NULL) return 0; @@ -1341,27 +1338,26 @@ int amdgpu_pm_sysfs_init(struct amdgpu_device *adev) return ret; } - if (adev->pp_enabled) { - ret = device_create_file(adev->dev, &dev_attr_pp_num_states); - if (ret) { - DRM_ERROR("failed to create device file pp_num_states\n"); - return ret; - } - ret = device_create_file(adev->dev, &dev_attr_pp_cur_state); - if (ret) { - DRM_ERROR("failed to create device file pp_cur_state\n"); - return ret; - } - ret = device_create_file(adev->dev, &dev_attr_pp_force_state); - if (ret) { - DRM_ERROR("failed to create device file pp_force_state\n"); - return ret; - } - ret = device_create_file(adev->dev, &dev_attr_pp_table); - if (ret) { - DRM_ERROR("failed to create device file pp_table\n"); - return ret; - } + + ret = device_create_file(adev->dev, &dev_attr_pp_num_states); + if (ret) { + DRM_ERROR("failed to create device file pp_num_states\n"); + return ret; + } + ret = device_create_file(adev->dev, &dev_attr_pp_cur_state); + if (ret) { + DRM_ERROR("failed to create device file pp_cur_state\n"); + return ret; + } + ret = device_create_file(adev->dev, &dev_attr_pp_force_state); + if (ret) { + DRM_ERROR("failed to create device file pp_force_state\n"); + return ret; + } + ret = device_create_file(adev->dev, &dev_attr_pp_table); + if (ret) { + DRM_ERROR("failed to create device file pp_table\n"); + return ret; } ret = device_create_file(adev->dev, &dev_attr_pp_dpm_sclk); @@ -1417,16 +1413,19 @@ int amdgpu_pm_sysfs_init(struct amdgpu_device *adev) void amdgpu_pm_sysfs_fini(struct amdgpu_device *adev) { + if (adev->pm.dpm_enabled == 0) + return; + if (adev->pm.int_hwmon_dev) hwmon_device_unregister(adev->pm.int_hwmon_dev); device_remove_file(adev->dev, &dev_attr_power_dpm_state); device_remove_file(adev->dev, &dev_attr_power_dpm_force_performance_level); - if (adev->pp_enabled) { - device_remove_file(adev->dev, &dev_attr_pp_num_states); - device_remove_file(adev->dev, &dev_attr_pp_cur_state); - device_remove_file(adev->dev, &dev_attr_pp_force_state); - device_remove_file(adev->dev, &dev_attr_pp_table); - } + + device_remove_file(adev->dev, &dev_attr_pp_num_states); + device_remove_file(adev->dev, &dev_attr_pp_cur_state); + device_remove_file(adev->dev, &dev_attr_pp_force_state); + device_remove_file(adev->dev, &dev_attr_pp_table); + device_remove_file(adev->dev, &dev_attr_pp_dpm_sclk); device_remove_file(adev->dev, &dev_attr_pp_dpm_mclk); device_remove_file(adev->dev, &dev_attr_pp_dpm_pcie); @@ -1457,7 +1456,7 @@ void amdgpu_pm_compute_clocks(struct amdgpu_device *adev) amdgpu_fence_wait_empty(ring); } - if (adev->pp_enabled) { + if (adev->powerplay.pp_funcs->dispatch_tasks) { amdgpu_dpm_dispatch_task(adev, AMD_PP_TASK_DISPLAY_CONFIG_CHANGE, NULL, NULL); } else { mutex_lock(&adev->pm.mutex); @@ -1592,15 +1591,15 @@ static int amdgpu_debugfs_pm_info(struct seq_file *m, void *data) if ((adev->flags & AMD_IS_PX) && (ddev->switch_power_state != DRM_SWITCH_POWER_ON)) { seq_printf(m, "PX asic powered off\n"); - } else if (adev->pp_enabled) { - return amdgpu_debugfs_pm_info_pp(m, adev); - } else { + } else if (adev->powerplay.pp_funcs->debugfs_print_current_performance_level) { mutex_lock(&adev->pm.mutex); if (adev->powerplay.pp_funcs->debugfs_print_current_performance_level) adev->powerplay.pp_funcs->debugfs_print_current_performance_level(adev, m); else seq_printf(m, "Debugfs support not implemented for this asic\n"); mutex_unlock(&adev->pm.mutex); + } else { + return amdgpu_debugfs_pm_info_pp(m, adev); } return 0; diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_powerplay.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_powerplay.c index 2d2f0960b025..3b42f407971d 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_powerplay.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_powerplay.c @@ -34,24 +34,6 @@ #include "cik_dpm.h" #include "vi_dpm.h" -static int amdgpu_create_pp_handle(struct amdgpu_device *adev) -{ - struct amd_pp_init pp_init; - struct amd_powerplay *amd_pp; - int ret; - - amd_pp = &(adev->powerplay); - pp_init.chip_family = adev->family; - pp_init.chip_id = adev->asic_type; - pp_init.pm_en = (amdgpu_dpm != 0 && !amdgpu_sriov_vf(adev)) ? true : false; - pp_init.feature_mask = amdgpu_pp_feature_mask; - pp_init.device = amdgpu_cgs_create_device(adev); - ret = amd_powerplay_create(&pp_init, &(amd_pp->pp_handle)); - if (ret) - return -EINVAL; - return 0; -} - static int amdgpu_pp_early_init(void *handle) { struct amdgpu_device *adev = (struct amdgpu_device *)handle; @@ -59,7 +41,6 @@ static int amdgpu_pp_early_init(void *handle) int ret = 0; amd_pp = &(adev->powerplay); - adev->pp_enabled = false; amd_pp->pp_handle = (void *)adev; switch (adev->asic_type) { @@ -73,9 +54,7 @@ static int amdgpu_pp_early_init(void *handle) case CHIP_STONEY: case CHIP_VEGA10: case CHIP_RAVEN: - adev->pp_enabled = true; - if (amdgpu_create_pp_handle(adev)) - return -EINVAL; + amd_pp->cgs_device = amdgpu_cgs_create_device(adev); amd_pp->ip_funcs = &pp_ip_funcs; amd_pp->pp_funcs = &pp_dpm_funcs; break; @@ -97,9 +76,7 @@ static int amdgpu_pp_early_init(void *handle) amd_pp->ip_funcs = &ci_dpm_ip_funcs; amd_pp->pp_funcs = &ci_dpm_funcs; } else { - adev->pp_enabled = true; - if (amdgpu_create_pp_handle(adev)) - return -EINVAL; + amd_pp->cgs_device = amdgpu_cgs_create_device(adev); amd_pp->ip_funcs = &pp_ip_funcs; amd_pp->pp_funcs = &pp_dpm_funcs; } @@ -118,12 +95,9 @@ static int amdgpu_pp_early_init(void *handle) if (adev->powerplay.ip_funcs->early_init) ret = adev->powerplay.ip_funcs->early_init( - adev->powerplay.pp_handle); + amd_pp->cgs_device ? amd_pp->cgs_device : + amd_pp->pp_handle); - if (ret == PP_DPM_DISABLED) { - adev->pm.dpm_enabled = false; - return 0; - } return ret; } @@ -137,11 +111,6 @@ static int amdgpu_pp_late_init(void *handle) ret = adev->powerplay.ip_funcs->late_init( adev->powerplay.pp_handle); - if (adev->pp_enabled && adev->pm.dpm_enabled) { - amdgpu_pm_sysfs_init(adev); - amdgpu_dpm_dispatch_task(adev, AMD_PP_TASK_COMPLETE_INIT, NULL, NULL); - } - return ret; } @@ -176,21 +145,11 @@ static int amdgpu_pp_hw_init(void *handle) int ret = 0; struct amdgpu_device *adev = (struct amdgpu_device *)handle; - if (adev->pp_enabled && adev->firmware.load_type == AMDGPU_FW_LOAD_SMU) - amdgpu_ucode_init_bo(adev); if (adev->powerplay.ip_funcs->hw_init) ret = adev->powerplay.ip_funcs->hw_init( adev->powerplay.pp_handle); - if (ret == PP_DPM_DISABLED) { - adev->pm.dpm_enabled = false; - return 0; - } - - if ((amdgpu_dpm != 0) && !amdgpu_sriov_vf(adev)) - adev->pm.dpm_enabled = true; - return ret; } @@ -199,16 +158,10 @@ static int amdgpu_pp_hw_fini(void *handle) int ret = 0; struct amdgpu_device *adev = (struct amdgpu_device *)handle; - if (adev->pp_enabled && adev->pm.dpm_enabled) - amdgpu_pm_sysfs_fini(adev); - if (adev->powerplay.ip_funcs->hw_fini) ret = adev->powerplay.ip_funcs->hw_fini( adev->powerplay.pp_handle); - if (adev->pp_enabled && adev->firmware.load_type == AMDGPU_FW_LOAD_SMU) - amdgpu_ucode_fini_bo(adev); - return ret; } @@ -220,9 +173,8 @@ static void amdgpu_pp_late_fini(void *handle) adev->powerplay.ip_funcs->late_fini( adev->powerplay.pp_handle); - - if (adev->pp_enabled) - amd_powerplay_destroy(adev->powerplay.pp_handle); + if (adev->powerplay.cgs_device) + amdgpu_cgs_destroy_device(adev->powerplay.cgs_device); } static int amdgpu_pp_suspend(void *handle) diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_psp.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_psp.c index 447d446b5015..f1035a689d35 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_psp.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_psp.c @@ -411,13 +411,6 @@ static int psp_hw_init(void *handle) return 0; mutex_lock(&adev->firmware.mutex); - /* - * This sequence is just used on hw_init only once, no need on - * resume. - */ - ret = amdgpu_ucode_init_bo(adev); - if (ret) - goto failed; ret = psp_load_fw(adev); if (ret) { @@ -442,8 +435,6 @@ static int psp_hw_fini(void *handle) if (adev->firmware.load_type != AMDGPU_FW_LOAD_PSP) return 0; - amdgpu_ucode_fini_bo(adev); - psp_ring_destroy(psp, PSP_RING_TYPE__KM); amdgpu_bo_free_kernel(&psp->tmr_bo, &psp->tmr_mc_addr, &psp->tmr_buf); diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_queue_mgr.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_queue_mgr.c index befc09b68543..190e28cb827e 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_queue_mgr.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_queue_mgr.c @@ -121,7 +121,7 @@ static enum amdgpu_ring_type amdgpu_hw_ip_to_ring_type(int hw_ip) static int amdgpu_lru_map(struct amdgpu_device *adev, struct amdgpu_queue_mapper *mapper, - int user_ring, + int user_ring, bool lru_pipe_order, struct amdgpu_ring **out_ring) { int r, i, j; @@ -139,7 +139,7 @@ static int amdgpu_lru_map(struct amdgpu_device *adev, } r = amdgpu_ring_lru_get(adev, ring_type, ring_blacklist, - j, out_ring); + j, lru_pipe_order, out_ring); if (r) return r; @@ -284,8 +284,10 @@ int amdgpu_queue_mgr_map(struct amdgpu_device *adev, r = amdgpu_identity_map(adev, mapper, ring, out_ring); break; case AMDGPU_HW_IP_DMA: + r = amdgpu_lru_map(adev, mapper, ring, false, out_ring); + break; case AMDGPU_HW_IP_COMPUTE: - r = amdgpu_lru_map(adev, mapper, ring, out_ring); + r = amdgpu_lru_map(adev, mapper, ring, true, out_ring); break; default: *out_ring = NULL; diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_ring.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_ring.c index 5ce65280b396..019932a7ea3a 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_ring.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_ring.c @@ -315,14 +315,16 @@ static bool amdgpu_ring_is_blacklisted(struct amdgpu_ring *ring, * @type: amdgpu_ring_type enum * @blacklist: blacklisted ring ids array * @num_blacklist: number of entries in @blacklist + * @lru_pipe_order: find a ring from the least recently used pipe * @ring: output ring * * Retrieve the amdgpu_ring structure for the least recently used ring of * a specific IP block (all asics). * Returns 0 on success, error on failure. */ -int amdgpu_ring_lru_get(struct amdgpu_device *adev, int type, int *blacklist, - int num_blacklist, struct amdgpu_ring **ring) +int amdgpu_ring_lru_get(struct amdgpu_device *adev, int type, + int *blacklist, int num_blacklist, + bool lru_pipe_order, struct amdgpu_ring **ring) { struct amdgpu_ring *entry; @@ -337,10 +339,23 @@ int amdgpu_ring_lru_get(struct amdgpu_device *adev, int type, int *blacklist, if (amdgpu_ring_is_blacklisted(entry, blacklist, num_blacklist)) continue; - *ring = entry; - amdgpu_ring_lru_touch_locked(adev, *ring); - break; + if (!*ring) { + *ring = entry; + + /* We are done for ring LRU */ + if (!lru_pipe_order) + break; + } + + /* Move all rings on the same pipe to the end of the list */ + if (entry->pipe == (*ring)->pipe) + amdgpu_ring_lru_touch_locked(adev, entry); } + + /* Move the ring we found to the end of the list */ + if (*ring) + amdgpu_ring_lru_touch_locked(adev, *ring); + spin_unlock(&adev->ring_lru_list_lock); if (!*ring) { diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_ring.h b/drivers/gpu/drm/amd/amdgpu/amdgpu_ring.h index 322d25299a00..491bd5512dcc 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_ring.h +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_ring.h @@ -201,8 +201,9 @@ int amdgpu_ring_init(struct amdgpu_device *adev, struct amdgpu_ring *ring, unsigned ring_size, struct amdgpu_irq_src *irq_src, unsigned irq_type); void amdgpu_ring_fini(struct amdgpu_ring *ring); -int amdgpu_ring_lru_get(struct amdgpu_device *adev, int type, int *blacklist, - int num_blacklist, struct amdgpu_ring **ring); +int amdgpu_ring_lru_get(struct amdgpu_device *adev, int type, + int *blacklist, int num_blacklist, + bool lru_pipe_order, struct amdgpu_ring **ring); void amdgpu_ring_lru_touch(struct amdgpu_device *adev, struct amdgpu_ring *ring); static inline void amdgpu_ring_clear_ring(struct amdgpu_ring *ring) { diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_vf_error.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_vf_error.c index 45ac91861965..746b81339835 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_vf_error.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_vf_error.c @@ -25,30 +25,21 @@ #include "amdgpu_vf_error.h" #include "mxgpu_ai.h" -#define AMDGPU_VF_ERROR_ENTRY_SIZE 16 - -/* struct error_entry - amdgpu VF error information. */ -struct amdgpu_vf_error_buffer { - int read_count; - int write_count; - uint16_t code[AMDGPU_VF_ERROR_ENTRY_SIZE]; - uint16_t flags[AMDGPU_VF_ERROR_ENTRY_SIZE]; - uint64_t data[AMDGPU_VF_ERROR_ENTRY_SIZE]; -}; - -struct amdgpu_vf_error_buffer admgpu_vf_errors; - - -void amdgpu_vf_error_put(uint16_t sub_error_code, uint16_t error_flags, uint64_t error_data) +void amdgpu_vf_error_put(struct amdgpu_device *adev, + uint16_t sub_error_code, + uint16_t error_flags, + uint64_t error_data) { int index; uint16_t error_code = AMDGIM_ERROR_CODE(AMDGIM_ERROR_CATEGORY_VF, sub_error_code); - index = admgpu_vf_errors.write_count % AMDGPU_VF_ERROR_ENTRY_SIZE; - admgpu_vf_errors.code [index] = error_code; - admgpu_vf_errors.flags [index] = error_flags; - admgpu_vf_errors.data [index] = error_data; - admgpu_vf_errors.write_count ++; + mutex_lock(&adev->virt.vf_errors.lock); + index = adev->virt.vf_errors.write_count % AMDGPU_VF_ERROR_ENTRY_SIZE; + adev->virt.vf_errors.code [index] = error_code; + adev->virt.vf_errors.flags [index] = error_flags; + adev->virt.vf_errors.data [index] = error_data; + adev->virt.vf_errors.write_count ++; + mutex_unlock(&adev->virt.vf_errors.lock); } @@ -58,7 +49,8 @@ void amdgpu_vf_error_trans_all(struct amdgpu_device *adev) u32 data1, data2, data3; int index; - if ((NULL == adev) || (!amdgpu_sriov_vf(adev)) || (!adev->virt.ops) || (!adev->virt.ops->trans_msg)) { + if ((NULL == adev) || (!amdgpu_sriov_vf(adev)) || + (!adev->virt.ops) || (!adev->virt.ops->trans_msg)) { return; } /* @@ -68,18 +60,22 @@ void amdgpu_vf_error_trans_all(struct amdgpu_device *adev) return; } */ + + mutex_lock(&adev->virt.vf_errors.lock); /* The errors are overlay of array, correct read_count as full. */ - if (admgpu_vf_errors.write_count - admgpu_vf_errors.read_count > AMDGPU_VF_ERROR_ENTRY_SIZE) { - admgpu_vf_errors.read_count = admgpu_vf_errors.write_count - AMDGPU_VF_ERROR_ENTRY_SIZE; + if (adev->virt.vf_errors.write_count - adev->virt.vf_errors.read_count > AMDGPU_VF_ERROR_ENTRY_SIZE) { + adev->virt.vf_errors.read_count = adev->virt.vf_errors.write_count - AMDGPU_VF_ERROR_ENTRY_SIZE; } - while (admgpu_vf_errors.read_count < admgpu_vf_errors.write_count) { - index =admgpu_vf_errors.read_count % AMDGPU_VF_ERROR_ENTRY_SIZE; - data1 = AMDGIM_ERROR_CODE_FLAGS_TO_MAILBOX (admgpu_vf_errors.code[index], admgpu_vf_errors.flags[index]); - data2 = admgpu_vf_errors.data[index] & 0xFFFFFFFF; - data3 = (admgpu_vf_errors.data[index] >> 32) & 0xFFFFFFFF; + while (adev->virt.vf_errors.read_count < adev->virt.vf_errors.write_count) { + index =adev->virt.vf_errors.read_count % AMDGPU_VF_ERROR_ENTRY_SIZE; + data1 = AMDGIM_ERROR_CODE_FLAGS_TO_MAILBOX(adev->virt.vf_errors.code[index], + adev->virt.vf_errors.flags[index]); + data2 = adev->virt.vf_errors.data[index] & 0xFFFFFFFF; + data3 = (adev->virt.vf_errors.data[index] >> 32) & 0xFFFFFFFF; adev->virt.ops->trans_msg(adev, IDH_LOG_VF_ERROR, data1, data2, data3); - admgpu_vf_errors.read_count ++; + adev->virt.vf_errors.read_count ++; } + mutex_unlock(&adev->virt.vf_errors.lock); } diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_vf_error.h b/drivers/gpu/drm/amd/amdgpu/amdgpu_vf_error.h index 2a3278ec76ba..6436bd053325 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_vf_error.h +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_vf_error.h @@ -56,7 +56,10 @@ enum AMDGIM_ERROR_CATEGORY { AMDGIM_ERROR_CATEGORY_MAX }; -void amdgpu_vf_error_put(uint16_t sub_error_code, uint16_t error_flags, uint64_t error_data); +void amdgpu_vf_error_put(struct amdgpu_device *adev, + uint16_t sub_error_code, + uint16_t error_flags, + uint64_t error_data); void amdgpu_vf_error_trans_all (struct amdgpu_device *adev); #endif /* __VF_ERROR_H__ */ diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_virt.h b/drivers/gpu/drm/amd/amdgpu/amdgpu_virt.h index afcfb8bcfb65..e5fd0ff6b29d 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_virt.h +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_virt.h @@ -36,6 +36,18 @@ struct amdgpu_mm_table { uint64_t gpu_addr; }; +#define AMDGPU_VF_ERROR_ENTRY_SIZE 16 + +/* struct error_entry - amdgpu VF error information. */ +struct amdgpu_vf_error_buffer { + struct mutex lock; + int read_count; + int write_count; + uint16_t code[AMDGPU_VF_ERROR_ENTRY_SIZE]; + uint16_t flags[AMDGPU_VF_ERROR_ENTRY_SIZE]; + uint64_t data[AMDGPU_VF_ERROR_ENTRY_SIZE]; +}; + /** * struct amdgpu_virt_ops - amdgpu device virt operations */ @@ -59,6 +71,7 @@ struct amdgpu_virt { struct work_struct flr_work; struct amdgpu_mm_table mm_table; const struct amdgpu_virt_ops *ops; + struct amdgpu_vf_error_buffer vf_errors; }; #define AMDGPU_CSA_SIZE (8 * 1024) diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_vm.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_vm.c index bbcc67038203..fee0a32ac56f 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_vm.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_vm.c @@ -2541,7 +2541,8 @@ static uint32_t amdgpu_vm_get_block_size(uint64_t vm_size) * @adev: amdgpu_device pointer * @fragment_size_default: the default fragment size if it's set auto */ -void amdgpu_vm_set_fragment_size(struct amdgpu_device *adev, uint32_t fragment_size_default) +void amdgpu_vm_set_fragment_size(struct amdgpu_device *adev, + uint32_t fragment_size_default) { if (amdgpu_vm_fragment_size == -1) adev->vm_manager.fragment_size = fragment_size_default; @@ -2555,7 +2556,8 @@ void amdgpu_vm_set_fragment_size(struct amdgpu_device *adev, uint32_t fragment_s * @adev: amdgpu_device pointer * @vm_size: the default vm size if it's set auto */ -void amdgpu_vm_adjust_size(struct amdgpu_device *adev, uint64_t vm_size, uint32_t fragment_size_default) +void amdgpu_vm_adjust_size(struct amdgpu_device *adev, uint64_t vm_size, + uint32_t fragment_size_default) { /* adjust vm size firstly */ if (amdgpu_vm_size == -1) @@ -2682,6 +2684,7 @@ int amdgpu_vm_init(struct amdgpu_device *adev, struct amdgpu_vm *vm, } INIT_KFIFO(vm->faults); + vm->fault_credit = 16; return 0; @@ -2777,6 +2780,36 @@ void amdgpu_vm_fini(struct amdgpu_device *adev, struct amdgpu_vm *vm) } /** + * amdgpu_vm_pasid_fault_credit - Check fault credit for given PASID + * + * @adev: amdgpu_device pointer + * @pasid: PASID do identify the VM + * + * This function is expected to be called in interrupt context. Returns + * true if there was fault credit, false otherwise + */ +bool amdgpu_vm_pasid_fault_credit(struct amdgpu_device *adev, + unsigned int pasid) +{ + struct amdgpu_vm *vm; + + spin_lock(&adev->vm_manager.pasid_lock); + vm = idr_find(&adev->vm_manager.pasid_idr, pasid); + spin_unlock(&adev->vm_manager.pasid_lock); + if (!vm) + /* VM not found, can't track fault credit */ + return true; + + /* No lock needed. only accessed by IRQ handler */ + if (!vm->fault_credit) + /* Too many faults in this VM */ + return false; + + vm->fault_credit--; + return true; +} + +/** * amdgpu_vm_manager_init - init the VM manager * * @adev: amdgpu_device pointer diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_vm.h b/drivers/gpu/drm/amd/amdgpu/amdgpu_vm.h index 0af090667dfc..d68f39b4e5e7 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_vm.h +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_vm.h @@ -165,8 +165,11 @@ struct amdgpu_vm { /* Flag to indicate ATS support from PTE for GFX9 */ bool pte_support_ats; - /* Up to 128 pending page faults */ + /* Up to 128 pending retry page faults */ DECLARE_KFIFO(faults, u64, 128); + + /* Limit non-retry fault storms */ + unsigned int fault_credit; }; struct amdgpu_vm_id { @@ -244,6 +247,8 @@ void amdgpu_vm_manager_fini(struct amdgpu_device *adev); int amdgpu_vm_init(struct amdgpu_device *adev, struct amdgpu_vm *vm, int vm_context, unsigned int pasid); void amdgpu_vm_fini(struct amdgpu_device *adev, struct amdgpu_vm *vm); +bool amdgpu_vm_pasid_fault_credit(struct amdgpu_device *adev, + unsigned int pasid); void amdgpu_vm_get_pd_bo(struct amdgpu_vm *vm, struct list_head *validated, struct amdgpu_bo_list_entry *entry); diff --git a/drivers/gpu/drm/amd/amdgpu/ci_dpm.c b/drivers/gpu/drm/amd/amdgpu/ci_dpm.c index 68ce1bdaf2fc..68b505c768ad 100644 --- a/drivers/gpu/drm/amd/amdgpu/ci_dpm.c +++ b/drivers/gpu/drm/amd/amdgpu/ci_dpm.c @@ -6365,7 +6365,6 @@ static int ci_dpm_sw_fini(void *handle) flush_work(&adev->pm.dpm.thermal.work); mutex_lock(&adev->pm.mutex); - amdgpu_pm_sysfs_fini(adev); ci_dpm_fini(adev); mutex_unlock(&adev->pm.mutex); diff --git a/drivers/gpu/drm/amd/amdgpu/cik_ih.c b/drivers/gpu/drm/amd/amdgpu/cik_ih.c index 07d3d895da10..a870b354e3f7 100644 --- a/drivers/gpu/drm/amd/amdgpu/cik_ih.c +++ b/drivers/gpu/drm/amd/amdgpu/cik_ih.c @@ -237,8 +237,23 @@ static u32 cik_ih_get_wptr(struct amdgpu_device *adev) */ static bool cik_ih_prescreen_iv(struct amdgpu_device *adev) { - /* Process all interrupts */ - return true; + u32 ring_index = adev->irq.ih.rptr >> 2; + u16 pasid; + + switch (le32_to_cpu(adev->irq.ih.ring[ring_index]) & 0xff) { + case 146: + case 147: + pasid = le32_to_cpu(adev->irq.ih.ring[ring_index + 2]) >> 16; + if (!pasid || amdgpu_vm_pasid_fault_credit(adev, pasid)) + return true; + break; + default: + /* Not a VM fault */ + return true; + } + + adev->irq.ih.rptr += 16; + return false; } /** diff --git a/drivers/gpu/drm/amd/amdgpu/cz_ih.c b/drivers/gpu/drm/amd/amdgpu/cz_ih.c index b6cdf4afaf46..fa61d649bb44 100644 --- a/drivers/gpu/drm/amd/amdgpu/cz_ih.c +++ b/drivers/gpu/drm/amd/amdgpu/cz_ih.c @@ -216,8 +216,23 @@ static u32 cz_ih_get_wptr(struct amdgpu_device *adev) */ static bool cz_ih_prescreen_iv(struct amdgpu_device *adev) { - /* Process all interrupts */ - return true; + u32 ring_index = adev->irq.ih.rptr >> 2; + u16 pasid; + + switch (le32_to_cpu(adev->irq.ih.ring[ring_index]) & 0xff) { + case 146: + case 147: + pasid = le32_to_cpu(adev->irq.ih.ring[ring_index + 2]) >> 16; + if (!pasid || amdgpu_vm_pasid_fault_credit(adev, pasid)) + return true; + break; + default: + /* Not a VM fault */ + return true; + } + + adev->irq.ih.rptr += 16; + return false; } /** diff --git a/drivers/gpu/drm/amd/amdgpu/gfx_v8_0.c b/drivers/gpu/drm/amd/amdgpu/gfx_v8_0.c index dfc10b1baea0..147e92b3a959 100644 --- a/drivers/gpu/drm/amd/amdgpu/gfx_v8_0.c +++ b/drivers/gpu/drm/amd/amdgpu/gfx_v8_0.c @@ -4132,18 +4132,12 @@ static int gfx_v8_0_rlc_resume(struct amdgpu_device *adev) gfx_v8_0_rlc_reset(adev); gfx_v8_0_init_pg(adev); - if (!adev->pp_enabled) { - if (adev->firmware.load_type != AMDGPU_FW_LOAD_SMU) { - /* legacy rlc firmware loading */ - r = gfx_v8_0_rlc_load_microcode(adev); - if (r) - return r; - } else { - r = adev->smu.smumgr_funcs->check_fw_load_finish(adev, - AMDGPU_UCODE_ID_RLC_G); - if (r) - return -EINVAL; - } + + if (adev->firmware.load_type == AMDGPU_FW_LOAD_DIRECT) { + /* legacy rlc firmware loading */ + r = gfx_v8_0_rlc_load_microcode(adev); + if (r) + return r; } gfx_v8_0_rlc_start(adev); @@ -4959,43 +4953,15 @@ static int gfx_v8_0_cp_resume(struct amdgpu_device *adev) if (!(adev->flags & AMD_IS_APU)) gfx_v8_0_enable_gui_idle_interrupt(adev, false); - if (!adev->pp_enabled) { - if (adev->firmware.load_type != AMDGPU_FW_LOAD_SMU) { + if (adev->firmware.load_type == AMDGPU_FW_LOAD_DIRECT) { /* legacy firmware loading */ - r = gfx_v8_0_cp_gfx_load_microcode(adev); - if (r) - return r; + r = gfx_v8_0_cp_gfx_load_microcode(adev); + if (r) + return r; - r = gfx_v8_0_cp_compute_load_microcode(adev); - if (r) - return r; - } else { - r = adev->smu.smumgr_funcs->check_fw_load_finish(adev, - AMDGPU_UCODE_ID_CP_CE); - if (r) - return -EINVAL; - - r = adev->smu.smumgr_funcs->check_fw_load_finish(adev, - AMDGPU_UCODE_ID_CP_PFP); - if (r) - return -EINVAL; - - r = adev->smu.smumgr_funcs->check_fw_load_finish(adev, - AMDGPU_UCODE_ID_CP_ME); - if (r) - return -EINVAL; - - if (adev->asic_type == CHIP_TOPAZ) { - r = gfx_v8_0_cp_compute_load_microcode(adev); - if (r) - return r; - } else { - r = adev->smu.smumgr_funcs->check_fw_load_finish(adev, - AMDGPU_UCODE_ID_CP_MEC1); - if (r) - return -EINVAL; - } - } + r = gfx_v8_0_cp_compute_load_microcode(adev); + if (r) + return r; } r = gfx_v8_0_cp_gfx_resume(adev); @@ -6018,7 +5984,6 @@ static int gfx_v8_0_tonga_update_gfx_clock_gating(struct amdgpu_device *adev, { uint32_t msg_id, pp_state = 0; uint32_t pp_support_state = 0; - void *pp_handle = adev->powerplay.pp_handle; if (adev->cg_flags & (AMD_CG_SUPPORT_GFX_CGCG | AMD_CG_SUPPORT_GFX_CGLS)) { if (adev->cg_flags & AMD_CG_SUPPORT_GFX_CGLS) { @@ -6036,7 +6001,8 @@ static int gfx_v8_0_tonga_update_gfx_clock_gating(struct amdgpu_device *adev, PP_BLOCK_GFX_CG, pp_support_state, pp_state); - amd_set_clockgating_by_smu(pp_handle, msg_id); + if (adev->powerplay.pp_funcs->set_clockgating_by_smu) + amdgpu_dpm_set_clockgating_by_smu(adev, msg_id); } if (adev->cg_flags & (AMD_CG_SUPPORT_GFX_MGCG | AMD_CG_SUPPORT_GFX_MGLS)) { @@ -6057,7 +6023,8 @@ static int gfx_v8_0_tonga_update_gfx_clock_gating(struct amdgpu_device *adev, PP_BLOCK_GFX_MG, pp_support_state, pp_state); - amd_set_clockgating_by_smu(pp_handle, msg_id); + if (adev->powerplay.pp_funcs->set_clockgating_by_smu) + amdgpu_dpm_set_clockgating_by_smu(adev, msg_id); } return 0; @@ -6069,7 +6036,6 @@ static int gfx_v8_0_polaris_update_gfx_clock_gating(struct amdgpu_device *adev, uint32_t msg_id, pp_state = 0; uint32_t pp_support_state = 0; - void *pp_handle = adev->powerplay.pp_handle; if (adev->cg_flags & (AMD_CG_SUPPORT_GFX_CGCG | AMD_CG_SUPPORT_GFX_CGLS)) { if (adev->cg_flags & AMD_CG_SUPPORT_GFX_CGLS) { @@ -6087,7 +6053,8 @@ static int gfx_v8_0_polaris_update_gfx_clock_gating(struct amdgpu_device *adev, PP_BLOCK_GFX_CG, pp_support_state, pp_state); - amd_set_clockgating_by_smu(pp_handle, msg_id); + if (adev->powerplay.pp_funcs->set_clockgating_by_smu) + amdgpu_dpm_set_clockgating_by_smu(adev, msg_id); } if (adev->cg_flags & (AMD_CG_SUPPORT_GFX_3D_CGCG | AMD_CG_SUPPORT_GFX_3D_CGLS)) { @@ -6106,7 +6073,8 @@ static int gfx_v8_0_polaris_update_gfx_clock_gating(struct amdgpu_device *adev, PP_BLOCK_GFX_3D, pp_support_state, pp_state); - amd_set_clockgating_by_smu(pp_handle, msg_id); + if (adev->powerplay.pp_funcs->set_clockgating_by_smu) + amdgpu_dpm_set_clockgating_by_smu(adev, msg_id); } if (adev->cg_flags & (AMD_CG_SUPPORT_GFX_MGCG | AMD_CG_SUPPORT_GFX_MGLS)) { @@ -6127,7 +6095,8 @@ static int gfx_v8_0_polaris_update_gfx_clock_gating(struct amdgpu_device *adev, PP_BLOCK_GFX_MG, pp_support_state, pp_state); - amd_set_clockgating_by_smu(pp_handle, msg_id); + if (adev->powerplay.pp_funcs->set_clockgating_by_smu) + amdgpu_dpm_set_clockgating_by_smu(adev, msg_id); } if (adev->cg_flags & AMD_CG_SUPPORT_GFX_RLC_LS) { @@ -6142,7 +6111,8 @@ static int gfx_v8_0_polaris_update_gfx_clock_gating(struct amdgpu_device *adev, PP_BLOCK_GFX_RLC, pp_support_state, pp_state); - amd_set_clockgating_by_smu(pp_handle, msg_id); + if (adev->powerplay.pp_funcs->set_clockgating_by_smu) + amdgpu_dpm_set_clockgating_by_smu(adev, msg_id); } if (adev->cg_flags & AMD_CG_SUPPORT_GFX_CP_LS) { @@ -6156,7 +6126,8 @@ static int gfx_v8_0_polaris_update_gfx_clock_gating(struct amdgpu_device *adev, PP_BLOCK_GFX_CP, pp_support_state, pp_state); - amd_set_clockgating_by_smu(pp_handle, msg_id); + if (adev->powerplay.pp_funcs->set_clockgating_by_smu) + amdgpu_dpm_set_clockgating_by_smu(adev, msg_id); } return 0; @@ -7076,7 +7047,7 @@ static void gfx_v8_0_ring_emit_ce_meta(struct amdgpu_ring *ring) { uint64_t ce_payload_addr; int cnt_ce; - static union { + union { struct vi_ce_ib_state regular; struct vi_ce_ib_state_chained_ib chained; } ce_payload = {}; @@ -7105,7 +7076,7 @@ static void gfx_v8_0_ring_emit_de_meta(struct amdgpu_ring *ring) { uint64_t de_payload_addr, gds_addr, csa_addr; int cnt_de; - static union { + union { struct vi_de_ib_state regular; struct vi_de_ib_state_chained_ib chained; } de_payload = {}; diff --git a/drivers/gpu/drm/amd/amdgpu/gfx_v9_0.c b/drivers/gpu/drm/amd/amdgpu/gfx_v9_0.c index deeaee1457ef..99a5b3b92e8e 100644 --- a/drivers/gpu/drm/amd/amdgpu/gfx_v9_0.c +++ b/drivers/gpu/drm/amd/amdgpu/gfx_v9_0.c @@ -3583,7 +3583,7 @@ static void gfx_v9_0_ring_set_wptr_gfx(struct amdgpu_ring *ring) static void gfx_v9_0_ring_emit_hdp_flush(struct amdgpu_ring *ring) { u32 ref_and_mask, reg_mem_engine; - struct nbio_hdp_flush_reg *nbio_hf_reg; + const struct nbio_hdp_flush_reg *nbio_hf_reg; if (ring->adev->flags & AMD_IS_APU) nbio_hf_reg = &nbio_v7_0_hdp_flush_reg; @@ -3806,7 +3806,7 @@ static void gfx_v9_ring_emit_sb(struct amdgpu_ring *ring) static void gfx_v9_0_ring_emit_ce_meta(struct amdgpu_ring *ring) { - static struct v9_ce_ib_state ce_payload = {0}; + struct v9_ce_ib_state ce_payload = {0}; uint64_t csa_addr; int cnt; @@ -3825,7 +3825,7 @@ static void gfx_v9_0_ring_emit_ce_meta(struct amdgpu_ring *ring) static void gfx_v9_0_ring_emit_de_meta(struct amdgpu_ring *ring) { - static struct v9_de_ib_state de_payload = {0}; + struct v9_de_ib_state de_payload = {0}; uint64_t csa_addr, gds_addr; int cnt; diff --git a/drivers/gpu/drm/amd/amdgpu/iceland_ih.c b/drivers/gpu/drm/amd/amdgpu/iceland_ih.c index 65ed6d3a8f05..bd592cb39f37 100644 --- a/drivers/gpu/drm/amd/amdgpu/iceland_ih.c +++ b/drivers/gpu/drm/amd/amdgpu/iceland_ih.c @@ -216,8 +216,23 @@ static u32 iceland_ih_get_wptr(struct amdgpu_device *adev) */ static bool iceland_ih_prescreen_iv(struct amdgpu_device *adev) { - /* Process all interrupts */ - return true; + u32 ring_index = adev->irq.ih.rptr >> 2; + u16 pasid; + + switch (le32_to_cpu(adev->irq.ih.ring[ring_index]) & 0xff) { + case 146: + case 147: + pasid = le32_to_cpu(adev->irq.ih.ring[ring_index + 2]) >> 16; + if (!pasid || amdgpu_vm_pasid_fault_credit(adev, pasid)) + return true; + break; + default: + /* Not a VM fault */ + return true; + } + + adev->irq.ih.rptr += 16; + return false; } /** diff --git a/drivers/gpu/drm/amd/amdgpu/kv_dpm.c b/drivers/gpu/drm/amd/amdgpu/kv_dpm.c index b57399a462c2..f33d1ffdb20b 100644 --- a/drivers/gpu/drm/amd/amdgpu/kv_dpm.c +++ b/drivers/gpu/drm/amd/amdgpu/kv_dpm.c @@ -2969,16 +2969,10 @@ static int kv_dpm_late_init(void *handle) { /* powerdown unused blocks for now */ struct amdgpu_device *adev = (struct amdgpu_device *)handle; - int ret; if (!amdgpu_dpm) return 0; - /* init the sysfs and debugfs files late */ - ret = amdgpu_pm_sysfs_init(adev); - if (ret) - return ret; - kv_dpm_powergate_acp(adev, true); kv_dpm_powergate_samu(adev, true); @@ -3040,7 +3034,6 @@ static int kv_dpm_sw_fini(void *handle) flush_work(&adev->pm.dpm.thermal.work); mutex_lock(&adev->pm.mutex); - amdgpu_pm_sysfs_fini(adev); kv_dpm_fini(adev); mutex_unlock(&adev->pm.mutex); diff --git a/drivers/gpu/drm/amd/amdgpu/nbio_v6_1.c b/drivers/gpu/drm/amd/amdgpu/nbio_v6_1.c index 045988b18bc3..904a1bab9b9f 100644 --- a/drivers/gpu/drm/amd/amdgpu/nbio_v6_1.c +++ b/drivers/gpu/drm/amd/amdgpu/nbio_v6_1.c @@ -215,31 +215,27 @@ void nbio_v6_1_get_clockgating_state(struct amdgpu_device *adev, u32 *flags) *flags |= AMD_CG_SUPPORT_BIF_LS; } -struct nbio_hdp_flush_reg nbio_v6_1_hdp_flush_reg; -struct nbio_pcie_index_data nbio_v6_1_pcie_index_data; +const struct nbio_hdp_flush_reg nbio_v6_1_hdp_flush_reg = { + .hdp_flush_req_offset = SOC15_REG_OFFSET(NBIO, 0, mmBIF_BX_PF0_GPU_HDP_FLUSH_REQ), + .hdp_flush_done_offset = SOC15_REG_OFFSET(NBIO, 0, mmBIF_BX_PF0_GPU_HDP_FLUSH_DONE), + .ref_and_mask_cp0 = BIF_BX_PF0_GPU_HDP_FLUSH_DONE__CP0_MASK, + .ref_and_mask_cp1 = BIF_BX_PF0_GPU_HDP_FLUSH_DONE__CP1_MASK, + .ref_and_mask_cp2 = BIF_BX_PF0_GPU_HDP_FLUSH_DONE__CP2_MASK, + .ref_and_mask_cp3 = BIF_BX_PF0_GPU_HDP_FLUSH_DONE__CP3_MASK, + .ref_and_mask_cp4 = BIF_BX_PF0_GPU_HDP_FLUSH_DONE__CP4_MASK, + .ref_and_mask_cp5 = BIF_BX_PF0_GPU_HDP_FLUSH_DONE__CP5_MASK, + .ref_and_mask_cp6 = BIF_BX_PF0_GPU_HDP_FLUSH_DONE__CP6_MASK, + .ref_and_mask_cp7 = BIF_BX_PF0_GPU_HDP_FLUSH_DONE__CP7_MASK, + .ref_and_mask_cp8 = BIF_BX_PF0_GPU_HDP_FLUSH_DONE__CP8_MASK, + .ref_and_mask_cp9 = BIF_BX_PF0_GPU_HDP_FLUSH_DONE__CP9_MASK, + .ref_and_mask_sdma0 = BIF_BX_PF0_GPU_HDP_FLUSH_DONE__SDMA0_MASK, + .ref_and_mask_sdma1 = BIF_BX_PF0_GPU_HDP_FLUSH_DONE__SDMA1_MASK +}; -int nbio_v6_1_init(struct amdgpu_device *adev) -{ - nbio_v6_1_hdp_flush_reg.hdp_flush_req_offset = SOC15_REG_OFFSET(NBIO, 0, mmBIF_BX_PF0_GPU_HDP_FLUSH_REQ); - nbio_v6_1_hdp_flush_reg.hdp_flush_done_offset = SOC15_REG_OFFSET(NBIO, 0, mmBIF_BX_PF0_GPU_HDP_FLUSH_DONE); - nbio_v6_1_hdp_flush_reg.ref_and_mask_cp0 = BIF_BX_PF0_GPU_HDP_FLUSH_DONE__CP0_MASK; - nbio_v6_1_hdp_flush_reg.ref_and_mask_cp1 = BIF_BX_PF0_GPU_HDP_FLUSH_DONE__CP1_MASK; - nbio_v6_1_hdp_flush_reg.ref_and_mask_cp2 = BIF_BX_PF0_GPU_HDP_FLUSH_DONE__CP2_MASK; - nbio_v6_1_hdp_flush_reg.ref_and_mask_cp3 = BIF_BX_PF0_GPU_HDP_FLUSH_DONE__CP3_MASK; - nbio_v6_1_hdp_flush_reg.ref_and_mask_cp4 = BIF_BX_PF0_GPU_HDP_FLUSH_DONE__CP4_MASK; - nbio_v6_1_hdp_flush_reg.ref_and_mask_cp5 = BIF_BX_PF0_GPU_HDP_FLUSH_DONE__CP5_MASK; - nbio_v6_1_hdp_flush_reg.ref_and_mask_cp6 = BIF_BX_PF0_GPU_HDP_FLUSH_DONE__CP6_MASK; - nbio_v6_1_hdp_flush_reg.ref_and_mask_cp7 = BIF_BX_PF0_GPU_HDP_FLUSH_DONE__CP7_MASK; - nbio_v6_1_hdp_flush_reg.ref_and_mask_cp8 = BIF_BX_PF0_GPU_HDP_FLUSH_DONE__CP8_MASK; - nbio_v6_1_hdp_flush_reg.ref_and_mask_cp9 = BIF_BX_PF0_GPU_HDP_FLUSH_DONE__CP9_MASK; - nbio_v6_1_hdp_flush_reg.ref_and_mask_sdma0 = BIF_BX_PF0_GPU_HDP_FLUSH_DONE__SDMA0_MASK; - nbio_v6_1_hdp_flush_reg.ref_and_mask_sdma1 = BIF_BX_PF0_GPU_HDP_FLUSH_DONE__SDMA1_MASK; - - nbio_v6_1_pcie_index_data.index_offset = SOC15_REG_OFFSET(NBIO, 0, mmPCIE_INDEX); - nbio_v6_1_pcie_index_data.data_offset = SOC15_REG_OFFSET(NBIO, 0, mmPCIE_DATA); - - return 0; -} +const struct nbio_pcie_index_data nbio_v6_1_pcie_index_data = { + .index_offset = SOC15_REG_OFFSET(NBIO, 0, mmPCIE_INDEX), + .data_offset = SOC15_REG_OFFSET(NBIO, 0, mmPCIE_DATA), +}; void nbio_v6_1_detect_hw_virt(struct amdgpu_device *adev) { diff --git a/drivers/gpu/drm/amd/amdgpu/nbio_v6_1.h b/drivers/gpu/drm/amd/amdgpu/nbio_v6_1.h index 686e4b4d296a..14ca8d45a46c 100644 --- a/drivers/gpu/drm/amd/amdgpu/nbio_v6_1.h +++ b/drivers/gpu/drm/amd/amdgpu/nbio_v6_1.h @@ -26,8 +26,8 @@ #include "soc15_common.h" -extern struct nbio_hdp_flush_reg nbio_v6_1_hdp_flush_reg; -extern struct nbio_pcie_index_data nbio_v6_1_pcie_index_data; +extern const struct nbio_hdp_flush_reg nbio_v6_1_hdp_flush_reg; +extern const struct nbio_pcie_index_data nbio_v6_1_pcie_index_data; int nbio_v6_1_init(struct amdgpu_device *adev); u32 nbio_v6_1_get_atombios_scratch_regs(struct amdgpu_device *adev, uint32_t idx); diff --git a/drivers/gpu/drm/amd/amdgpu/nbio_v7_0.c b/drivers/gpu/drm/amd/amdgpu/nbio_v7_0.c index 11b70d601922..f802b973410a 100644 --- a/drivers/gpu/drm/amd/amdgpu/nbio_v7_0.c +++ b/drivers/gpu/drm/amd/amdgpu/nbio_v7_0.c @@ -185,28 +185,24 @@ void nbio_v7_0_ih_control(struct amdgpu_device *adev) WREG32_SOC15(NBIO, 0, mmINTERRUPT_CNTL, interrupt_cntl); } -struct nbio_hdp_flush_reg nbio_v7_0_hdp_flush_reg; -struct nbio_pcie_index_data nbio_v7_0_pcie_index_data; +const struct nbio_hdp_flush_reg nbio_v7_0_hdp_flush_reg = { + .hdp_flush_req_offset = SOC15_REG_OFFSET(NBIO, 0, mmGPU_HDP_FLUSH_REQ), + .hdp_flush_done_offset = SOC15_REG_OFFSET(NBIO, 0, mmGPU_HDP_FLUSH_DONE), + .ref_and_mask_cp0 = GPU_HDP_FLUSH_DONE__CP0_MASK, + .ref_and_mask_cp1 = GPU_HDP_FLUSH_DONE__CP1_MASK, + .ref_and_mask_cp2 = GPU_HDP_FLUSH_DONE__CP2_MASK, + .ref_and_mask_cp3 = GPU_HDP_FLUSH_DONE__CP3_MASK, + .ref_and_mask_cp4 = GPU_HDP_FLUSH_DONE__CP4_MASK, + .ref_and_mask_cp5 = GPU_HDP_FLUSH_DONE__CP5_MASK, + .ref_and_mask_cp6 = GPU_HDP_FLUSH_DONE__CP6_MASK, + .ref_and_mask_cp7 = GPU_HDP_FLUSH_DONE__CP7_MASK, + .ref_and_mask_cp8 = GPU_HDP_FLUSH_DONE__CP8_MASK, + .ref_and_mask_cp9 = GPU_HDP_FLUSH_DONE__CP9_MASK, + .ref_and_mask_sdma0 = GPU_HDP_FLUSH_DONE__SDMA0_MASK, + .ref_and_mask_sdma1 = GPU_HDP_FLUSH_DONE__SDMA1_MASK, +}; -int nbio_v7_0_init(struct amdgpu_device *adev) -{ - nbio_v7_0_hdp_flush_reg.hdp_flush_req_offset = SOC15_REG_OFFSET(NBIO, 0, mmGPU_HDP_FLUSH_REQ); - nbio_v7_0_hdp_flush_reg.hdp_flush_done_offset = SOC15_REG_OFFSET(NBIO, 0, mmGPU_HDP_FLUSH_DONE); - nbio_v7_0_hdp_flush_reg.ref_and_mask_cp0 = GPU_HDP_FLUSH_DONE__CP0_MASK; - nbio_v7_0_hdp_flush_reg.ref_and_mask_cp1 = GPU_HDP_FLUSH_DONE__CP1_MASK; - nbio_v7_0_hdp_flush_reg.ref_and_mask_cp2 = GPU_HDP_FLUSH_DONE__CP2_MASK; - nbio_v7_0_hdp_flush_reg.ref_and_mask_cp3 = GPU_HDP_FLUSH_DONE__CP3_MASK; - nbio_v7_0_hdp_flush_reg.ref_and_mask_cp4 = GPU_HDP_FLUSH_DONE__CP4_MASK; - nbio_v7_0_hdp_flush_reg.ref_and_mask_cp5 = GPU_HDP_FLUSH_DONE__CP5_MASK; - nbio_v7_0_hdp_flush_reg.ref_and_mask_cp6 = GPU_HDP_FLUSH_DONE__CP6_MASK; - nbio_v7_0_hdp_flush_reg.ref_and_mask_cp7 = GPU_HDP_FLUSH_DONE__CP7_MASK; - nbio_v7_0_hdp_flush_reg.ref_and_mask_cp8 = GPU_HDP_FLUSH_DONE__CP8_MASK; - nbio_v7_0_hdp_flush_reg.ref_and_mask_cp9 = GPU_HDP_FLUSH_DONE__CP9_MASK; - nbio_v7_0_hdp_flush_reg.ref_and_mask_sdma0 = GPU_HDP_FLUSH_DONE__SDMA0_MASK; - nbio_v7_0_hdp_flush_reg.ref_and_mask_sdma1 = GPU_HDP_FLUSH_DONE__SDMA1_MASK; - - nbio_v7_0_pcie_index_data.index_offset = SOC15_REG_OFFSET(NBIO, 0, mmPCIE_INDEX2); - nbio_v7_0_pcie_index_data.data_offset = SOC15_REG_OFFSET(NBIO, 0, mmPCIE_DATA2); - - return 0; -} +const struct nbio_pcie_index_data nbio_v7_0_pcie_index_data = { + .index_offset = SOC15_REG_OFFSET(NBIO, 0, mmPCIE_INDEX2), + .data_offset = SOC15_REG_OFFSET(NBIO, 0, mmPCIE_DATA2) +}; diff --git a/drivers/gpu/drm/amd/amdgpu/nbio_v7_0.h b/drivers/gpu/drm/amd/amdgpu/nbio_v7_0.h index 054ff49427e6..df8fa90f40d7 100644 --- a/drivers/gpu/drm/amd/amdgpu/nbio_v7_0.h +++ b/drivers/gpu/drm/amd/amdgpu/nbio_v7_0.h @@ -26,8 +26,8 @@ #include "soc15_common.h" -extern struct nbio_hdp_flush_reg nbio_v7_0_hdp_flush_reg; -extern struct nbio_pcie_index_data nbio_v7_0_pcie_index_data; +extern const struct nbio_hdp_flush_reg nbio_v7_0_hdp_flush_reg; +extern const struct nbio_pcie_index_data nbio_v7_0_pcie_index_data; int nbio_v7_0_init(struct amdgpu_device *adev); u32 nbio_v7_0_get_atombios_scratch_regs(struct amdgpu_device *adev, uint32_t idx); diff --git a/drivers/gpu/drm/amd/amdgpu/sdma_v2_4.c b/drivers/gpu/drm/amd/amdgpu/sdma_v2_4.c index acdee3a4602c..67f375bfe452 100644 --- a/drivers/gpu/drm/amd/amdgpu/sdma_v2_4.c +++ b/drivers/gpu/drm/amd/amdgpu/sdma_v2_4.c @@ -561,21 +561,11 @@ static int sdma_v2_4_start(struct amdgpu_device *adev) { int r; - if (!adev->pp_enabled) { - if (adev->firmware.load_type != AMDGPU_FW_LOAD_SMU) { - r = sdma_v2_4_load_microcode(adev); - if (r) - return r; - } else { - r = adev->smu.smumgr_funcs->check_fw_load_finish(adev, - AMDGPU_UCODE_ID_SDMA0); - if (r) - return -EINVAL; - r = adev->smu.smumgr_funcs->check_fw_load_finish(adev, - AMDGPU_UCODE_ID_SDMA1); - if (r) - return -EINVAL; - } + + if (adev->firmware.load_type == AMDGPU_FW_LOAD_DIRECT) { + r = sdma_v2_4_load_microcode(adev); + if (r) + return r; } /* halt the engine before programing */ diff --git a/drivers/gpu/drm/amd/amdgpu/sdma_v3_0.c b/drivers/gpu/drm/amd/amdgpu/sdma_v3_0.c index 72f31cc7df00..6d06f8eb659f 100644 --- a/drivers/gpu/drm/amd/amdgpu/sdma_v3_0.c +++ b/drivers/gpu/drm/amd/amdgpu/sdma_v3_0.c @@ -819,23 +819,12 @@ static int sdma_v3_0_load_microcode(struct amdgpu_device *adev) */ static int sdma_v3_0_start(struct amdgpu_device *adev) { - int r, i; + int r; - if (!adev->pp_enabled) { - if (adev->firmware.load_type != AMDGPU_FW_LOAD_SMU) { - r = sdma_v3_0_load_microcode(adev); - if (r) - return r; - } else { - for (i = 0; i < adev->sdma.num_instances; i++) { - r = adev->smu.smumgr_funcs->check_fw_load_finish(adev, - (i == 0) ? - AMDGPU_UCODE_ID_SDMA0 : - AMDGPU_UCODE_ID_SDMA1); - if (r) - return -EINVAL; - } - } + if (adev->firmware.load_type == AMDGPU_FW_LOAD_DIRECT) { + r = sdma_v3_0_load_microcode(adev); + if (r) + return r; } /* disable sdma engine before programing it */ diff --git a/drivers/gpu/drm/amd/amdgpu/sdma_v4_0.c b/drivers/gpu/drm/amd/amdgpu/sdma_v4_0.c index c26d205ff3bf..46009db3d195 100644 --- a/drivers/gpu/drm/amd/amdgpu/sdma_v4_0.c +++ b/drivers/gpu/drm/amd/amdgpu/sdma_v4_0.c @@ -371,7 +371,7 @@ static void sdma_v4_0_ring_emit_ib(struct amdgpu_ring *ring, static void sdma_v4_0_ring_emit_hdp_flush(struct amdgpu_ring *ring) { u32 ref_and_mask = 0; - struct nbio_hdp_flush_reg *nbio_hf_reg; + const struct nbio_hdp_flush_reg *nbio_hf_reg; if (ring->adev->flags & AMD_IS_APU) nbio_hf_reg = &nbio_v7_0_hdp_flush_reg; diff --git a/drivers/gpu/drm/amd/amdgpu/si_dpm.c b/drivers/gpu/drm/amd/amdgpu/si_dpm.c index 9b8db6046271..51fd0c9a20a5 100644 --- a/drivers/gpu/drm/amd/amdgpu/si_dpm.c +++ b/drivers/gpu/drm/amd/amdgpu/si_dpm.c @@ -7604,11 +7604,6 @@ static int si_dpm_late_init(void *handle) if (!amdgpu_dpm) return 0; - /* init the sysfs and debugfs files late */ - ret = amdgpu_pm_sysfs_init(adev); - if (ret) - return ret; - ret = si_set_temperature_range(adev); if (ret) return ret; @@ -7764,7 +7759,6 @@ static int si_dpm_sw_fini(void *handle) flush_work(&adev->pm.dpm.thermal.work); mutex_lock(&adev->pm.mutex); - amdgpu_pm_sysfs_fini(adev); si_dpm_fini(adev); mutex_unlock(&adev->pm.mutex); diff --git a/drivers/gpu/drm/amd/amdgpu/soc15.c b/drivers/gpu/drm/amd/amdgpu/soc15.c index 245a18aeb389..1c006ba9d826 100644 --- a/drivers/gpu/drm/amd/amdgpu/soc15.c +++ b/drivers/gpu/drm/amd/amdgpu/soc15.c @@ -101,7 +101,7 @@ static u32 soc15_pcie_rreg(struct amdgpu_device *adev, u32 reg) { unsigned long flags, address, data; u32 r; - struct nbio_pcie_index_data *nbio_pcie_id; + const struct nbio_pcie_index_data *nbio_pcie_id; if (adev->flags & AMD_IS_APU) nbio_pcie_id = &nbio_v7_0_pcie_index_data; @@ -122,7 +122,7 @@ static u32 soc15_pcie_rreg(struct amdgpu_device *adev, u32 reg) static void soc15_pcie_wreg(struct amdgpu_device *adev, u32 reg, u32 v) { unsigned long flags, address, data; - struct nbio_pcie_index_data *nbio_pcie_id; + const struct nbio_pcie_index_data *nbio_pcie_id; if (adev->flags & AMD_IS_APU) nbio_pcie_id = &nbio_v7_0_pcie_index_data; @@ -604,21 +604,6 @@ static int soc15_common_early_init(void *handle) (amdgpu_ip_block_mask & (1 << AMD_IP_BLOCK_TYPE_PSP))) psp_enabled = true; - /* - * nbio need be used for both sdma and gfx9, but only - * initializes once - */ - switch(adev->asic_type) { - case CHIP_VEGA10: - nbio_v6_1_init(adev); - break; - case CHIP_RAVEN: - nbio_v7_0_init(adev); - break; - default: - return -EINVAL; - } - adev->rev_id = soc15_get_rev_id(adev); adev->external_rev_id = 0xFF; switch (adev->asic_type) { diff --git a/drivers/gpu/drm/amd/amdgpu/tonga_ih.c b/drivers/gpu/drm/amd/amdgpu/tonga_ih.c index 5ed00692618e..aa4e320e31f8 100644 --- a/drivers/gpu/drm/amd/amdgpu/tonga_ih.c +++ b/drivers/gpu/drm/amd/amdgpu/tonga_ih.c @@ -227,8 +227,23 @@ static u32 tonga_ih_get_wptr(struct amdgpu_device *adev) */ static bool tonga_ih_prescreen_iv(struct amdgpu_device *adev) { - /* Process all interrupts */ - return true; + u32 ring_index = adev->irq.ih.rptr >> 2; + u16 pasid; + + switch (le32_to_cpu(adev->irq.ih.ring[ring_index]) & 0xff) { + case 146: + case 147: + pasid = le32_to_cpu(adev->irq.ih.ring[ring_index + 2]) >> 16; + if (!pasid || amdgpu_vm_pasid_fault_credit(adev, pasid)) + return true; + break; + default: + /* Not a VM fault */ + return true; + } + + adev->irq.ih.rptr += 16; + return false; } /** diff --git a/drivers/gpu/drm/amd/amdgpu/uvd_v6_0.c b/drivers/gpu/drm/amd/amdgpu/uvd_v6_0.c index 31db356476f8..60af7310a234 100644 --- a/drivers/gpu/drm/amd/amdgpu/uvd_v6_0.c +++ b/drivers/gpu/drm/amd/amdgpu/uvd_v6_0.c @@ -38,6 +38,8 @@ #include "vi.h" static void uvd_v6_0_set_ring_funcs(struct amdgpu_device *adev); +static void uvd_v6_0_set_enc_ring_funcs(struct amdgpu_device *adev); + static void uvd_v6_0_set_irq_funcs(struct amdgpu_device *adev); static int uvd_v6_0_start(struct amdgpu_device *adev); static void uvd_v6_0_stop(struct amdgpu_device *adev); @@ -48,6 +50,18 @@ static void uvd_v6_0_enable_mgcg(struct amdgpu_device *adev, bool enable); /** +* uvd_v6_0_enc_support - get encode support status +* +* @adev: amdgpu_device pointer +* +* Returns the current hardware encode support status +*/ +static inline bool uvd_v6_0_enc_support(struct amdgpu_device *adev) +{ + return ((adev->asic_type >= CHIP_POLARIS10) && (adev->asic_type <= CHIP_POLARIS12)); +} + +/** * uvd_v6_0_ring_get_rptr - get read pointer * * @ring: amdgpu_ring pointer @@ -62,6 +76,22 @@ static uint64_t uvd_v6_0_ring_get_rptr(struct amdgpu_ring *ring) } /** + * uvd_v6_0_enc_ring_get_rptr - get enc read pointer + * + * @ring: amdgpu_ring pointer + * + * Returns the current hardware enc read pointer + */ +static uint64_t uvd_v6_0_enc_ring_get_rptr(struct amdgpu_ring *ring) +{ + struct amdgpu_device *adev = ring->adev; + + if (ring == &adev->uvd.ring_enc[0]) + return RREG32(mmUVD_RB_RPTR); + else + return RREG32(mmUVD_RB_RPTR2); +} +/** * uvd_v6_0_ring_get_wptr - get write pointer * * @ring: amdgpu_ring pointer @@ -76,6 +106,23 @@ static uint64_t uvd_v6_0_ring_get_wptr(struct amdgpu_ring *ring) } /** + * uvd_v6_0_enc_ring_get_wptr - get enc write pointer + * + * @ring: amdgpu_ring pointer + * + * Returns the current hardware enc write pointer + */ +static uint64_t uvd_v6_0_enc_ring_get_wptr(struct amdgpu_ring *ring) +{ + struct amdgpu_device *adev = ring->adev; + + if (ring == &adev->uvd.ring_enc[0]) + return RREG32(mmUVD_RB_WPTR); + else + return RREG32(mmUVD_RB_WPTR2); +} + +/** * uvd_v6_0_ring_set_wptr - set write pointer * * @ring: amdgpu_ring pointer @@ -89,11 +136,247 @@ static void uvd_v6_0_ring_set_wptr(struct amdgpu_ring *ring) WREG32(mmUVD_RBC_RB_WPTR, lower_32_bits(ring->wptr)); } +/** + * uvd_v6_0_enc_ring_set_wptr - set enc write pointer + * + * @ring: amdgpu_ring pointer + * + * Commits the enc write pointer to the hardware + */ +static void uvd_v6_0_enc_ring_set_wptr(struct amdgpu_ring *ring) +{ + struct amdgpu_device *adev = ring->adev; + + if (ring == &adev->uvd.ring_enc[0]) + WREG32(mmUVD_RB_WPTR, + lower_32_bits(ring->wptr)); + else + WREG32(mmUVD_RB_WPTR2, + lower_32_bits(ring->wptr)); +} + +/** + * uvd_v6_0_enc_ring_test_ring - test if UVD ENC ring is working + * + * @ring: the engine to test on + * + */ +static int uvd_v6_0_enc_ring_test_ring(struct amdgpu_ring *ring) +{ + struct amdgpu_device *adev = ring->adev; + uint32_t rptr = amdgpu_ring_get_rptr(ring); + unsigned i; + int r; + + r = amdgpu_ring_alloc(ring, 16); + if (r) { + DRM_ERROR("amdgpu: uvd enc failed to lock ring %d (%d).\n", + ring->idx, r); + return r; + } + amdgpu_ring_write(ring, HEVC_ENC_CMD_END); + amdgpu_ring_commit(ring); + + for (i = 0; i < adev->usec_timeout; i++) { + if (amdgpu_ring_get_rptr(ring) != rptr) + break; + DRM_UDELAY(1); + } + + if (i < adev->usec_timeout) { + DRM_INFO("ring test on %d succeeded in %d usecs\n", + ring->idx, i); + } else { + DRM_ERROR("amdgpu: ring %d test failed\n", + ring->idx); + r = -ETIMEDOUT; + } + + return r; +} + +/** + * uvd_v6_0_enc_get_create_msg - generate a UVD ENC create msg + * + * @adev: amdgpu_device pointer + * @ring: ring we should submit the msg to + * @handle: session handle to use + * @fence: optional fence to return + * + * Open up a stream for HW test + */ +static int uvd_v6_0_enc_get_create_msg(struct amdgpu_ring *ring, uint32_t handle, + struct dma_fence **fence) +{ + const unsigned ib_size_dw = 16; + struct amdgpu_job *job; + struct amdgpu_ib *ib; + struct dma_fence *f = NULL; + uint64_t dummy; + int i, r; + + r = amdgpu_job_alloc_with_ib(ring->adev, ib_size_dw * 4, &job); + if (r) + return r; + + ib = &job->ibs[0]; + dummy = ib->gpu_addr + 1024; + + ib->length_dw = 0; + ib->ptr[ib->length_dw++] = 0x00000018; + ib->ptr[ib->length_dw++] = 0x00000001; /* session info */ + ib->ptr[ib->length_dw++] = handle; + ib->ptr[ib->length_dw++] = 0x00010000; + ib->ptr[ib->length_dw++] = upper_32_bits(dummy); + ib->ptr[ib->length_dw++] = dummy; + + ib->ptr[ib->length_dw++] = 0x00000014; + ib->ptr[ib->length_dw++] = 0x00000002; /* task info */ + ib->ptr[ib->length_dw++] = 0x0000001c; + ib->ptr[ib->length_dw++] = 0x00000001; + ib->ptr[ib->length_dw++] = 0x00000000; + + ib->ptr[ib->length_dw++] = 0x00000008; + ib->ptr[ib->length_dw++] = 0x08000001; /* op initialize */ + + for (i = ib->length_dw; i < ib_size_dw; ++i) + ib->ptr[i] = 0x0; + + r = amdgpu_ib_schedule(ring, 1, ib, NULL, &f); + job->fence = dma_fence_get(f); + if (r) + goto err; + + amdgpu_job_free(job); + if (fence) + *fence = dma_fence_get(f); + dma_fence_put(f); + return 0; + +err: + amdgpu_job_free(job); + return r; +} + +/** + * uvd_v6_0_enc_get_destroy_msg - generate a UVD ENC destroy msg + * + * @adev: amdgpu_device pointer + * @ring: ring we should submit the msg to + * @handle: session handle to use + * @fence: optional fence to return + * + * Close up a stream for HW test or if userspace failed to do so + */ +int uvd_v6_0_enc_get_destroy_msg(struct amdgpu_ring *ring, uint32_t handle, + bool direct, struct dma_fence **fence) +{ + const unsigned ib_size_dw = 16; + struct amdgpu_job *job; + struct amdgpu_ib *ib; + struct dma_fence *f = NULL; + uint64_t dummy; + int i, r; + + r = amdgpu_job_alloc_with_ib(ring->adev, ib_size_dw * 4, &job); + if (r) + return r; + + ib = &job->ibs[0]; + dummy = ib->gpu_addr + 1024; + + ib->length_dw = 0; + ib->ptr[ib->length_dw++] = 0x00000018; + ib->ptr[ib->length_dw++] = 0x00000001; /* session info */ + ib->ptr[ib->length_dw++] = handle; + ib->ptr[ib->length_dw++] = 0x00010000; + ib->ptr[ib->length_dw++] = upper_32_bits(dummy); + ib->ptr[ib->length_dw++] = dummy; + + ib->ptr[ib->length_dw++] = 0x00000014; + ib->ptr[ib->length_dw++] = 0x00000002; /* task info */ + ib->ptr[ib->length_dw++] = 0x0000001c; + ib->ptr[ib->length_dw++] = 0x00000001; + ib->ptr[ib->length_dw++] = 0x00000000; + + ib->ptr[ib->length_dw++] = 0x00000008; + ib->ptr[ib->length_dw++] = 0x08000002; /* op close session */ + + for (i = ib->length_dw; i < ib_size_dw; ++i) + ib->ptr[i] = 0x0; + + if (direct) { + r = amdgpu_ib_schedule(ring, 1, ib, NULL, &f); + job->fence = dma_fence_get(f); + if (r) + goto err; + + amdgpu_job_free(job); + } else { + r = amdgpu_job_submit(job, ring, &ring->adev->vce.entity, + AMDGPU_FENCE_OWNER_UNDEFINED, &f); + if (r) + goto err; + } + + if (fence) + *fence = dma_fence_get(f); + dma_fence_put(f); + return 0; + +err: + amdgpu_job_free(job); + return r; +} + +/** + * uvd_v6_0_enc_ring_test_ib - test if UVD ENC IBs are working + * + * @ring: the engine to test on + * + */ +static int uvd_v6_0_enc_ring_test_ib(struct amdgpu_ring *ring, long timeout) +{ + struct dma_fence *fence = NULL; + long r; + + r = uvd_v6_0_enc_get_create_msg(ring, 1, NULL); + if (r) { + DRM_ERROR("amdgpu: failed to get create msg (%ld).\n", r); + goto error; + } + + r = uvd_v6_0_enc_get_destroy_msg(ring, 1, true, &fence); + if (r) { + DRM_ERROR("amdgpu: failed to get destroy ib (%ld).\n", r); + goto error; + } + + r = dma_fence_wait_timeout(fence, false, timeout); + if (r == 0) { + DRM_ERROR("amdgpu: IB test timed out.\n"); + r = -ETIMEDOUT; + } else if (r < 0) { + DRM_ERROR("amdgpu: fence wait failed (%ld).\n", r); + } else { + DRM_INFO("ib test on ring %d succeeded\n", ring->idx); + r = 0; + } +error: + dma_fence_put(fence); + return r; +} static int uvd_v6_0_early_init(void *handle) { struct amdgpu_device *adev = (struct amdgpu_device *)handle; uvd_v6_0_set_ring_funcs(adev); + + if (uvd_v6_0_enc_support(adev)) { + adev->uvd.num_enc_rings = 2; + uvd_v6_0_set_enc_ring_funcs(adev); + } + uvd_v6_0_set_irq_funcs(adev); return 0; @@ -102,7 +385,7 @@ static int uvd_v6_0_early_init(void *handle) static int uvd_v6_0_sw_init(void *handle) { struct amdgpu_ring *ring; - int r; + int i, r; struct amdgpu_device *adev = (struct amdgpu_device *)handle; /* UVD TRAP */ @@ -110,10 +393,31 @@ static int uvd_v6_0_sw_init(void *handle) if (r) return r; + /* UVD ENC TRAP */ + if (uvd_v6_0_enc_support(adev)) { + for (i = 0; i < adev->uvd.num_enc_rings; ++i) { + r = amdgpu_irq_add_id(adev, AMDGPU_IH_CLIENTID_LEGACY, i + 119, &adev->uvd.irq); + if (r) + return r; + } + } + r = amdgpu_uvd_sw_init(adev); if (r) return r; + if (uvd_v6_0_enc_support(adev)) { + struct amd_sched_rq *rq; + ring = &adev->uvd.ring_enc[0]; + rq = &ring->sched.sched_rq[AMD_SCHED_PRIORITY_NORMAL]; + r = amd_sched_entity_init(&ring->sched, &adev->uvd.entity_enc, + rq, amdgpu_sched_jobs); + if (r) { + DRM_ERROR("Failed setting up UVD ENC run queue.\n"); + return r; + } + } + r = amdgpu_uvd_resume(adev); if (r) return r; @@ -121,19 +425,38 @@ static int uvd_v6_0_sw_init(void *handle) ring = &adev->uvd.ring; sprintf(ring->name, "uvd"); r = amdgpu_ring_init(adev, ring, 512, &adev->uvd.irq, 0); + if (r) + return r; + + if (uvd_v6_0_enc_support(adev)) { + for (i = 0; i < adev->uvd.num_enc_rings; ++i) { + ring = &adev->uvd.ring_enc[i]; + sprintf(ring->name, "uvd_enc%d", i); + r = amdgpu_ring_init(adev, ring, 512, &adev->uvd.irq, 0); + if (r) + return r; + } + } return r; } static int uvd_v6_0_sw_fini(void *handle) { - int r; + int i, r; struct amdgpu_device *adev = (struct amdgpu_device *)handle; r = amdgpu_uvd_suspend(adev); if (r) return r; + if (uvd_v6_0_enc_support(adev)) { + amd_sched_entity_fini(&adev->uvd.ring_enc[0].sched, &adev->uvd.entity_enc); + + for (i = 0; i < adev->uvd.num_enc_rings; ++i) + amdgpu_ring_fini(&adev->uvd.ring_enc[i]); + } + return amdgpu_uvd_sw_fini(adev); } @@ -149,7 +472,7 @@ static int uvd_v6_0_hw_init(void *handle) struct amdgpu_device *adev = (struct amdgpu_device *)handle; struct amdgpu_ring *ring = &adev->uvd.ring; uint32_t tmp; - int r; + int i, r; amdgpu_asic_set_uvd_clocks(adev, 10000, 10000); uvd_v6_0_set_clockgating_state(adev, AMD_CG_STATE_UNGATE); @@ -189,9 +512,25 @@ static int uvd_v6_0_hw_init(void *handle) amdgpu_ring_commit(ring); + if (uvd_v6_0_enc_support(adev)) { + for (i = 0; i < adev->uvd.num_enc_rings; ++i) { + ring = &adev->uvd.ring_enc[i]; + ring->ready = true; + r = amdgpu_ring_test_ring(ring); + if (r) { + ring->ready = false; + goto done; + } + } + } + done: - if (!r) - DRM_INFO("UVD initialized successfully.\n"); + if (!r) { + if (uvd_v6_0_enc_support(adev)) + DRM_INFO("UVD and UVD ENC initialized successfully.\n"); + else + DRM_INFO("UVD initialized successfully.\n"); + } return r; } @@ -514,6 +853,22 @@ static int uvd_v6_0_start(struct amdgpu_device *adev) WREG32_FIELD(UVD_RBC_RB_CNTL, RB_NO_FETCH, 0); + if (uvd_v6_0_enc_support(adev)) { + ring = &adev->uvd.ring_enc[0]; + WREG32(mmUVD_RB_RPTR, lower_32_bits(ring->wptr)); + WREG32(mmUVD_RB_WPTR, lower_32_bits(ring->wptr)); + WREG32(mmUVD_RB_BASE_LO, ring->gpu_addr); + WREG32(mmUVD_RB_BASE_HI, upper_32_bits(ring->gpu_addr)); + WREG32(mmUVD_RB_SIZE, ring->ring_size / 4); + + ring = &adev->uvd.ring_enc[1]; + WREG32(mmUVD_RB_RPTR2, lower_32_bits(ring->wptr)); + WREG32(mmUVD_RB_WPTR2, lower_32_bits(ring->wptr)); + WREG32(mmUVD_RB_BASE_LO2, ring->gpu_addr); + WREG32(mmUVD_RB_BASE_HI2, upper_32_bits(ring->gpu_addr)); + WREG32(mmUVD_RB_SIZE2, ring->ring_size / 4); + } + return 0; } @@ -577,6 +932,26 @@ static void uvd_v6_0_ring_emit_fence(struct amdgpu_ring *ring, u64 addr, u64 seq } /** + * uvd_v6_0_enc_ring_emit_fence - emit an enc fence & trap command + * + * @ring: amdgpu_ring pointer + * @fence: fence to emit + * + * Write enc a fence and a trap command to the ring. + */ +static void uvd_v6_0_enc_ring_emit_fence(struct amdgpu_ring *ring, u64 addr, + u64 seq, unsigned flags) +{ + WARN_ON(flags & AMDGPU_FENCE_FLAG_64BIT); + + amdgpu_ring_write(ring, HEVC_ENC_CMD_FENCE); + amdgpu_ring_write(ring, addr); + amdgpu_ring_write(ring, upper_32_bits(addr)); + amdgpu_ring_write(ring, seq); + amdgpu_ring_write(ring, HEVC_ENC_CMD_TRAP); +} + +/** * uvd_v6_0_ring_emit_hdp_flush - emit an hdp flush * * @ring: amdgpu_ring pointer @@ -667,6 +1042,24 @@ static void uvd_v6_0_ring_emit_ib(struct amdgpu_ring *ring, amdgpu_ring_write(ring, ib->length_dw); } +/** + * uvd_v6_0_enc_ring_emit_ib - enc execute indirect buffer + * + * @ring: amdgpu_ring pointer + * @ib: indirect buffer to execute + * + * Write enc ring commands to execute the indirect buffer + */ +static void uvd_v6_0_enc_ring_emit_ib(struct amdgpu_ring *ring, + struct amdgpu_ib *ib, unsigned int vm_id, bool ctx_switch) +{ + amdgpu_ring_write(ring, HEVC_ENC_CMD_IB_VM); + amdgpu_ring_write(ring, vm_id); + amdgpu_ring_write(ring, lower_32_bits(ib->gpu_addr)); + amdgpu_ring_write(ring, upper_32_bits(ib->gpu_addr)); + amdgpu_ring_write(ring, ib->length_dw); +} + static void uvd_v6_0_ring_emit_vm_flush(struct amdgpu_ring *ring, unsigned vm_id, uint64_t pd_addr) { @@ -718,6 +1111,33 @@ static void uvd_v6_0_ring_emit_pipeline_sync(struct amdgpu_ring *ring) amdgpu_ring_write(ring, 0xE); } +static void uvd_v6_0_enc_ring_emit_pipeline_sync(struct amdgpu_ring *ring) +{ + uint32_t seq = ring->fence_drv.sync_seq; + uint64_t addr = ring->fence_drv.gpu_addr; + + amdgpu_ring_write(ring, HEVC_ENC_CMD_WAIT_GE); + amdgpu_ring_write(ring, lower_32_bits(addr)); + amdgpu_ring_write(ring, upper_32_bits(addr)); + amdgpu_ring_write(ring, seq); +} + +static void uvd_v6_0_enc_ring_insert_end(struct amdgpu_ring *ring) +{ + amdgpu_ring_write(ring, HEVC_ENC_CMD_END); +} + +static void uvd_v6_0_enc_ring_emit_vm_flush(struct amdgpu_ring *ring, + unsigned int vm_id, uint64_t pd_addr) +{ + amdgpu_ring_write(ring, HEVC_ENC_CMD_UPDATE_PTB); + amdgpu_ring_write(ring, vm_id); + amdgpu_ring_write(ring, pd_addr >> 12); + + amdgpu_ring_write(ring, HEVC_ENC_CMD_FLUSH_TLB); + amdgpu_ring_write(ring, vm_id); +} + static bool uvd_v6_0_is_idle(void *handle) { struct amdgpu_device *adev = (struct amdgpu_device *)handle; @@ -825,8 +1245,31 @@ static int uvd_v6_0_process_interrupt(struct amdgpu_device *adev, struct amdgpu_irq_src *source, struct amdgpu_iv_entry *entry) { + bool int_handled = true; DRM_DEBUG("IH: UVD TRAP\n"); - amdgpu_fence_process(&adev->uvd.ring); + + switch (entry->src_id) { + case 124: + amdgpu_fence_process(&adev->uvd.ring); + break; + case 119: + if (likely(uvd_v6_0_enc_support(adev))) + amdgpu_fence_process(&adev->uvd.ring_enc[0]); + else + int_handled = false; + break; + case 120: + if (likely(uvd_v6_0_enc_support(adev))) + amdgpu_fence_process(&adev->uvd.ring_enc[1]); + else + int_handled = false; + break; + } + + if (false == int_handled) + DRM_ERROR("Unhandled interrupt: %d %d\n", + entry->src_id, entry->src_data[0]); + return 0; } @@ -1153,6 +1596,33 @@ static const struct amdgpu_ring_funcs uvd_v6_0_ring_vm_funcs = { .end_use = amdgpu_uvd_ring_end_use, }; +static const struct amdgpu_ring_funcs uvd_v6_0_enc_ring_vm_funcs = { + .type = AMDGPU_RING_TYPE_UVD_ENC, + .align_mask = 0x3f, + .nop = HEVC_ENC_CMD_NO_OP, + .support_64bit_ptrs = false, + .get_rptr = uvd_v6_0_enc_ring_get_rptr, + .get_wptr = uvd_v6_0_enc_ring_get_wptr, + .set_wptr = uvd_v6_0_enc_ring_set_wptr, + .emit_frame_size = + 4 + /* uvd_v6_0_enc_ring_emit_pipeline_sync */ + 6 + /* uvd_v6_0_enc_ring_emit_vm_flush */ + 5 + 5 + /* uvd_v6_0_enc_ring_emit_fence x2 vm fence */ + 1, /* uvd_v6_0_enc_ring_insert_end */ + .emit_ib_size = 5, /* uvd_v6_0_enc_ring_emit_ib */ + .emit_ib = uvd_v6_0_enc_ring_emit_ib, + .emit_fence = uvd_v6_0_enc_ring_emit_fence, + .emit_vm_flush = uvd_v6_0_enc_ring_emit_vm_flush, + .emit_pipeline_sync = uvd_v6_0_enc_ring_emit_pipeline_sync, + .test_ring = uvd_v6_0_enc_ring_test_ring, + .test_ib = uvd_v6_0_enc_ring_test_ib, + .insert_nop = amdgpu_ring_insert_nop, + .insert_end = uvd_v6_0_enc_ring_insert_end, + .pad_ib = amdgpu_ring_generic_pad_ib, + .begin_use = amdgpu_uvd_ring_begin_use, + .end_use = amdgpu_uvd_ring_end_use, +}; + static void uvd_v6_0_set_ring_funcs(struct amdgpu_device *adev) { if (adev->asic_type >= CHIP_POLARIS10) { @@ -1164,6 +1634,16 @@ static void uvd_v6_0_set_ring_funcs(struct amdgpu_device *adev) } } +static void uvd_v6_0_set_enc_ring_funcs(struct amdgpu_device *adev) +{ + int i; + + for (i = 0; i < adev->uvd.num_enc_rings; ++i) + adev->uvd.ring_enc[i].funcs = &uvd_v6_0_enc_ring_vm_funcs; + + DRM_INFO("UVD ENC is enabled in VM mode\n"); +} + static const struct amdgpu_irq_src_funcs uvd_v6_0_irq_funcs = { .set = uvd_v6_0_set_interrupt_state, .process = uvd_v6_0_process_interrupt, @@ -1171,7 +1651,11 @@ static const struct amdgpu_irq_src_funcs uvd_v6_0_irq_funcs = { static void uvd_v6_0_set_irq_funcs(struct amdgpu_device *adev) { - adev->uvd.irq.num_types = 1; + if (uvd_v6_0_enc_support(adev)) + adev->uvd.irq.num_types = adev->uvd.num_enc_rings + 1; + else + adev->uvd.irq.num_types = 1; + adev->uvd.irq.funcs = &uvd_v6_0_irq_funcs; } diff --git a/drivers/gpu/drm/amd/amdgpu/vega10_ih.c b/drivers/gpu/drm/amd/amdgpu/vega10_ih.c index a3b30d84dbb3..697325737ba8 100644 --- a/drivers/gpu/drm/amd/amdgpu/vega10_ih.c +++ b/drivers/gpu/drm/amd/amdgpu/vega10_ih.c @@ -260,15 +260,18 @@ static bool vega10_ih_prescreen_iv(struct amdgpu_device *adev) return true; } - /* Not a retry fault */ - if (!(dw5 & 0x80)) - return true; - pasid = dw3 & 0xffff; /* No PASID, can't identify faulting process */ if (!pasid) return true; + /* Not a retry fault, check fault credit */ + if (!(dw5 & 0x80)) { + if (!amdgpu_vm_pasid_fault_credit(adev, pasid)) + goto ignore_iv; + return true; + } + addr = ((u64)(dw5 & 0xf) << 44) | ((u64)dw4 << 12); key = AMDGPU_VM_FAULT(pasid, addr); r = amdgpu_ih_add_fault(adev, key); diff --git a/drivers/gpu/drm/amd/amdgpu/vi.c b/drivers/gpu/drm/amd/amdgpu/vi.c index 9ff69b90df36..f3cfef48aa99 100644 --- a/drivers/gpu/drm/amd/amdgpu/vi.c +++ b/drivers/gpu/drm/amd/amdgpu/vi.c @@ -1254,7 +1254,6 @@ static int vi_common_set_clockgating_state_by_smu(void *handle, uint32_t msg_id, pp_state = 0; uint32_t pp_support_state = 0; struct amdgpu_device *adev = (struct amdgpu_device *)handle; - void *pp_handle = adev->powerplay.pp_handle; if (adev->cg_flags & (AMD_CG_SUPPORT_MC_LS | AMD_CG_SUPPORT_MC_MGCG)) { if (adev->cg_flags & AMD_CG_SUPPORT_MC_LS) { @@ -1271,7 +1270,8 @@ static int vi_common_set_clockgating_state_by_smu(void *handle, PP_BLOCK_SYS_MC, pp_support_state, pp_state); - amd_set_clockgating_by_smu(pp_handle, msg_id); + if (adev->powerplay.pp_funcs->set_clockgating_by_smu) + amdgpu_dpm_set_clockgating_by_smu(adev, msg_id); } if (adev->cg_flags & (AMD_CG_SUPPORT_SDMA_LS | AMD_CG_SUPPORT_SDMA_MGCG)) { @@ -1289,7 +1289,8 @@ static int vi_common_set_clockgating_state_by_smu(void *handle, PP_BLOCK_SYS_SDMA, pp_support_state, pp_state); - amd_set_clockgating_by_smu(pp_handle, msg_id); + if (adev->powerplay.pp_funcs->set_clockgating_by_smu) + amdgpu_dpm_set_clockgating_by_smu(adev, msg_id); } if (adev->cg_flags & (AMD_CG_SUPPORT_HDP_LS | AMD_CG_SUPPORT_HDP_MGCG)) { @@ -1307,7 +1308,8 @@ static int vi_common_set_clockgating_state_by_smu(void *handle, PP_BLOCK_SYS_HDP, pp_support_state, pp_state); - amd_set_clockgating_by_smu(pp_handle, msg_id); + if (adev->powerplay.pp_funcs->set_clockgating_by_smu) + amdgpu_dpm_set_clockgating_by_smu(adev, msg_id); } @@ -1321,7 +1323,8 @@ static int vi_common_set_clockgating_state_by_smu(void *handle, PP_BLOCK_SYS_BIF, PP_STATE_SUPPORT_LS, pp_state); - amd_set_clockgating_by_smu(pp_handle, msg_id); + if (adev->powerplay.pp_funcs->set_clockgating_by_smu) + amdgpu_dpm_set_clockgating_by_smu(adev, msg_id); } if (adev->cg_flags & AMD_CG_SUPPORT_BIF_MGCG) { if (state == AMD_CG_STATE_UNGATE) @@ -1333,7 +1336,8 @@ static int vi_common_set_clockgating_state_by_smu(void *handle, PP_BLOCK_SYS_BIF, PP_STATE_SUPPORT_CG, pp_state); - amd_set_clockgating_by_smu(pp_handle, msg_id); + if (adev->powerplay.pp_funcs->set_clockgating_by_smu) + amdgpu_dpm_set_clockgating_by_smu(adev, msg_id); } if (adev->cg_flags & AMD_CG_SUPPORT_DRM_LS) { @@ -1347,7 +1351,8 @@ static int vi_common_set_clockgating_state_by_smu(void *handle, PP_BLOCK_SYS_DRM, PP_STATE_SUPPORT_LS, pp_state); - amd_set_clockgating_by_smu(pp_handle, msg_id); + if (adev->powerplay.pp_funcs->set_clockgating_by_smu) + amdgpu_dpm_set_clockgating_by_smu(adev, msg_id); } if (adev->cg_flags & AMD_CG_SUPPORT_ROM_MGCG) { @@ -1361,7 +1366,8 @@ static int vi_common_set_clockgating_state_by_smu(void *handle, PP_BLOCK_SYS_ROM, PP_STATE_SUPPORT_CG, pp_state); - amd_set_clockgating_by_smu(pp_handle, msg_id); + if (adev->powerplay.pp_funcs->set_clockgating_by_smu) + amdgpu_dpm_set_clockgating_by_smu(adev, msg_id); } return 0; } diff --git a/drivers/gpu/drm/amd/amdgpu/vid.h b/drivers/gpu/drm/amd/amdgpu/vid.h index a6485254a169..dbf3703cbd1b 100644 --- a/drivers/gpu/drm/amd/amdgpu/vid.h +++ b/drivers/gpu/drm/amd/amdgpu/vid.h @@ -465,6 +465,16 @@ #define VCE_CMD_UPDATE_PTB 0x00000107 #define VCE_CMD_FLUSH_TLB 0x00000108 +/* HEVC ENC */ +#define HEVC_ENC_CMD_NO_OP 0x00000000 +#define HEVC_ENC_CMD_END 0x00000001 +#define HEVC_ENC_CMD_FENCE 0x00000003 +#define HEVC_ENC_CMD_TRAP 0x00000004 +#define HEVC_ENC_CMD_IB_VM 0x00000102 +#define HEVC_ENC_CMD_WAIT_GE 0x00000106 +#define HEVC_ENC_CMD_UPDATE_PTB 0x00000107 +#define HEVC_ENC_CMD_FLUSH_TLB 0x00000108 + /* mmPA_SC_RASTER_CONFIG mask */ #define RB_MAP_PKR0(x) ((x) << 0) #define RB_MAP_PKR0_MASK (0x3 << 0) |