From 01bb4a6c21c225e55a84ac606a3b213909023ba0 Mon Sep 17 00:00:00 2001 From: Daniel Baumann Date: Tue, 9 Apr 2024 11:59:14 +0200 Subject: Adding upstream version 2.4.120. Signed-off-by: Daniel Baumann --- amdgpu/.editorconfig | 13 + amdgpu/Android.mk | 19 + amdgpu/Makefile.sources | 14 + amdgpu/amdgpu-symbols.txt | 79 ++ amdgpu/amdgpu.h | 1897 ++++++++++++++++++++++++++++++++++++++++++++ amdgpu/amdgpu_asic_id.c | 161 ++++ amdgpu/amdgpu_bo.c | 791 ++++++++++++++++++ amdgpu/amdgpu_cs.c | 984 +++++++++++++++++++++++ amdgpu/amdgpu_device.c | 316 ++++++++ amdgpu/amdgpu_gpu_info.c | 362 +++++++++ amdgpu/amdgpu_internal.h | 175 ++++ amdgpu/amdgpu_vamgr.c | 302 +++++++ amdgpu/amdgpu_vm.c | 50 ++ amdgpu/handle_table.c | 72 ++ amdgpu/handle_table.h | 41 + amdgpu/libdrm_amdgpu.pc.in | 11 + amdgpu/meson.build | 70 ++ 17 files changed, 5357 insertions(+) create mode 100644 amdgpu/.editorconfig create mode 100644 amdgpu/Android.mk create mode 100644 amdgpu/Makefile.sources create mode 100644 amdgpu/amdgpu-symbols.txt create mode 100644 amdgpu/amdgpu.h create mode 100644 amdgpu/amdgpu_asic_id.c create mode 100644 amdgpu/amdgpu_bo.c create mode 100644 amdgpu/amdgpu_cs.c create mode 100644 amdgpu/amdgpu_device.c create mode 100644 amdgpu/amdgpu_gpu_info.c create mode 100644 amdgpu/amdgpu_internal.h create mode 100644 amdgpu/amdgpu_vamgr.c create mode 100644 amdgpu/amdgpu_vm.c create mode 100644 amdgpu/handle_table.c create mode 100644 amdgpu/handle_table.h create mode 100644 amdgpu/libdrm_amdgpu.pc.in create mode 100644 amdgpu/meson.build (limited to 'amdgpu') diff --git a/amdgpu/.editorconfig b/amdgpu/.editorconfig new file mode 100644 index 0000000..426273f --- /dev/null +++ b/amdgpu/.editorconfig @@ -0,0 +1,13 @@ +# To use this config with your editor, follow the instructions at: +# http://editorconfig.org + +[*] +charset = utf-8 +indent_style = tab +indent_size = 8 +tab_width = 8 +insert_final_newline = true + +[meson.build] +indent_style = space +indent_size = 2 diff --git a/amdgpu/Android.mk b/amdgpu/Android.mk new file mode 100644 index 0000000..1f028d0 --- /dev/null +++ b/amdgpu/Android.mk @@ -0,0 +1,19 @@ +LOCAL_PATH := $(call my-dir) +include $(CLEAR_VARS) + +# Import variables LIBDRM_AMDGPU_FILES, LIBDRM_AMDGPU_H_FILES +include $(LOCAL_PATH)/Makefile.sources + +LOCAL_MODULE := libdrm_amdgpu + +LOCAL_SHARED_LIBRARIES := libdrm + +LOCAL_SRC_FILES := $(LIBDRM_AMDGPU_FILES) + +LOCAL_CFLAGS := \ + -DAMDGPU_ASIC_ID_TABLE=\"/vendor/etc/hwdata/amdgpu.ids\" + +LOCAL_REQUIRED_MODULES := amdgpu.ids + +include $(LIBDRM_COMMON_MK) +include $(BUILD_SHARED_LIBRARY) diff --git a/amdgpu/Makefile.sources b/amdgpu/Makefile.sources new file mode 100644 index 0000000..d6df324 --- /dev/null +++ b/amdgpu/Makefile.sources @@ -0,0 +1,14 @@ +LIBDRM_AMDGPU_FILES := \ + amdgpu_asic_id.c \ + amdgpu_bo.c \ + amdgpu_cs.c \ + amdgpu_device.c \ + amdgpu_gpu_info.c \ + amdgpu_internal.h \ + amdgpu_vamgr.c \ + amdgpu_vm.c \ + handle_table.c \ + handle_table.h + +LIBDRM_AMDGPU_H_FILES := \ + amdgpu.h diff --git a/amdgpu/amdgpu-symbols.txt b/amdgpu/amdgpu-symbols.txt new file mode 100644 index 0000000..530b343 --- /dev/null +++ b/amdgpu/amdgpu-symbols.txt @@ -0,0 +1,79 @@ +amdgpu_bo_alloc +amdgpu_bo_cpu_map +amdgpu_bo_cpu_unmap +amdgpu_bo_export +amdgpu_bo_free +amdgpu_bo_import +amdgpu_bo_inc_ref +amdgpu_bo_list_create_raw +amdgpu_bo_list_destroy_raw +amdgpu_bo_list_create +amdgpu_bo_list_destroy +amdgpu_bo_list_update +amdgpu_bo_query_info +amdgpu_bo_set_metadata +amdgpu_bo_va_op +amdgpu_bo_va_op_raw +amdgpu_bo_wait_for_idle +amdgpu_create_bo_from_user_mem +amdgpu_cs_chunk_fence_info_to_data +amdgpu_cs_chunk_fence_to_dep +amdgpu_cs_create_semaphore +amdgpu_cs_create_syncobj +amdgpu_cs_create_syncobj2 +amdgpu_cs_ctx_create +amdgpu_cs_ctx_create2 +amdgpu_cs_ctx_free +amdgpu_cs_ctx_override_priority +amdgpu_cs_ctx_stable_pstate +amdgpu_cs_destroy_semaphore +amdgpu_cs_destroy_syncobj +amdgpu_cs_export_syncobj +amdgpu_cs_fence_to_handle +amdgpu_cs_import_syncobj +amdgpu_cs_query_fence_status +amdgpu_cs_query_reset_state +amdgpu_cs_query_reset_state2 +amdgpu_query_sw_info +amdgpu_cs_signal_semaphore +amdgpu_cs_submit +amdgpu_cs_submit_raw +amdgpu_cs_submit_raw2 +amdgpu_cs_syncobj_export_sync_file +amdgpu_cs_syncobj_export_sync_file2 +amdgpu_cs_syncobj_import_sync_file +amdgpu_cs_syncobj_import_sync_file2 +amdgpu_cs_syncobj_query +amdgpu_cs_syncobj_query2 +amdgpu_cs_syncobj_reset +amdgpu_cs_syncobj_signal +amdgpu_cs_syncobj_timeline_signal +amdgpu_cs_syncobj_timeline_wait +amdgpu_cs_syncobj_transfer +amdgpu_cs_syncobj_wait +amdgpu_cs_wait_fences +amdgpu_cs_wait_semaphore +amdgpu_device_deinitialize +amdgpu_device_get_fd +amdgpu_device_initialize +amdgpu_find_bo_by_cpu_mapping +amdgpu_get_marketing_name +amdgpu_query_buffer_size_alignment +amdgpu_query_crtc_from_id +amdgpu_query_firmware_version +amdgpu_query_gds_info +amdgpu_query_gpu_info +amdgpu_query_gpuvm_fault_info +amdgpu_query_heap_info +amdgpu_query_hw_ip_count +amdgpu_query_hw_ip_info +amdgpu_query_info +amdgpu_query_sensor_info +amdgpu_query_video_caps_info +amdgpu_read_mm_registers +amdgpu_va_range_alloc +amdgpu_va_range_free +amdgpu_va_get_start_addr +amdgpu_va_range_query +amdgpu_vm_reserve_vmid +amdgpu_vm_unreserve_vmid diff --git a/amdgpu/amdgpu.h b/amdgpu/amdgpu.h new file mode 100644 index 0000000..9bdbf36 --- /dev/null +++ b/amdgpu/amdgpu.h @@ -0,0 +1,1897 @@ +/* + * Copyright 2014 Advanced Micro Devices, Inc. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * the rights to use, copy, modify, merge, publish, distribute, sublicense, + * and/or sell copies of the Software, and to permit persons to whom the + * Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice shall be included in + * all copies or substantial portions of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL + * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR + * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, + * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR + * OTHER DEALINGS IN THE SOFTWARE. + * + */ + +/** + * \file amdgpu.h + * + * Declare public libdrm_amdgpu API + * + * This file define API exposed by libdrm_amdgpu library. + * User wanted to use libdrm_amdgpu functionality must include + * this file. + * + */ +#ifndef _AMDGPU_H_ +#define _AMDGPU_H_ + +#include +#include + +#ifdef __cplusplus +extern "C" { +#endif + +struct drm_amdgpu_info_hw_ip; +struct drm_amdgpu_bo_list_entry; + +/*--------------------------------------------------------------------------*/ +/* --------------------------- Defines ------------------------------------ */ +/*--------------------------------------------------------------------------*/ + +/** + * Define max. number of Command Buffers (IB) which could be sent to the single + * hardware IP to accommodate CE/DE requirements + * + * \sa amdgpu_cs_ib_info +*/ +#define AMDGPU_CS_MAX_IBS_PER_SUBMIT 4 + +/** + * Special timeout value meaning that the timeout is infinite. + */ +#define AMDGPU_TIMEOUT_INFINITE 0xffffffffffffffffull + +/** + * Used in amdgpu_cs_query_fence_status(), meaning that the given timeout + * is absolute. + */ +#define AMDGPU_QUERY_FENCE_TIMEOUT_IS_ABSOLUTE (1 << 0) + +/*--------------------------------------------------------------------------*/ +/* ----------------------------- Enums ------------------------------------ */ +/*--------------------------------------------------------------------------*/ + +/** + * Enum describing possible handle types + * + * \sa amdgpu_bo_import, amdgpu_bo_export + * +*/ +enum amdgpu_bo_handle_type { + /** GEM flink name (needs DRM authentication, used by DRI2) */ + amdgpu_bo_handle_type_gem_flink_name = 0, + + /** KMS handle which is used by all driver ioctls */ + amdgpu_bo_handle_type_kms = 1, + + /** DMA-buf fd handle */ + amdgpu_bo_handle_type_dma_buf_fd = 2, + + /** Deprecated in favour of and same behaviour as + * amdgpu_bo_handle_type_kms, use that instead of this + */ + amdgpu_bo_handle_type_kms_noimport = 3, +}; + +/** Define known types of GPU VM VA ranges */ +enum amdgpu_gpu_va_range +{ + /** Allocate from "normal"/general range */ + amdgpu_gpu_va_range_general = 0 +}; + +enum amdgpu_sw_info { + amdgpu_sw_info_address32_hi = 0, +}; + +/*--------------------------------------------------------------------------*/ +/* -------------------------- Datatypes ----------------------------------- */ +/*--------------------------------------------------------------------------*/ + +/** + * Define opaque pointer to context associated with fd. + * This context will be returned as the result of + * "initialize" function and should be pass as the first + * parameter to any API call + */ +typedef struct amdgpu_device *amdgpu_device_handle; + +/** + * Define GPU Context type as pointer to opaque structure + * Example of GPU Context is the "rendering" context associated + * with OpenGL context (glCreateContext) + */ +typedef struct amdgpu_context *amdgpu_context_handle; + +/** + * Define handle for amdgpu resources: buffer, GDS, etc. + */ +typedef struct amdgpu_bo *amdgpu_bo_handle; + +/** + * Define handle for list of BOs + */ +typedef struct amdgpu_bo_list *amdgpu_bo_list_handle; + +/** + * Define handle to be used to work with VA allocated ranges + */ +typedef struct amdgpu_va *amdgpu_va_handle; + +/** + * Define handle for semaphore + */ +typedef struct amdgpu_semaphore *amdgpu_semaphore_handle; + +/*--------------------------------------------------------------------------*/ +/* -------------------------- Structures ---------------------------------- */ +/*--------------------------------------------------------------------------*/ + +/** + * Structure describing memory allocation request + * + * \sa amdgpu_bo_alloc() + * +*/ +struct amdgpu_bo_alloc_request { + /** Allocation request. It must be aligned correctly. */ + uint64_t alloc_size; + + /** + * It may be required to have some specific alignment requirements + * for physical back-up storage (e.g. for displayable surface). + * If 0 there is no special alignment requirement + */ + uint64_t phys_alignment; + + /** + * UMD should specify where to allocate memory and how it + * will be accessed by the CPU. + */ + uint32_t preferred_heap; + + /** Additional flags passed on allocation */ + uint64_t flags; +}; + +/** + * Special UMD specific information associated with buffer. + * + * It may be need to pass some buffer charactersitic as part + * of buffer sharing. Such information are defined UMD and + * opaque for libdrm_amdgpu as well for kernel driver. + * + * \sa amdgpu_bo_set_metadata(), amdgpu_bo_query_info, + * amdgpu_bo_import(), amdgpu_bo_export + * +*/ +struct amdgpu_bo_metadata { + /** Special flag associated with surface */ + uint64_t flags; + + /** + * ASIC-specific tiling information (also used by DCE). + * The encoding is defined by the AMDGPU_TILING_* definitions. + */ + uint64_t tiling_info; + + /** Size of metadata associated with the buffer, in bytes. */ + uint32_t size_metadata; + + /** UMD specific metadata. Opaque for kernel */ + uint32_t umd_metadata[64]; +}; + +/** + * Structure describing allocated buffer. Client may need + * to query such information as part of 'sharing' buffers mechanism + * + * \sa amdgpu_bo_set_metadata(), amdgpu_bo_query_info(), + * amdgpu_bo_import(), amdgpu_bo_export() +*/ +struct amdgpu_bo_info { + /** Allocated memory size */ + uint64_t alloc_size; + + /** + * It may be required to have some specific alignment requirements + * for physical back-up storage. + */ + uint64_t phys_alignment; + + /** Heap where to allocate memory. */ + uint32_t preferred_heap; + + /** Additional allocation flags. */ + uint64_t alloc_flags; + + /** Metadata associated with buffer if any. */ + struct amdgpu_bo_metadata metadata; +}; + +/** + * Structure with information about "imported" buffer + * + * \sa amdgpu_bo_import() + * + */ +struct amdgpu_bo_import_result { + /** Handle of memory/buffer to use */ + amdgpu_bo_handle buf_handle; + + /** Buffer size */ + uint64_t alloc_size; +}; + +/** + * + * Structure to describe GDS partitioning information. + * \note OA and GWS resources are asscoiated with GDS partition + * + * \sa amdgpu_gpu_resource_query_gds_info + * +*/ +struct amdgpu_gds_resource_info { + uint32_t gds_gfx_partition_size; + uint32_t compute_partition_size; + uint32_t gds_total_size; + uint32_t gws_per_gfx_partition; + uint32_t gws_per_compute_partition; + uint32_t oa_per_gfx_partition; + uint32_t oa_per_compute_partition; +}; + +/** + * Structure describing CS fence + * + * \sa amdgpu_cs_query_fence_status(), amdgpu_cs_request, amdgpu_cs_submit() + * +*/ +struct amdgpu_cs_fence { + + /** In which context IB was sent to execution */ + amdgpu_context_handle context; + + /** To which HW IP type the fence belongs */ + uint32_t ip_type; + + /** IP instance index if there are several IPs of the same type. */ + uint32_t ip_instance; + + /** Ring index of the HW IP */ + uint32_t ring; + + /** Specify fence for which we need to check submission status.*/ + uint64_t fence; +}; + +/** + * Structure describing IB + * + * \sa amdgpu_cs_request, amdgpu_cs_submit() + * +*/ +struct amdgpu_cs_ib_info { + /** Special flags */ + uint64_t flags; + + /** Virtual MC address of the command buffer */ + uint64_t ib_mc_address; + + /** + * Size of Command Buffer to be submitted. + * - The size is in units of dwords (4 bytes). + * - Could be 0 + */ + uint32_t size; +}; + +/** + * Structure describing fence information + * + * \sa amdgpu_cs_request, amdgpu_cs_query_fence, + * amdgpu_cs_submit(), amdgpu_cs_query_fence_status() +*/ +struct amdgpu_cs_fence_info { + /** buffer object for the fence */ + amdgpu_bo_handle handle; + + /** fence offset in the unit of sizeof(uint64_t) */ + uint64_t offset; +}; + +/** + * Structure describing submission request + * + * \note We could have several IBs as packet. e.g. CE, CE, DE case for gfx + * + * \sa amdgpu_cs_submit() +*/ +struct amdgpu_cs_request { + /** Specify flags with additional information */ + uint64_t flags; + + /** Specify HW IP block type to which to send the IB. */ + unsigned ip_type; + + /** IP instance index if there are several IPs of the same type. */ + unsigned ip_instance; + + /** + * Specify ring index of the IP. We could have several rings + * in the same IP. E.g. 0 for SDMA0 and 1 for SDMA1. + */ + uint32_t ring; + + /** + * List handle with resources used by this request. + */ + amdgpu_bo_list_handle resources; + + /** + * Number of dependencies this Command submission needs to + * wait for before starting execution. + */ + uint32_t number_of_dependencies; + + /** + * Array of dependencies which need to be met before + * execution can start. + */ + struct amdgpu_cs_fence *dependencies; + + /** Number of IBs to submit in the field ibs. */ + uint32_t number_of_ibs; + + /** + * IBs to submit. Those IBs will be submit together as single entity + */ + struct amdgpu_cs_ib_info *ibs; + + /** + * The returned sequence number for the command submission + */ + uint64_t seq_no; + + /** + * The fence information + */ + struct amdgpu_cs_fence_info fence_info; +}; + +/** + * Structure which provide information about GPU VM MC Address space + * alignments requirements + * + * \sa amdgpu_query_buffer_size_alignment + */ +struct amdgpu_buffer_size_alignments { + /** Size alignment requirement for allocation in + * local memory */ + uint64_t size_local; + + /** + * Size alignment requirement for allocation in remote memory + */ + uint64_t size_remote; +}; + +/** + * Structure which provide information about heap + * + * \sa amdgpu_query_heap_info() + * + */ +struct amdgpu_heap_info { + /** Theoretical max. available memory in the given heap */ + uint64_t heap_size; + + /** + * Number of bytes allocated in the heap. This includes all processes + * and private allocations in the kernel. It changes when new buffers + * are allocated, freed, and moved. It cannot be larger than + * heap_size. + */ + uint64_t heap_usage; + + /** + * Theoretical possible max. size of buffer which + * could be allocated in the given heap + */ + uint64_t max_allocation; +}; + +/** + * Describe GPU h/w info needed for UMD correct initialization + * + * \sa amdgpu_query_gpu_info() +*/ +struct amdgpu_gpu_info { + /** Asic id */ + uint32_t asic_id; + /** Chip revision */ + uint32_t chip_rev; + /** Chip external revision */ + uint32_t chip_external_rev; + /** Family ID */ + uint32_t family_id; + /** Special flags */ + uint64_t ids_flags; + /** max engine clock*/ + uint64_t max_engine_clk; + /** max memory clock */ + uint64_t max_memory_clk; + /** number of shader engines */ + uint32_t num_shader_engines; + /** number of shader arrays per engine */ + uint32_t num_shader_arrays_per_engine; + /** Number of available good shader pipes */ + uint32_t avail_quad_shader_pipes; + /** Max. number of shader pipes.(including good and bad pipes */ + uint32_t max_quad_shader_pipes; + /** Number of parameter cache entries per shader quad pipe */ + uint32_t cache_entries_per_quad_pipe; + /** Number of available graphics context */ + uint32_t num_hw_gfx_contexts; + /** Number of render backend pipes */ + uint32_t rb_pipes; + /** Enabled render backend pipe mask */ + uint32_t enabled_rb_pipes_mask; + /** Frequency of GPU Counter */ + uint32_t gpu_counter_freq; + /** CC_RB_BACKEND_DISABLE.BACKEND_DISABLE per SE */ + uint32_t backend_disable[4]; + /** Value of MC_ARB_RAMCFG register*/ + uint32_t mc_arb_ramcfg; + /** Value of GB_ADDR_CONFIG */ + uint32_t gb_addr_cfg; + /** Values of the GB_TILE_MODE0..31 registers */ + uint32_t gb_tile_mode[32]; + /** Values of GB_MACROTILE_MODE0..15 registers */ + uint32_t gb_macro_tile_mode[16]; + /** Value of PA_SC_RASTER_CONFIG register per SE */ + uint32_t pa_sc_raster_cfg[4]; + /** Value of PA_SC_RASTER_CONFIG_1 register per SE */ + uint32_t pa_sc_raster_cfg1[4]; + /* CU info */ + uint32_t cu_active_number; + uint32_t cu_ao_mask; + uint32_t cu_bitmap[4][4]; + /* video memory type info*/ + uint32_t vram_type; + /* video memory bit width*/ + uint32_t vram_bit_width; + /** constant engine ram size*/ + uint32_t ce_ram_size; + /* vce harvesting instance */ + uint32_t vce_harvest_config; + /* PCI revision ID */ + uint32_t pci_rev_id; +}; + + +/*--------------------------------------------------------------------------*/ +/*------------------------- Functions --------------------------------------*/ +/*--------------------------------------------------------------------------*/ + +/* + * Initialization / Cleanup + * +*/ + +/** + * + * \param fd - \c [in] File descriptor for AMD GPU device + * received previously as the result of + * e.g. drmOpen() call. + * For legacy fd type, the DRI2/DRI3 + * authentication should be done before + * calling this function. + * \param major_version - \c [out] Major version of library. It is assumed + * that adding new functionality will cause + * increase in major version + * \param minor_version - \c [out] Minor version of library + * \param device_handle - \c [out] Pointer to opaque context which should + * be passed as the first parameter on each + * API call + * + * + * \return 0 on success\n + * <0 - Negative POSIX Error code + * + * + * \sa amdgpu_device_deinitialize() +*/ +int amdgpu_device_initialize(int fd, + uint32_t *major_version, + uint32_t *minor_version, + amdgpu_device_handle *device_handle); + +/** + * + * When access to such library does not needed any more the special + * function must be call giving opportunity to clean up any + * resources if needed. + * + * \param device_handle - \c [in] Context associated with file + * descriptor for AMD GPU device + * received previously as the + * result e.g. of drmOpen() call. + * + * \return 0 on success\n + * <0 - Negative POSIX Error code + * + * \sa amdgpu_device_initialize() + * +*/ +int amdgpu_device_deinitialize(amdgpu_device_handle device_handle); + +/** + * + * /param device_handle - \c [in] Device handle. + * See #amdgpu_device_initialize() + * + * \return Returns the drm fd used for operations on this + * device. This is still owned by the library and hence + * should not be closed. Guaranteed to be valid until + * #amdgpu_device_deinitialize gets called. + * +*/ +int amdgpu_device_get_fd(amdgpu_device_handle device_handle); + +/* + * Memory Management + * +*/ + +/** + * Allocate memory to be used by UMD for GPU related operations + * + * \param dev - \c [in] Device handle. + * See #amdgpu_device_initialize() + * \param alloc_buffer - \c [in] Pointer to the structure describing an + * allocation request + * \param buf_handle - \c [out] Allocated buffer handle + * + * \return 0 on success\n + * <0 - Negative POSIX Error code + * + * \sa amdgpu_bo_free() +*/ +int amdgpu_bo_alloc(amdgpu_device_handle dev, + struct amdgpu_bo_alloc_request *alloc_buffer, + amdgpu_bo_handle *buf_handle); + +/** + * Associate opaque data with buffer to be queried by another UMD + * + * \param dev - \c [in] Device handle. See #amdgpu_device_initialize() + * \param buf_handle - \c [in] Buffer handle + * \param info - \c [in] Metadata to associated with buffer + * + * \return 0 on success\n + * <0 - Negative POSIX Error code +*/ +int amdgpu_bo_set_metadata(amdgpu_bo_handle buf_handle, + struct amdgpu_bo_metadata *info); + +/** + * Query buffer information including metadata previusly associated with + * buffer. + * + * \param dev - \c [in] Device handle. + * See #amdgpu_device_initialize() + * \param buf_handle - \c [in] Buffer handle + * \param info - \c [out] Structure describing buffer + * + * \return 0 on success\n + * <0 - Negative POSIX Error code + * + * \sa amdgpu_bo_set_metadata(), amdgpu_bo_alloc() +*/ +int amdgpu_bo_query_info(amdgpu_bo_handle buf_handle, + struct amdgpu_bo_info *info); + +/** + * Allow others to get access to buffer + * + * \param dev - \c [in] Device handle. + * See #amdgpu_device_initialize() + * \param buf_handle - \c [in] Buffer handle + * \param type - \c [in] Type of handle requested + * \param shared_handle - \c [out] Special "shared" handle + * + * \return 0 on success\n + * <0 - Negative POSIX Error code + * + * \sa amdgpu_bo_import() + * +*/ +int amdgpu_bo_export(amdgpu_bo_handle buf_handle, + enum amdgpu_bo_handle_type type, + uint32_t *shared_handle); + +/** + * Request access to "shared" buffer + * + * \param dev - \c [in] Device handle. + * See #amdgpu_device_initialize() + * \param type - \c [in] Type of handle requested + * \param shared_handle - \c [in] Shared handle received as result "import" + * operation + * \param output - \c [out] Pointer to structure with information + * about imported buffer + * + * \return 0 on success\n + * <0 - Negative POSIX Error code + * + * \note Buffer must be "imported" only using new "fd" (different from + * one used by "exporter"). + * + * \sa amdgpu_bo_export() + * +*/ +int amdgpu_bo_import(amdgpu_device_handle dev, + enum amdgpu_bo_handle_type type, + uint32_t shared_handle, + struct amdgpu_bo_import_result *output); + +/** + * Request GPU access to user allocated memory e.g. via "malloc" + * + * \param dev - [in] Device handle. See #amdgpu_device_initialize() + * \param cpu - [in] CPU address of user allocated memory which we + * want to map to GPU address space (make GPU accessible) + * (This address must be correctly aligned). + * \param size - [in] Size of allocation (must be correctly aligned) + * \param buf_handle - [out] Buffer handle for the userptr memory + * resource on submission and be used in other operations. + * + * + * \return 0 on success\n + * <0 - Negative POSIX Error code + * + * \note + * This call doesn't guarantee that such memory will be persistently + * "locked" / make non-pageable. The purpose of this call is to provide + * opportunity for GPU get access to this resource during submission. + * + * The maximum amount of memory which could be mapped in this call depends + * if overcommit is disabled or not. If overcommit is disabled than the max. + * amount of memory to be pinned will be limited by left "free" size in total + * amount of memory which could be locked simultaneously ("GART" size). + * + * Supported (theoretical) max. size of mapping is restricted only by + * "GART" size. + * + * It is responsibility of caller to correctly specify access rights + * on VA assignment. +*/ +int amdgpu_create_bo_from_user_mem(amdgpu_device_handle dev, + void *cpu, uint64_t size, + amdgpu_bo_handle *buf_handle); + +/** + * Validate if the user memory comes from BO + * + * \param dev - [in] Device handle. See #amdgpu_device_initialize() + * \param cpu - [in] CPU address of user allocated memory which we + * want to map to GPU address space (make GPU accessible) + * (This address must be correctly aligned). + * \param size - [in] Size of allocation (must be correctly aligned) + * \param buf_handle - [out] Buffer handle for the userptr memory + * if the user memory is not from BO, the buf_handle will be NULL. + * \param offset_in_bo - [out] offset in this BO for this user memory + * + * + * \return 0 on success\n + * <0 - Negative POSIX Error code + * +*/ +int amdgpu_find_bo_by_cpu_mapping(amdgpu_device_handle dev, + void *cpu, + uint64_t size, + amdgpu_bo_handle *buf_handle, + uint64_t *offset_in_bo); + +/** + * Free previously allocated memory + * + * \param dev - \c [in] Device handle. See #amdgpu_device_initialize() + * \param buf_handle - \c [in] Buffer handle to free + * + * \return 0 on success\n + * <0 - Negative POSIX Error code + * + * \note In the case of memory shared between different applications all + * resources will be “physically” freed only all such applications + * will be terminated + * \note If is UMD responsibility to ‘free’ buffer only when there is no + * more GPU access + * + * \sa amdgpu_bo_set_metadata(), amdgpu_bo_alloc() + * +*/ +int amdgpu_bo_free(amdgpu_bo_handle buf_handle); + +/** + * Increase the reference count of a buffer object + * + * \param bo - \c [in] Buffer object handle to increase the reference count + * + * \sa amdgpu_bo_alloc(), amdgpu_bo_free() + * +*/ +void amdgpu_bo_inc_ref(amdgpu_bo_handle bo); + +/** + * Request CPU access to GPU accessible memory + * + * \param buf_handle - \c [in] Buffer handle + * \param cpu - \c [out] CPU address to be used for access + * + * \return 0 on success\n + * <0 - Negative POSIX Error code + * + * \sa amdgpu_bo_cpu_unmap() + * +*/ +int amdgpu_bo_cpu_map(amdgpu_bo_handle buf_handle, void **cpu); + +/** + * Release CPU access to GPU memory + * + * \param buf_handle - \c [in] Buffer handle + * + * \return 0 on success\n + * <0 - Negative POSIX Error code + * + * \sa amdgpu_bo_cpu_map() + * +*/ +int amdgpu_bo_cpu_unmap(amdgpu_bo_handle buf_handle); + +/** + * Wait until a buffer is not used by the device. + * + * \param dev - \c [in] Device handle. See #amdgpu_device_initialize() + * \param buf_handle - \c [in] Buffer handle. + * \param timeout_ns - Timeout in nanoseconds. + * \param buffer_busy - 0 if buffer is idle, all GPU access was completed + * and no GPU access is scheduled. + * 1 GPU access is in fly or scheduled + * + * \return 0 - on success + * <0 - Negative POSIX Error code + */ +int amdgpu_bo_wait_for_idle(amdgpu_bo_handle buf_handle, + uint64_t timeout_ns, + bool *buffer_busy); + +/** + * Creates a BO list handle for command submission. + * + * \param dev - \c [in] Device handle. + * See #amdgpu_device_initialize() + * \param number_of_buffers - \c [in] Number of BOs in the list + * \param buffers - \c [in] List of BO handles + * \param result - \c [out] Created BO list handle + * + * \return 0 on success\n + * <0 - Negative POSIX Error code + * + * \sa amdgpu_bo_list_destroy_raw(), amdgpu_cs_submit_raw2() +*/ +int amdgpu_bo_list_create_raw(amdgpu_device_handle dev, + uint32_t number_of_buffers, + struct drm_amdgpu_bo_list_entry *buffers, + uint32_t *result); + +/** + * Destroys a BO list handle. + * + * \param bo_list - \c [in] BO list handle. + * + * \return 0 on success\n + * <0 - Negative POSIX Error code + * + * \sa amdgpu_bo_list_create_raw(), amdgpu_cs_submit_raw2() +*/ +int amdgpu_bo_list_destroy_raw(amdgpu_device_handle dev, uint32_t bo_list); + +/** + * Creates a BO list handle for command submission. + * + * \param dev - \c [in] Device handle. + * See #amdgpu_device_initialize() + * \param number_of_resources - \c [in] Number of BOs in the list + * \param resources - \c [in] List of BO handles + * \param resource_prios - \c [in] Optional priority for each handle + * \param result - \c [out] Created BO list handle + * + * \return 0 on success\n + * <0 - Negative POSIX Error code + * + * \sa amdgpu_bo_list_destroy() +*/ +int amdgpu_bo_list_create(amdgpu_device_handle dev, + uint32_t number_of_resources, + amdgpu_bo_handle *resources, + uint8_t *resource_prios, + amdgpu_bo_list_handle *result); + +/** + * Destroys a BO list handle. + * + * \param handle - \c [in] BO list handle. + * + * \return 0 on success\n + * <0 - Negative POSIX Error code + * + * \sa amdgpu_bo_list_create() +*/ +int amdgpu_bo_list_destroy(amdgpu_bo_list_handle handle); + +/** + * Update resources for existing BO list + * + * \param handle - \c [in] BO list handle + * \param number_of_resources - \c [in] Number of BOs in the list + * \param resources - \c [in] List of BO handles + * \param resource_prios - \c [in] Optional priority for each handle + * + * \return 0 on success\n + * <0 - Negative POSIX Error code + * + * \sa amdgpu_bo_list_update() +*/ +int amdgpu_bo_list_update(amdgpu_bo_list_handle handle, + uint32_t number_of_resources, + amdgpu_bo_handle *resources, + uint8_t *resource_prios); + +/* + * GPU Execution context + * +*/ + +/** + * Create GPU execution Context + * + * For the purpose of GPU Scheduler and GPU Robustness extensions it is + * necessary to have information/identify rendering/compute contexts. + * It also may be needed to associate some specific requirements with such + * contexts. Kernel driver will guarantee that submission from the same + * context will always be executed in order (first come, first serve). + * + * + * \param dev - \c [in] Device handle. See #amdgpu_device_initialize() + * \param priority - \c [in] Context creation flags. See AMDGPU_CTX_PRIORITY_* + * \param context - \c [out] GPU Context handle + * + * \return 0 on success\n + * <0 - Negative POSIX Error code + * + * \sa amdgpu_cs_ctx_free() + * +*/ +int amdgpu_cs_ctx_create2(amdgpu_device_handle dev, + uint32_t priority, + amdgpu_context_handle *context); +/** + * Create GPU execution Context + * + * Refer to amdgpu_cs_ctx_create2 for full documentation. This call + * is missing the priority parameter. + * + * \sa amdgpu_cs_ctx_create2() + * +*/ +int amdgpu_cs_ctx_create(amdgpu_device_handle dev, + amdgpu_context_handle *context); + +/** + * + * Destroy GPU execution context when not needed any more + * + * \param context - \c [in] GPU Context handle + * + * \return 0 on success\n + * <0 - Negative POSIX Error code + * + * \sa amdgpu_cs_ctx_create() + * +*/ +int amdgpu_cs_ctx_free(amdgpu_context_handle context); + +/** + * Override the submission priority for the given context using a master fd. + * + * \param dev - \c [in] device handle + * \param context - \c [in] context handle for context id + * \param master_fd - \c [in] The master fd to authorize the override. + * \param priority - \c [in] The priority to assign to the context. + * + * \return 0 on success or a a negative Posix error code on failure. + */ +int amdgpu_cs_ctx_override_priority(amdgpu_device_handle dev, + amdgpu_context_handle context, + int master_fd, + unsigned priority); + +/** + * Set or query the stable power state for GPU profiling. + * + * \param dev - \c [in] device handle + * \param op - \c [in] AMDGPU_CTX_OP_{GET,SET}_STABLE_PSTATE + * \param flags - \c [in] AMDGPU_CTX_STABLE_PSTATE_* + * \param out_flags - \c [out] output current stable pstate + * + * \return 0 on success otherwise POSIX Error code. + */ +int amdgpu_cs_ctx_stable_pstate(amdgpu_context_handle context, + uint32_t op, + uint32_t flags, + uint32_t *out_flags); + +/** + * Query reset state for the specific GPU Context + * + * \param context - \c [in] GPU Context handle + * \param state - \c [out] One of AMDGPU_CTX_*_RESET + * \param hangs - \c [out] Number of hangs caused by the context. + * + * \return 0 on success\n + * <0 - Negative POSIX Error code + * + * \sa amdgpu_cs_ctx_create() + * +*/ +int amdgpu_cs_query_reset_state(amdgpu_context_handle context, + uint32_t *state, uint32_t *hangs); + +/** + * Query reset state for the specific GPU Context. + * + * \param context - \c [in] GPU Context handle + * \param flags - \c [out] A combination of AMDGPU_CTX_QUERY2_FLAGS_* + * + * \return 0 on success\n + * <0 - Negative POSIX Error code + * + * \sa amdgpu_cs_ctx_create() + * +*/ +int amdgpu_cs_query_reset_state2(amdgpu_context_handle context, + uint64_t *flags); + +/* + * Command Buffers Management + * +*/ + +/** + * Send request to submit command buffers to hardware. + * + * Kernel driver could use GPU Scheduler to make decision when physically + * sent this request to the hardware. Accordingly this request could be put + * in queue and sent for execution later. The only guarantee is that request + * from the same GPU context to the same ip:ip_instance:ring will be executed in + * order. + * + * The caller can specify the user fence buffer/location with the fence_info in the + * cs_request.The sequence number is returned via the 'seq_no' parameter + * in ibs_request structure. + * + * + * \param dev - \c [in] Device handle. + * See #amdgpu_device_initialize() + * \param context - \c [in] GPU Context + * \param flags - \c [in] Global submission flags + * \param ibs_request - \c [in/out] Pointer to submission requests. + * We could submit to the several + * engines/rings simulteniously as + * 'atomic' operation + * \param number_of_requests - \c [in] Number of submission requests + * + * \return 0 on success\n + * <0 - Negative POSIX Error code + * + * \note It is required to pass correct resource list with buffer handles + * which will be accessible by command buffers from submission + * This will allow kernel driver to correctly implement "paging". + * Failure to do so will have unpredictable results. + * + * \sa amdgpu_command_buffer_alloc(), amdgpu_command_buffer_free(), + * amdgpu_cs_query_fence_status() + * +*/ +int amdgpu_cs_submit(amdgpu_context_handle context, + uint64_t flags, + struct amdgpu_cs_request *ibs_request, + uint32_t number_of_requests); + +/** + * Query status of Command Buffer Submission + * + * \param fence - \c [in] Structure describing fence to query + * \param timeout_ns - \c [in] Timeout value to wait + * \param flags - \c [in] Flags for the query + * \param expired - \c [out] If fence expired or not.\n + * 0 – if fence is not expired\n + * !0 - otherwise + * + * \return 0 on success\n + * <0 - Negative POSIX Error code + * + * \note If UMD wants only to check operation status and returned immediately + * then timeout value as 0 must be passed. In this case success will be + * returned in the case if submission was completed or timeout error + * code. + * + * \sa amdgpu_cs_submit() +*/ +int amdgpu_cs_query_fence_status(struct amdgpu_cs_fence *fence, + uint64_t timeout_ns, + uint64_t flags, + uint32_t *expired); + +/** + * Wait for multiple fences + * + * \param fences - \c [in] The fence array to wait + * \param fence_count - \c [in] The fence count + * \param wait_all - \c [in] If true, wait all fences to be signaled, + * otherwise, wait at least one fence + * \param timeout_ns - \c [in] The timeout to wait, in nanoseconds + * \param status - \c [out] '1' for signaled, '0' for timeout + * \param first - \c [out] the index of the first signaled fence from @fences + * + * \return 0 on success + * <0 - Negative POSIX Error code + * + * \note Currently it supports only one amdgpu_device. All fences come from + * the same amdgpu_device with the same fd. +*/ +int amdgpu_cs_wait_fences(struct amdgpu_cs_fence *fences, + uint32_t fence_count, + bool wait_all, + uint64_t timeout_ns, + uint32_t *status, uint32_t *first); + +/* + * Query / Info API + * +*/ + +/** + * Query allocation size alignments + * + * UMD should query information about GPU VM MC size alignments requirements + * to be able correctly choose required allocation size and implement + * internal optimization if needed. + * + * \param dev - \c [in] Device handle. See #amdgpu_device_initialize() + * \param info - \c [out] Pointer to structure to get size alignment + * requirements + * + * \return 0 on success\n + * <0 - Negative POSIX Error code + * +*/ +int amdgpu_query_buffer_size_alignment(amdgpu_device_handle dev, + struct amdgpu_buffer_size_alignments + *info); + +/** + * Query firmware versions + * + * \param dev - \c [in] Device handle. See #amdgpu_device_initialize() + * \param fw_type - \c [in] AMDGPU_INFO_FW_* + * \param ip_instance - \c [in] Index of the IP block of the same type. + * \param index - \c [in] Index of the engine. (for SDMA and MEC) + * \param version - \c [out] Pointer to to the "version" return value + * \param feature - \c [out] Pointer to to the "feature" return value + * + * \return 0 on success\n + * <0 - Negative POSIX Error code + * +*/ +int amdgpu_query_firmware_version(amdgpu_device_handle dev, unsigned fw_type, + unsigned ip_instance, unsigned index, + uint32_t *version, uint32_t *feature); + +/** + * Query the number of HW IP instances of a certain type. + * + * \param dev - \c [in] Device handle. See #amdgpu_device_initialize() + * \param type - \c [in] Hardware IP block type = AMDGPU_HW_IP_* + * \param count - \c [out] Pointer to structure to get information + * + * \return 0 on success\n + * <0 - Negative POSIX Error code +*/ +int amdgpu_query_hw_ip_count(amdgpu_device_handle dev, unsigned type, + uint32_t *count); + +/** + * Query engine information + * + * This query allows UMD to query information different engines and their + * capabilities. + * + * \param dev - \c [in] Device handle. See #amdgpu_device_initialize() + * \param type - \c [in] Hardware IP block type = AMDGPU_HW_IP_* + * \param ip_instance - \c [in] Index of the IP block of the same type. + * \param info - \c [out] Pointer to structure to get information + * + * \return 0 on success\n + * <0 - Negative POSIX Error code +*/ +int amdgpu_query_hw_ip_info(amdgpu_device_handle dev, unsigned type, + unsigned ip_instance, + struct drm_amdgpu_info_hw_ip *info); + +/** + * Query heap information + * + * This query allows UMD to query potentially available memory resources and + * adjust their logic if necessary. + * + * \param dev - \c [in] Device handle. See #amdgpu_device_initialize() + * \param heap - \c [in] Heap type + * \param info - \c [in] Pointer to structure to get needed information + * + * \return 0 on success\n + * <0 - Negative POSIX Error code + * +*/ +int amdgpu_query_heap_info(amdgpu_device_handle dev, uint32_t heap, + uint32_t flags, struct amdgpu_heap_info *info); + +/** + * Get the CRTC ID from the mode object ID + * + * \param dev - \c [in] Device handle. See #amdgpu_device_initialize() + * \param id - \c [in] Mode object ID + * \param result - \c [in] Pointer to the CRTC ID + * + * \return 0 on success\n + * <0 - Negative POSIX Error code + * +*/ +int amdgpu_query_crtc_from_id(amdgpu_device_handle dev, unsigned id, + int32_t *result); + +/** + * Query GPU H/w Info + * + * Query hardware specific information + * + * \param dev - \c [in] Device handle. See #amdgpu_device_initialize() + * \param heap - \c [in] Heap type + * \param info - \c [in] Pointer to structure to get needed information + * + * \return 0 on success\n + * <0 - Negative POSIX Error code + * +*/ +int amdgpu_query_gpu_info(amdgpu_device_handle dev, + struct amdgpu_gpu_info *info); + +/** + * Query hardware or driver information. + * + * The return size is query-specific and depends on the "info_id" parameter. + * No more than "size" bytes is returned. + * + * \param dev - \c [in] Device handle. See #amdgpu_device_initialize() + * \param info_id - \c [in] AMDGPU_INFO_* + * \param size - \c [in] Size of the returned value. + * \param value - \c [out] Pointer to the return value. + * + * \return 0 on success\n + * <0 - Negative POSIX error code + * +*/ +int amdgpu_query_info(amdgpu_device_handle dev, unsigned info_id, + unsigned size, void *value); + +/** + * Query hardware or driver information. + * + * The return size is query-specific and depends on the "info_id" parameter. + * No more than "size" bytes is returned. + * + * \param dev - \c [in] Device handle. See #amdgpu_device_initialize() + * \param info - \c [in] amdgpu_sw_info_* + * \param value - \c [out] Pointer to the return value. + * + * \return 0 on success\n + * <0 - Negative POSIX error code + * +*/ +int amdgpu_query_sw_info(amdgpu_device_handle dev, enum amdgpu_sw_info info, + void *value); + +/** + * Query information about GDS + * + * \param dev - \c [in] Device handle. See #amdgpu_device_initialize() + * \param gds_info - \c [out] Pointer to structure to get GDS information + * + * \return 0 on success\n + * <0 - Negative POSIX Error code + * +*/ +int amdgpu_query_gds_info(amdgpu_device_handle dev, + struct amdgpu_gds_resource_info *gds_info); + +/** + * Query information about sensor. + * + * The return size is query-specific and depends on the "sensor_type" + * parameter. No more than "size" bytes is returned. + * + * \param dev - \c [in] Device handle. See #amdgpu_device_initialize() + * \param sensor_type - \c [in] AMDGPU_INFO_SENSOR_* + * \param size - \c [in] Size of the returned value. + * \param value - \c [out] Pointer to the return value. + * + * \return 0 on success\n + * <0 - Negative POSIX Error code + * +*/ +int amdgpu_query_sensor_info(amdgpu_device_handle dev, unsigned sensor_type, + unsigned size, void *value); + +/** + * Query information about video capabilities + * + * The return sizeof(struct drm_amdgpu_info_video_caps) + * + * \param dev - \c [in] Device handle. See #amdgpu_device_initialize() + * \param caps_type - \c [in] AMDGPU_INFO_VIDEO_CAPS_DECODE(ENCODE) + * \param size - \c [in] Size of the returned value. + * \param value - \c [out] Pointer to the return value. + * + * \return 0 on success\n + * <0 - Negative POSIX Error code + * +*/ +int amdgpu_query_video_caps_info(amdgpu_device_handle dev, unsigned cap_type, + unsigned size, void *value); + +/** + * Query information about VM faults + * + * The return sizeof(struct drm_amdgpu_info_gpuvm_fault) + * + * \param dev - \c [in] Device handle. See #amdgpu_device_initialize() + * \param size - \c [in] Size of the returned value. + * \param value - \c [out] Pointer to the return value. + * + * \return 0 on success\n + * <0 - Negative POSIX Error code + * +*/ +int amdgpu_query_gpuvm_fault_info(amdgpu_device_handle dev, unsigned size, + void *value); + +/** + * Read a set of consecutive memory-mapped registers. + * Not all registers are allowed to be read by userspace. + * + * \param dev - \c [in] Device handle. See #amdgpu_device_initialize( + * \param dword_offset - \c [in] Register offset in dwords + * \param count - \c [in] The number of registers to read starting + * from the offset + * \param instance - \c [in] GRBM_GFX_INDEX selector. It may have other + * uses. Set it to 0xffffffff if unsure. + * \param flags - \c [in] Flags with additional information. + * \param values - \c [out] The pointer to return values. + * + * \return 0 on success\n + * <0 - Negative POSIX error code + * +*/ +int amdgpu_read_mm_registers(amdgpu_device_handle dev, unsigned dword_offset, + unsigned count, uint32_t instance, uint32_t flags, + uint32_t *values); + +/** + * Flag to request VA address range in the 32bit address space +*/ +#define AMDGPU_VA_RANGE_32_BIT 0x1 +#define AMDGPU_VA_RANGE_HIGH 0x2 +#define AMDGPU_VA_RANGE_REPLAYABLE 0x4 + +/** + * Allocate virtual address range + * + * \param dev - [in] Device handle. See #amdgpu_device_initialize() + * \param va_range_type - \c [in] Type of MC va range from which to allocate + * \param size - \c [in] Size of range. Size must be correctly* aligned. + * It is client responsibility to correctly aligned size based on the future + * usage of allocated range. + * \param va_base_alignment - \c [in] Overwrite base address alignment + * requirement for GPU VM MC virtual + * address assignment. Must be multiple of size alignments received as + * 'amdgpu_buffer_size_alignments'. + * If 0 use the default one. + * \param va_base_required - \c [in] Specified required va base address. + * If 0 then library choose available one. + * If !0 value will be passed and those value already "in use" then + * corresponding error status will be returned. + * \param va_base_allocated - \c [out] On return: Allocated VA base to be used + * by client. + * \param va_range_handle - \c [out] On return: Handle assigned to allocation + * \param flags - \c [in] flags for special VA range + * + * \return 0 on success\n + * >0 - AMD specific error code\n + * <0 - Negative POSIX Error code + * + * \notes \n + * It is client responsibility to correctly handle VA assignments and usage. + * Neither kernel driver nor libdrm_amdpgu are able to prevent and + * detect wrong va assignment. + * + * It is client responsibility to correctly handle multi-GPU cases and to pass + * the corresponding arrays of all devices handles where corresponding VA will + * be used. + * +*/ +int amdgpu_va_range_alloc(amdgpu_device_handle dev, + enum amdgpu_gpu_va_range va_range_type, + uint64_t size, + uint64_t va_base_alignment, + uint64_t va_base_required, + uint64_t *va_base_allocated, + amdgpu_va_handle *va_range_handle, + uint64_t flags); + +/** + * Free previously allocated virtual address range + * + * + * \param va_range_handle - \c [in] Handle assigned to VA allocation + * + * \return 0 on success\n + * >0 - AMD specific error code\n + * <0 - Negative POSIX Error code + * +*/ +int amdgpu_va_range_free(amdgpu_va_handle va_range_handle); + +/** + * Return the starting address of the allocated virtual address range. + */ +uint64_t amdgpu_va_get_start_addr(amdgpu_va_handle va_handle); + +/** +* Query virtual address range +* +* UMD can query GPU VM range supported by each device +* to initialize its own VAM accordingly. +* +* \param dev - [in] Device handle. See #amdgpu_device_initialize() +* \param type - \c [in] Type of virtual address range +* \param offset - \c [out] Start offset of virtual address range +* \param size - \c [out] Size of virtual address range +* +* \return 0 on success\n +* <0 - Negative POSIX Error code +* +*/ + +int amdgpu_va_range_query(amdgpu_device_handle dev, + enum amdgpu_gpu_va_range type, + uint64_t *start, + uint64_t *end); + +/** + * VA mapping/unmapping for the buffer object + * + * \param bo - \c [in] BO handle + * \param offset - \c [in] Start offset to map + * \param size - \c [in] Size to map + * \param addr - \c [in] Start virtual address. + * \param flags - \c [in] Supported flags for mapping/unmapping + * \param ops - \c [in] AMDGPU_VA_OP_MAP or AMDGPU_VA_OP_UNMAP + * + * \return 0 on success\n + * <0 - Negative POSIX Error code + * +*/ + +int amdgpu_bo_va_op(amdgpu_bo_handle bo, + uint64_t offset, + uint64_t size, + uint64_t addr, + uint64_t flags, + uint32_t ops); + +/** + * VA mapping/unmapping for a buffer object or PRT region. + * + * This is not a simple drop-in extension for amdgpu_bo_va_op; instead, all + * parameters are treated "raw", i.e. size is not automatically aligned, and + * all flags must be specified explicitly. + * + * \param dev - \c [in] device handle + * \param bo - \c [in] BO handle (may be NULL) + * \param offset - \c [in] Start offset to map + * \param size - \c [in] Size to map + * \param addr - \c [in] Start virtual address. + * \param flags - \c [in] Supported flags for mapping/unmapping + * \param ops - \c [in] AMDGPU_VA_OP_MAP or AMDGPU_VA_OP_UNMAP + * + * \return 0 on success\n + * <0 - Negative POSIX Error code + * +*/ + +int amdgpu_bo_va_op_raw(amdgpu_device_handle dev, + amdgpu_bo_handle bo, + uint64_t offset, + uint64_t size, + uint64_t addr, + uint64_t flags, + uint32_t ops); + +/** + * create semaphore + * + * \param sem - \c [out] semaphore handle + * + * \return 0 on success\n + * <0 - Negative POSIX Error code + * +*/ +int amdgpu_cs_create_semaphore(amdgpu_semaphore_handle *sem); + +/** + * signal semaphore + * + * \param context - \c [in] GPU Context + * \param ip_type - \c [in] Hardware IP block type = AMDGPU_HW_IP_* + * \param ip_instance - \c [in] Index of the IP block of the same type + * \param ring - \c [in] Specify ring index of the IP + * \param sem - \c [in] semaphore handle + * + * \return 0 on success\n + * <0 - Negative POSIX Error code + * +*/ +int amdgpu_cs_signal_semaphore(amdgpu_context_handle ctx, + uint32_t ip_type, + uint32_t ip_instance, + uint32_t ring, + amdgpu_semaphore_handle sem); + +/** + * wait semaphore + * + * \param context - \c [in] GPU Context + * \param ip_type - \c [in] Hardware IP block type = AMDGPU_HW_IP_* + * \param ip_instance - \c [in] Index of the IP block of the same type + * \param ring - \c [in] Specify ring index of the IP + * \param sem - \c [in] semaphore handle + * + * \return 0 on success\n + * <0 - Negative POSIX Error code + * +*/ +int amdgpu_cs_wait_semaphore(amdgpu_context_handle ctx, + uint32_t ip_type, + uint32_t ip_instance, + uint32_t ring, + amdgpu_semaphore_handle sem); + +/** + * destroy semaphore + * + * \param sem - \c [in] semaphore handle + * + * \return 0 on success\n + * <0 - Negative POSIX Error code + * +*/ +int amdgpu_cs_destroy_semaphore(amdgpu_semaphore_handle sem); + +/** + * Get the ASIC marketing name + * + * \param dev - \c [in] Device handle. See #amdgpu_device_initialize() + * + * \return the constant string of the marketing name + * "NULL" means the ASIC is not found +*/ +const char *amdgpu_get_marketing_name(amdgpu_device_handle dev); + +/** + * Create kernel sync object + * + * \param dev - \c [in] device handle + * \param flags - \c [in] flags that affect creation + * \param syncobj - \c [out] sync object handle + * + * \return 0 on success\n + * <0 - Negative POSIX Error code + * +*/ +int amdgpu_cs_create_syncobj2(amdgpu_device_handle dev, + uint32_t flags, + uint32_t *syncobj); + +/** + * Create kernel sync object + * + * \param dev - \c [in] device handle + * \param syncobj - \c [out] sync object handle + * + * \return 0 on success\n + * <0 - Negative POSIX Error code + * +*/ +int amdgpu_cs_create_syncobj(amdgpu_device_handle dev, + uint32_t *syncobj); +/** + * Destroy kernel sync object + * + * \param dev - \c [in] device handle + * \param syncobj - \c [in] sync object handle + * + * \return 0 on success\n + * <0 - Negative POSIX Error code + * +*/ +int amdgpu_cs_destroy_syncobj(amdgpu_device_handle dev, + uint32_t syncobj); + +/** + * Reset kernel sync objects to unsignalled state. + * + * \param dev - \c [in] device handle + * \param syncobjs - \c [in] array of sync object handles + * \param syncobj_count - \c [in] number of handles in syncobjs + * + * \return 0 on success\n + * <0 - Negative POSIX Error code + * +*/ +int amdgpu_cs_syncobj_reset(amdgpu_device_handle dev, + const uint32_t *syncobjs, uint32_t syncobj_count); + +/** + * Signal kernel sync objects. + * + * \param dev - \c [in] device handle + * \param syncobjs - \c [in] array of sync object handles + * \param syncobj_count - \c [in] number of handles in syncobjs + * + * \return 0 on success\n + * <0 - Negative POSIX Error code + * +*/ +int amdgpu_cs_syncobj_signal(amdgpu_device_handle dev, + const uint32_t *syncobjs, uint32_t syncobj_count); + +/** + * Signal kernel timeline sync objects. + * + * \param dev - \c [in] device handle + * \param syncobjs - \c [in] array of sync object handles + * \param points - \c [in] array of timeline points + * \param syncobj_count - \c [in] number of handles in syncobjs + * + * \return 0 on success\n + * <0 - Negative POSIX Error code + * +*/ +int amdgpu_cs_syncobj_timeline_signal(amdgpu_device_handle dev, + const uint32_t *syncobjs, + uint64_t *points, + uint32_t syncobj_count); + +/** + * Wait for one or all sync objects to signal. + * + * \param dev - \c [in] self-explanatory + * \param handles - \c [in] array of sync object handles + * \param num_handles - \c [in] self-explanatory + * \param timeout_nsec - \c [in] self-explanatory + * \param flags - \c [in] a bitmask of DRM_SYNCOBJ_WAIT_FLAGS_* + * \param first_signaled - \c [in] self-explanatory + * + * \return 0 on success\n + * -ETIME - Timeout + * <0 - Negative POSIX Error code + * + */ +int amdgpu_cs_syncobj_wait(amdgpu_device_handle dev, + uint32_t *handles, unsigned num_handles, + int64_t timeout_nsec, unsigned flags, + uint32_t *first_signaled); + +/** + * Wait for one or all sync objects on their points to signal. + * + * \param dev - \c [in] self-explanatory + * \param handles - \c [in] array of sync object handles + * \param points - \c [in] array of sync points to wait + * \param num_handles - \c [in] self-explanatory + * \param timeout_nsec - \c [in] self-explanatory + * \param flags - \c [in] a bitmask of DRM_SYNCOBJ_WAIT_FLAGS_* + * \param first_signaled - \c [in] self-explanatory + * + * \return 0 on success\n + * -ETIME - Timeout + * <0 - Negative POSIX Error code + * + */ +int amdgpu_cs_syncobj_timeline_wait(amdgpu_device_handle dev, + uint32_t *handles, uint64_t *points, + unsigned num_handles, + int64_t timeout_nsec, unsigned flags, + uint32_t *first_signaled); +/** + * Query sync objects payloads. + * + * \param dev - \c [in] self-explanatory + * \param handles - \c [in] array of sync object handles + * \param points - \c [out] array of sync points returned, which presents + * syncobj payload. + * \param num_handles - \c [in] self-explanatory + * + * \return 0 on success\n + * -ETIME - Timeout + * <0 - Negative POSIX Error code + * + */ +int amdgpu_cs_syncobj_query(amdgpu_device_handle dev, + uint32_t *handles, uint64_t *points, + unsigned num_handles); +/** + * Query sync objects last signaled or submitted point. + * + * \param dev - \c [in] self-explanatory + * \param handles - \c [in] array of sync object handles + * \param points - \c [out] array of sync points returned, which presents + * syncobj payload. + * \param num_handles - \c [in] self-explanatory + * \param flags - \c [in] a bitmask of DRM_SYNCOBJ_QUERY_FLAGS_* + * + * \return 0 on success\n + * -ETIME - Timeout + * <0 - Negative POSIX Error code + * + */ +int amdgpu_cs_syncobj_query2(amdgpu_device_handle dev, + uint32_t *handles, uint64_t *points, + unsigned num_handles, uint32_t flags); + +/** + * Export kernel sync object to shareable fd. + * + * \param dev - \c [in] device handle + * \param syncobj - \c [in] sync object handle + * \param shared_fd - \c [out] shared file descriptor. + * + * \return 0 on success\n + * <0 - Negative POSIX Error code + * +*/ +int amdgpu_cs_export_syncobj(amdgpu_device_handle dev, + uint32_t syncobj, + int *shared_fd); +/** + * Import kernel sync object from shareable fd. + * + * \param dev - \c [in] device handle + * \param shared_fd - \c [in] shared file descriptor. + * \param syncobj - \c [out] sync object handle + * + * \return 0 on success\n + * <0 - Negative POSIX Error code + * +*/ +int amdgpu_cs_import_syncobj(amdgpu_device_handle dev, + int shared_fd, + uint32_t *syncobj); + +/** + * Export kernel sync object to a sync_file. + * + * \param dev - \c [in] device handle + * \param syncobj - \c [in] sync object handle + * \param sync_file_fd - \c [out] sync_file file descriptor. + * + * \return 0 on success\n + * <0 - Negative POSIX Error code + * + */ +int amdgpu_cs_syncobj_export_sync_file(amdgpu_device_handle dev, + uint32_t syncobj, + int *sync_file_fd); + +/** + * Import kernel sync object from a sync_file. + * + * \param dev - \c [in] device handle + * \param syncobj - \c [in] sync object handle + * \param sync_file_fd - \c [in] sync_file file descriptor. + * + * \return 0 on success\n + * <0 - Negative POSIX Error code + * + */ +int amdgpu_cs_syncobj_import_sync_file(amdgpu_device_handle dev, + uint32_t syncobj, + int sync_file_fd); +/** + * Export kernel timeline sync object to a sync_file. + * + * \param dev - \c [in] device handle + * \param syncobj - \c [in] sync object handle + * \param point - \c [in] timeline point + * \param flags - \c [in] flags + * \param sync_file_fd - \c [out] sync_file file descriptor. + * + * \return 0 on success\n + * <0 - Negative POSIX Error code + * + */ +int amdgpu_cs_syncobj_export_sync_file2(amdgpu_device_handle dev, + uint32_t syncobj, + uint64_t point, + uint32_t flags, + int *sync_file_fd); + +/** + * Import kernel timeline sync object from a sync_file. + * + * \param dev - \c [in] device handle + * \param syncobj - \c [in] sync object handle + * \param point - \c [in] timeline point + * \param sync_file_fd - \c [in] sync_file file descriptor. + * + * \return 0 on success\n + * <0 - Negative POSIX Error code + * + */ +int amdgpu_cs_syncobj_import_sync_file2(amdgpu_device_handle dev, + uint32_t syncobj, + uint64_t point, + int sync_file_fd); + +/** + * transfer between syncbojs. + * + * \param dev - \c [in] device handle + * \param dst_handle - \c [in] sync object handle + * \param dst_point - \c [in] timeline point, 0 presents dst is binary + * \param src_handle - \c [in] sync object handle + * \param src_point - \c [in] timeline point, 0 presents src is binary + * \param flags - \c [in] flags + * + * \return 0 on success\n + * <0 - Negative POSIX Error code + * + */ +int amdgpu_cs_syncobj_transfer(amdgpu_device_handle dev, + uint32_t dst_handle, + uint64_t dst_point, + uint32_t src_handle, + uint64_t src_point, + uint32_t flags); + +/** + * Export an amdgpu fence as a handle (syncobj or fd). + * + * \param what AMDGPU_FENCE_TO_HANDLE_GET_{SYNCOBJ, FD} + * \param out_handle returned handle + * + * \return 0 on success\n + * <0 - Negative POSIX Error code + */ +int amdgpu_cs_fence_to_handle(amdgpu_device_handle dev, + struct amdgpu_cs_fence *fence, + uint32_t what, + uint32_t *out_handle); + +/** + * Submit raw command submission to kernel + * + * \param dev - \c [in] device handle + * \param context - \c [in] context handle for context id + * \param bo_list_handle - \c [in] request bo list handle (0 for none) + * \param num_chunks - \c [in] number of CS chunks to submit + * \param chunks - \c [in] array of CS chunks + * \param seq_no - \c [out] output sequence number for submission. + * + * \return 0 on success\n + * <0 - Negative POSIX Error code + * + */ +struct drm_amdgpu_cs_chunk; +struct drm_amdgpu_cs_chunk_dep; +struct drm_amdgpu_cs_chunk_data; + +int amdgpu_cs_submit_raw(amdgpu_device_handle dev, + amdgpu_context_handle context, + amdgpu_bo_list_handle bo_list_handle, + int num_chunks, + struct drm_amdgpu_cs_chunk *chunks, + uint64_t *seq_no); + +/** + * Submit raw command submission to the kernel with a raw BO list handle. + * + * \param dev - \c [in] device handle + * \param context - \c [in] context handle for context id + * \param bo_list_handle - \c [in] raw bo list handle (0 for none) + * \param num_chunks - \c [in] number of CS chunks to submit + * \param chunks - \c [in] array of CS chunks + * \param seq_no - \c [out] output sequence number for submission. + * + * \return 0 on success\n + * <0 - Negative POSIX Error code + * + * \sa amdgpu_bo_list_create_raw(), amdgpu_bo_list_destroy_raw() + */ +int amdgpu_cs_submit_raw2(amdgpu_device_handle dev, + amdgpu_context_handle context, + uint32_t bo_list_handle, + int num_chunks, + struct drm_amdgpu_cs_chunk *chunks, + uint64_t *seq_no); + +void amdgpu_cs_chunk_fence_to_dep(struct amdgpu_cs_fence *fence, + struct drm_amdgpu_cs_chunk_dep *dep); +void amdgpu_cs_chunk_fence_info_to_data(struct amdgpu_cs_fence_info *fence_info, + struct drm_amdgpu_cs_chunk_data *data); + +/** + * Reserve VMID + * \param context - \c [in] GPU Context + * \param flags - \c [in] TBD + * + * \return 0 on success otherwise POSIX Error code +*/ +int amdgpu_vm_reserve_vmid(amdgpu_device_handle dev, uint32_t flags); + +/** + * Free reserved VMID + * \param context - \c [in] GPU Context + * \param flags - \c [in] TBD + * + * \return 0 on success otherwise POSIX Error code +*/ +int amdgpu_vm_unreserve_vmid(amdgpu_device_handle dev, uint32_t flags); + +#ifdef __cplusplus +} +#endif +#endif /* #ifdef _AMDGPU_H_ */ diff --git a/amdgpu/amdgpu_asic_id.c b/amdgpu/amdgpu_asic_id.c new file mode 100644 index 0000000..a5007ff --- /dev/null +++ b/amdgpu/amdgpu_asic_id.c @@ -0,0 +1,161 @@ +/* + * Copyright © 2017 Advanced Micro Devices, Inc. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * the rights to use, copy, modify, merge, publish, distribute, sublicense, + * and/or sell copies of the Software, and to permit persons to whom the + * Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice shall be included in + * all copies or substantial portions of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL + * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR + * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, + * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR + * OTHER DEALINGS IN THE SOFTWARE. + * + */ + +#include +#include +#include +#include +#include +#include +#include + +#include "xf86drm.h" +#include "amdgpu_drm.h" +#include "amdgpu_internal.h" + +static int parse_one_line(struct amdgpu_device *dev, const char *line) +{ + char *buf, *saveptr; + char *s_did; + uint32_t did; + char *s_rid; + uint32_t rid; + char *s_name; + char *endptr; + int r = -EINVAL; + + /* ignore empty line and commented line */ + if (strlen(line) == 0 || line[0] == '#') + return -EAGAIN; + + buf = strdup(line); + if (!buf) + return -ENOMEM; + + /* device id */ + s_did = strtok_r(buf, ",", &saveptr); + if (!s_did) + goto out; + + did = strtol(s_did, &endptr, 16); + if (*endptr) + goto out; + + if (did != dev->info.asic_id) { + r = -EAGAIN; + goto out; + } + + /* revision id */ + s_rid = strtok_r(NULL, ",", &saveptr); + if (!s_rid) + goto out; + + rid = strtol(s_rid, &endptr, 16); + if (*endptr) + goto out; + + if (rid != dev->info.pci_rev_id) { + r = -EAGAIN; + goto out; + } + + /* marketing name */ + s_name = strtok_r(NULL, ",", &saveptr); + if (!s_name) + goto out; + + /* trim leading whitespaces or tabs */ + while (isblank(*s_name)) + s_name++; + if (strlen(s_name) == 0) + goto out; + + dev->marketing_name = strdup(s_name); + if (dev->marketing_name) + r = 0; + else + r = -ENOMEM; + +out: + free(buf); + + return r; +} + +void amdgpu_parse_asic_ids(struct amdgpu_device *dev) +{ + FILE *fp; + char *line = NULL; + size_t len = 0; + ssize_t n; + int line_num = 1; + int r = 0; + + fp = fopen(AMDGPU_ASIC_ID_TABLE, "r"); + if (!fp) { + fprintf(stderr, "%s: %s\n", AMDGPU_ASIC_ID_TABLE, + strerror(errno)); + return; + } + + /* 1st valid line is file version */ + while ((n = getline(&line, &len, fp)) != -1) { + /* trim trailing newline */ + if (line[n - 1] == '\n') + line[n - 1] = '\0'; + + /* ignore empty line and commented line */ + if (strlen(line) == 0 || line[0] == '#') { + line_num++; + continue; + } + + drmMsg("%s version: %s\n", AMDGPU_ASIC_ID_TABLE, line); + break; + } + + while ((n = getline(&line, &len, fp)) != -1) { + /* trim trailing newline */ + if (line[n - 1] == '\n') + line[n - 1] = '\0'; + + r = parse_one_line(dev, line); + if (r != -EAGAIN) + break; + + line_num++; + } + + if (r == -EINVAL) { + fprintf(stderr, "Invalid format: %s: line %d: %s\n", + AMDGPU_ASIC_ID_TABLE, line_num, line); + } else if (r && r != -EAGAIN) { + fprintf(stderr, "%s: Cannot parse ASIC IDs: %s\n", + __func__, strerror(-r)); + } + + free(line); + fclose(fp); +} diff --git a/amdgpu/amdgpu_bo.c b/amdgpu/amdgpu_bo.c new file mode 100644 index 0000000..672f000 --- /dev/null +++ b/amdgpu/amdgpu_bo.c @@ -0,0 +1,791 @@ +/* + * Copyright © 2014 Advanced Micro Devices, Inc. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * the rights to use, copy, modify, merge, publish, distribute, sublicense, + * and/or sell copies of the Software, and to permit persons to whom the + * Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice shall be included in + * all copies or substantial portions of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL + * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR + * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, + * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR + * OTHER DEALINGS IN THE SOFTWARE. + * + */ + +#include +#include +#include +#include +#include +#include +#include +#include +#include +#include + +#include "libdrm_macros.h" +#include "xf86drm.h" +#include "amdgpu_drm.h" +#include "amdgpu_internal.h" +#include "util_math.h" + +static int amdgpu_bo_create(amdgpu_device_handle dev, + uint64_t size, + uint32_t handle, + amdgpu_bo_handle *buf_handle) +{ + struct amdgpu_bo *bo; + int r; + + bo = calloc(1, sizeof(struct amdgpu_bo)); + if (!bo) + return -ENOMEM; + + r = handle_table_insert(&dev->bo_handles, handle, bo); + if (r) { + free(bo); + return r; + } + + atomic_set(&bo->refcount, 1); + bo->dev = dev; + bo->alloc_size = size; + bo->handle = handle; + pthread_mutex_init(&bo->cpu_access_mutex, NULL); + + *buf_handle = bo; + return 0; +} + +drm_public int amdgpu_bo_alloc(amdgpu_device_handle dev, + struct amdgpu_bo_alloc_request *alloc_buffer, + amdgpu_bo_handle *buf_handle) +{ + union drm_amdgpu_gem_create args; + int r; + + memset(&args, 0, sizeof(args)); + args.in.bo_size = alloc_buffer->alloc_size; + args.in.alignment = alloc_buffer->phys_alignment; + + /* Set the placement. */ + args.in.domains = alloc_buffer->preferred_heap; + args.in.domain_flags = alloc_buffer->flags; + + /* Allocate the buffer with the preferred heap. */ + r = drmCommandWriteRead(dev->fd, DRM_AMDGPU_GEM_CREATE, + &args, sizeof(args)); + if (r) + goto out; + + pthread_mutex_lock(&dev->bo_table_mutex); + r = amdgpu_bo_create(dev, alloc_buffer->alloc_size, args.out.handle, + buf_handle); + pthread_mutex_unlock(&dev->bo_table_mutex); + if (r) { + drmCloseBufferHandle(dev->fd, args.out.handle); + } + +out: + return r; +} + +drm_public int amdgpu_bo_set_metadata(amdgpu_bo_handle bo, + struct amdgpu_bo_metadata *info) +{ + struct drm_amdgpu_gem_metadata args = {}; + + args.handle = bo->handle; + args.op = AMDGPU_GEM_METADATA_OP_SET_METADATA; + args.data.flags = info->flags; + args.data.tiling_info = info->tiling_info; + + if (info->size_metadata > sizeof(args.data.data)) + return -EINVAL; + + if (info->size_metadata) { + args.data.data_size_bytes = info->size_metadata; + memcpy(args.data.data, info->umd_metadata, info->size_metadata); + } + + return drmCommandWriteRead(bo->dev->fd, + DRM_AMDGPU_GEM_METADATA, + &args, sizeof(args)); +} + +drm_public int amdgpu_bo_query_info(amdgpu_bo_handle bo, + struct amdgpu_bo_info *info) +{ + struct drm_amdgpu_gem_metadata metadata = {}; + struct drm_amdgpu_gem_create_in bo_info = {}; + struct drm_amdgpu_gem_op gem_op = {}; + int r; + + /* Validate the BO passed in */ + if (!bo->handle) + return -EINVAL; + + /* Query metadata. */ + metadata.handle = bo->handle; + metadata.op = AMDGPU_GEM_METADATA_OP_GET_METADATA; + + r = drmCommandWriteRead(bo->dev->fd, DRM_AMDGPU_GEM_METADATA, + &metadata, sizeof(metadata)); + if (r) + return r; + + if (metadata.data.data_size_bytes > + sizeof(info->metadata.umd_metadata)) + return -EINVAL; + + /* Query buffer info. */ + gem_op.handle = bo->handle; + gem_op.op = AMDGPU_GEM_OP_GET_GEM_CREATE_INFO; + gem_op.value = (uintptr_t)&bo_info; + + r = drmCommandWriteRead(bo->dev->fd, DRM_AMDGPU_GEM_OP, + &gem_op, sizeof(gem_op)); + if (r) + return r; + + memset(info, 0, sizeof(*info)); + info->alloc_size = bo_info.bo_size; + info->phys_alignment = bo_info.alignment; + info->preferred_heap = bo_info.domains; + info->alloc_flags = bo_info.domain_flags; + info->metadata.flags = metadata.data.flags; + info->metadata.tiling_info = metadata.data.tiling_info; + + info->metadata.size_metadata = metadata.data.data_size_bytes; + if (metadata.data.data_size_bytes > 0) + memcpy(info->metadata.umd_metadata, metadata.data.data, + metadata.data.data_size_bytes); + + return 0; +} + +static int amdgpu_bo_export_flink(amdgpu_bo_handle bo) +{ + struct drm_gem_flink flink; + int fd, dma_fd; + uint32_t handle; + int r; + + fd = bo->dev->fd; + handle = bo->handle; + if (bo->flink_name) + return 0; + + + if (bo->dev->flink_fd != bo->dev->fd) { + r = drmPrimeHandleToFD(bo->dev->fd, bo->handle, DRM_CLOEXEC, + &dma_fd); + if (!r) { + r = drmPrimeFDToHandle(bo->dev->flink_fd, dma_fd, &handle); + close(dma_fd); + } + if (r) + return r; + fd = bo->dev->flink_fd; + } + memset(&flink, 0, sizeof(flink)); + flink.handle = handle; + + r = drmIoctl(fd, DRM_IOCTL_GEM_FLINK, &flink); + if (r) + return r; + + bo->flink_name = flink.name; + + if (bo->dev->flink_fd != bo->dev->fd) + drmCloseBufferHandle(bo->dev->flink_fd, handle); + + pthread_mutex_lock(&bo->dev->bo_table_mutex); + r = handle_table_insert(&bo->dev->bo_flink_names, bo->flink_name, bo); + pthread_mutex_unlock(&bo->dev->bo_table_mutex); + + return r; +} + +drm_public int amdgpu_bo_export(amdgpu_bo_handle bo, + enum amdgpu_bo_handle_type type, + uint32_t *shared_handle) +{ + int r; + + switch (type) { + case amdgpu_bo_handle_type_gem_flink_name: + r = amdgpu_bo_export_flink(bo); + if (r) + return r; + + *shared_handle = bo->flink_name; + return 0; + + case amdgpu_bo_handle_type_kms: + case amdgpu_bo_handle_type_kms_noimport: + *shared_handle = bo->handle; + return 0; + + case amdgpu_bo_handle_type_dma_buf_fd: + return drmPrimeHandleToFD(bo->dev->fd, bo->handle, + DRM_CLOEXEC | DRM_RDWR, + (int*)shared_handle); + } + return -EINVAL; +} + +drm_public int amdgpu_bo_import(amdgpu_device_handle dev, + enum amdgpu_bo_handle_type type, + uint32_t shared_handle, + struct amdgpu_bo_import_result *output) +{ + struct drm_gem_open open_arg = {}; + struct amdgpu_bo *bo = NULL; + uint32_t handle = 0, flink_name = 0; + uint64_t alloc_size = 0; + int r = 0; + int dma_fd; + uint64_t dma_buf_size = 0; + + /* We must maintain a list of pairs , so that we always + * return the same amdgpu_bo instance for the same handle. */ + pthread_mutex_lock(&dev->bo_table_mutex); + + /* Convert a DMA buf handle to a KMS handle now. */ + if (type == amdgpu_bo_handle_type_dma_buf_fd) { + off_t size; + + /* Get a KMS handle. */ + r = drmPrimeFDToHandle(dev->fd, shared_handle, &handle); + if (r) + goto unlock; + + /* Query the buffer size. */ + size = lseek(shared_handle, 0, SEEK_END); + if (size == (off_t)-1) { + r = -errno; + goto free_bo_handle; + } + lseek(shared_handle, 0, SEEK_SET); + + dma_buf_size = size; + shared_handle = handle; + } + + /* If we have already created a buffer with this handle, find it. */ + switch (type) { + case amdgpu_bo_handle_type_gem_flink_name: + bo = handle_table_lookup(&dev->bo_flink_names, shared_handle); + break; + + case amdgpu_bo_handle_type_dma_buf_fd: + bo = handle_table_lookup(&dev->bo_handles, shared_handle); + break; + + case amdgpu_bo_handle_type_kms: + case amdgpu_bo_handle_type_kms_noimport: + /* Importing a KMS handle in not allowed. */ + r = -EPERM; + goto unlock; + + default: + r = -EINVAL; + goto unlock; + } + + if (bo) { + /* The buffer already exists, just bump the refcount. */ + atomic_inc(&bo->refcount); + pthread_mutex_unlock(&dev->bo_table_mutex); + + output->buf_handle = bo; + output->alloc_size = bo->alloc_size; + return 0; + } + + /* Open the handle. */ + switch (type) { + case amdgpu_bo_handle_type_gem_flink_name: + open_arg.name = shared_handle; + r = drmIoctl(dev->flink_fd, DRM_IOCTL_GEM_OPEN, &open_arg); + if (r) + goto unlock; + + flink_name = shared_handle; + handle = open_arg.handle; + alloc_size = open_arg.size; + if (dev->flink_fd != dev->fd) { + r = drmPrimeHandleToFD(dev->flink_fd, handle, + DRM_CLOEXEC, &dma_fd); + if (r) + goto free_bo_handle; + r = drmPrimeFDToHandle(dev->fd, dma_fd, &handle); + close(dma_fd); + if (r) + goto free_bo_handle; + r = drmCloseBufferHandle(dev->flink_fd, + open_arg.handle); + if (r) + goto free_bo_handle; + } + open_arg.handle = 0; + break; + + case amdgpu_bo_handle_type_dma_buf_fd: + handle = shared_handle; + alloc_size = dma_buf_size; + break; + + case amdgpu_bo_handle_type_kms: + case amdgpu_bo_handle_type_kms_noimport: + assert(0); /* unreachable */ + } + + /* Initialize it. */ + r = amdgpu_bo_create(dev, alloc_size, handle, &bo); + if (r) + goto free_bo_handle; + + if (flink_name) { + bo->flink_name = flink_name; + r = handle_table_insert(&dev->bo_flink_names, flink_name, + bo); + if (r) + goto free_bo_handle; + + } + + output->buf_handle = bo; + output->alloc_size = bo->alloc_size; + pthread_mutex_unlock(&dev->bo_table_mutex); + return 0; + +free_bo_handle: + if (flink_name && open_arg.handle) + drmCloseBufferHandle(dev->flink_fd, open_arg.handle); + + if (bo) + amdgpu_bo_free(bo); + else + drmCloseBufferHandle(dev->fd, handle); +unlock: + pthread_mutex_unlock(&dev->bo_table_mutex); + return r; +} + +drm_public int amdgpu_bo_free(amdgpu_bo_handle buf_handle) +{ + struct amdgpu_device *dev; + struct amdgpu_bo *bo = buf_handle; + + assert(bo != NULL); + dev = bo->dev; + pthread_mutex_lock(&dev->bo_table_mutex); + + if (update_references(&bo->refcount, NULL)) { + /* Remove the buffer from the hash tables. */ + handle_table_remove(&dev->bo_handles, bo->handle); + + if (bo->flink_name) + handle_table_remove(&dev->bo_flink_names, + bo->flink_name); + + /* Release CPU access. */ + if (bo->cpu_map_count > 0) { + bo->cpu_map_count = 1; + amdgpu_bo_cpu_unmap(bo); + } + + drmCloseBufferHandle(dev->fd, bo->handle); + pthread_mutex_destroy(&bo->cpu_access_mutex); + free(bo); + } + + pthread_mutex_unlock(&dev->bo_table_mutex); + + return 0; +} + +drm_public void amdgpu_bo_inc_ref(amdgpu_bo_handle bo) +{ + atomic_inc(&bo->refcount); +} + +drm_public int amdgpu_bo_cpu_map(amdgpu_bo_handle bo, void **cpu) +{ + union drm_amdgpu_gem_mmap args; + void *ptr; + int r; + + pthread_mutex_lock(&bo->cpu_access_mutex); + + if (bo->cpu_ptr) { + /* already mapped */ + assert(bo->cpu_map_count > 0); + bo->cpu_map_count++; + *cpu = bo->cpu_ptr; + pthread_mutex_unlock(&bo->cpu_access_mutex); + return 0; + } + + assert(bo->cpu_map_count == 0); + + memset(&args, 0, sizeof(args)); + + /* Query the buffer address (args.addr_ptr). + * The kernel driver ignores the offset and size parameters. */ + args.in.handle = bo->handle; + + r = drmCommandWriteRead(bo->dev->fd, DRM_AMDGPU_GEM_MMAP, &args, + sizeof(args)); + if (r) { + pthread_mutex_unlock(&bo->cpu_access_mutex); + return r; + } + + /* Map the buffer. */ + ptr = drm_mmap(NULL, bo->alloc_size, PROT_READ | PROT_WRITE, MAP_SHARED, + bo->dev->fd, args.out.addr_ptr); + if (ptr == MAP_FAILED) { + pthread_mutex_unlock(&bo->cpu_access_mutex); + return -errno; + } + + bo->cpu_ptr = ptr; + bo->cpu_map_count = 1; + pthread_mutex_unlock(&bo->cpu_access_mutex); + + *cpu = ptr; + return 0; +} + +drm_public int amdgpu_bo_cpu_unmap(amdgpu_bo_handle bo) +{ + int r; + + pthread_mutex_lock(&bo->cpu_access_mutex); + assert(bo->cpu_map_count >= 0); + + if (bo->cpu_map_count == 0) { + /* not mapped */ + pthread_mutex_unlock(&bo->cpu_access_mutex); + return -EINVAL; + } + + bo->cpu_map_count--; + if (bo->cpu_map_count > 0) { + /* mapped multiple times */ + pthread_mutex_unlock(&bo->cpu_access_mutex); + return 0; + } + + r = drm_munmap(bo->cpu_ptr, bo->alloc_size) == 0 ? 0 : -errno; + bo->cpu_ptr = NULL; + pthread_mutex_unlock(&bo->cpu_access_mutex); + return r; +} + +drm_public int amdgpu_query_buffer_size_alignment(amdgpu_device_handle dev, + struct amdgpu_buffer_size_alignments *info) +{ + info->size_local = dev->dev_info.pte_fragment_size; + info->size_remote = dev->dev_info.gart_page_size; + return 0; +} + +drm_public int amdgpu_bo_wait_for_idle(amdgpu_bo_handle bo, + uint64_t timeout_ns, + bool *busy) +{ + union drm_amdgpu_gem_wait_idle args; + int r; + + memset(&args, 0, sizeof(args)); + args.in.handle = bo->handle; + args.in.timeout = amdgpu_cs_calculate_timeout(timeout_ns); + + r = drmCommandWriteRead(bo->dev->fd, DRM_AMDGPU_GEM_WAIT_IDLE, + &args, sizeof(args)); + + if (r == 0) { + *busy = args.out.status; + return 0; + } else { + fprintf(stderr, "amdgpu: GEM_WAIT_IDLE failed with %i\n", r); + return r; + } +} + +drm_public int amdgpu_find_bo_by_cpu_mapping(amdgpu_device_handle dev, + void *cpu, + uint64_t size, + amdgpu_bo_handle *buf_handle, + uint64_t *offset_in_bo) +{ + struct amdgpu_bo *bo = NULL; + uint32_t i; + int r = 0; + + if (cpu == NULL || size == 0) + return -EINVAL; + + /* + * Workaround for a buggy application which tries to import previously + * exposed CPU pointers. If we find a real world use case we should + * improve that by asking the kernel for the right handle. + */ + pthread_mutex_lock(&dev->bo_table_mutex); + for (i = 0; i < dev->bo_handles.max_key; i++) { + bo = handle_table_lookup(&dev->bo_handles, i); + if (!bo || !bo->cpu_ptr || size > bo->alloc_size) + continue; + if (cpu >= bo->cpu_ptr && + cpu < (void*)((uintptr_t)bo->cpu_ptr + (size_t)bo->alloc_size)) + break; + } + + if (i < dev->bo_handles.max_key) { + atomic_inc(&bo->refcount); + *buf_handle = bo; + *offset_in_bo = (uintptr_t)cpu - (uintptr_t)bo->cpu_ptr; + } else { + *buf_handle = NULL; + *offset_in_bo = 0; + r = -ENXIO; + } + pthread_mutex_unlock(&dev->bo_table_mutex); + + return r; +} + +drm_public int amdgpu_create_bo_from_user_mem(amdgpu_device_handle dev, + void *cpu, + uint64_t size, + amdgpu_bo_handle *buf_handle) +{ + int r; + struct drm_amdgpu_gem_userptr args; + + args.addr = (uintptr_t)cpu; + args.flags = AMDGPU_GEM_USERPTR_ANONONLY | AMDGPU_GEM_USERPTR_REGISTER | + AMDGPU_GEM_USERPTR_VALIDATE; + args.size = size; + r = drmCommandWriteRead(dev->fd, DRM_AMDGPU_GEM_USERPTR, + &args, sizeof(args)); + if (r) + goto out; + + pthread_mutex_lock(&dev->bo_table_mutex); + r = amdgpu_bo_create(dev, size, args.handle, buf_handle); + pthread_mutex_unlock(&dev->bo_table_mutex); + if (r) { + drmCloseBufferHandle(dev->fd, args.handle); + } + +out: + return r; +} + +drm_public int amdgpu_bo_list_create_raw(amdgpu_device_handle dev, + uint32_t number_of_buffers, + struct drm_amdgpu_bo_list_entry *buffers, + uint32_t *result) +{ + union drm_amdgpu_bo_list args; + int r; + + memset(&args, 0, sizeof(args)); + args.in.operation = AMDGPU_BO_LIST_OP_CREATE; + args.in.bo_number = number_of_buffers; + args.in.bo_info_size = sizeof(struct drm_amdgpu_bo_list_entry); + args.in.bo_info_ptr = (uint64_t)(uintptr_t)buffers; + + r = drmCommandWriteRead(dev->fd, DRM_AMDGPU_BO_LIST, + &args, sizeof(args)); + if (!r) + *result = args.out.list_handle; + return r; +} + +drm_public int amdgpu_bo_list_destroy_raw(amdgpu_device_handle dev, + uint32_t bo_list) +{ + union drm_amdgpu_bo_list args; + + memset(&args, 0, sizeof(args)); + args.in.operation = AMDGPU_BO_LIST_OP_DESTROY; + args.in.list_handle = bo_list; + + return drmCommandWriteRead(dev->fd, DRM_AMDGPU_BO_LIST, + &args, sizeof(args)); +} + +drm_public int amdgpu_bo_list_create(amdgpu_device_handle dev, + uint32_t number_of_resources, + amdgpu_bo_handle *resources, + uint8_t *resource_prios, + amdgpu_bo_list_handle *result) +{ + struct drm_amdgpu_bo_list_entry *list; + union drm_amdgpu_bo_list args; + unsigned i; + int r; + + if (!number_of_resources) + return -EINVAL; + + /* overflow check for multiplication */ + if (number_of_resources > UINT32_MAX / sizeof(struct drm_amdgpu_bo_list_entry)) + return -EINVAL; + + list = malloc(number_of_resources * sizeof(struct drm_amdgpu_bo_list_entry)); + if (!list) + return -ENOMEM; + + *result = malloc(sizeof(struct amdgpu_bo_list)); + if (!*result) { + free(list); + return -ENOMEM; + } + + memset(&args, 0, sizeof(args)); + args.in.operation = AMDGPU_BO_LIST_OP_CREATE; + args.in.bo_number = number_of_resources; + args.in.bo_info_size = sizeof(struct drm_amdgpu_bo_list_entry); + args.in.bo_info_ptr = (uint64_t)(uintptr_t)list; + + for (i = 0; i < number_of_resources; i++) { + list[i].bo_handle = resources[i]->handle; + if (resource_prios) + list[i].bo_priority = resource_prios[i]; + else + list[i].bo_priority = 0; + } + + r = drmCommandWriteRead(dev->fd, DRM_AMDGPU_BO_LIST, + &args, sizeof(args)); + free(list); + if (r) { + free(*result); + return r; + } + + (*result)->dev = dev; + (*result)->handle = args.out.list_handle; + return 0; +} + +drm_public int amdgpu_bo_list_destroy(amdgpu_bo_list_handle list) +{ + union drm_amdgpu_bo_list args; + int r; + + memset(&args, 0, sizeof(args)); + args.in.operation = AMDGPU_BO_LIST_OP_DESTROY; + args.in.list_handle = list->handle; + + r = drmCommandWriteRead(list->dev->fd, DRM_AMDGPU_BO_LIST, + &args, sizeof(args)); + + if (!r) + free(list); + + return r; +} + +drm_public int amdgpu_bo_list_update(amdgpu_bo_list_handle handle, + uint32_t number_of_resources, + amdgpu_bo_handle *resources, + uint8_t *resource_prios) +{ + struct drm_amdgpu_bo_list_entry *list; + union drm_amdgpu_bo_list args; + unsigned i; + int r; + + if (!number_of_resources) + return -EINVAL; + + /* overflow check for multiplication */ + if (number_of_resources > UINT32_MAX / sizeof(struct drm_amdgpu_bo_list_entry)) + return -EINVAL; + + list = malloc(number_of_resources * sizeof(struct drm_amdgpu_bo_list_entry)); + if (!list) + return -ENOMEM; + + args.in.operation = AMDGPU_BO_LIST_OP_UPDATE; + args.in.list_handle = handle->handle; + args.in.bo_number = number_of_resources; + args.in.bo_info_size = sizeof(struct drm_amdgpu_bo_list_entry); + args.in.bo_info_ptr = (uintptr_t)list; + + for (i = 0; i < number_of_resources; i++) { + list[i].bo_handle = resources[i]->handle; + if (resource_prios) + list[i].bo_priority = resource_prios[i]; + else + list[i].bo_priority = 0; + } + + r = drmCommandWriteRead(handle->dev->fd, DRM_AMDGPU_BO_LIST, + &args, sizeof(args)); + free(list); + return r; +} + +drm_public int amdgpu_bo_va_op(amdgpu_bo_handle bo, + uint64_t offset, + uint64_t size, + uint64_t addr, + uint64_t flags, + uint32_t ops) +{ + amdgpu_device_handle dev = bo->dev; + + size = ALIGN(size, getpagesize()); + + return amdgpu_bo_va_op_raw(dev, bo, offset, size, addr, + AMDGPU_VM_PAGE_READABLE | + AMDGPU_VM_PAGE_WRITEABLE | + AMDGPU_VM_PAGE_EXECUTABLE, ops); +} + +drm_public int amdgpu_bo_va_op_raw(amdgpu_device_handle dev, + amdgpu_bo_handle bo, + uint64_t offset, + uint64_t size, + uint64_t addr, + uint64_t flags, + uint32_t ops) +{ + struct drm_amdgpu_gem_va va; + int r; + + if (ops != AMDGPU_VA_OP_MAP && ops != AMDGPU_VA_OP_UNMAP && + ops != AMDGPU_VA_OP_REPLACE && ops != AMDGPU_VA_OP_CLEAR) + return -EINVAL; + + memset(&va, 0, sizeof(va)); + va.handle = bo ? bo->handle : 0; + va.operation = ops; + va.flags = flags; + va.va_address = addr; + va.offset_in_bo = offset; + va.map_size = size; + + r = drmCommandWriteRead(dev->fd, DRM_AMDGPU_GEM_VA, &va, sizeof(va)); + + return r; +} diff --git a/amdgpu/amdgpu_cs.c b/amdgpu/amdgpu_cs.c new file mode 100644 index 0000000..49fc16c --- /dev/null +++ b/amdgpu/amdgpu_cs.c @@ -0,0 +1,984 @@ +/* + * Copyright 2014 Advanced Micro Devices, Inc. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * the rights to use, copy, modify, merge, publish, distribute, sublicense, + * and/or sell copies of the Software, and to permit persons to whom the + * Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice shall be included in + * all copies or substantial portions of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL + * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR + * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, + * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR + * OTHER DEALINGS IN THE SOFTWARE. + * + */ + +#include +#include +#include +#include +#include +#include +#include +#if HAVE_ALLOCA_H +# include +#endif + +#include "xf86drm.h" +#include "amdgpu_drm.h" +#include "amdgpu_internal.h" + +static int amdgpu_cs_unreference_sem(amdgpu_semaphore_handle sem); +static int amdgpu_cs_reset_sem(amdgpu_semaphore_handle sem); + +/** + * Create command submission context + * + * \param dev - \c [in] Device handle. See #amdgpu_device_initialize() + * \param priority - \c [in] Context creation flags. See AMDGPU_CTX_PRIORITY_* + * \param context - \c [out] GPU Context handle + * + * \return 0 on success otherwise POSIX Error code +*/ +drm_public int amdgpu_cs_ctx_create2(amdgpu_device_handle dev, + uint32_t priority, + amdgpu_context_handle *context) +{ + struct amdgpu_context *gpu_context; + union drm_amdgpu_ctx args; + int i, j, k; + int r; + char *override_priority; + + if (!dev || !context) + return -EINVAL; + + override_priority = getenv("AMD_PRIORITY"); + if (override_priority) { + /* The priority is a signed integer. The variable type is + * wrong. If parsing fails, priority is unchanged. + */ + if (sscanf(override_priority, "%i", &priority) == 1) { + printf("amdgpu: context priority changed to %i\n", + priority); + } + } + + gpu_context = calloc(1, sizeof(struct amdgpu_context)); + if (!gpu_context) + return -ENOMEM; + + gpu_context->dev = dev; + + r = pthread_mutex_init(&gpu_context->sequence_mutex, NULL); + if (r) + goto error; + + /* Create the context */ + memset(&args, 0, sizeof(args)); + args.in.op = AMDGPU_CTX_OP_ALLOC_CTX; + args.in.priority = priority; + + r = drmCommandWriteRead(dev->fd, DRM_AMDGPU_CTX, &args, sizeof(args)); + if (r) + goto error; + + gpu_context->id = args.out.alloc.ctx_id; + for (i = 0; i < AMDGPU_HW_IP_NUM; i++) + for (j = 0; j < AMDGPU_HW_IP_INSTANCE_MAX_COUNT; j++) + for (k = 0; k < AMDGPU_CS_MAX_RINGS; k++) + list_inithead(&gpu_context->sem_list[i][j][k]); + *context = (amdgpu_context_handle)gpu_context; + + return 0; + +error: + pthread_mutex_destroy(&gpu_context->sequence_mutex); + free(gpu_context); + return r; +} + +drm_public int amdgpu_cs_ctx_create(amdgpu_device_handle dev, + amdgpu_context_handle *context) +{ + return amdgpu_cs_ctx_create2(dev, AMDGPU_CTX_PRIORITY_NORMAL, context); +} + +/** + * Release command submission context + * + * \param dev - \c [in] amdgpu device handle + * \param context - \c [in] amdgpu context handle + * + * \return 0 on success otherwise POSIX Error code +*/ +drm_public int amdgpu_cs_ctx_free(amdgpu_context_handle context) +{ + union drm_amdgpu_ctx args; + int i, j, k; + int r; + + if (!context) + return -EINVAL; + + pthread_mutex_destroy(&context->sequence_mutex); + + /* now deal with kernel side */ + memset(&args, 0, sizeof(args)); + args.in.op = AMDGPU_CTX_OP_FREE_CTX; + args.in.ctx_id = context->id; + r = drmCommandWriteRead(context->dev->fd, DRM_AMDGPU_CTX, + &args, sizeof(args)); + for (i = 0; i < AMDGPU_HW_IP_NUM; i++) { + for (j = 0; j < AMDGPU_HW_IP_INSTANCE_MAX_COUNT; j++) { + for (k = 0; k < AMDGPU_CS_MAX_RINGS; k++) { + amdgpu_semaphore_handle sem, tmp; + LIST_FOR_EACH_ENTRY_SAFE(sem, tmp, &context->sem_list[i][j][k], list) { + list_del(&sem->list); + amdgpu_cs_reset_sem(sem); + amdgpu_cs_unreference_sem(sem); + } + } + } + } + free(context); + + return r; +} + +drm_public int amdgpu_cs_ctx_override_priority(amdgpu_device_handle dev, + amdgpu_context_handle context, + int master_fd, + unsigned priority) +{ + union drm_amdgpu_sched args; + int r; + + if (!dev || !context || master_fd < 0) + return -EINVAL; + + memset(&args, 0, sizeof(args)); + + args.in.op = AMDGPU_SCHED_OP_CONTEXT_PRIORITY_OVERRIDE; + args.in.fd = dev->fd; + args.in.priority = priority; + args.in.ctx_id = context->id; + + r = drmCommandWrite(master_fd, DRM_AMDGPU_SCHED, &args, sizeof(args)); + if (r) + return r; + + return 0; +} + +drm_public int amdgpu_cs_ctx_stable_pstate(amdgpu_context_handle context, + uint32_t op, + uint32_t flags, + uint32_t *out_flags) +{ + union drm_amdgpu_ctx args; + int r; + + if (!context) + return -EINVAL; + + memset(&args, 0, sizeof(args)); + args.in.op = op; + args.in.ctx_id = context->id; + args.in.flags = flags; + r = drmCommandWriteRead(context->dev->fd, DRM_AMDGPU_CTX, + &args, sizeof(args)); + if (!r && out_flags) + *out_flags = args.out.pstate.flags; + return r; +} + +drm_public int amdgpu_cs_query_reset_state(amdgpu_context_handle context, + uint32_t *state, uint32_t *hangs) +{ + union drm_amdgpu_ctx args; + int r; + + if (!context) + return -EINVAL; + + memset(&args, 0, sizeof(args)); + args.in.op = AMDGPU_CTX_OP_QUERY_STATE; + args.in.ctx_id = context->id; + r = drmCommandWriteRead(context->dev->fd, DRM_AMDGPU_CTX, + &args, sizeof(args)); + if (!r) { + *state = args.out.state.reset_status; + *hangs = args.out.state.hangs; + } + return r; +} + +drm_public int amdgpu_cs_query_reset_state2(amdgpu_context_handle context, + uint64_t *flags) +{ + union drm_amdgpu_ctx args; + int r; + + if (!context) + return -EINVAL; + + memset(&args, 0, sizeof(args)); + args.in.op = AMDGPU_CTX_OP_QUERY_STATE2; + args.in.ctx_id = context->id; + r = drmCommandWriteRead(context->dev->fd, DRM_AMDGPU_CTX, + &args, sizeof(args)); + if (!r) + *flags = args.out.state.flags; + return r; +} + +/** + * Submit command to kernel DRM + * \param dev - \c [in] Device handle + * \param context - \c [in] GPU Context + * \param ibs_request - \c [in] Pointer to submission requests + * \param fence - \c [out] return fence for this submission + * + * \return 0 on success otherwise POSIX Error code + * \sa amdgpu_cs_submit() +*/ +static int amdgpu_cs_submit_one(amdgpu_context_handle context, + struct amdgpu_cs_request *ibs_request) +{ + struct drm_amdgpu_cs_chunk *chunks; + struct drm_amdgpu_cs_chunk_data *chunk_data; + struct drm_amdgpu_cs_chunk_dep *dependencies = NULL; + struct drm_amdgpu_cs_chunk_dep *sem_dependencies = NULL; + amdgpu_device_handle dev = context->dev; + struct list_head *sem_list; + amdgpu_semaphore_handle sem, tmp; + uint32_t i, size, num_chunks, bo_list_handle = 0, sem_count = 0; + uint64_t seq_no; + bool user_fence; + int r = 0; + + if (ibs_request->ip_type >= AMDGPU_HW_IP_NUM) + return -EINVAL; + if (ibs_request->ring >= AMDGPU_CS_MAX_RINGS) + return -EINVAL; + if (ibs_request->number_of_ibs == 0) { + ibs_request->seq_no = AMDGPU_NULL_SUBMIT_SEQ; + return 0; + } + user_fence = (ibs_request->fence_info.handle != NULL); + + size = ibs_request->number_of_ibs + (user_fence ? 2 : 1) + 1; + + chunks = alloca(sizeof(struct drm_amdgpu_cs_chunk) * size); + + size = ibs_request->number_of_ibs + (user_fence ? 1 : 0); + + chunk_data = alloca(sizeof(struct drm_amdgpu_cs_chunk_data) * size); + + if (ibs_request->resources) + bo_list_handle = ibs_request->resources->handle; + num_chunks = ibs_request->number_of_ibs; + /* IB chunks */ + for (i = 0; i < ibs_request->number_of_ibs; i++) { + struct amdgpu_cs_ib_info *ib; + chunks[i].chunk_id = AMDGPU_CHUNK_ID_IB; + chunks[i].length_dw = sizeof(struct drm_amdgpu_cs_chunk_ib) / 4; + chunks[i].chunk_data = (uint64_t)(uintptr_t)&chunk_data[i]; + + ib = &ibs_request->ibs[i]; + + chunk_data[i].ib_data._pad = 0; + chunk_data[i].ib_data.va_start = ib->ib_mc_address; + chunk_data[i].ib_data.ib_bytes = ib->size * 4; + chunk_data[i].ib_data.ip_type = ibs_request->ip_type; + chunk_data[i].ib_data.ip_instance = ibs_request->ip_instance; + chunk_data[i].ib_data.ring = ibs_request->ring; + chunk_data[i].ib_data.flags = ib->flags; + } + + pthread_mutex_lock(&context->sequence_mutex); + + if (user_fence) { + i = num_chunks++; + + /* fence chunk */ + chunks[i].chunk_id = AMDGPU_CHUNK_ID_FENCE; + chunks[i].length_dw = sizeof(struct drm_amdgpu_cs_chunk_fence) / 4; + chunks[i].chunk_data = (uint64_t)(uintptr_t)&chunk_data[i]; + + /* fence bo handle */ + chunk_data[i].fence_data.handle = ibs_request->fence_info.handle->handle; + /* offset */ + chunk_data[i].fence_data.offset = + ibs_request->fence_info.offset * sizeof(uint64_t); + } + + if (ibs_request->number_of_dependencies) { + dependencies = alloca(sizeof(struct drm_amdgpu_cs_chunk_dep) * + ibs_request->number_of_dependencies); + if (!dependencies) { + r = -ENOMEM; + goto error_unlock; + } + + for (i = 0; i < ibs_request->number_of_dependencies; ++i) { + struct amdgpu_cs_fence *info = &ibs_request->dependencies[i]; + struct drm_amdgpu_cs_chunk_dep *dep = &dependencies[i]; + dep->ip_type = info->ip_type; + dep->ip_instance = info->ip_instance; + dep->ring = info->ring; + dep->ctx_id = info->context->id; + dep->handle = info->fence; + } + + i = num_chunks++; + + /* dependencies chunk */ + chunks[i].chunk_id = AMDGPU_CHUNK_ID_DEPENDENCIES; + chunks[i].length_dw = sizeof(struct drm_amdgpu_cs_chunk_dep) / 4 + * ibs_request->number_of_dependencies; + chunks[i].chunk_data = (uint64_t)(uintptr_t)dependencies; + } + + sem_list = &context->sem_list[ibs_request->ip_type][ibs_request->ip_instance][ibs_request->ring]; + LIST_FOR_EACH_ENTRY(sem, sem_list, list) + sem_count++; + if (sem_count) { + sem_dependencies = alloca(sizeof(struct drm_amdgpu_cs_chunk_dep) * sem_count); + if (!sem_dependencies) { + r = -ENOMEM; + goto error_unlock; + } + sem_count = 0; + LIST_FOR_EACH_ENTRY_SAFE(sem, tmp, sem_list, list) { + struct amdgpu_cs_fence *info = &sem->signal_fence; + struct drm_amdgpu_cs_chunk_dep *dep = &sem_dependencies[sem_count++]; + dep->ip_type = info->ip_type; + dep->ip_instance = info->ip_instance; + dep->ring = info->ring; + dep->ctx_id = info->context->id; + dep->handle = info->fence; + + list_del(&sem->list); + amdgpu_cs_reset_sem(sem); + amdgpu_cs_unreference_sem(sem); + } + i = num_chunks++; + + /* dependencies chunk */ + chunks[i].chunk_id = AMDGPU_CHUNK_ID_DEPENDENCIES; + chunks[i].length_dw = sizeof(struct drm_amdgpu_cs_chunk_dep) / 4 * sem_count; + chunks[i].chunk_data = (uint64_t)(uintptr_t)sem_dependencies; + } + + r = amdgpu_cs_submit_raw2(dev, context, bo_list_handle, num_chunks, + chunks, &seq_no); + if (r) + goto error_unlock; + + ibs_request->seq_no = seq_no; + context->last_seq[ibs_request->ip_type][ibs_request->ip_instance][ibs_request->ring] = ibs_request->seq_no; +error_unlock: + pthread_mutex_unlock(&context->sequence_mutex); + return r; +} + +drm_public int amdgpu_cs_submit(amdgpu_context_handle context, + uint64_t flags, + struct amdgpu_cs_request *ibs_request, + uint32_t number_of_requests) +{ + uint32_t i; + int r; + + if (!context || !ibs_request) + return -EINVAL; + + r = 0; + for (i = 0; i < number_of_requests; i++) { + r = amdgpu_cs_submit_one(context, ibs_request); + if (r) + break; + ibs_request++; + } + + return r; +} + +/** + * Calculate absolute timeout. + * + * \param timeout - \c [in] timeout in nanoseconds. + * + * \return absolute timeout in nanoseconds +*/ +drm_private uint64_t amdgpu_cs_calculate_timeout(uint64_t timeout) +{ + int r; + + if (timeout != AMDGPU_TIMEOUT_INFINITE) { + struct timespec current; + uint64_t current_ns; + r = clock_gettime(CLOCK_MONOTONIC, ¤t); + if (r) { + fprintf(stderr, "clock_gettime() returned error (%d)!", errno); + return AMDGPU_TIMEOUT_INFINITE; + } + + current_ns = ((uint64_t)current.tv_sec) * 1000000000ull; + current_ns += current.tv_nsec; + timeout += current_ns; + if (timeout < current_ns) + timeout = AMDGPU_TIMEOUT_INFINITE; + } + return timeout; +} + +static int amdgpu_ioctl_wait_cs(amdgpu_context_handle context, + unsigned ip, + unsigned ip_instance, + uint32_t ring, + uint64_t handle, + uint64_t timeout_ns, + uint64_t flags, + bool *busy) +{ + amdgpu_device_handle dev = context->dev; + union drm_amdgpu_wait_cs args; + int r; + + memset(&args, 0, sizeof(args)); + args.in.handle = handle; + args.in.ip_type = ip; + args.in.ip_instance = ip_instance; + args.in.ring = ring; + args.in.ctx_id = context->id; + + if (flags & AMDGPU_QUERY_FENCE_TIMEOUT_IS_ABSOLUTE) + args.in.timeout = timeout_ns; + else + args.in.timeout = amdgpu_cs_calculate_timeout(timeout_ns); + + r = drmIoctl(dev->fd, DRM_IOCTL_AMDGPU_WAIT_CS, &args); + if (r) + return -errno; + + *busy = args.out.status; + return 0; +} + +drm_public int amdgpu_cs_query_fence_status(struct amdgpu_cs_fence *fence, + uint64_t timeout_ns, + uint64_t flags, + uint32_t *expired) +{ + bool busy = true; + int r; + + if (!fence || !expired || !fence->context) + return -EINVAL; + if (fence->ip_type >= AMDGPU_HW_IP_NUM) + return -EINVAL; + if (fence->ring >= AMDGPU_CS_MAX_RINGS) + return -EINVAL; + if (fence->fence == AMDGPU_NULL_SUBMIT_SEQ) { + *expired = true; + return 0; + } + + *expired = false; + + r = amdgpu_ioctl_wait_cs(fence->context, fence->ip_type, + fence->ip_instance, fence->ring, + fence->fence, timeout_ns, flags, &busy); + + if (!r && !busy) + *expired = true; + + return r; +} + +static int amdgpu_ioctl_wait_fences(struct amdgpu_cs_fence *fences, + uint32_t fence_count, + bool wait_all, + uint64_t timeout_ns, + uint32_t *status, + uint32_t *first) +{ + struct drm_amdgpu_fence *drm_fences; + amdgpu_device_handle dev = fences[0].context->dev; + union drm_amdgpu_wait_fences args; + int r; + uint32_t i; + + drm_fences = alloca(sizeof(struct drm_amdgpu_fence) * fence_count); + for (i = 0; i < fence_count; i++) { + drm_fences[i].ctx_id = fences[i].context->id; + drm_fences[i].ip_type = fences[i].ip_type; + drm_fences[i].ip_instance = fences[i].ip_instance; + drm_fences[i].ring = fences[i].ring; + drm_fences[i].seq_no = fences[i].fence; + } + + memset(&args, 0, sizeof(args)); + args.in.fences = (uint64_t)(uintptr_t)drm_fences; + args.in.fence_count = fence_count; + args.in.wait_all = wait_all; + args.in.timeout_ns = amdgpu_cs_calculate_timeout(timeout_ns); + + r = drmIoctl(dev->fd, DRM_IOCTL_AMDGPU_WAIT_FENCES, &args); + if (r) + return -errno; + + *status = args.out.status; + + if (first) + *first = args.out.first_signaled; + + return 0; +} + +drm_public int amdgpu_cs_wait_fences(struct amdgpu_cs_fence *fences, + uint32_t fence_count, + bool wait_all, + uint64_t timeout_ns, + uint32_t *status, + uint32_t *first) +{ + uint32_t i; + + /* Sanity check */ + if (!fences || !status || !fence_count) + return -EINVAL; + + for (i = 0; i < fence_count; i++) { + if (NULL == fences[i].context) + return -EINVAL; + if (fences[i].ip_type >= AMDGPU_HW_IP_NUM) + return -EINVAL; + if (fences[i].ring >= AMDGPU_CS_MAX_RINGS) + return -EINVAL; + } + + *status = 0; + + return amdgpu_ioctl_wait_fences(fences, fence_count, wait_all, + timeout_ns, status, first); +} + +drm_public int amdgpu_cs_create_semaphore(amdgpu_semaphore_handle *sem) +{ + struct amdgpu_semaphore *gpu_semaphore; + + if (!sem) + return -EINVAL; + + gpu_semaphore = calloc(1, sizeof(struct amdgpu_semaphore)); + if (!gpu_semaphore) + return -ENOMEM; + + atomic_set(&gpu_semaphore->refcount, 1); + *sem = gpu_semaphore; + + return 0; +} + +drm_public int amdgpu_cs_signal_semaphore(amdgpu_context_handle ctx, + uint32_t ip_type, + uint32_t ip_instance, + uint32_t ring, + amdgpu_semaphore_handle sem) +{ + if (!ctx || !sem) + return -EINVAL; + if (ip_type >= AMDGPU_HW_IP_NUM) + return -EINVAL; + if (ring >= AMDGPU_CS_MAX_RINGS) + return -EINVAL; + /* sem has been signaled */ + if (sem->signal_fence.context) + return -EINVAL; + pthread_mutex_lock(&ctx->sequence_mutex); + sem->signal_fence.context = ctx; + sem->signal_fence.ip_type = ip_type; + sem->signal_fence.ip_instance = ip_instance; + sem->signal_fence.ring = ring; + sem->signal_fence.fence = ctx->last_seq[ip_type][ip_instance][ring]; + update_references(NULL, &sem->refcount); + pthread_mutex_unlock(&ctx->sequence_mutex); + return 0; +} + +drm_public int amdgpu_cs_wait_semaphore(amdgpu_context_handle ctx, + uint32_t ip_type, + uint32_t ip_instance, + uint32_t ring, + amdgpu_semaphore_handle sem) +{ + if (!ctx || !sem) + return -EINVAL; + if (ip_type >= AMDGPU_HW_IP_NUM) + return -EINVAL; + if (ring >= AMDGPU_CS_MAX_RINGS) + return -EINVAL; + /* must signal first */ + if (!sem->signal_fence.context) + return -EINVAL; + + pthread_mutex_lock(&ctx->sequence_mutex); + list_add(&sem->list, &ctx->sem_list[ip_type][ip_instance][ring]); + pthread_mutex_unlock(&ctx->sequence_mutex); + return 0; +} + +static int amdgpu_cs_reset_sem(amdgpu_semaphore_handle sem) +{ + if (!sem || !sem->signal_fence.context) + return -EINVAL; + + sem->signal_fence.context = NULL; + sem->signal_fence.ip_type = 0; + sem->signal_fence.ip_instance = 0; + sem->signal_fence.ring = 0; + sem->signal_fence.fence = 0; + + return 0; +} + +static int amdgpu_cs_unreference_sem(amdgpu_semaphore_handle sem) +{ + if (!sem) + return -EINVAL; + + if (update_references(&sem->refcount, NULL)) + free(sem); + return 0; +} + +drm_public int amdgpu_cs_destroy_semaphore(amdgpu_semaphore_handle sem) +{ + return amdgpu_cs_unreference_sem(sem); +} + +drm_public int amdgpu_cs_create_syncobj2(amdgpu_device_handle dev, + uint32_t flags, + uint32_t *handle) +{ + if (NULL == dev) + return -EINVAL; + + return drmSyncobjCreate(dev->fd, flags, handle); +} + +drm_public int amdgpu_cs_create_syncobj(amdgpu_device_handle dev, + uint32_t *handle) +{ + if (NULL == dev) + return -EINVAL; + + return drmSyncobjCreate(dev->fd, 0, handle); +} + +drm_public int amdgpu_cs_destroy_syncobj(amdgpu_device_handle dev, + uint32_t handle) +{ + if (NULL == dev) + return -EINVAL; + + return drmSyncobjDestroy(dev->fd, handle); +} + +drm_public int amdgpu_cs_syncobj_reset(amdgpu_device_handle dev, + const uint32_t *syncobjs, + uint32_t syncobj_count) +{ + if (NULL == dev) + return -EINVAL; + + return drmSyncobjReset(dev->fd, syncobjs, syncobj_count); +} + +drm_public int amdgpu_cs_syncobj_signal(amdgpu_device_handle dev, + const uint32_t *syncobjs, + uint32_t syncobj_count) +{ + if (NULL == dev) + return -EINVAL; + + return drmSyncobjSignal(dev->fd, syncobjs, syncobj_count); +} + +drm_public int amdgpu_cs_syncobj_timeline_signal(amdgpu_device_handle dev, + const uint32_t *syncobjs, + uint64_t *points, + uint32_t syncobj_count) +{ + if (NULL == dev) + return -EINVAL; + + return drmSyncobjTimelineSignal(dev->fd, syncobjs, + points, syncobj_count); +} + +drm_public int amdgpu_cs_syncobj_wait(amdgpu_device_handle dev, + uint32_t *handles, unsigned num_handles, + int64_t timeout_nsec, unsigned flags, + uint32_t *first_signaled) +{ + if (NULL == dev) + return -EINVAL; + + return drmSyncobjWait(dev->fd, handles, num_handles, timeout_nsec, + flags, first_signaled); +} + +drm_public int amdgpu_cs_syncobj_timeline_wait(amdgpu_device_handle dev, + uint32_t *handles, uint64_t *points, + unsigned num_handles, + int64_t timeout_nsec, unsigned flags, + uint32_t *first_signaled) +{ + if (NULL == dev) + return -EINVAL; + + return drmSyncobjTimelineWait(dev->fd, handles, points, num_handles, + timeout_nsec, flags, first_signaled); +} + +drm_public int amdgpu_cs_syncobj_query(amdgpu_device_handle dev, + uint32_t *handles, uint64_t *points, + unsigned num_handles) +{ + if (NULL == dev) + return -EINVAL; + + return drmSyncobjQuery(dev->fd, handles, points, num_handles); +} + +drm_public int amdgpu_cs_syncobj_query2(amdgpu_device_handle dev, + uint32_t *handles, uint64_t *points, + unsigned num_handles, uint32_t flags) +{ + if (!dev) + return -EINVAL; + + return drmSyncobjQuery2(dev->fd, handles, points, num_handles, flags); +} + +drm_public int amdgpu_cs_export_syncobj(amdgpu_device_handle dev, + uint32_t handle, + int *shared_fd) +{ + if (NULL == dev) + return -EINVAL; + + return drmSyncobjHandleToFD(dev->fd, handle, shared_fd); +} + +drm_public int amdgpu_cs_import_syncobj(amdgpu_device_handle dev, + int shared_fd, + uint32_t *handle) +{ + if (NULL == dev) + return -EINVAL; + + return drmSyncobjFDToHandle(dev->fd, shared_fd, handle); +} + +drm_public int amdgpu_cs_syncobj_export_sync_file(amdgpu_device_handle dev, + uint32_t syncobj, + int *sync_file_fd) +{ + if (NULL == dev) + return -EINVAL; + + return drmSyncobjExportSyncFile(dev->fd, syncobj, sync_file_fd); +} + +drm_public int amdgpu_cs_syncobj_import_sync_file(amdgpu_device_handle dev, + uint32_t syncobj, + int sync_file_fd) +{ + if (NULL == dev) + return -EINVAL; + + return drmSyncobjImportSyncFile(dev->fd, syncobj, sync_file_fd); +} + +drm_public int amdgpu_cs_syncobj_export_sync_file2(amdgpu_device_handle dev, + uint32_t syncobj, + uint64_t point, + uint32_t flags, + int *sync_file_fd) +{ + uint32_t binary_handle; + int ret; + + if (NULL == dev) + return -EINVAL; + + if (!point) + return drmSyncobjExportSyncFile(dev->fd, syncobj, sync_file_fd); + + ret = drmSyncobjCreate(dev->fd, 0, &binary_handle); + if (ret) + return ret; + + ret = drmSyncobjTransfer(dev->fd, binary_handle, 0, + syncobj, point, flags); + if (ret) + goto out; + ret = drmSyncobjExportSyncFile(dev->fd, binary_handle, sync_file_fd); +out: + drmSyncobjDestroy(dev->fd, binary_handle); + return ret; +} + +drm_public int amdgpu_cs_syncobj_import_sync_file2(amdgpu_device_handle dev, + uint32_t syncobj, + uint64_t point, + int sync_file_fd) +{ + uint32_t binary_handle; + int ret; + + if (NULL == dev) + return -EINVAL; + + if (!point) + return drmSyncobjImportSyncFile(dev->fd, syncobj, sync_file_fd); + + ret = drmSyncobjCreate(dev->fd, 0, &binary_handle); + if (ret) + return ret; + ret = drmSyncobjImportSyncFile(dev->fd, binary_handle, sync_file_fd); + if (ret) + goto out; + ret = drmSyncobjTransfer(dev->fd, syncobj, point, + binary_handle, 0, 0); +out: + drmSyncobjDestroy(dev->fd, binary_handle); + return ret; +} + +drm_public int amdgpu_cs_syncobj_transfer(amdgpu_device_handle dev, + uint32_t dst_handle, + uint64_t dst_point, + uint32_t src_handle, + uint64_t src_point, + uint32_t flags) +{ + if (NULL == dev) + return -EINVAL; + + return drmSyncobjTransfer(dev->fd, + dst_handle, dst_point, + src_handle, src_point, + flags); +} + +drm_public int amdgpu_cs_submit_raw(amdgpu_device_handle dev, + amdgpu_context_handle context, + amdgpu_bo_list_handle bo_list_handle, + int num_chunks, + struct drm_amdgpu_cs_chunk *chunks, + uint64_t *seq_no) +{ + union drm_amdgpu_cs cs; + uint64_t *chunk_array; + int i, r; + if (num_chunks == 0) + return -EINVAL; + + memset(&cs, 0, sizeof(cs)); + chunk_array = alloca(sizeof(uint64_t) * num_chunks); + for (i = 0; i < num_chunks; i++) + chunk_array[i] = (uint64_t)(uintptr_t)&chunks[i]; + cs.in.chunks = (uint64_t)(uintptr_t)chunk_array; + cs.in.ctx_id = context->id; + cs.in.bo_list_handle = bo_list_handle ? bo_list_handle->handle : 0; + cs.in.num_chunks = num_chunks; + r = drmCommandWriteRead(dev->fd, DRM_AMDGPU_CS, + &cs, sizeof(cs)); + if (r) + return r; + + if (seq_no) + *seq_no = cs.out.handle; + return 0; +} + +drm_public int amdgpu_cs_submit_raw2(amdgpu_device_handle dev, + amdgpu_context_handle context, + uint32_t bo_list_handle, + int num_chunks, + struct drm_amdgpu_cs_chunk *chunks, + uint64_t *seq_no) +{ + union drm_amdgpu_cs cs; + uint64_t *chunk_array; + int i, r; + + memset(&cs, 0, sizeof(cs)); + chunk_array = alloca(sizeof(uint64_t) * num_chunks); + for (i = 0; i < num_chunks; i++) + chunk_array[i] = (uint64_t)(uintptr_t)&chunks[i]; + cs.in.chunks = (uint64_t)(uintptr_t)chunk_array; + cs.in.ctx_id = context->id; + cs.in.bo_list_handle = bo_list_handle; + cs.in.num_chunks = num_chunks; + r = drmCommandWriteRead(dev->fd, DRM_AMDGPU_CS, + &cs, sizeof(cs)); + if (!r && seq_no) + *seq_no = cs.out.handle; + return r; +} + +drm_public void amdgpu_cs_chunk_fence_info_to_data(struct amdgpu_cs_fence_info *fence_info, + struct drm_amdgpu_cs_chunk_data *data) +{ + data->fence_data.handle = fence_info->handle->handle; + data->fence_data.offset = fence_info->offset * sizeof(uint64_t); +} + +drm_public void amdgpu_cs_chunk_fence_to_dep(struct amdgpu_cs_fence *fence, + struct drm_amdgpu_cs_chunk_dep *dep) +{ + dep->ip_type = fence->ip_type; + dep->ip_instance = fence->ip_instance; + dep->ring = fence->ring; + dep->ctx_id = fence->context->id; + dep->handle = fence->fence; +} + +drm_public int amdgpu_cs_fence_to_handle(amdgpu_device_handle dev, + struct amdgpu_cs_fence *fence, + uint32_t what, + uint32_t *out_handle) +{ + union drm_amdgpu_fence_to_handle fth; + int r; + + memset(&fth, 0, sizeof(fth)); + fth.in.fence.ctx_id = fence->context->id; + fth.in.fence.ip_type = fence->ip_type; + fth.in.fence.ip_instance = fence->ip_instance; + fth.in.fence.ring = fence->ring; + fth.in.fence.seq_no = fence->fence; + fth.in.what = what; + + r = drmCommandWriteRead(dev->fd, DRM_AMDGPU_FENCE_TO_HANDLE, + &fth, sizeof(fth)); + if (r == 0) + *out_handle = fth.out.handle; + return r; +} diff --git a/amdgpu/amdgpu_device.c b/amdgpu/amdgpu_device.c new file mode 100644 index 0000000..aeb5e3c --- /dev/null +++ b/amdgpu/amdgpu_device.c @@ -0,0 +1,316 @@ +/* + * Copyright 2014 Advanced Micro Devices, Inc. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * the rights to use, copy, modify, merge, publish, distribute, sublicense, + * and/or sell copies of the Software, and to permit persons to whom the + * Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice shall be included in + * all copies or substantial portions of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL + * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR + * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, + * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR + * OTHER DEALINGS IN THE SOFTWARE. + * + */ + +/** + * \file amdgpu_device.c + * + * Implementation of functions for AMD GPU device + * + */ + +#include +#include +#include +#include +#include +#include +#include + +#include "xf86drm.h" +#include "amdgpu_drm.h" +#include "amdgpu_internal.h" +#include "util_math.h" + +#define PTR_TO_UINT(x) ((unsigned)((intptr_t)(x))) + +static pthread_mutex_t dev_mutex = PTHREAD_MUTEX_INITIALIZER; +static amdgpu_device_handle dev_list; + +static int fd_compare(int fd1, int fd2) +{ + char *name1 = drmGetPrimaryDeviceNameFromFd(fd1); + char *name2 = drmGetPrimaryDeviceNameFromFd(fd2); + int result; + + if (name1 == NULL || name2 == NULL) { + free(name1); + free(name2); + return 0; + } + + result = strcmp(name1, name2); + free(name1); + free(name2); + + return result; +} + +/** +* Get the authenticated form fd, +* +* \param fd - \c [in] File descriptor for AMD GPU device +* \param auth - \c [out] Pointer to output the fd is authenticated or not +* A render node fd, output auth = 0 +* A legacy fd, get the authenticated for compatibility root +* +* \return 0 on success\n +* >0 - AMD specific error code\n +* <0 - Negative POSIX Error code +*/ +static int amdgpu_get_auth(int fd, int *auth) +{ + int r = 0; + drm_client_t client = {}; + + if (drmGetNodeTypeFromFd(fd) == DRM_NODE_RENDER) + *auth = 0; + else { + client.idx = 0; + r = drmIoctl(fd, DRM_IOCTL_GET_CLIENT, &client); + if (!r) + *auth = client.auth; + } + return r; +} + +static void amdgpu_device_free_internal(amdgpu_device_handle dev) +{ + amdgpu_device_handle *node = &dev_list; + + pthread_mutex_lock(&dev_mutex); + while (*node != dev && (*node)->next) + node = &(*node)->next; + *node = (*node)->next; + pthread_mutex_unlock(&dev_mutex); + + close(dev->fd); + if ((dev->flink_fd >= 0) && (dev->fd != dev->flink_fd)) + close(dev->flink_fd); + + amdgpu_vamgr_deinit(&dev->vamgr_32); + amdgpu_vamgr_deinit(&dev->vamgr); + amdgpu_vamgr_deinit(&dev->vamgr_high_32); + amdgpu_vamgr_deinit(&dev->vamgr_high); + handle_table_fini(&dev->bo_handles); + handle_table_fini(&dev->bo_flink_names); + pthread_mutex_destroy(&dev->bo_table_mutex); + free(dev->marketing_name); + free(dev); +} + +/** + * Assignment between two amdgpu_device pointers with reference counting. + * + * Usage: + * struct amdgpu_device *dst = ... , *src = ...; + * + * dst = src; + * // No reference counting. Only use this when you need to move + * // a reference from one pointer to another. + * + * amdgpu_device_reference(&dst, src); + * // Reference counters are updated. dst is decremented and src is + * // incremented. dst is freed if its reference counter is 0. + */ +static void amdgpu_device_reference(struct amdgpu_device **dst, + struct amdgpu_device *src) +{ + if (update_references(&(*dst)->refcount, &src->refcount)) + amdgpu_device_free_internal(*dst); + *dst = src; +} + +drm_public int amdgpu_device_initialize(int fd, + uint32_t *major_version, + uint32_t *minor_version, + amdgpu_device_handle *device_handle) +{ + struct amdgpu_device *dev; + drmVersionPtr version; + int r; + int flag_auth = 0; + int flag_authexist=0; + uint32_t accel_working = 0; + uint64_t start, max; + + *device_handle = NULL; + + pthread_mutex_lock(&dev_mutex); + r = amdgpu_get_auth(fd, &flag_auth); + if (r) { + fprintf(stderr, "%s: amdgpu_get_auth (1) failed (%i)\n", + __func__, r); + pthread_mutex_unlock(&dev_mutex); + return r; + } + + for (dev = dev_list; dev; dev = dev->next) + if (fd_compare(dev->fd, fd) == 0) + break; + + if (dev) { + r = amdgpu_get_auth(dev->fd, &flag_authexist); + if (r) { + fprintf(stderr, "%s: amdgpu_get_auth (2) failed (%i)\n", + __func__, r); + pthread_mutex_unlock(&dev_mutex); + return r; + } + if ((flag_auth) && (!flag_authexist)) { + dev->flink_fd = fcntl(fd, F_DUPFD_CLOEXEC, 0); + } + *major_version = dev->major_version; + *minor_version = dev->minor_version; + amdgpu_device_reference(device_handle, dev); + pthread_mutex_unlock(&dev_mutex); + return 0; + } + + dev = calloc(1, sizeof(struct amdgpu_device)); + if (!dev) { + fprintf(stderr, "%s: calloc failed\n", __func__); + pthread_mutex_unlock(&dev_mutex); + return -ENOMEM; + } + + dev->fd = -1; + dev->flink_fd = -1; + + atomic_set(&dev->refcount, 1); + + version = drmGetVersion(fd); + if (version->version_major != 3) { + fprintf(stderr, "%s: DRM version is %d.%d.%d but this driver is " + "only compatible with 3.x.x.\n", + __func__, + version->version_major, + version->version_minor, + version->version_patchlevel); + drmFreeVersion(version); + r = -EBADF; + goto cleanup; + } + + dev->fd = fcntl(fd, F_DUPFD_CLOEXEC, 0); + dev->flink_fd = dev->fd; + dev->major_version = version->version_major; + dev->minor_version = version->version_minor; + drmFreeVersion(version); + + pthread_mutex_init(&dev->bo_table_mutex, NULL); + + /* Check if acceleration is working. */ + r = amdgpu_query_info(dev, AMDGPU_INFO_ACCEL_WORKING, 4, &accel_working); + if (r) { + fprintf(stderr, "%s: amdgpu_query_info(ACCEL_WORKING) failed (%i)\n", + __func__, r); + goto cleanup; + } + if (!accel_working) { + fprintf(stderr, "%s: AMDGPU_INFO_ACCEL_WORKING = 0\n", __func__); + r = -EBADF; + goto cleanup; + } + + r = amdgpu_query_gpu_info_init(dev); + if (r) { + fprintf(stderr, "%s: amdgpu_query_gpu_info_init failed\n", __func__); + goto cleanup; + } + + start = dev->dev_info.virtual_address_offset; + max = MIN2(dev->dev_info.virtual_address_max, 0x100000000ULL); + amdgpu_vamgr_init(&dev->vamgr_32, start, max, + dev->dev_info.virtual_address_alignment); + + start = max; + max = MAX2(dev->dev_info.virtual_address_max, 0x100000000ULL); + amdgpu_vamgr_init(&dev->vamgr, start, max, + dev->dev_info.virtual_address_alignment); + + start = dev->dev_info.high_va_offset; + max = MIN2(dev->dev_info.high_va_max, (start & ~0xffffffffULL) + + 0x100000000ULL); + amdgpu_vamgr_init(&dev->vamgr_high_32, start, max, + dev->dev_info.virtual_address_alignment); + + start = max; + max = MAX2(dev->dev_info.high_va_max, (start & ~0xffffffffULL) + + 0x100000000ULL); + amdgpu_vamgr_init(&dev->vamgr_high, start, max, + dev->dev_info.virtual_address_alignment); + + amdgpu_parse_asic_ids(dev); + + *major_version = dev->major_version; + *minor_version = dev->minor_version; + *device_handle = dev; + dev->next = dev_list; + dev_list = dev; + pthread_mutex_unlock(&dev_mutex); + + return 0; + +cleanup: + if (dev->fd >= 0) + close(dev->fd); + free(dev); + pthread_mutex_unlock(&dev_mutex); + return r; +} + +drm_public int amdgpu_device_deinitialize(amdgpu_device_handle dev) +{ + amdgpu_device_reference(&dev, NULL); + return 0; +} + +drm_public int amdgpu_device_get_fd(amdgpu_device_handle device_handle) +{ + return device_handle->fd; +} + +drm_public const char *amdgpu_get_marketing_name(amdgpu_device_handle dev) +{ + if (dev->marketing_name) + return dev->marketing_name; + else + return "AMD Radeon Graphics"; +} + +drm_public int amdgpu_query_sw_info(amdgpu_device_handle dev, + enum amdgpu_sw_info info, + void *value) +{ + uint32_t *val32 = (uint32_t*)value; + + switch (info) { + case amdgpu_sw_info_address32_hi: + if (dev->vamgr_high_32.va_max) + *val32 = (dev->vamgr_high_32.va_max - 1) >> 32; + else + *val32 = (dev->vamgr_32.va_max - 1) >> 32; + return 0; + } + return -EINVAL; +} diff --git a/amdgpu/amdgpu_gpu_info.c b/amdgpu/amdgpu_gpu_info.c new file mode 100644 index 0000000..1a5143a --- /dev/null +++ b/amdgpu/amdgpu_gpu_info.c @@ -0,0 +1,362 @@ +/* + * Copyright © 2014 Advanced Micro Devices, Inc. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * the rights to use, copy, modify, merge, publish, distribute, sublicense, + * and/or sell copies of the Software, and to permit persons to whom the + * Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice shall be included in + * all copies or substantial portions of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL + * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR + * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, + * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR + * OTHER DEALINGS IN THE SOFTWARE. + * + */ + +#include +#include + +#include "amdgpu.h" +#include "amdgpu_drm.h" +#include "amdgpu_internal.h" +#include "xf86drm.h" + +drm_public int amdgpu_query_info(amdgpu_device_handle dev, unsigned info_id, + unsigned size, void *value) +{ + struct drm_amdgpu_info request; + + memset(&request, 0, sizeof(request)); + request.return_pointer = (uintptr_t)value; + request.return_size = size; + request.query = info_id; + + return drmCommandWrite(dev->fd, DRM_AMDGPU_INFO, &request, + sizeof(struct drm_amdgpu_info)); +} + +drm_public int amdgpu_query_crtc_from_id(amdgpu_device_handle dev, unsigned id, + int32_t *result) +{ + struct drm_amdgpu_info request; + + memset(&request, 0, sizeof(request)); + request.return_pointer = (uintptr_t)result; + request.return_size = sizeof(*result); + request.query = AMDGPU_INFO_CRTC_FROM_ID; + request.mode_crtc.id = id; + + return drmCommandWrite(dev->fd, DRM_AMDGPU_INFO, &request, + sizeof(struct drm_amdgpu_info)); +} + +drm_public int amdgpu_read_mm_registers(amdgpu_device_handle dev, + unsigned dword_offset, unsigned count, uint32_t instance, + uint32_t flags, uint32_t *values) +{ + struct drm_amdgpu_info request; + + memset(&request, 0, sizeof(request)); + request.return_pointer = (uintptr_t)values; + request.return_size = count * sizeof(uint32_t); + request.query = AMDGPU_INFO_READ_MMR_REG; + request.read_mmr_reg.dword_offset = dword_offset; + request.read_mmr_reg.count = count; + request.read_mmr_reg.instance = instance; + request.read_mmr_reg.flags = flags; + + return drmCommandWrite(dev->fd, DRM_AMDGPU_INFO, &request, + sizeof(struct drm_amdgpu_info)); +} + +drm_public int amdgpu_query_hw_ip_count(amdgpu_device_handle dev, + unsigned type, + uint32_t *count) +{ + struct drm_amdgpu_info request; + + memset(&request, 0, sizeof(request)); + request.return_pointer = (uintptr_t)count; + request.return_size = sizeof(*count); + request.query = AMDGPU_INFO_HW_IP_COUNT; + request.query_hw_ip.type = type; + + return drmCommandWrite(dev->fd, DRM_AMDGPU_INFO, &request, + sizeof(struct drm_amdgpu_info)); +} + +drm_public int amdgpu_query_hw_ip_info(amdgpu_device_handle dev, unsigned type, + unsigned ip_instance, + struct drm_amdgpu_info_hw_ip *info) +{ + struct drm_amdgpu_info request; + + memset(&request, 0, sizeof(request)); + request.return_pointer = (uintptr_t)info; + request.return_size = sizeof(*info); + request.query = AMDGPU_INFO_HW_IP_INFO; + request.query_hw_ip.type = type; + request.query_hw_ip.ip_instance = ip_instance; + + return drmCommandWrite(dev->fd, DRM_AMDGPU_INFO, &request, + sizeof(struct drm_amdgpu_info)); +} + +drm_public int amdgpu_query_firmware_version(amdgpu_device_handle dev, + unsigned fw_type, unsigned ip_instance, unsigned index, + uint32_t *version, uint32_t *feature) +{ + struct drm_amdgpu_info request; + struct drm_amdgpu_info_firmware firmware = {}; + int r; + + memset(&request, 0, sizeof(request)); + request.return_pointer = (uintptr_t)&firmware; + request.return_size = sizeof(firmware); + request.query = AMDGPU_INFO_FW_VERSION; + request.query_fw.fw_type = fw_type; + request.query_fw.ip_instance = ip_instance; + request.query_fw.index = index; + + r = drmCommandWrite(dev->fd, DRM_AMDGPU_INFO, &request, + sizeof(struct drm_amdgpu_info)); + if (r) + return r; + + *version = firmware.ver; + *feature = firmware.feature; + return 0; +} + +drm_private int amdgpu_query_gpu_info_init(amdgpu_device_handle dev) +{ + int r, i; + + r = amdgpu_query_info(dev, AMDGPU_INFO_DEV_INFO, sizeof(dev->dev_info), + &dev->dev_info); + if (r) + return r; + + dev->info.asic_id = dev->dev_info.device_id; + dev->info.chip_rev = dev->dev_info.chip_rev; + dev->info.chip_external_rev = dev->dev_info.external_rev; + dev->info.family_id = dev->dev_info.family; + dev->info.max_engine_clk = dev->dev_info.max_engine_clock; + dev->info.max_memory_clk = dev->dev_info.max_memory_clock; + dev->info.gpu_counter_freq = dev->dev_info.gpu_counter_freq; + dev->info.enabled_rb_pipes_mask = dev->dev_info.enabled_rb_pipes_mask; + dev->info.rb_pipes = dev->dev_info.num_rb_pipes; + dev->info.ids_flags = dev->dev_info.ids_flags; + dev->info.num_hw_gfx_contexts = dev->dev_info.num_hw_gfx_contexts; + dev->info.num_shader_engines = dev->dev_info.num_shader_engines; + dev->info.num_shader_arrays_per_engine = + dev->dev_info.num_shader_arrays_per_engine; + dev->info.vram_type = dev->dev_info.vram_type; + dev->info.vram_bit_width = dev->dev_info.vram_bit_width; + dev->info.ce_ram_size = dev->dev_info.ce_ram_size; + dev->info.vce_harvest_config = dev->dev_info.vce_harvest_config; + dev->info.pci_rev_id = dev->dev_info.pci_rev; + + if (dev->info.family_id < AMDGPU_FAMILY_AI) { + for (i = 0; i < (int)dev->info.num_shader_engines; i++) { + unsigned instance = (i << AMDGPU_INFO_MMR_SE_INDEX_SHIFT) | + (AMDGPU_INFO_MMR_SH_INDEX_MASK << + AMDGPU_INFO_MMR_SH_INDEX_SHIFT); + + r = amdgpu_read_mm_registers(dev, 0x263d, 1, instance, 0, + &dev->info.backend_disable[i]); + if (r) + return r; + /* extract bitfield CC_RB_BACKEND_DISABLE.BACKEND_DISABLE */ + dev->info.backend_disable[i] = + (dev->info.backend_disable[i] >> 16) & 0xff; + + r = amdgpu_read_mm_registers(dev, 0xa0d4, 1, instance, 0, + &dev->info.pa_sc_raster_cfg[i]); + if (r) + return r; + + if (dev->info.family_id >= AMDGPU_FAMILY_CI) { + r = amdgpu_read_mm_registers(dev, 0xa0d5, 1, instance, 0, + &dev->info.pa_sc_raster_cfg1[i]); + if (r) + return r; + } + } + } + + r = amdgpu_read_mm_registers(dev, 0x263e, 1, 0xffffffff, 0, + &dev->info.gb_addr_cfg); + if (r) + return r; + + if (dev->info.family_id < AMDGPU_FAMILY_AI) { + r = amdgpu_read_mm_registers(dev, 0x2644, 32, 0xffffffff, 0, + dev->info.gb_tile_mode); + if (r) + return r; + + if (dev->info.family_id >= AMDGPU_FAMILY_CI) { + r = amdgpu_read_mm_registers(dev, 0x2664, 16, 0xffffffff, 0, + dev->info.gb_macro_tile_mode); + if (r) + return r; + } + + r = amdgpu_read_mm_registers(dev, 0x9d8, 1, 0xffffffff, 0, + &dev->info.mc_arb_ramcfg); + if (r) + return r; + } + + dev->info.cu_active_number = dev->dev_info.cu_active_number; + dev->info.cu_ao_mask = dev->dev_info.cu_ao_mask; + memcpy(&dev->info.cu_bitmap[0][0], &dev->dev_info.cu_bitmap[0][0], sizeof(dev->info.cu_bitmap)); + + /* TODO: info->max_quad_shader_pipes is not set */ + /* TODO: info->avail_quad_shader_pipes is not set */ + /* TODO: info->cache_entries_per_quad_pipe is not set */ + return 0; +} + +drm_public int amdgpu_query_gpu_info(amdgpu_device_handle dev, + struct amdgpu_gpu_info *info) +{ + if (!dev || !info) + return -EINVAL; + + /* Get ASIC info*/ + *info = dev->info; + + return 0; +} + +drm_public int amdgpu_query_heap_info(amdgpu_device_handle dev, + uint32_t heap, + uint32_t flags, + struct amdgpu_heap_info *info) +{ + struct drm_amdgpu_info_vram_gtt vram_gtt_info = {}; + int r; + + r = amdgpu_query_info(dev, AMDGPU_INFO_VRAM_GTT, + sizeof(vram_gtt_info), &vram_gtt_info); + if (r) + return r; + + /* Get heap information */ + switch (heap) { + case AMDGPU_GEM_DOMAIN_VRAM: + /* query visible only vram heap */ + if (flags & AMDGPU_GEM_CREATE_CPU_ACCESS_REQUIRED) + info->heap_size = vram_gtt_info.vram_cpu_accessible_size; + else /* query total vram heap */ + info->heap_size = vram_gtt_info.vram_size; + + info->max_allocation = vram_gtt_info.vram_cpu_accessible_size; + + if (flags & AMDGPU_GEM_CREATE_CPU_ACCESS_REQUIRED) + r = amdgpu_query_info(dev, AMDGPU_INFO_VIS_VRAM_USAGE, + sizeof(info->heap_usage), + &info->heap_usage); + else + r = amdgpu_query_info(dev, AMDGPU_INFO_VRAM_USAGE, + sizeof(info->heap_usage), + &info->heap_usage); + if (r) + return r; + break; + case AMDGPU_GEM_DOMAIN_GTT: + info->heap_size = vram_gtt_info.gtt_size; + info->max_allocation = vram_gtt_info.vram_cpu_accessible_size; + + r = amdgpu_query_info(dev, AMDGPU_INFO_GTT_USAGE, + sizeof(info->heap_usage), + &info->heap_usage); + if (r) + return r; + break; + default: + return -EINVAL; + } + + return 0; +} + +drm_public int amdgpu_query_gds_info(amdgpu_device_handle dev, + struct amdgpu_gds_resource_info *gds_info) +{ + struct drm_amdgpu_info_gds gds_config = {}; + int r; + + if (!gds_info) + return -EINVAL; + + r = amdgpu_query_info(dev, AMDGPU_INFO_GDS_CONFIG, + sizeof(gds_config), &gds_config); + if (r) + return r; + + gds_info->gds_gfx_partition_size = gds_config.gds_gfx_partition_size; + gds_info->compute_partition_size = gds_config.compute_partition_size; + gds_info->gds_total_size = gds_config.gds_total_size; + gds_info->gws_per_gfx_partition = gds_config.gws_per_gfx_partition; + gds_info->gws_per_compute_partition = gds_config.gws_per_compute_partition; + gds_info->oa_per_gfx_partition = gds_config.oa_per_gfx_partition; + gds_info->oa_per_compute_partition = gds_config.oa_per_compute_partition; + + return 0; +} + +drm_public int amdgpu_query_sensor_info(amdgpu_device_handle dev, unsigned sensor_type, + unsigned size, void *value) +{ + struct drm_amdgpu_info request; + + memset(&request, 0, sizeof(request)); + request.return_pointer = (uintptr_t)value; + request.return_size = size; + request.query = AMDGPU_INFO_SENSOR; + request.sensor_info.type = sensor_type; + + return drmCommandWrite(dev->fd, DRM_AMDGPU_INFO, &request, + sizeof(struct drm_amdgpu_info)); +} + +drm_public int amdgpu_query_video_caps_info(amdgpu_device_handle dev, unsigned cap_type, + unsigned size, void *value) +{ + struct drm_amdgpu_info request; + + memset(&request, 0, sizeof(request)); + request.return_pointer = (uintptr_t)value; + request.return_size = size; + request.query = AMDGPU_INFO_VIDEO_CAPS; + request.sensor_info.type = cap_type; + + return drmCommandWrite(dev->fd, DRM_AMDGPU_INFO, &request, + sizeof(struct drm_amdgpu_info)); +} + +drm_public int amdgpu_query_gpuvm_fault_info(amdgpu_device_handle dev, + unsigned size, void *value) +{ + struct drm_amdgpu_info request; + + memset(&request, 0, sizeof(request)); + request.return_pointer = (uintptr_t)value; + request.return_size = size; + request.query = AMDGPU_INFO_GPUVM_FAULT; + + return drmCommandWrite(dev->fd, DRM_AMDGPU_INFO, &request, + sizeof(struct drm_amdgpu_info)); +} diff --git a/amdgpu/amdgpu_internal.h b/amdgpu/amdgpu_internal.h new file mode 100644 index 0000000..2834c9c --- /dev/null +++ b/amdgpu/amdgpu_internal.h @@ -0,0 +1,175 @@ +/* + * Copyright © 2014 Advanced Micro Devices, Inc. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * the rights to use, copy, modify, merge, publish, distribute, sublicense, + * and/or sell copies of the Software, and to permit persons to whom the + * Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice shall be included in + * all copies or substantial portions of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL + * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR + * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, + * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR + * OTHER DEALINGS IN THE SOFTWARE. + * + */ + +#ifndef _AMDGPU_INTERNAL_H_ +#define _AMDGPU_INTERNAL_H_ + +#include +#include + +#include "libdrm_macros.h" +#include "xf86atomic.h" +#include "amdgpu.h" +#include "util_double_list.h" +#include "handle_table.h" + +#define AMDGPU_CS_MAX_RINGS 8 +/* do not use below macro if b is not power of 2 aligned value */ +#define __round_mask(x, y) ((__typeof__(x))((y)-1)) +#define ROUND_UP(x, y) ((((x)-1) | __round_mask(x, y))+1) +#define ROUND_DOWN(x, y) ((x) & ~__round_mask(x, y)) + +#define AMDGPU_INVALID_VA_ADDRESS 0xffffffffffffffff +#define AMDGPU_NULL_SUBMIT_SEQ 0 + +struct amdgpu_bo_va_hole { + struct list_head list; + uint64_t offset; + uint64_t size; +}; + +struct amdgpu_bo_va_mgr { + uint64_t va_max; + struct list_head va_holes; + pthread_mutex_t bo_va_mutex; + uint32_t va_alignment; +}; + +struct amdgpu_va { + uint64_t address; + uint64_t size; + enum amdgpu_gpu_va_range range; + struct amdgpu_bo_va_mgr *vamgr; +}; + +struct amdgpu_device { + atomic_t refcount; + struct amdgpu_device *next; + int fd; + int flink_fd; + unsigned major_version; + unsigned minor_version; + + char *marketing_name; + /** List of buffer handles. Protected by bo_table_mutex. */ + struct handle_table bo_handles; + /** List of buffer GEM flink names. Protected by bo_table_mutex. */ + struct handle_table bo_flink_names; + /** This protects all hash tables. */ + pthread_mutex_t bo_table_mutex; + struct drm_amdgpu_info_device dev_info; + struct amdgpu_gpu_info info; + /** The VA manager for the lower virtual address space */ + struct amdgpu_bo_va_mgr vamgr; + /** The VA manager for the 32bit address space */ + struct amdgpu_bo_va_mgr vamgr_32; + /** The VA manager for the high virtual address space */ + struct amdgpu_bo_va_mgr vamgr_high; + /** The VA manager for the 32bit high address space */ + struct amdgpu_bo_va_mgr vamgr_high_32; +}; + +struct amdgpu_bo { + atomic_t refcount; + struct amdgpu_device *dev; + + uint64_t alloc_size; + + uint32_t handle; + uint32_t flink_name; + + pthread_mutex_t cpu_access_mutex; + void *cpu_ptr; + int64_t cpu_map_count; +}; + +struct amdgpu_bo_list { + struct amdgpu_device *dev; + + uint32_t handle; +}; + +struct amdgpu_context { + struct amdgpu_device *dev; + /** Mutex for accessing fences and to maintain command submissions + in good sequence. */ + pthread_mutex_t sequence_mutex; + /* context id*/ + uint32_t id; + uint64_t last_seq[AMDGPU_HW_IP_NUM][AMDGPU_HW_IP_INSTANCE_MAX_COUNT][AMDGPU_CS_MAX_RINGS]; + struct list_head sem_list[AMDGPU_HW_IP_NUM][AMDGPU_HW_IP_INSTANCE_MAX_COUNT][AMDGPU_CS_MAX_RINGS]; +}; + +/** + * Structure describing sw semaphore based on scheduler + * + */ +struct amdgpu_semaphore { + atomic_t refcount; + struct list_head list; + struct amdgpu_cs_fence signal_fence; +}; + +/** + * Functions. + */ + +drm_private void amdgpu_vamgr_init(struct amdgpu_bo_va_mgr *mgr, uint64_t start, + uint64_t max, uint64_t alignment); + +drm_private void amdgpu_vamgr_deinit(struct amdgpu_bo_va_mgr *mgr); + +drm_private void amdgpu_parse_asic_ids(struct amdgpu_device *dev); + +drm_private int amdgpu_query_gpu_info_init(amdgpu_device_handle dev); + +drm_private uint64_t amdgpu_cs_calculate_timeout(uint64_t timeout); + +/** + * Inline functions. + */ + +/** + * Increment src and decrement dst as if we were updating references + * for an assignment between 2 pointers of some objects. + * + * \return true if dst is 0 + */ +static inline bool update_references(atomic_t *dst, atomic_t *src) +{ + if (dst != src) { + /* bump src first */ + if (src) { + assert(atomic_read(src) > 0); + atomic_inc(src); + } + if (dst) { + assert(atomic_read(dst) > 0); + return atomic_dec_and_test(dst); + } + } + return false; +} + +#endif diff --git a/amdgpu/amdgpu_vamgr.c b/amdgpu/amdgpu_vamgr.c new file mode 100644 index 0000000..2c4c9db --- /dev/null +++ b/amdgpu/amdgpu_vamgr.c @@ -0,0 +1,302 @@ +/* + * Copyright 2014 Advanced Micro Devices, Inc. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * the rights to use, copy, modify, merge, publish, distribute, sublicense, + * and/or sell copies of the Software, and to permit persons to whom the + * Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice shall be included in + * all copies or substantial portions of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL + * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR + * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, + * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR + * OTHER DEALINGS IN THE SOFTWARE. + * + */ + +#include +#include +#include +#include "amdgpu.h" +#include "amdgpu_drm.h" +#include "amdgpu_internal.h" +#include "util_math.h" + +drm_public int amdgpu_va_range_query(amdgpu_device_handle dev, + enum amdgpu_gpu_va_range type, + uint64_t *start, uint64_t *end) +{ + if (type != amdgpu_gpu_va_range_general) + return -EINVAL; + + *start = dev->dev_info.virtual_address_offset; + *end = dev->dev_info.virtual_address_max; + return 0; +} + +drm_private void amdgpu_vamgr_init(struct amdgpu_bo_va_mgr *mgr, uint64_t start, + uint64_t max, uint64_t alignment) +{ + struct amdgpu_bo_va_hole *n; + + mgr->va_max = max; + mgr->va_alignment = alignment; + + list_inithead(&mgr->va_holes); + pthread_mutex_init(&mgr->bo_va_mutex, NULL); + pthread_mutex_lock(&mgr->bo_va_mutex); + n = calloc(1, sizeof(struct amdgpu_bo_va_hole)); + n->size = mgr->va_max - start; + n->offset = start; + list_add(&n->list, &mgr->va_holes); + pthread_mutex_unlock(&mgr->bo_va_mutex); +} + +drm_private void amdgpu_vamgr_deinit(struct amdgpu_bo_va_mgr *mgr) +{ + struct amdgpu_bo_va_hole *hole, *tmp; + LIST_FOR_EACH_ENTRY_SAFE(hole, tmp, &mgr->va_holes, list) { + list_del(&hole->list); + free(hole); + } + pthread_mutex_destroy(&mgr->bo_va_mutex); +} + +static drm_private int +amdgpu_vamgr_subtract_hole(struct amdgpu_bo_va_hole *hole, uint64_t start_va, + uint64_t end_va) +{ + if (start_va > hole->offset && end_va - hole->offset < hole->size) { + struct amdgpu_bo_va_hole *n = calloc(1, sizeof(struct amdgpu_bo_va_hole)); + if (!n) + return -ENOMEM; + + n->size = start_va - hole->offset; + n->offset = hole->offset; + list_add(&n->list, &hole->list); + + hole->size -= (end_va - hole->offset); + hole->offset = end_va; + } else if (start_va > hole->offset) { + hole->size = start_va - hole->offset; + } else if (end_va - hole->offset < hole->size) { + hole->size -= (end_va - hole->offset); + hole->offset = end_va; + } else { + list_del(&hole->list); + free(hole); + } + + return 0; +} + +static drm_private int +amdgpu_vamgr_find_va(struct amdgpu_bo_va_mgr *mgr, uint64_t size, + uint64_t alignment, uint64_t base_required, + bool search_from_top, uint64_t *va_out) +{ + struct amdgpu_bo_va_hole *hole, *n; + uint64_t offset = 0; + int ret; + + + alignment = MAX2(alignment, mgr->va_alignment); + size = ALIGN(size, mgr->va_alignment); + + if (base_required % alignment) + return -EINVAL; + + pthread_mutex_lock(&mgr->bo_va_mutex); + if (!search_from_top) { + LIST_FOR_EACH_ENTRY_SAFE_REV(hole, n, &mgr->va_holes, list) { + if (base_required) { + if (hole->offset > base_required || + (hole->offset + hole->size) < (base_required + size)) + continue; + offset = base_required; + } else { + uint64_t waste = hole->offset % alignment; + waste = waste ? alignment - waste : 0; + offset = hole->offset + waste; + if (offset >= (hole->offset + hole->size) || + size > (hole->offset + hole->size) - offset) { + continue; + } + } + ret = amdgpu_vamgr_subtract_hole(hole, offset, offset + size); + pthread_mutex_unlock(&mgr->bo_va_mutex); + *va_out = offset; + return ret; + } + } else { + LIST_FOR_EACH_ENTRY_SAFE(hole, n, &mgr->va_holes, list) { + if (base_required) { + if (hole->offset > base_required || + (hole->offset + hole->size) < (base_required + size)) + continue; + offset = base_required; + } else { + if (size > hole->size) + continue; + + offset = hole->offset + hole->size - size; + offset -= offset % alignment; + if (offset < hole->offset) { + continue; + } + } + + ret = amdgpu_vamgr_subtract_hole(hole, offset, offset + size); + pthread_mutex_unlock(&mgr->bo_va_mutex); + *va_out = offset; + return ret; + } + } + + pthread_mutex_unlock(&mgr->bo_va_mutex); + return -ENOMEM; +} + +static drm_private void +amdgpu_vamgr_free_va(struct amdgpu_bo_va_mgr *mgr, uint64_t va, uint64_t size) +{ + struct amdgpu_bo_va_hole *hole, *next; + + if (va == AMDGPU_INVALID_VA_ADDRESS) + return; + + size = ALIGN(size, mgr->va_alignment); + + pthread_mutex_lock(&mgr->bo_va_mutex); + hole = container_of(&mgr->va_holes, hole, list); + LIST_FOR_EACH_ENTRY(next, &mgr->va_holes, list) { + if (next->offset < va) + break; + hole = next; + } + + if (&hole->list != &mgr->va_holes) { + /* Grow upper hole if it's adjacent */ + if (hole->offset == (va + size)) { + hole->offset = va; + hole->size += size; + /* Merge lower hole if it's adjacent */ + if (next != hole && + &next->list != &mgr->va_holes && + (next->offset + next->size) == va) { + next->size += hole->size; + list_del(&hole->list); + free(hole); + } + goto out; + } + } + + /* Grow lower hole if it's adjacent */ + if (next != hole && &next->list != &mgr->va_holes && + (next->offset + next->size) == va) { + next->size += size; + goto out; + } + + /* FIXME on allocation failure we just lose virtual address space + * maybe print a warning + */ + next = calloc(1, sizeof(struct amdgpu_bo_va_hole)); + if (next) { + next->size = size; + next->offset = va; + list_add(&next->list, &hole->list); + } + +out: + pthread_mutex_unlock(&mgr->bo_va_mutex); +} + +drm_public int amdgpu_va_range_alloc(amdgpu_device_handle dev, + enum amdgpu_gpu_va_range va_range_type, + uint64_t size, + uint64_t va_base_alignment, + uint64_t va_base_required, + uint64_t *va_base_allocated, + amdgpu_va_handle *va_range_handle, + uint64_t flags) +{ + struct amdgpu_bo_va_mgr *vamgr; + bool search_from_top = !!(flags & AMDGPU_VA_RANGE_REPLAYABLE); + int ret; + + /* Clear the flag when the high VA manager is not initialized */ + if (flags & AMDGPU_VA_RANGE_HIGH && !dev->vamgr_high_32.va_max) + flags &= ~AMDGPU_VA_RANGE_HIGH; + + if (flags & AMDGPU_VA_RANGE_HIGH) { + if (flags & AMDGPU_VA_RANGE_32_BIT) + vamgr = &dev->vamgr_high_32; + else + vamgr = &dev->vamgr_high; + } else { + if (flags & AMDGPU_VA_RANGE_32_BIT) + vamgr = &dev->vamgr_32; + else + vamgr = &dev->vamgr; + } + + va_base_alignment = MAX2(va_base_alignment, vamgr->va_alignment); + size = ALIGN(size, vamgr->va_alignment); + + ret = amdgpu_vamgr_find_va(vamgr, size, + va_base_alignment, va_base_required, + search_from_top, va_base_allocated); + + if (!(flags & AMDGPU_VA_RANGE_32_BIT) && ret) { + /* fallback to 32bit address */ + if (flags & AMDGPU_VA_RANGE_HIGH) + vamgr = &dev->vamgr_high_32; + else + vamgr = &dev->vamgr_32; + ret = amdgpu_vamgr_find_va(vamgr, size, + va_base_alignment, va_base_required, + search_from_top, va_base_allocated); + } + + if (!ret) { + struct amdgpu_va* va; + va = calloc(1, sizeof(struct amdgpu_va)); + if(!va){ + amdgpu_vamgr_free_va(vamgr, *va_base_allocated, size); + return -ENOMEM; + } + va->address = *va_base_allocated; + va->size = size; + va->range = va_range_type; + va->vamgr = vamgr; + *va_range_handle = va; + } + + return ret; +} + +drm_public int amdgpu_va_range_free(amdgpu_va_handle va_range_handle) +{ + if(!va_range_handle || !va_range_handle->address) + return 0; + + amdgpu_vamgr_free_va(va_range_handle->vamgr, + va_range_handle->address, + va_range_handle->size); + free(va_range_handle); + return 0; +} + +drm_public uint64_t amdgpu_va_get_start_addr(amdgpu_va_handle va_handle) +{ + return va_handle->address; +} diff --git a/amdgpu/amdgpu_vm.c b/amdgpu/amdgpu_vm.c new file mode 100644 index 0000000..7e6e28f --- /dev/null +++ b/amdgpu/amdgpu_vm.c @@ -0,0 +1,50 @@ +/* + * Copyright 2017 Advanced Micro Devices, Inc. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * the rights to use, copy, modify, merge, publish, distribute, sublicense, + * and/or sell copies of the Software, and to permit persons to whom the + * Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice shall be included in + * all copies or substantial portions of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL + * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR + * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, + * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR + * OTHER DEALINGS IN THE SOFTWARE. + * + */ + +#include "amdgpu.h" +#include "amdgpu_drm.h" +#include "xf86drm.h" +#include "amdgpu_internal.h" + +drm_public int amdgpu_vm_reserve_vmid(amdgpu_device_handle dev, uint32_t flags) +{ + union drm_amdgpu_vm vm; + + vm.in.op = AMDGPU_VM_OP_RESERVE_VMID; + vm.in.flags = flags; + + return drmCommandWriteRead(dev->fd, DRM_AMDGPU_VM, + &vm, sizeof(vm)); +} + +drm_public int amdgpu_vm_unreserve_vmid(amdgpu_device_handle dev, + uint32_t flags) +{ + union drm_amdgpu_vm vm; + + vm.in.op = AMDGPU_VM_OP_UNRESERVE_VMID; + vm.in.flags = flags; + + return drmCommandWriteRead(dev->fd, DRM_AMDGPU_VM, + &vm, sizeof(vm)); +} diff --git a/amdgpu/handle_table.c b/amdgpu/handle_table.c new file mode 100644 index 0000000..4fdd29d --- /dev/null +++ b/amdgpu/handle_table.c @@ -0,0 +1,72 @@ +/* + * Copyright 2018 Advanced Micro Devices, Inc. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * the rights to use, copy, modify, merge, publish, distribute, sublicense, + * and/or sell copies of the Software, and to permit persons to whom the + * Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice shall be included in + * all copies or substantial portions of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL + * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR + * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, + * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR + * OTHER DEALINGS IN THE SOFTWARE. + * + */ + +#include +#include +#include +#include +#include "handle_table.h" +#include "util_math.h" + +drm_private int handle_table_insert(struct handle_table *table, uint32_t key, + void *value) +{ + if (key >= table->max_key) { + uint32_t alignment = sysconf(_SC_PAGESIZE) / sizeof(void*); + uint32_t max_key = ALIGN(key + 1, alignment); + void **values; + + values = realloc(table->values, max_key * sizeof(void *)); + if (!values) + return -ENOMEM; + + memset(values + table->max_key, 0, (max_key - table->max_key) * + sizeof(void *)); + + table->max_key = max_key; + table->values = values; + } + table->values[key] = value; + return 0; +} + +drm_private void handle_table_remove(struct handle_table *table, uint32_t key) +{ + if (key < table->max_key) + table->values[key] = NULL; +} + +drm_private void *handle_table_lookup(struct handle_table *table, uint32_t key) +{ + if (key < table->max_key) + return table->values[key]; + else + return NULL; +} + +drm_private void handle_table_fini(struct handle_table *table) +{ + free(table->values); + table->max_key = 0; + table->values = NULL; +} diff --git a/amdgpu/handle_table.h b/amdgpu/handle_table.h new file mode 100644 index 0000000..461193f --- /dev/null +++ b/amdgpu/handle_table.h @@ -0,0 +1,41 @@ +/* + * Copyright 2018 Advanced Micro Devices, Inc. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * the rights to use, copy, modify, merge, publish, distribute, sublicense, + * and/or sell copies of the Software, and to permit persons to whom the + * Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice shall be included in + * all copies or substantial portions of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL + * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR + * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, + * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR + * OTHER DEALINGS IN THE SOFTWARE. + * + */ + +#ifndef _HANDLE_TABLE_H_ +#define _HANDLE_TABLE_H_ + +#include +#include "libdrm_macros.h" + +struct handle_table { + uint32_t max_key; + void **values; +}; + +drm_private int handle_table_insert(struct handle_table *table, uint32_t key, + void *value); +drm_private void handle_table_remove(struct handle_table *table, uint32_t key); +drm_private void *handle_table_lookup(struct handle_table *table, uint32_t key); +drm_private void handle_table_fini(struct handle_table *table); + +#endif /* _HANDLE_TABLE_H_ */ diff --git a/amdgpu/libdrm_amdgpu.pc.in b/amdgpu/libdrm_amdgpu.pc.in new file mode 100644 index 0000000..f1c552a --- /dev/null +++ b/amdgpu/libdrm_amdgpu.pc.in @@ -0,0 +1,11 @@ +prefix=@prefix@ +exec_prefix=@exec_prefix@ +libdir=@libdir@ +includedir=@includedir@ + +Name: libdrm_amdgpu +Description: Userspace interface to kernel DRM services for amdgpu +Version: @PACKAGE_VERSION@ +Libs: -L${libdir} -ldrm_amdgpu +Cflags: -I${includedir} -I${includedir}/libdrm +Requires.private: libdrm diff --git a/amdgpu/meson.build b/amdgpu/meson.build new file mode 100644 index 0000000..679de94 --- /dev/null +++ b/amdgpu/meson.build @@ -0,0 +1,70 @@ +# Copyright © 2017-2018 Intel Corporation + +# Permission is hereby granted, free of charge, to any person obtaining a copy +# of this software and associated documentation files (the "Software"), to deal +# in the Software without restriction, including without limitation the rights +# to use, copy, modify, merge, publish, distribute, sublicense, and/or sell +# copies of the Software, and to permit persons to whom the Software is +# furnished to do so, subject to the following conditions: + +# The above copyright notice and this permission notice shall be included in +# all copies or substantial portions of the Software. + +# THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR +# IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, +# FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE +# AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER +# LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, +# OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE +# SOFTWARE. + + +datadir_amdgpu = join_paths(get_option('prefix'), get_option('datadir'), 'libdrm') + +libdrm_amdgpu = library( + 'drm_amdgpu', + [ + files( + 'amdgpu_asic_id.c', 'amdgpu_bo.c', 'amdgpu_cs.c', 'amdgpu_device.c', + 'amdgpu_gpu_info.c', 'amdgpu_vamgr.c', 'amdgpu_vm.c', 'handle_table.c', + ), + config_file, + ], + c_args : [ + libdrm_c_args, + '-DAMDGPU_ASIC_ID_TABLE="@0@"'.format(join_paths(datadir_amdgpu, 'amdgpu.ids')), + ], + include_directories : [inc_root, inc_drm], + link_with : libdrm, + dependencies : [dep_threads, dep_atomic_ops, dep_rt], + version : '1.0.0', + install : true, +) + +install_headers('amdgpu.h', subdir : 'libdrm') + +pkg.generate( + libdrm_amdgpu, + name : 'libdrm_amdgpu', + subdirs : ['.', 'libdrm'], + description : 'Userspace interface to kernel DRM services for amdgpu', +) + +ext_libdrm_amdgpu = declare_dependency( + link_with : [libdrm, libdrm_amdgpu], + include_directories : [inc_drm, include_directories('.')], +) + +if meson.version().version_compare('>= 0.54.0') + meson.override_dependency('libdrm_amdgpu', ext_libdrm_amdgpu) +endif + +test( + 'amdgpu-symbols-check', + symbols_check, + args : [ + '--lib', libdrm_amdgpu, + '--symbols-file', files('amdgpu-symbols.txt'), + '--nm', prog_nm.full_path(), + ], +) -- cgit v1.2.3