diff options
Diffstat (limited to 'drivers/gpu/drm/amd/amdgpu/amdgpu_vcn.c')
| -rw-r--r-- | drivers/gpu/drm/amd/amdgpu/amdgpu_vcn.c | 1552 |
1 files changed, 1168 insertions, 384 deletions
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_vcn.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_vcn.c index ecf6f96df2ad..5e0786ea911b 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_vcn.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_vcn.c @@ -1,5 +1,5 @@ /* - * Copyright 2016 Advanced Micro Devices, Inc. + * Copyright 2016-2024 Advanced Micro Devices, Inc. * All Rights Reserved. * * Permission is hereby granted, free of charge, to any person obtaining a @@ -26,72 +26,147 @@ #include <linux/firmware.h> #include <linux/module.h> -#include <drm/drmP.h> -#include <drm/drm.h> +#include <linux/dmi.h> +#include <linux/pci.h> +#include <linux/debugfs.h> +#include <drm/drm_drv.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" -#define FIRMWARE_PICASSO "amdgpu/picasso_vcn.bin" -#define FIRMWARE_RAVEN2 "amdgpu/raven2_vcn.bin" +#define FIRMWARE_RAVEN "amdgpu/raven_vcn.bin" +#define FIRMWARE_PICASSO "amdgpu/picasso_vcn.bin" +#define FIRMWARE_RAVEN2 "amdgpu/raven2_vcn.bin" +#define FIRMWARE_ARCTURUS "amdgpu/arcturus_vcn.bin" +#define FIRMWARE_RENOIR "amdgpu/renoir_vcn.bin" +#define FIRMWARE_GREEN_SARDINE "amdgpu/green_sardine_vcn.bin" +#define FIRMWARE_NAVI10 "amdgpu/navi10_vcn.bin" +#define FIRMWARE_NAVI14 "amdgpu/navi14_vcn.bin" +#define FIRMWARE_NAVI12 "amdgpu/navi12_vcn.bin" +#define FIRMWARE_SIENNA_CICHLID "amdgpu/sienna_cichlid_vcn.bin" +#define FIRMWARE_NAVY_FLOUNDER "amdgpu/navy_flounder_vcn.bin" +#define FIRMWARE_VANGOGH "amdgpu/vangogh_vcn.bin" +#define FIRMWARE_DIMGREY_CAVEFISH "amdgpu/dimgrey_cavefish_vcn.bin" +#define FIRMWARE_ALDEBARAN "amdgpu/aldebaran_vcn.bin" +#define FIRMWARE_BEIGE_GOBY "amdgpu/beige_goby_vcn.bin" +#define FIRMWARE_YELLOW_CARP "amdgpu/yellow_carp_vcn.bin" +#define FIRMWARE_VCN_3_1_2 "amdgpu/vcn_3_1_2.bin" +#define FIRMWARE_VCN4_0_0 "amdgpu/vcn_4_0_0.bin" +#define FIRMWARE_VCN4_0_2 "amdgpu/vcn_4_0_2.bin" +#define FIRMWARE_VCN4_0_3 "amdgpu/vcn_4_0_3.bin" +#define FIRMWARE_VCN4_0_4 "amdgpu/vcn_4_0_4.bin" +#define FIRMWARE_VCN4_0_5 "amdgpu/vcn_4_0_5.bin" +#define FIRMWARE_VCN4_0_6 "amdgpu/vcn_4_0_6.bin" +#define FIRMWARE_VCN4_0_6_1 "amdgpu/vcn_4_0_6_1.bin" +#define FIRMWARE_VCN5_0_0 "amdgpu/vcn_5_0_0.bin" +#define FIRMWARE_VCN5_0_1 "amdgpu/vcn_5_0_1.bin" MODULE_FIRMWARE(FIRMWARE_RAVEN); MODULE_FIRMWARE(FIRMWARE_PICASSO); MODULE_FIRMWARE(FIRMWARE_RAVEN2); +MODULE_FIRMWARE(FIRMWARE_ARCTURUS); +MODULE_FIRMWARE(FIRMWARE_RENOIR); +MODULE_FIRMWARE(FIRMWARE_GREEN_SARDINE); +MODULE_FIRMWARE(FIRMWARE_ALDEBARAN); +MODULE_FIRMWARE(FIRMWARE_NAVI10); +MODULE_FIRMWARE(FIRMWARE_NAVI14); +MODULE_FIRMWARE(FIRMWARE_NAVI12); +MODULE_FIRMWARE(FIRMWARE_SIENNA_CICHLID); +MODULE_FIRMWARE(FIRMWARE_NAVY_FLOUNDER); +MODULE_FIRMWARE(FIRMWARE_VANGOGH); +MODULE_FIRMWARE(FIRMWARE_DIMGREY_CAVEFISH); +MODULE_FIRMWARE(FIRMWARE_BEIGE_GOBY); +MODULE_FIRMWARE(FIRMWARE_YELLOW_CARP); +MODULE_FIRMWARE(FIRMWARE_VCN_3_1_2); +MODULE_FIRMWARE(FIRMWARE_VCN4_0_0); +MODULE_FIRMWARE(FIRMWARE_VCN4_0_2); +MODULE_FIRMWARE(FIRMWARE_VCN4_0_3); +MODULE_FIRMWARE(FIRMWARE_VCN4_0_4); +MODULE_FIRMWARE(FIRMWARE_VCN4_0_5); +MODULE_FIRMWARE(FIRMWARE_VCN4_0_6); +MODULE_FIRMWARE(FIRMWARE_VCN4_0_6_1); +MODULE_FIRMWARE(FIRMWARE_VCN5_0_0); +MODULE_FIRMWARE(FIRMWARE_VCN5_0_1); static void amdgpu_vcn_idle_work_handler(struct work_struct *work); +static void amdgpu_vcn_reg_dump_fini(struct amdgpu_device *adev); -int amdgpu_vcn_sw_init(struct amdgpu_device *adev) +int amdgpu_vcn_early_init(struct amdgpu_device *adev, int i) { - unsigned long bo_size; - const char *fw_name; - const struct common_firmware_header *hdr; - unsigned char fw_check; + char ucode_prefix[25]; int r; - INIT_DELAYED_WORK(&adev->vcn.idle_work, amdgpu_vcn_idle_work_handler); + adev->vcn.inst[i].adev = adev; + adev->vcn.inst[i].inst = i; + amdgpu_ucode_ip_version_decode(adev, UVD_HWIP, ucode_prefix, sizeof(ucode_prefix)); - switch (adev->asic_type) { - case CHIP_RAVEN: - if (adev->rev_id >= 8) - fw_name = FIRMWARE_RAVEN2; - else if (adev->pdev->device == 0x15d8) - fw_name = FIRMWARE_PICASSO; - else - fw_name = FIRMWARE_RAVEN; - break; - default: - return -EINVAL; + if (i != 0 && adev->vcn.per_inst_fw) { + r = amdgpu_ucode_request(adev, &adev->vcn.inst[i].fw, + AMDGPU_UCODE_REQUIRED, + "amdgpu/%s_%d.bin", ucode_prefix, i); + if (r) + amdgpu_ucode_release(&adev->vcn.inst[i].fw); + } else { + if (!adev->vcn.inst[0].fw) { + r = amdgpu_ucode_request(adev, &adev->vcn.inst[0].fw, + AMDGPU_UCODE_REQUIRED, + "amdgpu/%s.bin", ucode_prefix); + if (r) + amdgpu_ucode_release(&adev->vcn.inst[0].fw); + } else { + r = 0; + } + adev->vcn.inst[i].fw = adev->vcn.inst[0].fw; } - r = request_firmware(&adev->vcn.fw, fw_name, adev->dev); - if (r) { - dev_err(adev->dev, "amdgpu_vcn: Can't load firmware \"%s\"\n", - fw_name); - return r; - } + return r; +} - r = amdgpu_ucode_validate(adev->vcn.fw); - if (r) { - dev_err(adev->dev, "amdgpu_vcn: Can't validate firmware \"%s\"\n", - fw_name); - release_firmware(adev->vcn.fw); - adev->vcn.fw = NULL; - return r; +int amdgpu_vcn_sw_init(struct amdgpu_device *adev, int i) +{ + unsigned long bo_size; + const struct common_firmware_header *hdr; + unsigned char fw_check; + unsigned int fw_shared_size, log_offset; + int r; + + mutex_init(&adev->vcn.inst[i].vcn1_jpeg1_workaround); + mutex_init(&adev->vcn.inst[i].vcn_pg_lock); + mutex_init(&adev->vcn.inst[i].engine_reset_mutex); + atomic_set(&adev->vcn.inst[i].total_submission_cnt, 0); + INIT_DELAYED_WORK(&adev->vcn.inst[i].idle_work, amdgpu_vcn_idle_work_handler); + atomic_set(&adev->vcn.inst[i].dpg_enc_submission_cnt, 0); + if ((adev->firmware.load_type == AMDGPU_FW_LOAD_PSP) && + (adev->pg_flags & AMD_PG_SUPPORT_VCN_DPG)) + adev->vcn.inst[i].indirect_sram = true; + + /* + * Some Steam Deck's BIOS versions are incompatible with the + * indirect SRAM mode, leading to amdgpu being unable to get + * properly probed (and even potentially crashing the kernel). + * Hence, check for these versions here - notice this is + * restricted to Vangogh (Deck's APU). + */ + if (amdgpu_ip_version(adev, UVD_HWIP, 0) == IP_VERSION(3, 0, 2)) { + const char *bios_ver = dmi_get_system_info(DMI_BIOS_VERSION); + + if (bios_ver && (!strncmp("F7A0113", bios_ver, 7) || + !strncmp("F7A0114", bios_ver, 7))) { + adev->vcn.inst[i].indirect_sram = false; + dev_info(adev->dev, + "Steam Deck quirk: indirect SRAM disabled on BIOS %s\n", bios_ver); + } } - hdr = (const struct common_firmware_header *)adev->vcn.fw->data; + /* from vcn4 and above, only unified queue is used */ + adev->vcn.inst[i].using_unified_queue = + amdgpu_ip_version(adev, UVD_HWIP, 0) >= IP_VERSION(4, 0, 0); + + hdr = (const struct common_firmware_header *)adev->vcn.inst[i].fw->data; + adev->vcn.inst[i].fw_version = le32_to_cpu(hdr->ucode_version); adev->vcn.fw_version = le32_to_cpu(hdr->ucode_version); /* Bit 20-23, it is encode major and non-zero for new naming convention. @@ -109,100 +184,219 @@ int amdgpu_vcn_sw_init(struct amdgpu_device *adev) enc_major = fw_check; dec_ver = (le32_to_cpu(hdr->ucode_version) >> 24) & 0xf; vep = (le32_to_cpu(hdr->ucode_version) >> 28) & 0xf; - DRM_INFO("Found VCN firmware Version ENC: %hu.%hu DEC: %hu VEP: %hu Revision: %hu\n", - enc_major, enc_minor, dec_ver, vep, fw_rev); + dev_info(adev->dev, + "[VCN instance %d] Found VCN firmware Version ENC: %u.%u DEC: %u VEP: %u Revision: %u\n", + i, enc_major, enc_minor, dec_ver, vep, fw_rev); } else { unsigned int version_major, version_minor, family_id; family_id = le32_to_cpu(hdr->ucode_version) & 0xff; version_major = (le32_to_cpu(hdr->ucode_version) >> 24) & 0xff; version_minor = (le32_to_cpu(hdr->ucode_version) >> 8) & 0xff; - DRM_INFO("Found VCN firmware Version: %hu.%hu Family ID: %hu\n", - version_major, version_minor, family_id); + dev_info(adev->dev, "[VCN instance %d] Found VCN firmware Version: %u.%u Family ID: %u\n", + i, version_major, version_minor, family_id); } bo_size = AMDGPU_VCN_STACK_SIZE + AMDGPU_VCN_CONTEXT_SIZE; if (adev->firmware.load_type != AMDGPU_FW_LOAD_PSP) bo_size += AMDGPU_GPU_PAGE_ALIGN(le32_to_cpu(hdr->ucode_size_bytes) + 8); + + if (amdgpu_ip_version(adev, UVD_HWIP, 0) >= IP_VERSION(5, 0, 0)) { + fw_shared_size = AMDGPU_GPU_PAGE_ALIGN(sizeof(struct amdgpu_vcn5_fw_shared)); + log_offset = offsetof(struct amdgpu_vcn5_fw_shared, fw_log); + } else if (amdgpu_ip_version(adev, UVD_HWIP, 0) >= IP_VERSION(4, 0, 0)) { + fw_shared_size = AMDGPU_GPU_PAGE_ALIGN(sizeof(struct amdgpu_vcn4_fw_shared)); + log_offset = offsetof(struct amdgpu_vcn4_fw_shared, fw_log); + } else { + fw_shared_size = AMDGPU_GPU_PAGE_ALIGN(sizeof(struct amdgpu_fw_shared)); + log_offset = offsetof(struct amdgpu_fw_shared, fw_log); + } + + bo_size += fw_shared_size; + + if (amdgpu_vcnfw_log) + bo_size += AMDGPU_VCNFW_LOG_SIZE; + r = amdgpu_bo_create_kernel(adev, bo_size, PAGE_SIZE, - AMDGPU_GEM_DOMAIN_VRAM, &adev->vcn.vcpu_bo, - &adev->vcn.gpu_addr, &adev->vcn.cpu_addr); + AMDGPU_GEM_DOMAIN_VRAM | + AMDGPU_GEM_DOMAIN_GTT, + &adev->vcn.inst[i].vcpu_bo, + &adev->vcn.inst[i].gpu_addr, + &adev->vcn.inst[i].cpu_addr); if (r) { dev_err(adev->dev, "(%d) failed to allocate vcn bo\n", r); return r; } + adev->vcn.inst[i].fw_shared.cpu_addr = adev->vcn.inst[i].cpu_addr + + bo_size - fw_shared_size; + adev->vcn.inst[i].fw_shared.gpu_addr = adev->vcn.inst[i].gpu_addr + + bo_size - fw_shared_size; + + adev->vcn.inst[i].fw_shared.mem_size = fw_shared_size; + + if (amdgpu_vcnfw_log) { + adev->vcn.inst[i].fw_shared.cpu_addr -= AMDGPU_VCNFW_LOG_SIZE; + adev->vcn.inst[i].fw_shared.gpu_addr -= AMDGPU_VCNFW_LOG_SIZE; + adev->vcn.inst[i].fw_shared.log_offset = log_offset; + } + + if (adev->vcn.inst[i].indirect_sram) { + r = amdgpu_bo_create_kernel(adev, 64 * 2 * 4, PAGE_SIZE, + AMDGPU_GEM_DOMAIN_VRAM | + AMDGPU_GEM_DOMAIN_GTT, + &adev->vcn.inst[i].dpg_sram_bo, + &adev->vcn.inst[i].dpg_sram_gpu_addr, + &adev->vcn.inst[i].dpg_sram_cpu_addr); + if (r) { + dev_err(adev->dev, "VCN %d (%d) failed to allocate DPG bo\n", i, r); + return r; + } + } + return 0; } -int amdgpu_vcn_sw_fini(struct amdgpu_device *adev) +void amdgpu_vcn_sw_fini(struct amdgpu_device *adev, int i) { - int i; + int j; - kvfree(adev->vcn.saved_bo); + if (adev->vcn.harvest_config & (1 << i)) + return; - amdgpu_bo_free_kernel(&adev->vcn.vcpu_bo, - &adev->vcn.gpu_addr, - (void **)&adev->vcn.cpu_addr); + amdgpu_bo_free_kernel( + &adev->vcn.inst[i].dpg_sram_bo, + &adev->vcn.inst[i].dpg_sram_gpu_addr, + (void **)&adev->vcn.inst[i].dpg_sram_cpu_addr); - amdgpu_ring_fini(&adev->vcn.ring_dec); + kvfree(adev->vcn.inst[i].saved_bo); - for (i = 0; i < adev->vcn.num_enc_rings; ++i) - amdgpu_ring_fini(&adev->vcn.ring_enc[i]); + amdgpu_bo_free_kernel(&adev->vcn.inst[i].vcpu_bo, + &adev->vcn.inst[i].gpu_addr, + (void **)&adev->vcn.inst[i].cpu_addr); - amdgpu_ring_fini(&adev->vcn.ring_jpeg); + amdgpu_ring_fini(&adev->vcn.inst[i].ring_dec); - release_firmware(adev->vcn.fw); + for (j = 0; j < adev->vcn.inst[i].num_enc_rings; ++j) + amdgpu_ring_fini(&adev->vcn.inst[i].ring_enc[j]); - return 0; + if (adev->vcn.per_inst_fw) { + amdgpu_ucode_release(&adev->vcn.inst[i].fw); + } else { + amdgpu_ucode_release(&adev->vcn.inst[0].fw); + adev->vcn.inst[i].fw = NULL; + } + + if (adev->vcn.reg_list) + amdgpu_vcn_reg_dump_fini(adev); + + mutex_destroy(&adev->vcn.inst[i].vcn_pg_lock); + mutex_destroy(&adev->vcn.inst[i].vcn1_jpeg1_workaround); } -int amdgpu_vcn_suspend(struct amdgpu_device *adev) +bool amdgpu_vcn_is_disabled_vcn(struct amdgpu_device *adev, enum vcn_ring_type type, uint32_t vcn_instance) { - unsigned size; - void *ptr; + bool ret = false; + int vcn_config = adev->vcn.inst[vcn_instance].vcn_config; + + if ((type == VCN_ENCODE_RING) && (vcn_config & VCN_BLOCK_ENCODE_DISABLE_MASK)) + ret = true; + else if ((type == VCN_DECODE_RING) && (vcn_config & VCN_BLOCK_DECODE_DISABLE_MASK)) + ret = true; + else if ((type == VCN_UNIFIED_RING) && (vcn_config & VCN_BLOCK_QUEUE_DISABLE_MASK)) + ret = true; - cancel_delayed_work_sync(&adev->vcn.idle_work); + return ret; +} + +static int amdgpu_vcn_save_vcpu_bo_inst(struct amdgpu_device *adev, int i) +{ + unsigned int size; + void *ptr; + int idx; - if (adev->vcn.vcpu_bo == NULL) + if (adev->vcn.harvest_config & (1 << i)) + return 0; + if (adev->vcn.inst[i].vcpu_bo == NULL) return 0; - size = amdgpu_bo_size(adev->vcn.vcpu_bo); - ptr = adev->vcn.cpu_addr; + size = amdgpu_bo_size(adev->vcn.inst[i].vcpu_bo); + ptr = adev->vcn.inst[i].cpu_addr; - adev->vcn.saved_bo = kvmalloc(size, GFP_KERNEL); - if (!adev->vcn.saved_bo) + adev->vcn.inst[i].saved_bo = kvmalloc(size, GFP_KERNEL); + if (!adev->vcn.inst[i].saved_bo) return -ENOMEM; - memcpy_fromio(adev->vcn.saved_bo, ptr, size); + if (drm_dev_enter(adev_to_drm(adev), &idx)) { + memcpy_fromio(adev->vcn.inst[i].saved_bo, ptr, size); + drm_dev_exit(idx); + } return 0; } -int amdgpu_vcn_resume(struct amdgpu_device *adev) +int amdgpu_vcn_save_vcpu_bo(struct amdgpu_device *adev) { - unsigned size; + int ret, i; + + for (i = 0; i < adev->vcn.num_vcn_inst; ++i) { + ret = amdgpu_vcn_save_vcpu_bo_inst(adev, i); + if (ret) + return ret; + } + + return 0; +} + +int amdgpu_vcn_suspend(struct amdgpu_device *adev, int i) +{ + bool in_ras_intr = amdgpu_ras_intr_triggered(); + + if (adev->vcn.harvest_config & (1 << i)) + return 0; + + /* err_event_athub and dpc recovery will corrupt VCPU buffer, so we need to + * restore fw data and clear buffer in amdgpu_vcn_resume() */ + if (in_ras_intr || adev->pcie_reset_ctx.in_link_reset) + return 0; + + return amdgpu_vcn_save_vcpu_bo_inst(adev, i); +} + +int amdgpu_vcn_resume(struct amdgpu_device *adev, int i) +{ + unsigned int size; void *ptr; + int idx; - if (adev->vcn.vcpu_bo == NULL) + if (adev->vcn.harvest_config & (1 << i)) + return 0; + if (adev->vcn.inst[i].vcpu_bo == NULL) return -EINVAL; - size = amdgpu_bo_size(adev->vcn.vcpu_bo); - ptr = adev->vcn.cpu_addr; + size = amdgpu_bo_size(adev->vcn.inst[i].vcpu_bo); + ptr = adev->vcn.inst[i].cpu_addr; - if (adev->vcn.saved_bo != NULL) { - memcpy_toio(ptr, adev->vcn.saved_bo, size); - kvfree(adev->vcn.saved_bo); - adev->vcn.saved_bo = NULL; + if (adev->vcn.inst[i].saved_bo != NULL) { + if (drm_dev_enter(adev_to_drm(adev), &idx)) { + memcpy_toio(ptr, adev->vcn.inst[i].saved_bo, size); + drm_dev_exit(idx); + } + kvfree(adev->vcn.inst[i].saved_bo); + adev->vcn.inst[i].saved_bo = NULL; } else { const struct common_firmware_header *hdr; - unsigned offset; + unsigned int offset; - hdr = (const struct common_firmware_header *)adev->vcn.fw->data; + hdr = (const struct common_firmware_header *)adev->vcn.inst[i].fw->data; if (adev->firmware.load_type != AMDGPU_FW_LOAD_PSP) { offset = le32_to_cpu(hdr->ucode_array_offset_bytes); - memcpy_toio(adev->vcn.cpu_addr, adev->vcn.fw->data + offset, - le32_to_cpu(hdr->ucode_size_bytes)); + if (drm_dev_enter(adev_to_drm(adev), &idx)) { + memcpy_toio(adev->vcn.inst[i].cpu_addr, + adev->vcn.inst[i].fw->data + offset, + le32_to_cpu(hdr->ucode_size_bytes)); + drm_dev_exit(idx); + } size -= le32_to_cpu(hdr->ucode_size_bytes); ptr += le32_to_cpu(hdr->ucode_size_bytes); } @@ -212,241 +406,205 @@ int amdgpu_vcn_resume(struct amdgpu_device *adev) return 0; } -static int amdgpu_vcn_pause_dpg_mode(struct amdgpu_device *adev, - struct dpg_pause_state *new_state) +void amdgpu_vcn_get_profile(struct amdgpu_device *adev) { - int ret_code; - uint32_t reg_data = 0; - uint32_t reg_data2 = 0; - struct amdgpu_ring *ring; + int r; - /* pause/unpause if state is changed */ - if (adev->vcn.pause_state.fw_based != new_state->fw_based) { - DRM_DEBUG("dpg pause state changed %d:%d -> %d:%d", - adev->vcn.pause_state.fw_based, adev->vcn.pause_state.jpeg, - new_state->fw_based, new_state->jpeg); - - reg_data = RREG32_SOC15(UVD, 0, mmUVD_DPG_PAUSE) & - (~UVD_DPG_PAUSE__NJ_PAUSE_DPG_ACK_MASK); - - if (new_state->fw_based == VCN_DPG_STATE__PAUSE) { - ret_code = 0; - - if (!(reg_data & UVD_DPG_PAUSE__JPEG_PAUSE_DPG_ACK_MASK)) - SOC15_WAIT_ON_RREG(UVD, 0, mmUVD_POWER_STATUS, - UVD_POWER_STATUS__UVD_POWER_STATUS_TILES_OFF, - UVD_POWER_STATUS__UVD_POWER_STATUS_MASK, ret_code); - - if (!ret_code) { - /* pause DPG non-jpeg */ - reg_data |= UVD_DPG_PAUSE__NJ_PAUSE_DPG_REQ_MASK; - WREG32_SOC15(UVD, 0, mmUVD_DPG_PAUSE, reg_data); - SOC15_WAIT_ON_RREG(UVD, 0, mmUVD_DPG_PAUSE, - UVD_DPG_PAUSE__NJ_PAUSE_DPG_ACK_MASK, - UVD_DPG_PAUSE__NJ_PAUSE_DPG_ACK_MASK, ret_code); - - /* Restore */ - ring = &adev->vcn.ring_enc[0]; - WREG32_SOC15(UVD, 0, mmUVD_RB_BASE_LO, ring->gpu_addr); - WREG32_SOC15(UVD, 0, mmUVD_RB_BASE_HI, upper_32_bits(ring->gpu_addr)); - WREG32_SOC15(UVD, 0, mmUVD_RB_SIZE, ring->ring_size / 4); - WREG32_SOC15(UVD, 0, mmUVD_RB_RPTR, lower_32_bits(ring->wptr)); - WREG32_SOC15(UVD, 0, mmUVD_RB_WPTR, lower_32_bits(ring->wptr)); - - ring = &adev->vcn.ring_enc[1]; - WREG32_SOC15(UVD, 0, mmUVD_RB_BASE_LO2, ring->gpu_addr); - WREG32_SOC15(UVD, 0, mmUVD_RB_BASE_HI2, upper_32_bits(ring->gpu_addr)); - WREG32_SOC15(UVD, 0, mmUVD_RB_SIZE2, ring->ring_size / 4); - WREG32_SOC15(UVD, 0, mmUVD_RB_RPTR2, lower_32_bits(ring->wptr)); - WREG32_SOC15(UVD, 0, mmUVD_RB_WPTR2, lower_32_bits(ring->wptr)); - - ring = &adev->vcn.ring_dec; - WREG32_SOC15(UVD, 0, mmUVD_RBC_RB_WPTR, - RREG32_SOC15(UVD, 0, mmUVD_SCRATCH2) & 0x7FFFFFFF); - SOC15_WAIT_ON_RREG(UVD, 0, mmUVD_POWER_STATUS, - UVD_PGFSM_CONFIG__UVDM_UVDU_PWR_ON, - UVD_POWER_STATUS__UVD_POWER_STATUS_MASK, ret_code); - } - } else { - /* unpause dpg non-jpeg, no need to wait */ - reg_data &= ~UVD_DPG_PAUSE__NJ_PAUSE_DPG_REQ_MASK; - WREG32_SOC15(UVD, 0, mmUVD_DPG_PAUSE, reg_data); - } - adev->vcn.pause_state.fw_based = new_state->fw_based; - } - - /* pause/unpause if state is changed */ - if (adev->vcn.pause_state.jpeg != new_state->jpeg) { - DRM_DEBUG("dpg pause state changed %d:%d -> %d:%d", - adev->vcn.pause_state.fw_based, adev->vcn.pause_state.jpeg, - new_state->fw_based, new_state->jpeg); - - reg_data = RREG32_SOC15(UVD, 0, mmUVD_DPG_PAUSE) & - (~UVD_DPG_PAUSE__JPEG_PAUSE_DPG_ACK_MASK); - - if (new_state->jpeg == VCN_DPG_STATE__PAUSE) { - ret_code = 0; - - if (!(reg_data & UVD_DPG_PAUSE__NJ_PAUSE_DPG_ACK_MASK)) - SOC15_WAIT_ON_RREG(UVD, 0, mmUVD_POWER_STATUS, - UVD_POWER_STATUS__UVD_POWER_STATUS_TILES_OFF, - UVD_POWER_STATUS__UVD_POWER_STATUS_MASK, ret_code); - - if (!ret_code) { - /* Make sure JPRG Snoop is disabled before sending the pause */ - reg_data2 = RREG32_SOC15(UVD, 0, mmUVD_POWER_STATUS); - reg_data2 |= UVD_POWER_STATUS__JRBC_SNOOP_DIS_MASK; - WREG32_SOC15(UVD, 0, mmUVD_POWER_STATUS, reg_data2); - - /* pause DPG jpeg */ - reg_data |= UVD_DPG_PAUSE__JPEG_PAUSE_DPG_REQ_MASK; - WREG32_SOC15(UVD, 0, mmUVD_DPG_PAUSE, reg_data); - SOC15_WAIT_ON_RREG(UVD, 0, mmUVD_DPG_PAUSE, - UVD_DPG_PAUSE__JPEG_PAUSE_DPG_ACK_MASK, - UVD_DPG_PAUSE__JPEG_PAUSE_DPG_ACK_MASK, ret_code); - - /* Restore */ - ring = &adev->vcn.ring_jpeg; - WREG32_SOC15(UVD, 0, mmUVD_LMI_JRBC_RB_VMID, 0); - WREG32_SOC15(UVD, 0, mmUVD_JRBC_RB_CNTL, - UVD_JRBC_RB_CNTL__RB_NO_FETCH_MASK | - UVD_JRBC_RB_CNTL__RB_RPTR_WR_EN_MASK); - WREG32_SOC15(UVD, 0, mmUVD_LMI_JRBC_RB_64BIT_BAR_LOW, - lower_32_bits(ring->gpu_addr)); - WREG32_SOC15(UVD, 0, mmUVD_LMI_JRBC_RB_64BIT_BAR_HIGH, - upper_32_bits(ring->gpu_addr)); - WREG32_SOC15(UVD, 0, mmUVD_JRBC_RB_RPTR, ring->wptr); - WREG32_SOC15(UVD, 0, mmUVD_JRBC_RB_WPTR, ring->wptr); - WREG32_SOC15(UVD, 0, mmUVD_JRBC_RB_CNTL, - UVD_JRBC_RB_CNTL__RB_RPTR_WR_EN_MASK); - - ring = &adev->vcn.ring_dec; - WREG32_SOC15(UVD, 0, mmUVD_RBC_RB_WPTR, - RREG32_SOC15(UVD, 0, mmUVD_SCRATCH2) & 0x7FFFFFFF); - SOC15_WAIT_ON_RREG(UVD, 0, mmUVD_POWER_STATUS, - UVD_PGFSM_CONFIG__UVDM_UVDU_PWR_ON, - UVD_POWER_STATUS__UVD_POWER_STATUS_MASK, ret_code); - } - } else { - /* unpause dpg jpeg, no need to wait */ - reg_data &= ~UVD_DPG_PAUSE__JPEG_PAUSE_DPG_REQ_MASK; - WREG32_SOC15(UVD, 0, mmUVD_DPG_PAUSE, reg_data); + mutex_lock(&adev->vcn.workload_profile_mutex); + + if (adev->vcn.workload_profile_active) { + mutex_unlock(&adev->vcn.workload_profile_mutex); + return; + } + r = amdgpu_dpm_switch_power_profile(adev, PP_SMC_POWER_PROFILE_VIDEO, + true); + if (r) + dev_warn(adev->dev, + "(%d) failed to enable video power profile mode\n", r); + else + adev->vcn.workload_profile_active = true; + mutex_unlock(&adev->vcn.workload_profile_mutex); +} + +void amdgpu_vcn_put_profile(struct amdgpu_device *adev) +{ + bool pg = true; + int r, i; + + mutex_lock(&adev->vcn.workload_profile_mutex); + for (i = 0; i < adev->vcn.num_vcn_inst; i++) { + if (adev->vcn.inst[i].cur_state != AMD_PG_STATE_GATE) { + pg = false; + break; } - adev->vcn.pause_state.jpeg = new_state->jpeg; } - return 0; + if (pg) { + r = amdgpu_dpm_switch_power_profile( + adev, PP_SMC_POWER_PROFILE_VIDEO, false); + if (r) + dev_warn( + adev->dev, + "(%d) failed to disable video power profile mode\n", + r); + else + adev->vcn.workload_profile_active = false; + } + + mutex_unlock(&adev->vcn.workload_profile_mutex); } static void amdgpu_vcn_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; - unsigned int i; + struct amdgpu_vcn_inst *vcn_inst = + container_of(work, struct amdgpu_vcn_inst, idle_work.work); + struct amdgpu_device *adev = vcn_inst->adev; + unsigned int fences = 0, fence[AMDGPU_MAX_VCN_INSTANCES] = {0}; + unsigned int i = vcn_inst->inst, j; - for (i = 0; i < adev->vcn.num_enc_rings; ++i) { - fences += amdgpu_fence_count_emitted(&adev->vcn.ring_enc[i]); - } + if (adev->vcn.harvest_config & (1 << i)) + return; + + for (j = 0; j < adev->vcn.inst[i].num_enc_rings; ++j) + fence[i] += amdgpu_fence_count_emitted(&vcn_inst->ring_enc[j]); - if (adev->pg_flags & AMD_PG_SUPPORT_VCN_DPG) { + /* Only set DPG pause for VCN3 or below, VCN4 and above will be handled by FW */ + if (adev->pg_flags & AMD_PG_SUPPORT_VCN_DPG && + !adev->vcn.inst[i].using_unified_queue) { struct dpg_pause_state new_state; - if (fences) + if (fence[i] || + unlikely(atomic_read(&vcn_inst->dpg_enc_submission_cnt))) new_state.fw_based = VCN_DPG_STATE__PAUSE; else new_state.fw_based = VCN_DPG_STATE__UNPAUSE; - if (amdgpu_fence_count_emitted(&adev->vcn.ring_jpeg)) - new_state.jpeg = VCN_DPG_STATE__PAUSE; - else - new_state.jpeg = VCN_DPG_STATE__UNPAUSE; - - amdgpu_vcn_pause_dpg_mode(adev, &new_state); + adev->vcn.inst[i].pause_dpg_mode(vcn_inst, &new_state); } - fences += amdgpu_fence_count_emitted(&adev->vcn.ring_jpeg); - fences += amdgpu_fence_count_emitted(&adev->vcn.ring_dec); + fence[i] += amdgpu_fence_count_emitted(&vcn_inst->ring_dec); + fences += fence[i]; + + if (!fences && !atomic_read(&vcn_inst->total_submission_cnt)) { + mutex_lock(&vcn_inst->vcn_pg_lock); + vcn_inst->set_pg_state(vcn_inst, AMD_PG_STATE_GATE); + mutex_unlock(&vcn_inst->vcn_pg_lock); + amdgpu_vcn_put_profile(adev); - 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); + schedule_delayed_work(&vcn_inst->idle_work, VCN_IDLE_TIMEOUT); } } void amdgpu_vcn_ring_begin_use(struct amdgpu_ring *ring) { struct amdgpu_device *adev = ring->adev; - bool set_clocks = !cancel_delayed_work_sync(&adev->vcn.idle_work); + struct amdgpu_vcn_inst *vcn_inst = &adev->vcn.inst[ring->me]; - 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); - } + atomic_inc(&vcn_inst->total_submission_cnt); + + cancel_delayed_work_sync(&vcn_inst->idle_work); - if (adev->pg_flags & AMD_PG_SUPPORT_VCN_DPG) { + mutex_lock(&vcn_inst->vcn_pg_lock); + vcn_inst->set_pg_state(vcn_inst, AMD_PG_STATE_UNGATE); + + /* Only set DPG pause for VCN3 or below, VCN4 and above will be handled by FW */ + if (adev->pg_flags & AMD_PG_SUPPORT_VCN_DPG && + !vcn_inst->using_unified_queue) { struct dpg_pause_state new_state; - unsigned int fences = 0; - unsigned int i; - for (i = 0; i < adev->vcn.num_enc_rings; ++i) { - fences += amdgpu_fence_count_emitted(&adev->vcn.ring_enc[i]); - } - if (fences) + if (ring->funcs->type == AMDGPU_RING_TYPE_VCN_ENC) { + atomic_inc(&vcn_inst->dpg_enc_submission_cnt); new_state.fw_based = VCN_DPG_STATE__PAUSE; - else - new_state.fw_based = VCN_DPG_STATE__UNPAUSE; + } else { + unsigned int fences = 0; + unsigned int i; - if (amdgpu_fence_count_emitted(&adev->vcn.ring_jpeg)) - new_state.jpeg = VCN_DPG_STATE__PAUSE; - else - new_state.jpeg = VCN_DPG_STATE__UNPAUSE; + for (i = 0; i < vcn_inst->num_enc_rings; ++i) + fences += amdgpu_fence_count_emitted(&vcn_inst->ring_enc[i]); - 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; + if (fences || atomic_read(&vcn_inst->dpg_enc_submission_cnt)) + new_state.fw_based = VCN_DPG_STATE__PAUSE; + else + new_state.fw_based = VCN_DPG_STATE__UNPAUSE; + } - amdgpu_vcn_pause_dpg_mode(adev, &new_state); + vcn_inst->pause_dpg_mode(vcn_inst, &new_state); } + mutex_unlock(&vcn_inst->vcn_pg_lock); + amdgpu_vcn_get_profile(adev); } void amdgpu_vcn_ring_end_use(struct amdgpu_ring *ring) { - schedule_delayed_work(&ring->adev->vcn.idle_work, VCN_IDLE_TIMEOUT); + struct amdgpu_device *adev = ring->adev; + + /* Only set DPG pause for VCN3 or below, VCN4 and above will be handled by FW */ + if (ring->adev->pg_flags & AMD_PG_SUPPORT_VCN_DPG && + ring->funcs->type == AMDGPU_RING_TYPE_VCN_ENC && + !adev->vcn.inst[ring->me].using_unified_queue) + atomic_dec(&ring->adev->vcn.inst[ring->me].dpg_enc_submission_cnt); + + atomic_dec(&ring->adev->vcn.inst[ring->me].total_submission_cnt); + + schedule_delayed_work(&ring->adev->vcn.inst[ring->me].idle_work, + VCN_IDLE_TIMEOUT); } int amdgpu_vcn_dec_ring_test_ring(struct amdgpu_ring *ring) { struct amdgpu_device *adev = ring->adev; uint32_t tmp = 0; - unsigned i; + unsigned int i; int r; - WREG32(SOC15_REG_OFFSET(UVD, 0, mmUVD_SCRATCH9), 0xCAFEDEAD); + /* VCN in SRIOV does not support direct register read/write */ + if (amdgpu_sriov_vf(adev)) + return 0; + + WREG32(adev->vcn.inst[ring->me].external.scratch9, 0xCAFEDEAD); r = amdgpu_ring_alloc(ring, 3); if (r) return r; - - amdgpu_ring_write(ring, - PACKET0(SOC15_REG_OFFSET(UVD, 0, mmUVD_SCRATCH9), 0)); + amdgpu_ring_write(ring, PACKET0(adev->vcn.inst[ring->me].internal.scratch9, 0)); amdgpu_ring_write(ring, 0xDEADBEEF); amdgpu_ring_commit(ring); for (i = 0; i < adev->usec_timeout; i++) { - tmp = RREG32(SOC15_REG_OFFSET(UVD, 0, mmUVD_SCRATCH9)); + tmp = RREG32(adev->vcn.inst[ring->me].external.scratch9); if (tmp == 0xDEADBEEF) break; - DRM_UDELAY(1); + udelay(1); + } + + if (i >= adev->usec_timeout) + r = -ETIMEDOUT; + + return r; +} + +int amdgpu_vcn_dec_sw_ring_test_ring(struct amdgpu_ring *ring) +{ + struct amdgpu_device *adev = ring->adev; + uint32_t rptr; + unsigned int i; + int r; + + if (amdgpu_sriov_vf(adev)) + return 0; + + r = amdgpu_ring_alloc(ring, 16); + if (r) + return r; + + rptr = amdgpu_ring_get_rptr(ring); + + amdgpu_ring_write(ring, VCN_DEC_SW_CMD_END); + amdgpu_ring_commit(ring); + + for (i = 0; i < adev->usec_timeout; i++) { + if (amdgpu_ring_get_rptr(ring) != rptr) + break; + udelay(1); } if (i >= adev->usec_timeout) @@ -456,30 +614,31 @@ int amdgpu_vcn_dec_ring_test_ring(struct amdgpu_ring *ring) } static int amdgpu_vcn_dec_send_msg(struct amdgpu_ring *ring, - struct amdgpu_bo *bo, + struct amdgpu_ib *ib_msg, struct dma_fence **fence) { + u64 addr = AMDGPU_GPU_PAGE_ALIGN(ib_msg->gpu_addr); struct amdgpu_device *adev = ring->adev; struct dma_fence *f = NULL; struct amdgpu_job *job; struct amdgpu_ib *ib; - uint64_t addr; int i, r; - r = amdgpu_job_alloc_with_ib(adev, 64, &job); + r = amdgpu_job_alloc_with_ib(ring->adev, NULL, NULL, + 64, AMDGPU_IB_POOL_DIRECT, + &job, AMDGPU_KERNEL_JOB_ID_VCN_RING_TEST); if (r) goto err; ib = &job->ibs[0]; - addr = amdgpu_bo_gpu_offset(bo); - ib->ptr[0] = PACKET0(SOC15_REG_OFFSET(UVD, 0, mmUVD_GPCOM_VCPU_DATA0), 0); + ib->ptr[0] = PACKET0(adev->vcn.inst[ring->me].internal.data0, 0); ib->ptr[1] = addr; - ib->ptr[2] = PACKET0(SOC15_REG_OFFSET(UVD, 0, mmUVD_GPCOM_VCPU_DATA1), 0); + ib->ptr[2] = PACKET0(adev->vcn.inst[ring->me].internal.data1, 0); ib->ptr[3] = addr >> 32; - ib->ptr[4] = PACKET0(SOC15_REG_OFFSET(UVD, 0, mmUVD_GPCOM_VCPU_CMD), 0); + ib->ptr[4] = PACKET0(adev->vcn.inst[ring->me].internal.cmd, 0); ib->ptr[5] = 0; for (i = 6; i < 16; i += 2) { - ib->ptr[i] = PACKET0(SOC15_REG_OFFSET(UVD, 0, mmUVD_NO_OP), 0); + ib->ptr[i] = PACKET0(adev->vcn.inst[ring->me].internal.nop, 0); ib->ptr[i+1] = 0; } ib->length_dw = 16; @@ -488,9 +647,7 @@ static int amdgpu_vcn_dec_send_msg(struct amdgpu_ring *ring, if (r) goto err_free; - amdgpu_bo_fence(bo, f, false); - amdgpu_bo_unreserve(bo); - amdgpu_bo_unref(&bo); + amdgpu_ib_free(ib_msg, f); if (fence) *fence = dma_fence_get(f); @@ -500,27 +657,26 @@ static int amdgpu_vcn_dec_send_msg(struct amdgpu_ring *ring, err_free: amdgpu_job_free(job); - err: - amdgpu_bo_unreserve(bo); - amdgpu_bo_unref(&bo); + amdgpu_ib_free(ib_msg, f); return r; } static int amdgpu_vcn_dec_get_create_msg(struct amdgpu_ring *ring, uint32_t handle, - struct dma_fence **fence) + struct amdgpu_ib *ib) { struct amdgpu_device *adev = ring->adev; - struct amdgpu_bo *bo = NULL; uint32_t *msg; int r, i; - r = amdgpu_bo_create_reserved(adev, 1024, PAGE_SIZE, - AMDGPU_GEM_DOMAIN_VRAM, - &bo, NULL, (void **)&msg); + memset(ib, 0, sizeof(*ib)); + r = amdgpu_ib_get(adev, NULL, AMDGPU_GPU_PAGE_SIZE * 2, + AMDGPU_IB_POOL_DIRECT, + ib); if (r) return r; + msg = (uint32_t *)AMDGPU_GPU_PAGE_ALIGN((unsigned long)ib->ptr); msg[0] = cpu_to_le32(0x00000028); msg[1] = cpu_to_le32(0x00000038); msg[2] = cpu_to_le32(0x00000001); @@ -538,23 +694,24 @@ static int amdgpu_vcn_dec_get_create_msg(struct amdgpu_ring *ring, uint32_t hand for (i = 14; i < 1024; ++i) msg[i] = cpu_to_le32(0x0); - return amdgpu_vcn_dec_send_msg(ring, bo, fence); + return 0; } static int amdgpu_vcn_dec_get_destroy_msg(struct amdgpu_ring *ring, uint32_t handle, - struct dma_fence **fence) + struct amdgpu_ib *ib) { struct amdgpu_device *adev = ring->adev; - struct amdgpu_bo *bo = NULL; uint32_t *msg; int r, i; - r = amdgpu_bo_create_reserved(adev, 1024, PAGE_SIZE, - AMDGPU_GEM_DOMAIN_VRAM, - &bo, NULL, (void **)&msg); + memset(ib, 0, sizeof(*ib)); + r = amdgpu_ib_get(adev, NULL, AMDGPU_GPU_PAGE_SIZE * 2, + AMDGPU_IB_POOL_DIRECT, + ib); if (r) return r; + msg = (uint32_t *)AMDGPU_GPU_PAGE_ALIGN((unsigned long)ib->ptr); msg[0] = cpu_to_le32(0x00000028); msg[1] = cpu_to_le32(0x00000018); msg[2] = cpu_to_le32(0x00000000); @@ -564,19 +721,158 @@ static int amdgpu_vcn_dec_get_destroy_msg(struct amdgpu_ring *ring, uint32_t han for (i = 6; i < 1024; ++i) msg[i] = cpu_to_le32(0x0); - return amdgpu_vcn_dec_send_msg(ring, bo, fence); + return 0; } int amdgpu_vcn_dec_ring_test_ib(struct amdgpu_ring *ring, long timeout) { - struct dma_fence *fence; + struct dma_fence *fence = NULL; + struct amdgpu_ib ib; long r; - r = amdgpu_vcn_dec_get_create_msg(ring, 1, NULL); + r = amdgpu_vcn_dec_get_create_msg(ring, 1, &ib); + if (r) + goto error; + + r = amdgpu_vcn_dec_send_msg(ring, &ib, NULL); + if (r) + goto error; + r = amdgpu_vcn_dec_get_destroy_msg(ring, 1, &ib); + if (r) + goto error; + + r = amdgpu_vcn_dec_send_msg(ring, &ib, &fence); + if (r) + goto error; + + r = dma_fence_wait_timeout(fence, false, timeout); + if (r == 0) + r = -ETIMEDOUT; + else if (r > 0) + r = 0; + + dma_fence_put(fence); +error: + return r; +} + +static uint32_t *amdgpu_vcn_unified_ring_ib_header(struct amdgpu_ib *ib, + uint32_t ib_pack_in_dw, bool enc) +{ + uint32_t *ib_checksum; + + ib->ptr[ib->length_dw++] = 0x00000010; /* single queue checksum */ + ib->ptr[ib->length_dw++] = 0x30000002; + ib_checksum = &ib->ptr[ib->length_dw++]; + ib->ptr[ib->length_dw++] = ib_pack_in_dw; + + ib->ptr[ib->length_dw++] = 0x00000010; /* engine info */ + ib->ptr[ib->length_dw++] = 0x30000001; + ib->ptr[ib->length_dw++] = enc ? 0x2 : 0x3; + ib->ptr[ib->length_dw++] = ib_pack_in_dw * sizeof(uint32_t); + + return ib_checksum; +} + +static void amdgpu_vcn_unified_ring_ib_checksum(uint32_t **ib_checksum, + uint32_t ib_pack_in_dw) +{ + uint32_t i; + uint32_t checksum = 0; + + for (i = 0; i < ib_pack_in_dw; i++) + checksum += *(*ib_checksum + 2 + i); + + **ib_checksum = checksum; +} + +static int amdgpu_vcn_dec_sw_send_msg(struct amdgpu_ring *ring, + struct amdgpu_ib *ib_msg, + struct dma_fence **fence) +{ + struct amdgpu_vcn_decode_buffer *decode_buffer = NULL; + unsigned int ib_size_dw = 64; + struct amdgpu_device *adev = ring->adev; + struct dma_fence *f = NULL; + struct amdgpu_job *job; + struct amdgpu_ib *ib; + uint64_t addr = AMDGPU_GPU_PAGE_ALIGN(ib_msg->gpu_addr); + uint32_t *ib_checksum; + uint32_t ib_pack_in_dw; + int i, r; + + if (adev->vcn.inst[ring->me].using_unified_queue) + ib_size_dw += 8; + + r = amdgpu_job_alloc_with_ib(ring->adev, NULL, NULL, + ib_size_dw * 4, AMDGPU_IB_POOL_DIRECT, + &job, AMDGPU_KERNEL_JOB_ID_VCN_RING_TEST); + if (r) + goto err; + + ib = &job->ibs[0]; + ib->length_dw = 0; + + /* single queue headers */ + if (adev->vcn.inst[ring->me].using_unified_queue) { + ib_pack_in_dw = sizeof(struct amdgpu_vcn_decode_buffer) / sizeof(uint32_t) + + 4 + 2; /* engine info + decoding ib in dw */ + ib_checksum = amdgpu_vcn_unified_ring_ib_header(ib, ib_pack_in_dw, false); + } + + ib->ptr[ib->length_dw++] = sizeof(struct amdgpu_vcn_decode_buffer) + 8; + ib->ptr[ib->length_dw++] = cpu_to_le32(AMDGPU_VCN_IB_FLAG_DECODE_BUFFER); + decode_buffer = (struct amdgpu_vcn_decode_buffer *)&(ib->ptr[ib->length_dw]); + ib->length_dw += sizeof(struct amdgpu_vcn_decode_buffer) / 4; + memset(decode_buffer, 0, sizeof(struct amdgpu_vcn_decode_buffer)); + + decode_buffer->valid_buf_flag |= cpu_to_le32(AMDGPU_VCN_CMD_FLAG_MSG_BUFFER); + decode_buffer->msg_buffer_address_hi = cpu_to_le32(addr >> 32); + decode_buffer->msg_buffer_address_lo = cpu_to_le32(addr); + + for (i = ib->length_dw; i < ib_size_dw; ++i) + ib->ptr[i] = 0x0; + + if (adev->vcn.inst[ring->me].using_unified_queue) + amdgpu_vcn_unified_ring_ib_checksum(&ib_checksum, ib_pack_in_dw); + + r = amdgpu_job_submit_direct(job, ring, &f); + if (r) + goto err_free; + + amdgpu_ib_free(ib_msg, f); + + if (fence) + *fence = dma_fence_get(f); + dma_fence_put(f); + + return 0; + +err_free: + amdgpu_job_free(job); +err: + amdgpu_ib_free(ib_msg, f); + return r; +} + +int amdgpu_vcn_dec_sw_ring_test_ib(struct amdgpu_ring *ring, long timeout) +{ + struct dma_fence *fence = NULL; + struct amdgpu_ib ib; + long r; + + r = amdgpu_vcn_dec_get_create_msg(ring, 1, &ib); + if (r) + goto error; + + r = amdgpu_vcn_dec_sw_send_msg(ring, &ib, NULL); + if (r) + goto error; + r = amdgpu_vcn_dec_get_destroy_msg(ring, 1, &ib); if (r) goto error; - r = amdgpu_vcn_dec_get_destroy_msg(ring, 1, &fence); + r = amdgpu_vcn_dec_sw_send_msg(ring, &ib, &fence); if (r) goto error; @@ -594,21 +890,26 @@ error: int amdgpu_vcn_enc_ring_test_ring(struct amdgpu_ring *ring) { struct amdgpu_device *adev = ring->adev; - uint32_t rptr = amdgpu_ring_get_rptr(ring); - unsigned i; + uint32_t rptr; + unsigned int i; int r; + if (amdgpu_sriov_vf(adev)) + return 0; + r = amdgpu_ring_alloc(ring, 16); if (r) return r; + rptr = amdgpu_ring_get_rptr(ring); + amdgpu_ring_write(ring, VCN_ENC_CMD_END); amdgpu_ring_commit(ring); for (i = 0; i < adev->usec_timeout; i++) { if (amdgpu_ring_get_rptr(ring) != rptr) break; - DRM_UDELAY(1); + udelay(1); } if (i >= adev->usec_timeout) @@ -618,29 +919,41 @@ int amdgpu_vcn_enc_ring_test_ring(struct amdgpu_ring *ring) } static int amdgpu_vcn_enc_get_create_msg(struct amdgpu_ring *ring, uint32_t handle, - struct dma_fence **fence) + struct amdgpu_ib *ib_msg, + struct dma_fence **fence) { - const unsigned ib_size_dw = 16; + unsigned int ib_size_dw = 16; + struct amdgpu_device *adev = ring->adev; struct amdgpu_job *job; struct amdgpu_ib *ib; struct dma_fence *f = NULL; - uint64_t dummy; + uint32_t *ib_checksum = NULL; + uint64_t addr; int i, r; - r = amdgpu_job_alloc_with_ib(ring->adev, ib_size_dw * 4, &job); + if (adev->vcn.inst[ring->me].using_unified_queue) + ib_size_dw += 8; + + r = amdgpu_job_alloc_with_ib(ring->adev, NULL, NULL, + ib_size_dw * 4, AMDGPU_IB_POOL_DIRECT, + &job, AMDGPU_KERNEL_JOB_ID_VCN_RING_TEST); if (r) return r; ib = &job->ibs[0]; - dummy = ib->gpu_addr + 1024; + addr = AMDGPU_GPU_PAGE_ALIGN(ib_msg->gpu_addr); ib->length_dw = 0; + + if (adev->vcn.inst[ring->me].using_unified_queue) + ib_checksum = amdgpu_vcn_unified_ring_ib_header(ib, 0x11, true); + ib->ptr[ib->length_dw++] = 0x00000018; ib->ptr[ib->length_dw++] = 0x00000001; /* session info */ ib->ptr[ib->length_dw++] = handle; - ib->ptr[ib->length_dw++] = upper_32_bits(dummy); - ib->ptr[ib->length_dw++] = dummy; - ib->ptr[ib->length_dw++] = 0x0000000b; + ib->ptr[ib->length_dw++] = upper_32_bits(addr); + ib->ptr[ib->length_dw++] = addr; + ib->ptr[ib->length_dw++] = 0x00000000; ib->ptr[ib->length_dw++] = 0x00000014; ib->ptr[ib->length_dw++] = 0x00000002; /* task info */ @@ -654,6 +967,9 @@ static int amdgpu_vcn_enc_get_create_msg(struct amdgpu_ring *ring, uint32_t hand for (i = ib->length_dw; i < ib_size_dw; ++i) ib->ptr[i] = 0x0; + if (adev->vcn.inst[ring->me].using_unified_queue) + amdgpu_vcn_unified_ring_ib_checksum(&ib_checksum, 0x11); + r = amdgpu_job_submit_direct(job, ring, &f); if (r) goto err; @@ -670,29 +986,41 @@ err: } static int amdgpu_vcn_enc_get_destroy_msg(struct amdgpu_ring *ring, uint32_t handle, - struct dma_fence **fence) + struct amdgpu_ib *ib_msg, + struct dma_fence **fence) { - const unsigned ib_size_dw = 16; + unsigned int ib_size_dw = 16; + struct amdgpu_device *adev = ring->adev; struct amdgpu_job *job; struct amdgpu_ib *ib; struct dma_fence *f = NULL; - uint64_t dummy; + uint32_t *ib_checksum = NULL; + uint64_t addr; int i, r; - r = amdgpu_job_alloc_with_ib(ring->adev, ib_size_dw * 4, &job); + if (adev->vcn.inst[ring->me].using_unified_queue) + ib_size_dw += 8; + + r = amdgpu_job_alloc_with_ib(ring->adev, NULL, NULL, + ib_size_dw * 4, AMDGPU_IB_POOL_DIRECT, + &job, AMDGPU_KERNEL_JOB_ID_VCN_RING_TEST); if (r) return r; ib = &job->ibs[0]; - dummy = ib->gpu_addr + 1024; + addr = AMDGPU_GPU_PAGE_ALIGN(ib_msg->gpu_addr); ib->length_dw = 0; + + if (adev->vcn.inst[ring->me].using_unified_queue) + ib_checksum = amdgpu_vcn_unified_ring_ib_header(ib, 0x11, true); + ib->ptr[ib->length_dw++] = 0x00000018; ib->ptr[ib->length_dw++] = 0x00000001; ib->ptr[ib->length_dw++] = handle; - ib->ptr[ib->length_dw++] = upper_32_bits(dummy); - ib->ptr[ib->length_dw++] = dummy; - ib->ptr[ib->length_dw++] = 0x0000000b; + ib->ptr[ib->length_dw++] = upper_32_bits(addr); + ib->ptr[ib->length_dw++] = addr; + ib->ptr[ib->length_dw++] = 0x00000000; ib->ptr[ib->length_dw++] = 0x00000014; ib->ptr[ib->length_dw++] = 0x00000002; @@ -706,6 +1034,9 @@ static int amdgpu_vcn_enc_get_destroy_msg(struct amdgpu_ring *ring, uint32_t han for (i = ib->length_dw; i < ib_size_dw; ++i) ib->ptr[i] = 0x0; + if (adev->vcn.inst[ring->me].using_unified_queue) + amdgpu_vcn_unified_ring_ib_checksum(&ib_checksum, 0x11); + r = amdgpu_job_submit_direct(job, ring, &f); if (r) goto err; @@ -723,14 +1054,23 @@ 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_ib ib; long r; - r = amdgpu_vcn_enc_get_create_msg(ring, 1, NULL); + memset(&ib, 0, sizeof(ib)); + r = amdgpu_ib_get(adev, NULL, (128 << 10) + AMDGPU_GPU_PAGE_SIZE, + AMDGPU_IB_POOL_DIRECT, + &ib); + if (r) + return r; + + r = amdgpu_vcn_enc_get_create_msg(ring, 1, &ib, NULL); if (r) goto error; - r = amdgpu_vcn_enc_get_destroy_msg(ring, 1, &fence); + r = amdgpu_vcn_enc_get_destroy_msg(ring, 1, &ib, &fence); if (r) goto error; @@ -741,113 +1081,557 @@ int amdgpu_vcn_enc_ring_test_ib(struct amdgpu_ring *ring, long timeout) r = 0; error: + amdgpu_ib_free(&ib, fence); dma_fence_put(fence); + return r; } -int amdgpu_vcn_jpeg_ring_test_ring(struct amdgpu_ring *ring) +int amdgpu_vcn_unified_ring_test_ib(struct amdgpu_ring *ring, long timeout) { struct amdgpu_device *adev = ring->adev; - uint32_t tmp = 0; - unsigned i; - int r; + long r; - WREG32(SOC15_REG_OFFSET(UVD, 0, mmUVD_SCRATCH9), 0xCAFEDEAD); - r = amdgpu_ring_alloc(ring, 3); + if ((amdgpu_ip_version(adev, UVD_HWIP, 0) != IP_VERSION(4, 0, 3)) && + (amdgpu_ip_version(adev, UVD_HWIP, 0) != IP_VERSION(5, 0, 1))) { + r = amdgpu_vcn_enc_ring_test_ib(ring, timeout); + if (r) + goto error; + } - if (r) - return r; + r = amdgpu_vcn_dec_sw_ring_test_ib(ring, timeout); - amdgpu_ring_write(ring, - PACKETJ(SOC15_REG_OFFSET(UVD, 0, mmUVD_SCRATCH9), 0, 0, 0)); - amdgpu_ring_write(ring, 0xDEADBEEF); - amdgpu_ring_commit(ring); +error: + return r; +} - for (i = 0; i < adev->usec_timeout; i++) { - tmp = RREG32(SOC15_REG_OFFSET(UVD, 0, mmUVD_SCRATCH9)); - if (tmp == 0xDEADBEEF) - break; - DRM_UDELAY(1); +enum amdgpu_ring_priority_level amdgpu_vcn_get_enc_ring_prio(int ring) +{ + switch (ring) { + case 0: + return AMDGPU_RING_PRIO_0; + case 1: + return AMDGPU_RING_PRIO_1; + case 2: + return AMDGPU_RING_PRIO_2; + default: + return AMDGPU_RING_PRIO_0; } +} - if (i >= adev->usec_timeout) - r = -ETIMEDOUT; +void amdgpu_vcn_setup_ucode(struct amdgpu_device *adev, int i) +{ + unsigned int idx; - return r; + if (adev->firmware.load_type == AMDGPU_FW_LOAD_PSP) { + const struct common_firmware_header *hdr; + + if (adev->vcn.harvest_config & (1 << i)) + return; + + if ((amdgpu_ip_version(adev, UVD_HWIP, 0) == IP_VERSION(4, 0, 3) || + amdgpu_ip_version(adev, UVD_HWIP, 0) == IP_VERSION(5, 0, 1)) + && (i > 0)) + return; + + hdr = (const struct common_firmware_header *)adev->vcn.inst[i].fw->data; + /* currently only support 2 FW instances */ + if (i >= 2) { + dev_info(adev->dev, "More then 2 VCN FW instances!\n"); + return; + } + idx = AMDGPU_UCODE_ID_VCN + i; + adev->firmware.ucode[idx].ucode_id = idx; + adev->firmware.ucode[idx].fw = adev->vcn.inst[i].fw; + adev->firmware.fw_size += + ALIGN(le32_to_cpu(hdr->ucode_size_bytes), PAGE_SIZE); + } } -static int amdgpu_vcn_jpeg_set_reg(struct amdgpu_ring *ring, uint32_t handle, - struct dma_fence **fence) +/* + * debugfs for mapping vcn firmware log buffer. + */ +#if defined(CONFIG_DEBUG_FS) +static ssize_t amdgpu_debugfs_vcn_fwlog_read(struct file *f, char __user *buf, + size_t size, loff_t *pos) { - struct amdgpu_device *adev = ring->adev; - struct amdgpu_job *job; - struct amdgpu_ib *ib; - struct dma_fence *f = NULL; - const unsigned ib_size_dw = 16; - int i, r; + struct amdgpu_vcn_inst *vcn; + void *log_buf; + struct amdgpu_vcn_fwlog *plog; + unsigned int read_pos, write_pos, available, i, read_bytes = 0; + unsigned int read_num[2] = {0}; + + vcn = file_inode(f)->i_private; + if (!vcn) + return -ENODEV; + + if (!vcn->fw_shared.cpu_addr || !amdgpu_vcnfw_log) + return -EFAULT; + + log_buf = vcn->fw_shared.cpu_addr + vcn->fw_shared.mem_size; + + plog = (struct amdgpu_vcn_fwlog *)log_buf; + read_pos = plog->rptr; + write_pos = plog->wptr; + + if (read_pos > AMDGPU_VCNFW_LOG_SIZE || write_pos > AMDGPU_VCNFW_LOG_SIZE) + return -EFAULT; + + if (!size || (read_pos == write_pos)) + return 0; + + if (write_pos > read_pos) { + available = write_pos - read_pos; + read_num[0] = min_t(size_t, size, available); + } else { + read_num[0] = AMDGPU_VCNFW_LOG_SIZE - read_pos; + available = read_num[0] + write_pos - plog->header_size; + if (size > available) + read_num[1] = write_pos - plog->header_size; + else if (size > read_num[0]) + read_num[1] = size - read_num[0]; + else + read_num[0] = size; + } + + for (i = 0; i < 2; i++) { + if (read_num[i]) { + if (read_pos == AMDGPU_VCNFW_LOG_SIZE) + read_pos = plog->header_size; + if (read_num[i] == copy_to_user((buf + read_bytes), + (log_buf + read_pos), read_num[i])) + return -EFAULT; + + read_bytes += read_num[i]; + read_pos += read_num[i]; + } + } + + plog->rptr = read_pos; + *pos += read_bytes; + return read_bytes; +} + +static const struct file_operations amdgpu_debugfs_vcnfwlog_fops = { + .owner = THIS_MODULE, + .read = amdgpu_debugfs_vcn_fwlog_read, + .llseek = default_llseek +}; +#endif + +void amdgpu_debugfs_vcn_fwlog_init(struct amdgpu_device *adev, uint8_t i, + struct amdgpu_vcn_inst *vcn) +{ +#if defined(CONFIG_DEBUG_FS) + struct drm_minor *minor = adev_to_drm(adev)->primary; + struct dentry *root = minor->debugfs_root; + char name[32]; + + sprintf(name, "amdgpu_vcn_%d_fwlog", i); + debugfs_create_file_size(name, S_IFREG | 0444, root, vcn, + &amdgpu_debugfs_vcnfwlog_fops, + AMDGPU_VCNFW_LOG_SIZE); +#endif +} + +void amdgpu_vcn_fwlog_init(struct amdgpu_vcn_inst *vcn) +{ +#if defined(CONFIG_DEBUG_FS) + uint32_t *flag = vcn->fw_shared.cpu_addr; + void *fw_log_cpu_addr = vcn->fw_shared.cpu_addr + vcn->fw_shared.mem_size; + uint64_t fw_log_gpu_addr = vcn->fw_shared.gpu_addr + vcn->fw_shared.mem_size; + struct amdgpu_vcn_fwlog *log_buf = fw_log_cpu_addr; + struct amdgpu_fw_shared_fw_logging *fw_log = vcn->fw_shared.cpu_addr + + vcn->fw_shared.log_offset; + *flag |= cpu_to_le32(AMDGPU_VCN_FW_LOGGING_FLAG); + fw_log->is_enabled = 1; + fw_log->addr_lo = cpu_to_le32(fw_log_gpu_addr & 0xFFFFFFFF); + fw_log->addr_hi = cpu_to_le32(fw_log_gpu_addr >> 32); + fw_log->size = cpu_to_le32(AMDGPU_VCNFW_LOG_SIZE); + + log_buf->header_size = sizeof(struct amdgpu_vcn_fwlog); + log_buf->buffer_size = AMDGPU_VCNFW_LOG_SIZE; + log_buf->rptr = log_buf->header_size; + log_buf->wptr = log_buf->header_size; + log_buf->wrapped = 0; +#endif +} + +int amdgpu_vcn_process_poison_irq(struct amdgpu_device *adev, + struct amdgpu_irq_src *source, + struct amdgpu_iv_entry *entry) +{ + struct ras_common_if *ras_if = adev->vcn.ras_if; + struct ras_dispatch_if ih_data = { + .entry = entry, + }; + + if (!ras_if) + return 0; + + if (!amdgpu_sriov_vf(adev)) { + ih_data.head = *ras_if; + amdgpu_ras_interrupt_dispatch(adev, &ih_data); + } else { + if (adev->virt.ops && adev->virt.ops->ras_poison_handler) + adev->virt.ops->ras_poison_handler(adev, ras_if->block); + else + dev_warn(adev->dev, + "No ras_poison_handler interface in SRIOV for VCN!\n"); + } - r = amdgpu_job_alloc_with_ib(ring->adev, ib_size_dw * 4, &job); + return 0; +} + +int amdgpu_vcn_ras_late_init(struct amdgpu_device *adev, struct ras_common_if *ras_block) +{ + int r, i; + + r = amdgpu_ras_block_late_init(adev, ras_block); if (r) return r; - ib = &job->ibs[0]; + if (amdgpu_ras_is_supported(adev, ras_block->block)) { + for (i = 0; i < adev->vcn.num_vcn_inst; i++) { + if (adev->vcn.harvest_config & (1 << i) || + !adev->vcn.inst[i].ras_poison_irq.funcs) + continue; - ib->ptr[0] = PACKETJ(SOC15_REG_OFFSET(UVD, 0, mmUVD_SCRATCH9), 0, 0, PACKETJ_TYPE0); - ib->ptr[1] = 0xDEADBEEF; - for (i = 2; i < 16; i += 2) { - ib->ptr[i] = PACKETJ(0, 0, 0, PACKETJ_TYPE6); - ib->ptr[i+1] = 0; + r = amdgpu_irq_get(adev, &adev->vcn.inst[i].ras_poison_irq, 0); + if (r) + goto late_fini; + } } - ib->length_dw = 16; + return 0; - r = amdgpu_job_submit_direct(job, ring, &f); - if (r) - goto err; +late_fini: + amdgpu_ras_block_late_fini(adev, ras_block); + return r; +} - if (fence) - *fence = dma_fence_get(f); - dma_fence_put(f); +int amdgpu_vcn_ras_sw_init(struct amdgpu_device *adev) +{ + int err; + struct amdgpu_vcn_ras *ras; + + if (!adev->vcn.ras) + return 0; + + ras = adev->vcn.ras; + err = amdgpu_ras_register_ras_block(adev, &ras->ras_block); + if (err) { + dev_err(adev->dev, "Failed to register vcn ras block!\n"); + return err; + } + + strcpy(ras->ras_block.ras_comm.name, "vcn"); + ras->ras_block.ras_comm.block = AMDGPU_RAS_BLOCK__VCN; + ras->ras_block.ras_comm.type = AMDGPU_RAS_ERROR__POISON; + adev->vcn.ras_if = &ras->ras_block.ras_comm; + + if (!ras->ras_block.ras_late_init) + ras->ras_block.ras_late_init = amdgpu_vcn_ras_late_init; return 0; +} + +int amdgpu_vcn_psp_update_sram(struct amdgpu_device *adev, int inst_idx, + enum AMDGPU_UCODE_ID ucode_id) +{ + struct amdgpu_firmware_info ucode = { + .ucode_id = (ucode_id ? ucode_id : + (inst_idx ? AMDGPU_UCODE_ID_VCN1_RAM : + AMDGPU_UCODE_ID_VCN0_RAM)), + .mc_addr = adev->vcn.inst[inst_idx].dpg_sram_gpu_addr, + .ucode_size = ((uintptr_t)adev->vcn.inst[inst_idx].dpg_sram_curr_addr - + (uintptr_t)adev->vcn.inst[inst_idx].dpg_sram_cpu_addr), + }; + + return psp_execute_ip_fw_load(&adev->psp, &ucode); +} + +static ssize_t amdgpu_get_vcn_reset_mask(struct device *dev, + struct device_attribute *attr, + char *buf) +{ + struct drm_device *ddev = dev_get_drvdata(dev); + struct amdgpu_device *adev = drm_to_adev(ddev); + + if (!adev) + return -ENODEV; + + return amdgpu_show_reset_mask(buf, adev->vcn.supported_reset); +} + +static DEVICE_ATTR(vcn_reset_mask, 0444, + amdgpu_get_vcn_reset_mask, NULL); + +int amdgpu_vcn_sysfs_reset_mask_init(struct amdgpu_device *adev) +{ + int r = 0; + + if (adev->vcn.num_vcn_inst) { + r = device_create_file(adev->dev, &dev_attr_vcn_reset_mask); + if (r) + return r; + } -err: - amdgpu_job_free(job); return r; } -int amdgpu_vcn_jpeg_ring_test_ib(struct amdgpu_ring *ring, long timeout) +void amdgpu_vcn_sysfs_reset_mask_fini(struct amdgpu_device *adev) { - struct amdgpu_device *adev = ring->adev; - uint32_t tmp = 0; - unsigned i; - struct dma_fence *fence = NULL; - long r = 0; + if (adev->dev->kobj.sd) { + if (adev->vcn.num_vcn_inst) + device_remove_file(adev->dev, &dev_attr_vcn_reset_mask); + } +} - r = amdgpu_vcn_jpeg_set_reg(ring, 1, &fence); - if (r) - goto error; +/* + * debugfs to enable/disable vcn job submission to specific core or + * instance. It is created only if the queue type is unified. + */ +#if defined(CONFIG_DEBUG_FS) +static int amdgpu_debugfs_vcn_sched_mask_set(void *data, u64 val) +{ + struct amdgpu_device *adev = (struct amdgpu_device *)data; + u32 i; + u64 mask; + struct amdgpu_ring *ring; - r = dma_fence_wait_timeout(fence, false, timeout); - if (r == 0) { - r = -ETIMEDOUT; - goto error; - } else if (r < 0) { - goto error; - } else { - r = 0; + if (!adev) + return -ENODEV; + + mask = (1ULL << adev->vcn.num_vcn_inst) - 1; + if ((val & mask) == 0) + return -EINVAL; + for (i = 0; i < adev->vcn.num_vcn_inst; ++i) { + ring = &adev->vcn.inst[i].ring_enc[0]; + if (val & (1ULL << i)) + ring->sched.ready = true; + else + ring->sched.ready = false; } + /* publish sched.ready flag update effective immediately across smp */ + smp_rmb(); + return 0; +} - for (i = 0; i < adev->usec_timeout; i++) { - tmp = RREG32(SOC15_REG_OFFSET(UVD, 0, mmUVD_SCRATCH9)); - if (tmp == 0xDEADBEEF) - break; - DRM_UDELAY(1); +static int amdgpu_debugfs_vcn_sched_mask_get(void *data, u64 *val) +{ + struct amdgpu_device *adev = (struct amdgpu_device *)data; + u32 i; + u64 mask = 0; + struct amdgpu_ring *ring; + + if (!adev) + return -ENODEV; + for (i = 0; i < adev->vcn.num_vcn_inst; ++i) { + ring = &adev->vcn.inst[i].ring_enc[0]; + if (ring->sched.ready) + mask |= 1ULL << i; + } + *val = mask; + return 0; +} + +DEFINE_DEBUGFS_ATTRIBUTE(amdgpu_debugfs_vcn_sched_mask_fops, + amdgpu_debugfs_vcn_sched_mask_get, + amdgpu_debugfs_vcn_sched_mask_set, "%llx\n"); +#endif + +void amdgpu_debugfs_vcn_sched_mask_init(struct amdgpu_device *adev) +{ +#if defined(CONFIG_DEBUG_FS) + struct drm_minor *minor = adev_to_drm(adev)->primary; + struct dentry *root = minor->debugfs_root; + char name[32]; + + if (adev->vcn.num_vcn_inst <= 1 || !adev->vcn.inst[0].using_unified_queue) + return; + sprintf(name, "amdgpu_vcn_sched_mask"); + debugfs_create_file(name, 0600, root, adev, + &amdgpu_debugfs_vcn_sched_mask_fops); +#endif +} + +/** + * vcn_set_powergating_state - set VCN block powergating state + * + * @ip_block: amdgpu_ip_block pointer + * @state: power gating state + * + * Set VCN block powergating state + */ +int vcn_set_powergating_state(struct amdgpu_ip_block *ip_block, + enum amd_powergating_state state) +{ + struct amdgpu_device *adev = ip_block->adev; + int ret = 0, i; + + for (i = 0; i < adev->vcn.num_vcn_inst; ++i) { + struct amdgpu_vcn_inst *vinst = &adev->vcn.inst[i]; + + ret |= vinst->set_pg_state(vinst, state); } - if (i >= adev->usec_timeout) - r = -ETIMEDOUT; + return ret; +} + +/** + * amdgpu_vcn_reset_engine - Reset a specific VCN engine + * @adev: Pointer to the AMDGPU device + * @instance_id: VCN engine instance to reset + * + * Returns: 0 on success, or a negative error code on failure. + */ +static int amdgpu_vcn_reset_engine(struct amdgpu_device *adev, + uint32_t instance_id) +{ + struct amdgpu_vcn_inst *vinst = &adev->vcn.inst[instance_id]; + int r, i; + + mutex_lock(&vinst->engine_reset_mutex); + /* Stop the scheduler's work queue for the dec and enc rings if they are running. + * This ensures that no new tasks are submitted to the queues while + * the reset is in progress. + */ + drm_sched_wqueue_stop(&vinst->ring_dec.sched); + for (i = 0; i < vinst->num_enc_rings; i++) + drm_sched_wqueue_stop(&vinst->ring_enc[i].sched); + + /* Perform the VCN reset for the specified instance */ + r = vinst->reset(vinst); + if (r) + goto unlock; + r = amdgpu_ring_test_ring(&vinst->ring_dec); + if (r) + goto unlock; + for (i = 0; i < vinst->num_enc_rings; i++) { + r = amdgpu_ring_test_ring(&vinst->ring_enc[i]); + if (r) + goto unlock; + } + amdgpu_fence_driver_force_completion(&vinst->ring_dec); + for (i = 0; i < vinst->num_enc_rings; i++) + amdgpu_fence_driver_force_completion(&vinst->ring_enc[i]); + + /* Restart the scheduler's work queue for the dec and enc rings + * if they were stopped by this function. This allows new tasks + * to be submitted to the queues after the reset is complete. + */ + drm_sched_wqueue_start(&vinst->ring_dec.sched); + for (i = 0; i < vinst->num_enc_rings; i++) + drm_sched_wqueue_start(&vinst->ring_enc[i].sched); + +unlock: + mutex_unlock(&vinst->engine_reset_mutex); - dma_fence_put(fence); -error: return r; } + +/** + * amdgpu_vcn_ring_reset - Reset a VCN ring + * @ring: ring to reset + * @vmid: vmid of guilty job + * @timedout_fence: fence of timed out job + * + * This helper is for VCN blocks without unified queues because + * resetting the engine resets all queues in that case. With + * unified queues we have one queue per engine. + * Returns: 0 on success, or a negative error code on failure. + */ +int amdgpu_vcn_ring_reset(struct amdgpu_ring *ring, + unsigned int vmid, + struct amdgpu_fence *timedout_fence) +{ + struct amdgpu_device *adev = ring->adev; + + if (adev->vcn.inst[ring->me].using_unified_queue) + return -EINVAL; + + return amdgpu_vcn_reset_engine(adev, ring->me); +} + +int amdgpu_vcn_reg_dump_init(struct amdgpu_device *adev, + const struct amdgpu_hwip_reg_entry *reg, u32 count) +{ + adev->vcn.ip_dump = kcalloc(adev->vcn.num_vcn_inst * count, + sizeof(uint32_t), GFP_KERNEL); + if (!adev->vcn.ip_dump) + return -ENOMEM; + adev->vcn.reg_list = reg; + adev->vcn.reg_count = count; + + return 0; +} + +static void amdgpu_vcn_reg_dump_fini(struct amdgpu_device *adev) +{ + kfree(adev->vcn.ip_dump); + adev->vcn.ip_dump = NULL; + adev->vcn.reg_list = NULL; + adev->vcn.reg_count = 0; +} + +void amdgpu_vcn_dump_ip_state(struct amdgpu_ip_block *ip_block) +{ + struct amdgpu_device *adev = ip_block->adev; + int i, j; + bool is_powered; + u32 inst_off; + + if (!adev->vcn.ip_dump) + return; + + for (i = 0; i < adev->vcn.num_vcn_inst; i++) { + if (adev->vcn.harvest_config & (1 << i)) + continue; + + inst_off = i * adev->vcn.reg_count; + /* mmUVD_POWER_STATUS is always readable and is the first in reg_list */ + adev->vcn.ip_dump[inst_off] = + RREG32(SOC15_REG_ENTRY_OFFSET_INST(adev->vcn.reg_list[0], i)); + is_powered = (adev->vcn.ip_dump[inst_off] & + UVD_POWER_STATUS__UVD_POWER_STATUS_TILES_OFF) != + UVD_POWER_STATUS__UVD_POWER_STATUS_TILES_OFF; + + if (is_powered) + for (j = 1; j < adev->vcn.reg_count; j++) + adev->vcn.ip_dump[inst_off + j] = + RREG32(SOC15_REG_ENTRY_OFFSET_INST(adev->vcn.reg_list[j], i)); + } +} + +void amdgpu_vcn_print_ip_state(struct amdgpu_ip_block *ip_block, struct drm_printer *p) +{ + struct amdgpu_device *adev = ip_block->adev; + int i, j; + bool is_powered; + u32 inst_off; + + if (!adev->vcn.ip_dump) + return; + + drm_printf(p, "num_instances:%d\n", adev->vcn.num_vcn_inst); + for (i = 0; i < adev->vcn.num_vcn_inst; i++) { + if (adev->vcn.harvest_config & (1 << i)) { + drm_printf(p, "\nHarvested Instance:VCN%d Skipping dump\n", i); + continue; + } + + inst_off = i * adev->vcn.reg_count; + is_powered = (adev->vcn.ip_dump[inst_off] & + UVD_POWER_STATUS__UVD_POWER_STATUS_TILES_OFF) != + UVD_POWER_STATUS__UVD_POWER_STATUS_TILES_OFF; + + if (is_powered) { + drm_printf(p, "\nActive Instance:VCN%d\n", i); + for (j = 0; j < adev->vcn.reg_count; j++) + drm_printf(p, "%-50s \t 0x%08x\n", adev->vcn.reg_list[j].reg_name, + adev->vcn.ip_dump[inst_off + j]); + } else { + drm_printf(p, "\nInactive Instance:VCN%d\n", i); + } + } +} |