1 /* 2 * Copyright 2018 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_RAS_H 25 #define _AMDGPU_RAS_H 26 27 #include <linux/debugfs.h> 28 #include <linux/list.h> 29 #include "amdgpu.h" 30 #include "amdgpu_psp.h" 31 #include "ta_ras_if.h" 32 33 enum amdgpu_ras_block { 34 AMDGPU_RAS_BLOCK__UMC = 0, 35 AMDGPU_RAS_BLOCK__SDMA, 36 AMDGPU_RAS_BLOCK__GFX, 37 AMDGPU_RAS_BLOCK__MMHUB, 38 AMDGPU_RAS_BLOCK__ATHUB, 39 AMDGPU_RAS_BLOCK__PCIE_BIF, 40 AMDGPU_RAS_BLOCK__HDP, 41 AMDGPU_RAS_BLOCK__XGMI_WAFL, 42 AMDGPU_RAS_BLOCK__DF, 43 AMDGPU_RAS_BLOCK__SMN, 44 AMDGPU_RAS_BLOCK__SEM, 45 AMDGPU_RAS_BLOCK__MP0, 46 AMDGPU_RAS_BLOCK__MP1, 47 AMDGPU_RAS_BLOCK__FUSE, 48 49 AMDGPU_RAS_BLOCK__LAST 50 }; 51 52 #define AMDGPU_RAS_BLOCK_COUNT AMDGPU_RAS_BLOCK__LAST 53 #define AMDGPU_RAS_BLOCK_MASK ((1ULL << AMDGPU_RAS_BLOCK_COUNT) - 1) 54 55 enum amdgpu_ras_error_type { 56 AMDGPU_RAS_ERROR__NONE = 0, 57 AMDGPU_RAS_ERROR__PARITY = 1, 58 AMDGPU_RAS_ERROR__SINGLE_CORRECTABLE = 2, 59 AMDGPU_RAS_ERROR__MULTI_UNCORRECTABLE = 4, 60 AMDGPU_RAS_ERROR__POISON = 8, 61 }; 62 63 enum amdgpu_ras_ret { 64 AMDGPU_RAS_SUCCESS = 0, 65 AMDGPU_RAS_FAIL, 66 AMDGPU_RAS_UE, 67 AMDGPU_RAS_CE, 68 AMDGPU_RAS_PT, 69 }; 70 71 struct ras_common_if { 72 enum amdgpu_ras_block block; 73 enum amdgpu_ras_error_type type; 74 uint32_t sub_block_index; 75 /* block name */ 76 char name[32]; 77 }; 78 79 typedef int (*ras_ih_cb)(struct amdgpu_device *adev, 80 struct amdgpu_iv_entry *entry); 81 82 struct amdgpu_ras { 83 /* ras infrastructure */ 84 /* for ras itself. */ 85 uint32_t hw_supported; 86 /* for IP to check its ras ability. */ 87 uint32_t supported; 88 uint32_t features; 89 struct list_head head; 90 /* debugfs */ 91 struct dentry *dir; 92 /* debugfs ctrl */ 93 struct dentry *ent; 94 /* sysfs */ 95 struct device_attribute features_attr; 96 struct bin_attribute badpages_attr; 97 /* block array */ 98 struct ras_manager *objs; 99 100 /* gpu recovery */ 101 struct work_struct recovery_work; 102 atomic_t in_recovery; 103 struct amdgpu_device *adev; 104 /* error handler data */ 105 struct ras_err_handler_data *eh_data; 106 struct mutex recovery_lock; 107 108 uint32_t flags; 109 }; 110 111 /* interfaces for IP */ 112 113 struct ras_fs_if { 114 struct ras_common_if head; 115 char sysfs_name[32]; 116 char debugfs_name[32]; 117 }; 118 119 struct ras_query_if { 120 struct ras_common_if head; 121 unsigned long ue_count; 122 unsigned long ce_count; 123 }; 124 125 struct ras_inject_if { 126 struct ras_common_if head; 127 uint64_t address; 128 uint64_t value; 129 }; 130 131 struct ras_cure_if { 132 struct ras_common_if head; 133 uint64_t address; 134 }; 135 136 struct ras_ih_if { 137 struct ras_common_if head; 138 ras_ih_cb cb; 139 }; 140 141 struct ras_dispatch_if { 142 struct ras_common_if head; 143 struct amdgpu_iv_entry *entry; 144 }; 145 146 struct ras_debug_if { 147 union { 148 struct ras_common_if head; 149 struct ras_inject_if inject; 150 }; 151 int op; 152 }; 153 /* work flow 154 * vbios 155 * 1: ras feature enable (enabled by default) 156 * psp 157 * 2: ras framework init (in ip_init) 158 * IP 159 * 3: IH add 160 * 4: debugfs/sysfs create 161 * 5: query/inject 162 * 6: debugfs/sysfs remove 163 * 7: IH remove 164 * 8: feature disable 165 */ 166 167 #define amdgpu_ras_get_context(adev) ((adev)->psp.ras.ras) 168 #define amdgpu_ras_set_context(adev, ras_con) ((adev)->psp.ras.ras = (ras_con)) 169 170 /* check if ras is supported on block, say, sdma, gfx */ 171 static inline int amdgpu_ras_is_supported(struct amdgpu_device *adev, 172 unsigned int block) 173 { 174 struct amdgpu_ras *ras = amdgpu_ras_get_context(adev); 175 176 if (block >= AMDGPU_RAS_BLOCK_COUNT) 177 return 0; 178 return ras && (ras->supported & (1 << block)); 179 } 180 181 int amdgpu_ras_request_reset_on_boot(struct amdgpu_device *adev, 182 unsigned int block); 183 184 void amdgpu_ras_resume(struct amdgpu_device *adev); 185 void amdgpu_ras_suspend(struct amdgpu_device *adev); 186 187 int amdgpu_ras_query_error_count(struct amdgpu_device *adev, 188 bool is_ce); 189 190 /* error handling functions */ 191 int amdgpu_ras_add_bad_pages(struct amdgpu_device *adev, 192 unsigned long *bps, int pages); 193 194 int amdgpu_ras_reserve_bad_pages(struct amdgpu_device *adev); 195 196 static inline int amdgpu_ras_reset_gpu(struct amdgpu_device *adev, 197 bool is_baco) 198 { 199 struct amdgpu_ras *ras = amdgpu_ras_get_context(adev); 200 201 if (atomic_cmpxchg(&ras->in_recovery, 0, 1) == 0) 202 schedule_work(&ras->recovery_work); 203 return 0; 204 } 205 206 static inline enum ta_ras_block 207 amdgpu_ras_block_to_ta(enum amdgpu_ras_block block) { 208 switch (block) { 209 case AMDGPU_RAS_BLOCK__UMC: 210 return TA_RAS_BLOCK__UMC; 211 case AMDGPU_RAS_BLOCK__SDMA: 212 return TA_RAS_BLOCK__SDMA; 213 case AMDGPU_RAS_BLOCK__GFX: 214 return TA_RAS_BLOCK__GFX; 215 case AMDGPU_RAS_BLOCK__MMHUB: 216 return TA_RAS_BLOCK__MMHUB; 217 case AMDGPU_RAS_BLOCK__ATHUB: 218 return TA_RAS_BLOCK__ATHUB; 219 case AMDGPU_RAS_BLOCK__PCIE_BIF: 220 return TA_RAS_BLOCK__PCIE_BIF; 221 case AMDGPU_RAS_BLOCK__HDP: 222 return TA_RAS_BLOCK__HDP; 223 case AMDGPU_RAS_BLOCK__XGMI_WAFL: 224 return TA_RAS_BLOCK__XGMI_WAFL; 225 case AMDGPU_RAS_BLOCK__DF: 226 return TA_RAS_BLOCK__DF; 227 case AMDGPU_RAS_BLOCK__SMN: 228 return TA_RAS_BLOCK__SMN; 229 case AMDGPU_RAS_BLOCK__SEM: 230 return TA_RAS_BLOCK__SEM; 231 case AMDGPU_RAS_BLOCK__MP0: 232 return TA_RAS_BLOCK__MP0; 233 case AMDGPU_RAS_BLOCK__MP1: 234 return TA_RAS_BLOCK__MP1; 235 case AMDGPU_RAS_BLOCK__FUSE: 236 return TA_RAS_BLOCK__FUSE; 237 default: 238 WARN_ONCE(1, "RAS ERROR: unexpected block id %d\n", block); 239 return TA_RAS_BLOCK__UMC; 240 } 241 } 242 243 static inline enum ta_ras_error_type 244 amdgpu_ras_error_to_ta(enum amdgpu_ras_error_type error) { 245 switch (error) { 246 case AMDGPU_RAS_ERROR__NONE: 247 return TA_RAS_ERROR__NONE; 248 case AMDGPU_RAS_ERROR__PARITY: 249 return TA_RAS_ERROR__PARITY; 250 case AMDGPU_RAS_ERROR__SINGLE_CORRECTABLE: 251 return TA_RAS_ERROR__SINGLE_CORRECTABLE; 252 case AMDGPU_RAS_ERROR__MULTI_UNCORRECTABLE: 253 return TA_RAS_ERROR__MULTI_UNCORRECTABLE; 254 case AMDGPU_RAS_ERROR__POISON: 255 return TA_RAS_ERROR__POISON; 256 default: 257 WARN_ONCE(1, "RAS ERROR: unexpected error type %d\n", error); 258 return TA_RAS_ERROR__NONE; 259 } 260 } 261 262 /* called in ip_init and ip_fini */ 263 int amdgpu_ras_init(struct amdgpu_device *adev); 264 int amdgpu_ras_fini(struct amdgpu_device *adev); 265 int amdgpu_ras_pre_fini(struct amdgpu_device *adev); 266 267 int amdgpu_ras_feature_enable(struct amdgpu_device *adev, 268 struct ras_common_if *head, bool enable); 269 270 int amdgpu_ras_feature_enable_on_boot(struct amdgpu_device *adev, 271 struct ras_common_if *head, bool enable); 272 273 int amdgpu_ras_sysfs_create(struct amdgpu_device *adev, 274 struct ras_fs_if *head); 275 276 int amdgpu_ras_sysfs_remove(struct amdgpu_device *adev, 277 struct ras_common_if *head); 278 279 void amdgpu_ras_debugfs_create(struct amdgpu_device *adev, 280 struct ras_fs_if *head); 281 282 void amdgpu_ras_debugfs_remove(struct amdgpu_device *adev, 283 struct ras_common_if *head); 284 285 int amdgpu_ras_error_query(struct amdgpu_device *adev, 286 struct ras_query_if *info); 287 288 int amdgpu_ras_error_inject(struct amdgpu_device *adev, 289 struct ras_inject_if *info); 290 291 int amdgpu_ras_interrupt_add_handler(struct amdgpu_device *adev, 292 struct ras_ih_if *info); 293 294 int amdgpu_ras_interrupt_remove_handler(struct amdgpu_device *adev, 295 struct ras_ih_if *info); 296 297 int amdgpu_ras_interrupt_dispatch(struct amdgpu_device *adev, 298 struct ras_dispatch_if *info); 299 #endif 300