1 #undef TRACE_SYSTEM 2 #define TRACE_SYSTEM ras 3 #define TRACE_INCLUDE_FILE ras_event 4 5 #if !defined(_TRACE_HW_EVENT_MC_H) || defined(TRACE_HEADER_MULTI_READ) 6 #define _TRACE_HW_EVENT_MC_H 7 8 #include <linux/tracepoint.h> 9 #include <linux/edac.h> 10 #include <linux/ktime.h> 11 #include <linux/pci.h> 12 #include <linux/aer.h> 13 #include <linux/cper.h> 14 #include <linux/mm.h> 15 16 /* 17 * MCE Extended Error Log trace event 18 * 19 * These events are generated when hardware detects a corrected or 20 * uncorrected event. 21 */ 22 23 /* memory trace event */ 24 25 #if defined(CONFIG_ACPI_EXTLOG) || defined(CONFIG_ACPI_EXTLOG_MODULE) 26 TRACE_EVENT(extlog_mem_event, 27 TP_PROTO(struct cper_sec_mem_err *mem, 28 u32 err_seq, 29 const uuid_le *fru_id, 30 const char *fru_text, 31 u8 sev), 32 33 TP_ARGS(mem, err_seq, fru_id, fru_text, sev), 34 35 TP_STRUCT__entry( 36 __field(u32, err_seq) 37 __field(u8, etype) 38 __field(u8, sev) 39 __field(u64, pa) 40 __field(u8, pa_mask_lsb) 41 __field_struct(uuid_le, fru_id) 42 __string(fru_text, fru_text) 43 __field_struct(struct cper_mem_err_compact, data) 44 ), 45 46 TP_fast_assign( 47 __entry->err_seq = err_seq; 48 if (mem->validation_bits & CPER_MEM_VALID_ERROR_TYPE) 49 __entry->etype = mem->error_type; 50 else 51 __entry->etype = ~0; 52 __entry->sev = sev; 53 if (mem->validation_bits & CPER_MEM_VALID_PA) 54 __entry->pa = mem->physical_addr; 55 else 56 __entry->pa = ~0ull; 57 58 if (mem->validation_bits & CPER_MEM_VALID_PA_MASK) 59 __entry->pa_mask_lsb = (u8)__ffs64(mem->physical_addr_mask); 60 else 61 __entry->pa_mask_lsb = ~0; 62 __entry->fru_id = *fru_id; 63 __assign_str(fru_text, fru_text); 64 cper_mem_err_pack(mem, &__entry->data); 65 ), 66 67 TP_printk("{%d} %s error: %s physical addr: %016llx (mask lsb: %x) %sFRU: %pUl %.20s", 68 __entry->err_seq, 69 cper_severity_str(__entry->sev), 70 cper_mem_err_type_str(__entry->etype), 71 __entry->pa, 72 __entry->pa_mask_lsb, 73 cper_mem_err_unpack(p, &__entry->data), 74 &__entry->fru_id, 75 __get_str(fru_text)) 76 ); 77 #endif 78 79 /* 80 * Hardware Events Report 81 * 82 * Those events are generated when hardware detected a corrected or 83 * uncorrected event, and are meant to replace the current API to report 84 * errors defined on both EDAC and MCE subsystems. 85 * 86 * FIXME: Add events for handling memory errors originated from the 87 * MCE subsystem. 88 */ 89 90 /* 91 * Hardware-independent Memory Controller specific events 92 */ 93 94 /* 95 * Default error mechanisms for Memory Controller errors (CE and UE) 96 */ 97 TRACE_EVENT(mc_event, 98 99 TP_PROTO(const unsigned int err_type, 100 const char *error_msg, 101 const char *label, 102 const int error_count, 103 const u8 mc_index, 104 const s8 top_layer, 105 const s8 mid_layer, 106 const s8 low_layer, 107 unsigned long address, 108 const u8 grain_bits, 109 unsigned long syndrome, 110 const char *driver_detail), 111 112 TP_ARGS(err_type, error_msg, label, error_count, mc_index, 113 top_layer, mid_layer, low_layer, address, grain_bits, 114 syndrome, driver_detail), 115 116 TP_STRUCT__entry( 117 __field( unsigned int, error_type ) 118 __string( msg, error_msg ) 119 __string( label, label ) 120 __field( u16, error_count ) 121 __field( u8, mc_index ) 122 __field( s8, top_layer ) 123 __field( s8, middle_layer ) 124 __field( s8, lower_layer ) 125 __field( long, address ) 126 __field( u8, grain_bits ) 127 __field( long, syndrome ) 128 __string( driver_detail, driver_detail ) 129 ), 130 131 TP_fast_assign( 132 __entry->error_type = err_type; 133 __assign_str(msg, error_msg); 134 __assign_str(label, label); 135 __entry->error_count = error_count; 136 __entry->mc_index = mc_index; 137 __entry->top_layer = top_layer; 138 __entry->middle_layer = mid_layer; 139 __entry->lower_layer = low_layer; 140 __entry->address = address; 141 __entry->grain_bits = grain_bits; 142 __entry->syndrome = syndrome; 143 __assign_str(driver_detail, driver_detail); 144 ), 145 146 TP_printk("%d %s error%s:%s%s on %s (mc:%d location:%d:%d:%d address:0x%08lx grain:%d syndrome:0x%08lx%s%s)", 147 __entry->error_count, 148 mc_event_error_type(__entry->error_type), 149 __entry->error_count > 1 ? "s" : "", 150 __get_str(msg)[0] ? " " : "", 151 __get_str(msg), 152 __get_str(label), 153 __entry->mc_index, 154 __entry->top_layer, 155 __entry->middle_layer, 156 __entry->lower_layer, 157 __entry->address, 158 1 << __entry->grain_bits, 159 __entry->syndrome, 160 __get_str(driver_detail)[0] ? " " : "", 161 __get_str(driver_detail)) 162 ); 163 164 /* 165 * PCIe AER Trace event 166 * 167 * These events are generated when hardware detects a corrected or 168 * uncorrected event on a PCIe device. The event report has 169 * the following structure: 170 * 171 * char * dev_name - The name of the slot where the device resides 172 * ([domain:]bus:device.function). 173 * u32 status - Either the correctable or uncorrectable register 174 * indicating what error or errors have been seen 175 * u8 severity - error severity 0:NONFATAL 1:FATAL 2:CORRECTED 176 */ 177 178 #define aer_correctable_errors \ 179 {PCI_ERR_COR_RCVR, "Receiver Error"}, \ 180 {PCI_ERR_COR_BAD_TLP, "Bad TLP"}, \ 181 {PCI_ERR_COR_BAD_DLLP, "Bad DLLP"}, \ 182 {PCI_ERR_COR_REP_ROLL, "RELAY_NUM Rollover"}, \ 183 {PCI_ERR_COR_REP_TIMER, "Replay Timer Timeout"}, \ 184 {PCI_ERR_COR_ADV_NFAT, "Advisory Non-Fatal Error"}, \ 185 {PCI_ERR_COR_INTERNAL, "Corrected Internal Error"}, \ 186 {PCI_ERR_COR_LOG_OVER, "Header Log Overflow"} 187 188 #define aer_uncorrectable_errors \ 189 {PCI_ERR_UNC_UND, "Undefined"}, \ 190 {PCI_ERR_UNC_DLP, "Data Link Protocol Error"}, \ 191 {PCI_ERR_UNC_SURPDN, "Surprise Down Error"}, \ 192 {PCI_ERR_UNC_POISON_TLP,"Poisoned TLP"}, \ 193 {PCI_ERR_UNC_FCP, "Flow Control Protocol Error"}, \ 194 {PCI_ERR_UNC_COMP_TIME, "Completion Timeout"}, \ 195 {PCI_ERR_UNC_COMP_ABORT,"Completer Abort"}, \ 196 {PCI_ERR_UNC_UNX_COMP, "Unexpected Completion"}, \ 197 {PCI_ERR_UNC_RX_OVER, "Receiver Overflow"}, \ 198 {PCI_ERR_UNC_MALF_TLP, "Malformed TLP"}, \ 199 {PCI_ERR_UNC_ECRC, "ECRC Error"}, \ 200 {PCI_ERR_UNC_UNSUP, "Unsupported Request Error"}, \ 201 {PCI_ERR_UNC_ACSV, "ACS Violation"}, \ 202 {PCI_ERR_UNC_INTN, "Uncorrectable Internal Error"},\ 203 {PCI_ERR_UNC_MCBTLP, "MC Blocked TLP"}, \ 204 {PCI_ERR_UNC_ATOMEG, "AtomicOp Egress Blocked"}, \ 205 {PCI_ERR_UNC_TLPPRE, "TLP Prefix Blocked Error"} 206 207 TRACE_EVENT(aer_event, 208 TP_PROTO(const char *dev_name, 209 const u32 status, 210 const u8 severity), 211 212 TP_ARGS(dev_name, status, severity), 213 214 TP_STRUCT__entry( 215 __string( dev_name, dev_name ) 216 __field( u32, status ) 217 __field( u8, severity ) 218 ), 219 220 TP_fast_assign( 221 __assign_str(dev_name, dev_name); 222 __entry->status = status; 223 __entry->severity = severity; 224 ), 225 226 TP_printk("%s PCIe Bus Error: severity=%s, %s\n", 227 __get_str(dev_name), 228 __entry->severity == AER_CORRECTABLE ? "Corrected" : 229 __entry->severity == AER_FATAL ? 230 "Fatal" : "Uncorrected, non-fatal", 231 __entry->severity == AER_CORRECTABLE ? 232 __print_flags(__entry->status, "|", aer_correctable_errors) : 233 __print_flags(__entry->status, "|", aer_uncorrectable_errors)) 234 ); 235 236 /* 237 * memory-failure recovery action result event 238 * 239 * unsigned long pfn - Page Frame Number of the corrupted page 240 * int type - Page types of the corrupted page 241 * int result - Result of recovery action 242 */ 243 244 #ifdef CONFIG_MEMORY_FAILURE 245 #define MF_ACTION_RESULT \ 246 EM ( MF_IGNORED, "Ignored" ) \ 247 EM ( MF_FAILED, "Failed" ) \ 248 EM ( MF_DELAYED, "Delayed" ) \ 249 EMe ( MF_RECOVERED, "Recovered" ) 250 251 #define MF_PAGE_TYPE \ 252 EM ( MF_MSG_KERNEL, "reserved kernel page" ) \ 253 EM ( MF_MSG_KERNEL_HIGH_ORDER, "high-order kernel page" ) \ 254 EM ( MF_MSG_SLAB, "kernel slab page" ) \ 255 EM ( MF_MSG_DIFFERENT_COMPOUND, "different compound page after locking" ) \ 256 EM ( MF_MSG_POISONED_HUGE, "huge page already hardware poisoned" ) \ 257 EM ( MF_MSG_HUGE, "huge page" ) \ 258 EM ( MF_MSG_FREE_HUGE, "free huge page" ) \ 259 EM ( MF_MSG_UNMAP_FAILED, "unmapping failed page" ) \ 260 EM ( MF_MSG_DIRTY_SWAPCACHE, "dirty swapcache page" ) \ 261 EM ( MF_MSG_CLEAN_SWAPCACHE, "clean swapcache page" ) \ 262 EM ( MF_MSG_DIRTY_MLOCKED_LRU, "dirty mlocked LRU page" ) \ 263 EM ( MF_MSG_CLEAN_MLOCKED_LRU, "clean mlocked LRU page" ) \ 264 EM ( MF_MSG_DIRTY_UNEVICTABLE_LRU, "dirty unevictable LRU page" ) \ 265 EM ( MF_MSG_CLEAN_UNEVICTABLE_LRU, "clean unevictable LRU page" ) \ 266 EM ( MF_MSG_DIRTY_LRU, "dirty LRU page" ) \ 267 EM ( MF_MSG_CLEAN_LRU, "clean LRU page" ) \ 268 EM ( MF_MSG_TRUNCATED_LRU, "already truncated LRU page" ) \ 269 EM ( MF_MSG_BUDDY, "free buddy page" ) \ 270 EM ( MF_MSG_BUDDY_2ND, "free buddy page (2nd try)" ) \ 271 EMe ( MF_MSG_UNKNOWN, "unknown page" ) 272 273 /* 274 * First define the enums in MM_ACTION_RESULT to be exported to userspace 275 * via TRACE_DEFINE_ENUM(). 276 */ 277 #undef EM 278 #undef EMe 279 #define EM(a, b) TRACE_DEFINE_ENUM(a); 280 #define EMe(a, b) TRACE_DEFINE_ENUM(a); 281 282 MF_ACTION_RESULT 283 MF_PAGE_TYPE 284 285 /* 286 * Now redefine the EM() and EMe() macros to map the enums to the strings 287 * that will be printed in the output. 288 */ 289 #undef EM 290 #undef EMe 291 #define EM(a, b) { a, b }, 292 #define EMe(a, b) { a, b } 293 294 TRACE_EVENT(memory_failure_event, 295 TP_PROTO(unsigned long pfn, 296 int type, 297 int result), 298 299 TP_ARGS(pfn, type, result), 300 301 TP_STRUCT__entry( 302 __field(unsigned long, pfn) 303 __field(int, type) 304 __field(int, result) 305 ), 306 307 TP_fast_assign( 308 __entry->pfn = pfn; 309 __entry->type = type; 310 __entry->result = result; 311 ), 312 313 TP_printk("pfn %#lx: recovery action for %s: %s", 314 __entry->pfn, 315 __print_symbolic(__entry->type, MF_PAGE_TYPE), 316 __print_symbolic(__entry->result, MF_ACTION_RESULT) 317 ) 318 ); 319 #endif /* CONFIG_MEMORY_FAILURE */ 320 #endif /* _TRACE_HW_EVENT_MC_H */ 321 322 /* This part must be outside protection */ 323 #include <trace/define_trace.h> 324