1 /* 2 * Copyright 2021 Advanced Micro Devices, Inc. 3 * 4 * Permission is hereby granted, free of charge, to any person obtaining a 5 * copy of this software and associated documentation files (the "Software"), 6 * to deal in the Software without restriction, including without limitation 7 * the rights to use, copy, modify, merge, publish, distribute, sublicense, 8 * and/or sell copies of the Software, and to permit persons to whom the 9 * Software is furnished to do so, subject to the following conditions: 10 * 11 * The above copyright notice and this permission notice shall be included in 12 * all copies or substantial portions of the Software. 13 * 14 * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR 15 * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, 16 * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL 17 * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR 18 * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, 19 * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR 20 * OTHER DEALINGS IN THE SOFTWARE. 21 * 22 */ 23 24 #ifndef __AMDGPU_RESET_H__ 25 #define __AMDGPU_RESET_H__ 26 27 #include "amdgpu.h" 28 29 #define AMDGPU_RESET_MAX_HANDLERS 5 30 31 enum AMDGPU_RESET_FLAGS { 32 33 AMDGPU_NEED_FULL_RESET = 0, 34 AMDGPU_SKIP_HW_RESET = 1, 35 AMDGPU_RESET_FOR_DEVICE_REMOVE = 2, 36 }; 37 38 struct amdgpu_reset_context { 39 enum amd_reset_method method; 40 struct amdgpu_device *reset_req_dev; 41 struct amdgpu_job *job; 42 struct amdgpu_hive_info *hive; 43 struct list_head *reset_device_list; 44 unsigned long flags; 45 }; 46 47 struct amdgpu_reset_handler { 48 enum amd_reset_method reset_method; 49 int (*prepare_env)(struct amdgpu_reset_control *reset_ctl, 50 struct amdgpu_reset_context *context); 51 int (*prepare_hwcontext)(struct amdgpu_reset_control *reset_ctl, 52 struct amdgpu_reset_context *context); 53 int (*perform_reset)(struct amdgpu_reset_control *reset_ctl, 54 struct amdgpu_reset_context *context); 55 int (*restore_hwcontext)(struct amdgpu_reset_control *reset_ctl, 56 struct amdgpu_reset_context *context); 57 int (*restore_env)(struct amdgpu_reset_control *reset_ctl, 58 struct amdgpu_reset_context *context); 59 60 int (*do_reset)(struct amdgpu_device *adev); 61 }; 62 63 struct amdgpu_reset_control { 64 void *handle; 65 struct work_struct reset_work; 66 struct mutex reset_lock; 67 struct amdgpu_reset_handler *( 68 *reset_handlers)[AMDGPU_RESET_MAX_HANDLERS]; 69 atomic_t in_reset; 70 enum amd_reset_method active_reset; 71 struct amdgpu_reset_handler *(*get_reset_handler)( 72 struct amdgpu_reset_control *reset_ctl, 73 struct amdgpu_reset_context *context); 74 void (*async_reset)(struct work_struct *work); 75 }; 76 77 78 enum amdgpu_reset_domain_type { 79 SINGLE_DEVICE, 80 XGMI_HIVE 81 }; 82 83 struct amdgpu_reset_domain { 84 struct kref refcount; 85 struct workqueue_struct *wq; 86 enum amdgpu_reset_domain_type type; 87 struct rw_semaphore sem; 88 atomic_t in_gpu_reset; 89 atomic_t reset_res; 90 }; 91 92 #ifdef CONFIG_DEV_COREDUMP 93 94 #define AMDGPU_COREDUMP_VERSION "1" 95 96 struct amdgpu_coredump_info { 97 struct amdgpu_device *adev; 98 struct amdgpu_task_info reset_task_info; 99 struct timespec64 reset_time; 100 bool reset_vram_lost; 101 }; 102 #endif 103 104 int amdgpu_reset_init(struct amdgpu_device *adev); 105 int amdgpu_reset_fini(struct amdgpu_device *adev); 106 107 int amdgpu_reset_prepare_hwcontext(struct amdgpu_device *adev, 108 struct amdgpu_reset_context *reset_context); 109 110 int amdgpu_reset_perform_reset(struct amdgpu_device *adev, 111 struct amdgpu_reset_context *reset_context); 112 113 int amdgpu_reset_prepare_env(struct amdgpu_device *adev, 114 struct amdgpu_reset_context *reset_context); 115 int amdgpu_reset_restore_env(struct amdgpu_device *adev, 116 struct amdgpu_reset_context *reset_context); 117 118 struct amdgpu_reset_domain *amdgpu_reset_create_reset_domain(enum amdgpu_reset_domain_type type, 119 char *wq_name); 120 121 void amdgpu_reset_destroy_reset_domain(struct kref *ref); 122 123 static inline bool amdgpu_reset_get_reset_domain(struct amdgpu_reset_domain *domain) 124 { 125 return kref_get_unless_zero(&domain->refcount) != 0; 126 } 127 128 static inline void amdgpu_reset_put_reset_domain(struct amdgpu_reset_domain *domain) 129 { 130 if (domain) 131 kref_put(&domain->refcount, amdgpu_reset_destroy_reset_domain); 132 } 133 134 static inline bool amdgpu_reset_domain_schedule(struct amdgpu_reset_domain *domain, 135 struct work_struct *work) 136 { 137 return queue_work(domain->wq, work); 138 } 139 140 void amdgpu_device_lock_reset_domain(struct amdgpu_reset_domain *reset_domain); 141 142 void amdgpu_device_unlock_reset_domain(struct amdgpu_reset_domain *reset_domain); 143 144 void amdgpu_coredump(struct amdgpu_device *adev, bool vram_lost, 145 struct amdgpu_reset_context *reset_context); 146 147 #define for_each_handler(i, handler, reset_ctl) \ 148 for (i = 0; (i < AMDGPU_RESET_MAX_HANDLERS) && \ 149 (handler = (*reset_ctl->reset_handlers)[i]); \ 150 ++i) 151 #endif 152