diff options
Diffstat (limited to 'drivers/gpu/drm/radeon/radeon_uvd.c')
| -rw-r--r-- | drivers/gpu/drm/radeon/radeon_uvd.c | 325 |
1 files changed, 229 insertions, 96 deletions
diff --git a/drivers/gpu/drm/radeon/radeon_uvd.c b/drivers/gpu/drm/radeon/radeon_uvd.c index 906e5c0ca3b..a4ad270e826 100644 --- a/drivers/gpu/drm/radeon/radeon_uvd.c +++ b/drivers/gpu/drm/radeon/radeon_uvd.c @@ -44,30 +44,24 @@ #define FIRMWARE_CYPRESS "radeon/CYPRESS_uvd.bin" #define FIRMWARE_SUMO "radeon/SUMO_uvd.bin" #define FIRMWARE_TAHITI "radeon/TAHITI_uvd.bin" +#define FIRMWARE_BONAIRE "radeon/BONAIRE_uvd.bin" MODULE_FIRMWARE(FIRMWARE_RV710); MODULE_FIRMWARE(FIRMWARE_CYPRESS); MODULE_FIRMWARE(FIRMWARE_SUMO); MODULE_FIRMWARE(FIRMWARE_TAHITI); +MODULE_FIRMWARE(FIRMWARE_BONAIRE); static void radeon_uvd_idle_work_handler(struct work_struct *work); int radeon_uvd_init(struct radeon_device *rdev) { - struct platform_device *pdev; unsigned long bo_size; const char *fw_name; int i, r; INIT_DELAYED_WORK(&rdev->uvd.idle_work, radeon_uvd_idle_work_handler); - pdev = platform_device_register_simple("radeon_uvd", 0, NULL, 0); - r = IS_ERR(pdev); - if (r) { - dev_err(rdev->dev, "radeon_uvd: Failed to register firmware\n"); - return -EINVAL; - } - switch (rdev->family) { case CHIP_RV710: case CHIP_RV730: @@ -97,23 +91,29 @@ int radeon_uvd_init(struct radeon_device *rdev) case CHIP_VERDE: case CHIP_PITCAIRN: case CHIP_ARUBA: + case CHIP_OLAND: fw_name = FIRMWARE_TAHITI; break; + case CHIP_BONAIRE: + case CHIP_KABINI: + case CHIP_KAVERI: + case CHIP_HAWAII: + case CHIP_MULLINS: + fw_name = FIRMWARE_BONAIRE; + break; + default: return -EINVAL; } - r = request_firmware(&rdev->uvd_fw, fw_name, &pdev->dev); + r = request_firmware(&rdev->uvd_fw, fw_name, rdev->dev); if (r) { dev_err(rdev->dev, "radeon_uvd: Can't load firmware \"%s\"\n", fw_name); - platform_device_unregister(pdev); return r; } - platform_device_unregister(pdev); - bo_size = RADEON_GPU_PAGE_ALIGN(rdev->uvd_fw->size + 8) + RADEON_UVD_STACK_SIZE + RADEON_UVD_HEAP_SIZE; r = radeon_bo_create(rdev, bo_size, PAGE_SIZE, true, @@ -123,20 +123,34 @@ int radeon_uvd_init(struct radeon_device *rdev) return r; } - r = radeon_uvd_resume(rdev); - if (r) + r = radeon_bo_reserve(rdev->uvd.vcpu_bo, false); + if (r) { + radeon_bo_unref(&rdev->uvd.vcpu_bo); + dev_err(rdev->dev, "(%d) failed to reserve UVD bo\n", r); return r; + } - memset(rdev->uvd.cpu_addr, 0, bo_size); - memcpy(rdev->uvd.cpu_addr, rdev->uvd_fw->data, rdev->uvd_fw->size); + r = radeon_bo_pin(rdev->uvd.vcpu_bo, RADEON_GEM_DOMAIN_VRAM, + &rdev->uvd.gpu_addr); + if (r) { + radeon_bo_unreserve(rdev->uvd.vcpu_bo); + radeon_bo_unref(&rdev->uvd.vcpu_bo); + dev_err(rdev->dev, "(%d) UVD bo pin failed\n", r); + return r; + } - r = radeon_uvd_suspend(rdev); - if (r) + r = radeon_bo_kmap(rdev->uvd.vcpu_bo, &rdev->uvd.cpu_addr); + if (r) { + dev_err(rdev->dev, "(%d) UVD map failed\n", r); return r; + } + + radeon_bo_unreserve(rdev->uvd.vcpu_bo); for (i = 0; i < RADEON_MAX_UVD_HANDLES; ++i) { atomic_set(&rdev->uvd.handles[i], 0); rdev->uvd.filp[i] = NULL; + rdev->uvd.img_size[i] = 0; } return 0; @@ -144,16 +158,10 @@ int radeon_uvd_init(struct radeon_device *rdev) void radeon_uvd_fini(struct radeon_device *rdev) { - radeon_uvd_suspend(rdev); - radeon_bo_unref(&rdev->uvd.vcpu_bo); -} - -int radeon_uvd_suspend(struct radeon_device *rdev) -{ int r; if (rdev->uvd.vcpu_bo == NULL) - return 0; + return; r = radeon_bo_reserve(rdev->uvd.vcpu_bo, false); if (!r) { @@ -161,39 +169,64 @@ int radeon_uvd_suspend(struct radeon_device *rdev) radeon_bo_unpin(rdev->uvd.vcpu_bo); radeon_bo_unreserve(rdev->uvd.vcpu_bo); } - return r; + + radeon_bo_unref(&rdev->uvd.vcpu_bo); + + radeon_ring_fini(rdev, &rdev->ring[R600_RING_TYPE_UVD_INDEX]); + + release_firmware(rdev->uvd_fw); +} + +int radeon_uvd_suspend(struct radeon_device *rdev) +{ + unsigned size; + void *ptr; + int i; + + if (rdev->uvd.vcpu_bo == NULL) + return 0; + + for (i = 0; i < RADEON_MAX_UVD_HANDLES; ++i) + if (atomic_read(&rdev->uvd.handles[i])) + break; + + if (i == RADEON_MAX_UVD_HANDLES) + return 0; + + size = radeon_bo_size(rdev->uvd.vcpu_bo); + size -= rdev->uvd_fw->size; + + ptr = rdev->uvd.cpu_addr; + ptr += rdev->uvd_fw->size; + + rdev->uvd.saved_bo = kmalloc(size, GFP_KERNEL); + memcpy(rdev->uvd.saved_bo, ptr, size); + + return 0; } int radeon_uvd_resume(struct radeon_device *rdev) { - int r; + unsigned size; + void *ptr; if (rdev->uvd.vcpu_bo == NULL) return -EINVAL; - r = radeon_bo_reserve(rdev->uvd.vcpu_bo, false); - if (r) { - radeon_bo_unref(&rdev->uvd.vcpu_bo); - dev_err(rdev->dev, "(%d) failed to reserve UVD bo\n", r); - return r; - } + memcpy(rdev->uvd.cpu_addr, rdev->uvd_fw->data, rdev->uvd_fw->size); - r = radeon_bo_pin(rdev->uvd.vcpu_bo, RADEON_GEM_DOMAIN_VRAM, - &rdev->uvd.gpu_addr); - if (r) { - radeon_bo_unreserve(rdev->uvd.vcpu_bo); - radeon_bo_unref(&rdev->uvd.vcpu_bo); - dev_err(rdev->dev, "(%d) UVD bo pin failed\n", r); - return r; - } + size = radeon_bo_size(rdev->uvd.vcpu_bo); + size -= rdev->uvd_fw->size; - r = radeon_bo_kmap(rdev->uvd.vcpu_bo, &rdev->uvd.cpu_addr); - if (r) { - dev_err(rdev->dev, "(%d) UVD map failed\n", r); - return r; - } + ptr = rdev->uvd.cpu_addr; + ptr += rdev->uvd_fw->size; - radeon_bo_unreserve(rdev->uvd.vcpu_bo); + if (rdev->uvd.saved_bo != NULL) { + memcpy(ptr, rdev->uvd.saved_bo, size); + kfree(rdev->uvd.saved_bo); + rdev->uvd.saved_bo = NULL; + } else + memset(ptr, 0, size); return 0; } @@ -208,10 +241,12 @@ void radeon_uvd_free_handles(struct radeon_device *rdev, struct drm_file *filp) { int i, r; for (i = 0; i < RADEON_MAX_UVD_HANDLES; ++i) { - if (rdev->uvd.filp[i] == filp) { - uint32_t handle = atomic_read(&rdev->uvd.handles[i]); + uint32_t handle = atomic_read(&rdev->uvd.handles[i]); + if (handle != 0 && rdev->uvd.filp[i] == filp) { struct radeon_fence *fence; + radeon_uvd_note_usage(rdev); + r = radeon_uvd_get_destroy_msg(rdev, R600_RING_TYPE_UVD_INDEX, handle, &fence); if (r) { @@ -320,6 +355,7 @@ static int radeon_uvd_cs_msg(struct radeon_cs_parser *p, struct radeon_bo *bo, unsigned offset, unsigned buf_sizes[]) { int32_t *msg, msg_type, handle; + unsigned img_size = 0; void *ptr; int i, r; @@ -329,9 +365,19 @@ static int radeon_uvd_cs_msg(struct radeon_cs_parser *p, struct radeon_bo *bo, return -EINVAL; } + if (bo->tbo.sync_obj) { + r = radeon_fence_wait(bo->tbo.sync_obj, false); + if (r) { + DRM_ERROR("Failed waiting for UVD message (%d)!\n", r); + return r; + } + } + r = radeon_bo_kmap(bo, &ptr); - if (r) + if (r) { + DRM_ERROR("Failed mapping the UVD message (%d)!\n", r); return r; + } msg = ptr + offset; @@ -346,6 +392,8 @@ static int radeon_uvd_cs_msg(struct radeon_cs_parser *p, struct radeon_bo *bo, if (msg_type == 1) { /* it's a decode msg, calc buffer sizes */ r = radeon_uvd_cs_msg_decode(msg, buf_sizes); + /* calc image size (width * height) */ + img_size = msg[6] * msg[7]; radeon_bo_kunmap(bo); if (r) return r; @@ -357,8 +405,16 @@ static int radeon_uvd_cs_msg(struct radeon_cs_parser *p, struct radeon_bo *bo, radeon_bo_kunmap(bo); return 0; } else { - /* it's a create msg, no special handling needed */ + /* it's a create msg, calc image size (width * height) */ + img_size = msg[7] * msg[8]; radeon_bo_kunmap(bo); + + if (msg_type != 0) { + DRM_ERROR("Illegal UVD message type (%d)!\n", msg_type); + return -EINVAL; + } + + /* it's a create msg, no special handling needed */ } /* create or decode, validate the handle */ @@ -371,6 +427,7 @@ static int radeon_uvd_cs_msg(struct radeon_cs_parser *p, struct radeon_bo *bo, for (i = 0; i < RADEON_MAX_UVD_HANDLES; ++i) { if (!atomic_cmpxchg(&p->rdev->uvd.handles[i], 0, handle)) { p->rdev->uvd.filp[i] = p->filp; + p->rdev->uvd.img_size[i] = img_size; return 0; } } @@ -381,7 +438,7 @@ static int radeon_uvd_cs_msg(struct radeon_cs_parser *p, struct radeon_bo *bo, static int radeon_uvd_cs_reloc(struct radeon_cs_parser *p, int data0, int data1, - unsigned buf_sizes[]) + unsigned buf_sizes[], bool *has_msg_cmd) { struct radeon_cs_chunk *relocs_chunk; struct radeon_cs_reloc *reloc; @@ -399,7 +456,7 @@ static int radeon_uvd_cs_reloc(struct radeon_cs_parser *p, } reloc = p->relocs_ptr[(idx / 4)]; - start = reloc->lobj.gpu_offset; + start = reloc->gpu_offset; end = start + radeon_bo_size(reloc->robj); start += offset; @@ -409,8 +466,12 @@ static int radeon_uvd_cs_reloc(struct radeon_cs_parser *p, cmd = radeon_get_ib_value(p, p->idx) >> 1; if (cmd < 0x4) { + if (end <= start) { + DRM_ERROR("invalid reloc offset %X!\n", offset); + return -EINVAL; + } if ((end - start) < buf_sizes[cmd]) { - DRM_ERROR("buffer to small (%d / %d)!\n", + DRM_ERROR("buffer (%d) to small (%d / %d)!\n", cmd, (unsigned)(end - start), buf_sizes[cmd]); return -EINVAL; } @@ -420,7 +481,7 @@ static int radeon_uvd_cs_reloc(struct radeon_cs_parser *p, return -EINVAL; } - if ((start >> 28) != (end >> 28)) { + if ((start >> 28) != ((end - 1) >> 28)) { DRM_ERROR("reloc %LX-%LX crossing 256MB boundary!\n", start, end); return -EINVAL; @@ -435,9 +496,17 @@ static int radeon_uvd_cs_reloc(struct radeon_cs_parser *p, } if (cmd == 0) { + if (*has_msg_cmd) { + DRM_ERROR("More than one message in a UVD-IB!\n"); + return -EINVAL; + } + *has_msg_cmd = true; r = radeon_uvd_cs_msg(p, reloc->robj, offset, buf_sizes); if (r) return r; + } else if (!*has_msg_cmd) { + DRM_ERROR("Message needed before other commands are send!\n"); + return -EINVAL; } return 0; @@ -446,7 +515,8 @@ static int radeon_uvd_cs_reloc(struct radeon_cs_parser *p, static int radeon_uvd_cs_reg(struct radeon_cs_parser *p, struct radeon_cs_packet *pkt, int *data0, int *data1, - unsigned buf_sizes[]) + unsigned buf_sizes[], + bool *has_msg_cmd) { int i, r; @@ -460,7 +530,8 @@ static int radeon_uvd_cs_reg(struct radeon_cs_parser *p, *data1 = p->idx; break; case UVD_GPCOM_VCPU_CMD: - r = radeon_uvd_cs_reloc(p, *data0, *data1, buf_sizes); + r = radeon_uvd_cs_reloc(p, *data0, *data1, + buf_sizes, has_msg_cmd); if (r) return r; break; @@ -481,6 +552,9 @@ int radeon_uvd_cs_parse(struct radeon_cs_parser *p) struct radeon_cs_packet pkt; int r, data0 = 0, data1 = 0; + /* does the IB has a msg command */ + bool has_msg_cmd = false; + /* minimum buffer sizes */ unsigned buf_sizes[] = { [0x00000000] = 2048, @@ -507,8 +581,8 @@ int radeon_uvd_cs_parse(struct radeon_cs_parser *p) return r; switch (pkt.type) { case RADEON_PACKET_TYPE0: - r = radeon_uvd_cs_reg(p, &pkt, &data0, - &data1, buf_sizes); + r = radeon_uvd_cs_reg(p, &pkt, &data0, &data1, + buf_sizes, &has_msg_cmd); if (r) return r; break; @@ -520,6 +594,12 @@ int radeon_uvd_cs_parse(struct radeon_cs_parser *p) return -EINVAL; } } while (p->idx < p->chunks[p->chunk_ib_idx].length_dw); + + if (!has_msg_cmd) { + DRM_ERROR("UVD-IBs need a msg command!\n"); + return -EINVAL; + } + return 0; } @@ -528,6 +608,7 @@ static int radeon_uvd_send_msg(struct radeon_device *rdev, struct radeon_fence **fence) { struct ttm_validate_buffer tv; + struct ww_acquire_ctx ticket; struct list_head head; struct radeon_ib ib; uint64_t addr; @@ -539,7 +620,7 @@ static int radeon_uvd_send_msg(struct radeon_device *rdev, INIT_LIST_HEAD(&head); list_add(&tv.head, &head); - r = ttm_eu_reserve_buffers(&head); + r = ttm_eu_reserve_buffers(&ticket, &head); if (r) return r; @@ -547,16 +628,12 @@ static int radeon_uvd_send_msg(struct radeon_device *rdev, radeon_uvd_force_into_uvd_segment(bo); r = ttm_bo_validate(&bo->tbo, &bo->placement, true, false); - if (r) { - ttm_eu_backoff_reservation(&head); - return r; - } + if (r) + goto err; - r = radeon_ib_get(rdev, ring, &ib, NULL, 16); - if (r) { - ttm_eu_backoff_reservation(&head); - return r; - } + r = radeon_ib_get(rdev, ring, &ib, NULL, 64); + if (r) + goto err; addr = radeon_bo_gpu_offset(bo); ib.ptr[0] = PACKET0(UVD_GPCOM_VCPU_DATA0, 0); @@ -570,11 +647,9 @@ static int radeon_uvd_send_msg(struct radeon_device *rdev, ib.length_dw = 16; r = radeon_ib_schedule(rdev, &ib, NULL); - if (r) { - ttm_eu_backoff_reservation(&head); - return r; - } - ttm_eu_fence_buffer_objects(&head, ib.fence); + if (r) + goto err; + ttm_eu_fence_buffer_objects(&ticket, &head, ib.fence); if (fence) *fence = radeon_fence_ref(ib.fence); @@ -582,6 +657,10 @@ static int radeon_uvd_send_msg(struct radeon_device *rdev, radeon_ib_free(rdev, &ib); radeon_bo_unref(&bo); return 0; + +err: + ttm_eu_backoff_reservation(&ticket, &head); + return r; } /* multiple fence commands without any stream commands in between can @@ -613,19 +692,19 @@ int radeon_uvd_get_create_msg(struct radeon_device *rdev, int ring, } /* stitch together an UVD create msg */ - msg[0] = 0x00000de4; - msg[1] = 0x00000000; - msg[2] = handle; - msg[3] = 0x00000000; - msg[4] = 0x00000000; - msg[5] = 0x00000000; - msg[6] = 0x00000000; - msg[7] = 0x00000780; - msg[8] = 0x00000440; - msg[9] = 0x00000000; - msg[10] = 0x01b37000; + msg[0] = cpu_to_le32(0x00000de4); + msg[1] = cpu_to_le32(0x00000000); + msg[2] = cpu_to_le32(handle); + msg[3] = cpu_to_le32(0x00000000); + msg[4] = cpu_to_le32(0x00000000); + msg[5] = cpu_to_le32(0x00000000); + msg[6] = cpu_to_le32(0x00000000); + msg[7] = cpu_to_le32(0x00000780); + msg[8] = cpu_to_le32(0x00000440); + msg[9] = cpu_to_le32(0x00000000); + msg[10] = cpu_to_le32(0x01b37000); for (i = 11; i < 1024; ++i) - msg[i] = 0x0; + msg[i] = cpu_to_le32(0x0); radeon_bo_kunmap(bo); radeon_bo_unreserve(bo); @@ -659,12 +738,12 @@ int radeon_uvd_get_destroy_msg(struct radeon_device *rdev, int ring, } /* stitch together an UVD destroy msg */ - msg[0] = 0x00000de4; - msg[1] = 0x00000002; - msg[2] = handle; - msg[3] = 0x00000000; + msg[0] = cpu_to_le32(0x00000de4); + msg[1] = cpu_to_le32(0x00000002); + msg[2] = cpu_to_le32(handle); + msg[3] = cpu_to_le32(0x00000000); for (i = 4; i < 1024; ++i) - msg[i] = 0x0; + msg[i] = cpu_to_le32(0x0); radeon_bo_kunmap(bo); radeon_bo_unreserve(bo); @@ -672,25 +751,79 @@ int radeon_uvd_get_destroy_msg(struct radeon_device *rdev, int ring, return radeon_uvd_send_msg(rdev, ring, bo, fence); } +/** + * radeon_uvd_count_handles - count number of open streams + * + * @rdev: radeon_device pointer + * @sd: number of SD streams + * @hd: number of HD streams + * + * Count the number of open SD/HD streams as a hint for power mangement + */ +static void radeon_uvd_count_handles(struct radeon_device *rdev, + unsigned *sd, unsigned *hd) +{ + unsigned i; + + *sd = 0; + *hd = 0; + + for (i = 0; i < RADEON_MAX_UVD_HANDLES; ++i) { + if (!atomic_read(&rdev->uvd.handles[i])) + continue; + + if (rdev->uvd.img_size[i] >= 720*576) + ++(*hd); + else + ++(*sd); + } +} + static void radeon_uvd_idle_work_handler(struct work_struct *work) { struct radeon_device *rdev = container_of(work, struct radeon_device, uvd.idle_work.work); - if (radeon_fence_count_emitted(rdev, R600_RING_TYPE_UVD_INDEX) == 0) - radeon_set_uvd_clocks(rdev, 0, 0); - else + if (radeon_fence_count_emitted(rdev, R600_RING_TYPE_UVD_INDEX) == 0) { + if ((rdev->pm.pm_method == PM_METHOD_DPM) && rdev->pm.dpm_enabled) { + radeon_uvd_count_handles(rdev, &rdev->pm.dpm.sd, + &rdev->pm.dpm.hd); + radeon_dpm_enable_uvd(rdev, false); + } else { + radeon_set_uvd_clocks(rdev, 0, 0); + } + } else { schedule_delayed_work(&rdev->uvd.idle_work, msecs_to_jiffies(UVD_IDLE_TIMEOUT_MS)); + } } void radeon_uvd_note_usage(struct radeon_device *rdev) { + bool streams_changed = false; bool set_clocks = !cancel_delayed_work_sync(&rdev->uvd.idle_work); set_clocks &= schedule_delayed_work(&rdev->uvd.idle_work, msecs_to_jiffies(UVD_IDLE_TIMEOUT_MS)); - if (set_clocks) - radeon_set_uvd_clocks(rdev, 53300, 40000); + + if ((rdev->pm.pm_method == PM_METHOD_DPM) && rdev->pm.dpm_enabled) { + unsigned hd = 0, sd = 0; + radeon_uvd_count_handles(rdev, &sd, &hd); + if ((rdev->pm.dpm.sd != sd) || + (rdev->pm.dpm.hd != hd)) { + rdev->pm.dpm.sd = sd; + rdev->pm.dpm.hd = hd; + /* disable this for now */ + /*streams_changed = true;*/ + } + } + + if (set_clocks || streams_changed) { + if ((rdev->pm.pm_method == PM_METHOD_DPM) && rdev->pm.dpm_enabled) { + radeon_dpm_enable_uvd(rdev, true); + } else { + radeon_set_uvd_clocks(rdev, 53300, 40000); + } + } } static unsigned radeon_uvd_calc_upll_post_div(unsigned vco_freq, |
