1 /* 2 * Machine check exception handling. 3 * 4 * This program is free software; you can redistribute it and/or modify 5 * it under the terms of the GNU General Public License as published by 6 * the Free Software Foundation; either version 2 of the License, or 7 * (at your option) any later version. 8 * 9 * This program is distributed in the hope that it will be useful, 10 * but WITHOUT ANY WARRANTY; without even the implied warranty of 11 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the 12 * GNU General Public License for more details. 13 * 14 * You should have received a copy of the GNU General Public License 15 * along with this program; if not, write to the Free Software 16 * Foundation, Inc., 59 Temple Place - Suite 330, Boston, MA 02111-1307, USA. 17 * 18 * Copyright 2013 IBM Corporation 19 * Author: Mahesh Salgaonkar <mahesh@linux.vnet.ibm.com> 20 */ 21 22 #undef DEBUG 23 #define pr_fmt(fmt) "mce: " fmt 24 25 #include <linux/types.h> 26 #include <linux/ptrace.h> 27 #include <linux/percpu.h> 28 #include <linux/export.h> 29 #include <asm/mce.h> 30 31 static DEFINE_PER_CPU(int, mce_nest_count); 32 static DEFINE_PER_CPU(struct machine_check_event[MAX_MC_EVT], mce_event); 33 34 /* Queue for delayed MCE events. */ 35 static DEFINE_PER_CPU(int, mce_queue_count); 36 static DEFINE_PER_CPU(struct machine_check_event[MAX_MC_EVT], mce_event_queue); 37 38 static void mce_set_error_info(struct machine_check_event *mce, 39 struct mce_error_info *mce_err) 40 { 41 mce->error_type = mce_err->error_type; 42 switch (mce_err->error_type) { 43 case MCE_ERROR_TYPE_UE: 44 mce->u.ue_error.ue_error_type = mce_err->u.ue_error_type; 45 break; 46 case MCE_ERROR_TYPE_SLB: 47 mce->u.slb_error.slb_error_type = mce_err->u.slb_error_type; 48 break; 49 case MCE_ERROR_TYPE_ERAT: 50 mce->u.erat_error.erat_error_type = mce_err->u.erat_error_type; 51 break; 52 case MCE_ERROR_TYPE_TLB: 53 mce->u.tlb_error.tlb_error_type = mce_err->u.tlb_error_type; 54 break; 55 case MCE_ERROR_TYPE_UNKNOWN: 56 default: 57 break; 58 } 59 } 60 61 /* 62 * Decode and save high level MCE information into per cpu buffer which 63 * is an array of machine_check_event structure. 64 */ 65 void save_mce_event(struct pt_regs *regs, long handled, 66 struct mce_error_info *mce_err, 67 uint64_t addr) 68 { 69 uint64_t srr1; 70 int index = __get_cpu_var(mce_nest_count)++; 71 struct machine_check_event *mce = &__get_cpu_var(mce_event[index]); 72 73 /* 74 * Return if we don't have enough space to log mce event. 75 * mce_nest_count may go beyond MAX_MC_EVT but that's ok, 76 * the check below will stop buffer overrun. 77 */ 78 if (index >= MAX_MC_EVT) 79 return; 80 81 /* Populate generic machine check info */ 82 mce->version = MCE_V1; 83 mce->srr0 = regs->nip; 84 mce->srr1 = regs->msr; 85 mce->gpr3 = regs->gpr[3]; 86 mce->in_use = 1; 87 88 mce->initiator = MCE_INITIATOR_CPU; 89 if (handled) 90 mce->disposition = MCE_DISPOSITION_RECOVERED; 91 else 92 mce->disposition = MCE_DISPOSITION_NOT_RECOVERED; 93 mce->severity = MCE_SEV_ERROR_SYNC; 94 95 srr1 = regs->msr; 96 97 /* 98 * Populate the mce error_type and type-specific error_type. 99 */ 100 mce_set_error_info(mce, mce_err); 101 102 if (!addr) 103 return; 104 105 if (mce->error_type == MCE_ERROR_TYPE_TLB) { 106 mce->u.tlb_error.effective_address_provided = true; 107 mce->u.tlb_error.effective_address = addr; 108 } else if (mce->error_type == MCE_ERROR_TYPE_SLB) { 109 mce->u.slb_error.effective_address_provided = true; 110 mce->u.slb_error.effective_address = addr; 111 } else if (mce->error_type == MCE_ERROR_TYPE_ERAT) { 112 mce->u.erat_error.effective_address_provided = true; 113 mce->u.erat_error.effective_address = addr; 114 } else if (mce->error_type == MCE_ERROR_TYPE_UE) { 115 mce->u.ue_error.effective_address_provided = true; 116 mce->u.ue_error.effective_address = addr; 117 } 118 return; 119 } 120 121 /* 122 * get_mce_event: 123 * mce Pointer to machine_check_event structure to be filled. 124 * release Flag to indicate whether to free the event slot or not. 125 * 0 <= do not release the mce event. Caller will invoke 126 * release_mce_event() once event has been consumed. 127 * 1 <= release the slot. 128 * 129 * return 1 = success 130 * 0 = failure 131 * 132 * get_mce_event() will be called by platform specific machine check 133 * handle routine and in KVM. 134 * When we call get_mce_event(), we are still in interrupt context and 135 * preemption will not be scheduled until ret_from_expect() routine 136 * is called. 137 */ 138 int get_mce_event(struct machine_check_event *mce, bool release) 139 { 140 int index = __get_cpu_var(mce_nest_count) - 1; 141 struct machine_check_event *mc_evt; 142 int ret = 0; 143 144 /* Sanity check */ 145 if (index < 0) 146 return ret; 147 148 /* Check if we have MCE info to process. */ 149 if (index < MAX_MC_EVT) { 150 mc_evt = &__get_cpu_var(mce_event[index]); 151 /* Copy the event structure and release the original */ 152 if (mce) 153 *mce = *mc_evt; 154 if (release) 155 mc_evt->in_use = 0; 156 ret = 1; 157 } 158 /* Decrement the count to free the slot. */ 159 if (release) 160 __get_cpu_var(mce_nest_count)--; 161 162 return ret; 163 } 164 165 void release_mce_event(void) 166 { 167 get_mce_event(NULL, true); 168 } 169 170 /* 171 * Queue up the MCE event which then can be handled later. 172 */ 173 void machine_check_queue_event(void) 174 { 175 int index; 176 struct machine_check_event evt; 177 178 if (!get_mce_event(&evt, MCE_EVENT_RELEASE)) 179 return; 180 181 index = __get_cpu_var(mce_queue_count)++; 182 /* If queue is full, just return for now. */ 183 if (index >= MAX_MC_EVT) { 184 __get_cpu_var(mce_queue_count)--; 185 return; 186 } 187 __get_cpu_var(mce_event_queue[index]) = evt; 188 } 189 190 /* 191 * process pending MCE event from the mce event queue. This function will be 192 * called during syscall exit. 193 */ 194 void machine_check_process_queued_event(void) 195 { 196 int index; 197 198 preempt_disable(); 199 /* 200 * For now just print it to console. 201 * TODO: log this error event to FSP or nvram. 202 */ 203 while (__get_cpu_var(mce_queue_count) > 0) { 204 index = __get_cpu_var(mce_queue_count) - 1; 205 machine_check_print_event_info( 206 &__get_cpu_var(mce_event_queue[index])); 207 __get_cpu_var(mce_queue_count)--; 208 } 209 preempt_enable(); 210 } 211 212 void machine_check_print_event_info(struct machine_check_event *evt) 213 { 214 const char *level, *sevstr, *subtype; 215 static const char *mc_ue_types[] = { 216 "Indeterminate", 217 "Instruction fetch", 218 "Page table walk ifetch", 219 "Load/Store", 220 "Page table walk Load/Store", 221 }; 222 static const char *mc_slb_types[] = { 223 "Indeterminate", 224 "Parity", 225 "Multihit", 226 }; 227 static const char *mc_erat_types[] = { 228 "Indeterminate", 229 "Parity", 230 "Multihit", 231 }; 232 static const char *mc_tlb_types[] = { 233 "Indeterminate", 234 "Parity", 235 "Multihit", 236 }; 237 238 /* Print things out */ 239 if (evt->version != MCE_V1) { 240 pr_err("Machine Check Exception, Unknown event version %d !\n", 241 evt->version); 242 return; 243 } 244 switch (evt->severity) { 245 case MCE_SEV_NO_ERROR: 246 level = KERN_INFO; 247 sevstr = "Harmless"; 248 break; 249 case MCE_SEV_WARNING: 250 level = KERN_WARNING; 251 sevstr = ""; 252 break; 253 case MCE_SEV_ERROR_SYNC: 254 level = KERN_ERR; 255 sevstr = "Severe"; 256 break; 257 case MCE_SEV_FATAL: 258 default: 259 level = KERN_ERR; 260 sevstr = "Fatal"; 261 break; 262 } 263 264 printk("%s%s Machine check interrupt [%s]\n", level, sevstr, 265 evt->disposition == MCE_DISPOSITION_RECOVERED ? 266 "Recovered" : "[Not recovered"); 267 printk("%s Initiator: %s\n", level, 268 evt->initiator == MCE_INITIATOR_CPU ? "CPU" : "Unknown"); 269 switch (evt->error_type) { 270 case MCE_ERROR_TYPE_UE: 271 subtype = evt->u.ue_error.ue_error_type < 272 ARRAY_SIZE(mc_ue_types) ? 273 mc_ue_types[evt->u.ue_error.ue_error_type] 274 : "Unknown"; 275 printk("%s Error type: UE [%s]\n", level, subtype); 276 if (evt->u.ue_error.effective_address_provided) 277 printk("%s Effective address: %016llx\n", 278 level, evt->u.ue_error.effective_address); 279 if (evt->u.ue_error.physical_address_provided) 280 printk("%s Physial address: %016llx\n", 281 level, evt->u.ue_error.physical_address); 282 break; 283 case MCE_ERROR_TYPE_SLB: 284 subtype = evt->u.slb_error.slb_error_type < 285 ARRAY_SIZE(mc_slb_types) ? 286 mc_slb_types[evt->u.slb_error.slb_error_type] 287 : "Unknown"; 288 printk("%s Error type: SLB [%s]\n", level, subtype); 289 if (evt->u.slb_error.effective_address_provided) 290 printk("%s Effective address: %016llx\n", 291 level, evt->u.slb_error.effective_address); 292 break; 293 case MCE_ERROR_TYPE_ERAT: 294 subtype = evt->u.erat_error.erat_error_type < 295 ARRAY_SIZE(mc_erat_types) ? 296 mc_erat_types[evt->u.erat_error.erat_error_type] 297 : "Unknown"; 298 printk("%s Error type: ERAT [%s]\n", level, subtype); 299 if (evt->u.erat_error.effective_address_provided) 300 printk("%s Effective address: %016llx\n", 301 level, evt->u.erat_error.effective_address); 302 break; 303 case MCE_ERROR_TYPE_TLB: 304 subtype = evt->u.tlb_error.tlb_error_type < 305 ARRAY_SIZE(mc_tlb_types) ? 306 mc_tlb_types[evt->u.tlb_error.tlb_error_type] 307 : "Unknown"; 308 printk("%s Error type: TLB [%s]\n", level, subtype); 309 if (evt->u.tlb_error.effective_address_provided) 310 printk("%s Effective address: %016llx\n", 311 level, evt->u.tlb_error.effective_address); 312 break; 313 default: 314 case MCE_ERROR_TYPE_UNKNOWN: 315 printk("%s Error type: Unknown\n", level); 316 break; 317 } 318 } 319