diff options
Diffstat (limited to 'drivers/gpu/drm/amd')
| -rw-r--r-- | drivers/gpu/drm/amd/amdgpu/amdgpu.h | 12 | ||||
| -rw-r--r-- | drivers/gpu/drm/amd/amdgpu/amdgpu_device.c | 38 | ||||
| -rw-r--r-- | drivers/gpu/drm/amd/amdgpu/amdgpu_gem.c | 8 | ||||
| -rw-r--r-- | drivers/gpu/drm/amd/amdgpu/amdgpu_ib.c | 16 | ||||
| -rw-r--r-- | drivers/gpu/drm/amd/amdgpu/amdgpu_kms.c | 19 | ||||
| -rw-r--r-- | drivers/gpu/drm/amd/amdgpu/amdgpu_uvd.c | 17 | ||||
| -rw-r--r-- | drivers/gpu/drm/amd/amdgpu/cik_sdma.c | 1 | ||||
| -rw-r--r-- | drivers/gpu/drm/amd/amdgpu/gfx_v7_0.c | 54 | ||||
| -rw-r--r-- | drivers/gpu/drm/amd/amdgpu/gfx_v8_0.c | 85 | ||||
| -rw-r--r-- | drivers/gpu/drm/amd/amdgpu/sdma_v2_4.c | 6 | ||||
| -rw-r--r-- | drivers/gpu/drm/amd/amdgpu/sdma_v3_0.c | 6 |
11 files changed, 181 insertions, 81 deletions
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu.h b/drivers/gpu/drm/amd/amdgpu/amdgpu.h index e9fde72cf038..f7b49d5ce4b8 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu.h +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu.h @@ -1130,6 +1130,9 @@ struct amdgpu_gfx { uint32_t me_feature_version; uint32_t ce_feature_version; uint32_t pfp_feature_version; + uint32_t rlc_feature_version; + uint32_t mec_feature_version; + uint32_t mec2_feature_version; struct amdgpu_ring gfx_ring[AMDGPU_MAX_GFX_RINGS]; unsigned num_gfx_rings; struct amdgpu_ring compute_ring[AMDGPU_MAX_COMPUTE_RINGS]; @@ -1639,6 +1642,7 @@ struct amdgpu_sdma { /* SDMA firmware */ const struct firmware *fw; uint32_t fw_version; + uint32_t feature_version; struct amdgpu_ring ring; }; @@ -1866,6 +1870,12 @@ typedef void (*amdgpu_wreg_t)(struct amdgpu_device*, uint32_t, uint32_t); typedef uint32_t (*amdgpu_block_rreg_t)(struct amdgpu_device*, uint32_t, uint32_t); typedef void (*amdgpu_block_wreg_t)(struct amdgpu_device*, uint32_t, uint32_t, uint32_t); +struct amdgpu_ip_block_status { + bool valid; + bool sw; + bool hw; +}; + struct amdgpu_device { struct device *dev; struct drm_device *ddev; @@ -2008,7 +2018,7 @@ struct amdgpu_device { const struct amdgpu_ip_block_version *ip_blocks; int num_ip_blocks; - bool *ip_block_enabled; + struct amdgpu_ip_block_status *ip_block_status; struct mutex mn_lock; DECLARE_HASHTABLE(mn_hash, 7); diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_device.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_device.c index d79009b65867..99f158e1baff 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_device.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_device.c @@ -1191,8 +1191,9 @@ static int amdgpu_early_init(struct amdgpu_device *adev) return -EINVAL; } - adev->ip_block_enabled = kcalloc(adev->num_ip_blocks, sizeof(bool), GFP_KERNEL); - if (adev->ip_block_enabled == NULL) + adev->ip_block_status = kcalloc(adev->num_ip_blocks, + sizeof(struct amdgpu_ip_block_status), GFP_KERNEL); + if (adev->ip_block_status == NULL) return -ENOMEM; if (adev->ip_blocks == NULL) { @@ -1203,18 +1204,18 @@ static int amdgpu_early_init(struct amdgpu_device *adev) for (i = 0; i < adev->num_ip_blocks; i++) { if ((amdgpu_ip_block_mask & (1 << i)) == 0) { DRM_ERROR("disabled ip block: %d\n", i); - adev->ip_block_enabled[i] = false; + adev->ip_block_status[i].valid = false; } else { if (adev->ip_blocks[i].funcs->early_init) { r = adev->ip_blocks[i].funcs->early_init((void *)adev); if (r == -ENOENT) - adev->ip_block_enabled[i] = false; + adev->ip_block_status[i].valid = false; else if (r) return r; else - adev->ip_block_enabled[i] = true; + adev->ip_block_status[i].valid = true; } else { - adev->ip_block_enabled[i] = true; + adev->ip_block_status[i].valid = true; } } } @@ -1227,11 +1228,12 @@ static int amdgpu_init(struct amdgpu_device *adev) int i, r; for (i = 0; i < adev->num_ip_blocks; i++) { - if (!adev->ip_block_enabled[i]) + if (!adev->ip_block_status[i].valid) continue; r = adev->ip_blocks[i].funcs->sw_init((void *)adev); if (r) return r; + adev->ip_block_status[i].sw = true; /* need to do gmc hw init early so we can allocate gpu mem */ if (adev->ip_blocks[i].type == AMD_IP_BLOCK_TYPE_GMC) { r = amdgpu_vram_scratch_init(adev); @@ -1243,11 +1245,12 @@ static int amdgpu_init(struct amdgpu_device *adev) r = amdgpu_wb_init(adev); if (r) return r; + adev->ip_block_status[i].hw = true; } } for (i = 0; i < adev->num_ip_blocks; i++) { - if (!adev->ip_block_enabled[i]) + if (!adev->ip_block_status[i].sw) continue; /* gmc hw init is done early */ if (adev->ip_blocks[i].type == AMD_IP_BLOCK_TYPE_GMC) @@ -1255,6 +1258,7 @@ static int amdgpu_init(struct amdgpu_device *adev) r = adev->ip_blocks[i].funcs->hw_init((void *)adev); if (r) return r; + adev->ip_block_status[i].hw = true; } return 0; @@ -1265,7 +1269,7 @@ static int amdgpu_late_init(struct amdgpu_device *adev) int i = 0, r; for (i = 0; i < adev->num_ip_blocks; i++) { - if (!adev->ip_block_enabled[i]) + if (!adev->ip_block_status[i].valid) continue; /* enable clockgating to save power */ r = adev->ip_blocks[i].funcs->set_clockgating_state((void *)adev, @@ -1287,7 +1291,7 @@ static int amdgpu_fini(struct amdgpu_device *adev) int i, r; for (i = adev->num_ip_blocks - 1; i >= 0; i--) { - if (!adev->ip_block_enabled[i]) + if (!adev->ip_block_status[i].hw) continue; if (adev->ip_blocks[i].type == AMD_IP_BLOCK_TYPE_GMC) { amdgpu_wb_fini(adev); @@ -1300,14 +1304,16 @@ static int amdgpu_fini(struct amdgpu_device *adev) return r; r = adev->ip_blocks[i].funcs->hw_fini((void *)adev); /* XXX handle errors */ + adev->ip_block_status[i].hw = false; } for (i = adev->num_ip_blocks - 1; i >= 0; i--) { - if (!adev->ip_block_enabled[i]) + if (!adev->ip_block_status[i].sw) continue; r = adev->ip_blocks[i].funcs->sw_fini((void *)adev); /* XXX handle errors */ - adev->ip_block_enabled[i] = false; + adev->ip_block_status[i].sw = false; + adev->ip_block_status[i].valid = false; } return 0; @@ -1318,7 +1324,7 @@ static int amdgpu_suspend(struct amdgpu_device *adev) int i, r; for (i = adev->num_ip_blocks - 1; i >= 0; i--) { - if (!adev->ip_block_enabled[i]) + if (!adev->ip_block_status[i].valid) continue; /* ungate blocks so that suspend can properly shut them down */ r = adev->ip_blocks[i].funcs->set_clockgating_state((void *)adev, @@ -1336,7 +1342,7 @@ static int amdgpu_resume(struct amdgpu_device *adev) int i, r; for (i = 0; i < adev->num_ip_blocks; i++) { - if (!adev->ip_block_enabled[i]) + if (!adev->ip_block_status[i].valid) continue; r = adev->ip_blocks[i].funcs->resume(adev); if (r) @@ -1582,8 +1588,8 @@ void amdgpu_device_fini(struct amdgpu_device *adev) amdgpu_fence_driver_fini(adev); amdgpu_fbdev_fini(adev); r = amdgpu_fini(adev); - kfree(adev->ip_block_enabled); - adev->ip_block_enabled = NULL; + kfree(adev->ip_block_status); + adev->ip_block_status = NULL; adev->accel_working = false; /* free i2c buses */ amdgpu_i2c_fini(adev); diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_gem.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_gem.c index ae43b58c9733..4afc507820c0 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_gem.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_gem.c @@ -449,7 +449,7 @@ out: * vital here, so they are not reported back to userspace. */ static void amdgpu_gem_va_update_vm(struct amdgpu_device *adev, - struct amdgpu_bo_va *bo_va) + struct amdgpu_bo_va *bo_va, uint32_t operation) { struct ttm_validate_buffer tv, *entry; struct amdgpu_bo_list_entry *vm_bos; @@ -485,7 +485,9 @@ static void amdgpu_gem_va_update_vm(struct amdgpu_device *adev, if (r) goto error_unlock; - r = amdgpu_vm_bo_update(adev, bo_va, &bo_va->bo->tbo.mem); + + if (operation == AMDGPU_VA_OP_MAP) + r = amdgpu_vm_bo_update(adev, bo_va, &bo_va->bo->tbo.mem); error_unlock: mutex_unlock(&bo_va->vm->mutex); @@ -580,7 +582,7 @@ int amdgpu_gem_va_ioctl(struct drm_device *dev, void *data, } if (!r && !(args->flags & AMDGPU_VM_DELAY_UPDATE)) - amdgpu_gem_va_update_vm(adev, bo_va); + amdgpu_gem_va_update_vm(adev, bo_va, args->operation); drm_gem_object_unreference_unlocked(gobj); return r; diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_ib.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_ib.c index 52dff75aac6f..bc0fac618a3f 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_ib.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_ib.c @@ -180,16 +180,16 @@ int amdgpu_ib_schedule(struct amdgpu_device *adev, unsigned num_ibs, if (vm) { /* do context switch */ amdgpu_vm_flush(ring, vm, ib->sync.last_vm_update); - } - if (vm && ring->funcs->emit_gds_switch) - amdgpu_ring_emit_gds_switch(ring, ib->vm->ids[ring->idx].id, - ib->gds_base, ib->gds_size, - ib->gws_base, ib->gws_size, - ib->oa_base, ib->oa_size); + if (ring->funcs->emit_gds_switch) + amdgpu_ring_emit_gds_switch(ring, ib->vm->ids[ring->idx].id, + ib->gds_base, ib->gds_size, + ib->gws_base, ib->gws_size, + ib->oa_base, ib->oa_size); - if (ring->funcs->emit_hdp_flush) - amdgpu_ring_emit_hdp_flush(ring); + if (ring->funcs->emit_hdp_flush) + amdgpu_ring_emit_hdp_flush(ring); + } old_ctx = ring->current_ctx; for (i = 0; i < num_ibs; ++i) { diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_kms.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_kms.c index 31ad444c6386..3bfe67de8349 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_kms.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_kms.c @@ -235,7 +235,7 @@ static int amdgpu_info_ioctl(struct drm_device *dev, void *data, struct drm_file for (i = 0; i < adev->num_ip_blocks; i++) { if (adev->ip_blocks[i].type == type && - adev->ip_block_enabled[i]) { + adev->ip_block_status[i].valid) { ip.hw_ip_version_major = adev->ip_blocks[i].major; ip.hw_ip_version_minor = adev->ip_blocks[i].minor; ip.capabilities_flags = 0; @@ -274,7 +274,7 @@ static int amdgpu_info_ioctl(struct drm_device *dev, void *data, struct drm_file for (i = 0; i < adev->num_ip_blocks; i++) if (adev->ip_blocks[i].type == type && - adev->ip_block_enabled[i] && + adev->ip_block_status[i].valid && count < AMDGPU_HW_IP_INSTANCE_MAX_COUNT) count++; @@ -317,16 +317,17 @@ static int amdgpu_info_ioctl(struct drm_device *dev, void *data, struct drm_file break; case AMDGPU_INFO_FW_GFX_RLC: fw_info.ver = adev->gfx.rlc_fw_version; - fw_info.feature = 0; + fw_info.feature = adev->gfx.rlc_feature_version; break; case AMDGPU_INFO_FW_GFX_MEC: - if (info->query_fw.index == 0) + if (info->query_fw.index == 0) { fw_info.ver = adev->gfx.mec_fw_version; - else if (info->query_fw.index == 1) + fw_info.feature = adev->gfx.mec_feature_version; + } else if (info->query_fw.index == 1) { fw_info.ver = adev->gfx.mec2_fw_version; - else + fw_info.feature = adev->gfx.mec2_feature_version; + } else return -EINVAL; - fw_info.feature = 0; break; case AMDGPU_INFO_FW_SMC: fw_info.ver = adev->pm.fw_version; @@ -336,7 +337,7 @@ static int amdgpu_info_ioctl(struct drm_device *dev, void *data, struct drm_file if (info->query_fw.index >= 2) return -EINVAL; fw_info.ver = adev->sdma[info->query_fw.index].fw_version; - fw_info.feature = 0; + fw_info.feature = adev->sdma[info->query_fw.index].feature_version; break; default: return -EINVAL; @@ -416,7 +417,7 @@ static int amdgpu_info_ioctl(struct drm_device *dev, void *data, struct drm_file return n ? -EFAULT : 0; } case AMDGPU_INFO_DEV_INFO: { - struct drm_amdgpu_info_device dev_info; + struct drm_amdgpu_info_device dev_info = {}; struct amdgpu_cu_info cu_info; dev_info.device_id = dev->pdev->device; diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_uvd.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_uvd.c index 2f7a5efa21c2..f5c22556ec2c 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_uvd.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_uvd.c @@ -374,7 +374,7 @@ static int amdgpu_uvd_cs_msg_decode(uint32_t *msg, unsigned buf_sizes[]) unsigned height_in_mb = ALIGN(height / 16, 2); unsigned fs_in_mb = width_in_mb * height_in_mb; - unsigned image_size, tmp, min_dpb_size, num_dpb_buffer; + unsigned image_size, tmp, min_dpb_size, num_dpb_buffer, min_ctx_size; image_size = width * height; image_size += image_size / 2; @@ -466,6 +466,8 @@ static int amdgpu_uvd_cs_msg_decode(uint32_t *msg, unsigned buf_sizes[]) num_dpb_buffer = (le32_to_cpu(msg[59]) & 0xff) + 2; min_dpb_size = image_size * num_dpb_buffer; + min_ctx_size = ((width + 255) / 16) * ((height + 255) / 16) + * 16 * num_dpb_buffer + 52 * 1024; break; default: @@ -486,6 +488,7 @@ static int amdgpu_uvd_cs_msg_decode(uint32_t *msg, unsigned buf_sizes[]) buf_sizes[0x1] = dpb_size; buf_sizes[0x2] = image_size; + buf_sizes[0x4] = min_ctx_size; return 0; } @@ -628,6 +631,13 @@ static int amdgpu_uvd_cs_pass2(struct amdgpu_uvd_cs_ctx *ctx) return -EINVAL; } + } else if (cmd == 0x206) { + if ((end - start) < ctx->buf_sizes[4]) { + DRM_ERROR("buffer (%d) to small (%d / %d)!\n", cmd, + (unsigned)(end - start), + ctx->buf_sizes[4]); + return -EINVAL; + } } else if ((cmd != 0x100) && (cmd != 0x204)) { DRM_ERROR("invalid UVD command %X!\n", cmd); return -EINVAL; @@ -755,9 +765,10 @@ int amdgpu_uvd_ring_parse_cs(struct amdgpu_cs_parser *parser, uint32_t ib_idx) struct amdgpu_uvd_cs_ctx ctx = {}; unsigned buf_sizes[] = { [0x00000000] = 2048, - [0x00000001] = 32 * 1024 * 1024, - [0x00000002] = 2048 * 1152 * 3, + [0x00000001] = 0xFFFFFFFF, + [0x00000002] = 0xFFFFFFFF, [0x00000003] = 2048, + [0x00000004] = 0xFFFFFFFF, }; struct amdgpu_ib *ib = &parser->ibs[ib_idx]; int r; diff --git a/drivers/gpu/drm/amd/amdgpu/cik_sdma.c b/drivers/gpu/drm/amd/amdgpu/cik_sdma.c index ab83cc1ca4cc..15df46c93f0a 100644 --- a/drivers/gpu/drm/amd/amdgpu/cik_sdma.c +++ b/drivers/gpu/drm/amd/amdgpu/cik_sdma.c @@ -500,6 +500,7 @@ static int cik_sdma_load_microcode(struct amdgpu_device *adev) amdgpu_ucode_print_sdma_hdr(&hdr->header); fw_size = le32_to_cpu(hdr->header.ucode_size_bytes) / 4; adev->sdma[i].fw_version = le32_to_cpu(hdr->header.ucode_version); + adev->sdma[i].feature_version = le32_to_cpu(hdr->ucode_feature_version); fw_data = (const __le32 *) (adev->sdma[i].fw->data + le32_to_cpu(hdr->header.ucode_array_offset_bytes)); WREG32(mmSDMA0_UCODE_ADDR + sdma_offsets[i], 0); diff --git a/drivers/gpu/drm/amd/amdgpu/gfx_v7_0.c b/drivers/gpu/drm/amd/amdgpu/gfx_v7_0.c index 2c188fb9fd22..0d8bf2cb1956 100644 --- a/drivers/gpu/drm/amd/amdgpu/gfx_v7_0.c +++ b/drivers/gpu/drm/amd/amdgpu/gfx_v7_0.c @@ -2561,7 +2561,7 @@ static bool gfx_v7_0_ring_emit_semaphore(struct amdgpu_ring *ring, * sheduling on the ring. This function schedules the IB * on the gfx ring for execution by the GPU. */ -static void gfx_v7_0_ring_emit_ib(struct amdgpu_ring *ring, +static void gfx_v7_0_ring_emit_ib_gfx(struct amdgpu_ring *ring, struct amdgpu_ib *ib) { bool need_ctx_switch = ring->current_ctx != ib->ctx; @@ -2569,15 +2569,10 @@ static void gfx_v7_0_ring_emit_ib(struct amdgpu_ring *ring, u32 next_rptr = ring->wptr + 5; /* drop the CE preamble IB for the same context */ - if ((ring->type == AMDGPU_RING_TYPE_GFX) && - (ib->flags & AMDGPU_IB_FLAG_PREAMBLE) && - !need_ctx_switch) + if ((ib->flags & AMDGPU_IB_FLAG_PREAMBLE) && !need_ctx_switch) return; - if (ring->type == AMDGPU_RING_TYPE_COMPUTE) - control |= INDIRECT_BUFFER_VALID; - - if (need_ctx_switch && ring->type == AMDGPU_RING_TYPE_GFX) + if (need_ctx_switch) next_rptr += 2; next_rptr += 4; @@ -2588,7 +2583,7 @@ static void gfx_v7_0_ring_emit_ib(struct amdgpu_ring *ring, amdgpu_ring_write(ring, next_rptr); /* insert SWITCH_BUFFER packet before first IB in the ring frame */ - if (need_ctx_switch && ring->type == AMDGPU_RING_TYPE_GFX) { + if (need_ctx_switch) { amdgpu_ring_write(ring, PACKET3(PACKET3_SWITCH_BUFFER, 0)); amdgpu_ring_write(ring, 0); } @@ -2611,6 +2606,35 @@ static void gfx_v7_0_ring_emit_ib(struct amdgpu_ring *ring, amdgpu_ring_write(ring, control); } +static void gfx_v7_0_ring_emit_ib_compute(struct amdgpu_ring *ring, + struct amdgpu_ib *ib) +{ + u32 header, control = 0; + u32 next_rptr = ring->wptr + 5; + + control |= INDIRECT_BUFFER_VALID; + next_rptr += 4; + amdgpu_ring_write(ring, PACKET3(PACKET3_WRITE_DATA, 3)); + amdgpu_ring_write(ring, WRITE_DATA_DST_SEL(5) | WR_CONFIRM); + amdgpu_ring_write(ring, ring->next_rptr_gpu_addr & 0xfffffffc); + amdgpu_ring_write(ring, upper_32_bits(ring->next_rptr_gpu_addr) & 0xffffffff); + amdgpu_ring_write(ring, next_rptr); + + header = PACKET3(PACKET3_INDIRECT_BUFFER, 2); + + control |= ib->length_dw | + (ib->vm ? (ib->vm->ids[ring->idx].id << 24) : 0); + + amdgpu_ring_write(ring, header); + amdgpu_ring_write(ring, +#ifdef __BIG_ENDIAN + (2 << 0) | +#endif + (ib->gpu_addr & 0xFFFFFFFC)); + amdgpu_ring_write(ring, upper_32_bits(ib->gpu_addr) & 0xFFFF); + amdgpu_ring_write(ring, control); +} + /** * gfx_v7_0_ring_test_ib - basic ring IB test * @@ -3056,6 +3080,8 @@ static int gfx_v7_0_cp_compute_load_microcode(struct amdgpu_device *adev) mec_hdr = (const struct gfx_firmware_header_v1_0 *)adev->gfx.mec_fw->data; amdgpu_ucode_print_gfx_hdr(&mec_hdr->header); adev->gfx.mec_fw_version = le32_to_cpu(mec_hdr->header.ucode_version); + adev->gfx.mec_feature_version = le32_to_cpu( + mec_hdr->ucode_feature_version); gfx_v7_0_cp_compute_enable(adev, false); @@ -3078,6 +3104,8 @@ static int gfx_v7_0_cp_compute_load_microcode(struct amdgpu_device *adev) mec2_hdr = (const struct gfx_firmware_header_v1_0 *)adev->gfx.mec2_fw->data; amdgpu_ucode_print_gfx_hdr(&mec2_hdr->header); adev->gfx.mec2_fw_version = le32_to_cpu(mec2_hdr->header.ucode_version); + adev->gfx.mec2_feature_version = le32_to_cpu( + mec2_hdr->ucode_feature_version); /* MEC2 */ fw_data = (const __le32 *) @@ -4042,6 +4070,8 @@ static int gfx_v7_0_rlc_resume(struct amdgpu_device *adev) hdr = (const struct rlc_firmware_header_v1_0 *)adev->gfx.rlc_fw->data; amdgpu_ucode_print_rlc_hdr(&hdr->header); adev->gfx.rlc_fw_version = le32_to_cpu(hdr->header.ucode_version); + adev->gfx.rlc_feature_version = le32_to_cpu( + hdr->ucode_feature_version); gfx_v7_0_rlc_stop(adev); @@ -5098,7 +5128,7 @@ static void gfx_v7_0_print_status(void *handle) dev_info(adev->dev, " CP_HPD_EOP_CONTROL=0x%08X\n", RREG32(mmCP_HPD_EOP_CONTROL)); - for (queue = 0; queue < 8; i++) { + for (queue = 0; queue < 8; queue++) { cik_srbm_select(adev, me, pipe, queue, 0); dev_info(adev->dev, " queue: %d\n", queue); dev_info(adev->dev, " CP_PQ_WPTR_POLL_CNTL=0x%08X\n", @@ -5555,7 +5585,7 @@ static const struct amdgpu_ring_funcs gfx_v7_0_ring_funcs_gfx = { .get_wptr = gfx_v7_0_ring_get_wptr_gfx, .set_wptr = gfx_v7_0_ring_set_wptr_gfx, .parse_cs = NULL, - .emit_ib = gfx_v7_0_ring_emit_ib, + .emit_ib = gfx_v7_0_ring_emit_ib_gfx, .emit_fence = gfx_v7_0_ring_emit_fence_gfx, .emit_semaphore = gfx_v7_0_ring_emit_semaphore, .emit_vm_flush = gfx_v7_0_ring_emit_vm_flush, @@ -5571,7 +5601,7 @@ static const struct amdgpu_ring_funcs gfx_v7_0_ring_funcs_compute = { .get_wptr = gfx_v7_0_ring_get_wptr_compute, .set_wptr = gfx_v7_0_ring_set_wptr_compute, .parse_cs = NULL, - .emit_ib = gfx_v7_0_ring_emit_ib, + .emit_ib = gfx_v7_0_ring_emit_ib_compute, .emit_fence = gfx_v7_0_ring_emit_fence_compute, .emit_semaphore = gfx_v7_0_ring_emit_semaphore, .emit_vm_flush = gfx_v7_0_ring_emit_vm_flush, diff --git a/drivers/gpu/drm/amd/amdgpu/gfx_v8_0.c b/drivers/gpu/drm/amd/amdgpu/gfx_v8_0.c index 1c7c992dea37..20e2cfd521d5 100644 --- a/drivers/gpu/drm/amd/amdgpu/gfx_v8_0.c +++ b/drivers/gpu/drm/amd/amdgpu/gfx_v8_0.c @@ -587,6 +587,7 @@ static int gfx_v8_0_init_microcode(struct amdgpu_device *adev) int err; struct amdgpu_firmware_info *info = NULL; const struct common_firmware_header *header = NULL; + const struct gfx_firmware_header_v1_0 *cp_hdr; DRM_DEBUG("\n"); @@ -611,6 +612,9 @@ static int gfx_v8_0_init_microcode(struct amdgpu_device *adev) err = amdgpu_ucode_validate(adev->gfx.pfp_fw); if (err) goto out; + cp_hdr = (const struct gfx_firmware_header_v1_0 *)adev->gfx.pfp_fw->data; + adev->gfx.pfp_fw_version = le32_to_cpu(cp_hdr->header.ucode_version); + adev->gfx.pfp_feature_version = le32_to_cpu(cp_hdr->ucode_feature_version); snprintf(fw_name, sizeof(fw_name), "amdgpu/%s_me.bin", chip_name); err = request_firmware(&adev->gfx.me_fw, fw_name, adev->dev); @@ -619,6 +623,9 @@ static int gfx_v8_0_init_microcode(struct amdgpu_device *adev) err = amdgpu_ucode_validate(adev->gfx.me_fw); if (err) goto out; + cp_hdr = (const struct gfx_firmware_header_v1_0 *)adev->gfx.me_fw->data; + adev->gfx.me_fw_version = le32_to_cpu(cp_hdr->header.ucode_version); + adev->gfx.me_feature_version = le32_to_cpu(cp_hdr->ucode_feature_version); snprintf(fw_name, sizeof(fw_name), "amdgpu/%s_ce.bin", chip_name); err = request_firmware(&adev->gfx.ce_fw, fw_name, adev->dev); @@ -627,12 +634,18 @@ static int gfx_v8_0_init_microcode(struct amdgpu_device *adev) err = amdgpu_ucode_validate(adev->gfx.ce_fw); if (err) goto out; + cp_hdr = (const struct gfx_firmware_header_v1_0 *)adev->gfx.ce_fw->data; + adev->gfx.ce_fw_version = le32_to_cpu(cp_hdr->header.ucode_version); + adev->gfx.ce_feature_version = le32_to_cpu(cp_hdr->ucode_feature_version); snprintf(fw_name, sizeof(fw_name), "amdgpu/%s_rlc.bin", chip_name); err = request_firmware(&adev->gfx.rlc_fw, fw_name, adev->dev); if (err) goto out; err = amdgpu_ucode_validate(adev->gfx.rlc_fw); + cp_hdr = (const struct gfx_firmware_header_v1_0 *)adev->gfx.rlc_fw->data; + adev->gfx.rlc_fw_version = le32_to_cpu(cp_hdr->header.ucode_version); + adev->gfx.rlc_feature_version = le32_to_cpu(cp_hdr->ucode_feature_version); snprintf(fw_name, sizeof(fw_name), "amdgpu/%s_mec.bin", chip_name); err = request_firmware(&adev->gfx.mec_fw, fw_name, adev->dev); @@ -641,6 +654,9 @@ static int gfx_v8_0_init_microcode(struct amdgpu_device *adev) err = amdgpu_ucode_validate(adev->gfx.mec_fw); if (err) goto out; + cp_hdr = (const struct gfx_firmware_header_v1_0 *)adev->gfx.mec_fw->data; + adev->gfx.mec_fw_version = le32_to_cpu(cp_hdr->header.ucode_version); + adev->gfx.mec_feature_version = le32_to_cpu(cp_hdr->ucode_feature_version); snprintf(fw_name, sizeof(fw_name), "amdgpu/%s_mec2.bin", chip_name); err = request_firmware(&adev->gfx.mec2_fw, fw_name, adev->dev); @@ -648,6 +664,12 @@ static int gfx_v8_0_init_microcode(struct amdgpu_device *adev) err = amdgpu_ucode_validate(adev->gfx.mec2_fw); if (err) goto out; + cp_hdr = (const struct gfx_firmware_header_v1_0 *) + adev->gfx.mec2_fw->data; + adev->gfx.mec2_fw_version = le32_to_cpu( + cp_hdr->header.ucode_version); + adev->gfx.mec2_feature_version = le32_to_cpu( + cp_hdr->ucode_feature_version); } else { err = 0; adev->gfx.mec2_fw = NULL; @@ -1983,6 +2005,7 @@ static void gfx_v8_0_gpu_init(struct amdgpu_device *adev) adev->gfx.config.max_shader_engines = 1; adev->gfx.config.max_tile_pipes = 2; adev->gfx.config.max_sh_per_se = 1; + adev->gfx.config.max_backends_per_se = 2; switch (adev->pdev->revision) { case 0xc4: @@ -1991,7 +2014,6 @@ static void gfx_v8_0_gpu_init(struct amdgpu_device *adev) case 0xcc: /* B10 */ adev->gfx.config.max_cu_per_sh = 8; - adev->gfx.config.max_backends_per_se = 2; break; case 0xc5: case 0x81: @@ -2000,14 +2022,12 @@ static void gfx_v8_0_gpu_init(struct amdgpu_device *adev) case 0xcd: /* B8 */ adev->gfx.config.max_cu_per_sh = 6; - adev->gfx.config.max_backends_per_se = 2; break; case 0xc6: case 0xca: case 0xce: /* B6 */ adev->gfx.config.max_cu_per_sh = 6; - adev->gfx.config.max_backends_per_se = 2; break; case 0xc7: case 0x87: @@ -2015,7 +2035,6 @@ static void gfx_v8_0_gpu_init(struct amdgpu_device *adev) default: /* B4 */ adev->gfx.config.max_cu_per_sh = 4; - adev->gfx.config.max_backends_per_se = 1; break; } @@ -2275,7 +2294,6 @@ static int gfx_v8_0_rlc_load_microcode(struct amdgpu_device *adev) hdr = (const struct rlc_firmware_header_v2_0 *)adev->gfx.rlc_fw->data; amdgpu_ucode_print_rlc_hdr(&hdr->header); - adev->gfx.rlc_fw_version = le32_to_cpu(hdr->header.ucode_version); fw_data = (const __le32 *)(adev->gfx.rlc_fw->data + le32_to_cpu(hdr->header.ucode_array_offset_bytes)); @@ -2361,12 +2379,6 @@ static int gfx_v8_0_cp_gfx_load_microcode(struct amdgpu_device *adev) amdgpu_ucode_print_gfx_hdr(&pfp_hdr->header); amdgpu_ucode_print_gfx_hdr(&ce_hdr->header); amdgpu_ucode_print_gfx_hdr(&me_hdr->header); - adev->gfx.pfp_fw_version = le32_to_cpu(pfp_hdr->header.ucode_version); - adev->gfx.ce_fw_version = le32_to_cpu(ce_hdr->header.ucode_version); - adev->gfx.me_fw_version = le32_to_cpu(me_hdr->header.ucode_version); - adev->gfx.me_feature_version = le32_to_cpu(me_hdr->ucode_feature_version); - adev->gfx.ce_feature_version = le32_to_cpu(ce_hdr->ucode_feature_version); - adev->gfx.pfp_feature_version = le32_to_cpu(pfp_hdr->ucode_feature_version); gfx_v8_0_cp_gfx_enable(adev, false); @@ -2622,7 +2634,6 @@ static int gfx_v8_0_cp_compute_load_microcode(struct amdgpu_device *adev) mec_hdr = (const struct gfx_firmware_header_v1_0 *)adev->gfx.mec_fw->data; amdgpu_ucode_print_gfx_hdr(&mec_hdr->header); - adev->gfx.mec_fw_version = le32_to_cpu(mec_hdr->header.ucode_version); fw_data = (const __le32 *) (adev->gfx.mec_fw->data + @@ -2641,7 +2652,6 @@ static int gfx_v8_0_cp_compute_load_microcode(struct amdgpu_device *adev) mec2_hdr = (const struct gfx_firmware_header_v1_0 *)adev->gfx.mec2_fw->data; amdgpu_ucode_print_gfx_hdr(&mec2_hdr->header); - adev->gfx.mec2_fw_version = le32_to_cpu(mec2_hdr->header.ucode_version); fw_data = (const __le32 *) (adev->gfx.mec2_fw->data + @@ -3125,7 +3135,7 @@ static int gfx_v8_0_cp_compute_resume(struct amdgpu_device *adev) WREG32(mmCP_MEC_DOORBELL_RANGE_LOWER, AMDGPU_DOORBELL_KIQ << 2); WREG32(mmCP_MEC_DOORBELL_RANGE_UPPER, - 0x7FFFF << 2); + AMDGPU_DOORBELL_MEC_RING7 << 2); } tmp = RREG32(mmCP_HQD_PQ_DOORBELL_CONTROL); tmp = REG_SET_FIELD(tmp, CP_HQD_PQ_DOORBELL_CONTROL, @@ -3753,7 +3763,7 @@ static void gfx_v8_0_ring_emit_hdp_flush(struct amdgpu_ring *ring) amdgpu_ring_write(ring, 0x20); /* poll interval */ } -static void gfx_v8_0_ring_emit_ib(struct amdgpu_ring *ring, +static void gfx_v8_0_ring_emit_ib_gfx(struct amdgpu_ring *ring, struct amdgpu_ib *ib) { bool need_ctx_switch = ring->current_ctx != ib->ctx; @@ -3761,15 +3771,10 @@ static void gfx_v8_0_ring_emit_ib(struct amdgpu_ring *ring, u32 next_rptr = ring->wptr + 5; /* drop the CE preamble IB for the same context */ - if ((ring->type == AMDGPU_RING_TYPE_GFX) && - (ib->flags & AMDGPU_IB_FLAG_PREAMBLE) && - !need_ctx_switch) + if ((ib->flags & AMDGPU_IB_FLAG_PREAMBLE) && !need_ctx_switch) return; - if (ring->type == AMDGPU_RING_TYPE_COMPUTE) - control |= INDIRECT_BUFFER_VALID; - - if (need_ctx_switch && ring->type == AMDGPU_RING_TYPE_GFX) + if (need_ctx_switch) next_rptr += 2; next_rptr += 4; @@ -3780,7 +3785,7 @@ static void gfx_v8_0_ring_emit_ib(struct amdgpu_ring *ring, amdgpu_ring_write(ring, next_rptr); /* insert SWITCH_BUFFER packet before first IB in the ring frame */ - if (need_ctx_switch && ring->type == AMDGPU_RING_TYPE_GFX) { + if (need_ctx_switch) { amdgpu_ring_write(ring, PACKET3(PACKET3_SWITCH_BUFFER, 0)); amdgpu_ring_write(ring, 0); } @@ -3803,6 +3808,36 @@ static void gfx_v8_0_ring_emit_ib(struct amdgpu_ring *ring, amdgpu_ring_write(ring, control); } +static void gfx_v8_0_ring_emit_ib_compute(struct amdgpu_ring *ring, + struct amdgpu_ib *ib) +{ + u32 header, control = 0; + u32 next_rptr = ring->wptr + 5; + + control |= INDIRECT_BUFFER_VALID; + + next_rptr += 4; + amdgpu_ring_write(ring, PACKET3(PACKET3_WRITE_DATA, 3)); + amdgpu_ring_write(ring, WRITE_DATA_DST_SEL(5) | WR_CONFIRM); + amdgpu_ring_write(ring, ring->next_rptr_gpu_addr & 0xfffffffc); + amdgpu_ring_write(ring, upper_32_bits(ring->next_rptr_gpu_addr) & 0xffffffff); + amdgpu_ring_write(ring, next_rptr); + + header = PACKET3(PACKET3_INDIRECT_BUFFER, 2); + + control |= ib->length_dw | + (ib->vm ? (ib->vm->ids[ring->idx].id << 24) : 0); + + amdgpu_ring_write(ring, header); + amdgpu_ring_write(ring, +#ifdef __BIG_ENDIAN + (2 << 0) | +#endif + (ib->gpu_addr & 0xFFFFFFFC)); + amdgpu_ring_write(ring, upper_32_bits(ib->gpu_addr) & 0xFFFF); + amdgpu_ring_write(ring, control); +} + static void gfx_v8_0_ring_emit_fence_gfx(struct amdgpu_ring *ring, u64 addr, u64 seq, unsigned flags) { @@ -4224,7 +4259,7 @@ static const struct amdgpu_ring_funcs gfx_v8_0_ring_funcs_gfx = { .get_wptr = gfx_v8_0_ring_get_wptr_gfx, .set_wptr = gfx_v8_0_ring_set_wptr_gfx, .parse_cs = NULL, - .emit_ib = gfx_v8_0_ring_emit_ib, + .emit_ib = gfx_v8_0_ring_emit_ib_gfx, .emit_fence = gfx_v8_0_ring_emit_fence_gfx, .emit_semaphore = gfx_v8_0_ring_emit_semaphore, .emit_vm_flush = gfx_v8_0_ring_emit_vm_flush, @@ -4240,7 +4275,7 @@ static const struct amdgpu_ring_funcs gfx_v8_0_ring_funcs_compute = { .get_wptr = gfx_v8_0_ring_get_wptr_compute, .set_wptr = gfx_v8_0_ring_set_wptr_compute, .parse_cs = NULL, - .emit_ib = gfx_v8_0_ring_emit_ib, + .emit_ib = gfx_v8_0_ring_emit_ib_compute, .emit_fence = gfx_v8_0_ring_emit_fence_compute, .emit_semaphore = gfx_v8_0_ring_emit_semaphore, .emit_vm_flush = gfx_v8_0_ring_emit_vm_flush, diff --git a/drivers/gpu/drm/amd/amdgpu/sdma_v2_4.c b/drivers/gpu/drm/amd/amdgpu/sdma_v2_4.c index d7895885fe0c..a988dfb1d394 100644 --- a/drivers/gpu/drm/amd/amdgpu/sdma_v2_4.c +++ b/drivers/gpu/drm/amd/amdgpu/sdma_v2_4.c @@ -121,6 +121,7 @@ static int sdma_v2_4_init_microcode(struct amdgpu_device *adev) int err, i; struct amdgpu_firmware_info *info = NULL; const struct common_firmware_header *header = NULL; + const struct sdma_firmware_header_v1_0 *hdr; DRM_DEBUG("\n"); @@ -142,6 +143,9 @@ static int sdma_v2_4_init_microcode(struct amdgpu_device *adev) err = amdgpu_ucode_validate(adev->sdma[i].fw); if (err) goto out; + hdr = (const struct sdma_firmware_header_v1_0 *)adev->sdma[i].fw->data; + adev->sdma[i].fw_version = le32_to_cpu(hdr->header.ucode_version); + adev->sdma[i].feature_version = le32_to_cpu(hdr->ucode_feature_version); if (adev->firmware.smu_load) { info = &adev->firmware.ucode[AMDGPU_UCODE_ID_SDMA0 + i]; @@ -541,8 +545,6 @@ static int sdma_v2_4_load_microcode(struct amdgpu_device *adev) hdr = (const struct sdma_firmware_header_v1_0 *)adev->sdma[i].fw->data; amdgpu_ucode_print_sdma_hdr(&hdr->header); fw_size = le32_to_cpu(hdr->header.ucode_size_bytes) / 4; - adev->sdma[i].fw_version = le32_to_cpu(hdr->header.ucode_version); - fw_data = (const __le32 *) (adev->sdma[i].fw->data + le32_to_cpu(hdr->header.ucode_array_offset_bytes)); diff --git a/drivers/gpu/drm/amd/amdgpu/sdma_v3_0.c b/drivers/gpu/drm/amd/amdgpu/sdma_v3_0.c index 7bb37b93993f..2b86569b18d3 100644 --- a/drivers/gpu/drm/amd/amdgpu/sdma_v3_0.c +++ b/drivers/gpu/drm/amd/amdgpu/sdma_v3_0.c @@ -159,6 +159,7 @@ static int sdma_v3_0_init_microcode(struct amdgpu_device *adev) int err, i; struct amdgpu_firmware_info *info = NULL; const struct common_firmware_header *header = NULL; + const struct sdma_firmware_header_v1_0 *hdr; DRM_DEBUG("\n"); @@ -183,6 +184,9 @@ static int sdma_v3_0_init_microcode(struct amdgpu_device *adev) err = amdgpu_ucode_validate(adev->sdma[i].fw); if (err) goto out; + hdr = (const struct sdma_firmware_header_v1_0 *)adev->sdma[i].fw->data; + adev->sdma[i].fw_version = le32_to_cpu(hdr->header.ucode_version); + adev->sdma[i].feature_version = le32_to_cpu(hdr->ucode_feature_version); if (adev->firmware.smu_load) { info = &adev->firmware.ucode[AMDGPU_UCODE_ID_SDMA0 + i]; @@ -630,8 +634,6 @@ static int sdma_v3_0_load_microcode(struct amdgpu_device *adev) hdr = (const struct sdma_firmware_header_v1_0 *)adev->sdma[i].fw->data; amdgpu_ucode_print_sdma_hdr(&hdr->header); fw_size = le32_to_cpu(hdr->header.ucode_size_bytes) / 4; - adev->sdma[i].fw_version = le32_to_cpu(hdr->header.ucode_version); - fw_data = (const __le32 *) (adev->sdma[i].fw->data + le32_to_cpu(hdr->header.ucode_array_offset_bytes)); |