2 * Copyright 2018 Advanced Micro Devices, Inc.
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:
11 * The above copyright notice and this permission notice shall be included in
12 * all copies or substantial portions of the Software.
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.
27 #include <linux/debugfs.h>
28 #include <linux/list.h>
29 #include <linux/kfifo.h>
30 #include <linux/radix-tree.h>
31 #include "ta_ras_if.h"
32 #include "amdgpu_ras_eeprom.h"
33 #include "amdgpu_smuio.h"
34 #include "amdgpu_aca.h"
36 struct amdgpu_iv_entry;
38 #define AMDGPU_RAS_GPU_ERR_MEM_TRAINING(x) AMDGPU_GET_REG_FIELD(x, 0, 0)
39 #define AMDGPU_RAS_GPU_ERR_FW_LOAD(x) AMDGPU_GET_REG_FIELD(x, 1, 1)
40 #define AMDGPU_RAS_GPU_ERR_WAFL_LINK_TRAINING(x) AMDGPU_GET_REG_FIELD(x, 2, 2)
41 #define AMDGPU_RAS_GPU_ERR_XGMI_LINK_TRAINING(x) AMDGPU_GET_REG_FIELD(x, 3, 3)
42 #define AMDGPU_RAS_GPU_ERR_USR_CP_LINK_TRAINING(x) AMDGPU_GET_REG_FIELD(x, 4, 4)
43 #define AMDGPU_RAS_GPU_ERR_USR_DP_LINK_TRAINING(x) AMDGPU_GET_REG_FIELD(x, 5, 5)
44 #define AMDGPU_RAS_GPU_ERR_HBM_MEM_TEST(x) AMDGPU_GET_REG_FIELD(x, 6, 6)
45 #define AMDGPU_RAS_GPU_ERR_HBM_BIST_TEST(x) AMDGPU_GET_REG_FIELD(x, 7, 7)
46 #define AMDGPU_RAS_GPU_ERR_SOCKET_ID(x) AMDGPU_GET_REG_FIELD(x, 10, 8)
47 #define AMDGPU_RAS_GPU_ERR_AID_ID(x) AMDGPU_GET_REG_FIELD(x, 12, 11)
48 #define AMDGPU_RAS_GPU_ERR_HBM_ID(x) AMDGPU_GET_REG_FIELD(x, 14, 13)
49 #define AMDGPU_RAS_GPU_ERR_DATA_ABORT(x) AMDGPU_GET_REG_FIELD(x, 29, 29)
50 #define AMDGPU_RAS_GPU_ERR_UNKNOWN(x) AMDGPU_GET_REG_FIELD(x, 30, 30)
52 #define AMDGPU_RAS_BOOT_STATUS_POLLING_LIMIT 100
53 #define AMDGPU_RAS_BOOT_STEADY_STATUS 0xBA
54 #define AMDGPU_RAS_BOOT_STATUS_MASK 0xFF
56 #define AMDGPU_RAS_FLAG_INIT_BY_VBIOS (0x1 << 0)
57 /* position of instance value in sub_block_index of
58 * ta_ras_trigger_error_input, the sub block uses lower 12 bits
60 #define AMDGPU_RAS_INST_MASK 0xfffff000
61 #define AMDGPU_RAS_INST_SHIFT 0xc
63 #define AMDGPU_RAS_FEATURES_SOCKETID_SHIFT 29
64 #define AMDGPU_RAS_FEATURES_SOCKETID_MASK 0xe0000000
66 /* Reserve 8 physical dram row for possible retirement.
67 * In worst cases, it will lose 8 * 2MB memory in vram domain */
68 #define AMDGPU_RAS_RESERVED_VRAM_SIZE (16ULL << 20)
69 /* The high three bits indicates socketid */
70 #define AMDGPU_RAS_GET_FEATURES(val) ((val) & ~AMDGPU_RAS_FEATURES_SOCKETID_MASK)
72 #define RAS_EVENT_INVALID_ID (BIT_ULL(63))
73 #define RAS_EVENT_ID_IS_VALID(x) (!((x) & BIT_ULL(63)))
75 #define RAS_EVENT_LOG(adev, id, fmt, ...) \
76 amdgpu_ras_event_log_print((adev), (id), (fmt), ##__VA_ARGS__)
78 #define amdgpu_ras_mark_ras_event(adev, type) \
79 (amdgpu_ras_mark_ras_event_caller((adev), (type), __builtin_return_address(0)))
81 enum amdgpu_ras_block {
82 AMDGPU_RAS_BLOCK__UMC = 0,
83 AMDGPU_RAS_BLOCK__SDMA,
84 AMDGPU_RAS_BLOCK__GFX,
85 AMDGPU_RAS_BLOCK__MMHUB,
86 AMDGPU_RAS_BLOCK__ATHUB,
87 AMDGPU_RAS_BLOCK__PCIE_BIF,
88 AMDGPU_RAS_BLOCK__HDP,
89 AMDGPU_RAS_BLOCK__XGMI_WAFL,
91 AMDGPU_RAS_BLOCK__SMN,
92 AMDGPU_RAS_BLOCK__SEM,
93 AMDGPU_RAS_BLOCK__MP0,
94 AMDGPU_RAS_BLOCK__MP1,
95 AMDGPU_RAS_BLOCK__FUSE,
96 AMDGPU_RAS_BLOCK__MCA,
97 AMDGPU_RAS_BLOCK__VCN,
98 AMDGPU_RAS_BLOCK__JPEG,
100 AMDGPU_RAS_BLOCK__MPIO,
102 AMDGPU_RAS_BLOCK__LAST
105 enum amdgpu_ras_mca_block {
106 AMDGPU_RAS_MCA_BLOCK__MP0 = 0,
107 AMDGPU_RAS_MCA_BLOCK__MP1,
108 AMDGPU_RAS_MCA_BLOCK__MPIO,
109 AMDGPU_RAS_MCA_BLOCK__IOHC,
111 AMDGPU_RAS_MCA_BLOCK__LAST
114 #define AMDGPU_RAS_BLOCK_COUNT AMDGPU_RAS_BLOCK__LAST
115 #define AMDGPU_RAS_MCA_BLOCK_COUNT AMDGPU_RAS_MCA_BLOCK__LAST
116 #define AMDGPU_RAS_BLOCK_MASK ((1ULL << AMDGPU_RAS_BLOCK_COUNT) - 1)
118 enum amdgpu_ras_gfx_subblock {
120 AMDGPU_RAS_BLOCK__GFX_CPC_INDEX_START = 0,
121 AMDGPU_RAS_BLOCK__GFX_CPC_SCRATCH =
122 AMDGPU_RAS_BLOCK__GFX_CPC_INDEX_START,
123 AMDGPU_RAS_BLOCK__GFX_CPC_UCODE,
124 AMDGPU_RAS_BLOCK__GFX_DC_STATE_ME1,
125 AMDGPU_RAS_BLOCK__GFX_DC_CSINVOC_ME1,
126 AMDGPU_RAS_BLOCK__GFX_DC_RESTORE_ME1,
127 AMDGPU_RAS_BLOCK__GFX_DC_STATE_ME2,
128 AMDGPU_RAS_BLOCK__GFX_DC_CSINVOC_ME2,
129 AMDGPU_RAS_BLOCK__GFX_DC_RESTORE_ME2,
130 AMDGPU_RAS_BLOCK__GFX_CPC_INDEX_END =
131 AMDGPU_RAS_BLOCK__GFX_DC_RESTORE_ME2,
133 AMDGPU_RAS_BLOCK__GFX_CPF_INDEX_START,
134 AMDGPU_RAS_BLOCK__GFX_CPF_ROQ_ME2 =
135 AMDGPU_RAS_BLOCK__GFX_CPF_INDEX_START,
136 AMDGPU_RAS_BLOCK__GFX_CPF_ROQ_ME1,
137 AMDGPU_RAS_BLOCK__GFX_CPF_TAG,
138 AMDGPU_RAS_BLOCK__GFX_CPF_INDEX_END = AMDGPU_RAS_BLOCK__GFX_CPF_TAG,
140 AMDGPU_RAS_BLOCK__GFX_CPG_INDEX_START,
141 AMDGPU_RAS_BLOCK__GFX_CPG_DMA_ROQ =
142 AMDGPU_RAS_BLOCK__GFX_CPG_INDEX_START,
143 AMDGPU_RAS_BLOCK__GFX_CPG_DMA_TAG,
144 AMDGPU_RAS_BLOCK__GFX_CPG_TAG,
145 AMDGPU_RAS_BLOCK__GFX_CPG_INDEX_END = AMDGPU_RAS_BLOCK__GFX_CPG_TAG,
147 AMDGPU_RAS_BLOCK__GFX_GDS_INDEX_START,
148 AMDGPU_RAS_BLOCK__GFX_GDS_MEM = AMDGPU_RAS_BLOCK__GFX_GDS_INDEX_START,
149 AMDGPU_RAS_BLOCK__GFX_GDS_INPUT_QUEUE,
150 AMDGPU_RAS_BLOCK__GFX_GDS_OA_PHY_CMD_RAM_MEM,
151 AMDGPU_RAS_BLOCK__GFX_GDS_OA_PHY_DATA_RAM_MEM,
152 AMDGPU_RAS_BLOCK__GFX_GDS_OA_PIPE_MEM,
153 AMDGPU_RAS_BLOCK__GFX_GDS_INDEX_END =
154 AMDGPU_RAS_BLOCK__GFX_GDS_OA_PIPE_MEM,
156 AMDGPU_RAS_BLOCK__GFX_SPI_SR_MEM,
158 AMDGPU_RAS_BLOCK__GFX_SQ_INDEX_START,
159 AMDGPU_RAS_BLOCK__GFX_SQ_SGPR = AMDGPU_RAS_BLOCK__GFX_SQ_INDEX_START,
160 AMDGPU_RAS_BLOCK__GFX_SQ_LDS_D,
161 AMDGPU_RAS_BLOCK__GFX_SQ_LDS_I,
162 AMDGPU_RAS_BLOCK__GFX_SQ_VGPR,
163 AMDGPU_RAS_BLOCK__GFX_SQ_INDEX_END = AMDGPU_RAS_BLOCK__GFX_SQ_VGPR,
165 AMDGPU_RAS_BLOCK__GFX_SQC_INDEX_START,
167 AMDGPU_RAS_BLOCK__GFX_SQC_INDEX0_START =
168 AMDGPU_RAS_BLOCK__GFX_SQC_INDEX_START,
169 AMDGPU_RAS_BLOCK__GFX_SQC_INST_UTCL1_LFIFO =
170 AMDGPU_RAS_BLOCK__GFX_SQC_INDEX0_START,
171 AMDGPU_RAS_BLOCK__GFX_SQC_DATA_CU0_WRITE_DATA_BUF,
172 AMDGPU_RAS_BLOCK__GFX_SQC_DATA_CU0_UTCL1_LFIFO,
173 AMDGPU_RAS_BLOCK__GFX_SQC_DATA_CU1_WRITE_DATA_BUF,
174 AMDGPU_RAS_BLOCK__GFX_SQC_DATA_CU1_UTCL1_LFIFO,
175 AMDGPU_RAS_BLOCK__GFX_SQC_DATA_CU2_WRITE_DATA_BUF,
176 AMDGPU_RAS_BLOCK__GFX_SQC_DATA_CU2_UTCL1_LFIFO,
177 AMDGPU_RAS_BLOCK__GFX_SQC_INDEX0_END =
178 AMDGPU_RAS_BLOCK__GFX_SQC_DATA_CU2_UTCL1_LFIFO,
180 AMDGPU_RAS_BLOCK__GFX_SQC_INDEX1_START,
181 AMDGPU_RAS_BLOCK__GFX_SQC_INST_BANKA_TAG_RAM =
182 AMDGPU_RAS_BLOCK__GFX_SQC_INDEX1_START,
183 AMDGPU_RAS_BLOCK__GFX_SQC_INST_BANKA_UTCL1_MISS_FIFO,
184 AMDGPU_RAS_BLOCK__GFX_SQC_INST_BANKA_MISS_FIFO,
185 AMDGPU_RAS_BLOCK__GFX_SQC_INST_BANKA_BANK_RAM,
186 AMDGPU_RAS_BLOCK__GFX_SQC_DATA_BANKA_TAG_RAM,
187 AMDGPU_RAS_BLOCK__GFX_SQC_DATA_BANKA_HIT_FIFO,
188 AMDGPU_RAS_BLOCK__GFX_SQC_DATA_BANKA_MISS_FIFO,
189 AMDGPU_RAS_BLOCK__GFX_SQC_DATA_BANKA_DIRTY_BIT_RAM,
190 AMDGPU_RAS_BLOCK__GFX_SQC_DATA_BANKA_BANK_RAM,
191 AMDGPU_RAS_BLOCK__GFX_SQC_INDEX1_END =
192 AMDGPU_RAS_BLOCK__GFX_SQC_DATA_BANKA_BANK_RAM,
194 AMDGPU_RAS_BLOCK__GFX_SQC_INDEX2_START,
195 AMDGPU_RAS_BLOCK__GFX_SQC_INST_BANKB_TAG_RAM =
196 AMDGPU_RAS_BLOCK__GFX_SQC_INDEX2_START,
197 AMDGPU_RAS_BLOCK__GFX_SQC_INST_BANKB_UTCL1_MISS_FIFO,
198 AMDGPU_RAS_BLOCK__GFX_SQC_INST_BANKB_MISS_FIFO,
199 AMDGPU_RAS_BLOCK__GFX_SQC_INST_BANKB_BANK_RAM,
200 AMDGPU_RAS_BLOCK__GFX_SQC_DATA_BANKB_TAG_RAM,
201 AMDGPU_RAS_BLOCK__GFX_SQC_DATA_BANKB_HIT_FIFO,
202 AMDGPU_RAS_BLOCK__GFX_SQC_DATA_BANKB_MISS_FIFO,
203 AMDGPU_RAS_BLOCK__GFX_SQC_DATA_BANKB_DIRTY_BIT_RAM,
204 AMDGPU_RAS_BLOCK__GFX_SQC_DATA_BANKB_BANK_RAM,
205 AMDGPU_RAS_BLOCK__GFX_SQC_INDEX2_END =
206 AMDGPU_RAS_BLOCK__GFX_SQC_DATA_BANKB_BANK_RAM,
207 AMDGPU_RAS_BLOCK__GFX_SQC_INDEX_END =
208 AMDGPU_RAS_BLOCK__GFX_SQC_INDEX2_END,
210 AMDGPU_RAS_BLOCK__GFX_TA_INDEX_START,
211 AMDGPU_RAS_BLOCK__GFX_TA_FS_DFIFO =
212 AMDGPU_RAS_BLOCK__GFX_TA_INDEX_START,
213 AMDGPU_RAS_BLOCK__GFX_TA_FS_AFIFO,
214 AMDGPU_RAS_BLOCK__GFX_TA_FL_LFIFO,
215 AMDGPU_RAS_BLOCK__GFX_TA_FX_LFIFO,
216 AMDGPU_RAS_BLOCK__GFX_TA_FS_CFIFO,
217 AMDGPU_RAS_BLOCK__GFX_TA_INDEX_END = AMDGPU_RAS_BLOCK__GFX_TA_FS_CFIFO,
219 AMDGPU_RAS_BLOCK__GFX_TCA_INDEX_START,
220 AMDGPU_RAS_BLOCK__GFX_TCA_HOLE_FIFO =
221 AMDGPU_RAS_BLOCK__GFX_TCA_INDEX_START,
222 AMDGPU_RAS_BLOCK__GFX_TCA_REQ_FIFO,
223 AMDGPU_RAS_BLOCK__GFX_TCA_INDEX_END =
224 AMDGPU_RAS_BLOCK__GFX_TCA_REQ_FIFO,
225 /* TCC (5 sub-ranges) */
226 AMDGPU_RAS_BLOCK__GFX_TCC_INDEX_START,
228 AMDGPU_RAS_BLOCK__GFX_TCC_INDEX0_START =
229 AMDGPU_RAS_BLOCK__GFX_TCC_INDEX_START,
230 AMDGPU_RAS_BLOCK__GFX_TCC_CACHE_DATA =
231 AMDGPU_RAS_BLOCK__GFX_TCC_INDEX0_START,
232 AMDGPU_RAS_BLOCK__GFX_TCC_CACHE_DATA_BANK_0_1,
233 AMDGPU_RAS_BLOCK__GFX_TCC_CACHE_DATA_BANK_1_0,
234 AMDGPU_RAS_BLOCK__GFX_TCC_CACHE_DATA_BANK_1_1,
235 AMDGPU_RAS_BLOCK__GFX_TCC_CACHE_DIRTY_BANK_0,
236 AMDGPU_RAS_BLOCK__GFX_TCC_CACHE_DIRTY_BANK_1,
237 AMDGPU_RAS_BLOCK__GFX_TCC_HIGH_RATE_TAG,
238 AMDGPU_RAS_BLOCK__GFX_TCC_LOW_RATE_TAG,
239 AMDGPU_RAS_BLOCK__GFX_TCC_INDEX0_END =
240 AMDGPU_RAS_BLOCK__GFX_TCC_LOW_RATE_TAG,
242 AMDGPU_RAS_BLOCK__GFX_TCC_INDEX1_START,
243 AMDGPU_RAS_BLOCK__GFX_TCC_IN_USE_DEC =
244 AMDGPU_RAS_BLOCK__GFX_TCC_INDEX1_START,
245 AMDGPU_RAS_BLOCK__GFX_TCC_IN_USE_TRANSFER,
246 AMDGPU_RAS_BLOCK__GFX_TCC_INDEX1_END =
247 AMDGPU_RAS_BLOCK__GFX_TCC_IN_USE_TRANSFER,
249 AMDGPU_RAS_BLOCK__GFX_TCC_INDEX2_START,
250 AMDGPU_RAS_BLOCK__GFX_TCC_RETURN_DATA =
251 AMDGPU_RAS_BLOCK__GFX_TCC_INDEX2_START,
252 AMDGPU_RAS_BLOCK__GFX_TCC_RETURN_CONTROL,
253 AMDGPU_RAS_BLOCK__GFX_TCC_UC_ATOMIC_FIFO,
254 AMDGPU_RAS_BLOCK__GFX_TCC_WRITE_RETURN,
255 AMDGPU_RAS_BLOCK__GFX_TCC_WRITE_CACHE_READ,
256 AMDGPU_RAS_BLOCK__GFX_TCC_SRC_FIFO,
257 AMDGPU_RAS_BLOCK__GFX_TCC_SRC_FIFO_NEXT_RAM,
258 AMDGPU_RAS_BLOCK__GFX_TCC_CACHE_TAG_PROBE_FIFO,
259 AMDGPU_RAS_BLOCK__GFX_TCC_INDEX2_END =
260 AMDGPU_RAS_BLOCK__GFX_TCC_CACHE_TAG_PROBE_FIFO,
262 AMDGPU_RAS_BLOCK__GFX_TCC_INDEX3_START,
263 AMDGPU_RAS_BLOCK__GFX_TCC_LATENCY_FIFO =
264 AMDGPU_RAS_BLOCK__GFX_TCC_INDEX3_START,
265 AMDGPU_RAS_BLOCK__GFX_TCC_LATENCY_FIFO_NEXT_RAM,
266 AMDGPU_RAS_BLOCK__GFX_TCC_INDEX3_END =
267 AMDGPU_RAS_BLOCK__GFX_TCC_LATENCY_FIFO_NEXT_RAM,
269 AMDGPU_RAS_BLOCK__GFX_TCC_INDEX4_START,
270 AMDGPU_RAS_BLOCK__GFX_TCC_WRRET_TAG_WRITE_RETURN =
271 AMDGPU_RAS_BLOCK__GFX_TCC_INDEX4_START,
272 AMDGPU_RAS_BLOCK__GFX_TCC_ATOMIC_RETURN_BUFFER,
273 AMDGPU_RAS_BLOCK__GFX_TCC_INDEX4_END =
274 AMDGPU_RAS_BLOCK__GFX_TCC_ATOMIC_RETURN_BUFFER,
275 AMDGPU_RAS_BLOCK__GFX_TCC_INDEX_END =
276 AMDGPU_RAS_BLOCK__GFX_TCC_INDEX4_END,
278 AMDGPU_RAS_BLOCK__GFX_TCI_WRITE_RAM,
280 AMDGPU_RAS_BLOCK__GFX_TCP_INDEX_START,
281 AMDGPU_RAS_BLOCK__GFX_TCP_CACHE_RAM =
282 AMDGPU_RAS_BLOCK__GFX_TCP_INDEX_START,
283 AMDGPU_RAS_BLOCK__GFX_TCP_LFIFO_RAM,
284 AMDGPU_RAS_BLOCK__GFX_TCP_CMD_FIFO,
285 AMDGPU_RAS_BLOCK__GFX_TCP_VM_FIFO,
286 AMDGPU_RAS_BLOCK__GFX_TCP_DB_RAM,
287 AMDGPU_RAS_BLOCK__GFX_TCP_UTCL1_LFIFO0,
288 AMDGPU_RAS_BLOCK__GFX_TCP_UTCL1_LFIFO1,
289 AMDGPU_RAS_BLOCK__GFX_TCP_INDEX_END =
290 AMDGPU_RAS_BLOCK__GFX_TCP_UTCL1_LFIFO1,
292 AMDGPU_RAS_BLOCK__GFX_TD_INDEX_START,
293 AMDGPU_RAS_BLOCK__GFX_TD_SS_FIFO_LO =
294 AMDGPU_RAS_BLOCK__GFX_TD_INDEX_START,
295 AMDGPU_RAS_BLOCK__GFX_TD_SS_FIFO_HI,
296 AMDGPU_RAS_BLOCK__GFX_TD_CS_FIFO,
297 AMDGPU_RAS_BLOCK__GFX_TD_INDEX_END = AMDGPU_RAS_BLOCK__GFX_TD_CS_FIFO,
298 /* EA (3 sub-ranges) */
299 AMDGPU_RAS_BLOCK__GFX_EA_INDEX_START,
301 AMDGPU_RAS_BLOCK__GFX_EA_INDEX0_START =
302 AMDGPU_RAS_BLOCK__GFX_EA_INDEX_START,
303 AMDGPU_RAS_BLOCK__GFX_EA_DRAMRD_CMDMEM =
304 AMDGPU_RAS_BLOCK__GFX_EA_INDEX0_START,
305 AMDGPU_RAS_BLOCK__GFX_EA_DRAMWR_CMDMEM,
306 AMDGPU_RAS_BLOCK__GFX_EA_DRAMWR_DATAMEM,
307 AMDGPU_RAS_BLOCK__GFX_EA_RRET_TAGMEM,
308 AMDGPU_RAS_BLOCK__GFX_EA_WRET_TAGMEM,
309 AMDGPU_RAS_BLOCK__GFX_EA_GMIRD_CMDMEM,
310 AMDGPU_RAS_BLOCK__GFX_EA_GMIWR_CMDMEM,
311 AMDGPU_RAS_BLOCK__GFX_EA_GMIWR_DATAMEM,
312 AMDGPU_RAS_BLOCK__GFX_EA_INDEX0_END =
313 AMDGPU_RAS_BLOCK__GFX_EA_GMIWR_DATAMEM,
315 AMDGPU_RAS_BLOCK__GFX_EA_INDEX1_START,
316 AMDGPU_RAS_BLOCK__GFX_EA_DRAMRD_PAGEMEM =
317 AMDGPU_RAS_BLOCK__GFX_EA_INDEX1_START,
318 AMDGPU_RAS_BLOCK__GFX_EA_DRAMWR_PAGEMEM,
319 AMDGPU_RAS_BLOCK__GFX_EA_IORD_CMDMEM,
320 AMDGPU_RAS_BLOCK__GFX_EA_IOWR_CMDMEM,
321 AMDGPU_RAS_BLOCK__GFX_EA_IOWR_DATAMEM,
322 AMDGPU_RAS_BLOCK__GFX_EA_GMIRD_PAGEMEM,
323 AMDGPU_RAS_BLOCK__GFX_EA_GMIWR_PAGEMEM,
324 AMDGPU_RAS_BLOCK__GFX_EA_INDEX1_END =
325 AMDGPU_RAS_BLOCK__GFX_EA_GMIWR_PAGEMEM,
327 AMDGPU_RAS_BLOCK__GFX_EA_INDEX2_START,
328 AMDGPU_RAS_BLOCK__GFX_EA_MAM_D0MEM =
329 AMDGPU_RAS_BLOCK__GFX_EA_INDEX2_START,
330 AMDGPU_RAS_BLOCK__GFX_EA_MAM_D1MEM,
331 AMDGPU_RAS_BLOCK__GFX_EA_MAM_D2MEM,
332 AMDGPU_RAS_BLOCK__GFX_EA_MAM_D3MEM,
333 AMDGPU_RAS_BLOCK__GFX_EA_INDEX2_END =
334 AMDGPU_RAS_BLOCK__GFX_EA_MAM_D3MEM,
335 AMDGPU_RAS_BLOCK__GFX_EA_INDEX_END =
336 AMDGPU_RAS_BLOCK__GFX_EA_INDEX2_END,
338 AMDGPU_RAS_BLOCK__UTC_VML2_BANK_CACHE,
340 AMDGPU_RAS_BLOCK__UTC_VML2_WALKER,
341 /* UTC ATC L2 2MB cache */
342 AMDGPU_RAS_BLOCK__UTC_ATCL2_CACHE_2M_BANK,
343 /* UTC ATC L2 4KB cache */
344 AMDGPU_RAS_BLOCK__UTC_ATCL2_CACHE_4K_BANK,
345 AMDGPU_RAS_BLOCK__GFX_MAX
348 enum amdgpu_ras_error_type {
349 AMDGPU_RAS_ERROR__NONE = 0,
350 AMDGPU_RAS_ERROR__PARITY = 1,
351 AMDGPU_RAS_ERROR__SINGLE_CORRECTABLE = 2,
352 AMDGPU_RAS_ERROR__MULTI_UNCORRECTABLE = 4,
353 AMDGPU_RAS_ERROR__POISON = 8,
356 enum amdgpu_ras_ret {
357 AMDGPU_RAS_SUCCESS = 0,
364 enum amdgpu_ras_error_query_mode {
365 AMDGPU_RAS_INVALID_ERROR_QUERY = 0,
366 AMDGPU_RAS_DIRECT_ERROR_QUERY = 1,
367 AMDGPU_RAS_FIRMWARE_ERROR_QUERY = 2,
370 /* ras error status reisger fields */
371 #define ERR_STATUS_LO__ERR_STATUS_VALID_FLAG__SHIFT 0x0
372 #define ERR_STATUS_LO__ERR_STATUS_VALID_FLAG_MASK 0x00000001L
373 #define ERR_STATUS_LO__MEMORY_ID__SHIFT 0x18
374 #define ERR_STATUS_LO__MEMORY_ID_MASK 0xFF000000L
375 #define ERR_STATUS_HI__ERR_INFO_VALID_FLAG__SHIFT 0x2
376 #define ERR_STATUS_HI__ERR_INFO_VALID_FLAG_MASK 0x00000004L
377 #define ERR_STATUS__ERR_CNT__SHIFT 0x17
378 #define ERR_STATUS__ERR_CNT_MASK 0x03800000L
380 #define AMDGPU_RAS_REG_ENTRY(ip, inst, reg_lo, reg_hi) \
381 ip##_HWIP, inst, reg_lo##_BASE_IDX, reg_lo, reg_hi##_BASE_IDX, reg_hi
383 #define AMDGPU_RAS_REG_ENTRY_OFFSET(hwip, ip_inst, segment, reg) \
384 (adev->reg_offset[hwip][ip_inst][segment] + (reg))
386 #define AMDGPU_RAS_ERR_INFO_VALID (1 << 0)
387 #define AMDGPU_RAS_ERR_STATUS_VALID (1 << 1)
388 #define AMDGPU_RAS_ERR_ADDRESS_VALID (1 << 2)
390 #define AMDGPU_RAS_GPU_RESET_MODE2_RESET (0x1 << 0)
391 #define AMDGPU_RAS_GPU_RESET_MODE1_RESET (0x1 << 1)
393 struct amdgpu_ras_err_status_reg_entry {
402 const char *block_name;
405 struct amdgpu_ras_memory_id_entry {
410 struct ras_common_if {
411 enum amdgpu_ras_block block;
412 enum amdgpu_ras_error_type type;
413 uint32_t sub_block_index;
417 #define MAX_UMC_CHANNEL_NUM 32
419 struct ecc_info_per_ch {
420 uint16_t ce_count_lo_chip;
421 uint16_t ce_count_hi_chip;
422 uint64_t mca_umc_status;
423 uint64_t mca_umc_addr;
424 uint64_t mca_ceumc_addr;
427 struct umc_ecc_info {
428 struct ecc_info_per_ch ecc[MAX_UMC_CHANNEL_NUM];
430 /* Determine smu ecctable whether support
431 * record correctable error address
433 int record_ce_addr_supported;
436 enum ras_event_type {
437 RAS_EVENT_TYPE_INVALID = 0,
438 RAS_EVENT_TYPE_FATAL,
439 RAS_EVENT_TYPE_POISON_CREATION,
440 RAS_EVENT_TYPE_POISON_CONSUMPTION,
441 RAS_EVENT_TYPE_COUNT,
444 struct ras_event_state {
449 struct ras_event_manager {
451 struct ras_event_state event_state[RAS_EVENT_TYPE_COUNT];
454 struct ras_event_id {
455 enum ras_event_type type;
459 struct ras_query_context {
460 struct ras_event_id evid;
463 typedef int (*pasid_notify)(struct amdgpu_device *adev,
464 uint16_t pasid, void *data);
466 struct ras_poison_msg {
467 enum amdgpu_ras_block block;
470 pasid_notify pasid_fn;
474 struct ras_err_pages {
484 struct ras_err_pages err_pages;
487 struct ras_ecc_log_info {
489 struct radix_tree_root de_page_tree;
490 uint64_t de_queried_count;
491 uint64_t prev_de_queried_count;
495 /* ras infrastructure */
496 /* for ras itself. */
499 struct list_head head;
501 struct device_attribute features_attr;
502 struct device_attribute version_attr;
503 struct device_attribute schema_attr;
504 struct device_attribute event_state_attr;
505 struct bin_attribute badpages_attr;
506 struct dentry *de_ras_eeprom_table;
508 struct ras_manager *objs;
511 struct work_struct recovery_work;
512 atomic_t in_recovery;
513 struct amdgpu_device *adev;
514 /* error handler data */
515 struct ras_err_handler_data *eh_data;
516 struct mutex recovery_lock;
520 struct amdgpu_ras_eeprom_control eeprom_control;
522 bool error_query_ready;
524 /* bad page count threshold */
525 uint32_t bad_page_cnt_threshold;
527 /* disable ras error count harvest in recovery */
528 bool disable_ras_err_cnt_harvest;
530 /* is poison mode supported */
531 bool poison_supported;
533 /* RAS count errors delayed work */
534 struct delayed_work ras_counte_delay_work;
535 atomic_t ras_ue_count;
536 atomic_t ras_ce_count;
538 /* record umc error info queried from smu */
539 struct umc_ecc_info umc_ecc;
541 /* Indicates smu whether need update bad channel info */
542 bool update_channel_flag;
543 /* Record status of smu mca debug mode */
544 bool is_aca_debug_mode;
547 /* Record special requirements of gpu reset caller */
548 uint32_t gpu_reset_flags;
550 struct task_struct *page_retirement_thread;
551 wait_queue_head_t page_retirement_wq;
552 struct mutex page_retirement_lock;
553 atomic_t page_retirement_req_cnt;
554 atomic_t poison_creation_count;
555 struct mutex page_rsv_lock;
556 DECLARE_KFIFO(poison_fifo, struct ras_poison_msg, 128);
557 struct ras_ecc_log_info umc_ecc_log;
558 struct delayed_work page_retirement_dwork;
560 /* Fatal error detected flag */
563 /* RAS event manager */
564 struct ras_event_manager __event_mgr;
565 struct ras_event_manager *event_mgr;
567 uint64_t reserved_pages_in_bytes;
572 char debugfs_name[32];
575 struct ras_err_info {
576 struct amdgpu_smuio_mcm_config_info mcm_info;
582 struct ras_err_node {
583 struct list_head node;
584 struct ras_err_info err_info;
587 struct ras_err_data {
588 unsigned long ue_count;
589 unsigned long ce_count;
590 unsigned long de_count;
591 unsigned long err_addr_cnt;
592 struct eeprom_table_record *err_addr;
593 unsigned long err_addr_len;
595 struct list_head err_node_list;
598 #define for_each_ras_error(err_node, err_data) \
599 list_for_each_entry(err_node, &(err_data)->err_node_list, node)
601 struct ras_err_handler_data {
602 /* point to bad page records array */
603 struct eeprom_table_record *bps;
604 /* the count of entries */
606 /* the space can place new entries */
610 typedef int (*ras_ih_cb)(struct amdgpu_device *adev,
612 struct amdgpu_iv_entry *entry);
615 /* interrupt bottom half */
616 struct work_struct ih_work;
620 /* full of entries */
622 unsigned int ring_size;
623 unsigned int element_size;
624 unsigned int aligned_element_size;
630 struct ras_common_if head;
631 /* reference count */
634 struct list_head node;
636 struct amdgpu_device *adev;
638 struct device_attribute sysfs_attr;
642 struct ras_fs_data fs_data;
645 struct ras_ih_data ih_data;
647 struct ras_err_data err_data;
649 struct aca_handle aca_handle;
658 /* interfaces for IP */
660 struct ras_common_if head;
661 const char* sysfs_name;
662 char debugfs_name[32];
665 struct ras_query_if {
666 struct ras_common_if head;
667 unsigned long ue_count;
668 unsigned long ce_count;
669 unsigned long de_count;
672 struct ras_inject_if {
673 struct ras_common_if head;
676 uint32_t instance_mask;
680 struct ras_common_if head;
685 struct ras_common_if head;
689 struct ras_dispatch_if {
690 struct ras_common_if head;
691 struct amdgpu_iv_entry *entry;
694 struct ras_debug_if {
696 struct ras_common_if head;
697 struct ras_inject_if inject;
702 struct amdgpu_ras_block_object {
703 struct ras_common_if ras_comm;
705 int (*ras_block_match)(struct amdgpu_ras_block_object *block_obj,
706 enum amdgpu_ras_block block, uint32_t sub_block_index);
707 int (*ras_late_init)(struct amdgpu_device *adev, struct ras_common_if *ras_block);
708 void (*ras_fini)(struct amdgpu_device *adev, struct ras_common_if *ras_block);
710 const struct amdgpu_ras_block_hw_ops *hw_ops;
713 struct amdgpu_ras_block_hw_ops {
714 int (*ras_error_inject)(struct amdgpu_device *adev,
715 void *inject_if, uint32_t instance_mask);
716 void (*query_ras_error_count)(struct amdgpu_device *adev, void *ras_error_status);
717 void (*query_ras_error_status)(struct amdgpu_device *adev);
718 void (*query_ras_error_address)(struct amdgpu_device *adev, void *ras_error_status);
719 void (*reset_ras_error_count)(struct amdgpu_device *adev);
720 void (*reset_ras_error_status)(struct amdgpu_device *adev);
721 bool (*query_poison_status)(struct amdgpu_device *adev);
722 bool (*handle_poison_consumption)(struct amdgpu_device *adev);
727 * 1: ras feature enable (enabled by default)
729 * 2: ras framework init (in ip_init)
732 * 4: debugfs/sysfs create
734 * 6: debugfs/sysfs remove
740 int amdgpu_ras_recovery_init(struct amdgpu_device *adev);
742 void amdgpu_ras_resume(struct amdgpu_device *adev);
743 void amdgpu_ras_suspend(struct amdgpu_device *adev);
745 int amdgpu_ras_query_error_count(struct amdgpu_device *adev,
746 unsigned long *ce_count,
747 unsigned long *ue_count,
748 struct ras_query_if *query_info);
750 /* error handling functions */
751 int amdgpu_ras_add_bad_pages(struct amdgpu_device *adev,
752 struct eeprom_table_record *bps, int pages);
754 int amdgpu_ras_save_bad_pages(struct amdgpu_device *adev,
755 unsigned long *new_cnt);
757 static inline enum ta_ras_block
758 amdgpu_ras_block_to_ta(enum amdgpu_ras_block block) {
760 case AMDGPU_RAS_BLOCK__UMC:
761 return TA_RAS_BLOCK__UMC;
762 case AMDGPU_RAS_BLOCK__SDMA:
763 return TA_RAS_BLOCK__SDMA;
764 case AMDGPU_RAS_BLOCK__GFX:
765 return TA_RAS_BLOCK__GFX;
766 case AMDGPU_RAS_BLOCK__MMHUB:
767 return TA_RAS_BLOCK__MMHUB;
768 case AMDGPU_RAS_BLOCK__ATHUB:
769 return TA_RAS_BLOCK__ATHUB;
770 case AMDGPU_RAS_BLOCK__PCIE_BIF:
771 return TA_RAS_BLOCK__PCIE_BIF;
772 case AMDGPU_RAS_BLOCK__HDP:
773 return TA_RAS_BLOCK__HDP;
774 case AMDGPU_RAS_BLOCK__XGMI_WAFL:
775 return TA_RAS_BLOCK__XGMI_WAFL;
776 case AMDGPU_RAS_BLOCK__DF:
777 return TA_RAS_BLOCK__DF;
778 case AMDGPU_RAS_BLOCK__SMN:
779 return TA_RAS_BLOCK__SMN;
780 case AMDGPU_RAS_BLOCK__SEM:
781 return TA_RAS_BLOCK__SEM;
782 case AMDGPU_RAS_BLOCK__MP0:
783 return TA_RAS_BLOCK__MP0;
784 case AMDGPU_RAS_BLOCK__MP1:
785 return TA_RAS_BLOCK__MP1;
786 case AMDGPU_RAS_BLOCK__FUSE:
787 return TA_RAS_BLOCK__FUSE;
788 case AMDGPU_RAS_BLOCK__MCA:
789 return TA_RAS_BLOCK__MCA;
790 case AMDGPU_RAS_BLOCK__VCN:
791 return TA_RAS_BLOCK__VCN;
792 case AMDGPU_RAS_BLOCK__JPEG:
793 return TA_RAS_BLOCK__JPEG;
795 WARN_ONCE(1, "RAS ERROR: unexpected block id %d\n", block);
796 return TA_RAS_BLOCK__UMC;
800 static inline enum ta_ras_error_type
801 amdgpu_ras_error_to_ta(enum amdgpu_ras_error_type error) {
803 case AMDGPU_RAS_ERROR__NONE:
804 return TA_RAS_ERROR__NONE;
805 case AMDGPU_RAS_ERROR__PARITY:
806 return TA_RAS_ERROR__PARITY;
807 case AMDGPU_RAS_ERROR__SINGLE_CORRECTABLE:
808 return TA_RAS_ERROR__SINGLE_CORRECTABLE;
809 case AMDGPU_RAS_ERROR__MULTI_UNCORRECTABLE:
810 return TA_RAS_ERROR__MULTI_UNCORRECTABLE;
811 case AMDGPU_RAS_ERROR__POISON:
812 return TA_RAS_ERROR__POISON;
814 WARN_ONCE(1, "RAS ERROR: unexpected error type %d\n", error);
815 return TA_RAS_ERROR__NONE;
819 /* called in ip_init and ip_fini */
820 int amdgpu_ras_init(struct amdgpu_device *adev);
821 int amdgpu_ras_late_init(struct amdgpu_device *adev);
822 int amdgpu_ras_fini(struct amdgpu_device *adev);
823 int amdgpu_ras_pre_fini(struct amdgpu_device *adev);
825 int amdgpu_ras_block_late_init(struct amdgpu_device *adev,
826 struct ras_common_if *ras_block);
828 void amdgpu_ras_block_late_fini(struct amdgpu_device *adev,
829 struct ras_common_if *ras_block);
831 int amdgpu_ras_feature_enable(struct amdgpu_device *adev,
832 struct ras_common_if *head, bool enable);
834 int amdgpu_ras_feature_enable_on_boot(struct amdgpu_device *adev,
835 struct ras_common_if *head, bool enable);
837 int amdgpu_ras_sysfs_create(struct amdgpu_device *adev,
838 struct ras_common_if *head);
840 int amdgpu_ras_sysfs_remove(struct amdgpu_device *adev,
841 struct ras_common_if *head);
843 void amdgpu_ras_debugfs_create_all(struct amdgpu_device *adev);
845 int amdgpu_ras_query_error_status(struct amdgpu_device *adev,
846 struct ras_query_if *info);
848 int amdgpu_ras_reset_error_count(struct amdgpu_device *adev,
849 enum amdgpu_ras_block block);
850 int amdgpu_ras_reset_error_status(struct amdgpu_device *adev,
851 enum amdgpu_ras_block block);
853 int amdgpu_ras_error_inject(struct amdgpu_device *adev,
854 struct ras_inject_if *info);
856 int amdgpu_ras_interrupt_add_handler(struct amdgpu_device *adev,
857 struct ras_common_if *head);
859 int amdgpu_ras_interrupt_remove_handler(struct amdgpu_device *adev,
860 struct ras_common_if *head);
862 int amdgpu_ras_interrupt_dispatch(struct amdgpu_device *adev,
863 struct ras_dispatch_if *info);
865 struct ras_manager *amdgpu_ras_find_obj(struct amdgpu_device *adev,
866 struct ras_common_if *head);
868 extern atomic_t amdgpu_ras_in_intr;
870 static inline bool amdgpu_ras_intr_triggered(void)
872 return !!atomic_read(&amdgpu_ras_in_intr);
875 static inline void amdgpu_ras_intr_cleared(void)
877 atomic_set(&amdgpu_ras_in_intr, 0);
880 void amdgpu_ras_global_ras_isr(struct amdgpu_device *adev);
882 void amdgpu_ras_set_error_query_ready(struct amdgpu_device *adev, bool ready);
884 bool amdgpu_ras_need_emergency_restart(struct amdgpu_device *adev);
886 void amdgpu_release_ras_context(struct amdgpu_device *adev);
888 int amdgpu_persistent_edc_harvesting_supported(struct amdgpu_device *adev);
890 const char *get_ras_block_str(struct ras_common_if *ras_block);
892 bool amdgpu_ras_is_poison_mode_supported(struct amdgpu_device *adev);
894 int amdgpu_ras_is_supported(struct amdgpu_device *adev, unsigned int block);
896 int amdgpu_ras_reset_gpu(struct amdgpu_device *adev);
898 struct amdgpu_ras* amdgpu_ras_get_context(struct amdgpu_device *adev);
900 int amdgpu_ras_set_context(struct amdgpu_device *adev, struct amdgpu_ras *ras_con);
902 int amdgpu_ras_set_mca_debug_mode(struct amdgpu_device *adev, bool enable);
903 int amdgpu_ras_set_aca_debug_mode(struct amdgpu_device *adev, bool enable);
904 bool amdgpu_ras_get_aca_debug_mode(struct amdgpu_device *adev);
905 bool amdgpu_ras_get_error_query_mode(struct amdgpu_device *adev,
908 int amdgpu_ras_register_ras_block(struct amdgpu_device *adev,
909 struct amdgpu_ras_block_object *ras_block_obj);
910 void amdgpu_ras_interrupt_fatal_error_handler(struct amdgpu_device *adev);
911 void amdgpu_ras_get_error_type_name(uint32_t err_type, char *err_type_name);
912 bool amdgpu_ras_inst_get_memory_id_field(struct amdgpu_device *adev,
913 const struct amdgpu_ras_err_status_reg_entry *reg_entry,
915 uint32_t *memory_id);
916 bool amdgpu_ras_inst_get_err_cnt_field(struct amdgpu_device *adev,
917 const struct amdgpu_ras_err_status_reg_entry *reg_entry,
919 unsigned long *err_cnt);
920 void amdgpu_ras_inst_query_ras_error_count(struct amdgpu_device *adev,
921 const struct amdgpu_ras_err_status_reg_entry *reg_list,
922 uint32_t reg_list_size,
923 const struct amdgpu_ras_memory_id_entry *mem_list,
924 uint32_t mem_list_size,
927 unsigned long *err_count);
928 void amdgpu_ras_inst_reset_ras_error_count(struct amdgpu_device *adev,
929 const struct amdgpu_ras_err_status_reg_entry *reg_list,
930 uint32_t reg_list_size,
933 int amdgpu_ras_error_data_init(struct ras_err_data *err_data);
934 void amdgpu_ras_error_data_fini(struct ras_err_data *err_data);
935 int amdgpu_ras_error_statistic_ce_count(struct ras_err_data *err_data,
936 struct amdgpu_smuio_mcm_config_info *mcm_info,
938 int amdgpu_ras_error_statistic_ue_count(struct ras_err_data *err_data,
939 struct amdgpu_smuio_mcm_config_info *mcm_info,
941 int amdgpu_ras_error_statistic_de_count(struct ras_err_data *err_data,
942 struct amdgpu_smuio_mcm_config_info *mcm_info,
944 void amdgpu_ras_query_boot_status(struct amdgpu_device *adev, u32 num_instances);
945 int amdgpu_ras_bind_aca(struct amdgpu_device *adev, enum amdgpu_ras_block blk,
946 const struct aca_info *aca_info, void *data);
947 int amdgpu_ras_unbind_aca(struct amdgpu_device *adev, enum amdgpu_ras_block blk);
949 ssize_t amdgpu_ras_aca_sysfs_read(struct device *dev, struct device_attribute *attr,
950 struct aca_handle *handle, char *buf, void *data);
952 void amdgpu_ras_set_fed(struct amdgpu_device *adev, bool status);
953 bool amdgpu_ras_get_fed_status(struct amdgpu_device *adev);
955 u64 amdgpu_ras_acquire_event_id(struct amdgpu_device *adev, enum ras_event_type type);
956 int amdgpu_ras_mark_ras_event_caller(struct amdgpu_device *adev, enum ras_event_type type,
959 int amdgpu_ras_reserve_page(struct amdgpu_device *adev, uint64_t pfn);
961 int amdgpu_ras_put_poison_req(struct amdgpu_device *adev,
962 enum amdgpu_ras_block block, uint16_t pasid,
963 pasid_notify pasid_fn, void *data, uint32_t reset);
965 bool amdgpu_ras_in_recovery(struct amdgpu_device *adev);
968 void amdgpu_ras_event_log_print(struct amdgpu_device *adev, u64 event_id,
969 const char *fmt, ...);
971 bool amdgpu_ras_is_rma(struct amdgpu_device *adev);