diff options
Diffstat (limited to 'drivers/edac')
-rw-r--r-- | drivers/edac/amd64_edac.c | 4 | ||||
-rw-r--r-- | drivers/edac/mce_amd.c | 450 | ||||
-rw-r--r-- | drivers/edac/mce_amd.h | 14 | ||||
-rw-r--r-- | drivers/edac/mce_amd_inj.c | 9 |
4 files changed, 359 insertions, 118 deletions
diff --git a/drivers/edac/amd64_edac.c b/drivers/edac/amd64_edac.c index 6bf7e248e758..4a5ecc58025d 100644 --- a/drivers/edac/amd64_edac.c +++ b/drivers/edac/amd64_edac.c | |||
@@ -1950,8 +1950,8 @@ static void amd64_handle_ue(struct mem_ctl_info *mci, | |||
1950 | static inline void __amd64_decode_bus_error(struct mem_ctl_info *mci, | 1950 | static inline void __amd64_decode_bus_error(struct mem_ctl_info *mci, |
1951 | struct err_regs *info) | 1951 | struct err_regs *info) |
1952 | { | 1952 | { |
1953 | u32 ec = ERROR_CODE(info->nbsl); | 1953 | u16 ec = EC(info->nbsl); |
1954 | u32 xec = EXT_ERROR_CODE(info->nbsl); | 1954 | u8 xec = XEC(info->nbsl, 0x1f); |
1955 | int ecc_type = (info->nbsh >> 13) & 0x3; | 1955 | int ecc_type = (info->nbsh >> 13) & 0x3; |
1956 | 1956 | ||
1957 | /* Bail early out if this was an 'observed' error */ | 1957 | /* Bail early out if this was an 'observed' error */ |
diff --git a/drivers/edac/mce_amd.c b/drivers/edac/mce_amd.c index c0181093b490..f6cf73d93359 100644 --- a/drivers/edac/mce_amd.c +++ b/drivers/edac/mce_amd.c | |||
@@ -5,6 +5,7 @@ | |||
5 | 5 | ||
6 | static struct amd_decoder_ops *fam_ops; | 6 | static struct amd_decoder_ops *fam_ops; |
7 | 7 | ||
8 | static u8 xec_mask = 0xf; | ||
8 | static u8 nb_err_cpumask = 0xf; | 9 | static u8 nb_err_cpumask = 0xf; |
9 | 10 | ||
10 | static bool report_gart_errors; | 11 | static bool report_gart_errors; |
@@ -74,57 +75,104 @@ static const char *f10h_nb_mce_desc[] = { | |||
74 | "ECC Error in the Probe Filter directory" | 75 | "ECC Error in the Probe Filter directory" |
75 | }; | 76 | }; |
76 | 77 | ||
77 | static bool f12h_dc_mce(u16 ec) | 78 | static const char * const f15h_ic_mce_desc[] = { |
79 | "UC during a demand linefill from L2", | ||
80 | "Parity error during data load from IC", | ||
81 | "Parity error for IC valid bit", | ||
82 | "Main tag parity error", | ||
83 | "Parity error in prediction queue", | ||
84 | "PFB data/address parity error", | ||
85 | "Parity error in the branch status reg", | ||
86 | "PFB promotion address error", | ||
87 | "Tag error during probe/victimization", | ||
88 | "Parity error for IC probe tag valid bit", | ||
89 | "PFB non-cacheable bit parity error", | ||
90 | "PFB valid bit parity error", /* xec = 0xd */ | ||
91 | "patch RAM", /* xec = 010 */ | ||
92 | "uop queue", | ||
93 | "insn buffer", | ||
94 | "predecode buffer", | ||
95 | "fetch address FIFO" | ||
96 | }; | ||
97 | |||
98 | static const char * const f15h_cu_mce_desc[] = { | ||
99 | "Fill ECC error on data fills", /* xec = 0x4 */ | ||
100 | "Fill parity error on insn fills", | ||
101 | "Prefetcher request FIFO parity error", | ||
102 | "PRQ address parity error", | ||
103 | "PRQ data parity error", | ||
104 | "WCC Tag ECC error", | ||
105 | "WCC Data ECC error", | ||
106 | "WCB Data parity error", | ||
107 | "VB Data/ECC error", | ||
108 | "L2 Tag ECC error", /* xec = 0x10 */ | ||
109 | "Hard L2 Tag ECC error", | ||
110 | "Multiple hits on L2 tag", | ||
111 | "XAB parity error", | ||
112 | "PRB address parity error" | ||
113 | }; | ||
114 | |||
115 | static const char * const fr_ex_mce_desc[] = { | ||
116 | "CPU Watchdog timer expire", | ||
117 | "Wakeup array dest tag", | ||
118 | "AG payload array", | ||
119 | "EX payload array", | ||
120 | "IDRF array", | ||
121 | "Retire dispatch queue", | ||
122 | "Mapper checkpoint array", | ||
123 | "Physical register file EX0 port", | ||
124 | "Physical register file EX1 port", | ||
125 | "Physical register file AG0 port", | ||
126 | "Physical register file AG1 port", | ||
127 | "Flag register file", | ||
128 | "DE correctable error could not be corrected" | ||
129 | }; | ||
130 | |||
131 | static bool f12h_dc_mce(u16 ec, u8 xec) | ||
78 | { | 132 | { |
79 | bool ret = false; | 133 | bool ret = false; |
80 | 134 | ||
81 | if (MEM_ERROR(ec)) { | 135 | if (MEM_ERROR(ec)) { |
82 | u8 ll = ec & 0x3; | 136 | u8 ll = LL(ec); |
83 | ret = true; | 137 | ret = true; |
84 | 138 | ||
85 | if (ll == LL_L2) | 139 | if (ll == LL_L2) |
86 | pr_cont("during L1 linefill from L2.\n"); | 140 | pr_cont("during L1 linefill from L2.\n"); |
87 | else if (ll == LL_L1) | 141 | else if (ll == LL_L1) |
88 | pr_cont("Data/Tag %s error.\n", RRRR_MSG(ec)); | 142 | pr_cont("Data/Tag %s error.\n", R4_MSG(ec)); |
89 | else | 143 | else |
90 | ret = false; | 144 | ret = false; |
91 | } | 145 | } |
92 | return ret; | 146 | return ret; |
93 | } | 147 | } |
94 | 148 | ||
95 | static bool f10h_dc_mce(u16 ec) | 149 | static bool f10h_dc_mce(u16 ec, u8 xec) |
96 | { | 150 | { |
97 | u8 r4 = (ec >> 4) & 0xf; | 151 | if (R4(ec) == R4_GEN && LL(ec) == LL_L1) { |
98 | u8 ll = ec & 0x3; | ||
99 | |||
100 | if (r4 == R4_GEN && ll == LL_L1) { | ||
101 | pr_cont("during data scrub.\n"); | 152 | pr_cont("during data scrub.\n"); |
102 | return true; | 153 | return true; |
103 | } | 154 | } |
104 | return f12h_dc_mce(ec); | 155 | return f12h_dc_mce(ec, xec); |
105 | } | 156 | } |
106 | 157 | ||
107 | static bool k8_dc_mce(u16 ec) | 158 | static bool k8_dc_mce(u16 ec, u8 xec) |
108 | { | 159 | { |
109 | if (BUS_ERROR(ec)) { | 160 | if (BUS_ERROR(ec)) { |
110 | pr_cont("during system linefill.\n"); | 161 | pr_cont("during system linefill.\n"); |
111 | return true; | 162 | return true; |
112 | } | 163 | } |
113 | 164 | ||
114 | return f10h_dc_mce(ec); | 165 | return f10h_dc_mce(ec, xec); |
115 | } | 166 | } |
116 | 167 | ||
117 | static bool f14h_dc_mce(u16 ec) | 168 | static bool f14h_dc_mce(u16 ec, u8 xec) |
118 | { | 169 | { |
119 | u8 r4 = (ec >> 4) & 0xf; | 170 | u8 r4 = R4(ec); |
120 | u8 ll = ec & 0x3; | ||
121 | u8 tt = (ec >> 2) & 0x3; | ||
122 | u8 ii = tt; | ||
123 | bool ret = true; | 171 | bool ret = true; |
124 | 172 | ||
125 | if (MEM_ERROR(ec)) { | 173 | if (MEM_ERROR(ec)) { |
126 | 174 | ||
127 | if (tt != TT_DATA || ll != LL_L1) | 175 | if (TT(ec) != TT_DATA || LL(ec) != LL_L1) |
128 | return false; | 176 | return false; |
129 | 177 | ||
130 | switch (r4) { | 178 | switch (r4) { |
@@ -144,7 +192,7 @@ static bool f14h_dc_mce(u16 ec) | |||
144 | } | 192 | } |
145 | } else if (BUS_ERROR(ec)) { | 193 | } else if (BUS_ERROR(ec)) { |
146 | 194 | ||
147 | if ((ii != II_MEM && ii != II_IO) || ll != LL_LG) | 195 | if ((II(ec) != II_MEM && II(ec) != II_IO) || LL(ec) != LL_LG) |
148 | return false; | 196 | return false; |
149 | 197 | ||
150 | pr_cont("System read data error on a "); | 198 | pr_cont("System read data error on a "); |
@@ -169,39 +217,78 @@ static bool f14h_dc_mce(u16 ec) | |||
169 | return ret; | 217 | return ret; |
170 | } | 218 | } |
171 | 219 | ||
220 | static bool f15h_dc_mce(u16 ec, u8 xec) | ||
221 | { | ||
222 | bool ret = true; | ||
223 | |||
224 | if (MEM_ERROR(ec)) { | ||
225 | |||
226 | switch (xec) { | ||
227 | case 0x0: | ||
228 | pr_cont("Data Array access error.\n"); | ||
229 | break; | ||
230 | |||
231 | case 0x1: | ||
232 | pr_cont("UC error during a linefill from L2/NB.\n"); | ||
233 | break; | ||
234 | |||
235 | case 0x2: | ||
236 | case 0x11: | ||
237 | pr_cont("STQ access error.\n"); | ||
238 | break; | ||
239 | |||
240 | case 0x3: | ||
241 | pr_cont("SCB access error.\n"); | ||
242 | break; | ||
243 | |||
244 | case 0x10: | ||
245 | pr_cont("Tag error.\n"); | ||
246 | break; | ||
247 | |||
248 | case 0x12: | ||
249 | pr_cont("LDQ access error.\n"); | ||
250 | break; | ||
251 | |||
252 | default: | ||
253 | ret = false; | ||
254 | } | ||
255 | } else if (BUS_ERROR(ec)) { | ||
256 | |||
257 | if (!xec) | ||
258 | pr_cont("during system linefill.\n"); | ||
259 | else | ||
260 | pr_cont(" Internal %s condition.\n", | ||
261 | ((xec == 1) ? "livelock" : "deadlock")); | ||
262 | } else | ||
263 | ret = false; | ||
264 | |||
265 | return ret; | ||
266 | } | ||
267 | |||
172 | static void amd_decode_dc_mce(struct mce *m) | 268 | static void amd_decode_dc_mce(struct mce *m) |
173 | { | 269 | { |
174 | u16 ec = m->status & 0xffff; | 270 | u16 ec = EC(m->status); |
175 | u8 xec = (m->status >> 16) & 0xf; | 271 | u8 xec = XEC(m->status, xec_mask); |
176 | 272 | ||
177 | pr_emerg(HW_ERR "Data Cache Error: "); | 273 | pr_emerg(HW_ERR "Data Cache Error: "); |
178 | 274 | ||
179 | /* TLB error signatures are the same across families */ | 275 | /* TLB error signatures are the same across families */ |
180 | if (TLB_ERROR(ec)) { | 276 | if (TLB_ERROR(ec)) { |
181 | u8 tt = (ec >> 2) & 0x3; | 277 | if (TT(ec) == TT_DATA) { |
182 | |||
183 | if (tt == TT_DATA) { | ||
184 | pr_cont("%s TLB %s.\n", LL_MSG(ec), | 278 | pr_cont("%s TLB %s.\n", LL_MSG(ec), |
185 | (xec ? "multimatch" : "parity error")); | 279 | ((xec == 2) ? "locked miss" |
280 | : (xec ? "multimatch" : "parity"))); | ||
186 | return; | 281 | return; |
187 | } | 282 | } |
188 | else | 283 | } else if (fam_ops->dc_mce(ec, xec)) |
189 | goto wrong_dc_mce; | 284 | ; |
190 | } | 285 | else |
191 | 286 | pr_emerg(HW_ERR "Corrupted DC MCE info?\n"); | |
192 | if (!fam_ops->dc_mce(ec)) | ||
193 | goto wrong_dc_mce; | ||
194 | |||
195 | return; | ||
196 | |||
197 | wrong_dc_mce: | ||
198 | pr_emerg(HW_ERR "Corrupted DC MCE info?\n"); | ||
199 | } | 287 | } |
200 | 288 | ||
201 | static bool k8_ic_mce(u16 ec) | 289 | static bool k8_ic_mce(u16 ec, u8 xec) |
202 | { | 290 | { |
203 | u8 ll = ec & 0x3; | 291 | u8 ll = LL(ec); |
204 | u8 r4 = (ec >> 4) & 0xf; | ||
205 | bool ret = true; | 292 | bool ret = true; |
206 | 293 | ||
207 | if (!MEM_ERROR(ec)) | 294 | if (!MEM_ERROR(ec)) |
@@ -210,7 +297,7 @@ static bool k8_ic_mce(u16 ec) | |||
210 | if (ll == 0x2) | 297 | if (ll == 0x2) |
211 | pr_cont("during a linefill from L2.\n"); | 298 | pr_cont("during a linefill from L2.\n"); |
212 | else if (ll == 0x1) { | 299 | else if (ll == 0x1) { |
213 | switch (r4) { | 300 | switch (R4(ec)) { |
214 | case R4_IRD: | 301 | case R4_IRD: |
215 | pr_cont("Parity error during data load.\n"); | 302 | pr_cont("Parity error during data load.\n"); |
216 | break; | 303 | break; |
@@ -233,15 +320,13 @@ static bool k8_ic_mce(u16 ec) | |||
233 | return ret; | 320 | return ret; |
234 | } | 321 | } |
235 | 322 | ||
236 | static bool f14h_ic_mce(u16 ec) | 323 | static bool f14h_ic_mce(u16 ec, u8 xec) |
237 | { | 324 | { |
238 | u8 ll = ec & 0x3; | 325 | u8 r4 = R4(ec); |
239 | u8 tt = (ec >> 2) & 0x3; | ||
240 | u8 r4 = (ec >> 4) & 0xf; | ||
241 | bool ret = true; | 326 | bool ret = true; |
242 | 327 | ||
243 | if (MEM_ERROR(ec)) { | 328 | if (MEM_ERROR(ec)) { |
244 | if (tt != 0 || ll != 1) | 329 | if (TT(ec) != 0 || LL(ec) != 1) |
245 | ret = false; | 330 | ret = false; |
246 | 331 | ||
247 | if (r4 == R4_IRD) | 332 | if (r4 == R4_IRD) |
@@ -254,10 +339,36 @@ static bool f14h_ic_mce(u16 ec) | |||
254 | return ret; | 339 | return ret; |
255 | } | 340 | } |
256 | 341 | ||
342 | static bool f15h_ic_mce(u16 ec, u8 xec) | ||
343 | { | ||
344 | bool ret = true; | ||
345 | |||
346 | if (!MEM_ERROR(ec)) | ||
347 | return false; | ||
348 | |||
349 | switch (xec) { | ||
350 | case 0x0 ... 0xa: | ||
351 | pr_cont("%s.\n", f15h_ic_mce_desc[xec]); | ||
352 | break; | ||
353 | |||
354 | case 0xd: | ||
355 | pr_cont("%s.\n", f15h_ic_mce_desc[xec-2]); | ||
356 | break; | ||
357 | |||
358 | case 0x10 ... 0x14: | ||
359 | pr_cont("Decoder %s parity error.\n", f15h_ic_mce_desc[xec-4]); | ||
360 | break; | ||
361 | |||
362 | default: | ||
363 | ret = false; | ||
364 | } | ||
365 | return ret; | ||
366 | } | ||
367 | |||
257 | static void amd_decode_ic_mce(struct mce *m) | 368 | static void amd_decode_ic_mce(struct mce *m) |
258 | { | 369 | { |
259 | u16 ec = m->status & 0xffff; | 370 | u16 ec = EC(m->status); |
260 | u8 xec = (m->status >> 16) & 0xf; | 371 | u8 xec = XEC(m->status, xec_mask); |
261 | 372 | ||
262 | pr_emerg(HW_ERR "Instruction Cache Error: "); | 373 | pr_emerg(HW_ERR "Instruction Cache Error: "); |
263 | 374 | ||
@@ -268,7 +379,7 @@ static void amd_decode_ic_mce(struct mce *m) | |||
268 | bool k8 = (boot_cpu_data.x86 == 0xf && (m->status & BIT_64(58))); | 379 | bool k8 = (boot_cpu_data.x86 == 0xf && (m->status & BIT_64(58))); |
269 | 380 | ||
270 | pr_cont("during %s.\n", (k8 ? "system linefill" : "NB data read")); | 381 | pr_cont("during %s.\n", (k8 ? "system linefill" : "NB data read")); |
271 | } else if (fam_ops->ic_mce(ec)) | 382 | } else if (fam_ops->ic_mce(ec, xec)) |
272 | ; | 383 | ; |
273 | else | 384 | else |
274 | pr_emerg(HW_ERR "Corrupted IC MCE info?\n"); | 385 | pr_emerg(HW_ERR "Corrupted IC MCE info?\n"); |
@@ -276,8 +387,8 @@ static void amd_decode_ic_mce(struct mce *m) | |||
276 | 387 | ||
277 | static void amd_decode_bu_mce(struct mce *m) | 388 | static void amd_decode_bu_mce(struct mce *m) |
278 | { | 389 | { |
279 | u32 ec = m->status & 0xffff; | 390 | u16 ec = EC(m->status); |
280 | u32 xec = (m->status >> 16) & 0xf; | 391 | u8 xec = XEC(m->status, xec_mask); |
281 | 392 | ||
282 | pr_emerg(HW_ERR "Bus Unit Error"); | 393 | pr_emerg(HW_ERR "Bus Unit Error"); |
283 | 394 | ||
@@ -286,23 +397,23 @@ static void amd_decode_bu_mce(struct mce *m) | |||
286 | else if (xec == 0x3) | 397 | else if (xec == 0x3) |
287 | pr_cont(" in the victim data buffers.\n"); | 398 | pr_cont(" in the victim data buffers.\n"); |
288 | else if (xec == 0x2 && MEM_ERROR(ec)) | 399 | else if (xec == 0x2 && MEM_ERROR(ec)) |
289 | pr_cont(": %s error in the L2 cache tags.\n", RRRR_MSG(ec)); | 400 | pr_cont(": %s error in the L2 cache tags.\n", R4_MSG(ec)); |
290 | else if (xec == 0x0) { | 401 | else if (xec == 0x0) { |
291 | if (TLB_ERROR(ec)) | 402 | if (TLB_ERROR(ec)) |
292 | pr_cont(": %s error in a Page Descriptor Cache or " | 403 | pr_cont(": %s error in a Page Descriptor Cache or " |
293 | "Guest TLB.\n", TT_MSG(ec)); | 404 | "Guest TLB.\n", TT_MSG(ec)); |
294 | else if (BUS_ERROR(ec)) | 405 | else if (BUS_ERROR(ec)) |
295 | pr_cont(": %s/ECC error in data read from NB: %s.\n", | 406 | pr_cont(": %s/ECC error in data read from NB: %s.\n", |
296 | RRRR_MSG(ec), PP_MSG(ec)); | 407 | R4_MSG(ec), PP_MSG(ec)); |
297 | else if (MEM_ERROR(ec)) { | 408 | else if (MEM_ERROR(ec)) { |
298 | u8 rrrr = (ec >> 4) & 0xf; | 409 | u8 r4 = R4(ec); |
299 | 410 | ||
300 | if (rrrr >= 0x7) | 411 | if (r4 >= 0x7) |
301 | pr_cont(": %s error during data copyback.\n", | 412 | pr_cont(": %s error during data copyback.\n", |
302 | RRRR_MSG(ec)); | 413 | R4_MSG(ec)); |
303 | else if (rrrr <= 0x1) | 414 | else if (r4 <= 0x1) |
304 | pr_cont(": %s parity/ECC error during data " | 415 | pr_cont(": %s parity/ECC error during data " |
305 | "access from L2.\n", RRRR_MSG(ec)); | 416 | "access from L2.\n", R4_MSG(ec)); |
306 | else | 417 | else |
307 | goto wrong_bu_mce; | 418 | goto wrong_bu_mce; |
308 | } else | 419 | } else |
@@ -316,12 +427,52 @@ wrong_bu_mce: | |||
316 | pr_emerg(HW_ERR "Corrupted BU MCE info?\n"); | 427 | pr_emerg(HW_ERR "Corrupted BU MCE info?\n"); |
317 | } | 428 | } |
318 | 429 | ||
430 | static void amd_decode_cu_mce(struct mce *m) | ||
431 | { | ||
432 | u16 ec = EC(m->status); | ||
433 | u8 xec = XEC(m->status, xec_mask); | ||
434 | |||
435 | pr_emerg(HW_ERR "Combined Unit Error: "); | ||
436 | |||
437 | if (TLB_ERROR(ec)) { | ||
438 | if (xec == 0x0) | ||
439 | pr_cont("Data parity TLB read error.\n"); | ||
440 | else if (xec == 0x1) | ||
441 | pr_cont("Poison data provided for TLB fill.\n"); | ||
442 | else | ||
443 | goto wrong_cu_mce; | ||
444 | } else if (BUS_ERROR(ec)) { | ||
445 | if (xec > 2) | ||
446 | goto wrong_cu_mce; | ||
447 | |||
448 | pr_cont("Error during attempted NB data read.\n"); | ||
449 | } else if (MEM_ERROR(ec)) { | ||
450 | switch (xec) { | ||
451 | case 0x4 ... 0xc: | ||
452 | pr_cont("%s.\n", f15h_cu_mce_desc[xec - 0x4]); | ||
453 | break; | ||
454 | |||
455 | case 0x10 ... 0x14: | ||
456 | pr_cont("%s.\n", f15h_cu_mce_desc[xec - 0x7]); | ||
457 | break; | ||
458 | |||
459 | default: | ||
460 | goto wrong_cu_mce; | ||
461 | } | ||
462 | } | ||
463 | |||
464 | return; | ||
465 | |||
466 | wrong_cu_mce: | ||
467 | pr_emerg(HW_ERR "Corrupted CU MCE info?\n"); | ||
468 | } | ||
469 | |||
319 | static void amd_decode_ls_mce(struct mce *m) | 470 | static void amd_decode_ls_mce(struct mce *m) |
320 | { | 471 | { |
321 | u16 ec = m->status & 0xffff; | 472 | u16 ec = EC(m->status); |
322 | u8 xec = (m->status >> 16) & 0xf; | 473 | u8 xec = XEC(m->status, xec_mask); |
323 | 474 | ||
324 | if (boot_cpu_data.x86 == 0x14) { | 475 | if (boot_cpu_data.x86 >= 0x14) { |
325 | pr_emerg("You shouldn't be seeing an LS MCE on this cpu family," | 476 | pr_emerg("You shouldn't be seeing an LS MCE on this cpu family," |
326 | " please report on LKML.\n"); | 477 | " please report on LKML.\n"); |
327 | return; | 478 | return; |
@@ -330,12 +481,12 @@ static void amd_decode_ls_mce(struct mce *m) | |||
330 | pr_emerg(HW_ERR "Load Store Error"); | 481 | pr_emerg(HW_ERR "Load Store Error"); |
331 | 482 | ||
332 | if (xec == 0x0) { | 483 | if (xec == 0x0) { |
333 | u8 r4 = (ec >> 4) & 0xf; | 484 | u8 r4 = R4(ec); |
334 | 485 | ||
335 | if (!BUS_ERROR(ec) || (r4 != R4_DRD && r4 != R4_DWR)) | 486 | if (!BUS_ERROR(ec) || (r4 != R4_DRD && r4 != R4_DWR)) |
336 | goto wrong_ls_mce; | 487 | goto wrong_ls_mce; |
337 | 488 | ||
338 | pr_cont(" during %s.\n", RRRR_MSG(ec)); | 489 | pr_cont(" during %s.\n", R4_MSG(ec)); |
339 | } else | 490 | } else |
340 | goto wrong_ls_mce; | 491 | goto wrong_ls_mce; |
341 | 492 | ||
@@ -410,6 +561,15 @@ static bool f10h_nb_mce(u16 ec, u8 xec) | |||
410 | goto out; | 561 | goto out; |
411 | break; | 562 | break; |
412 | 563 | ||
564 | case 0x19: | ||
565 | if (boot_cpu_data.x86 == 0x15) | ||
566 | pr_cont("Compute Unit Data Error.\n"); | ||
567 | else | ||
568 | ret = false; | ||
569 | |||
570 | goto out; | ||
571 | break; | ||
572 | |||
413 | case 0x1c ... 0x1f: | 573 | case 0x1c ... 0x1f: |
414 | offset = 24; | 574 | offset = 24; |
415 | break; | 575 | break; |
@@ -434,27 +594,30 @@ static bool nb_noop_mce(u16 ec, u8 xec) | |||
434 | 594 | ||
435 | void amd_decode_nb_mce(int node_id, struct mce *m, u32 nbcfg) | 595 | void amd_decode_nb_mce(int node_id, struct mce *m, u32 nbcfg) |
436 | { | 596 | { |
437 | u8 xec = (m->status >> 16) & 0x1f; | 597 | u16 ec = EC(m->status); |
438 | u16 ec = m->status & 0xffff; | 598 | u8 xec = XEC(m->status, 0x1f); |
439 | u32 nbsh = (u32)(m->status >> 32); | 599 | u32 nbsh = (u32)(m->status >> 32); |
600 | int core = -1; | ||
440 | 601 | ||
441 | pr_emerg(HW_ERR "Northbridge Error, node %d: ", node_id); | 602 | pr_emerg(HW_ERR "Northbridge Error (node %d", node_id); |
442 | 603 | ||
443 | /* | 604 | /* F10h, revD can disable ErrCpu[3:0] through ErrCpuVal */ |
444 | * F10h, revD can disable ErrCpu[3:0] so check that first and also the | ||
445 | * value encoding has changed so interpret those differently | ||
446 | */ | ||
447 | if ((boot_cpu_data.x86 == 0x10) && | 605 | if ((boot_cpu_data.x86 == 0x10) && |
448 | (boot_cpu_data.x86_model > 7)) { | 606 | (boot_cpu_data.x86_model > 7)) { |
449 | if (nbsh & K8_NBSH_ERR_CPU_VAL) | 607 | if (nbsh & K8_NBSH_ERR_CPU_VAL) |
450 | pr_cont(", core: %u", (u8)(nbsh & nb_err_cpumask)); | 608 | core = nbsh & nb_err_cpumask; |
451 | } else { | 609 | } else { |
452 | u8 assoc_cpus = nbsh & nb_err_cpumask; | 610 | u8 assoc_cpus = nbsh & nb_err_cpumask; |
453 | 611 | ||
454 | if (assoc_cpus > 0) | 612 | if (assoc_cpus > 0) |
455 | pr_cont(", core: %d", fls(assoc_cpus) - 1); | 613 | core = fls(assoc_cpus) - 1; |
456 | } | 614 | } |
457 | 615 | ||
616 | if (core >= 0) | ||
617 | pr_cont(", core %d): ", core); | ||
618 | else | ||
619 | pr_cont("): "); | ||
620 | |||
458 | switch (xec) { | 621 | switch (xec) { |
459 | case 0x2: | 622 | case 0x2: |
460 | pr_cont("Sync error (sync packets on HT link detected).\n"); | 623 | pr_cont("Sync error (sync packets on HT link detected).\n"); |
@@ -496,35 +659,89 @@ EXPORT_SYMBOL_GPL(amd_decode_nb_mce); | |||
496 | 659 | ||
497 | static void amd_decode_fr_mce(struct mce *m) | 660 | static void amd_decode_fr_mce(struct mce *m) |
498 | { | 661 | { |
499 | if (boot_cpu_data.x86 == 0xf || | 662 | struct cpuinfo_x86 *c = &boot_cpu_data; |
500 | boot_cpu_data.x86 == 0x11) | 663 | u8 xec = XEC(m->status, xec_mask); |
664 | |||
665 | if (c->x86 == 0xf || c->x86 == 0x11) | ||
501 | goto wrong_fr_mce; | 666 | goto wrong_fr_mce; |
502 | 667 | ||
503 | /* we have only one error signature so match all fields at once. */ | 668 | if (c->x86 != 0x15 && xec != 0x0) |
504 | if ((m->status & 0xffff) == 0x0f0f) { | 669 | goto wrong_fr_mce; |
505 | pr_emerg(HW_ERR "FR Error: CPU Watchdog timer expire.\n"); | 670 | |
506 | return; | 671 | pr_emerg(HW_ERR "%s Error: ", |
507 | } | 672 | (c->x86 == 0x15 ? "Execution Unit" : "FIROB")); |
673 | |||
674 | if (xec == 0x0 || xec == 0xc) | ||
675 | pr_cont("%s.\n", fr_ex_mce_desc[xec]); | ||
676 | else if (xec < 0xd) | ||
677 | pr_cont("%s parity error.\n", fr_ex_mce_desc[xec]); | ||
678 | else | ||
679 | goto wrong_fr_mce; | ||
680 | |||
681 | return; | ||
508 | 682 | ||
509 | wrong_fr_mce: | 683 | wrong_fr_mce: |
510 | pr_emerg(HW_ERR "Corrupted FR MCE info?\n"); | 684 | pr_emerg(HW_ERR "Corrupted FR MCE info?\n"); |
511 | } | 685 | } |
512 | 686 | ||
687 | static void amd_decode_fp_mce(struct mce *m) | ||
688 | { | ||
689 | u8 xec = XEC(m->status, xec_mask); | ||
690 | |||
691 | pr_emerg(HW_ERR "Floating Point Unit Error: "); | ||
692 | |||
693 | switch (xec) { | ||
694 | case 0x1: | ||
695 | pr_cont("Free List"); | ||
696 | break; | ||
697 | |||
698 | case 0x2: | ||
699 | pr_cont("Physical Register File"); | ||
700 | break; | ||
701 | |||
702 | case 0x3: | ||
703 | pr_cont("Retire Queue"); | ||
704 | break; | ||
705 | |||
706 | case 0x4: | ||
707 | pr_cont("Scheduler table"); | ||
708 | break; | ||
709 | |||
710 | case 0x5: | ||
711 | pr_cont("Status Register File"); | ||
712 | break; | ||
713 | |||
714 | default: | ||
715 | goto wrong_fp_mce; | ||
716 | break; | ||
717 | } | ||
718 | |||
719 | pr_cont(" parity error.\n"); | ||
720 | |||
721 | return; | ||
722 | |||
723 | wrong_fp_mce: | ||
724 | pr_emerg(HW_ERR "Corrupted FP MCE info?\n"); | ||
725 | } | ||
726 | |||
513 | static inline void amd_decode_err_code(u16 ec) | 727 | static inline void amd_decode_err_code(u16 ec) |
514 | { | 728 | { |
515 | if (TLB_ERROR(ec)) { | 729 | |
516 | pr_emerg(HW_ERR "Transaction: %s, Cache Level: %s\n", | 730 | pr_emerg(HW_ERR "cache level: %s", LL_MSG(ec)); |
517 | TT_MSG(ec), LL_MSG(ec)); | 731 | |
518 | } else if (MEM_ERROR(ec)) { | 732 | if (BUS_ERROR(ec)) |
519 | pr_emerg(HW_ERR "Transaction: %s, Type: %s, Cache Level: %s\n", | 733 | pr_cont(", mem/io: %s", II_MSG(ec)); |
520 | RRRR_MSG(ec), TT_MSG(ec), LL_MSG(ec)); | 734 | else |
521 | } else if (BUS_ERROR(ec)) { | 735 | pr_cont(", tx: %s", TT_MSG(ec)); |
522 | pr_emerg(HW_ERR "Transaction: %s (%s), %s, Cache Level: %s, " | 736 | |
523 | "Participating Processor: %s\n", | 737 | if (MEM_ERROR(ec) || BUS_ERROR(ec)) { |
524 | RRRR_MSG(ec), II_MSG(ec), TO_MSG(ec), LL_MSG(ec), | 738 | pr_cont(", mem-tx: %s", R4_MSG(ec)); |
525 | PP_MSG(ec)); | 739 | |
526 | } else | 740 | if (BUS_ERROR(ec)) |
527 | pr_emerg(HW_ERR "Huh? Unknown MCE error 0x%x\n", ec); | 741 | pr_cont(", part-proc: %s (%s)", PP_MSG(ec), TO_MSG(ec)); |
742 | } | ||
743 | |||
744 | pr_cont("\n"); | ||
528 | } | 745 | } |
529 | 746 | ||
530 | /* | 747 | /* |
@@ -546,25 +763,32 @@ static bool amd_filter_mce(struct mce *m) | |||
546 | int amd_decode_mce(struct notifier_block *nb, unsigned long val, void *data) | 763 | int amd_decode_mce(struct notifier_block *nb, unsigned long val, void *data) |
547 | { | 764 | { |
548 | struct mce *m = (struct mce *)data; | 765 | struct mce *m = (struct mce *)data; |
766 | struct cpuinfo_x86 *c = &boot_cpu_data; | ||
549 | int node, ecc; | 767 | int node, ecc; |
550 | 768 | ||
551 | if (amd_filter_mce(m)) | 769 | if (amd_filter_mce(m)) |
552 | return NOTIFY_STOP; | 770 | return NOTIFY_STOP; |
553 | 771 | ||
554 | pr_emerg(HW_ERR "MC%d_STATUS: ", m->bank); | 772 | pr_emerg(HW_ERR "MC%d_STATUS[%s|%s|%s|%s|%s", |
773 | m->bank, | ||
774 | ((m->status & MCI_STATUS_OVER) ? "Over" : "-"), | ||
775 | ((m->status & MCI_STATUS_UC) ? "UE" : "CE"), | ||
776 | ((m->status & MCI_STATUS_MISCV) ? "MiscV" : "-"), | ||
777 | ((m->status & MCI_STATUS_PCC) ? "PCC" : "-"), | ||
778 | ((m->status & MCI_STATUS_ADDRV) ? "AddrV" : "-")); | ||
555 | 779 | ||
556 | pr_cont("%sorrected error, other errors lost: %s, " | 780 | if (c->x86 == 0x15) |
557 | "CPU context corrupt: %s", | 781 | pr_cont("|%s|%s", |
558 | ((m->status & MCI_STATUS_UC) ? "Unc" : "C"), | 782 | ((m->status & BIT_64(44)) ? "Deferred" : "-"), |
559 | ((m->status & MCI_STATUS_OVER) ? "yes" : "no"), | 783 | ((m->status & BIT_64(43)) ? "Poison" : "-")); |
560 | ((m->status & MCI_STATUS_PCC) ? "yes" : "no")); | ||
561 | 784 | ||
562 | /* do the two bits[14:13] together */ | 785 | /* do the two bits[14:13] together */ |
563 | ecc = (m->status >> 45) & 0x3; | 786 | ecc = (m->status >> 45) & 0x3; |
564 | if (ecc) | 787 | if (ecc) |
565 | pr_cont(", %sECC Error", ((ecc == 2) ? "C" : "U")); | 788 | pr_cont("|%sECC", ((ecc == 2) ? "C" : "U")); |
789 | |||
790 | pr_cont("]: 0x%016llx\n", m->status); | ||
566 | 791 | ||
567 | pr_cont("\n"); | ||
568 | 792 | ||
569 | switch (m->bank) { | 793 | switch (m->bank) { |
570 | case 0: | 794 | case 0: |
@@ -576,7 +800,10 @@ int amd_decode_mce(struct notifier_block *nb, unsigned long val, void *data) | |||
576 | break; | 800 | break; |
577 | 801 | ||
578 | case 2: | 802 | case 2: |
579 | amd_decode_bu_mce(m); | 803 | if (c->x86 == 0x15) |
804 | amd_decode_cu_mce(m); | ||
805 | else | ||
806 | amd_decode_bu_mce(m); | ||
580 | break; | 807 | break; |
581 | 808 | ||
582 | case 3: | 809 | case 3: |
@@ -592,6 +819,10 @@ int amd_decode_mce(struct notifier_block *nb, unsigned long val, void *data) | |||
592 | amd_decode_fr_mce(m); | 819 | amd_decode_fr_mce(m); |
593 | break; | 820 | break; |
594 | 821 | ||
822 | case 6: | ||
823 | amd_decode_fp_mce(m); | ||
824 | break; | ||
825 | |||
595 | default: | 826 | default: |
596 | break; | 827 | break; |
597 | } | 828 | } |
@@ -608,18 +839,21 @@ static struct notifier_block amd_mce_dec_nb = { | |||
608 | 839 | ||
609 | static int __init mce_amd_init(void) | 840 | static int __init mce_amd_init(void) |
610 | { | 841 | { |
611 | if (boot_cpu_data.x86_vendor != X86_VENDOR_AMD) | 842 | struct cpuinfo_x86 *c = &boot_cpu_data; |
843 | |||
844 | if (c->x86_vendor != X86_VENDOR_AMD) | ||
612 | return 0; | 845 | return 0; |
613 | 846 | ||
614 | if ((boot_cpu_data.x86 < 0xf || boot_cpu_data.x86 > 0x12) && | 847 | if ((c->x86 < 0xf || c->x86 > 0x12) && |
615 | (boot_cpu_data.x86 != 0x14 || boot_cpu_data.x86_model > 0xf)) | 848 | (c->x86 != 0x14 || c->x86_model > 0xf) && |
849 | (c->x86 != 0x15 || c->x86_model > 0xf)) | ||
616 | return 0; | 850 | return 0; |
617 | 851 | ||
618 | fam_ops = kzalloc(sizeof(struct amd_decoder_ops), GFP_KERNEL); | 852 | fam_ops = kzalloc(sizeof(struct amd_decoder_ops), GFP_KERNEL); |
619 | if (!fam_ops) | 853 | if (!fam_ops) |
620 | return -ENOMEM; | 854 | return -ENOMEM; |
621 | 855 | ||
622 | switch (boot_cpu_data.x86) { | 856 | switch (c->x86) { |
623 | case 0xf: | 857 | case 0xf: |
624 | fam_ops->dc_mce = k8_dc_mce; | 858 | fam_ops->dc_mce = k8_dc_mce; |
625 | fam_ops->ic_mce = k8_ic_mce; | 859 | fam_ops->ic_mce = k8_ic_mce; |
@@ -651,9 +885,15 @@ static int __init mce_amd_init(void) | |||
651 | fam_ops->nb_mce = nb_noop_mce; | 885 | fam_ops->nb_mce = nb_noop_mce; |
652 | break; | 886 | break; |
653 | 887 | ||
888 | case 0x15: | ||
889 | xec_mask = 0x1f; | ||
890 | fam_ops->dc_mce = f15h_dc_mce; | ||
891 | fam_ops->ic_mce = f15h_ic_mce; | ||
892 | fam_ops->nb_mce = f10h_nb_mce; | ||
893 | break; | ||
894 | |||
654 | default: | 895 | default: |
655 | printk(KERN_WARNING "Huh? What family is that: %d?!\n", | 896 | printk(KERN_WARNING "Huh? What family is that: %d?!\n", c->x86); |
656 | boot_cpu_data.x86); | ||
657 | kfree(fam_ops); | 897 | kfree(fam_ops); |
658 | return -EINVAL; | 898 | return -EINVAL; |
659 | } | 899 | } |
diff --git a/drivers/edac/mce_amd.h b/drivers/edac/mce_amd.h index 35f6e0e3b297..45dda47173f2 100644 --- a/drivers/edac/mce_amd.h +++ b/drivers/edac/mce_amd.h | |||
@@ -7,8 +7,8 @@ | |||
7 | 7 | ||
8 | #define BIT_64(n) (U64_C(1) << (n)) | 8 | #define BIT_64(n) (U64_C(1) << (n)) |
9 | 9 | ||
10 | #define ERROR_CODE(x) ((x) & 0xffff) | 10 | #define EC(x) ((x) & 0xffff) |
11 | #define EXT_ERROR_CODE(x) (((x) >> 16) & 0x1f) | 11 | #define XEC(x, mask) (((x) >> 16) & mask) |
12 | 12 | ||
13 | #define LOW_SYNDROME(x) (((x) >> 15) & 0xff) | 13 | #define LOW_SYNDROME(x) (((x) >> 15) & 0xff) |
14 | #define HIGH_SYNDROME(x) (((x) >> 24) & 0xff) | 14 | #define HIGH_SYNDROME(x) (((x) >> 24) & 0xff) |
@@ -21,15 +21,15 @@ | |||
21 | #define TT_MSG(x) tt_msgs[TT(x)] | 21 | #define TT_MSG(x) tt_msgs[TT(x)] |
22 | #define II(x) (((x) >> 2) & 0x3) | 22 | #define II(x) (((x) >> 2) & 0x3) |
23 | #define II_MSG(x) ii_msgs[II(x)] | 23 | #define II_MSG(x) ii_msgs[II(x)] |
24 | #define LL(x) (((x) >> 0) & 0x3) | 24 | #define LL(x) ((x) & 0x3) |
25 | #define LL_MSG(x) ll_msgs[LL(x)] | 25 | #define LL_MSG(x) ll_msgs[LL(x)] |
26 | #define TO(x) (((x) >> 8) & 0x1) | 26 | #define TO(x) (((x) >> 8) & 0x1) |
27 | #define TO_MSG(x) to_msgs[TO(x)] | 27 | #define TO_MSG(x) to_msgs[TO(x)] |
28 | #define PP(x) (((x) >> 9) & 0x3) | 28 | #define PP(x) (((x) >> 9) & 0x3) |
29 | #define PP_MSG(x) pp_msgs[PP(x)] | 29 | #define PP_MSG(x) pp_msgs[PP(x)] |
30 | 30 | ||
31 | #define RRRR(x) (((x) >> 4) & 0xf) | 31 | #define R4(x) (((x) >> 4) & 0xf) |
32 | #define RRRR_MSG(x) ((RRRR(x) < 9) ? rrrr_msgs[RRRR(x)] : "Wrong R4!") | 32 | #define R4_MSG(x) ((R4(x) < 9) ? rrrr_msgs[R4(x)] : "Wrong R4!") |
33 | 33 | ||
34 | #define K8_NBSH 0x4C | 34 | #define K8_NBSH 0x4C |
35 | 35 | ||
@@ -100,8 +100,8 @@ struct err_regs { | |||
100 | * per-family decoder ops | 100 | * per-family decoder ops |
101 | */ | 101 | */ |
102 | struct amd_decoder_ops { | 102 | struct amd_decoder_ops { |
103 | bool (*dc_mce)(u16); | 103 | bool (*dc_mce)(u16, u8); |
104 | bool (*ic_mce)(u16); | 104 | bool (*ic_mce)(u16, u8); |
105 | bool (*nb_mce)(u16, u8); | 105 | bool (*nb_mce)(u16, u8); |
106 | }; | 106 | }; |
107 | 107 | ||
diff --git a/drivers/edac/mce_amd_inj.c b/drivers/edac/mce_amd_inj.c index 39faded3cadd..733a7e7a8d6f 100644 --- a/drivers/edac/mce_amd_inj.c +++ b/drivers/edac/mce_amd_inj.c | |||
@@ -88,10 +88,11 @@ static ssize_t edac_inject_bank_store(struct kobject *kobj, | |||
88 | return -EINVAL; | 88 | return -EINVAL; |
89 | } | 89 | } |
90 | 90 | ||
91 | if (value > 5) { | 91 | if (value > 5) |
92 | printk(KERN_ERR "Non-existant MCE bank: %lu\n", value); | 92 | if (boot_cpu_data.x86 != 0x15 || value > 6) { |
93 | return -EINVAL; | 93 | printk(KERN_ERR "Non-existant MCE bank: %lu\n", value); |
94 | } | 94 | return -EINVAL; |
95 | } | ||
95 | 96 | ||
96 | i_mce.bank = value; | 97 | i_mce.bank = value; |
97 | 98 | ||