1 /* 2 * CDDL HEADER START 3 * 4 * The contents of this file are subject to the terms of the 5 * Common Development and Distribution License (the "License"). 6 * You may not use this file except in compliance with the License. 7 * 8 * You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE 9 * or http://www.opensolaris.org/os/licensing. 10 * See the License for the specific language governing permissions 11 * and limitations under the License. 12 * 13 * When distributing Covered Code, include this CDDL HEADER in each 14 * file and include the License file at usr/src/OPENSOLARIS.LICENSE. 15 * If applicable, add the following below this CDDL HEADER, with the 16 * fields enclosed by brackets "[]" replaced with your own identifying 17 * information: Portions Copyright [yyyy] [name of copyright owner] 18 * 19 * CDDL HEADER END 20 */ 21 /* 22 * Copyright (c) 1999, 2010, Oracle and/or its affiliates. All rights reserved. 23 */ 24 25 /* 26 * Copyright (c) 2011, Joyent, Inc. All rights reserved. 27 * Copyright (c) 2016 by Delphix. All rights reserved. 28 */ 29 30 /* 31 * When the operating system detects that it is in an invalid state, a panic 32 * is initiated in order to minimize potential damage to user data and to 33 * facilitate debugging. There are three major tasks to be performed in 34 * a system panic: recording information about the panic in memory (and thus 35 * making it part of the crash dump), synchronizing the file systems to 36 * preserve user file data, and generating the crash dump. We define the 37 * system to be in one of four states with respect to the panic code: 38 * 39 * CALM - the state of the system prior to any thread initiating a panic 40 * 41 * QUIESCE - the state of the system when the first thread to initiate 42 * a system panic records information about the cause of the panic 43 * and renders the system quiescent by stopping other processors 44 * 45 * SYNC - the state of the system when we synchronize the file systems 46 * DUMP - the state when we generate the crash dump. 47 * 48 * The transitions between these states are irreversible: once we begin 49 * panicking, we only make one attempt to perform the actions associated with 50 * each state. 51 * 52 * The panic code itself must be re-entrant because actions taken during any 53 * state may lead to another system panic. Additionally, any Solaris 54 * thread may initiate a panic at any time, and so we must have synchronization 55 * between threads which attempt to initiate a state transition simultaneously. 56 * The panic code makes use of a special locking primitive, a trigger, to 57 * perform this synchronization. A trigger is simply a word which is set 58 * atomically and can only be set once. We declare three triggers, one for 59 * each transition between the four states. When a thread enters the panic 60 * code it attempts to set each trigger; if it fails it moves on to the 61 * next trigger. A special case is the first trigger: if two threads race 62 * to perform the transition to QUIESCE, the losing thread may execute before 63 * the winner has a chance to stop its CPU. To solve this problem, we have 64 * the loser look ahead to see if any other triggers are set; if not, it 65 * presumes a panic is underway and simply spins. Unfortunately, since we 66 * are panicking, it is not possible to know this with absolute certainty. 67 * 68 * There are two common reasons for re-entering the panic code once a panic 69 * has been initiated: (1) after we debug_enter() at the end of QUIESCE, 70 * the operator may type "sync" instead of "go", and the PROM's sync callback 71 * routine will invoke panic(); (2) if the clock routine decides that sync 72 * or dump is not making progress, it will invoke panic() to force a timeout. 73 * The design assumes that a third possibility, another thread causing an 74 * unrelated panic while sync or dump is still underway, is extremely unlikely. 75 * If this situation occurs, we may end up triggering dump while sync is 76 * still in progress. This third case is considered extremely unlikely because 77 * all other CPUs are stopped and low-level interrupts have been blocked. 78 * 79 * The panic code is entered via a call directly to the vpanic() function, 80 * or its varargs wrappers panic() and cmn_err(9F). The vpanic routine 81 * is implemented in assembly language to record the current machine 82 * registers, attempt to set the trigger for the QUIESCE state, and 83 * if successful, switch stacks on to the panic_stack before calling into 84 * the common panicsys() routine. The first thread to initiate a panic 85 * is allowed to make use of the reserved panic_stack so that executing 86 * the panic code itself does not overwrite valuable data on that thread's 87 * stack *ahead* of the current stack pointer. This data will be preserved 88 * in the crash dump and may prove invaluable in determining what this 89 * thread has previously been doing. The first thread, saved in panic_thread, 90 * is also responsible for stopping the other CPUs as quickly as possible, 91 * and then setting the various panic_* variables. Most important among 92 * these is panicstr, which allows threads to subsequently bypass held 93 * locks so that we can proceed without ever blocking. We must stop the 94 * other CPUs *prior* to setting panicstr in case threads running there are 95 * currently spinning to acquire a lock; we want that state to be preserved. 96 * Every thread which initiates a panic has its T_PANIC flag set so we can 97 * identify all such threads in the crash dump. 98 * 99 * The panic_thread is also allowed to make use of the special memory buffer 100 * panicbuf, which on machines with appropriate hardware is preserved across 101 * reboots. We allow the panic_thread to store its register set and panic 102 * message in this buffer, so even if we fail to obtain a crash dump we will 103 * be able to examine the machine after reboot and determine some of the 104 * state at the time of the panic. If we do get a dump, the panic buffer 105 * data is structured so that a debugger can easily consume the information 106 * therein (see <sys/panic.h>). 107 * 108 * Each platform or architecture is required to implement the functions 109 * panic_savetrap() to record trap-specific information to panicbuf, 110 * panic_saveregs() to record a register set to panicbuf, panic_stopcpus() 111 * to halt all CPUs but the panicking CPU, panic_quiesce_hw() to perform 112 * miscellaneous platform-specific tasks *after* panicstr is set, 113 * panic_showtrap() to print trap-specific information to the console, 114 * and panic_dump_hw() to perform platform tasks prior to calling dumpsys(). 115 * 116 * A Note on Word Formation, courtesy of the Oxford Guide to English Usage: 117 * 118 * Words ending in -c interpose k before suffixes which otherwise would 119 * indicate a soft c, and thus the verb and adjective forms of 'panic' are 120 * spelled "panicked", "panicking", and "panicky" respectively. Use of 121 * the ill-conceived "panicing" and "panic'd" is discouraged. 122 */ 123 124 #include <sys/types.h> 125 #include <sys/varargs.h> 126 #include <sys/sysmacros.h> 127 #include <sys/cmn_err.h> 128 #include <sys/cpuvar.h> 129 #include <sys/thread.h> 130 #include <sys/t_lock.h> 131 #include <sys/cred.h> 132 #include <sys/systm.h> 133 #include <sys/archsystm.h> 134 #include <sys/uadmin.h> 135 #include <sys/callb.h> 136 #include <sys/vfs.h> 137 #include <sys/log.h> 138 #include <sys/disp.h> 139 #include <sys/param.h> 140 #include <sys/dumphdr.h> 141 #include <sys/ftrace.h> 142 #include <sys/reboot.h> 143 #include <sys/debug.h> 144 #include <sys/stack.h> 145 #include <sys/spl.h> 146 #include <sys/errorq.h> 147 #include <sys/panic.h> 148 #include <sys/fm/util.h> 149 #include <sys/clock_impl.h> 150 #include <sys/sunddi.h> 151 152 /* 153 * Panic variables which are set once during the QUIESCE state by the 154 * first thread to initiate a panic. These are examined by post-mortem 155 * debugging tools; the inconsistent use of 'panic' versus 'panic_' in 156 * the variable naming is historical and allows legacy tools to work. 157 */ 158 #pragma align STACK_ALIGN(panic_stack) 159 char panic_stack[PANICSTKSIZE]; /* reserved stack for panic_thread */ 160 kthread_t *panic_thread; /* first thread to call panicsys() */ 161 cpu_t panic_cpu; /* cpu from first call to panicsys() */ 162 label_t panic_regs; /* setjmp label from panic_thread */ 163 label_t panic_pcb; /* t_pcb at time of panic */ 164 struct regs *panic_reg; /* regs struct from first panicsys() */ 165 char *volatile panicstr; /* format string to first panicsys() */ 166 va_list panicargs; /* arguments to first panicsys() */ 167 clock_t panic_lbolt; /* lbolt at time of panic */ 168 int64_t panic_lbolt64; /* lbolt64 at time of panic */ 169 hrtime_t panic_hrtime; /* hrtime at time of panic */ 170 timespec_t panic_hrestime; /* hrestime at time of panic */ 171 int panic_ipl; /* ipl on panic_cpu at time of panic */ 172 ushort_t panic_schedflag; /* t_schedflag for panic_thread */ 173 cpu_t *panic_bound_cpu; /* t_bound_cpu for panic_thread */ 174 char panic_preempt; /* t_preempt for panic_thread */ 175 dev_info_t *panic_dip; /* from dev_err(dip, CE_PANIC, ...) */ 176 177 /* 178 * Panic variables which can be set via /etc/system or patched while 179 * the system is in operation. Again, the stupid names are historic. 180 */ 181 char *panic_bootstr = NULL; /* mdboot string to use after panic */ 182 int panic_bootfcn = AD_BOOT; /* mdboot function to use after panic */ 183 int halt_on_panic = 0; /* halt after dump instead of reboot? */ 184 int nopanicdebug = 0; /* reboot instead of call debugger? */ 185 int in_sync = 0; /* skip vfs_syncall() and just dump? */ 186 187 /* 188 * The do_polled_io flag is set by the panic code to inform the SCSI subsystem 189 * to use polled mode instead of interrupt-driven i/o. 190 */ 191 int do_polled_io = 0; 192 193 /* 194 * The panic_forced flag is set by the uadmin A_DUMP code to inform the 195 * panic subsystem that it should not attempt an initial debug_enter. 196 */ 197 int panic_forced = 0; 198 199 /* 200 * Triggers for panic state transitions: 201 */ 202 int panic_quiesce; /* trigger for CALM -> QUIESCE */ 203 int panic_dump; /* trigger for QUIESCE -> DUMP */ 204 205 /* 206 * Variable signifying quiesce(9E) is in progress. 207 */ 208 volatile int quiesce_active = 0; 209 210 void 211 panicsys(const char *format, va_list alist, struct regs *rp, int on_panic_stack) 212 { 213 int s = spl8(); 214 kthread_t *t = curthread; 215 cpu_t *cp = CPU; 216 217 caddr_t intr_stack = NULL; 218 volatile uint_t intr_actv; 219 220 ushort_t schedflag = t->t_schedflag; 221 cpu_t *bound_cpu = t->t_bound_cpu; 222 char preempt = t->t_preempt; 223 label_t pcb = t->t_pcb; 224 225 (void) setjmp(&t->t_pcb); 226 t->t_flag |= T_PANIC; 227 228 t->t_schedflag |= TS_DONT_SWAP; 229 t->t_bound_cpu = cp; 230 t->t_preempt++; 231 232 panic_enter_hw(s); 233 234 /* 235 * If we're on the interrupt stack and an interrupt thread is available 236 * in this CPU's pool, preserve the interrupt stack by detaching an 237 * interrupt thread and making its stack the intr_stack. 238 */ 239 if (CPU_ON_INTR(cp) && cp->cpu_intr_thread != NULL) { 240 kthread_t *it = cp->cpu_intr_thread; 241 242 intr_stack = cp->cpu_intr_stack; 243 intr_actv = cp->cpu_intr_actv; 244 245 cp->cpu_intr_stack = thread_stk_init(it->t_stk); 246 cp->cpu_intr_thread = it->t_link; 247 248 /* 249 * Clear only the high level bits of cpu_intr_actv. 250 * We want to indicate that high-level interrupts are 251 * not active without destroying the low-level interrupt 252 * information stored there. 253 */ 254 cp->cpu_intr_actv &= ((1 << (LOCK_LEVEL + 1)) - 1); 255 } 256 257 /* 258 * Record one-time panic information and quiesce the other CPUs. 259 * Then print out the panic message and stack trace. 260 */ 261 if (on_panic_stack) { 262 panic_data_t *pdp = (panic_data_t *)panicbuf; 263 264 pdp->pd_version = PANICBUFVERS; 265 pdp->pd_msgoff = sizeof (panic_data_t) - sizeof (panic_nv_t); 266 267 (void) strncpy(pdp->pd_uuid, dump_get_uuid(), 268 sizeof (pdp->pd_uuid)); 269 270 if (t->t_panic_trap != NULL) 271 panic_savetrap(pdp, t->t_panic_trap); 272 else 273 panic_saveregs(pdp, rp); 274 275 (void) vsnprintf(&panicbuf[pdp->pd_msgoff], 276 PANICBUFSIZE - pdp->pd_msgoff, format, alist); 277 278 /* 279 * Call into the platform code to stop the other CPUs. 280 * We currently have all interrupts blocked, and expect that 281 * the platform code will lower ipl only as far as needed to 282 * perform cross-calls, and will acquire as *few* locks as is 283 * possible -- panicstr is not set so we can still deadlock. 284 */ 285 panic_stopcpus(cp, t, s); 286 287 panicstr = (char *)format; 288 va_copy(panicargs, alist); 289 panic_lbolt = LBOLT_NO_ACCOUNT; 290 panic_lbolt64 = LBOLT_NO_ACCOUNT64; 291 panic_hrestime = hrestime; 292 panic_hrtime = gethrtime_waitfree(); 293 panic_thread = t; 294 panic_regs = t->t_pcb; 295 panic_reg = rp; 296 panic_cpu = *cp; 297 panic_ipl = spltoipl(s); 298 panic_schedflag = schedflag; 299 panic_bound_cpu = bound_cpu; 300 panic_preempt = preempt; 301 panic_pcb = pcb; 302 303 if (intr_stack != NULL) { 304 panic_cpu.cpu_intr_stack = intr_stack; 305 panic_cpu.cpu_intr_actv = intr_actv; 306 } 307 308 /* 309 * Lower ipl to 10 to keep clock() from running, but allow 310 * keyboard interrupts to enter the debugger. These callbacks 311 * are executed with panicstr set so they can bypass locks. 312 */ 313 splx(ipltospl(CLOCK_LEVEL)); 314 panic_quiesce_hw(pdp); 315 (void) FTRACE_STOP(); 316 (void) callb_execute_class(CB_CL_PANIC, 0); 317 318 if (log_intrq != NULL) 319 log_flushq(log_intrq); 320 321 /* 322 * If log_consq has been initialized and syslogd has started, 323 * print any messages in log_consq that haven't been consumed. 324 */ 325 if (log_consq != NULL && log_consq != log_backlogq) 326 log_printq(log_consq); 327 328 fm_banner(); 329 330 #if defined(__x86) 331 /* 332 * A hypervisor panic originates outside of Solaris, so we 333 * don't want to prepend the panic message with misleading 334 * pointers from within Solaris. 335 */ 336 if (!IN_XPV_PANIC()) 337 #endif 338 printf("\n\rpanic[cpu%d]/thread=%p: ", cp->cpu_id, 339 (void *)t); 340 if (panic_dip != NULL) { 341 printf("%s%d: ", ddi_driver_name(panic_dip), 342 ddi_get_instance(panic_dip)); 343 } 344 vprintf(format, alist); 345 printf("\n\n"); 346 347 if (t->t_panic_trap != NULL) { 348 panic_showtrap(t->t_panic_trap); 349 printf("\n"); 350 } 351 352 traceregs(rp); 353 printf("\n"); 354 355 if (((boothowto & RB_DEBUG) || obpdebug) && 356 !nopanicdebug && !panic_forced) { 357 if (dumpvp != NULL) { 358 debug_enter("panic: entering debugger " 359 "(continue to save dump)"); 360 } else { 361 debug_enter("panic: entering debugger " 362 "(no dump device, continue to reboot)"); 363 } 364 } 365 366 } else if (panic_dump != 0 || panicstr != NULL) { 367 printf("\n\rpanic[cpu%d]/thread=%p: ", cp->cpu_id, (void *)t); 368 vprintf(format, alist); 369 printf("\n"); 370 } else 371 goto spin; 372 373 /* 374 * Prior to performing dump, we make sure that do_polled_io is 375 * set, but we'll leave ipl at 10; deadman(), a CY_HIGH_LEVEL cyclic, 376 * will re-enter panic if we are not making progress with dump. 377 */ 378 /* 379 * Take the crash dump. If the dump trigger is already set, try to 380 * enter the debugger again before rebooting the system. 381 */ 382 if (panic_trigger(&panic_dump)) { 383 panic_dump_hw(s); 384 splx(ipltospl(CLOCK_LEVEL)); 385 errorq_panic(); 386 do_polled_io = 1; 387 dumpsys(); 388 } else if (((boothowto & RB_DEBUG) || obpdebug) && !nopanicdebug) { 389 debug_enter("panic: entering debugger (continue to reboot)"); 390 } else 391 printf("dump aborted: please record the above information!\n"); 392 393 if (halt_on_panic) 394 mdboot(A_REBOOT, AD_HALT, NULL, B_FALSE); 395 else 396 mdboot(A_REBOOT, panic_bootfcn, panic_bootstr, B_FALSE); 397 spin: 398 /* 399 * Restore ipl to at most CLOCK_LEVEL so we don't end up spinning 400 * and unable to jump into the debugger. 401 */ 402 splx(MIN(s, ipltospl(CLOCK_LEVEL))); 403 for (;;) 404 ; 405 } 406 407 void 408 panic(const char *format, ...) 409 { 410 va_list alist; 411 412 va_start(alist, format); 413 vpanic(format, alist); 414 va_end(alist); 415 } 416