diff options
Diffstat (limited to 'drivers/gpu/drm/amd/amdgpu')
68 files changed, 1315 insertions, 1015 deletions
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu.h b/drivers/gpu/drm/amd/amdgpu/amdgpu.h index 81a531b652aa..f42e8d467c12 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu.h +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu.h @@ -636,9 +636,8 @@ struct amdgpu_fw_vram_usage { struct amdgpu_bo *reserved_bo; void *va; - /* Offset on the top of VRAM, used as c2p write buffer. + /* GDDR6 training support flag. */ - u64 mem_train_fb_loc; bool mem_train_support; }; @@ -994,8 +993,6 @@ struct amdgpu_device { bool pm_sysfs_en; bool ucode_sysfs_en; - - bool in_baco; }; static inline struct amdgpu_device *amdgpu_ttm_adev(struct ttm_bo_device *bdev) diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_arcturus.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_arcturus.c index b6713e0ed1b2..3c119407dc34 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_arcturus.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_arcturus.c @@ -46,6 +46,8 @@ #include "soc15.h" #include "soc15d.h" #include "amdgpu_amdkfd_gfx_v9.h" +#include "gfxhub_v1_0.h" +#include "mmhub_v9_4.h" #define HQD_N_REGS 56 #define DUMP_REG(addr) do { \ @@ -258,6 +260,22 @@ static int kgd_hqd_sdma_destroy(struct kgd_dev *kgd, void *mqd, return 0; } +static void kgd_set_vm_context_page_table_base(struct kgd_dev *kgd, uint32_t vmid, + uint64_t page_table_base) +{ + struct amdgpu_device *adev = get_amdgpu_device(kgd); + + if (!amdgpu_amdkfd_is_kfd_vmid(adev, vmid)) { + pr_err("trying to set page table base for wrong VMID %u\n", + vmid); + return; + } + + mmhub_v9_4_setup_vm_pt_regs(adev, vmid, page_table_base); + + gfxhub_v1_0_setup_vm_pt_regs(adev, vmid, page_table_base); +} + const struct kfd2kgd_calls arcturus_kfd2kgd = { .program_sh_mem_settings = kgd_gfx_v9_program_sh_mem_settings, .set_pasid_vmid_mapping = kgd_gfx_v9_set_pasid_vmid_mapping, @@ -277,7 +295,7 @@ const struct kfd2kgd_calls arcturus_kfd2kgd = { .get_atc_vmid_pasid_mapping_info = kgd_gfx_v9_get_atc_vmid_pasid_mapping_info, .get_tile_config = kgd_gfx_v9_get_tile_config, - .set_vm_context_page_table_base = kgd_gfx_v9_set_vm_context_page_table_base, + .set_vm_context_page_table_base = kgd_set_vm_context_page_table_base, .invalidate_tlbs = kgd_gfx_v9_invalidate_tlbs, .invalidate_tlbs_vmid = kgd_gfx_v9_invalidate_tlbs_vmid, .get_hive_id = amdgpu_amdkfd_get_hive_id, 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 6f1a4676ddde..e7861f0ef415 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gfx_v9.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gfx_v9.c @@ -40,7 +40,6 @@ #include "soc15d.h" #include "mmhub_v1_0.h" #include "gfxhub_v1_0.h" -#include "mmhub_v9_4.h" enum hqd_dequeue_request_type { @@ -758,8 +757,8 @@ uint32_t kgd_gfx_v9_address_watch_get_offset(struct kgd_dev *kgd, return 0; } -void kgd_gfx_v9_set_vm_context_page_table_base(struct kgd_dev *kgd, uint32_t vmid, - uint64_t page_table_base) +static void kgd_gfx_v9_set_vm_context_page_table_base(struct kgd_dev *kgd, + uint32_t vmid, uint64_t page_table_base) { struct amdgpu_device *adev = get_amdgpu_device(kgd); @@ -769,14 +768,7 @@ void kgd_gfx_v9_set_vm_context_page_table_base(struct kgd_dev *kgd, uint32_t vmi return; } - /* TODO: take advantage of per-process address space size. For - * now, all processes share the same address space size, like - * on GFX8 and older. - */ - if (adev->asic_type == CHIP_ARCTURUS) { - mmhub_v9_4_setup_vm_pt_regs(adev, vmid, page_table_base); - } else - mmhub_v1_0_setup_vm_pt_regs(adev, vmid, page_table_base); + mmhub_v1_0_setup_vm_pt_regs(adev, vmid, page_table_base); gfxhub_v1_0_setup_vm_pt_regs(adev, vmid, page_table_base); } diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gfx_v9.h b/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gfx_v9.h index d9e9ad22b2bd..02b1426d17d1 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gfx_v9.h +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gfx_v9.h @@ -57,8 +57,6 @@ uint32_t kgd_gfx_v9_address_watch_get_offset(struct kgd_dev *kgd, bool kgd_gfx_v9_get_atc_vmid_pasid_mapping_info(struct kgd_dev *kgd, uint8_t vmid, uint16_t *p_pasid); -void kgd_gfx_v9_set_vm_context_page_table_base(struct kgd_dev *kgd, uint32_t vmid, - uint64_t page_table_base); int kgd_gfx_v9_invalidate_tlbs(struct kgd_dev *kgd, uint16_t pasid); int kgd_gfx_v9_invalidate_tlbs_vmid(struct kgd_dev *kgd, uint16_t vmid); int kgd_gfx_v9_get_tile_config(struct kgd_dev *kgd, diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_atombios.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_atombios.c index 9ba80d828876..fdd52d86a4d7 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_atombios.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_atombios.c @@ -2022,7 +2022,7 @@ int amdgpu_atombios_init(struct amdgpu_device *adev) if (adev->is_atom_fw) { amdgpu_atomfirmware_scratch_regs_init(adev); amdgpu_atomfirmware_allocate_fb_scratch(adev); - ret = amdgpu_atomfirmware_get_mem_train_fb_loc(adev); + ret = amdgpu_atomfirmware_get_mem_train_info(adev); if (ret) { DRM_ERROR("Failed to get mem train fb location.\n"); return ret; diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_atomfirmware.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_atomfirmware.c index ff4eb96bdfb5..58f9d8c3a17a 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_atomfirmware.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_atomfirmware.c @@ -525,16 +525,12 @@ static int gddr6_mem_train_support(struct amdgpu_device *adev) return ret; } -int amdgpu_atomfirmware_get_mem_train_fb_loc(struct amdgpu_device *adev) +int amdgpu_atomfirmware_get_mem_train_info(struct amdgpu_device *adev) { struct atom_context *ctx = adev->mode_info.atom_context; - unsigned char *bios = ctx->bios; - struct vram_reserve_block *reserved_block; - int index, block_number; + int index; uint8_t frev, crev; uint16_t data_offset, size; - uint32_t start_address_in_kb; - uint64_t offset; int ret; adev->fw_vram_usage.mem_train_support = false; @@ -569,32 +565,6 @@ int amdgpu_atomfirmware_get_mem_train_fb_loc(struct amdgpu_device *adev) return -EINVAL; } - reserved_block = (struct vram_reserve_block *) - (bios + data_offset + sizeof(struct atom_common_table_header)); - block_number = ((unsigned int)size - sizeof(struct atom_common_table_header)) - / sizeof(struct vram_reserve_block); - reserved_block += (block_number > 0) ? block_number-1 : 0; - DRM_DEBUG("block_number:0x%04x, last block: 0x%08xkb sz, %dkb fw, %dkb drv.\n", - block_number, - le32_to_cpu(reserved_block->start_address_in_kb), - le16_to_cpu(reserved_block->used_by_firmware_in_kb), - le16_to_cpu(reserved_block->used_by_driver_in_kb)); - if (reserved_block->used_by_firmware_in_kb > 0) { - start_address_in_kb = le32_to_cpu(reserved_block->start_address_in_kb); - offset = (uint64_t)start_address_in_kb * ONE_KiB; - if ((offset & (ONE_MiB - 1)) < (4 * ONE_KiB + 1) ) { - offset -= ONE_MiB; - } - - offset &= ~(ONE_MiB - 1); - adev->fw_vram_usage.mem_train_fb_loc = offset; - adev->fw_vram_usage.mem_train_support = true; - DRM_DEBUG("mem_train_fb_loc:0x%09llx.\n", offset); - ret = 0; - } else { - DRM_ERROR("used_by_firmware_in_kb is 0!\n"); - ret = -EINVAL; - } - - return ret; + adev->fw_vram_usage.mem_train_support = true; + return 0; } diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_atomfirmware.h b/drivers/gpu/drm/amd/amdgpu/amdgpu_atomfirmware.h index f871af5ea6f3..434fe2fa0089 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_atomfirmware.h +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_atomfirmware.h @@ -31,7 +31,7 @@ void amdgpu_atomfirmware_scratch_regs_init(struct amdgpu_device *adev); int amdgpu_atomfirmware_allocate_fb_scratch(struct amdgpu_device *adev); int amdgpu_atomfirmware_get_vram_info(struct amdgpu_device *adev, int *vram_width, int *vram_type, int *vram_vendor); -int amdgpu_atomfirmware_get_mem_train_fb_loc(struct amdgpu_device *adev); +int amdgpu_atomfirmware_get_mem_train_info(struct amdgpu_device *adev); int amdgpu_atomfirmware_get_clock_info(struct amdgpu_device *adev); int amdgpu_atomfirmware_get_gfx_info(struct amdgpu_device *adev); bool amdgpu_atomfirmware_mem_ecc_supported(struct amdgpu_device *adev); diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_atpx_handler.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_atpx_handler.c index a97fb759e2f4..3e35a8f2c5e5 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_atpx_handler.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_atpx_handler.c @@ -613,7 +613,17 @@ static bool amdgpu_atpx_detect(void) bool d3_supported = false; struct pci_dev *parent_pdev; - while ((pdev = pci_get_class(PCI_BASE_CLASS_DISPLAY << 16, pdev)) != NULL) { + while ((pdev = pci_get_class(PCI_CLASS_DISPLAY_VGA << 8, pdev)) != NULL) { + vga_count++; + + has_atpx |= (amdgpu_atpx_pci_probe_handle(pdev) == true); + + parent_pdev = pci_upstream_bridge(pdev); + d3_supported |= parent_pdev && parent_pdev->bridge_d3; + amdgpu_atpx_get_quirks(pdev); + } + + while ((pdev = pci_get_class(PCI_CLASS_DISPLAY_OTHER << 8, pdev)) != NULL) { vga_count++; has_atpx |= (amdgpu_atpx_pci_probe_handle(pdev) == true); diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_ctx.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_ctx.c index 1d2bbf10614e..64e2babbc36e 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_ctx.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_ctx.c @@ -74,7 +74,7 @@ static int amdgpu_ctx_init(struct amdgpu_device *adev, struct amdgpu_ctx *ctx) { unsigned num_entities = amdgpu_ctx_total_num_entities(); - unsigned i, j, k; + unsigned i, j; int r; if (priority < 0 || priority >= DRM_SCHED_PRIORITY_MAX) @@ -121,72 +121,57 @@ static int amdgpu_ctx_init(struct amdgpu_device *adev, ctx->override_priority = DRM_SCHED_PRIORITY_UNSET; for (i = 0; i < AMDGPU_HW_IP_NUM; ++i) { - struct amdgpu_ring *rings[AMDGPU_MAX_RINGS]; - struct drm_sched_rq *rqs[AMDGPU_MAX_RINGS]; - unsigned num_rings = 0; - unsigned num_rqs = 0; + struct drm_gpu_scheduler **scheds; + struct drm_gpu_scheduler *sched; + unsigned num_scheds = 0; switch (i) { case AMDGPU_HW_IP_GFX: - rings[0] = &adev->gfx.gfx_ring[0]; - num_rings = 1; + sched = &adev->gfx.gfx_ring[0].sched; + scheds = &sched; + num_scheds = 1; break; case AMDGPU_HW_IP_COMPUTE: - for (j = 0; j < adev->gfx.num_compute_rings; ++j) - rings[j] = &adev->gfx.compute_ring[j]; - num_rings = adev->gfx.num_compute_rings; + scheds = adev->gfx.compute_sched; + num_scheds = adev->gfx.num_compute_sched; break; case AMDGPU_HW_IP_DMA: - for (j = 0; j < adev->sdma.num_instances; ++j) - rings[j] = &adev->sdma.instance[j].ring; - num_rings = adev->sdma.num_instances; + scheds = adev->sdma.sdma_sched; + num_scheds = adev->sdma.num_sdma_sched; break; case AMDGPU_HW_IP_UVD: - rings[0] = &adev->uvd.inst[0].ring; - num_rings = 1; + sched = &adev->uvd.inst[0].ring.sched; + scheds = &sched; + num_scheds = 1; break; case AMDGPU_HW_IP_VCE: - rings[0] = &adev->vce.ring[0]; - num_rings = 1; + sched = &adev->vce.ring[0].sched; + scheds = &sched; + num_scheds = 1; break; case AMDGPU_HW_IP_UVD_ENC: - rings[0] = &adev->uvd.inst[0].ring_enc[0]; - num_rings = 1; + sched = &adev->uvd.inst[0].ring_enc[0].sched; + scheds = &sched; + num_scheds = 1; break; case AMDGPU_HW_IP_VCN_DEC: - for (j = 0; j < adev->vcn.num_vcn_inst; ++j) { - if (adev->vcn.harvest_config & (1 << j)) - continue; - rings[num_rings++] = &adev->vcn.inst[j].ring_dec; - } + scheds = adev->vcn.vcn_dec_sched; + num_scheds = adev->vcn.num_vcn_dec_sched; break; case AMDGPU_HW_IP_VCN_ENC: - for (j = 0; j < adev->vcn.num_vcn_inst; ++j) { - if (adev->vcn.harvest_config & (1 << j)) - continue; - for (k = 0; k < adev->vcn.num_enc_rings; ++k) - rings[num_rings++] = &adev->vcn.inst[j].ring_enc[k]; - } + scheds = adev->vcn.vcn_enc_sched; + num_scheds = adev->vcn.num_vcn_enc_sched; break; case AMDGPU_HW_IP_VCN_JPEG: - for (j = 0; j < adev->jpeg.num_jpeg_inst; ++j) { - if (adev->jpeg.harvest_config & (1 << j)) - continue; - rings[num_rings++] = &adev->jpeg.inst[j].ring_dec; - } + scheds = adev->jpeg.jpeg_sched; + num_scheds = adev->jpeg.num_jpeg_sched; break; } - for (j = 0; j < num_rings; ++j) { - if (!rings[j]->adev) - continue; - - rqs[num_rqs++] = &rings[j]->sched.sched_rq[priority]; - } - for (j = 0; j < amdgpu_ctx_num_entities[i]; ++j) r = drm_sched_entity_init(&ctx->entities[i][j].entity, - rqs, num_rqs, &ctx->guilty); + priority, scheds, + num_scheds, &ctx->guilty); if (r) goto error_cleanup_entities; } @@ -627,3 +612,45 @@ void amdgpu_ctx_mgr_fini(struct amdgpu_ctx_mgr *mgr) idr_destroy(&mgr->ctx_handles); mutex_destroy(&mgr->lock); } + +void amdgpu_ctx_init_sched(struct amdgpu_device *adev) +{ + int i, j; + + for (i = 0; i < adev->gfx.num_gfx_rings; i++) { + adev->gfx.gfx_sched[i] = &adev->gfx.gfx_ring[i].sched; + adev->gfx.num_gfx_sched++; + } + + for (i = 0; i < adev->gfx.num_compute_rings; i++) { + adev->gfx.compute_sched[i] = &adev->gfx.compute_ring[i].sched; + adev->gfx.num_compute_sched++; + } + + for (i = 0; i < adev->sdma.num_instances; i++) { + adev->sdma.sdma_sched[i] = &adev->sdma.instance[i].ring.sched; + adev->sdma.num_sdma_sched++; + } + + for (i = 0; i < adev->vcn.num_vcn_inst; ++i) { + if (adev->vcn.harvest_config & (1 << i)) + continue; + adev->vcn.vcn_dec_sched[adev->vcn.num_vcn_dec_sched++] = + &adev->vcn.inst[i].ring_dec.sched; + } + + for (i = 0; i < adev->vcn.num_vcn_inst; ++i) { + if (adev->vcn.harvest_config & (1 << i)) + continue; + for (j = 0; j < adev->vcn.num_enc_rings; ++j) + adev->vcn.vcn_enc_sched[adev->vcn.num_vcn_enc_sched++] = + &adev->vcn.inst[i].ring_enc[j].sched; + } + + for (i = 0; i < adev->jpeg.num_jpeg_inst; ++i) { + if (adev->jpeg.harvest_config & (1 << i)) + continue; + adev->jpeg.jpeg_sched[adev->jpeg.num_jpeg_sched++] = + &adev->jpeg.inst[i].ring_dec.sched; + } +} diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_ctx.h b/drivers/gpu/drm/amd/amdgpu/amdgpu_ctx.h index da808633732b..4ad90a44dc3c 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_ctx.h +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_ctx.h @@ -87,4 +87,7 @@ void amdgpu_ctx_mgr_entity_fini(struct amdgpu_ctx_mgr *mgr); long amdgpu_ctx_mgr_entity_flush(struct amdgpu_ctx_mgr *mgr, long timeout); void amdgpu_ctx_mgr_fini(struct amdgpu_ctx_mgr *mgr); +void amdgpu_ctx_init_sched(struct amdgpu_device *adev); + + #endif diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_debugfs.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_debugfs.c index 8e6726e0d035..63343bb43049 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_debugfs.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_debugfs.c @@ -129,7 +129,7 @@ static int amdgpu_debugfs_process_reg_op(bool read, struct file *f, sh_bank = 0xFFFFFFFF; if (instance_bank == 0x3FF) instance_bank = 0xFFFFFFFF; - use_bank = 1; + use_bank = true; } else if (*pos & (1ULL << 61)) { me = (*pos & GENMASK_ULL(33, 24)) >> 24; @@ -137,9 +137,9 @@ static int amdgpu_debugfs_process_reg_op(bool read, struct file *f, queue = (*pos & GENMASK_ULL(53, 44)) >> 44; vmid = (*pos & GENMASK_ULL(58, 54)) >> 54; - use_ring = 1; + use_ring = true; } else { - use_bank = use_ring = 0; + use_bank = use_ring = false; } *pos &= (1UL << 22) - 1; diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_device.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_device.c index a97946878024..9b4c18b3546f 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_device.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_device.c @@ -66,6 +66,7 @@ #include "amdgpu_pmu.h" #include <linux/suspend.h> +#include <drm/task_barrier.h> MODULE_FIRMWARE("amdgpu/vega10_gpu_info.bin"); MODULE_FIRMWARE("amdgpu/vega12_gpu_info.bin"); @@ -1031,8 +1032,6 @@ def_value: */ static int amdgpu_device_check_arguments(struct amdgpu_device *adev) { - int ret = 0; - if (amdgpu_sched_jobs < 4) { dev_warn(adev->dev, "sched jobs (%d) must be at least 4\n", amdgpu_sched_jobs); @@ -1072,7 +1071,7 @@ static int amdgpu_device_check_arguments(struct amdgpu_device *adev) adev->firmware.load_type = amdgpu_ucode_get_load_type(adev, amdgpu_fw_load_type); - return ret; + return 0; } /** @@ -1810,7 +1809,8 @@ static int amdgpu_device_fw_loading(struct amdgpu_device *adev) } } - r = amdgpu_pm_load_smu_firmware(adev, &smu_version); + if (!amdgpu_sriov_vf(adev) || adev->asic_type == CHIP_TONGA) + r = amdgpu_pm_load_smu_firmware(adev, &smu_version); return r; } @@ -2439,7 +2439,8 @@ static int amdgpu_device_ip_reinit_late_sriov(struct amdgpu_device *adev) AMD_IP_BLOCK_TYPE_GFX, AMD_IP_BLOCK_TYPE_SDMA, AMD_IP_BLOCK_TYPE_UVD, - AMD_IP_BLOCK_TYPE_VCE + AMD_IP_BLOCK_TYPE_VCE, + AMD_IP_BLOCK_TYPE_VCN }; for (i = 0; i < ARRAY_SIZE(ip_order); i++) { @@ -2454,7 +2455,11 @@ static int amdgpu_device_ip_reinit_late_sriov(struct amdgpu_device *adev) block->status.hw) continue; - r = block->version->funcs->hw_init(adev); + if (block->version->type == AMD_IP_BLOCK_TYPE_SMC) + r = block->version->funcs->resume(adev); + else + r = block->version->funcs->hw_init(adev); + DRM_INFO("RE-INIT-late: %s %s\n", block->version->funcs->name, r?"failed":"succeeded"); if (r) return r; @@ -2663,14 +2668,38 @@ static void amdgpu_device_xgmi_reset_func(struct work_struct *__work) { struct amdgpu_device *adev = container_of(__work, struct amdgpu_device, xgmi_reset_work); + struct amdgpu_hive_info *hive = amdgpu_get_xgmi_hive(adev, 0); - if (amdgpu_asic_reset_method(adev) == AMD_RESET_METHOD_BACO) - adev->asic_reset_res = (adev->in_baco == false) ? - amdgpu_device_baco_enter(adev->ddev) : - amdgpu_device_baco_exit(adev->ddev); - else - adev->asic_reset_res = amdgpu_asic_reset(adev); + /* It's a bug to not have a hive within this function */ + if (WARN_ON(!hive)) + return; + + /* + * Use task barrier to synchronize all xgmi reset works across the + * hive. task_barrier_enter and task_barrier_exit will block + * until all the threads running the xgmi reset works reach + * those points. task_barrier_full will do both blocks. + */ + if (amdgpu_asic_reset_method(adev) == AMD_RESET_METHOD_BACO) { + + task_barrier_enter(&hive->tb); + adev->asic_reset_res = amdgpu_device_baco_enter(adev->ddev); + + if (adev->asic_reset_res) + goto fail; + task_barrier_exit(&hive->tb); + adev->asic_reset_res = amdgpu_device_baco_exit(adev->ddev); + + if (adev->asic_reset_res) + goto fail; + } else { + + task_barrier_full(&hive->tb); + adev->asic_reset_res = amdgpu_asic_reset(adev); + } + +fail: if (adev->asic_reset_res) DRM_WARN("ASIC reset failed with error, %d for drm dev, %s", adev->asic_reset_res, adev->ddev->unique); @@ -2785,7 +2814,7 @@ int amdgpu_device_init(struct amdgpu_device *adev, adev->mman.buffer_funcs = NULL; adev->mman.buffer_funcs_ring = NULL; adev->vm_manager.vm_pte_funcs = NULL; - adev->vm_manager.vm_pte_num_rqs = 0; + adev->vm_manager.vm_pte_num_scheds = 0; adev->gmc.gmc_funcs = NULL; adev->fence_context = dma_fence_context_alloc(AMDGPU_MAX_RINGS); bitmap_zero(adev->gfx.pipe_reserve_bitmap, AMDGPU_MAX_COMPUTE_QUEUES); @@ -3029,6 +3058,14 @@ fence_driver_init: goto failed; } + DRM_DEBUG("SE %d, SH per SE %d, CU per SH %d, active_cu_number %d\n", + adev->gfx.config.max_shader_engines, + adev->gfx.config.max_sh_per_se, + adev->gfx.config.max_cu_per_sh, + adev->gfx.cu_info.number); + + amdgpu_ctx_init_sched(adev); + adev->accel_working = true; amdgpu_vm_check_compute_bug(adev); @@ -3660,8 +3697,6 @@ static int amdgpu_device_reset_sriov(struct amdgpu_device *adev, if (r) return r; - amdgpu_amdkfd_pre_reset(adev); - /* Resume IP prior to SMC */ r = amdgpu_device_ip_reinit_early_sriov(adev); if (r) @@ -3790,18 +3825,13 @@ static int amdgpu_device_pre_asic_reset(struct amdgpu_device *adev, return r; } -static int amdgpu_do_asic_reset(struct amdgpu_device *adev, - struct amdgpu_hive_info *hive, +static int amdgpu_do_asic_reset(struct amdgpu_hive_info *hive, struct list_head *device_list_handle, bool *need_full_reset_arg) { struct amdgpu_device *tmp_adev = NULL; bool need_full_reset = *need_full_reset_arg, vram_lost = false; int r = 0; - int cpu = smp_processor_id(); - bool use_baco = - (amdgpu_asic_reset_method(adev) == AMD_RESET_METHOD_BACO) ? - true : false; /* * ASIC reset has to be done on all HGMI hive nodes ASAP @@ -3809,62 +3839,22 @@ static int amdgpu_do_asic_reset(struct amdgpu_device *adev, */ if (need_full_reset) { list_for_each_entry(tmp_adev, device_list_handle, gmc.xgmi.head) { - /* - * For XGMI run all resets in parallel to speed up the - * process by scheduling the highpri wq on different - * cpus. For XGMI with baco reset, all nodes must enter - * baco within close proximity before anyone exit. - */ + /* For XGMI run all resets in parallel to speed up the process */ if (tmp_adev->gmc.xgmi.num_physical_nodes > 1) { - if (!queue_work_on(cpu, system_highpri_wq, - &tmp_adev->xgmi_reset_work)) + if (!queue_work(system_unbound_wq, &tmp_adev->xgmi_reset_work)) r = -EALREADY; - cpu = cpumask_next(cpu, cpu_online_mask); } else r = amdgpu_asic_reset(tmp_adev); - if (r) - break; - } - /* For XGMI wait for all work to complete before proceed */ - if (!r) { - list_for_each_entry(tmp_adev, device_list_handle, - gmc.xgmi.head) { - if (tmp_adev->gmc.xgmi.num_physical_nodes > 1) { - flush_work(&tmp_adev->xgmi_reset_work); - r = tmp_adev->asic_reset_res; - if (r) - break; - if (use_baco) - tmp_adev->in_baco = true; - } - } - } - - /* - * For XGMI with baco reset, need exit baco phase by scheduling - * xgmi_reset_work one more time. PSP reset and sGPU skips this - * phase. Not assume the situation that PSP reset and baco reset - * coexist within an XGMI hive. - */ - - if (!r && use_baco) { - cpu = smp_processor_id(); - list_for_each_entry(tmp_adev, device_list_handle, - gmc.xgmi.head) { - if (tmp_adev->gmc.xgmi.num_physical_nodes > 1) { - if (!queue_work_on(cpu, - system_highpri_wq, - &tmp_adev->xgmi_reset_work)) - r = -EALREADY; - if (r) - break; - cpu = cpumask_next(cpu, cpu_online_mask); - } + if (r) { + DRM_ERROR("ASIC reset failed with error, %d for drm dev, %s", + r, tmp_adev->ddev->unique); + break; } } - if (!r && use_baco) { + /* For XGMI wait for all resets to complete before proceed */ + if (!r) { list_for_each_entry(tmp_adev, device_list_handle, gmc.xgmi.head) { if (tmp_adev->gmc.xgmi.num_physical_nodes > 1) { @@ -3872,16 +3862,9 @@ static int amdgpu_do_asic_reset(struct amdgpu_device *adev, r = tmp_adev->asic_reset_res; if (r) break; - tmp_adev->in_baco = false; } } } - - if (r) { - DRM_ERROR("ASIC reset failed with error, %d for drm dev, %s", - r, tmp_adev->ddev->unique); - goto end; - } } if (!r && amdgpu_ras_intr_triggered()) @@ -3974,7 +3957,7 @@ static bool amdgpu_device_lock_adev(struct amdgpu_device *adev, bool trylock) mutex_lock(&adev->lock_reset); atomic_inc(&adev->gpu_reset_counter); - adev->in_gpu_reset = 1; + adev->in_gpu_reset = true; switch (amdgpu_asic_reset_method(adev)) { case AMD_RESET_METHOD_MODE1: adev->mp1_state = PP_MP1_STATE_SHUTDOWN; @@ -3994,7 +3977,7 @@ static void amdgpu_device_unlock_adev(struct amdgpu_device *adev) { amdgpu_vf_error_trans_all(adev); adev->mp1_state = PP_MP1_STATE_NONE; - adev->in_gpu_reset = 0; + adev->in_gpu_reset = false; mutex_unlock(&adev->lock_reset); } @@ -4175,8 +4158,7 @@ retry: /* Rest of adevs pre asic reset from XGMI hive. */ if (r) adev->asic_reset_res = r; } else { - r = amdgpu_do_asic_reset(adev, hive, device_list_handle, - &need_full_reset); + r = amdgpu_do_asic_reset(hive, device_list_handle, &need_full_reset); if (r && r == -EAGAIN) goto retry; } diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_dpm.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_dpm.c index 9cc270efee7c..cd76fbf4385d 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_dpm.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_dpm.c @@ -951,16 +951,31 @@ int amdgpu_dpm_set_powergating_by_smu(struct amdgpu_device *adev, uint32_t block case AMD_IP_BLOCK_TYPE_VCN: case AMD_IP_BLOCK_TYPE_VCE: case AMD_IP_BLOCK_TYPE_SDMA: + if (swsmu) { + ret = smu_dpm_set_power_gate(&adev->smu, block_type, gate); + } else { + if (adev->powerplay.pp_funcs && + adev->powerplay.pp_funcs->set_powergating_by_smu) { + mutex_lock(&adev->pm.mutex); + ret = ((adev)->powerplay.pp_funcs->set_powergating_by_smu( + (adev)->powerplay.pp_handle, block_type, gate)); + mutex_unlock(&adev->pm.mutex); + } + } + break; + case AMD_IP_BLOCK_TYPE_JPEG: if (swsmu) ret = smu_dpm_set_power_gate(&adev->smu, block_type, gate); - else - ret = ((adev)->powerplay.pp_funcs->set_powergating_by_smu( - (adev)->powerplay.pp_handle, block_type, gate)); break; case AMD_IP_BLOCK_TYPE_GMC: case AMD_IP_BLOCK_TYPE_ACP: - ret = ((adev)->powerplay.pp_funcs->set_powergating_by_smu( + if (adev->powerplay.pp_funcs && + adev->powerplay.pp_funcs->set_powergating_by_smu) { + mutex_lock(&adev->pm.mutex); + ret = ((adev)->powerplay.pp_funcs->set_powergating_by_smu( (adev)->powerplay.pp_handle, block_type, gate)); + mutex_unlock(&adev->pm.mutex); + } break; default: break; diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_drv.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_drv.c index 3f6f14ce1511..a9c4edca70c9 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_drv.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_drv.c @@ -142,7 +142,7 @@ int amdgpu_async_gfx_ring = 1; int amdgpu_mcbp = 0; int amdgpu_discovery = -1; int amdgpu_mes = 0; -int amdgpu_noretry = 1; +int amdgpu_noretry; int amdgpu_force_asic_type = -1; struct amdgpu_mgpu_info mgpu_info = { @@ -588,7 +588,7 @@ MODULE_PARM_DESC(mes, module_param_named(mes, amdgpu_mes, int, 0444); MODULE_PARM_DESC(noretry, - "Disable retry faults (0 = retry enabled, 1 = retry disabled (default))"); + "Disable retry faults (0 = retry enabled (default), 1 = retry disabled)"); module_param_named(noretry, amdgpu_noretry, int, 0644); /** @@ -1203,13 +1203,23 @@ static int amdgpu_pmops_runtime_suspend(struct device *dev) struct pci_dev *pdev = to_pci_dev(dev); struct drm_device *drm_dev = pci_get_drvdata(pdev); struct amdgpu_device *adev = drm_dev->dev_private; - int ret; + int ret, i; if (!adev->runpm) { pm_runtime_forbid(dev); return -EBUSY; } + /* wait for all rings to drain before suspending */ + for (i = 0; i < AMDGPU_MAX_RINGS; i++) { + struct amdgpu_ring *ring = adev->rings[i]; + if (ring && ring->sched.ready) { + ret = amdgpu_fence_wait_empty(ring); + if (ret) + return -EBUSY; + } + } + if (amdgpu_device_supports_boco(drm_dev)) drm_dev->switch_power_state = DRM_SWITCH_POWER_CHANGING; drm_kms_helper_poll_disable(drm_dev); @@ -1381,7 +1391,8 @@ static struct drm_driver kms_driver = { .driver_features = DRIVER_USE_AGP | DRIVER_ATOMIC | DRIVER_GEM | - DRIVER_RENDER | DRIVER_MODESET | DRIVER_SYNCOBJ, + DRIVER_RENDER | DRIVER_MODESET | DRIVER_SYNCOBJ | + DRIVER_SYNCOBJ_TIMELINE, .load = amdgpu_driver_load_kms, .open = amdgpu_driver_open_kms, .postclose = amdgpu_driver_postclose_kms, diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_fence.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_fence.c index 377fe20bce23..e9efee04ca23 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_fence.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_fence.c @@ -34,6 +34,7 @@ #include <linux/kref.h> #include <linux/slab.h> #include <linux/firmware.h> +#include <linux/pm_runtime.h> #include <drm/drm_debugfs.h> @@ -154,7 +155,7 @@ int amdgpu_fence_emit(struct amdgpu_ring *ring, struct dma_fence **f, seq); amdgpu_ring_emit_fence(ring, ring->fence_drv.gpu_addr, seq, flags | AMDGPU_FENCE_FLAG_INT); - + pm_runtime_get_noresume(adev->ddev->dev); ptr = &ring->fence_drv.fences[seq & ring->fence_drv.num_fences_mask]; if (unlikely(rcu_dereference_protected(*ptr, 1))) { struct dma_fence *old; @@ -234,6 +235,7 @@ static void amdgpu_fence_schedule_fallback(struct amdgpu_ring *ring) bool amdgpu_fence_process(struct amdgpu_ring *ring) { struct amdgpu_fence_driver *drv = &ring->fence_drv; + struct amdgpu_device *adev = ring->adev; uint32_t seq, last_seq; int r; @@ -274,6 +276,8 @@ bool amdgpu_fence_process(struct amdgpu_ring *ring) BUG(); dma_fence_put(fence); + pm_runtime_mark_last_busy(adev->ddev->dev); + pm_runtime_put_autosuspend(adev->ddev->dev); } while (last_seq != seq); return true; diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_gfx.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_gfx.c index e00b46180d2e..db7b2b3f9966 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_gfx.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_gfx.c @@ -641,7 +641,7 @@ int amdgpu_gfx_process_ras_data_cb(struct amdgpu_device *adev, kgd2kfd_set_sram_ecc_flag(adev->kfd.dev); if (adev->gfx.funcs->query_ras_error_count) adev->gfx.funcs->query_ras_error_count(adev, err_data); - amdgpu_ras_reset_gpu(adev, 0); + amdgpu_ras_reset_gpu(adev); } return AMDGPU_RAS_SUCCESS; } diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_gfx.h b/drivers/gpu/drm/amd/amdgpu/amdgpu_gfx.h index 0ae0a2715b0d..8e88e0411662 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_gfx.h +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_gfx.h @@ -269,8 +269,12 @@ struct amdgpu_gfx { bool me_fw_write_wait; bool cp_fw_write_wait; struct amdgpu_ring gfx_ring[AMDGPU_MAX_GFX_RINGS]; + struct drm_gpu_scheduler *gfx_sched[AMDGPU_MAX_GFX_RINGS]; + uint32_t num_gfx_sched; unsigned num_gfx_rings; struct amdgpu_ring compute_ring[AMDGPU_MAX_COMPUTE_RINGS]; + struct drm_gpu_scheduler *compute_sched[AMDGPU_MAX_COMPUTE_RINGS]; + uint32_t num_compute_sched; unsigned num_compute_rings; struct amdgpu_irq_src eop_irq; struct amdgpu_irq_src priv_reg_irq; diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_gmc.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_gmc.c index a12f33c0f5df..5884ab590486 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_gmc.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_gmc.c @@ -223,7 +223,7 @@ void amdgpu_gmc_agp_location(struct amdgpu_device *adev, struct amdgpu_gmc *mc) u64 size_af, size_bf; if (amdgpu_sriov_vf(adev)) { - mc->agp_start = 0xffffffff; + mc->agp_start = 0xffffffffffff; mc->agp_end = 0x0; mc->agp_size = 0; @@ -333,3 +333,43 @@ void amdgpu_gmc_ras_fini(struct amdgpu_device *adev) amdgpu_mmhub_ras_fini(adev); amdgpu_xgmi_ras_fini(adev); } + + /* + * The latest engine allocation on gfx9/10 is: + * Engine 2, 3: firmware + * Engine 0, 1, 4~16: amdgpu ring, + * subject to change when ring number changes + * Engine 17: Gart flushes + */ +#define GFXHUB_FREE_VM_INV_ENGS_BITMAP 0x1FFF3 +#define MMHUB_FREE_VM_INV_ENGS_BITMAP 0x1FFF3 + +int amdgpu_gmc_allocate_vm_inv_eng(struct amdgpu_device *adev) +{ + struct amdgpu_ring *ring; + unsigned vm_inv_engs[AMDGPU_MAX_VMHUBS] = + {GFXHUB_FREE_VM_INV_ENGS_BITMAP, MMHUB_FREE_VM_INV_ENGS_BITMAP, + GFXHUB_FREE_VM_INV_ENGS_BITMAP}; + unsigned i; + unsigned vmhub, inv_eng; + + for (i = 0; i < adev->num_rings; ++i) { + ring = adev->rings[i]; + vmhub = ring->funcs->vmhub; + + inv_eng = ffs(vm_inv_engs[vmhub]); + if (!inv_eng) { + dev_err(adev->dev, "no VM inv eng for ring %s\n", + ring->name); + return -EINVAL; + } + + ring->vm_inv_eng = inv_eng - 1; + vm_inv_engs[vmhub] &= ~(1 << ring->vm_inv_eng); + + dev_info(adev->dev, "ring %s uses VM inv eng %u on hub %u\n", + ring->name, ring->vm_inv_eng, ring->funcs->vmhub); + } + + return 0; +} diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_gmc.h b/drivers/gpu/drm/amd/amdgpu/amdgpu_gmc.h index b499a3de8bb6..c91dd602d5f1 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_gmc.h +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_gmc.h @@ -267,5 +267,6 @@ bool amdgpu_gmc_filter_faults(struct amdgpu_device *adev, uint64_t addr, uint16_t pasid, uint64_t timestamp); int amdgpu_gmc_ras_late_init(struct amdgpu_device *adev); void amdgpu_gmc_ras_fini(struct amdgpu_device *adev); +int amdgpu_gmc_allocate_vm_inv_eng(struct amdgpu_device *adev); #endif diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_jpeg.h b/drivers/gpu/drm/amd/amdgpu/amdgpu_jpeg.h index 5131a0a1bc8a..bd9ef9cc86de 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_jpeg.h +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_jpeg.h @@ -43,6 +43,8 @@ struct amdgpu_jpeg { uint8_t num_jpeg_inst; struct amdgpu_jpeg_inst inst[AMDGPU_MAX_JPEG_INSTANCES]; struct amdgpu_jpeg_reg internal; + struct drm_gpu_scheduler *jpeg_sched[AMDGPU_MAX_JPEG_INSTANCES]; + uint32_t num_jpeg_sched; unsigned harvest_config; struct delayed_work idle_work; enum amd_powergating_state cur_state; diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_pm.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_pm.c index b32adda70bbc..285d460624c8 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_pm.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_pm.c @@ -2762,17 +2762,12 @@ static void amdgpu_dpm_change_power_state_locked(struct amdgpu_device *adev) void amdgpu_dpm_enable_uvd(struct amdgpu_device *adev, bool enable) { int ret = 0; - if (is_support_sw_smu(adev)) { - ret = smu_dpm_set_power_gate(&adev->smu, AMD_IP_BLOCK_TYPE_UVD, enable); - if (ret) - DRM_ERROR("[SW SMU]: dpm enable uvd failed, state = %s, ret = %d. \n", - enable ? "true" : "false", ret); - } else if (adev->powerplay.pp_funcs->set_powergating_by_smu) { - /* enable/disable UVD */ - mutex_lock(&adev->pm.mutex); - amdgpu_dpm_set_powergating_by_smu(adev, AMD_IP_BLOCK_TYPE_UVD, !enable); - mutex_unlock(&adev->pm.mutex); - } + + ret = amdgpu_dpm_set_powergating_by_smu(adev, AMD_IP_BLOCK_TYPE_UVD, !enable); + if (ret) + DRM_ERROR("Dpm %s uvd failed, ret = %d. \n", + enable ? "enable" : "disable", ret); + /* enable/disable Low Memory PState for UVD (4k videos) */ if (adev->asic_type == CHIP_STONEY && adev->uvd.decode_image_width >= WIDTH_4K) { @@ -2789,17 +2784,11 @@ void amdgpu_dpm_enable_uvd(struct amdgpu_device *adev, bool enable) void amdgpu_dpm_enable_vce(struct amdgpu_device *adev, bool enable) { int ret = 0; - if (is_support_sw_smu(adev)) { - ret = smu_dpm_set_power_gate(&adev->smu, AMD_IP_BLOCK_TYPE_VCE, enable); - if (ret) - DRM_ERROR("[SW SMU]: dpm enable vce failed, state = %s, ret = %d. \n", - enable ? "true" : "false", ret); - } else if (adev->powerplay.pp_funcs->set_powergating_by_smu) { - /* enable/disable VCE */ - mutex_lock(&adev->pm.mutex); - amdgpu_dpm_set_powergating_by_smu(adev, AMD_IP_BLOCK_TYPE_VCE, !enable); - mutex_unlock(&adev->pm.mutex); - } + + ret = amdgpu_dpm_set_powergating_by_smu(adev, AMD_IP_BLOCK_TYPE_VCE, !enable); + if (ret) + DRM_ERROR("Dpm %s vce failed, ret = %d. \n", + enable ? "enable" : "disable", ret); } void amdgpu_pm_print_power_states(struct amdgpu_device *adev) @@ -2818,12 +2807,10 @@ void amdgpu_dpm_enable_jpeg(struct amdgpu_device *adev, bool enable) { int ret = 0; - if (is_support_sw_smu(adev)) { - ret = smu_dpm_set_power_gate(&adev->smu, AMD_IP_BLOCK_TYPE_JPEG, enable); - if (ret) - DRM_ERROR("[SW SMU]: dpm enable jpeg failed, state = %s, ret = %d. \n", - enable ? "true" : "false", ret); - } + ret = amdgpu_dpm_set_powergating_by_smu(adev, AMD_IP_BLOCK_TYPE_JPEG, !enable); + if (ret) + DRM_ERROR("Dpm %s jpeg failed, ret = %d. \n", + enable ? "enable" : "disable", ret); } int amdgpu_pm_load_smu_firmware(struct amdgpu_device *adev, uint32_t *smu_version) diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_pmu.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_pmu.c index 0e6dba9f60f0..cf21ad0cad9a 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_pmu.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_pmu.c @@ -107,7 +107,7 @@ static void amdgpu_perf_read(struct perf_event *event) default: count = 0; break; - }; + } } while (local64_cmpxchg(&hwc->prev_count, prev, count) != prev); local64_add(count - prev, &event->count); @@ -130,7 +130,7 @@ static void amdgpu_perf_stop(struct perf_event *event, int flags) break; default: break; - }; + } WARN_ON_ONCE(hwc->state & PERF_HES_STOPPED); hwc->state |= PERF_HES_STOPPED; @@ -160,7 +160,7 @@ static int amdgpu_perf_add(struct perf_event *event, int flags) break; default: return 0; - }; + } if (retval) return retval; @@ -188,7 +188,7 @@ static void amdgpu_perf_del(struct perf_event *event, int flags) break; default: break; - }; + } perf_event_update_userpage(event); } diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_psp.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_psp.c index c14f2ccd0677..281d89640344 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_psp.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_psp.c @@ -191,9 +191,9 @@ psp_cmd_submit_buf(struct psp_context *psp, if (ucode) DRM_WARN("failed to load ucode id (%d) ", ucode->ucode_id); - DRM_DEBUG_DRIVER("psp command (0x%X) failed and response status is (0x%X)\n", + DRM_WARN("psp command (0x%X) failed and response status is (0x%X)\n", psp->cmd_buf_mem->cmd_id, - psp->cmd_buf_mem->resp.status & GFX_CMD_STATUS_MASK); + psp->cmd_buf_mem->resp.status); if (!timeout) { mutex_unlock(&psp->mutex); return -EINVAL; @@ -365,11 +365,11 @@ static int psp_asd_load(struct psp_context *psp) return ret; } -static void psp_prep_asd_unload_cmd_buf(struct psp_gfx_cmd_resp *cmd, - uint32_t asd_session_id) +static void psp_prep_ta_unload_cmd_buf(struct psp_gfx_cmd_resp *cmd, + uint32_t session_id) { cmd->cmd_id = GFX_CMD_ID_UNLOAD_TA; - cmd->cmd.cmd_unload_ta.session_id = asd_session_id; + cmd->cmd.cmd_unload_ta.session_id = session_id; } static int psp_asd_unload(struct psp_context *psp) @@ -387,7 +387,7 @@ static int psp_asd_unload(struct psp_context *psp) if (!cmd) return -ENOMEM; - psp_prep_asd_unload_cmd_buf(cmd, psp->asd_context.session_id); + psp_prep_ta_unload_cmd_buf(cmd, psp->asd_context.session_id); ret = psp_cmd_submit_buf(psp, NULL, cmd, psp->fence_buf_mc_addr); @@ -427,18 +427,20 @@ int psp_reg_program(struct psp_context *psp, enum psp_reg_prog_id reg, return ret; } -static void psp_prep_xgmi_ta_load_cmd_buf(struct psp_gfx_cmd_resp *cmd, - uint64_t xgmi_ta_mc, uint64_t xgmi_mc_shared, - uint32_t xgmi_ta_size, uint32_t shared_size) +static void psp_prep_ta_load_cmd_buf(struct psp_gfx_cmd_resp *cmd, + uint64_t ta_bin_mc, + uint32_t ta_bin_size, + uint64_t ta_shared_mc, + uint32_t ta_shared_size) { - cmd->cmd_id = GFX_CMD_ID_LOAD_TA; - cmd->cmd.cmd_load_ta.app_phy_addr_lo = lower_32_bits(xgmi_ta_mc); - cmd->cmd.cmd_load_ta.app_phy_addr_hi = upper_32_bits(xgmi_ta_mc); - cmd->cmd.cmd_load_ta.app_len = xgmi_ta_size; + cmd->cmd_id = GFX_CMD_ID_LOAD_TA; + cmd->cmd.cmd_load_ta.app_phy_addr_lo = lower_32_bits(ta_bin_mc); + cmd->cmd.cmd_load_ta.app_phy_addr_hi = upper_32_bits(ta_bin_mc); + cmd->cmd.cmd_load_ta.app_len = ta_bin_size; - cmd->cmd.cmd_load_ta.cmd_buf_phy_addr_lo = lower_32_bits(xgmi_mc_shared); - cmd->cmd.cmd_load_ta.cmd_buf_phy_addr_hi = upper_32_bits(xgmi_mc_shared); - cmd->cmd.cmd_load_ta.cmd_buf_len = shared_size; + cmd->cmd.cmd_load_ta.cmd_buf_phy_addr_lo = lower_32_bits(ta_shared_mc); + cmd->cmd.cmd_load_ta.cmd_buf_phy_addr_hi = upper_32_bits(ta_shared_mc); + cmd->cmd.cmd_load_ta.cmd_buf_len = ta_shared_size; } static int psp_xgmi_init_shared_buf(struct psp_context *psp) @@ -458,6 +460,36 @@ static int psp_xgmi_init_shared_buf(struct psp_context *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) +{ + cmd->cmd_id = GFX_CMD_ID_INVOKE_CMD; + cmd->cmd.cmd_invoke_cmd.session_id = session_id; + cmd->cmd.cmd_invoke_cmd.ta_cmd_id = ta_cmd_id; +} + +int psp_ta_invoke(struct psp_context *psp, + uint32_t ta_cmd_id, + uint32_t session_id) +{ + int ret; + struct psp_gfx_cmd_resp *cmd; + + cmd = kzalloc(sizeof(struct psp_gfx_cmd_resp), GFP_KERNEL); + if (!cmd) + return -ENOMEM; + + psp_prep_ta_invoke_cmd_buf(cmd, ta_cmd_id, session_id); + + ret = psp_cmd_submit_buf(psp, NULL, cmd, + psp->fence_buf_mc_addr); + + kfree(cmd); + + return ret; +} + static int psp_xgmi_load(struct psp_context *psp) { int ret; @@ -466,8 +498,6 @@ static int psp_xgmi_load(struct psp_context *psp) /* * TODO: bypass the loading in sriov for now */ - if (amdgpu_sriov_vf(psp->adev)) - return 0; cmd = kzalloc(sizeof(struct psp_gfx_cmd_resp), GFP_KERNEL); if (!cmd) @@ -476,9 +506,11 @@ static int psp_xgmi_load(struct psp_context *psp) memset(psp->fw_pri_buf, 0, PSP_1_MEG); memcpy(psp->fw_pri_buf, psp->ta_xgmi_start_addr, psp->ta_xgmi_ucode_size); - psp_prep_xgmi_ta_load_cmd_buf(cmd, psp->fw_pri_mc_addr, - psp->xgmi_context.xgmi_shared_mc_addr, - psp->ta_xgmi_ucode_size, PSP_XGMI_SHARED_MEM_SIZE); + psp_prep_ta_load_cmd_buf(cmd, + psp->fw_pri_mc_addr, + psp->ta_xgmi_ucode_size, + psp->xgmi_context.xgmi_shared_mc_addr, + PSP_XGMI_SHARED_MEM_SIZE); ret = psp_cmd_submit_buf(psp, NULL, cmd, psp->fence_buf_mc_addr); @@ -493,13 +525,6 @@ static int psp_xgmi_load(struct psp_context *psp) return ret; } -static void psp_prep_xgmi_ta_unload_cmd_buf(struct psp_gfx_cmd_resp *cmd, - uint32_t xgmi_session_id) -{ - cmd->cmd_id = GFX_CMD_ID_UNLOAD_TA; - cmd->cmd.cmd_unload_ta.session_id = xgmi_session_id; -} - static int psp_xgmi_unload(struct psp_context *psp) { int ret; @@ -508,14 +533,12 @@ static int psp_xgmi_unload(struct psp_context *psp) /* * TODO: bypass the unloading in sriov for now */ - if (amdgpu_sriov_vf(psp->adev)) - return 0; cmd = kzalloc(sizeof(struct psp_gfx_cmd_resp), GFP_KERNEL); if (!cmd) return -ENOMEM; - psp_prep_xgmi_ta_unload_cmd_buf(cmd, psp->xgmi_context.session_id); + psp_prep_ta_unload_cmd_buf(cmd, psp->xgmi_context.session_id); ret = psp_cmd_submit_buf(psp, NULL, cmd, psp->fence_buf_mc_addr); @@ -525,40 +548,9 @@ static int psp_xgmi_unload(struct psp_context *psp) return ret; } -static void psp_prep_xgmi_ta_invoke_cmd_buf(struct psp_gfx_cmd_resp *cmd, - uint32_t ta_cmd_id, - uint32_t xgmi_session_id) -{ - cmd->cmd_id = GFX_CMD_ID_INVOKE_CMD; - cmd->cmd.cmd_invoke_cmd.session_id = xgmi_session_id; - cmd->cmd.cmd_invoke_cmd.ta_cmd_id = ta_cmd_id; - /* Note: cmd_invoke_cmd.buf is not used for now */ -} - int psp_xgmi_invoke(struct psp_context *psp, uint32_t ta_cmd_id) { - int ret; - struct psp_gfx_cmd_resp *cmd; - - /* - * TODO: bypass the loading in sriov for now - */ - if (amdgpu_sriov_vf(psp->adev)) - return 0; - - cmd = kzalloc(sizeof(struct psp_gfx_cmd_resp), GFP_KERNEL); - if (!cmd) - return -ENOMEM; - - psp_prep_xgmi_ta_invoke_cmd_buf(cmd, ta_cmd_id, - psp->xgmi_context.session_id); - - ret = psp_cmd_submit_buf(psp, NULL, cmd, - psp->fence_buf_mc_addr); - - kfree(cmd); - - return ret; + return psp_ta_invoke(psp, ta_cmd_id, psp->xgmi_context.session_id); } static int psp_xgmi_terminate(struct psp_context *psp) @@ -614,20 +606,6 @@ static int psp_xgmi_initialize(struct psp_context *psp) } // ras begin -static void psp_prep_ras_ta_load_cmd_buf(struct psp_gfx_cmd_resp *cmd, - uint64_t ras_ta_mc, uint64_t ras_mc_shared, - uint32_t ras_ta_size, uint32_t shared_size) -{ - cmd->cmd_id = GFX_CMD_ID_LOAD_TA; - cmd->cmd.cmd_load_ta.app_phy_addr_lo = lower_32_bits(ras_ta_mc); - cmd->cmd.cmd_load_ta.app_phy_addr_hi = upper_32_bits(ras_ta_mc); - cmd->cmd.cmd_load_ta.app_len = ras_ta_size; - - cmd->cmd.cmd_load_ta.cmd_buf_phy_addr_lo = lower_32_bits(ras_mc_shared); - cmd->cmd.cmd_load_ta.cmd_buf_phy_addr_hi = upper_32_bits(ras_mc_shared); - cmd->cmd.cmd_load_ta.cmd_buf_len = shared_size; -} - static int psp_ras_init_shared_buf(struct psp_context *psp) { int ret; @@ -663,15 +641,17 @@ static int psp_ras_load(struct psp_context *psp) memset(psp->fw_pri_buf, 0, PSP_1_MEG); memcpy(psp->fw_pri_buf, psp->ta_ras_start_addr, psp->ta_ras_ucode_size); - psp_prep_ras_ta_load_cmd_buf(cmd, psp->fw_pri_mc_addr, - psp->ras.ras_shared_mc_addr, - psp->ta_ras_ucode_size, PSP_RAS_SHARED_MEM_SIZE); + psp_prep_ta_load_cmd_buf(cmd, + psp->fw_pri_mc_addr, + psp->ta_ras_ucode_size, + psp->ras.ras_shared_mc_addr, + PSP_RAS_SHARED_MEM_SIZE); ret = psp_cmd_submit_buf(psp, NULL, cmd, psp->fence_buf_mc_addr); if (!ret) { - psp->ras.ras_initialized = 1; + psp->ras.ras_initialized = true; psp->ras.session_id = cmd->resp.session_id; } @@ -680,13 +660,6 @@ static int psp_ras_load(struct psp_context *psp) return ret; } -static void psp_prep_ras_ta_unload_cmd_buf(struct psp_gfx_cmd_resp *cmd, - uint32_t ras_session_id) -{ - cmd->cmd_id = GFX_CMD_ID_UNLOAD_TA; - cmd->cmd.cmd_unload_ta.session_id = ras_session_id; -} - static int psp_ras_unload(struct psp_context *psp) { int ret; @@ -702,7 +675,7 @@ static int psp_ras_unload(struct psp_context *psp) if (!cmd) return -ENOMEM; - psp_prep_ras_ta_unload_cmd_buf(cmd, psp->ras.session_id); + psp_prep_ta_unload_cmd_buf(cmd, psp->ras.session_id); ret = psp_cmd_submit_buf(psp, NULL, cmd, psp->fence_buf_mc_addr); @@ -712,40 +685,15 @@ static int psp_ras_unload(struct psp_context *psp) return ret; } -static void psp_prep_ras_ta_invoke_cmd_buf(struct psp_gfx_cmd_resp *cmd, - uint32_t ta_cmd_id, - uint32_t ras_session_id) -{ - cmd->cmd_id = GFX_CMD_ID_INVOKE_CMD; - cmd->cmd.cmd_invoke_cmd.session_id = ras_session_id; - cmd->cmd.cmd_invoke_cmd.ta_cmd_id = ta_cmd_id; - /* Note: cmd_invoke_cmd.buf is not used for now */ -} - int psp_ras_invoke(struct psp_context *psp, uint32_t ta_cmd_id) { - int ret; - struct psp_gfx_cmd_resp *cmd; - /* * TODO: bypass the loading in sriov for now */ if (amdgpu_sriov_vf(psp->adev)) return 0; - cmd = kzalloc(sizeof(struct psp_gfx_cmd_resp), GFP_KERNEL); - if (!cmd) - return -ENOMEM; - - psp_prep_ras_ta_invoke_cmd_buf(cmd, ta_cmd_id, - psp->ras.session_id); - - ret = psp_cmd_submit_buf(psp, NULL, cmd, - psp->fence_buf_mc_addr); - - kfree(cmd); - - return ret; + return psp_ta_invoke(psp, ta_cmd_id, psp->ras.session_id); } int psp_ras_enable_features(struct psp_context *psp, @@ -791,7 +739,7 @@ static int psp_ras_terminate(struct psp_context *psp) if (ret) return ret; - psp->ras.ras_initialized = 0; + psp->ras.ras_initialized = false; /* free ras shared memory */ amdgpu_bo_free_kernel(&psp->ras.ras_shared_bo, @@ -832,24 +780,6 @@ static int psp_ras_initialize(struct psp_context *psp) // ras end // HDCP start -static void psp_prep_hdcp_ta_load_cmd_buf(struct psp_gfx_cmd_resp *cmd, - uint64_t hdcp_ta_mc, - uint64_t hdcp_mc_shared, - uint32_t hdcp_ta_size, - uint32_t shared_size) -{ - cmd->cmd_id = GFX_CMD_ID_LOAD_TA; - cmd->cmd.cmd_load_ta.app_phy_addr_lo = lower_32_bits(hdcp_ta_mc); - cmd->cmd.cmd_load_ta.app_phy_addr_hi = upper_32_bits(hdcp_ta_mc); - cmd->cmd.cmd_load_ta.app_len = hdcp_ta_size; - - cmd->cmd.cmd_load_ta.cmd_buf_phy_addr_lo = - lower_32_bits(hdcp_mc_shared); - cmd->cmd.cmd_load_ta.cmd_buf_phy_addr_hi = - upper_32_bits(hdcp_mc_shared); - cmd->cmd.cmd_load_ta.cmd_buf_len = shared_size; -} - static int psp_hdcp_init_shared_buf(struct psp_context *psp) { int ret; @@ -886,15 +816,16 @@ static int psp_hdcp_load(struct psp_context *psp) memcpy(psp->fw_pri_buf, psp->ta_hdcp_start_addr, psp->ta_hdcp_ucode_size); - psp_prep_hdcp_ta_load_cmd_buf(cmd, psp->fw_pri_mc_addr, - psp->hdcp_context.hdcp_shared_mc_addr, - psp->ta_hdcp_ucode_size, - PSP_HDCP_SHARED_MEM_SIZE); + psp_prep_ta_load_cmd_buf(cmd, + psp->fw_pri_mc_addr, + psp->ta_hdcp_ucode_size, + psp->hdcp_context.hdcp_shared_mc_addr, + PSP_HDCP_SHARED_MEM_SIZE); ret = psp_cmd_submit_buf(psp, NULL, cmd, psp->fence_buf_mc_addr); if (!ret) { - psp->hdcp_context.hdcp_initialized = 1; + psp->hdcp_context.hdcp_initialized = true; psp->hdcp_context.session_id = cmd->resp.session_id; } @@ -930,12 +861,6 @@ static int psp_hdcp_initialize(struct psp_context *psp) return 0; } -static void psp_prep_hdcp_ta_unload_cmd_buf(struct psp_gfx_cmd_resp *cmd, - uint32_t hdcp_session_id) -{ - cmd->cmd_id = GFX_CMD_ID_UNLOAD_TA; - cmd->cmd.cmd_unload_ta.session_id = hdcp_session_id; -} static int psp_hdcp_unload(struct psp_context *psp) { @@ -952,7 +877,7 @@ static int psp_hdcp_unload(struct psp_context *psp) if (!cmd) return -ENOMEM; - psp_prep_hdcp_ta_unload_cmd_buf(cmd, psp->hdcp_context.session_id); + psp_prep_ta_unload_cmd_buf(cmd, psp->hdcp_context.session_id); ret = psp_cmd_submit_buf(psp, NULL, cmd, psp->fence_buf_mc_addr); @@ -961,39 +886,15 @@ static int psp_hdcp_unload(struct psp_context *psp) return ret; } -static void psp_prep_hdcp_ta_invoke_cmd_buf(struct psp_gfx_cmd_resp *cmd, - uint32_t ta_cmd_id, - uint32_t hdcp_session_id) -{ - cmd->cmd_id = GFX_CMD_ID_INVOKE_CMD; - cmd->cmd.cmd_invoke_cmd.session_id = hdcp_session_id; - cmd->cmd.cmd_invoke_cmd.ta_cmd_id = ta_cmd_id; - /* Note: cmd_invoke_cmd.buf is not used for now */ -} - int psp_hdcp_invoke(struct psp_context *psp, uint32_t ta_cmd_id) { - int ret; - struct psp_gfx_cmd_resp *cmd; - /* * TODO: bypass the loading in sriov for now */ if (amdgpu_sriov_vf(psp->adev)) return 0; - cmd = kzalloc(sizeof(struct psp_gfx_cmd_resp), GFP_KERNEL); - if (!cmd) - return -ENOMEM; - - psp_prep_hdcp_ta_invoke_cmd_buf(cmd, ta_cmd_id, - psp->hdcp_context.session_id); - - ret = psp_cmd_submit_buf(psp, NULL, cmd, psp->fence_buf_mc_addr); - - kfree(cmd); - - return ret; + return psp_ta_invoke(psp, ta_cmd_id, psp->hdcp_context.session_id); } static int psp_hdcp_terminate(struct psp_context *psp) @@ -1013,7 +914,7 @@ static int psp_hdcp_terminate(struct psp_context *psp) if (ret) return ret; - psp->hdcp_context.hdcp_initialized = 0; + psp->hdcp_context.hdcp_initialized = false; /* free hdcp shared memory */ amdgpu_bo_free_kernel(&psp->hdcp_context.hdcp_shared_bo, @@ -1025,22 +926,6 @@ static int psp_hdcp_terminate(struct psp_context *psp) // HDCP end // DTM start -static void psp_prep_dtm_ta_load_cmd_buf(struct psp_gfx_cmd_resp *cmd, - uint64_t dtm_ta_mc, - uint64_t dtm_mc_shared, - uint32_t dtm_ta_size, - uint32_t shared_size) -{ - cmd->cmd_id = GFX_CMD_ID_LOAD_TA; - cmd->cmd.cmd_load_ta.app_phy_addr_lo = lower_32_bits(dtm_ta_mc); - cmd->cmd.cmd_load_ta.app_phy_addr_hi = upper_32_bits(dtm_ta_mc); - cmd->cmd.cmd_load_ta.app_len = dtm_ta_size; - - cmd->cmd.cmd_load_ta.cmd_buf_phy_addr_lo = lower_32_bits(dtm_mc_shared); - cmd->cmd.cmd_load_ta.cmd_buf_phy_addr_hi = upper_32_bits(dtm_mc_shared); - cmd->cmd.cmd_load_ta.cmd_buf_len = shared_size; -} - static int psp_dtm_init_shared_buf(struct psp_context *psp) { int ret; @@ -1076,15 +961,16 @@ static int psp_dtm_load(struct psp_context *psp) memset(psp->fw_pri_buf, 0, PSP_1_MEG); memcpy(psp->fw_pri_buf, psp->ta_dtm_start_addr, psp->ta_dtm_ucode_size); - psp_prep_dtm_ta_load_cmd_buf(cmd, psp->fw_pri_mc_addr, - psp->dtm_context.dtm_shared_mc_addr, - psp->ta_dtm_ucode_size, - PSP_DTM_SHARED_MEM_SIZE); + psp_prep_ta_load_cmd_buf(cmd, + psp->fw_pri_mc_addr, + psp->ta_dtm_ucode_size, + psp->dtm_context.dtm_shared_mc_addr, + PSP_DTM_SHARED_MEM_SIZE); ret = psp_cmd_submit_buf(psp, NULL, cmd, psp->fence_buf_mc_addr); if (!ret) { - psp->dtm_context.dtm_initialized = 1; + psp->dtm_context.dtm_initialized = true; psp->dtm_context.session_id = cmd->resp.session_id; } @@ -1122,39 +1008,15 @@ static int psp_dtm_initialize(struct psp_context *psp) return 0; } -static void psp_prep_dtm_ta_invoke_cmd_buf(struct psp_gfx_cmd_resp *cmd, - uint32_t ta_cmd_id, - uint32_t dtm_session_id) -{ - cmd->cmd_id = GFX_CMD_ID_INVOKE_CMD; - cmd->cmd.cmd_invoke_cmd.session_id = dtm_session_id; - cmd->cmd.cmd_invoke_cmd.ta_cmd_id = ta_cmd_id; - /* Note: cmd_invoke_cmd.buf is not used for now */ -} - int psp_dtm_invoke(struct psp_context *psp, uint32_t ta_cmd_id) { - int ret; - struct psp_gfx_cmd_resp *cmd; - /* * TODO: bypass the loading in sriov for now */ if (amdgpu_sriov_vf(psp->adev)) return 0; - cmd = kzalloc(sizeof(struct psp_gfx_cmd_resp), GFP_KERNEL); - if (!cmd) - return -ENOMEM; - - psp_prep_dtm_ta_invoke_cmd_buf(cmd, ta_cmd_id, - psp->dtm_context.session_id); - - ret = psp_cmd_submit_buf(psp, NULL, cmd, psp->fence_buf_mc_addr); - - kfree(cmd); - - return ret; + return psp_ta_invoke(psp, ta_cmd_id, psp->dtm_context.session_id); } static int psp_dtm_terminate(struct psp_context *psp) @@ -1174,7 +1036,7 @@ static int psp_dtm_terminate(struct psp_context *psp) if (ret) return ret; - psp->dtm_context.dtm_initialized = 0; + psp->dtm_context.dtm_initialized = false; /* free hdcp shared memory */ amdgpu_bo_free_kernel(&psp->dtm_context.dtm_shared_bo, @@ -1310,6 +1172,9 @@ static int psp_get_fw_type(struct amdgpu_firmware_info *ucode, case AMDGPU_UCODE_ID_VCN: *type = GFX_FW_TYPE_VCN; break; + case AMDGPU_UCODE_ID_VCN1: + *type = GFX_FW_TYPE_VCN1; + break; case AMDGPU_UCODE_ID_DMCU_ERAM: *type = GFX_FW_TYPE_DMCU_ERAM; break; @@ -1454,7 +1319,8 @@ out: || ucode->ucode_id == AMDGPU_UCODE_ID_RLC_G || ucode->ucode_id == AMDGPU_UCODE_ID_RLC_RESTORE_LIST_CNTL || ucode->ucode_id == AMDGPU_UCODE_ID_RLC_RESTORE_LIST_GPM_MEM - || ucode->ucode_id == AMDGPU_UCODE_ID_RLC_RESTORE_LIST_SRM_MEM)) + || ucode->ucode_id == AMDGPU_UCODE_ID_RLC_RESTORE_LIST_SRM_MEM + || ucode->ucode_id == AMDGPU_UCODE_ID_SMC)) /*skip ucode loading in SRIOV VF */ continue; @@ -1472,7 +1338,7 @@ out: /* Start rlc autoload after psp recieved all the gfx firmware */ if (psp->autoload_supported && ucode->ucode_id == (amdgpu_sriov_vf(adev) ? - AMDGPU_UCODE_ID_CP_MEC2 : AMDGPU_UCODE_ID_RLC_RESTORE_LIST_SRM_MEM)) { + AMDGPU_UCODE_ID_CP_MEC2 : AMDGPU_UCODE_ID_RLC_G)) { ret = psp_rlc_autoload(psp); if (ret) { DRM_ERROR("Failed to start rlc autoload\n"); @@ -1503,16 +1369,13 @@ static int psp_load_fw(struct amdgpu_device *adev) if (!psp->cmd) return -ENOMEM; - /* this fw pri bo is not used under SRIOV */ - if (!amdgpu_sriov_vf(psp->adev)) { - ret = amdgpu_bo_create_kernel(adev, PSP_1_MEG, PSP_1_MEG, - AMDGPU_GEM_DOMAIN_GTT, - &psp->fw_pri_bo, - &psp->fw_pri_mc_addr, - &psp->fw_pri_buf); - if (ret) - goto failed; - } + ret = amdgpu_bo_create_kernel(adev, PSP_1_MEG, PSP_1_MEG, + AMDGPU_GEM_DOMAIN_GTT, + &psp->fw_pri_bo, + &psp->fw_pri_mc_addr, + &psp->fw_pri_buf); + if (ret) + goto failed; ret = amdgpu_bo_create_kernel(adev, PSP_FENCE_BUFFER_SIZE, PAGE_SIZE, AMDGPU_GEM_DOMAIN_VRAM, diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_psp.h b/drivers/gpu/drm/amd/amdgpu/amdgpu_psp.h index 5f8fd3e3535b..3265487b859f 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_psp.h +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_psp.h @@ -202,7 +202,6 @@ struct psp_memory_training_context { /*vram offset of the p2c training data*/ u64 p2c_train_data_offset; - struct amdgpu_bo *p2c_bo; /*vram offset of the c2p training data*/ u64 c2p_train_data_offset; diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_ras.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_ras.c index 04394c45aa03..96fc538ec824 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_ras.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_ras.c @@ -315,7 +315,7 @@ static ssize_t amdgpu_ras_debugfs_ctrl_write(struct file *f, const char __user * default: ret = -EINVAL; break; - }; + } if (ret) return -EINVAL; @@ -1311,6 +1311,7 @@ static int amdgpu_ras_badpages_read(struct amdgpu_device *adev, data = con->eh_data; if (!data || data->count == 0) { *bps = NULL; + ret = -EINVAL; goto out; } @@ -1870,7 +1871,7 @@ void amdgpu_ras_resume(struct amdgpu_device *adev) * See feature_enable_on_boot */ amdgpu_ras_disable_all_features(adev, 1); - amdgpu_ras_reset_gpu(adev, 0); + amdgpu_ras_reset_gpu(adev); } } @@ -1933,6 +1934,6 @@ void amdgpu_ras_global_ras_isr(struct amdgpu_device *adev) if (atomic_cmpxchg(&amdgpu_ras_in_intr, 0, 1) == 0) { DRM_WARN("RAS event of type ERREVENT_ATHUB_INTERRUPT detected!\n"); - amdgpu_ras_reset_gpu(adev, false); + amdgpu_ras_reset_gpu(adev); } } diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_ras.h b/drivers/gpu/drm/amd/amdgpu/amdgpu_ras.h index d4ade4739245..a5fe29a9373e 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_ras.h +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_ras.h @@ -494,8 +494,7 @@ int amdgpu_ras_add_bad_pages(struct amdgpu_device *adev, int amdgpu_ras_reserve_bad_pages(struct amdgpu_device *adev); -static inline int amdgpu_ras_reset_gpu(struct amdgpu_device *adev, - bool is_baco) +static inline int amdgpu_ras_reset_gpu(struct amdgpu_device *adev) { struct amdgpu_ras *ras = amdgpu_ras_get_context(adev); diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_sdma.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_sdma.c index 6010999d9020..a2ee30b16212 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_sdma.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_sdma.c @@ -160,7 +160,7 @@ int amdgpu_sdma_process_ras_data_cb(struct amdgpu_device *adev, struct amdgpu_iv_entry *entry) { kgd2kfd_set_sram_ecc_flag(adev->kfd.dev); - amdgpu_ras_reset_gpu(adev, 0); + amdgpu_ras_reset_gpu(adev); return AMDGPU_RAS_SUCCESS; } diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_sdma.h b/drivers/gpu/drm/amd/amdgpu/amdgpu_sdma.h index 761ff8be6314..346dcb1f7146 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_sdma.h +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_sdma.h @@ -52,6 +52,8 @@ struct amdgpu_sdma_instance { struct amdgpu_sdma { struct amdgpu_sdma_instance instance[AMDGPU_MAX_SDMA_INSTANCES]; + struct drm_gpu_scheduler *sdma_sched[AMDGPU_MAX_SDMA_INSTANCES]; + uint32_t num_sdma_sched; struct amdgpu_irq_src trap_irq; struct amdgpu_irq_src illegal_inst_irq; struct amdgpu_irq_src ecc_irq; diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_ttm.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_ttm.c index 445de594c214..3114d8a47e88 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_ttm.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_ttm.c @@ -1714,12 +1714,17 @@ static int amdgpu_ttm_training_reserve_vram_fini(struct amdgpu_device *adev) amdgpu_bo_free_kernel(&ctx->c2p_bo, NULL, NULL); ctx->c2p_bo = NULL; - amdgpu_bo_free_kernel(&ctx->p2c_bo, NULL, NULL); - ctx->p2c_bo = NULL; - return 0; } +static u64 amdgpu_ttm_training_get_c2p_offset(u64 vram_size) +{ + if ((vram_size & (SZ_1M - 1)) < (SZ_4K + 1) ) + vram_size -= SZ_1M; + + return ALIGN(vram_size, SZ_1M); +} + /** * amdgpu_ttm_training_reserve_vram_init - create bo vram reservation from memory training * @@ -1738,7 +1743,7 @@ static int amdgpu_ttm_training_reserve_vram_init(struct amdgpu_device *adev) return 0; } - ctx->c2p_train_data_offset = adev->fw_vram_usage.mem_train_fb_loc; + ctx->c2p_train_data_offset = amdgpu_ttm_training_get_c2p_offset(adev->gmc.mc_vram_size); ctx->p2c_train_data_offset = (adev->gmc.mc_vram_size - GDDR6_MEM_TRAINING_OFFSET); ctx->train_data_size = GDDR6_MEM_TRAINING_DATA_SIZE_IN_BYTES; @@ -1748,17 +1753,6 @@ static int amdgpu_ttm_training_reserve_vram_init(struct amdgpu_device *adev) ctx->c2p_train_data_offset); ret = amdgpu_bo_create_kernel_at(adev, - ctx->p2c_train_data_offset, - ctx->train_data_size, - AMDGPU_GEM_DOMAIN_VRAM, - &ctx->p2c_bo, - NULL); - if (ret) { - DRM_ERROR("alloc p2c_bo failed(%d)!\n", ret); - goto Err_out; - } - - ret = amdgpu_bo_create_kernel_at(adev, ctx->c2p_train_data_offset, ctx->train_data_size, AMDGPU_GEM_DOMAIN_VRAM, @@ -1766,15 +1760,12 @@ static int amdgpu_ttm_training_reserve_vram_init(struct amdgpu_device *adev) NULL); if (ret) { DRM_ERROR("alloc c2p_bo failed(%d)!\n", ret); - goto Err_out; + amdgpu_ttm_training_reserve_vram_fini(adev); + return ret; } ctx->init = PSP_MEM_TRAIN_RESERVE_SUCCESS; return 0; - -Err_out: - amdgpu_ttm_training_reserve_vram_fini(adev); - return ret; } /** @@ -1987,11 +1978,13 @@ void amdgpu_ttm_set_buffer_funcs_status(struct amdgpu_device *adev, bool enable) if (enable) { struct amdgpu_ring *ring; - struct drm_sched_rq *rq; + struct drm_gpu_scheduler *sched; ring = adev->mman.buffer_funcs_ring; - rq = &ring->sched.sched_rq[DRM_SCHED_PRIORITY_KERNEL]; - r = drm_sched_entity_init(&adev->mman.entity, &rq, 1, NULL); + sched = &ring->sched; + r = drm_sched_entity_init(&adev->mman.entity, + DRM_SCHED_PRIORITY_KERNEL, &sched, + 1, NULL); if (r) { DRM_ERROR("Failed setting up TTM BO move entity (%d)\n", r); diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_ucode.h b/drivers/gpu/drm/amd/amdgpu/amdgpu_ucode.h index eaf2d5b9c92f..b0e656409c03 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_ucode.h +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_ucode.h @@ -300,10 +300,10 @@ enum AMDGPU_UCODE_ID { AMDGPU_UCODE_ID_CP_MEC2_JT, AMDGPU_UCODE_ID_CP_MES, AMDGPU_UCODE_ID_CP_MES_DATA, - AMDGPU_UCODE_ID_RLC_G, AMDGPU_UCODE_ID_RLC_RESTORE_LIST_CNTL, AMDGPU_UCODE_ID_RLC_RESTORE_LIST_GPM_MEM, AMDGPU_UCODE_ID_RLC_RESTORE_LIST_SRM_MEM, + AMDGPU_UCODE_ID_RLC_G, AMDGPU_UCODE_ID_STORAGE, AMDGPU_UCODE_ID_SMC, AMDGPU_UCODE_ID_UVD, diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_umc.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_umc.c index d4fb9cf27e21..f4d40855147b 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_umc.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_umc.c @@ -95,13 +95,6 @@ int amdgpu_umc_process_ras_data_cb(struct amdgpu_device *adev, { struct ras_err_data *err_data = (struct ras_err_data *)ras_error_status; - /* When “Full RAS” is enabled, the per-IP interrupt sources should - * be disabled and the driver should only look for the aggregated - * interrupt via sync flood - */ - if (amdgpu_ras_is_supported(adev, AMDGPU_RAS_BLOCK__GFX)) - return AMDGPU_RAS_SUCCESS; - kgd2kfd_set_sram_ecc_flag(adev->kfd.dev); if (adev->umc.funcs && adev->umc.funcs->query_ras_error_count) @@ -113,6 +106,7 @@ int amdgpu_umc_process_ras_data_cb(struct amdgpu_device *adev, err_data->err_addr = kcalloc(adev->umc.max_ras_err_cnt_per_query, sizeof(struct eeprom_table_record), GFP_KERNEL); + /* still call query_ras_error_address to clear error status * even NOMEM error is encountered */ @@ -132,7 +126,7 @@ int amdgpu_umc_process_ras_data_cb(struct amdgpu_device *adev, err_data->err_addr_cnt)) DRM_WARN("Failed to add ras bad page!\n"); - amdgpu_ras_reset_gpu(adev, 0); + amdgpu_ras_reset_gpu(adev); } kfree(err_data->err_addr); diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_umc.h b/drivers/gpu/drm/amd/amdgpu/amdgpu_umc.h index 3283032a78e5..a615a1eb750b 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_umc.h +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_umc.h @@ -21,38 +21,6 @@ #ifndef __AMDGPU_UMC_H__ #define __AMDGPU_UMC_H__ -/* implement 64 bits REG operations via 32 bits interface */ -#define RREG64_UMC(reg) (RREG32(reg) | \ - ((uint64_t)RREG32((reg) + 1) << 32)) -#define WREG64_UMC(reg, v) \ - do { \ - WREG32((reg), lower_32_bits(v)); \ - WREG32((reg) + 1, upper_32_bits(v)); \ - } while (0) - -/* - * void (*func)(struct amdgpu_device *adev, struct ras_err_data *err_data, - * uint32_t umc_reg_offset, uint32_t channel_index) - */ -#define amdgpu_umc_for_each_channel(func) \ - struct ras_err_data *err_data = (struct ras_err_data *)ras_error_status; \ - uint32_t umc_inst, channel_inst, umc_reg_offset, channel_index; \ - for (umc_inst = 0; umc_inst < adev->umc.umc_inst_num; umc_inst++) { \ - /* enable the index mode to query eror count per channel */ \ - adev->umc.funcs->enable_umc_index_mode(adev, umc_inst); \ - for (channel_inst = 0; \ - channel_inst < adev->umc.channel_inst_num; \ - channel_inst++) { \ - /* calc the register offset according to channel instance */ \ - umc_reg_offset = adev->umc.channel_offs * channel_inst; \ - /* get channel index of interleaved memory */ \ - channel_index = adev->umc.channel_idx_tbl[ \ - umc_inst * adev->umc.channel_inst_num + channel_inst]; \ - (func)(adev, err_data, umc_reg_offset, channel_index); \ - } \ - } \ - adev->umc.funcs->disable_umc_index_mode(adev); - struct amdgpu_umc_funcs { void (*err_cnt_init)(struct amdgpu_device *adev); int (*ras_late_init)(struct amdgpu_device *adev); @@ -60,9 +28,6 @@ struct amdgpu_umc_funcs { void *ras_error_status); void (*query_ras_error_address)(struct amdgpu_device *adev, void *ras_error_status); - void (*enable_umc_index_mode)(struct amdgpu_device *adev, - uint32_t umc_instance); - void (*disable_umc_index_mode)(struct amdgpu_device *adev); void (*init_registers)(struct amdgpu_device *adev); }; diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_uvd.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_uvd.c index d587ffe2af8e..a92f3b18e657 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_uvd.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_uvd.c @@ -330,12 +330,13 @@ int amdgpu_uvd_sw_fini(struct amdgpu_device *adev) int amdgpu_uvd_entity_init(struct amdgpu_device *adev) { struct amdgpu_ring *ring; - struct drm_sched_rq *rq; + struct drm_gpu_scheduler *sched; int r; ring = &adev->uvd.inst[0].ring; - rq = &ring->sched.sched_rq[DRM_SCHED_PRIORITY_NORMAL]; - r = drm_sched_entity_init(&adev->uvd.entity, &rq, 1, NULL); + sched = &ring->sched; + r = drm_sched_entity_init(&adev->uvd.entity, DRM_SCHED_PRIORITY_NORMAL, + &sched, 1, NULL); if (r) { DRM_ERROR("Failed setting up UVD kernel entity.\n"); return r; diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_vce.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_vce.c index 46b590af2fd2..ceb0dbf685f1 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_vce.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_vce.c @@ -240,12 +240,13 @@ int amdgpu_vce_sw_fini(struct amdgpu_device *adev) int amdgpu_vce_entity_init(struct amdgpu_device *adev) { struct amdgpu_ring *ring; - struct drm_sched_rq *rq; + struct drm_gpu_scheduler *sched; int r; ring = &adev->vce.ring[0]; - rq = &ring->sched.sched_rq[DRM_SCHED_PRIORITY_NORMAL]; - r = drm_sched_entity_init(&adev->vce.entity, &rq, 1, NULL); + sched = &ring->sched; + r = drm_sched_entity_init(&adev->vce.entity, DRM_SCHED_PRIORITY_NORMAL, + &sched, 1, NULL); if (r != 0) { DRM_ERROR("Failed setting up VCE run queue.\n"); return r; diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_vcn.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_vcn.c index 428cfd58b37d..ed106d99d078 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_vcn.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_vcn.c @@ -28,19 +28,10 @@ #include <linux/module.h> #include <linux/pci.h> -#include <drm/drm.h> - #include "amdgpu.h" #include "amdgpu_pm.h" #include "amdgpu_vcn.h" #include "soc15d.h" -#include "soc15_common.h" - -#include "vcn/vcn_1_0_offset.h" -#include "vcn/vcn_1_0_sh_mask.h" - -/* 1 second timeout */ -#define VCN_IDLE_TIMEOUT msecs_to_jiffies(1000) /* Firmware Names */ #define FIRMWARE_RAVEN "amdgpu/raven_vcn.bin" @@ -294,6 +285,7 @@ static void amdgpu_vcn_idle_work_handler(struct work_struct *work) for (j = 0; j < adev->vcn.num_vcn_inst; ++j) { if (adev->vcn.harvest_config & (1 << j)) continue; + for (i = 0; i < adev->vcn.num_enc_rings; ++i) { fence[j] += amdgpu_fence_count_emitted(&adev->vcn.inst[j].ring_enc[i]); } @@ -306,26 +298,17 @@ static void amdgpu_vcn_idle_work_handler(struct work_struct *work) else new_state.fw_based = VCN_DPG_STATE__UNPAUSE; - if (amdgpu_fence_count_emitted(&adev->jpeg.inst[j].ring_dec)) - new_state.jpeg = VCN_DPG_STATE__PAUSE; - else - new_state.jpeg = VCN_DPG_STATE__UNPAUSE; - adev->vcn.pause_dpg_mode(adev, &new_state); } - fence[j] += amdgpu_fence_count_emitted(&adev->jpeg.inst[j].ring_dec); fence[j] += amdgpu_fence_count_emitted(&adev->vcn.inst[j].ring_dec); fences += fence[j]; } if (fences == 0) { amdgpu_gfx_off_ctrl(adev, true); - if (adev->asic_type < CHIP_ARCTURUS && adev->pm.dpm_enabled) - amdgpu_dpm_enable_uvd(adev, false); - else - amdgpu_device_ip_set_powergating_state(adev, AMD_IP_BLOCK_TYPE_VCN, - AMD_PG_STATE_GATE); + amdgpu_device_ip_set_powergating_state(adev, AMD_IP_BLOCK_TYPE_VCN, + AMD_PG_STATE_GATE); } else { schedule_delayed_work(&adev->vcn.idle_work, VCN_IDLE_TIMEOUT); } @@ -338,11 +321,8 @@ void amdgpu_vcn_ring_begin_use(struct amdgpu_ring *ring) if (set_clocks) { amdgpu_gfx_off_ctrl(adev, false); - if (adev->asic_type < CHIP_ARCTURUS && adev->pm.dpm_enabled) - amdgpu_dpm_enable_uvd(adev, true); - else - amdgpu_device_ip_set_powergating_state(adev, AMD_IP_BLOCK_TYPE_VCN, - AMD_PG_STATE_UNGATE); + amdgpu_device_ip_set_powergating_state(adev, AMD_IP_BLOCK_TYPE_VCN, + AMD_PG_STATE_UNGATE); } if (adev->pg_flags & AMD_PG_SUPPORT_VCN_DPG) { @@ -358,15 +338,8 @@ void amdgpu_vcn_ring_begin_use(struct amdgpu_ring *ring) else new_state.fw_based = VCN_DPG_STATE__UNPAUSE; - if (amdgpu_fence_count_emitted(&adev->jpeg.inst[ring->me].ring_dec)) - new_state.jpeg = VCN_DPG_STATE__PAUSE; - else - new_state.jpeg = VCN_DPG_STATE__UNPAUSE; - if (ring->funcs->type == AMDGPU_RING_TYPE_VCN_ENC) new_state.fw_based = VCN_DPG_STATE__PAUSE; - else if (ring->funcs->type == AMDGPU_RING_TYPE_VCN_JPEG) - new_state.jpeg = VCN_DPG_STATE__PAUSE; adev->vcn.pause_dpg_mode(adev, &new_state); } @@ -518,9 +491,14 @@ static int amdgpu_vcn_dec_get_destroy_msg(struct amdgpu_ring *ring, uint32_t han int amdgpu_vcn_dec_ring_test_ib(struct amdgpu_ring *ring, long timeout) { + struct amdgpu_device *adev = ring->adev; struct dma_fence *fence; long r; + /* temporarily disable ib test for sriov */ + if (amdgpu_sriov_vf(adev)) + return 0; + r = amdgpu_vcn_dec_get_create_msg(ring, 1, NULL); if (r) goto error; @@ -676,10 +654,15 @@ err: int amdgpu_vcn_enc_ring_test_ib(struct amdgpu_ring *ring, long timeout) { + struct amdgpu_device *adev = ring->adev; struct dma_fence *fence = NULL; struct amdgpu_bo *bo = NULL; long r; + /* temporarily disable ib test for sriov */ + if (amdgpu_sriov_vf(adev)) + return 0; + r = amdgpu_bo_create_reserved(ring->adev, 128 * 1024, PAGE_SIZE, AMDGPU_GEM_DOMAIN_VRAM, &bo, NULL, NULL); diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_vcn.h b/drivers/gpu/drm/amd/amdgpu/amdgpu_vcn.h index 402a5046b985..e6dee8224d33 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_vcn.h +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_vcn.h @@ -31,6 +31,7 @@ #define AMDGPU_VCN_MAX_ENC_RINGS 3 #define AMDGPU_MAX_VCN_INSTANCES 2 +#define AMDGPU_MAX_VCN_ENC_RINGS AMDGPU_VCN_MAX_ENC_RINGS * AMDGPU_MAX_VCN_INSTANCES #define AMDGPU_VCN_HARVEST_VCN0 (1 << 0) #define AMDGPU_VCN_HARVEST_VCN1 (1 << 1) @@ -56,6 +57,9 @@ #define VCN_VID_IP_ADDRESS_2_0 0x0 #define VCN_AON_IP_ADDRESS_2_0 0x30000 +/* 1 second timeout */ +#define VCN_IDLE_TIMEOUT msecs_to_jiffies(1000) + #define RREG32_SOC15_DPG_MODE(ip, inst, reg, mask, sram_sel) \ ({ WREG32_SOC15(ip, inst, mmUVD_DPG_LMA_MASK, mask); \ WREG32_SOC15(ip, inst, mmUVD_DPG_LMA_CTL, \ @@ -186,8 +190,12 @@ struct amdgpu_vcn { uint32_t *dpg_sram_curr_addr; uint8_t num_vcn_inst; - struct amdgpu_vcn_inst inst[AMDGPU_MAX_VCN_INSTANCES]; - struct amdgpu_vcn_reg internal; + struct amdgpu_vcn_inst inst[AMDGPU_MAX_VCN_INSTANCES]; + struct amdgpu_vcn_reg internal; + struct drm_gpu_scheduler *vcn_enc_sched[AMDGPU_MAX_VCN_ENC_RINGS]; + struct drm_gpu_scheduler *vcn_dec_sched[AMDGPU_MAX_VCN_INSTANCES]; + uint32_t num_vcn_enc_sched; + uint32_t num_vcn_dec_sched; unsigned harvest_config; int (*pause_dpg_mode)(struct amdgpu_device *adev, diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_vm.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_vm.c index 8f26504a59a7..4dc75eda1d91 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_vm.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_vm.c @@ -2753,14 +2753,17 @@ int amdgpu_vm_init(struct amdgpu_device *adev, struct amdgpu_vm *vm, spin_lock_init(&vm->invalidated_lock); INIT_LIST_HEAD(&vm->freed); + /* create scheduler entities for page table updates */ - r = drm_sched_entity_init(&vm->direct, adev->vm_manager.vm_pte_rqs, - adev->vm_manager.vm_pte_num_rqs, NULL); + r = drm_sched_entity_init(&vm->direct, DRM_SCHED_PRIORITY_NORMAL, + adev->vm_manager.vm_pte_scheds, + adev->vm_manager.vm_pte_num_scheds, NULL); if (r) return r; - r = drm_sched_entity_init(&vm->delayed, adev->vm_manager.vm_pte_rqs, - adev->vm_manager.vm_pte_num_rqs, NULL); + r = drm_sched_entity_init(&vm->delayed, DRM_SCHED_PRIORITY_NORMAL, + adev->vm_manager.vm_pte_scheds, + adev->vm_manager.vm_pte_num_scheds, NULL); if (r) goto error_free_direct; diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_vm.h b/drivers/gpu/drm/amd/amdgpu/amdgpu_vm.h index 7e0eb36da27d..fade4f45320c 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_vm.h +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_vm.h @@ -327,8 +327,8 @@ struct amdgpu_vm_manager { u64 vram_base_offset; /* vm pte handling */ const struct amdgpu_vm_pte_funcs *vm_pte_funcs; - struct drm_sched_rq *vm_pte_rqs[AMDGPU_MAX_RINGS]; - unsigned vm_pte_num_rqs; + struct drm_gpu_scheduler *vm_pte_scheds[AMDGPU_MAX_RINGS]; + unsigned vm_pte_num_scheds; struct amdgpu_ring *page_fault; /* partial resident texture handling */ diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_xgmi.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_xgmi.c index 61d13d8b7b20..5cf920d9358b 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_xgmi.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_xgmi.c @@ -261,6 +261,7 @@ struct amdgpu_hive_info *amdgpu_get_xgmi_hive(struct amdgpu_device *adev, int lo INIT_LIST_HEAD(&tmp->device_list); mutex_init(&tmp->hive_lock); mutex_init(&tmp->reset_lock); + task_barrier_init(&tmp->tb); if (lock) mutex_lock(&tmp->hive_lock); @@ -408,6 +409,8 @@ int amdgpu_xgmi_add_device(struct amdgpu_device *adev) top_info->num_nodes = count; hive->number_devices = count; + task_barrier_add_task(&hive->tb); + if (amdgpu_device_ip_get_ip_block(adev, AMD_IP_BLOCK_TYPE_PSP)) { list_for_each_entry(tmp_adev, &hive->device_list, gmc.xgmi.head) { /* update node list for other device in the hive */ @@ -470,6 +473,7 @@ void amdgpu_xgmi_remove_device(struct amdgpu_device *adev) mutex_destroy(&hive->hive_lock); mutex_destroy(&hive->reset_lock); } else { + task_barrier_rem_task(&hive->tb); amdgpu_xgmi_sysfs_rem_dev_info(adev, hive); mutex_unlock(&hive->hive_lock); } diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_xgmi.h b/drivers/gpu/drm/amd/amdgpu/amdgpu_xgmi.h index bbf504ff7051..74011fbc2251 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_xgmi.h +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_xgmi.h @@ -22,6 +22,7 @@ #ifndef __AMDGPU_XGMI_H__ #define __AMDGPU_XGMI_H__ +#include <drm/task_barrier.h> #include "amdgpu_psp.h" struct amdgpu_hive_info { @@ -33,6 +34,7 @@ struct amdgpu_hive_info { struct device_attribute dev_attr; struct amdgpu_device *adev; int pstate; /*0 -- low , 1 -- high , -1 unknown*/ + struct task_barrier tb; }; struct amdgpu_hive_info *amdgpu_get_xgmi_hive(struct amdgpu_device *adev, int lock); diff --git a/drivers/gpu/drm/amd/amdgpu/cik_sdma.c b/drivers/gpu/drm/amd/amdgpu/cik_sdma.c index c45304f1047c..580d3f93d670 100644 --- a/drivers/gpu/drm/amd/amdgpu/cik_sdma.c +++ b/drivers/gpu/drm/amd/amdgpu/cik_sdma.c @@ -228,7 +228,7 @@ static void cik_sdma_ring_emit_ib(struct amdgpu_ring *ring, u32 extra_bits = vmid & 0xf; /* IB packet must end on a 8 DW boundary */ - cik_sdma_ring_insert_nop(ring, (12 - (lower_32_bits(ring->wptr) & 7)) % 8); + cik_sdma_ring_insert_nop(ring, (4 - lower_32_bits(ring->wptr)) & 7); amdgpu_ring_write(ring, SDMA_PACKET(SDMA_OPCODE_INDIRECT_BUFFER, 0, extra_bits)); amdgpu_ring_write(ring, ib->gpu_addr & 0xffffffe0); /* base must be 32 byte aligned */ @@ -811,7 +811,7 @@ static void cik_sdma_ring_pad_ib(struct amdgpu_ring *ring, struct amdgpu_ib *ib) u32 pad_count; int i; - pad_count = (8 - (ib->length_dw & 0x7)) % 8; + pad_count = (-ib->length_dw) & 7; for (i = 0; i < pad_count; i++) if (sdma && sdma->burst_nop && (i == 0)) ib->ptr[ib->length_dw++] = @@ -1372,16 +1372,14 @@ static const struct amdgpu_vm_pte_funcs cik_sdma_vm_pte_funcs = { static void cik_sdma_set_vm_pte_funcs(struct amdgpu_device *adev) { - struct drm_gpu_scheduler *sched; unsigned i; adev->vm_manager.vm_pte_funcs = &cik_sdma_vm_pte_funcs; for (i = 0; i < adev->sdma.num_instances; i++) { - sched = &adev->sdma.instance[i].ring.sched; - adev->vm_manager.vm_pte_rqs[i] = - &sched->sched_rq[DRM_SCHED_PRIORITY_KERNEL]; + adev->vm_manager.vm_pte_scheds[i] = + &adev->sdma.instance[i].ring.sched; } - adev->vm_manager.vm_pte_num_rqs = adev->sdma.num_instances; + adev->vm_manager.vm_pte_num_scheds = adev->sdma.num_instances; } const struct amdgpu_ip_block_version cik_sdma_ip_block = diff --git a/drivers/gpu/drm/amd/amdgpu/df_v3_6.c b/drivers/gpu/drm/amd/amdgpu/df_v3_6.c index 4043ebcea5de..2f884d941e8d 100644 --- a/drivers/gpu/drm/amd/amdgpu/df_v3_6.c +++ b/drivers/gpu/drm/amd/amdgpu/df_v3_6.c @@ -183,6 +183,61 @@ static void df_v3_6_perfmon_wreg(struct amdgpu_device *adev, uint32_t lo_addr, spin_unlock_irqrestore(&adev->pcie_idx_lock, flags); } +/* same as perfmon_wreg but return status on write value check */ +static int df_v3_6_perfmon_arm_with_status(struct amdgpu_device *adev, + uint32_t lo_addr, uint32_t lo_val, + uint32_t hi_addr, uint32_t hi_val) +{ + unsigned long flags, address, data; + uint32_t lo_val_rb, hi_val_rb; + + address = adev->nbio.funcs->get_pcie_index_offset(adev); + data = adev->nbio.funcs->get_pcie_data_offset(adev); + + spin_lock_irqsave(&adev->pcie_idx_lock, flags); + WREG32(address, lo_addr); + WREG32(data, lo_val); + WREG32(address, hi_addr); + WREG32(data, hi_val); + + WREG32(address, lo_addr); + lo_val_rb = RREG32(data); + WREG32(address, hi_addr); + hi_val_rb = RREG32(data); + spin_unlock_irqrestore(&adev->pcie_idx_lock, flags); + + if (!(lo_val == lo_val_rb && hi_val == hi_val_rb)) + return -EBUSY; + + return 0; +} + + +/* + * retry arming counters every 100 usecs within 1 millisecond interval. + * if retry fails after time out, return error. + */ +#define ARM_RETRY_USEC_TIMEOUT 1000 +#define ARM_RETRY_USEC_INTERVAL 100 +static int df_v3_6_perfmon_arm_with_retry(struct amdgpu_device *adev, + uint32_t lo_addr, uint32_t lo_val, + uint32_t hi_addr, uint32_t hi_val) +{ + int countdown = ARM_RETRY_USEC_TIMEOUT; + + while (countdown) { + + if (!df_v3_6_perfmon_arm_with_status(adev, lo_addr, lo_val, + hi_addr, hi_val)) + break; + + countdown -= ARM_RETRY_USEC_INTERVAL; + udelay(ARM_RETRY_USEC_INTERVAL); + } + + return countdown > 0 ? 0 : -ETIME; +} + /* get the number of df counters available */ static ssize_t df_v3_6_get_df_cntr_avail(struct device *dev, struct device_attribute *attr, @@ -334,20 +389,20 @@ static void df_v3_6_pmc_get_addr(struct amdgpu_device *adev, switch (target_cntr) { case 0: - *lo_base_addr = is_ctrl ? smnPerfMonCtlLo0 : smnPerfMonCtrLo0; - *hi_base_addr = is_ctrl ? smnPerfMonCtlHi0 : smnPerfMonCtrHi0; + *lo_base_addr = is_ctrl ? smnPerfMonCtlLo4 : smnPerfMonCtrLo4; + *hi_base_addr = is_ctrl ? smnPerfMonCtlHi4 : smnPerfMonCtrHi4; break; case 1: - *lo_base_addr = is_ctrl ? smnPerfMonCtlLo1 : smnPerfMonCtrLo1; - *hi_base_addr = is_ctrl ? smnPerfMonCtlHi1 : smnPerfMonCtrHi1; + *lo_base_addr = is_ctrl ? smnPerfMonCtlLo5 : smnPerfMonCtrLo5; + *hi_base_addr = is_ctrl ? smnPerfMonCtlHi5 : smnPerfMonCtrHi5; break; case 2: - *lo_base_addr = is_ctrl ? smnPerfMonCtlLo2 : smnPerfMonCtrLo2; - *hi_base_addr = is_ctrl ? smnPerfMonCtlHi2 : smnPerfMonCtrHi2; + *lo_base_addr = is_ctrl ? smnPerfMonCtlLo6 : smnPerfMonCtrLo6; + *hi_base_addr = is_ctrl ? smnPerfMonCtlHi6 : smnPerfMonCtrHi6; break; case 3: - *lo_base_addr = is_ctrl ? smnPerfMonCtlLo3 : smnPerfMonCtrLo3; - *hi_base_addr = is_ctrl ? smnPerfMonCtlHi3 : smnPerfMonCtrHi3; + *lo_base_addr = is_ctrl ? smnPerfMonCtlLo7 : smnPerfMonCtrLo7; + *hi_base_addr = is_ctrl ? smnPerfMonCtlHi7 : smnPerfMonCtrHi7; break; } @@ -422,6 +477,44 @@ static int df_v3_6_pmc_add_cntr(struct amdgpu_device *adev, return -ENOSPC; } +#define DEFERRED_ARM_MASK (1 << 31) +static int df_v3_6_pmc_set_deferred(struct amdgpu_device *adev, + uint64_t config, bool is_deferred) +{ + int target_cntr; + + target_cntr = df_v3_6_pmc_config_2_cntr(adev, config); + + if (target_cntr < 0) + return -EINVAL; + + if (is_deferred) + adev->df_perfmon_config_assign_mask[target_cntr] |= + DEFERRED_ARM_MASK; + else + adev->df_perfmon_config_assign_mask[target_cntr] &= + ~DEFERRED_ARM_MASK; + + return 0; +} + +static bool df_v3_6_pmc_is_deferred(struct amdgpu_device *adev, + uint64_t config) +{ + int target_cntr; + + target_cntr = df_v3_6_pmc_config_2_cntr(adev, config); + + /* + * we never get target_cntr < 0 since this funciton is only called in + * pmc_count for now but we should check anyways. + */ + return (target_cntr >= 0 && + (adev->df_perfmon_config_assign_mask[target_cntr] + & DEFERRED_ARM_MASK)); + +} + /* release performance counter */ static void df_v3_6_pmc_release_cntr(struct amdgpu_device *adev, uint64_t config) @@ -451,29 +544,33 @@ static int df_v3_6_pmc_start(struct amdgpu_device *adev, uint64_t config, int is_enable) { uint32_t lo_base_addr, hi_base_addr, lo_val, hi_val; - int ret = 0; + int err = 0, ret = 0; switch (adev->asic_type) { case CHIP_VEGA20: + if (is_enable) + return df_v3_6_pmc_add_cntr(adev, config); df_v3_6_reset_perfmon_cntr(adev, config); - if (is_enable) { - ret = df_v3_6_pmc_add_cntr(adev, config); - } else { - ret = df_v3_6_pmc_get_ctrl_settings(adev, + ret = df_v3_6_pmc_get_ctrl_settings(adev, config, &lo_base_addr, &hi_base_addr, &lo_val, &hi_val); - if (ret) - return ret; + if (ret) + return ret; + + err = df_v3_6_perfmon_arm_with_retry(adev, + lo_base_addr, + lo_val, + hi_base_addr, + hi_val); - df_v3_6_perfmon_wreg(adev, lo_base_addr, lo_val, - hi_base_addr, hi_val); - } + if (err) + ret = df_v3_6_pmc_set_deferred(adev, config, true); break; default: @@ -501,7 +598,7 @@ static int df_v3_6_pmc_stop(struct amdgpu_device *adev, uint64_t config, if (ret) return ret; - df_v3_6_perfmon_wreg(adev, lo_base_addr, 0, hi_base_addr, 0); + df_v3_6_reset_perfmon_cntr(adev, config); if (is_disable) df_v3_6_pmc_release_cntr(adev, config); @@ -518,18 +615,29 @@ static void df_v3_6_pmc_get_count(struct amdgpu_device *adev, uint64_t config, uint64_t *count) { - uint32_t lo_base_addr, hi_base_addr, lo_val, hi_val; + uint32_t lo_base_addr, hi_base_addr, lo_val = 0, hi_val = 0; *count = 0; switch (adev->asic_type) { case CHIP_VEGA20: - df_v3_6_pmc_get_read_settings(adev, config, &lo_base_addr, &hi_base_addr); if ((lo_base_addr == 0) || (hi_base_addr == 0)) return; + /* rearm the counter or throw away count value on failure */ + if (df_v3_6_pmc_is_deferred(adev, config)) { + int rearm_err = df_v3_6_perfmon_arm_with_status(adev, + lo_base_addr, lo_val, + hi_base_addr, hi_val); + + if (rearm_err) + return; + + df_v3_6_pmc_set_deferred(adev, config, false); + } + df_v3_6_perfmon_rreg(adev, lo_base_addr, &lo_val, hi_base_addr, &hi_val); @@ -542,7 +650,6 @@ static void df_v3_6_pmc_get_count(struct amdgpu_device *adev, config, lo_base_addr, hi_base_addr, lo_val, hi_val); break; - default: break; } diff --git a/drivers/gpu/drm/amd/amdgpu/gfx_v10_0.c b/drivers/gpu/drm/amd/amdgpu/gfx_v10_0.c index 98db25215d0d..6bc3b937fba2 100644 --- a/drivers/gpu/drm/amd/amdgpu/gfx_v10_0.c +++ b/drivers/gpu/drm/amd/amdgpu/gfx_v10_0.c @@ -471,18 +471,10 @@ static int gfx_v10_0_ring_test_ring(struct amdgpu_ring *ring) else udelay(1); } - if (i < adev->usec_timeout) { - if (amdgpu_emu_mode == 1) - DRM_INFO("ring test on %d succeeded in %d msecs\n", - ring->idx, i); - else - DRM_INFO("ring test on %d succeeded in %d usecs\n", - ring->idx, i); - } else { - DRM_ERROR("amdgpu: ring %d test failed (scratch(0x%04X)=0x%08X)\n", - ring->idx, scratch, tmp); - r = -EINVAL; - } + + if (i >= adev->usec_timeout) + r = -ETIMEDOUT; + amdgpu_gfx_scratch_free(adev, scratch); return r; @@ -532,14 +524,10 @@ static int gfx_v10_0_ring_test_ib(struct amdgpu_ring *ring, long timeout) } tmp = RREG32(scratch); - if (tmp == 0xDEADBEEF) { - DRM_INFO("ib test on ring %d succeeded\n", ring->idx); + if (tmp == 0xDEADBEEF) r = 0; - } else { - DRM_ERROR("amdgpu: ib test failed (scratch(0x%04X)=0x%08X)\n", - scratch, tmp); + else r = -EINVAL; - } err2: amdgpu_ib_free(adev, &ib, NULL); dma_fence_put(f); @@ -588,8 +576,7 @@ static void gfx_v10_0_check_fw_write_wait(struct amdgpu_device *adev) } if (adev->gfx.cp_fw_write_wait == false) - DRM_WARN_ONCE("Warning: check cp_fw_version and update it to realize \ - GRBM requires 1-cycle delay in cp firmware\n"); + DRM_WARN_ONCE("CP firmware version too old, please update!"); } @@ -1963,7 +1950,7 @@ static int gfx_v10_0_parse_rlc_toc(struct amdgpu_device *adev) rlc_autoload_info[rlc_toc->id].size = rlc_toc->size * 4; rlc_toc++; - }; + } return 0; } @@ -3606,23 +3593,16 @@ static int gfx_v10_0_cp_resume(struct amdgpu_device *adev) for (i = 0; i < adev->gfx.num_gfx_rings; i++) { ring = &adev->gfx.gfx_ring[i]; - DRM_INFO("gfx %d ring me %d pipe %d q %d\n", - i, ring->me, ring->pipe, ring->queue); - r = amdgpu_ring_test_ring(ring); - if (r) { - ring->sched.ready = false; + r = amdgpu_ring_test_helper(ring); + if (r) return r; - } } for (i = 0; i < adev->gfx.num_compute_rings; i++) { ring = &adev->gfx.compute_ring[i]; - ring->sched.ready = true; - DRM_INFO("compute ring %d mec %d pipe %d q %d\n", - i, ring->me, ring->pipe, ring->queue); - r = amdgpu_ring_test_ring(ring); + r = amdgpu_ring_test_helper(ring); if (r) - ring->sched.ready = false; + return r; } return 0; diff --git a/drivers/gpu/drm/amd/amdgpu/gfx_v9_0.c b/drivers/gpu/drm/amd/amdgpu/gfx_v9_0.c index 2616f1b59bbd..a5492e375f29 100644 --- a/drivers/gpu/drm/amd/amdgpu/gfx_v9_0.c +++ b/drivers/gpu/drm/amd/amdgpu/gfx_v9_0.c @@ -48,8 +48,15 @@ #include "amdgpu_ras.h" -#include "sdma0/sdma0_4_0_offset.h" -#include "sdma1/sdma1_4_0_offset.h" +#include "sdma0/sdma0_4_2_offset.h" +#include "sdma1/sdma1_4_2_offset.h" +#include "sdma2/sdma2_4_2_2_offset.h" +#include "sdma3/sdma3_4_2_2_offset.h" +#include "sdma4/sdma4_4_2_2_offset.h" +#include "sdma5/sdma5_4_2_2_offset.h" +#include "sdma6/sdma6_4_2_2_offset.h" +#include "sdma7/sdma7_4_2_2_offset.h" + #define GFX9_NUM_GFX_RINGS 1 #define GFX9_MEC_HPD_SIZE 4096 #define RLCG_UCODE_LOADING_START_ADDRESS 0x00002000L @@ -981,8 +988,7 @@ static void gfx_v9_0_check_fw_write_wait(struct amdgpu_device *adev) (adev->gfx.mec_feature_version < 46) || (adev->gfx.pfp_fw_version < 0x000000b7) || (adev->gfx.pfp_feature_version < 46)) - DRM_WARN_ONCE("Warning: check cp_fw_version and update it to realize \ - GRBM requires 1-cycle delay in cp firmware\n"); + DRM_WARN_ONCE("CP firmware version too old, please update!"); switch (adev->asic_type) { case CHIP_VEGA10: @@ -1042,17 +1048,10 @@ static void gfx_v9_0_check_if_need_gfxoff(struct amdgpu_device *adev) case CHIP_VEGA20: break; case CHIP_RAVEN: - /* Disable GFXOFF on original raven. There are combinations - * of sbios and platforms that are not stable. - */ - if (!(adev->rev_id >= 0x8 || adev->pdev->device == 0x15d8)) - adev->pm.pp_feature &= ~PP_GFXOFF_MASK; - else if (!(adev->rev_id >= 0x8 || adev->pdev->device == 0x15d8) - &&((adev->gfx.rlc_fw_version != 106 && - adev->gfx.rlc_fw_version < 531) || - (adev->gfx.rlc_fw_version == 53815) || - (adev->gfx.rlc_feature_version < 1) || - !adev->gfx.rlc.is_rlc_v2_1)) + if (!(adev->rev_id >= 0x8 || + adev->pdev->device == 0x15d8) && + (adev->pm.fw_version < 0x41e2b || /* not raven1 fresh */ + !adev->gfx.rlc.is_rlc_v2_1)) /* without rlc save restore ucodes */ adev->pm.pp_feature &= ~PP_GFXOFF_MASK; if (adev->pm.pp_feature & PP_GFXOFF_MASK) @@ -3933,43 +3932,58 @@ static const u32 sgpr_init_compute_shader[] = 0xbe800080, 0xbf810000, }; +/* When below register arrays changed, please update gpr_reg_size, + and sec_ded_counter_reg_size in function gfx_v9_0_do_edc_gpr_workarounds, + to cover all gfx9 ASICs */ static const struct soc15_reg_entry vgpr_init_regs[] = { - { SOC15_REG_ENTRY(GC, 0, mmCOMPUTE_STATIC_THREAD_MGMT_SE0), 0xffffffff }, - { SOC15_REG_ENTRY(GC, 0, mmCOMPUTE_STATIC_THREAD_MGMT_SE1), 0xffffffff }, - { SOC15_REG_ENTRY(GC, 0, mmCOMPUTE_STATIC_THREAD_MGMT_SE2), 0xffffffff }, - { SOC15_REG_ENTRY(GC, 0, mmCOMPUTE_STATIC_THREAD_MGMT_SE3), 0xffffffff }, { SOC15_REG_ENTRY(GC, 0, mmCOMPUTE_RESOURCE_LIMITS), 0x0000000 }, { SOC15_REG_ENTRY(GC, 0, mmCOMPUTE_NUM_THREAD_X), 0x40 }, { SOC15_REG_ENTRY(GC, 0, mmCOMPUTE_NUM_THREAD_Y), 4 }, { SOC15_REG_ENTRY(GC, 0, mmCOMPUTE_NUM_THREAD_Z), 1 }, { SOC15_REG_ENTRY(GC, 0, mmCOMPUTE_PGM_RSRC1), 0x3f }, { SOC15_REG_ENTRY(GC, 0, mmCOMPUTE_PGM_RSRC2), 0x400000 }, /* 64KB LDS */ + { SOC15_REG_ENTRY(GC, 0, mmCOMPUTE_STATIC_THREAD_MGMT_SE0), 0xffffffff }, + { SOC15_REG_ENTRY(GC, 0, mmCOMPUTE_STATIC_THREAD_MGMT_SE1), 0xffffffff }, + { SOC15_REG_ENTRY(GC, 0, mmCOMPUTE_STATIC_THREAD_MGMT_SE2), 0xffffffff }, + { SOC15_REG_ENTRY(GC, 0, mmCOMPUTE_STATIC_THREAD_MGMT_SE3), 0xffffffff }, + { SOC15_REG_ENTRY(GC, 0, mmCOMPUTE_STATIC_THREAD_MGMT_SE4), 0xffffffff }, + { SOC15_REG_ENTRY(GC, 0, mmCOMPUTE_STATIC_THREAD_MGMT_SE5), 0xffffffff }, + { SOC15_REG_ENTRY(GC, 0, mmCOMPUTE_STATIC_THREAD_MGMT_SE6), 0xffffffff }, + { SOC15_REG_ENTRY(GC, 0, mmCOMPUTE_STATIC_THREAD_MGMT_SE7), 0xffffffff }, }; static const struct soc15_reg_entry sgpr1_init_regs[] = { - { SOC15_REG_ENTRY(GC, 0, mmCOMPUTE_STATIC_THREAD_MGMT_SE0), 0x000000ff }, - { SOC15_REG_ENTRY(GC, 0, mmCOMPUTE_STATIC_THREAD_MGMT_SE1), 0x000000ff }, - { SOC15_REG_ENTRY(GC, 0, mmCOMPUTE_STATIC_THREAD_MGMT_SE2), 0x000000ff }, - { SOC15_REG_ENTRY(GC, 0, mmCOMPUTE_STATIC_THREAD_MGMT_SE3), 0x000000ff }, { SOC15_REG_ENTRY(GC, 0, mmCOMPUTE_RESOURCE_LIMITS), 0x0000000 }, { SOC15_REG_ENTRY(GC, 0, mmCOMPUTE_NUM_THREAD_X), 0x40 }, { SOC15_REG_ENTRY(GC, 0, mmCOMPUTE_NUM_THREAD_Y), 8 }, { SOC15_REG_ENTRY(GC, 0, mmCOMPUTE_NUM_THREAD_Z), 1 }, { SOC15_REG_ENTRY(GC, 0, mmCOMPUTE_PGM_RSRC1), 0x240 }, /* (80 GPRS) */ { SOC15_REG_ENTRY(GC, 0, mmCOMPUTE_PGM_RSRC2), 0x0 }, + { SOC15_REG_ENTRY(GC, 0, mmCOMPUTE_STATIC_THREAD_MGMT_SE0), 0x000000ff }, + { SOC15_REG_ENTRY(GC, 0, mmCOMPUTE_STATIC_THREAD_MGMT_SE1), 0x000000ff }, + { SOC15_REG_ENTRY(GC, 0, mmCOMPUTE_STATIC_THREAD_MGMT_SE2), 0x000000ff }, + { SOC15_REG_ENTRY(GC, 0, mmCOMPUTE_STATIC_THREAD_MGMT_SE3), 0x000000ff }, + { SOC15_REG_ENTRY(GC, 0, mmCOMPUTE_STATIC_THREAD_MGMT_SE4), 0x000000ff }, + { SOC15_REG_ENTRY(GC, 0, mmCOMPUTE_STATIC_THREAD_MGMT_SE5), 0x000000ff }, + { SOC15_REG_ENTRY(GC, 0, mmCOMPUTE_STATIC_THREAD_MGMT_SE6), 0x000000ff }, + { SOC15_REG_ENTRY(GC, 0, mmCOMPUTE_STATIC_THREAD_MGMT_SE7), 0x000000ff }, }; static const struct soc15_reg_entry sgpr2_init_regs[] = { - { SOC15_REG_ENTRY(GC, 0, mmCOMPUTE_STATIC_THREAD_MGMT_SE0), 0x0000ff00 }, - { SOC15_REG_ENTRY(GC, 0, mmCOMPUTE_STATIC_THREAD_MGMT_SE1), 0x0000ff00 }, - { SOC15_REG_ENTRY(GC, 0, mmCOMPUTE_STATIC_THREAD_MGMT_SE2), 0x0000ff00 }, - { SOC15_REG_ENTRY(GC, 0, mmCOMPUTE_STATIC_THREAD_MGMT_SE3), 0x0000ff00 }, { SOC15_REG_ENTRY(GC, 0, mmCOMPUTE_RESOURCE_LIMITS), 0x0000000 }, { SOC15_REG_ENTRY(GC, 0, mmCOMPUTE_NUM_THREAD_X), 0x40 }, { SOC15_REG_ENTRY(GC, 0, mmCOMPUTE_NUM_THREAD_Y), 8 }, { SOC15_REG_ENTRY(GC, 0, mmCOMPUTE_NUM_THREAD_Z), 1 }, { SOC15_REG_ENTRY(GC, 0, mmCOMPUTE_PGM_RSRC1), 0x240 }, /* (80 GPRS) */ { SOC15_REG_ENTRY(GC, 0, mmCOMPUTE_PGM_RSRC2), 0x0 }, + { SOC15_REG_ENTRY(GC, 0, mmCOMPUTE_STATIC_THREAD_MGMT_SE0), 0x0000ff00 }, + { SOC15_REG_ENTRY(GC, 0, mmCOMPUTE_STATIC_THREAD_MGMT_SE1), 0x0000ff00 }, + { SOC15_REG_ENTRY(GC, 0, mmCOMPUTE_STATIC_THREAD_MGMT_SE2), 0x0000ff00 }, + { SOC15_REG_ENTRY(GC, 0, mmCOMPUTE_STATIC_THREAD_MGMT_SE3), 0x0000ff00 }, + { SOC15_REG_ENTRY(GC, 0, mmCOMPUTE_STATIC_THREAD_MGMT_SE4), 0x0000ff00 }, + { SOC15_REG_ENTRY(GC, 0, mmCOMPUTE_STATIC_THREAD_MGMT_SE5), 0x0000ff00 }, + { SOC15_REG_ENTRY(GC, 0, mmCOMPUTE_STATIC_THREAD_MGMT_SE6), 0x0000ff00 }, + { SOC15_REG_ENTRY(GC, 0, mmCOMPUTE_STATIC_THREAD_MGMT_SE7), 0x0000ff00 }, }; static const struct soc15_reg_entry sec_ded_counter_registers[] = { @@ -4006,9 +4020,15 @@ static const struct soc15_reg_entry sec_ded_counter_registers[] = { { SOC15_REG_ENTRY(GC, 0, mmTCC_EDC_CNT2), 0, 1, 16}, { SOC15_REG_ENTRY(GC, 0, mmTCA_EDC_CNT), 0, 1, 2}, { SOC15_REG_ENTRY(GC, 0, mmSQC_EDC_CNT3), 0, 4, 6}, + { SOC15_REG_ENTRY(HDP, 0, mmHDP_EDC_CNT), 0, 1, 1}, { SOC15_REG_ENTRY(SDMA0, 0, mmSDMA0_EDC_COUNTER), 0, 1, 1}, { SOC15_REG_ENTRY(SDMA1, 0, mmSDMA1_EDC_COUNTER), 0, 1, 1}, - { SOC15_REG_ENTRY(HDP, 0, mmHDP_EDC_CNT), 0, 1, 1}, + { SOC15_REG_ENTRY(SDMA2, 0, mmSDMA2_EDC_COUNTER), 0, 1, 1}, + { SOC15_REG_ENTRY(SDMA3, 0, mmSDMA3_EDC_COUNTER), 0, 1, 1}, + { SOC15_REG_ENTRY(SDMA4, 0, mmSDMA4_EDC_COUNTER), 0, 1, 1}, + { SOC15_REG_ENTRY(SDMA5, 0, mmSDMA5_EDC_COUNTER), 0, 1, 1}, + { SOC15_REG_ENTRY(SDMA6, 0, mmSDMA6_EDC_COUNTER), 0, 1, 1}, + { SOC15_REG_ENTRY(SDMA7, 0, mmSDMA7_EDC_COUNTER), 0, 1, 1}, }; static int gfx_v9_0_do_edc_gds_workarounds(struct amdgpu_device *adev) @@ -4067,6 +4087,13 @@ static int gfx_v9_0_do_edc_gpr_workarounds(struct amdgpu_device *adev) unsigned total_size, vgpr_offset, sgpr_offset; u64 gpu_addr; + int compute_dim_x = adev->gfx.config.max_shader_engines * + adev->gfx.config.max_cu_per_sh * + adev->gfx.config.max_sh_per_se; + int sgpr_work_group_size = 5; + int gpr_reg_size = compute_dim_x / 16 + 6; + int sec_ded_counter_reg_size = adev->sdma.num_instances + 34; + /* only support when RAS is enabled */ if (!amdgpu_ras_is_supported(adev, AMDGPU_RAS_BLOCK__GFX)) return 0; @@ -4076,11 +4103,11 @@ static int gfx_v9_0_do_edc_gpr_workarounds(struct amdgpu_device *adev) return 0; total_size = - ((ARRAY_SIZE(vgpr_init_regs) * 3) + 4 + 5 + 2) * 4; + (gpr_reg_size * 3 + 4 + 5 + 2) * 4; /* VGPRS */ total_size += - ((ARRAY_SIZE(sgpr1_init_regs) * 3) + 4 + 5 + 2) * 4; + (gpr_reg_size * 3 + 4 + 5 + 2) * 4; /* SGPRS1 */ total_size += - ((ARRAY_SIZE(sgpr2_init_regs) * 3) + 4 + 5 + 2) * 4; + (gpr_reg_size * 3 + 4 + 5 + 2) * 4; /* SGPRS2 */ total_size = ALIGN(total_size, 256); vgpr_offset = total_size; total_size += ALIGN(sizeof(vgpr_init_compute_shader), 256); @@ -4107,7 +4134,7 @@ static int gfx_v9_0_do_edc_gpr_workarounds(struct amdgpu_device *adev) /* VGPR */ /* write the register state for the compute dispatch */ - for (i = 0; i < ARRAY_SIZE(vgpr_init_regs); i++) { + for (i = 0; i < gpr_reg_size; i++) { ib.ptr[ib.length_dw++] = PACKET3(PACKET3_SET_SH_REG, 1); ib.ptr[ib.length_dw++] = SOC15_REG_ENTRY_OFFSET(vgpr_init_regs[i]) - PACKET3_SET_SH_REG_START; @@ -4123,7 +4150,7 @@ static int gfx_v9_0_do_edc_gpr_workarounds(struct amdgpu_device *adev) /* write dispatch packet */ ib.ptr[ib.length_dw++] = PACKET3(PACKET3_DISPATCH_DIRECT, 3); - ib.ptr[ib.length_dw++] = 0x40*2; /* x */ + ib.ptr[ib.length_dw++] = compute_dim_x; /* x */ ib.ptr[ib.length_dw++] = 1; /* y */ ib.ptr[ib.length_dw++] = 1; /* z */ ib.ptr[ib.length_dw++] = @@ -4135,7 +4162,7 @@ static int gfx_v9_0_do_edc_gpr_workarounds(struct amdgpu_device *adev) /* SGPR1 */ /* write the register state for the compute dispatch */ - for (i = 0; i < ARRAY_SIZE(sgpr1_init_regs); i++) { + for (i = 0; i < gpr_reg_size; i++) { ib.ptr[ib.length_dw++] = PACKET3(PACKET3_SET_SH_REG, 1); ib.ptr[ib.length_dw++] = SOC15_REG_ENTRY_OFFSET(sgpr1_init_regs[i]) - PACKET3_SET_SH_REG_START; @@ -4151,7 +4178,7 @@ static int gfx_v9_0_do_edc_gpr_workarounds(struct amdgpu_device *adev) /* write dispatch packet */ ib.ptr[ib.length_dw++] = PACKET3(PACKET3_DISPATCH_DIRECT, 3); - ib.ptr[ib.length_dw++] = 0xA0*2; /* x */ + ib.ptr[ib.length_dw++] = compute_dim_x / 2 * sgpr_work_group_size; /* x */ ib.ptr[ib.length_dw++] = 1; /* y */ ib.ptr[ib.length_dw++] = 1; /* z */ ib.ptr[ib.length_dw++] = @@ -4163,7 +4190,7 @@ static int gfx_v9_0_do_edc_gpr_workarounds(struct amdgpu_device *adev) /* SGPR2 */ /* write the register state for the compute dispatch */ - for (i = 0; i < ARRAY_SIZE(sgpr2_init_regs); i++) { + for (i = 0; i < gpr_reg_size; i++) { ib.ptr[ib.length_dw++] = PACKET3(PACKET3_SET_SH_REG, 1); ib.ptr[ib.length_dw++] = SOC15_REG_ENTRY_OFFSET(sgpr2_init_regs[i]) - PACKET3_SET_SH_REG_START; @@ -4179,7 +4206,7 @@ static int gfx_v9_0_do_edc_gpr_workarounds(struct amdgpu_device *adev) /* write dispatch packet */ ib.ptr[ib.length_dw++] = PACKET3(PACKET3_DISPATCH_DIRECT, 3); - ib.ptr[ib.length_dw++] = 0xA0*2; /* x */ + ib.ptr[ib.length_dw++] = compute_dim_x / 2 * sgpr_work_group_size; /* x */ ib.ptr[ib.length_dw++] = 1; /* y */ ib.ptr[ib.length_dw++] = 1; /* z */ ib.ptr[ib.length_dw++] = @@ -4205,7 +4232,7 @@ static int gfx_v9_0_do_edc_gpr_workarounds(struct amdgpu_device *adev) /* read back registers to clear the counters */ mutex_lock(&adev->grbm_idx_mutex); - for (i = 0; i < ARRAY_SIZE(sec_ded_counter_registers); i++) { + for (i = 0; i < sec_ded_counter_reg_size; i++) { for (j = 0; j < sec_ded_counter_registers[i].se_num; j++) { for (k = 0; k < sec_ded_counter_registers[i].instance; k++) { gfx_v9_0_select_se_sh(adev, j, 0x0, k); diff --git a/drivers/gpu/drm/amd/amdgpu/gfxhub_v1_0.c b/drivers/gpu/drm/amd/amdgpu/gfxhub_v1_0.c index e91bd7945777..1a2f18b908fe 100644 --- a/drivers/gpu/drm/amd/amdgpu/gfxhub_v1_0.c +++ b/drivers/gpu/drm/amd/amdgpu/gfxhub_v1_0.c @@ -75,40 +75,45 @@ static void gfxhub_v1_0_init_system_aperture_regs(struct amdgpu_device *adev) WREG32_SOC15_RLC(GC, 0, mmMC_VM_AGP_BOT, adev->gmc.agp_start >> 24); WREG32_SOC15_RLC(GC, 0, mmMC_VM_AGP_TOP, adev->gmc.agp_end >> 24); - /* Program the system aperture low logical page number. */ - WREG32_SOC15_RLC(GC, 0, mmMC_VM_SYSTEM_APERTURE_LOW_ADDR, - min(adev->gmc.fb_start, adev->gmc.agp_start) >> 18); - - if (adev->asic_type == CHIP_RAVEN && adev->rev_id >= 0x8) - /* - * Raven2 has a HW issue that it is unable to use the vram which - * is out of MC_VM_SYSTEM_APERTURE_HIGH_ADDR. So here is the - * workaround that increase system aperture high address (add 1) - * to get rid of the VM fault and hardware hang. - */ - WREG32_SOC15_RLC(GC, 0, mmMC_VM_SYSTEM_APERTURE_HIGH_ADDR, - max((adev->gmc.fb_end >> 18) + 0x1, - adev->gmc.agp_end >> 18)); - else - WREG32_SOC15_RLC(GC, 0, mmMC_VM_SYSTEM_APERTURE_HIGH_ADDR, - max(adev->gmc.fb_end, adev->gmc.agp_end) >> 18); - - /* Set default page address. */ - value = adev->vram_scratch.gpu_addr - adev->gmc.vram_start - + adev->vm_manager.vram_base_offset; - WREG32_SOC15(GC, 0, mmMC_VM_SYSTEM_APERTURE_DEFAULT_ADDR_LSB, - (u32)(value >> 12)); - WREG32_SOC15(GC, 0, mmMC_VM_SYSTEM_APERTURE_DEFAULT_ADDR_MSB, - (u32)(value >> 44)); - - /* Program "protection fault". */ - WREG32_SOC15(GC, 0, mmVM_L2_PROTECTION_FAULT_DEFAULT_ADDR_LO32, - (u32)(adev->dummy_page_addr >> 12)); - WREG32_SOC15(GC, 0, mmVM_L2_PROTECTION_FAULT_DEFAULT_ADDR_HI32, - (u32)((u64)adev->dummy_page_addr >> 44)); - - WREG32_FIELD15(GC, 0, VM_L2_PROTECTION_FAULT_CNTL2, - ACTIVE_PAGE_MIGRATION_PTE_READ_RETRY, 1); + if (!amdgpu_sriov_vf(adev) || adev->asic_type <= CHIP_VEGA10) { + /* Program the system aperture low logical page number. */ + WREG32_SOC15_RLC(GC, 0, mmMC_VM_SYSTEM_APERTURE_LOW_ADDR, + min(adev->gmc.fb_start, adev->gmc.agp_start) >> 18); + + if (adev->asic_type == CHIP_RAVEN && adev->rev_id >= 0x8) + /* + * Raven2 has a HW issue that it is unable to use the + * vram which is out of MC_VM_SYSTEM_APERTURE_HIGH_ADDR. + * So here is the workaround that increase system + * aperture high address (add 1) to get rid of the VM + * fault and hardware hang. + */ + WREG32_SOC15_RLC(GC, 0, + mmMC_VM_SYSTEM_APERTURE_HIGH_ADDR, + max((adev->gmc.fb_end >> 18) + 0x1, + adev->gmc.agp_end >> 18)); + else + WREG32_SOC15_RLC( + GC, 0, mmMC_VM_SYSTEM_APERTURE_HIGH_ADDR, + max(adev->gmc.fb_end, adev->gmc.agp_end) >> 18); + + /* Set default page address. */ + value = adev->vram_scratch.gpu_addr - adev->gmc.vram_start + + adev->vm_manager.vram_base_offset; + WREG32_SOC15(GC, 0, mmMC_VM_SYSTEM_APERTURE_DEFAULT_ADDR_LSB, + (u32)(value >> 12)); + WREG32_SOC15(GC, 0, mmMC_VM_SYSTEM_APERTURE_DEFAULT_ADDR_MSB, + (u32)(value >> 44)); + + /* Program "protection fault". */ + WREG32_SOC15(GC, 0, mmVM_L2_PROTECTION_FAULT_DEFAULT_ADDR_LO32, + (u32)(adev->dummy_page_addr >> 12)); + WREG32_SOC15(GC, 0, mmVM_L2_PROTECTION_FAULT_DEFAULT_ADDR_HI32, + (u32)((u64)adev->dummy_page_addr >> 44)); + + WREG32_FIELD15(GC, 0, VM_L2_PROTECTION_FAULT_CNTL2, + ACTIVE_PAGE_MIGRATION_PTE_READ_RETRY, 1); + } } static void gfxhub_v1_0_init_tlb_regs(struct amdgpu_device *adev) @@ -264,7 +269,7 @@ static void gfxhub_v1_0_program_invalidation(struct amdgpu_device *adev) int gfxhub_v1_0_gart_enable(struct amdgpu_device *adev) { - if (amdgpu_sriov_vf(adev)) { + if (amdgpu_sriov_vf(adev) && adev->asic_type != CHIP_ARCTURUS) { /* * MC_VM_FB_LOCATION_BASE/TOP is NULL for VF, becuase they are * VF copy registers so vbios post doesn't program them, for @@ -280,10 +285,12 @@ int gfxhub_v1_0_gart_enable(struct amdgpu_device *adev) gfxhub_v1_0_init_gart_aperture_regs(adev); gfxhub_v1_0_init_system_aperture_regs(adev); gfxhub_v1_0_init_tlb_regs(adev); - gfxhub_v1_0_init_cache_regs(adev); + if (!amdgpu_sriov_vf(adev)) + gfxhub_v1_0_init_cache_regs(adev); gfxhub_v1_0_enable_system_domain(adev); - gfxhub_v1_0_disable_identity_aperture(adev); + if (!amdgpu_sriov_vf(adev)) + gfxhub_v1_0_disable_identity_aperture(adev); gfxhub_v1_0_setup_vmid_config(adev); gfxhub_v1_0_program_invalidation(adev); diff --git a/drivers/gpu/drm/amd/amdgpu/gmc_v10_0.c b/drivers/gpu/drm/amd/amdgpu/gmc_v10_0.c index f5725336a5f2..da9765ff45d6 100644 --- a/drivers/gpu/drm/amd/amdgpu/gmc_v10_0.c +++ b/drivers/gpu/drm/amd/amdgpu/gmc_v10_0.c @@ -564,22 +564,11 @@ static int gmc_v10_0_early_init(void *handle) static int gmc_v10_0_late_init(void *handle) { struct amdgpu_device *adev = (struct amdgpu_device *)handle; - unsigned vm_inv_eng[AMDGPU_MAX_VMHUBS] = { 4, 4 }; - unsigned i; - - for(i = 0; i < adev->num_rings; ++i) { - struct amdgpu_ring *ring = adev->rings[i]; - unsigned vmhub = ring->funcs->vmhub; - - ring->vm_inv_eng = vm_inv_eng[vmhub]++; - dev_info(adev->dev, "ring %u(%s) uses VM inv eng %u on hub %u\n", - ring->idx, ring->name, ring->vm_inv_eng, - ring->funcs->vmhub); - } + int r; - /* Engine 17 is used for GART flushes */ - for(i = 0; i < AMDGPU_MAX_VMHUBS; ++i) - BUG_ON(vm_inv_eng[i] > 17); + r = amdgpu_gmc_allocate_vm_inv_eng(adev); + if (r) + return r; return amdgpu_irq_get(adev, &adev->gmc.vm_fault, 0); } diff --git a/drivers/gpu/drm/amd/amdgpu/gmc_v9_0.c b/drivers/gpu/drm/amd/amdgpu/gmc_v9_0.c index fa025ceeea0f..26194ac9af98 100644 --- a/drivers/gpu/drm/amd/amdgpu/gmc_v9_0.c +++ b/drivers/gpu/drm/amd/amdgpu/gmc_v9_0.c @@ -207,6 +207,11 @@ static int gmc_v9_0_ecc_interrupt_state(struct amdgpu_device *adev, { u32 bits, i, tmp, reg; + /* Devices newer then VEGA10/12 shall have these programming + sequences performed by PSP BL */ + if (adev->asic_type >= CHIP_VEGA20) + return 0; + bits = 0x7f; switch (state) { @@ -393,8 +398,10 @@ static void gmc_v9_0_set_irq_funcs(struct amdgpu_device *adev) adev->gmc.vm_fault.num_types = 1; adev->gmc.vm_fault.funcs = &gmc_v9_0_irq_funcs; - adev->gmc.ecc_irq.num_types = 1; - adev->gmc.ecc_irq.funcs = &gmc_v9_0_ecc_funcs; + if (!amdgpu_sriov_vf(adev)) { + adev->gmc.ecc_irq.num_types = 1; + adev->gmc.ecc_irq.funcs = &gmc_v9_0_ecc_funcs; + } } static uint32_t gmc_v9_0_get_invalidate_req(unsigned int vmid, @@ -790,36 +797,6 @@ static bool gmc_v9_0_keep_stolen_memory(struct amdgpu_device *adev) } } -static int gmc_v9_0_allocate_vm_inv_eng(struct amdgpu_device *adev) -{ - struct amdgpu_ring *ring; - unsigned vm_inv_engs[AMDGPU_MAX_VMHUBS] = - {GFXHUB_FREE_VM_INV_ENGS_BITMAP, MMHUB_FREE_VM_INV_ENGS_BITMAP, - GFXHUB_FREE_VM_INV_ENGS_BITMAP}; - unsigned i; - unsigned vmhub, inv_eng; - - for (i = 0; i < adev->num_rings; ++i) { - ring = adev->rings[i]; - vmhub = ring->funcs->vmhub; - - inv_eng = ffs(vm_inv_engs[vmhub]); - if (!inv_eng) { - dev_err(adev->dev, "no VM inv eng for ring %s\n", - ring->name); - return -EINVAL; - } - - ring->vm_inv_eng = inv_eng - 1; - vm_inv_engs[vmhub] &= ~(1 << ring->vm_inv_eng); - - dev_info(adev->dev, "ring %s uses VM inv eng %u on hub %u\n", - ring->name, ring->vm_inv_eng, ring->funcs->vmhub); - } - - return 0; -} - static int gmc_v9_0_late_init(void *handle) { struct amdgpu_device *adev = (struct amdgpu_device *)handle; @@ -828,7 +805,7 @@ static int gmc_v9_0_late_init(void *handle) if (!gmc_v9_0_keep_stolen_memory(adev)) amdgpu_bo_late_init(adev); - r = gmc_v9_0_allocate_vm_inv_eng(adev); + r = amdgpu_gmc_allocate_vm_inv_eng(adev); if (r) return r; /* Check if ecc is available */ @@ -1112,11 +1089,13 @@ static int gmc_v9_0_sw_init(void *handle) if (r) return r; - /* interrupt sent to DF. */ - r = amdgpu_irq_add_id(adev, SOC15_IH_CLIENTID_DF, 0, - &adev->gmc.ecc_irq); - if (r) - return r; + if (!amdgpu_sriov_vf(adev)) { + /* interrupt sent to DF. */ + r = amdgpu_irq_add_id(adev, SOC15_IH_CLIENTID_DF, 0, + &adev->gmc.ecc_irq); + if (r) + return r; + } /* Set the internal MC address mask * This is the max address of the GPU's @@ -1302,12 +1281,13 @@ static int gmc_v9_0_hw_init(void *handle) else value = true; - gfxhub_v1_0_set_fault_enable_default(adev, value); - if (adev->asic_type == CHIP_ARCTURUS) - mmhub_v9_4_set_fault_enable_default(adev, value); - else - mmhub_v1_0_set_fault_enable_default(adev, value); - + if (!amdgpu_sriov_vf(adev)) { + gfxhub_v1_0_set_fault_enable_default(adev, value); + if (adev->asic_type == CHIP_ARCTURUS) + mmhub_v9_4_set_fault_enable_default(adev, value); + else + mmhub_v1_0_set_fault_enable_default(adev, value); + } for (i = 0; i < adev->num_vmhubs; ++i) gmc_v9_0_flush_gpu_tlb(adev, 0, i, 0); diff --git a/drivers/gpu/drm/amd/amdgpu/gmc_v9_0.h b/drivers/gpu/drm/amd/amdgpu/gmc_v9_0.h index 49e8be761214..e0585e8c6c1b 100644 --- a/drivers/gpu/drm/amd/amdgpu/gmc_v9_0.h +++ b/drivers/gpu/drm/amd/amdgpu/gmc_v9_0.h @@ -24,16 +24,6 @@ #ifndef __GMC_V9_0_H__ #define __GMC_V9_0_H__ - /* - * The latest engine allocation on gfx9 is: - * Engine 2, 3: firmware - * Engine 0, 1, 4~16: amdgpu ring, - * subject to change when ring number changes - * Engine 17: Gart flushes - */ -#define GFXHUB_FREE_VM_INV_ENGS_BITMAP 0x1FFF3 -#define MMHUB_FREE_VM_INV_ENGS_BITMAP 0x1FFF3 - extern const struct amd_ip_funcs gmc_v9_0_ip_funcs; extern const struct amdgpu_ip_block_version gmc_v9_0_ip_block; #endif diff --git a/drivers/gpu/drm/amd/amdgpu/jpeg_v1_0.c b/drivers/gpu/drm/amd/amdgpu/jpeg_v1_0.c index a141408dfb23..0debfd9f428c 100644 --- a/drivers/gpu/drm/amd/amdgpu/jpeg_v1_0.c +++ b/drivers/gpu/drm/amd/amdgpu/jpeg_v1_0.c @@ -25,6 +25,7 @@ #include "amdgpu_jpeg.h" #include "soc15.h" #include "soc15d.h" +#include "vcn_v1_0.h" #include "vcn/vcn_1_0_offset.h" #include "vcn/vcn_1_0_sh_mask.h" @@ -561,7 +562,7 @@ static const struct amdgpu_ring_funcs jpeg_v1_0_decode_ring_vm_funcs = { .insert_start = jpeg_v1_0_decode_ring_insert_start, .insert_end = jpeg_v1_0_decode_ring_insert_end, .pad_ib = amdgpu_ring_generic_pad_ib, - .begin_use = amdgpu_vcn_ring_begin_use, + .begin_use = vcn_v1_0_ring_begin_use, .end_use = amdgpu_vcn_ring_end_use, .emit_wreg = jpeg_v1_0_decode_ring_emit_wreg, .emit_reg_wait = jpeg_v1_0_decode_ring_emit_reg_wait, diff --git a/drivers/gpu/drm/amd/amdgpu/mmhub_v9_4.c b/drivers/gpu/drm/amd/amdgpu/mmhub_v9_4.c index d9301e80522a..5c42387c9274 100644 --- a/drivers/gpu/drm/amd/amdgpu/mmhub_v9_4.c +++ b/drivers/gpu/drm/amd/amdgpu/mmhub_v9_4.c @@ -128,45 +128,53 @@ static void mmhub_v9_4_init_system_aperture_regs(struct amdgpu_device *adev, hubid * MMHUB_INSTANCE_REGISTER_OFFSET, adev->gmc.agp_start >> 24); - /* Program the system aperture low logical page number. */ - WREG32_SOC15_OFFSET(MMHUB, 0, - mmVMSHAREDVC0_MC_VM_SYSTEM_APERTURE_LOW_ADDR, - hubid * MMHUB_INSTANCE_REGISTER_OFFSET, - min(adev->gmc.fb_start, adev->gmc.agp_start) >> 18); - WREG32_SOC15_OFFSET(MMHUB, 0, - mmVMSHAREDVC0_MC_VM_SYSTEM_APERTURE_HIGH_ADDR, - hubid * MMHUB_INSTANCE_REGISTER_OFFSET, - max(adev->gmc.fb_end, adev->gmc.agp_end) >> 18); + if (!amdgpu_sriov_vf(adev)) { + /* Program the system aperture low logical page number. */ + WREG32_SOC15_OFFSET( + MMHUB, 0, mmVMSHAREDVC0_MC_VM_SYSTEM_APERTURE_LOW_ADDR, + hubid * MMHUB_INSTANCE_REGISTER_OFFSET, + min(adev->gmc.fb_start, adev->gmc.agp_start) >> 18); + WREG32_SOC15_OFFSET( + MMHUB, 0, mmVMSHAREDVC0_MC_VM_SYSTEM_APERTURE_HIGH_ADDR, + hubid * MMHUB_INSTANCE_REGISTER_OFFSET, + max(adev->gmc.fb_end, adev->gmc.agp_end) >> 18); - /* Set default page address. */ - value = adev->vram_scratch.gpu_addr - adev->gmc.vram_start + - adev->vm_manager.vram_base_offset; - WREG32_SOC15_OFFSET(MMHUB, 0, + /* Set default page address. */ + value = adev->vram_scratch.gpu_addr - adev->gmc.vram_start + + adev->vm_manager.vram_base_offset; + WREG32_SOC15_OFFSET( + MMHUB, 0, mmVMSHAREDPF0_MC_VM_SYSTEM_APERTURE_DEFAULT_ADDR_LSB, hubid * MMHUB_INSTANCE_REGISTER_OFFSET, (u32)(value >> 12)); - WREG32_SOC15_OFFSET(MMHUB, 0, + WREG32_SOC15_OFFSET( + MMHUB, 0, mmVMSHAREDPF0_MC_VM_SYSTEM_APERTURE_DEFAULT_ADDR_MSB, hubid * MMHUB_INSTANCE_REGISTER_OFFSET, (u32)(value >> 44)); - /* Program "protection fault". */ - WREG32_SOC15_OFFSET(MMHUB, 0, - mmVML2PF0_VM_L2_PROTECTION_FAULT_DEFAULT_ADDR_LO32, - hubid * MMHUB_INSTANCE_REGISTER_OFFSET, - (u32)(adev->dummy_page_addr >> 12)); - WREG32_SOC15_OFFSET(MMHUB, 0, - mmVML2PF0_VM_L2_PROTECTION_FAULT_DEFAULT_ADDR_HI32, - hubid * MMHUB_INSTANCE_REGISTER_OFFSET, - (u32)((u64)adev->dummy_page_addr >> 44)); + /* Program "protection fault". */ + WREG32_SOC15_OFFSET( + MMHUB, 0, + mmVML2PF0_VM_L2_PROTECTION_FAULT_DEFAULT_ADDR_LO32, + hubid * MMHUB_INSTANCE_REGISTER_OFFSET, + (u32)(adev->dummy_page_addr >> 12)); + WREG32_SOC15_OFFSET( + MMHUB, 0, + mmVML2PF0_VM_L2_PROTECTION_FAULT_DEFAULT_ADDR_HI32, + hubid * MMHUB_INSTANCE_REGISTER_OFFSET, + (u32)((u64)adev->dummy_page_addr >> 44)); - tmp = RREG32_SOC15_OFFSET(MMHUB, 0, - mmVML2PF0_VM_L2_PROTECTION_FAULT_CNTL2, - hubid * MMHUB_INSTANCE_REGISTER_OFFSET); - tmp = REG_SET_FIELD(tmp, VML2PF0_VM_L2_PROTECTION_FAULT_CNTL2, - ACTIVE_PAGE_MIGRATION_PTE_READ_RETRY, 1); - WREG32_SOC15_OFFSET(MMHUB, 0, mmVML2PF0_VM_L2_PROTECTION_FAULT_CNTL2, - hubid * MMHUB_INSTANCE_REGISTER_OFFSET, tmp); + tmp = RREG32_SOC15_OFFSET( + MMHUB, 0, mmVML2PF0_VM_L2_PROTECTION_FAULT_CNTL2, + hubid * MMHUB_INSTANCE_REGISTER_OFFSET); + tmp = REG_SET_FIELD(tmp, VML2PF0_VM_L2_PROTECTION_FAULT_CNTL2, + ACTIVE_PAGE_MIGRATION_PTE_READ_RETRY, 1); + WREG32_SOC15_OFFSET(MMHUB, 0, + mmVML2PF0_VM_L2_PROTECTION_FAULT_CNTL2, + hubid * MMHUB_INSTANCE_REGISTER_OFFSET, + tmp); + } } static void mmhub_v9_4_init_tlb_regs(struct amdgpu_device *adev, int hubid) @@ -368,30 +376,16 @@ int mmhub_v9_4_gart_enable(struct amdgpu_device *adev) int i; for (i = 0; i < MMHUB_NUM_INSTANCES; i++) { - if (amdgpu_sriov_vf(adev)) { - /* - * MC_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_OFFSET(MMHUB, 0, - mmVMSHAREDVC0_MC_VM_FB_LOCATION_BASE, - i * MMHUB_INSTANCE_REGISTER_OFFSET, - adev->gmc.vram_start >> 24); - WREG32_SOC15_OFFSET(MMHUB, 0, - mmVMSHAREDVC0_MC_VM_FB_LOCATION_TOP, - i * MMHUB_INSTANCE_REGISTER_OFFSET, - adev->gmc.vram_end >> 24); - } - /* GART Enable. */ mmhub_v9_4_init_gart_aperture_regs(adev, i); mmhub_v9_4_init_system_aperture_regs(adev, i); mmhub_v9_4_init_tlb_regs(adev, i); - mmhub_v9_4_init_cache_regs(adev, i); + if (!amdgpu_sriov_vf(adev)) + mmhub_v9_4_init_cache_regs(adev, i); mmhub_v9_4_enable_system_domain(adev, i); - mmhub_v9_4_disable_identity_aperture(adev, i); + if (!amdgpu_sriov_vf(adev)) + mmhub_v9_4_disable_identity_aperture(adev, i); mmhub_v9_4_setup_vmid_config(adev, i); mmhub_v9_4_program_invalidation(adev, i); } diff --git a/drivers/gpu/drm/amd/amdgpu/mmsch_v1_0.h b/drivers/gpu/drm/amd/amdgpu/mmsch_v1_0.h index 8af0bddf85e4..20958639b601 100644 --- a/drivers/gpu/drm/amd/amdgpu/mmsch_v1_0.h +++ b/drivers/gpu/drm/amd/amdgpu/mmsch_v1_0.h @@ -47,6 +47,18 @@ struct mmsch_v1_0_init_header { uint32_t uvd_table_size; }; +struct mmsch_vf_eng_init_header { + uint32_t init_status; + uint32_t table_offset; + uint32_t table_size; +}; + +struct mmsch_v1_1_init_header { + uint32_t version; + uint32_t total_size; + struct mmsch_vf_eng_init_header eng[2]; +}; + struct mmsch_v1_0_cmd_direct_reg_header { uint32_t reg_offset : 28; uint32_t command_type : 4; diff --git a/drivers/gpu/drm/amd/amdgpu/mxgpu_ai.c b/drivers/gpu/drm/amd/amdgpu/mxgpu_ai.c index 43305afa3d6f..5fd67e1cc2a0 100644 --- a/drivers/gpu/drm/amd/amdgpu/mxgpu_ai.c +++ b/drivers/gpu/drm/amd/amdgpu/mxgpu_ai.c @@ -250,7 +250,7 @@ static void xgpu_ai_mailbox_flr_work(struct work_struct *work) */ locked = mutex_trylock(&adev->lock_reset); if (locked) - adev->in_gpu_reset = 1; + adev->in_gpu_reset = true; do { if (xgpu_ai_mailbox_peek_msg(adev) == IDH_FLR_NOTIFICATION_CMPL) @@ -262,7 +262,7 @@ static void xgpu_ai_mailbox_flr_work(struct work_struct *work) flr_done: if (locked) { - adev->in_gpu_reset = 0; + adev->in_gpu_reset = false; mutex_unlock(&adev->lock_reset); } diff --git a/drivers/gpu/drm/amd/amdgpu/mxgpu_nv.c b/drivers/gpu/drm/amd/amdgpu/mxgpu_nv.c index 0d8767eb7a70..237fa5e16b7c 100644 --- a/drivers/gpu/drm/amd/amdgpu/mxgpu_nv.c +++ b/drivers/gpu/drm/amd/amdgpu/mxgpu_nv.c @@ -252,7 +252,7 @@ static void xgpu_nv_mailbox_flr_work(struct work_struct *work) */ locked = mutex_trylock(&adev->lock_reset); if (locked) - adev->in_gpu_reset = 1; + adev->in_gpu_reset = true; do { if (xgpu_nv_mailbox_peek_msg(adev) == IDH_FLR_NOTIFICATION_CMPL) @@ -264,12 +264,16 @@ static void xgpu_nv_mailbox_flr_work(struct work_struct *work) flr_done: if (locked) { - adev->in_gpu_reset = 0; + adev->in_gpu_reset = false; mutex_unlock(&adev->lock_reset); } /* Trigger recovery for world switch failure if no TDR */ - if (amdgpu_device_should_recover_gpu(adev)) + if (amdgpu_device_should_recover_gpu(adev) + && (adev->sdma_timeout == MAX_SCHEDULE_TIMEOUT || + adev->gfx_timeout == MAX_SCHEDULE_TIMEOUT || + adev->compute_timeout == MAX_SCHEDULE_TIMEOUT || + adev->video_timeout == MAX_SCHEDULE_TIMEOUT)) amdgpu_device_gpu_recover(adev, NULL); } diff --git a/drivers/gpu/drm/amd/amdgpu/navi10_ih.c b/drivers/gpu/drm/amd/amdgpu/navi10_ih.c index 9af73567e716..f737ce459c28 100644 --- a/drivers/gpu/drm/amd/amdgpu/navi10_ih.c +++ b/drivers/gpu/drm/amd/amdgpu/navi10_ih.c @@ -110,7 +110,6 @@ static uint32_t navi10_ih_rb_cntl(struct amdgpu_ih_ring *ih, uint32_t ih_rb_cntl static int navi10_ih_irq_init(struct amdgpu_device *adev) { struct amdgpu_ih_ring *ih = &adev->irq.ih; - int ret = 0; u32 ih_rb_cntl, ih_doorbell_rtpr, ih_chicken; u32 tmp; @@ -179,7 +178,7 @@ static int navi10_ih_irq_init(struct amdgpu_device *adev) /* enable interrupts */ navi10_ih_enable_interrupts(adev); - return ret; + return 0; } /** diff --git a/drivers/gpu/drm/amd/amdgpu/nbio_v7_4.c b/drivers/gpu/drm/amd/amdgpu/nbio_v7_4.c index bb701dbfd472..65eb378fa035 100644 --- a/drivers/gpu/drm/amd/amdgpu/nbio_v7_4.c +++ b/drivers/gpu/drm/amd/amdgpu/nbio_v7_4.c @@ -339,7 +339,7 @@ static void nbio_v7_4_handle_ras_controller_intr_no_bifring(struct amdgpu_device /* ras_controller_int is dedicated for nbif ras error, * not the global interrupt for sync flood */ - amdgpu_ras_reset_gpu(adev, true); + amdgpu_ras_reset_gpu(adev); } } @@ -456,10 +456,8 @@ static int nbio_v7_4_init_ras_controller_interrupt (struct amdgpu_device *adev) r = amdgpu_irq_add_id(adev, SOC15_IH_CLIENTID_BIF, NBIF_7_4__SRCID__RAS_CONTROLLER_INTERRUPT, &adev->nbio.ras_controller_irq); - if (r) - return r; - return 0; + return r; } static int nbio_v7_4_init_ras_err_event_athub_interrupt (struct amdgpu_device *adev) @@ -476,10 +474,8 @@ static int nbio_v7_4_init_ras_err_event_athub_interrupt (struct amdgpu_device *a r = amdgpu_irq_add_id(adev, SOC15_IH_CLIENTID_BIF, NBIF_7_4__SRCID__ERREVENT_ATHUB_INTERRUPT, &adev->nbio.ras_err_event_athub_irq); - if (r) - return r; - return 0; + return r; } #define smnPARITY_ERROR_STATUS_UNCORR_GRP2 0x13a20030 diff --git a/drivers/gpu/drm/amd/amdgpu/psp_gfx_if.h b/drivers/gpu/drm/amd/amdgpu/psp_gfx_if.h index 74a9fe8e0cfb..36b65797434e 100644 --- a/drivers/gpu/drm/amd/amdgpu/psp_gfx_if.h +++ b/drivers/gpu/drm/amd/amdgpu/psp_gfx_if.h @@ -242,6 +242,7 @@ enum psp_gfx_fw_type { GFX_FW_TYPE_SDMA5 = 55, /* SDMA5 MI */ GFX_FW_TYPE_SDMA6 = 56, /* SDMA6 MI */ GFX_FW_TYPE_SDMA7 = 57, /* SDMA7 MI */ + GFX_FW_TYPE_VCN1 = 58, /* VCN1 MI */ GFX_FW_TYPE_MAX }; diff --git a/drivers/gpu/drm/amd/amdgpu/psp_v11_0.c b/drivers/gpu/drm/amd/amdgpu/psp_v11_0.c index c66ca8cc2ebd..a57f3d737677 100644 --- a/drivers/gpu/drm/amd/amdgpu/psp_v11_0.c +++ b/drivers/gpu/drm/amd/amdgpu/psp_v11_0.c @@ -233,6 +233,29 @@ out: return err; } +int psp_v11_0_wait_for_bootloader(struct psp_context *psp) +{ + struct amdgpu_device *adev = psp->adev; + + int ret; + int retry_loop; + + for (retry_loop = 0; retry_loop < 10; retry_loop++) { + /* Wait for bootloader to signify that is + ready having bit 31 of C2PMSG_35 set to 1 */ + ret = psp_wait_for(psp, + SOC15_REG_OFFSET(MP0, 0, mmMP0_SMN_C2PMSG_35), + 0x80000000, + 0x80000000, + false); + + if (ret == 0) + return 0; + } + + return ret; +} + static bool psp_v11_0_is_sos_alive(struct psp_context *psp) { struct amdgpu_device *adev = psp->adev; @@ -258,9 +281,7 @@ static int psp_v11_0_bootloader_load_kdb(struct psp_context *psp) return 0; } - /* Wait for bootloader to signify that is ready having bit 31 of C2PMSG_35 set to 1 */ - ret = psp_wait_for(psp, SOC15_REG_OFFSET(MP0, 0, mmMP0_SMN_C2PMSG_35), - 0x80000000, 0x80000000, false); + ret = psp_v11_0_wait_for_bootloader(psp); if (ret) return ret; @@ -276,9 +297,7 @@ static int psp_v11_0_bootloader_load_kdb(struct psp_context *psp) WREG32_SOC15(MP0, 0, mmMP0_SMN_C2PMSG_35, psp_gfxdrv_command_reg); - /* Wait for bootloader to signify that is ready having bit 31 of C2PMSG_35 set to 1*/ - ret = psp_wait_for(psp, SOC15_REG_OFFSET(MP0, 0, mmMP0_SMN_C2PMSG_35), - 0x80000000, 0x80000000, false); + ret = psp_v11_0_wait_for_bootloader(psp); return ret; } @@ -298,9 +317,7 @@ static int psp_v11_0_bootloader_load_sysdrv(struct psp_context *psp) return 0; } - /* Wait for bootloader to signify that is ready having bit 31 of C2PMSG_35 set to 1 */ - ret = psp_wait_for(psp, SOC15_REG_OFFSET(MP0, 0, mmMP0_SMN_C2PMSG_35), - 0x80000000, 0x80000000, false); + ret = psp_v11_0_wait_for_bootloader(psp); if (ret) return ret; @@ -319,8 +336,7 @@ static int psp_v11_0_bootloader_load_sysdrv(struct psp_context *psp) /* there might be handshake issue with hardware which needs delay */ mdelay(20); - ret = psp_wait_for(psp, SOC15_REG_OFFSET(MP0, 0, mmMP0_SMN_C2PMSG_35), - 0x80000000, 0x80000000, false); + ret = psp_v11_0_wait_for_bootloader(psp); return ret; } @@ -337,9 +353,7 @@ static int psp_v11_0_bootloader_load_sos(struct psp_context *psp) if (psp_v11_0_is_sos_alive(psp)) return 0; - /* Wait for bootloader to signify that is ready having bit 31 of C2PMSG_35 set to 1 */ - ret = psp_wait_for(psp, SOC15_REG_OFFSET(MP0, 0, mmMP0_SMN_C2PMSG_35), - 0x80000000, 0x80000000, false); + ret = psp_v11_0_wait_for_bootloader(psp); if (ret) return ret; diff --git a/drivers/gpu/drm/amd/amdgpu/sdma_v2_4.c b/drivers/gpu/drm/amd/amdgpu/sdma_v2_4.c index a10175838013..7d509a40076f 100644 --- a/drivers/gpu/drm/amd/amdgpu/sdma_v2_4.c +++ b/drivers/gpu/drm/amd/amdgpu/sdma_v2_4.c @@ -255,7 +255,7 @@ static void sdma_v2_4_ring_emit_ib(struct amdgpu_ring *ring, unsigned vmid = AMDGPU_JOB_GET_VMID(job); /* IB packet must end on a 8 DW boundary */ - sdma_v2_4_ring_insert_nop(ring, (10 - (lower_32_bits(ring->wptr) & 7)) % 8); + sdma_v2_4_ring_insert_nop(ring, (2 - lower_32_bits(ring->wptr)) & 7); amdgpu_ring_write(ring, SDMA_PKT_HEADER_OP(SDMA_OP_INDIRECT) | SDMA_PKT_INDIRECT_HEADER_VMID(vmid & 0xf)); @@ -750,7 +750,7 @@ static void sdma_v2_4_ring_pad_ib(struct amdgpu_ring *ring, struct amdgpu_ib *ib u32 pad_count; int i; - pad_count = (8 - (ib->length_dw & 0x7)) % 8; + pad_count = (-ib->length_dw) & 7; for (i = 0; i < pad_count; i++) if (sdma && sdma->burst_nop && (i == 0)) ib->ptr[ib->length_dw++] = @@ -1260,16 +1260,14 @@ static const struct amdgpu_vm_pte_funcs sdma_v2_4_vm_pte_funcs = { static void sdma_v2_4_set_vm_pte_funcs(struct amdgpu_device *adev) { - struct drm_gpu_scheduler *sched; unsigned i; adev->vm_manager.vm_pte_funcs = &sdma_v2_4_vm_pte_funcs; for (i = 0; i < adev->sdma.num_instances; i++) { - sched = &adev->sdma.instance[i].ring.sched; - adev->vm_manager.vm_pte_rqs[i] = - &sched->sched_rq[DRM_SCHED_PRIORITY_KERNEL]; + adev->vm_manager.vm_pte_scheds[i] = + &adev->sdma.instance[i].ring.sched; } - adev->vm_manager.vm_pte_num_rqs = adev->sdma.num_instances; + adev->vm_manager.vm_pte_num_scheds = adev->sdma.num_instances; } const struct amdgpu_ip_block_version sdma_v2_4_ip_block = diff --git a/drivers/gpu/drm/amd/amdgpu/sdma_v3_0.c b/drivers/gpu/drm/amd/amdgpu/sdma_v3_0.c index 5f4e2c616241..b6109a99fc43 100644 --- a/drivers/gpu/drm/amd/amdgpu/sdma_v3_0.c +++ b/drivers/gpu/drm/amd/amdgpu/sdma_v3_0.c @@ -429,7 +429,7 @@ static void sdma_v3_0_ring_emit_ib(struct amdgpu_ring *ring, unsigned vmid = AMDGPU_JOB_GET_VMID(job); /* IB packet must end on a 8 DW boundary */ - sdma_v3_0_ring_insert_nop(ring, (10 - (lower_32_bits(ring->wptr) & 7)) % 8); + sdma_v3_0_ring_insert_nop(ring, (2 - lower_32_bits(ring->wptr)) & 7); amdgpu_ring_write(ring, SDMA_PKT_HEADER_OP(SDMA_OP_INDIRECT) | SDMA_PKT_INDIRECT_HEADER_VMID(vmid & 0xf)); @@ -1021,7 +1021,7 @@ static void sdma_v3_0_ring_pad_ib(struct amdgpu_ring *ring, struct amdgpu_ib *ib u32 pad_count; int i; - pad_count = (8 - (ib->length_dw & 0x7)) % 8; + pad_count = (-ib->length_dw) & 7; for (i = 0; i < pad_count; i++) if (sdma && sdma->burst_nop && (i == 0)) ib->ptr[ib->length_dw++] = @@ -1698,16 +1698,14 @@ static const struct amdgpu_vm_pte_funcs sdma_v3_0_vm_pte_funcs = { static void sdma_v3_0_set_vm_pte_funcs(struct amdgpu_device *adev) { - struct drm_gpu_scheduler *sched; unsigned i; adev->vm_manager.vm_pte_funcs = &sdma_v3_0_vm_pte_funcs; for (i = 0; i < adev->sdma.num_instances; i++) { - sched = &adev->sdma.instance[i].ring.sched; - adev->vm_manager.vm_pte_rqs[i] = - &sched->sched_rq[DRM_SCHED_PRIORITY_KERNEL]; + adev->vm_manager.vm_pte_scheds[i] = + &adev->sdma.instance[i].ring.sched; } - adev->vm_manager.vm_pte_num_rqs = adev->sdma.num_instances; + adev->vm_manager.vm_pte_num_scheds = adev->sdma.num_instances; } const struct amdgpu_ip_block_version sdma_v3_0_ip_block = diff --git a/drivers/gpu/drm/amd/amdgpu/sdma_v4_0.c b/drivers/gpu/drm/amd/amdgpu/sdma_v4_0.c index 4ef4d31f5231..ce0753a9d241 100644 --- a/drivers/gpu/drm/amd/amdgpu/sdma_v4_0.c +++ b/drivers/gpu/drm/amd/amdgpu/sdma_v4_0.c @@ -698,7 +698,7 @@ static void sdma_v4_0_ring_emit_ib(struct amdgpu_ring *ring, unsigned vmid = AMDGPU_JOB_GET_VMID(job); /* IB packet must end on a 8 DW boundary */ - sdma_v4_0_ring_insert_nop(ring, (10 - (lower_32_bits(ring->wptr) & 7)) % 8); + sdma_v4_0_ring_insert_nop(ring, (2 - lower_32_bits(ring->wptr)) & 7); amdgpu_ring_write(ring, SDMA_PKT_HEADER_OP(SDMA_OP_INDIRECT) | SDMA_PKT_INDIRECT_HEADER_VMID(vmid & 0xf)); @@ -1579,7 +1579,7 @@ static void sdma_v4_0_ring_pad_ib(struct amdgpu_ring *ring, struct amdgpu_ib *ib u32 pad_count; int i; - pad_count = (8 - (ib->length_dw & 0x7)) % 8; + pad_count = (-ib->length_dw) & 7; for (i = 0; i < pad_count; i++) if (sdma && sdma->burst_nop && (i == 0)) ib->ptr[ib->length_dw++] = @@ -2409,10 +2409,9 @@ static void sdma_v4_0_set_vm_pte_funcs(struct amdgpu_device *adev) sched = &adev->sdma.instance[i].page.sched; else sched = &adev->sdma.instance[i].ring.sched; - adev->vm_manager.vm_pte_rqs[i] = - &sched->sched_rq[DRM_SCHED_PRIORITY_KERNEL]; + adev->vm_manager.vm_pte_scheds[i] = sched; } - adev->vm_manager.vm_pte_num_rqs = adev->sdma.num_instances; + adev->vm_manager.vm_pte_num_scheds = adev->sdma.num_instances; } const struct amdgpu_ip_block_version sdma_v4_0_ip_block = { diff --git a/drivers/gpu/drm/amd/amdgpu/sdma_v5_0.c b/drivers/gpu/drm/amd/amdgpu/sdma_v5_0.c index f4ad2990f973..4c6bf1f8a528 100644 --- a/drivers/gpu/drm/amd/amdgpu/sdma_v5_0.c +++ b/drivers/gpu/drm/amd/amdgpu/sdma_v5_0.c @@ -382,8 +382,15 @@ static void sdma_v5_0_ring_emit_ib(struct amdgpu_ring *ring, unsigned vmid = AMDGPU_JOB_GET_VMID(job); uint64_t csa_mc_addr = amdgpu_sdma_get_csa_mc_addr(ring, vmid); - /* IB packet must end on a 8 DW boundary */ - sdma_v5_0_ring_insert_nop(ring, (10 - (lower_32_bits(ring->wptr) & 7)) % 8); + /* An IB packet must end on a 8 DW boundary--the next dword + * must be on a 8-dword boundary. Our IB packet below is 6 + * dwords long, thus add x number of NOPs, such that, in + * modular arithmetic, + * wptr + 6 + x = 8k, k >= 0, which in C is, + * (wptr + 6 + x) % 8 = 0. + * The expression below, is a solution of x. + */ + sdma_v5_0_ring_insert_nop(ring, (2 - lower_32_bits(ring->wptr)) & 7); amdgpu_ring_write(ring, SDMA_PKT_HEADER_OP(SDMA_OP_INDIRECT) | SDMA_PKT_INDIRECT_HEADER_VMID(vmid & 0xf)); @@ -907,16 +914,9 @@ static int sdma_v5_0_ring_test_ring(struct amdgpu_ring *ring) udelay(1); } - if (i < adev->usec_timeout) { - if (amdgpu_emu_mode == 1) - DRM_INFO("ring test on %d succeeded in %d msecs\n", ring->idx, i); - else - DRM_INFO("ring test on %d succeeded in %d usecs\n", ring->idx, i); - } else { - DRM_ERROR("amdgpu: ring %d test failed (0x%08X)\n", - ring->idx, tmp); - r = -EINVAL; - } + if (i >= adev->usec_timeout) + r = -ETIMEDOUT; + amdgpu_device_wb_free(adev, index); return r; @@ -981,13 +981,10 @@ static int sdma_v5_0_ring_test_ib(struct amdgpu_ring *ring, long timeout) goto err1; } tmp = le32_to_cpu(adev->wb.wb[index]); - if (tmp == 0xDEADBEEF) { - DRM_INFO("ib test on ring %d succeeded\n", ring->idx); + if (tmp == 0xDEADBEEF) r = 0; - } else { - DRM_ERROR("amdgpu: ib test failed (0x%08X)\n", tmp); + else r = -EINVAL; - } err1: amdgpu_ib_free(adev, &ib, NULL); @@ -1086,10 +1083,10 @@ static void sdma_v5_0_vm_set_pte_pde(struct amdgpu_ib *ib, } /** - * sdma_v5_0_ring_pad_ib - pad the IB to the required number of dw - * + * sdma_v5_0_ring_pad_ib - pad the IB * @ib: indirect buffer to fill with padding * + * Pad the IB with NOPs to a boundary multiple of 8. */ static void sdma_v5_0_ring_pad_ib(struct amdgpu_ring *ring, struct amdgpu_ib *ib) { @@ -1097,7 +1094,7 @@ static void sdma_v5_0_ring_pad_ib(struct amdgpu_ring *ring, struct amdgpu_ib *ib u32 pad_count; int i; - pad_count = (8 - (ib->length_dw & 0x7)) % 8; + pad_count = (-ib->length_dw) & 0x7; for (i = 0; i < pad_count; i++) if (sdma && sdma->burst_nop && (i == 0)) ib->ptr[ib->length_dw++] = @@ -1721,17 +1718,15 @@ static const struct amdgpu_vm_pte_funcs sdma_v5_0_vm_pte_funcs = { static void sdma_v5_0_set_vm_pte_funcs(struct amdgpu_device *adev) { - struct drm_gpu_scheduler *sched; unsigned i; if (adev->vm_manager.vm_pte_funcs == NULL) { adev->vm_manager.vm_pte_funcs = &sdma_v5_0_vm_pte_funcs; for (i = 0; i < adev->sdma.num_instances; i++) { - sched = &adev->sdma.instance[i].ring.sched; - adev->vm_manager.vm_pte_rqs[i] = - &sched->sched_rq[DRM_SCHED_PRIORITY_KERNEL]; + adev->vm_manager.vm_pte_scheds[i] = + &adev->sdma.instance[i].ring.sched; } - adev->vm_manager.vm_pte_num_rqs = adev->sdma.num_instances; + adev->vm_manager.vm_pte_num_scheds = adev->sdma.num_instances; } } diff --git a/drivers/gpu/drm/amd/amdgpu/si_dma.c b/drivers/gpu/drm/amd/amdgpu/si_dma.c index bdda8b4e03f0..9aac9f9c50bb 100644 --- a/drivers/gpu/drm/amd/amdgpu/si_dma.c +++ b/drivers/gpu/drm/amd/amdgpu/si_dma.c @@ -834,16 +834,14 @@ static const struct amdgpu_vm_pte_funcs si_dma_vm_pte_funcs = { static void si_dma_set_vm_pte_funcs(struct amdgpu_device *adev) { - struct drm_gpu_scheduler *sched; unsigned i; adev->vm_manager.vm_pte_funcs = &si_dma_vm_pte_funcs; for (i = 0; i < adev->sdma.num_instances; i++) { - sched = &adev->sdma.instance[i].ring.sched; - adev->vm_manager.vm_pte_rqs[i] = - &sched->sched_rq[DRM_SCHED_PRIORITY_KERNEL]; + adev->vm_manager.vm_pte_scheds[i] = + &adev->sdma.instance[i].ring.sched; } - adev->vm_manager.vm_pte_num_rqs = adev->sdma.num_instances; + adev->vm_manager.vm_pte_num_scheds = adev->sdma.num_instances; } const struct amdgpu_ip_block_version si_dma_ip_block = diff --git a/drivers/gpu/drm/amd/amdgpu/soc15.c b/drivers/gpu/drm/amd/amdgpu/soc15.c index 5bd6ae7a52fe..714cf4dfd0a7 100644 --- a/drivers/gpu/drm/amd/amdgpu/soc15.c +++ b/drivers/gpu/drm/amd/amdgpu/soc15.c @@ -613,6 +613,7 @@ static bool soc15_supports_baco(struct amdgpu_device *adev) switch (adev->asic_type) { case CHIP_VEGA10: case CHIP_VEGA12: + case CHIP_ARCTURUS: soc15_asic_get_baco_capability(adev, &baco_support); break; case CHIP_VEGA20: @@ -827,11 +828,15 @@ int soc15_set_ip_blocks(struct amdgpu_device *adev) amdgpu_device_ip_block_add(adev, &dce_virtual_ip_block); amdgpu_device_ip_block_add(adev, &gfx_v9_0_ip_block); amdgpu_device_ip_block_add(adev, &sdma_v4_0_ip_block); - if (!amdgpu_sriov_vf(adev)) - amdgpu_device_ip_block_add(adev, &smu_v11_0_ip_block); + amdgpu_device_ip_block_add(adev, &smu_v11_0_ip_block); - if (unlikely(adev->firmware.load_type == AMDGPU_FW_LOAD_DIRECT)) - amdgpu_device_ip_block_add(adev, &vcn_v2_5_ip_block); + if (amdgpu_sriov_vf(adev)) { + if (likely(adev->firmware.load_type == AMDGPU_FW_LOAD_PSP)) + amdgpu_device_ip_block_add(adev, &vcn_v2_5_ip_block); + } else { + if (unlikely(adev->firmware.load_type == AMDGPU_FW_LOAD_DIRECT)) + amdgpu_device_ip_block_add(adev, &vcn_v2_5_ip_block); + } if (!amdgpu_sriov_vf(adev)) amdgpu_device_ip_block_add(adev, &jpeg_v2_5_ip_block); break; diff --git a/drivers/gpu/drm/amd/amdgpu/umc_v6_1.c b/drivers/gpu/drm/amd/amdgpu/umc_v6_1.c index 515eb50cd0f8..11e924dd88ff 100644 --- a/drivers/gpu/drm/amd/amdgpu/umc_v6_1.c +++ b/drivers/gpu/drm/amd/amdgpu/umc_v6_1.c @@ -28,27 +28,26 @@ #include "rsmu/rsmu_0_0_2_sh_mask.h" #include "umc/umc_6_1_1_offset.h" #include "umc/umc_6_1_1_sh_mask.h" +#include "umc/umc_6_1_2_offset.h" #define smnMCA_UMC0_MCUMC_ADDRT0 0x50f10 -/* UMC 6_1_2 register offsets */ -#define mmUMCCH0_0_EccErrCntSel_ARCT 0x0360 -#define mmUMCCH0_0_EccErrCntSel_ARCT_BASE_IDX 1 -#define mmUMCCH0_0_EccErrCnt_ARCT 0x0361 -#define mmUMCCH0_0_EccErrCnt_ARCT_BASE_IDX 1 -#define mmMCA_UMC_UMC0_MCUMC_STATUST0_ARCT 0x03c2 -#define mmMCA_UMC_UMC0_MCUMC_STATUST0_ARCT_BASE_IDX 1 +#define UMC_6_INST_DIST 0x40000 /* * (addr / 256) * 8192, the higher 26 bits in ErrorAddr * is the index of 8KB block */ -#define ADDR_OF_8KB_BLOCK(addr) (((addr) & ~0xffULL) << 5) +#define ADDR_OF_8KB_BLOCK(addr) (((addr) & ~0xffULL) << 5) /* channel index is the index of 256B block */ #define ADDR_OF_256B_BLOCK(channel_index) ((channel_index) << 8) /* offset in 256B block */ #define OFFSET_IN_256B_BLOCK(addr) ((addr) & 0xffULL) +#define LOOP_UMC_INST(umc_inst) for ((umc_inst) = 0; (umc_inst) < adev->umc.umc_inst_num; (umc_inst)++) +#define LOOP_UMC_CH_INST(ch_inst) for ((ch_inst) = 0; (ch_inst) < adev->umc.channel_inst_num; (ch_inst)++) +#define LOOP_UMC_INST_AND_CH(umc_inst, ch_inst) LOOP_UMC_INST((umc_inst)) LOOP_UMC_CH_INST((ch_inst)) + const uint32_t umc_v6_1_channel_idx_tbl[UMC_V6_1_UMC_INSTANCE_NUM][UMC_V6_1_CHANNEL_INSTANCE_NUM] = { {2, 18, 11, 27}, {4, 20, 13, 29}, @@ -57,41 +56,17 @@ const uint32_t {9, 25, 0, 16}, {15, 31, 6, 22} }; -static void umc_v6_1_enable_umc_index_mode(struct amdgpu_device *adev, - uint32_t umc_instance) -{ - uint32_t rsmu_umc_index; - - rsmu_umc_index = RREG32_SOC15(RSMU, 0, - mmRSMU_UMC_INDEX_REGISTER_NBIF_VG20_GPU); - rsmu_umc_index = REG_SET_FIELD(rsmu_umc_index, - RSMU_UMC_INDEX_REGISTER_NBIF_VG20_GPU, - RSMU_UMC_INDEX_MODE_EN, 1); - rsmu_umc_index = REG_SET_FIELD(rsmu_umc_index, - RSMU_UMC_INDEX_REGISTER_NBIF_VG20_GPU, - RSMU_UMC_INDEX_INSTANCE, umc_instance); - rsmu_umc_index = REG_SET_FIELD(rsmu_umc_index, - RSMU_UMC_INDEX_REGISTER_NBIF_VG20_GPU, - RSMU_UMC_INDEX_WREN, 1 << umc_instance); - WREG32_SOC15(RSMU, 0, mmRSMU_UMC_INDEX_REGISTER_NBIF_VG20_GPU, - rsmu_umc_index); -} - static void umc_v6_1_disable_umc_index_mode(struct amdgpu_device *adev) { WREG32_FIELD15(RSMU, 0, RSMU_UMC_INDEX_REGISTER_NBIF_VG20_GPU, RSMU_UMC_INDEX_MODE_EN, 0); } -static uint32_t umc_v6_1_get_umc_inst(struct amdgpu_device *adev) +static inline uint32_t get_umc_6_reg_offset(struct amdgpu_device *adev, + uint32_t umc_inst, + uint32_t ch_inst) { - uint32_t rsmu_umc_index; - - rsmu_umc_index = RREG32_SOC15(RSMU, 0, - mmRSMU_UMC_INDEX_REGISTER_NBIF_VG20_GPU); - return REG_GET_FIELD(rsmu_umc_index, - RSMU_UMC_INDEX_REGISTER_NBIF_VG20_GPU, - RSMU_UMC_INDEX_INSTANCE); + return adev->umc.channel_offs*ch_inst + UMC_6_INST_DIST*umc_inst; } static void umc_v6_1_query_correctable_error_count(struct amdgpu_device *adev, @@ -105,7 +80,6 @@ static void umc_v6_1_query_correctable_error_count(struct amdgpu_device *adev, if (adev->asic_type == CHIP_ARCTURUS) { /* UMC 6_1_2 registers */ - ecc_err_cnt_sel_addr = SOC15_REG_OFFSET(UMC, 0, mmUMCCH0_0_EccErrCntSel_ARCT); ecc_err_cnt_addr = @@ -114,7 +88,6 @@ static void umc_v6_1_query_correctable_error_count(struct amdgpu_device *adev, SOC15_REG_OFFSET(UMC, 0, mmMCA_UMC_UMC0_MCUMC_STATUST0_ARCT); } else { /* UMC 6_1_1 registers */ - ecc_err_cnt_sel_addr = SOC15_REG_OFFSET(UMC, 0, mmUMCCH0_0_EccErrCntSel); ecc_err_cnt_addr = @@ -124,31 +97,31 @@ static void umc_v6_1_query_correctable_error_count(struct amdgpu_device *adev, } /* select the lower chip and check the error count */ - ecc_err_cnt_sel = RREG32(ecc_err_cnt_sel_addr + umc_reg_offset); + ecc_err_cnt_sel = RREG32_PCIE((ecc_err_cnt_sel_addr + umc_reg_offset) * 4); ecc_err_cnt_sel = REG_SET_FIELD(ecc_err_cnt_sel, UMCCH0_0_EccErrCntSel, EccErrCntCsSel, 0); - WREG32(ecc_err_cnt_sel_addr + umc_reg_offset, ecc_err_cnt_sel); - ecc_err_cnt = RREG32(ecc_err_cnt_addr + umc_reg_offset); + WREG32_PCIE((ecc_err_cnt_sel_addr + umc_reg_offset) * 4, ecc_err_cnt_sel); + ecc_err_cnt = RREG32_PCIE((ecc_err_cnt_addr + umc_reg_offset) * 4); *error_count += (REG_GET_FIELD(ecc_err_cnt, UMCCH0_0_EccErrCnt, EccErrCnt) - UMC_V6_1_CE_CNT_INIT); /* clear the lower chip err count */ - WREG32(ecc_err_cnt_addr + umc_reg_offset, UMC_V6_1_CE_CNT_INIT); + WREG32_PCIE((ecc_err_cnt_addr + umc_reg_offset) * 4, UMC_V6_1_CE_CNT_INIT); /* select the higher chip and check the err counter */ ecc_err_cnt_sel = REG_SET_FIELD(ecc_err_cnt_sel, UMCCH0_0_EccErrCntSel, EccErrCntCsSel, 1); - WREG32(ecc_err_cnt_sel_addr + umc_reg_offset, ecc_err_cnt_sel); - ecc_err_cnt = RREG32(ecc_err_cnt_addr + umc_reg_offset); + WREG32_PCIE((ecc_err_cnt_sel_addr + umc_reg_offset) * 4, ecc_err_cnt_sel); + ecc_err_cnt = RREG32_PCIE((ecc_err_cnt_addr + umc_reg_offset) * 4); *error_count += (REG_GET_FIELD(ecc_err_cnt, UMCCH0_0_EccErrCnt, EccErrCnt) - UMC_V6_1_CE_CNT_INIT); /* clear the higher chip err count */ - WREG32(ecc_err_cnt_addr + umc_reg_offset, UMC_V6_1_CE_CNT_INIT); + WREG32_PCIE((ecc_err_cnt_addr + umc_reg_offset) * 4, UMC_V6_1_CE_CNT_INIT); /* check for SRAM correctable error MCUMC_STATUS is a 64 bit register */ - mc_umc_status = RREG64_UMC(mc_umc_status_addr + umc_reg_offset); + mc_umc_status = RREG64_PCIE((mc_umc_status_addr + umc_reg_offset) * 4); if (REG_GET_FIELD(mc_umc_status, MCA_UMC_UMC0_MCUMC_STATUST0, ErrorCodeExt) == 6 && REG_GET_FIELD(mc_umc_status, MCA_UMC_UMC0_MCUMC_STATUST0, Val) == 1 && REG_GET_FIELD(mc_umc_status, MCA_UMC_UMC0_MCUMC_STATUST0, CECC) == 1) @@ -164,18 +137,16 @@ static void umc_v6_1_querry_uncorrectable_error_count(struct amdgpu_device *adev if (adev->asic_type == CHIP_ARCTURUS) { /* UMC 6_1_2 registers */ - mc_umc_status_addr = SOC15_REG_OFFSET(UMC, 0, mmMCA_UMC_UMC0_MCUMC_STATUST0_ARCT); } else { /* UMC 6_1_1 registers */ - mc_umc_status_addr = SOC15_REG_OFFSET(UMC, 0, mmMCA_UMC_UMC0_MCUMC_STATUST0); } /* check the MCUMC_STATUS */ - mc_umc_status = RREG64_UMC(mc_umc_status_addr + umc_reg_offset); + mc_umc_status = RREG64_PCIE((mc_umc_status_addr + umc_reg_offset) * 4); if ((REG_GET_FIELD(mc_umc_status, MCA_UMC_UMC0_MCUMC_STATUST0, Val) == 1) && (REG_GET_FIELD(mc_umc_status, MCA_UMC_UMC0_MCUMC_STATUST0, Deferred) == 1 || REG_GET_FIELD(mc_umc_status, MCA_UMC_UMC0_MCUMC_STATUST0, UECC) == 1 || @@ -185,38 +156,46 @@ static void umc_v6_1_querry_uncorrectable_error_count(struct amdgpu_device *adev *error_count += 1; } -static void umc_v6_1_query_error_count(struct amdgpu_device *adev, - struct ras_err_data *err_data, uint32_t umc_reg_offset, - uint32_t channel_index) -{ - umc_v6_1_query_correctable_error_count(adev, umc_reg_offset, - &(err_data->ce_count)); - umc_v6_1_querry_uncorrectable_error_count(adev, umc_reg_offset, - &(err_data->ue_count)); -} - static void umc_v6_1_query_ras_error_count(struct amdgpu_device *adev, void *ras_error_status) { - amdgpu_umc_for_each_channel(umc_v6_1_query_error_count); + struct ras_err_data* err_data = (struct ras_err_data*)ras_error_status; + + uint32_t umc_inst = 0; + uint32_t ch_inst = 0; + uint32_t umc_reg_offset = 0; + + LOOP_UMC_INST_AND_CH(umc_inst, ch_inst) { + umc_reg_offset = get_umc_6_reg_offset(adev, + umc_inst, + ch_inst); + + umc_v6_1_query_correctable_error_count(adev, + umc_reg_offset, + &(err_data->ce_count)); + umc_v6_1_querry_uncorrectable_error_count(adev, + umc_reg_offset, + &(err_data->ue_count)); + } } static void umc_v6_1_query_error_address(struct amdgpu_device *adev, struct ras_err_data *err_data, - uint32_t umc_reg_offset, uint32_t channel_index) + uint32_t umc_reg_offset, + uint32_t ch_inst, + uint32_t umc_inst) { uint32_t lsb, mc_umc_status_addr; uint64_t mc_umc_status, err_addr, retired_page; struct eeprom_table_record *err_rec; + uint32_t channel_index = adev->umc.channel_idx_tbl[umc_inst * adev->umc.channel_inst_num + ch_inst]; if (adev->asic_type == CHIP_ARCTURUS) { /* UMC 6_1_2 registers */ - mc_umc_status_addr = SOC15_REG_OFFSET(UMC, 0, mmMCA_UMC_UMC0_MCUMC_STATUST0_ARCT); } else { /* UMC 6_1_1 registers */ - mc_umc_status_addr = SOC15_REG_OFFSET(UMC, 0, mmMCA_UMC_UMC0_MCUMC_STATUST0); } @@ -224,12 +203,12 @@ static void umc_v6_1_query_error_address(struct amdgpu_device *adev, /* skip error address process if -ENOMEM */ if (!err_data->err_addr) { /* clear umc status */ - WREG64_UMC(mc_umc_status_addr + umc_reg_offset, 0x0ULL); + WREG64_PCIE((mc_umc_status_addr + umc_reg_offset) * 4, 0x0ULL); return; } err_rec = &err_data->err_addr[err_data->err_addr_cnt]; - mc_umc_status = RREG64_UMC(mc_umc_status_addr + umc_reg_offset); + mc_umc_status = RREG64_PCIE((mc_umc_status_addr + umc_reg_offset) * 4); /* calculate error address if ue/ce error is detected */ if (REG_GET_FIELD(mc_umc_status, MCA_UMC_UMC0_MCUMC_STATUST0, Val) == 1 && @@ -257,39 +236,53 @@ static void umc_v6_1_query_error_address(struct amdgpu_device *adev, err_rec->err_type = AMDGPU_RAS_EEPROM_ERR_NON_RECOVERABLE; err_rec->cu = 0; err_rec->mem_channel = channel_index; - err_rec->mcumc_id = umc_v6_1_get_umc_inst(adev); + err_rec->mcumc_id = umc_inst; err_data->err_addr_cnt++; } } /* clear umc status */ - WREG64_UMC(mc_umc_status_addr + umc_reg_offset, 0x0ULL); + WREG64_PCIE((mc_umc_status_addr + umc_reg_offset) * 4, 0x0ULL); } static void umc_v6_1_query_ras_error_address(struct amdgpu_device *adev, void *ras_error_status) { - amdgpu_umc_for_each_channel(umc_v6_1_query_error_address); + struct ras_err_data* err_data = (struct ras_err_data*)ras_error_status; + + uint32_t umc_inst = 0; + uint32_t ch_inst = 0; + uint32_t umc_reg_offset = 0; + + LOOP_UMC_INST_AND_CH(umc_inst, ch_inst) { + umc_reg_offset = get_umc_6_reg_offset(adev, + umc_inst, + ch_inst); + + umc_v6_1_query_error_address(adev, + err_data, + umc_reg_offset, + ch_inst, + umc_inst); + } + } static void umc_v6_1_err_cnt_init_per_channel(struct amdgpu_device *adev, - struct ras_err_data *err_data, - uint32_t umc_reg_offset, uint32_t channel_index) + uint32_t umc_reg_offset) { uint32_t ecc_err_cnt_sel, ecc_err_cnt_sel_addr; uint32_t ecc_err_cnt_addr; if (adev->asic_type == CHIP_ARCTURUS) { /* UMC 6_1_2 registers */ - ecc_err_cnt_sel_addr = SOC15_REG_OFFSET(UMC, 0, mmUMCCH0_0_EccErrCntSel_ARCT); ecc_err_cnt_addr = SOC15_REG_OFFSET(UMC, 0, mmUMCCH0_0_EccErrCnt_ARCT); } else { /* UMC 6_1_1 registers */ - ecc_err_cnt_sel_addr = SOC15_REG_OFFSET(UMC, 0, mmUMCCH0_0_EccErrCntSel); ecc_err_cnt_addr = @@ -297,28 +290,38 @@ static void umc_v6_1_err_cnt_init_per_channel(struct amdgpu_device *adev, } /* select the lower chip and check the error count */ - ecc_err_cnt_sel = RREG32(ecc_err_cnt_sel_addr + umc_reg_offset); + ecc_err_cnt_sel = RREG32_PCIE((ecc_err_cnt_sel_addr + umc_reg_offset) * 4); ecc_err_cnt_sel = REG_SET_FIELD(ecc_err_cnt_sel, UMCCH0_0_EccErrCntSel, EccErrCntCsSel, 0); /* set ce error interrupt type to APIC based interrupt */ ecc_err_cnt_sel = REG_SET_FIELD(ecc_err_cnt_sel, UMCCH0_0_EccErrCntSel, EccErrInt, 0x1); - WREG32(ecc_err_cnt_sel_addr + umc_reg_offset, ecc_err_cnt_sel); + WREG32_PCIE((ecc_err_cnt_sel_addr + umc_reg_offset) * 4, ecc_err_cnt_sel); /* set error count to initial value */ - WREG32(ecc_err_cnt_addr + umc_reg_offset, UMC_V6_1_CE_CNT_INIT); + WREG32_PCIE((ecc_err_cnt_addr + umc_reg_offset) * 4, UMC_V6_1_CE_CNT_INIT); /* select the higher chip and check the err counter */ ecc_err_cnt_sel = REG_SET_FIELD(ecc_err_cnt_sel, UMCCH0_0_EccErrCntSel, EccErrCntCsSel, 1); - WREG32(ecc_err_cnt_sel_addr + umc_reg_offset, ecc_err_cnt_sel); - WREG32(ecc_err_cnt_addr + umc_reg_offset, UMC_V6_1_CE_CNT_INIT); + WREG32_PCIE((ecc_err_cnt_sel_addr + umc_reg_offset) * 4, ecc_err_cnt_sel); + WREG32_PCIE((ecc_err_cnt_addr + umc_reg_offset) * 4, UMC_V6_1_CE_CNT_INIT); } static void umc_v6_1_err_cnt_init(struct amdgpu_device *adev) { - void *ras_error_status = NULL; + uint32_t umc_inst = 0; + uint32_t ch_inst = 0; + uint32_t umc_reg_offset = 0; + + umc_v6_1_disable_umc_index_mode(adev); - amdgpu_umc_for_each_channel(umc_v6_1_err_cnt_init_per_channel); + LOOP_UMC_INST_AND_CH(umc_inst, ch_inst) { + umc_reg_offset = get_umc_6_reg_offset(adev, + umc_inst, + ch_inst); + + umc_v6_1_err_cnt_init_per_channel(adev, umc_reg_offset); + } } const struct amdgpu_umc_funcs umc_v6_1_funcs = { @@ -326,6 +329,4 @@ const struct amdgpu_umc_funcs umc_v6_1_funcs = { .ras_late_init = amdgpu_umc_ras_late_init, .query_ras_error_count = umc_v6_1_query_ras_error_count, .query_ras_error_address = umc_v6_1_query_ras_error_address, - .enable_umc_index_mode = umc_v6_1_enable_umc_index_mode, - .disable_umc_index_mode = umc_v6_1_disable_umc_index_mode, }; diff --git a/drivers/gpu/drm/amd/amdgpu/vcn_v1_0.c b/drivers/gpu/drm/amd/amdgpu/vcn_v1_0.c index 652cecc030b3..3b025a3f8c7d 100644 --- a/drivers/gpu/drm/amd/amdgpu/vcn_v1_0.c +++ b/drivers/gpu/drm/amd/amdgpu/vcn_v1_0.c @@ -25,6 +25,7 @@ #include "amdgpu.h" #include "amdgpu_vcn.h" +#include "amdgpu_pm.h" #include "soc15.h" #include "soc15d.h" #include "soc15_common.h" @@ -51,6 +52,8 @@ static int vcn_v1_0_set_powergating_state(void *handle, enum amd_powergating_sta static int vcn_v1_0_pause_dpg_mode(struct amdgpu_device *adev, struct dpg_pause_state *new_state); +static void vcn_v1_0_idle_work_handler(struct work_struct *work); + /** * vcn_v1_0_early_init - set function pointers * @@ -105,6 +108,9 @@ static int vcn_v1_0_sw_init(void *handle) if (r) return r; + /* Override the work func */ + adev->vcn.idle_work.work.func = vcn_v1_0_idle_work_handler; + if (adev->firmware.load_type == AMDGPU_FW_LOAD_PSP) { const struct common_firmware_header *hdr; hdr = (const struct common_firmware_header *)adev->vcn.fw->data; @@ -1758,6 +1764,86 @@ static int vcn_v1_0_set_powergating_state(void *handle, return ret; } +static void vcn_v1_0_idle_work_handler(struct work_struct *work) +{ + struct amdgpu_device *adev = + container_of(work, struct amdgpu_device, vcn.idle_work.work); + unsigned int fences = 0, i; + + for (i = 0; i < adev->vcn.num_enc_rings; ++i) + fences += amdgpu_fence_count_emitted(&adev->vcn.inst->ring_enc[i]); + + if (adev->pg_flags & AMD_PG_SUPPORT_VCN_DPG) { + struct dpg_pause_state new_state; + + if (fences) + new_state.fw_based = VCN_DPG_STATE__PAUSE; + else + new_state.fw_based = VCN_DPG_STATE__UNPAUSE; + + if (amdgpu_fence_count_emitted(&adev->jpeg.inst->ring_dec)) + new_state.jpeg = VCN_DPG_STATE__PAUSE; + else + new_state.jpeg = VCN_DPG_STATE__UNPAUSE; + + adev->vcn.pause_dpg_mode(adev, &new_state); + } + + fences += amdgpu_fence_count_emitted(&adev->jpeg.inst->ring_dec); + fences += amdgpu_fence_count_emitted(&adev->vcn.inst->ring_dec); + + if (fences == 0) { + amdgpu_gfx_off_ctrl(adev, true); + if (adev->pm.dpm_enabled) + amdgpu_dpm_enable_uvd(adev, false); + else + amdgpu_device_ip_set_powergating_state(adev, AMD_IP_BLOCK_TYPE_VCN, + AMD_PG_STATE_GATE); + } else { + schedule_delayed_work(&adev->vcn.idle_work, VCN_IDLE_TIMEOUT); + } +} + +void vcn_v1_0_ring_begin_use(struct amdgpu_ring *ring) +{ + struct amdgpu_device *adev = ring->adev; + bool set_clocks = !cancel_delayed_work_sync(&adev->vcn.idle_work); + + if (set_clocks) { + amdgpu_gfx_off_ctrl(adev, false); + if (adev->pm.dpm_enabled) + amdgpu_dpm_enable_uvd(adev, true); + else + amdgpu_device_ip_set_powergating_state(adev, AMD_IP_BLOCK_TYPE_VCN, + AMD_PG_STATE_UNGATE); + } + + if (adev->pg_flags & AMD_PG_SUPPORT_VCN_DPG) { + struct dpg_pause_state new_state; + unsigned int fences = 0, i; + + for (i = 0; i < adev->vcn.num_enc_rings; ++i) + fences += amdgpu_fence_count_emitted(&adev->vcn.inst->ring_enc[i]); + + if (fences) + new_state.fw_based = VCN_DPG_STATE__PAUSE; + else + new_state.fw_based = VCN_DPG_STATE__UNPAUSE; + + if (amdgpu_fence_count_emitted(&adev->jpeg.inst->ring_dec)) + new_state.jpeg = VCN_DPG_STATE__PAUSE; + else + new_state.jpeg = VCN_DPG_STATE__UNPAUSE; + + if (ring->funcs->type == AMDGPU_RING_TYPE_VCN_ENC) + new_state.fw_based = VCN_DPG_STATE__PAUSE; + else if (ring->funcs->type == AMDGPU_RING_TYPE_VCN_JPEG) + new_state.jpeg = VCN_DPG_STATE__PAUSE; + + adev->vcn.pause_dpg_mode(adev, &new_state); + } +} + static const struct amd_ip_funcs vcn_v1_0_ip_funcs = { .name = "vcn_v1_0", .early_init = vcn_v1_0_early_init, @@ -1804,7 +1890,7 @@ static const struct amdgpu_ring_funcs vcn_v1_0_dec_ring_vm_funcs = { .insert_start = vcn_v1_0_dec_ring_insert_start, .insert_end = vcn_v1_0_dec_ring_insert_end, .pad_ib = amdgpu_ring_generic_pad_ib, - .begin_use = amdgpu_vcn_ring_begin_use, + .begin_use = vcn_v1_0_ring_begin_use, .end_use = amdgpu_vcn_ring_end_use, .emit_wreg = vcn_v1_0_dec_ring_emit_wreg, .emit_reg_wait = vcn_v1_0_dec_ring_emit_reg_wait, @@ -1836,7 +1922,7 @@ static const struct amdgpu_ring_funcs vcn_v1_0_enc_ring_vm_funcs = { .insert_nop = amdgpu_ring_insert_nop, .insert_end = vcn_v1_0_enc_ring_insert_end, .pad_ib = amdgpu_ring_generic_pad_ib, - .begin_use = amdgpu_vcn_ring_begin_use, + .begin_use = vcn_v1_0_ring_begin_use, .end_use = amdgpu_vcn_ring_end_use, .emit_wreg = vcn_v1_0_enc_ring_emit_wreg, .emit_reg_wait = vcn_v1_0_enc_ring_emit_reg_wait, diff --git a/drivers/gpu/drm/amd/amdgpu/vcn_v1_0.h b/drivers/gpu/drm/amd/amdgpu/vcn_v1_0.h index 2a497a7a4840..f67d7391fc21 100644 --- a/drivers/gpu/drm/amd/amdgpu/vcn_v1_0.h +++ b/drivers/gpu/drm/amd/amdgpu/vcn_v1_0.h @@ -24,6 +24,8 @@ #ifndef __VCN_V1_0_H__ #define __VCN_V1_0_H__ +void vcn_v1_0_ring_begin_use(struct amdgpu_ring *ring); + extern const struct amdgpu_ip_block_version vcn_v1_0_ip_block; #endif diff --git a/drivers/gpu/drm/amd/amdgpu/vcn_v2_5.c b/drivers/gpu/drm/amd/amdgpu/vcn_v2_5.c index f67fca38c1a9..4ea8e20ed15d 100644 --- a/drivers/gpu/drm/amd/amdgpu/vcn_v2_5.c +++ b/drivers/gpu/drm/amd/amdgpu/vcn_v2_5.c @@ -29,6 +29,7 @@ #include "soc15.h" #include "soc15d.h" #include "vcn_v2_0.h" +#include "mmsch_v1_0.h" #include "vcn/vcn_2_5_offset.h" #include "vcn/vcn_2_5_sh_mask.h" @@ -54,6 +55,7 @@ static void vcn_v2_5_set_enc_ring_funcs(struct amdgpu_device *adev); static void vcn_v2_5_set_irq_funcs(struct amdgpu_device *adev); static int vcn_v2_5_set_powergating_state(void *handle, enum amd_powergating_state state); +static int vcn_v2_5_sriov_start(struct amdgpu_device *adev); static int amdgpu_ih_clientid_vcns[] = { SOC15_IH_CLIENTID_VCN, @@ -88,7 +90,13 @@ static int vcn_v2_5_early_init(void *handle) } else adev->vcn.num_vcn_inst = 1; - adev->vcn.num_enc_rings = 2; + if (amdgpu_sriov_vf(adev)) { + adev->vcn.num_vcn_inst = 2; + adev->vcn.harvest_config = 0; + adev->vcn.num_enc_rings = 1; + } else { + adev->vcn.num_enc_rings = 2; + } vcn_v2_5_set_dec_ring_funcs(adev); vcn_v2_5_set_enc_ring_funcs(adev); @@ -176,7 +184,9 @@ static int vcn_v2_5_sw_init(void *handle) ring = &adev->vcn.inst[j].ring_dec; ring->use_doorbell = true; - ring->doorbell_index = (adev->doorbell_index.vcn.vcn_ring0_1 << 1) + 8*j; + + ring->doorbell_index = (adev->doorbell_index.vcn.vcn_ring0_1 << 1) + + (amdgpu_sriov_vf(adev) ? 2*j : 8*j); sprintf(ring->name, "vcn_dec_%d", j); r = amdgpu_ring_init(adev, ring, 512, &adev->vcn.inst[j].irq, 0); if (r) @@ -185,7 +195,10 @@ static int vcn_v2_5_sw_init(void *handle) for (i = 0; i < adev->vcn.num_enc_rings; ++i) { ring = &adev->vcn.inst[j].ring_enc[i]; ring->use_doorbell = true; - ring->doorbell_index = (adev->doorbell_index.vcn.vcn_ring0_1 << 1) + 2 + i + 8*j; + + ring->doorbell_index = (adev->doorbell_index.vcn.vcn_ring0_1 << 1) + + (amdgpu_sriov_vf(adev) ? (1 + i + 2*j) : (2 + i + 8*j)); + sprintf(ring->name, "vcn_enc_%d.%d", j, i); r = amdgpu_ring_init(adev, ring, 512, &adev->vcn.inst[j].irq, 0); if (r) @@ -193,6 +206,12 @@ static int vcn_v2_5_sw_init(void *handle) } } + if (amdgpu_sriov_vf(adev)) { + r = amdgpu_virt_alloc_mm_table(adev); + if (r) + return r; + } + return 0; } @@ -208,6 +227,9 @@ static int vcn_v2_5_sw_fini(void *handle) int r; struct amdgpu_device *adev = (struct amdgpu_device *)handle; + if (amdgpu_sriov_vf(adev)) + amdgpu_virt_free_mm_table(adev); + r = amdgpu_vcn_suspend(adev); if (r) return r; @@ -228,25 +250,37 @@ static int vcn_v2_5_hw_init(void *handle) { struct amdgpu_device *adev = (struct amdgpu_device *)handle; struct amdgpu_ring *ring; - int i, j, r; + int i, j, r = 0; + + if (amdgpu_sriov_vf(adev)) + r = vcn_v2_5_sriov_start(adev); for (j = 0; j < adev->vcn.num_vcn_inst; ++j) { if (adev->vcn.harvest_config & (1 << j)) continue; - ring = &adev->vcn.inst[j].ring_dec; - adev->nbio.funcs->vcn_doorbell_range(adev, ring->use_doorbell, - ring->doorbell_index, j); + if (amdgpu_sriov_vf(adev)) { + adev->vcn.inst[j].ring_enc[0].sched.ready = true; + adev->vcn.inst[j].ring_enc[1].sched.ready = false; + adev->vcn.inst[j].ring_enc[2].sched.ready = false; + adev->vcn.inst[j].ring_dec.sched.ready = true; + } else { - r = amdgpu_ring_test_helper(ring); - if (r) - goto done; + ring = &adev->vcn.inst[j].ring_dec; + + adev->nbio.funcs->vcn_doorbell_range(adev, ring->use_doorbell, + ring->doorbell_index, j); - for (i = 0; i < adev->vcn.num_enc_rings; ++i) { - ring = &adev->vcn.inst[j].ring_enc[i]; r = amdgpu_ring_test_helper(ring); if (r) goto done; + + for (i = 0; i < adev->vcn.num_enc_rings; ++i) { + ring = &adev->vcn.inst[j].ring_enc[i]; + r = amdgpu_ring_test_helper(ring); + if (r) + goto done; + } } } @@ -741,6 +775,204 @@ static int vcn_v2_5_start(struct amdgpu_device *adev) return 0; } +static int vcn_v2_5_mmsch_start(struct amdgpu_device *adev, + struct amdgpu_mm_table *table) +{ + uint32_t data = 0, loop = 0, size = 0; + uint64_t addr = table->gpu_addr; + struct mmsch_v1_1_init_header *header = NULL;; + + header = (struct mmsch_v1_1_init_header *)table->cpu_addr; + size = header->total_size; + + /* + * 1, write to vce_mmsch_vf_ctx_addr_lo/hi register with GPU mc addr of + * memory descriptor location + */ + WREG32_SOC15(UVD, 0, mmMMSCH_VF_CTX_ADDR_LO, lower_32_bits(addr)); + WREG32_SOC15(UVD, 0, mmMMSCH_VF_CTX_ADDR_HI, upper_32_bits(addr)); + + /* 2, update vmid of descriptor */ + data = RREG32_SOC15(UVD, 0, mmMMSCH_VF_VMID); + data &= ~MMSCH_VF_VMID__VF_CTX_VMID_MASK; + /* use domain0 for MM scheduler */ + data |= (0 << MMSCH_VF_VMID__VF_CTX_VMID__SHIFT); + WREG32_SOC15(UVD, 0, mmMMSCH_VF_VMID, data); + + /* 3, notify mmsch about the size of this descriptor */ + WREG32_SOC15(UVD, 0, mmMMSCH_VF_CTX_SIZE, size); + + /* 4, set resp to zero */ + WREG32_SOC15(UVD, 0, mmMMSCH_VF_MAILBOX_RESP, 0); + + /* + * 5, kick off the initialization and wait until + * VCE_MMSCH_VF_MAILBOX_RESP becomes non-zero + */ + WREG32_SOC15(UVD, 0, mmMMSCH_VF_MAILBOX_HOST, 0x10000001); + + data = RREG32_SOC15(UVD, 0, mmMMSCH_VF_MAILBOX_RESP); + loop = 10; + while ((data & 0x10000002) != 0x10000002) { + udelay(100); + data = RREG32_SOC15(UVD, 0, mmMMSCH_VF_MAILBOX_RESP); + loop--; + if (!loop) + break; + } + + if (!loop) { + dev_err(adev->dev, + "failed to init MMSCH, mmMMSCH_VF_MAILBOX_RESP = %x\n", + data); + return -EBUSY; + } + + return 0; +} + +static int vcn_v2_5_sriov_start(struct amdgpu_device *adev) +{ + struct amdgpu_ring *ring; + uint32_t offset, size, tmp, i, rb_bufsz; + uint32_t table_size = 0; + struct mmsch_v1_0_cmd_direct_write direct_wt = { { 0 } }; + struct mmsch_v1_0_cmd_direct_read_modify_write direct_rd_mod_wt = { { 0 } }; + struct mmsch_v1_0_cmd_direct_polling direct_poll = { { 0 } }; + struct mmsch_v1_0_cmd_end end = { { 0 } }; + uint32_t *init_table = adev->virt.mm_table.cpu_addr; + struct mmsch_v1_1_init_header *header = (struct mmsch_v1_1_init_header *)init_table; + + direct_wt.cmd_header.command_type = MMSCH_COMMAND__DIRECT_REG_WRITE; + direct_rd_mod_wt.cmd_header.command_type = MMSCH_COMMAND__DIRECT_REG_READ_MODIFY_WRITE; + direct_poll.cmd_header.command_type = MMSCH_COMMAND__DIRECT_REG_POLLING; + end.cmd_header.command_type = MMSCH_COMMAND__END; + + header->version = MMSCH_VERSION; + header->total_size = sizeof(struct mmsch_v1_1_init_header) >> 2; + init_table += header->total_size; + + for (i = 0; i < adev->vcn.num_vcn_inst; ++i) { + header->eng[i].table_offset = header->total_size; + header->eng[i].init_status = 0; + header->eng[i].table_size = 0; + + table_size = 0; + + MMSCH_V1_0_INSERT_DIRECT_RD_MOD_WT( + SOC15_REG_OFFSET(UVD, i, mmUVD_STATUS), + ~UVD_STATUS__UVD_BUSY, UVD_STATUS__UVD_BUSY); + + size = AMDGPU_GPU_PAGE_ALIGN(adev->vcn.fw->size + 4); + /* mc resume*/ + if (adev->firmware.load_type == AMDGPU_FW_LOAD_PSP) { + MMSCH_V1_0_INSERT_DIRECT_WT( + SOC15_REG_OFFSET(UVD, i, + mmUVD_LMI_VCPU_CACHE_64BIT_BAR_LOW), + adev->firmware.ucode[AMDGPU_UCODE_ID_VCN + i].tmr_mc_addr_lo); + MMSCH_V1_0_INSERT_DIRECT_WT( + SOC15_REG_OFFSET(UVD, i, + mmUVD_LMI_VCPU_CACHE_64BIT_BAR_HIGH), + adev->firmware.ucode[AMDGPU_UCODE_ID_VCN + i].tmr_mc_addr_hi); + offset = 0; + MMSCH_V1_0_INSERT_DIRECT_WT( + SOC15_REG_OFFSET(UVD, i, mmUVD_VCPU_CACHE_OFFSET0), 0); + } else { + MMSCH_V1_0_INSERT_DIRECT_WT( + SOC15_REG_OFFSET(UVD, i, + mmUVD_LMI_VCPU_CACHE_64BIT_BAR_LOW), + lower_32_bits(adev->vcn.inst[i].gpu_addr)); + MMSCH_V1_0_INSERT_DIRECT_WT( + SOC15_REG_OFFSET(UVD, i, + mmUVD_LMI_VCPU_CACHE_64BIT_BAR_HIGH), + upper_32_bits(adev->vcn.inst[i].gpu_addr)); + offset = size; + MMSCH_V1_0_INSERT_DIRECT_WT( + SOC15_REG_OFFSET(UVD, i, mmUVD_VCPU_CACHE_OFFSET0), + AMDGPU_UVD_FIRMWARE_OFFSET >> 3); + } + + MMSCH_V1_0_INSERT_DIRECT_WT( + SOC15_REG_OFFSET(UVD, i, mmUVD_VCPU_CACHE_SIZE0), + size); + MMSCH_V1_0_INSERT_DIRECT_WT( + SOC15_REG_OFFSET(UVD, i, + mmUVD_LMI_VCPU_CACHE1_64BIT_BAR_LOW), + lower_32_bits(adev->vcn.inst[i].gpu_addr + offset)); + MMSCH_V1_0_INSERT_DIRECT_WT( + SOC15_REG_OFFSET(UVD, i, + mmUVD_LMI_VCPU_CACHE1_64BIT_BAR_HIGH), + upper_32_bits(adev->vcn.inst[i].gpu_addr + offset)); + MMSCH_V1_0_INSERT_DIRECT_WT( + SOC15_REG_OFFSET(UVD, i, mmUVD_VCPU_CACHE_OFFSET1), + 0); + MMSCH_V1_0_INSERT_DIRECT_WT( + SOC15_REG_OFFSET(UVD, i, mmUVD_VCPU_CACHE_SIZE1), + AMDGPU_VCN_STACK_SIZE); + MMSCH_V1_0_INSERT_DIRECT_WT( + SOC15_REG_OFFSET(UVD, i, + mmUVD_LMI_VCPU_CACHE2_64BIT_BAR_LOW), + lower_32_bits(adev->vcn.inst[i].gpu_addr + offset + + AMDGPU_VCN_STACK_SIZE)); + MMSCH_V1_0_INSERT_DIRECT_WT( + SOC15_REG_OFFSET(UVD, i, + mmUVD_LMI_VCPU_CACHE2_64BIT_BAR_HIGH), + upper_32_bits(adev->vcn.inst[i].gpu_addr + offset + + AMDGPU_VCN_STACK_SIZE)); + MMSCH_V1_0_INSERT_DIRECT_WT( + SOC15_REG_OFFSET(UVD, i, mmUVD_VCPU_CACHE_OFFSET2), + 0); + MMSCH_V1_0_INSERT_DIRECT_WT( + SOC15_REG_OFFSET(UVD, i, mmUVD_VCPU_CACHE_SIZE2), + AMDGPU_VCN_CONTEXT_SIZE); + + ring = &adev->vcn.inst[i].ring_enc[0]; + ring->wptr = 0; + + MMSCH_V1_0_INSERT_DIRECT_WT( + SOC15_REG_OFFSET(UVD, i, mmUVD_RB_BASE_LO), + lower_32_bits(ring->gpu_addr)); + MMSCH_V1_0_INSERT_DIRECT_WT( + SOC15_REG_OFFSET(UVD, i, mmUVD_RB_BASE_HI), + upper_32_bits(ring->gpu_addr)); + MMSCH_V1_0_INSERT_DIRECT_WT( + SOC15_REG_OFFSET(UVD, i, mmUVD_RB_SIZE), + ring->ring_size / 4); + + ring = &adev->vcn.inst[i].ring_dec; + ring->wptr = 0; + MMSCH_V1_0_INSERT_DIRECT_WT( + SOC15_REG_OFFSET(UVD, i, + mmUVD_LMI_RBC_RB_64BIT_BAR_LOW), + lower_32_bits(ring->gpu_addr)); + MMSCH_V1_0_INSERT_DIRECT_WT( + SOC15_REG_OFFSET(UVD, i, + mmUVD_LMI_RBC_RB_64BIT_BAR_HIGH), + upper_32_bits(ring->gpu_addr)); + + /* force RBC into idle state */ + rb_bufsz = order_base_2(ring->ring_size); + tmp = REG_SET_FIELD(0, UVD_RBC_RB_CNTL, RB_BUFSZ, rb_bufsz); + tmp = REG_SET_FIELD(tmp, UVD_RBC_RB_CNTL, RB_BLKSZ, 1); + tmp = REG_SET_FIELD(tmp, UVD_RBC_RB_CNTL, RB_NO_FETCH, 1); + tmp = REG_SET_FIELD(tmp, UVD_RBC_RB_CNTL, RB_NO_UPDATE, 1); + tmp = REG_SET_FIELD(tmp, UVD_RBC_RB_CNTL, RB_RPTR_WR_EN, 1); + MMSCH_V1_0_INSERT_DIRECT_WT( + SOC15_REG_OFFSET(UVD, i, mmUVD_RBC_RB_CNTL), tmp); + + /* add end packet */ + memcpy((void *)init_table, &end, sizeof(struct mmsch_v1_0_cmd_end)); + table_size += sizeof(struct mmsch_v1_0_cmd_end) / 4; + init_table += sizeof(struct mmsch_v1_0_cmd_end) / 4; + + /* refine header */ + header->eng[i].table_size = table_size; + header->total_size += table_size; + } + + return vcn_v2_5_mmsch_start(adev, &adev->virt.mm_table); +} + static int vcn_v2_5_stop(struct amdgpu_device *adev) { uint32_t tmp; @@ -1048,6 +1280,9 @@ static int vcn_v2_5_set_clockgating_state(void *handle, struct amdgpu_device *adev = (struct amdgpu_device *)handle; bool enable = (state == AMD_CG_STATE_GATE) ? true : false; + if (amdgpu_sriov_vf(adev)) + return 0; + if (enable) { if (vcn_v2_5_is_idle(handle)) return -EBUSY; @@ -1065,6 +1300,9 @@ static int vcn_v2_5_set_powergating_state(void *handle, struct amdgpu_device *adev = (struct amdgpu_device *)handle; int ret; + if (amdgpu_sriov_vf(adev)) + return 0; + if(state == adev->vcn.cur_state) return 0; diff --git a/drivers/gpu/drm/amd/amdgpu/vega10_ih.c b/drivers/gpu/drm/amd/amdgpu/vega10_ih.c index 5cb7e231de5f..d9e331084ea0 100644 --- a/drivers/gpu/drm/amd/amdgpu/vega10_ih.c +++ b/drivers/gpu/drm/amd/amdgpu/vega10_ih.c @@ -234,16 +234,9 @@ static int vega10_ih_irq_init(struct amdgpu_device *adev) WREG32_SOC15(OSSSYS, 0, mmIH_RB_BASE_HI, (ih->gpu_addr >> 40) & 0xff); ih_rb_cntl = RREG32_SOC15(OSSSYS, 0, mmIH_RB_CNTL); - ih_chicken = RREG32_SOC15(OSSSYS, 0, mmIH_CHICKEN); ih_rb_cntl = vega10_ih_rb_cntl(ih, ih_rb_cntl); - if (adev->irq.ih.use_bus_addr) { - ih_chicken = REG_SET_FIELD(ih_chicken, IH_CHICKEN, MC_SPACE_GPA_ENABLE, 1); - } else { - ih_chicken = REG_SET_FIELD(ih_chicken, IH_CHICKEN, MC_SPACE_FBPA_ENABLE, 1); - } ih_rb_cntl = REG_SET_FIELD(ih_rb_cntl, IH_RB_CNTL, RPTR_REARM, !!adev->irq.msi_enabled); - if (amdgpu_sriov_vf(adev)) { if (psp_reg_program(&adev->psp, PSP_REG_IH_RB_CNTL, ih_rb_cntl)) { DRM_ERROR("PSP program IH_RB_CNTL failed!\n"); @@ -253,10 +246,19 @@ static int vega10_ih_irq_init(struct amdgpu_device *adev) WREG32_SOC15(OSSSYS, 0, mmIH_RB_CNTL, ih_rb_cntl); } - if ((adev->asic_type == CHIP_ARCTURUS - && adev->firmware.load_type == AMDGPU_FW_LOAD_DIRECT) - || adev->asic_type == CHIP_RENOIR) + if ((adev->asic_type == CHIP_ARCTURUS && + adev->firmware.load_type == AMDGPU_FW_LOAD_DIRECT) || + adev->asic_type == CHIP_RENOIR) { + ih_chicken = RREG32_SOC15(OSSSYS, 0, mmIH_CHICKEN); + if (adev->irq.ih.use_bus_addr) { + ih_chicken = REG_SET_FIELD(ih_chicken, IH_CHICKEN, + MC_SPACE_GPA_ENABLE, 1); + } else { + ih_chicken = REG_SET_FIELD(ih_chicken, IH_CHICKEN, + MC_SPACE_FBPA_ENABLE, 1); + } WREG32_SOC15(OSSSYS, 0, mmIH_CHICKEN, ih_chicken); + } /* set the writeback address whether it's enabled or not */ WREG32_SOC15(OSSSYS, 0, mmIH_RB_WPTR_ADDR_LO, |