arch/x86/kernel/dumpstack.c

   1 /*
   2  *  Copyright (C) 1991, 1992  Linus Torvalds
   3  *  Copyright (C) 2000, 2001, 2002 Andi Kleen, SuSE Labs
   4  */
   5 #include <linux/kallsyms.h>
   6 #include <linux/kprobes.h>
   7 #include <linux/uaccess.h>
   8 #include <linux/utsname.h>
   9 #include <linux/hardirq.h>
  10 #include <linux/kdebug.h>
  11 #include <linux/module.h>
  12 #include <linux/ptrace.h>
  13 #include <linux/ftrace.h>
  14 #include <linux/kexec.h>
  15 #include <linux/bug.h>
  16 #include <linux/nmi.h>
  17 #include <linux/sysfs.h>
  18
  19 #include <asm/stacktrace.h>
  20
  21
  22 int panic_on_unrecovered_nmi;
  23 int panic_on_io_nmi;
  24 unsigned int code_bytes = 64;
  25 int kstack_depth_to_print = 3 * STACKSLOTS_PER_LINE;
  26 static int die_counter;
  27
  28 static void printk_stack_address(unsigned long address, int reliable,
  29                 void *data)
  30 {
  31         printk("%s [<%p>] %s%pB\n",
  32                 (char *)data, (void *)address, reliable ? "" : "? ",
  33                 (void *)address);
  34 }
  35
  36 void printk_address(unsigned long address)
  37 {
  38         pr_cont(" [<%p>] %pS\n", (void *)address, (void *)address);
  39 }
  40
  41 /*
  42  * x86-64 can have up to three kernel stacks:
  43  * process stack
  44  * interrupt stack
  45  * severe exception (double fault, nmi, stack fault, debug, mce) hardware stack
  46  */
  47
  48 static inline int valid_stack_ptr(struct task_struct *task,
  49                         void *p, unsigned int size, void *end)
  50 {
  51         void *t = task_stack_page(task);
  52         if (end) {
  53                 if (p < end && p >= (end-THREAD_SIZE))
  54                         return 1;
  55                 else
  56                         return 0;
  57         }
  58         return p >= t && p < t + THREAD_SIZE - size;
  59 }
  60
  61 unsigned long
  62 print_context_stack(struct task_struct *task,
  63                 unsigned long *stack, unsigned long bp,
  64                 const struct stacktrace_ops *ops, void *data,
  65                 unsigned long *end, int *graph)
  66 {
  67         struct stack_frame *frame = (struct stack_frame *)bp;
  68
  69         /*
  70          * If we overflowed the stack into a guard page, jump back to the
  71          * bottom of the usable stack.
  72          */
  73         if ((unsigned long)task_stack_page(task) - (unsigned long)stack <
  74             PAGE_SIZE)
  75                 stack = (unsigned long *)task_stack_page(task);
  76
  77         while (valid_stack_ptr(task, stack, sizeof(*stack), end)) {
  78                 unsigned long addr = *stack;
  79
  80                 if (__kernel_text_address(addr)) {
  81                         unsigned long real_addr;
  82                         int reliable = 0;
  83
  84                         if ((unsigned long) stack == bp + sizeof(long)) {
  85                                 reliable = 1;
  86                                 frame = frame->next_frame;
  87                                 bp = (unsigned long) frame;
  88                         }
  89
  90                         /*
  91                          * When function graph tracing is enabled for a
  92                          * function, its return address on the stack is
  93                          * replaced with the address of an ftrace handler
  94                          * (return_to_handler).  In that case, before printing
  95                          * the "real" address, we want to print the handler
  96                          * address as an "unreliable" hint that function graph
  97                          * tracing was involved.
  98                          */
  99                         real_addr = ftrace_graph_ret_addr(task, graph, addr,
 100                                                           stack);
 101                         if (real_addr != addr)
 102                                 ops->address(data, addr, 0);
 103
 104                         ops->address(data, real_addr, reliable);
 105                 }
 106                 stack++;
 107         }
 108         return bp;
 109 }
 110 EXPORT_SYMBOL_GPL(print_context_stack);
 111
 112 unsigned long
 113 print_context_stack_bp(struct task_struct *task,
 114                        unsigned long *stack, unsigned long bp,
 115                        const struct stacktrace_ops *ops, void *data,
 116                        unsigned long *end, int *graph)
 117 {
 118         struct stack_frame *frame = (struct stack_frame *)bp;
 119         unsigned long *retp = &frame->return_address;
 120
 121         while (valid_stack_ptr(task, retp, sizeof(*retp), end)) {
 122                 unsigned long addr = *retp;
 123                 unsigned long real_addr;
 124
 125                 if (!__kernel_text_address(addr))
 126                         break;
 127
 128                 real_addr = ftrace_graph_ret_addr(task, graph, addr, retp);
 129                 if (ops->address(data, real_addr, 1))
 130                         break;
 131
 132                 frame = frame->next_frame;
 133                 retp = &frame->return_address;
 134         }
 135
 136         return (unsigned long)frame;
 137 }
 138 EXPORT_SYMBOL_GPL(print_context_stack_bp);
 139
 140 static int print_trace_stack(void *data, char *name)
 141 {
 142         printk("%s <%s> ", (char *)data, name);
 143         return 0;
 144 }
 145
 146 /*
 147  * Print one address/symbol entries per line.
 148  */
 149 static int print_trace_address(void *data, unsigned long addr, int reliable)
 150 {
 151         touch_nmi_watchdog();
 152         printk_stack_address(addr, reliable, data);
 153         return 0;
 154 }
 155
 156 static const struct stacktrace_ops print_trace_ops = {
 157         .stack                  = print_trace_stack,
 158         .address                = print_trace_address,
 159         .walk_stack             = print_context_stack,
 160 };
 161
 162 void
 163 show_trace_log_lvl(struct task_struct *task, struct pt_regs *regs,
 164                 unsigned long *stack, unsigned long bp, char *log_lvl)
 165 {
 166         printk("%sCall Trace:\n", log_lvl);
 167         dump_trace(task, regs, stack, bp, &print_trace_ops, log_lvl);
 168 }
 169
 170 void show_stack(struct task_struct *task, unsigned long *sp)
 171 {
 172         unsigned long bp = 0;
 173         unsigned long stack;
 174
 175         /*
 176          * Stack frames below this one aren't interesting.  Don't show them
 177          * if we're printing for %current.
 178          */
 179         if (!sp && (!task || task == current)) {
 180                 sp = &stack;
 181                 bp = stack_frame(current, NULL);
 182         }
 183
 184         show_stack_log_lvl(task, NULL, sp, bp, "");
 185 }
 186
 187 void show_stack_regs(struct pt_regs *regs)
 188 {
 189         show_stack_log_lvl(current, regs, (unsigned long *)regs->sp, regs->bp, "");
 190 }
 191
 192 static arch_spinlock_t die_lock = __ARCH_SPIN_LOCK_UNLOCKED;
 193 static int die_owner = -1;
 194 static unsigned int die_nest_count;
 195
 196 unsigned long oops_begin(void)
 197 {
 198         int cpu;
 199         unsigned long flags;
 200
 201         oops_enter();
 202
 203         /* racy, but better than risking deadlock. */
 204         raw_local_irq_save(flags);
 205         cpu = smp_processor_id();
 206         if (!arch_spin_trylock(&die_lock)) {
 207                 if (cpu == die_owner)
 208                         /* nested oops. should stop eventually */;
 209                 else
 210                         arch_spin_lock(&die_lock);
 211         }
 212         die_nest_count++;
 213         die_owner = cpu;
 214         console_verbose();
 215         bust_spinlocks(1);
 216         return flags;
 217 }
 218 EXPORT_SYMBOL_GPL(oops_begin);
 219 NOKPROBE_SYMBOL(oops_begin);
 220
 221 void __noreturn rewind_stack_do_exit(int signr);
 222
 223 void oops_end(unsigned long flags, struct pt_regs *regs, int signr)
 224 {
 225         if (regs && kexec_should_crash(current))
 226                 crash_kexec(regs);
 227
 228         bust_spinlocks(0);
 229         die_owner = -1;
 230         add_taint(TAINT_DIE, LOCKDEP_NOW_UNRELIABLE);
 231         die_nest_count--;
 232         if (!die_nest_count)
 233                 /* Nest count reaches zero, release the lock. */
 234                 arch_spin_unlock(&die_lock);
 235         raw_local_irq_restore(flags);
 236         oops_exit();
 237
 238         if (!signr)
 239                 return;
 240         if (in_interrupt())
 241                 panic("Fatal exception in interrupt");
 242         if (panic_on_oops)
 243                 panic("Fatal exception");
 244
 245         /*
 246          * We're not going to return, but we might be on an IST stack or
 247          * have very little stack space left.  Rewind the stack and kill
 248          * the task.
 249          */
 250         rewind_stack_do_exit(signr);
 251 }
 252 NOKPROBE_SYMBOL(oops_end);
 253
 254 int __die(const char *str, struct pt_regs *regs, long err)
 255 {
 256 #ifdef CONFIG_X86_32
 257         unsigned short ss;
 258         unsigned long sp;
 259 #endif
 260         printk(KERN_DEFAULT
 261                "%s: %04lx [#%d]%s%s%s%s\n", str, err & 0xffff, ++die_counter,
 262                IS_ENABLED(CONFIG_PREEMPT) ? " PREEMPT"         : "",
 263                IS_ENABLED(CONFIG_SMP)     ? " SMP"             : "",
 264                debug_pagealloc_enabled()  ? " DEBUG_PAGEALLOC" : "",
 265                IS_ENABLED(CONFIG_KASAN)   ? " KASAN"           : "");
 266
 267         if (notify_die(DIE_OOPS, str, regs, err,
 268                         current->thread.trap_nr, SIGSEGV) == NOTIFY_STOP)
 269                 return 1;
 270
 271         print_modules();
 272         show_regs(regs);
 273 #ifdef CONFIG_X86_32
 274         if (user_mode(regs)) {
 275                 sp = regs->sp;
 276                 ss = regs->ss & 0xffff;
 277         } else {
 278                 sp = kernel_stack_pointer(regs);
 279                 savesegment(ss, ss);
 280         }
 281         printk(KERN_EMERG "EIP: [<%08lx>] ", regs->ip);
 282         print_symbol("%s", regs->ip);
 283         printk(" SS:ESP %04x:%08lx\n", ss, sp);
 284 #else
 285         /* Executive summary in case the oops scrolled away */
 286         printk(KERN_ALERT "RIP ");
 287         printk_address(regs->ip);
 288         printk(" RSP <%016lx>\n", regs->sp);
 289 #endif
 290         return 0;
 291 }
 292 NOKPROBE_SYMBOL(__die);
 293
 294 /*
 295  * This is gone through when something in the kernel has done something bad
 296  * and is about to be terminated:
 297  */
 298 void die(const char *str, struct pt_regs *regs, long err)
 299 {
 300         unsigned long flags = oops_begin();
 301         int sig = SIGSEGV;
 302
 303         if (!user_mode(regs))
 304                 report_bug(regs->ip, regs);
 305
 306         if (__die(str, regs, err))
 307                 sig = 0;
 308         oops_end(flags, regs, sig);
 309 }
 310
 311 static int __init kstack_setup(char *s)
 312 {
 313         ssize_t ret;
 314         unsigned long val;
 315
 316         if (!s)
 317                 return -EINVAL;
 318
 319         ret = kstrtoul(s, 0, &val);
 320         if (ret)
 321                 return ret;
 322         kstack_depth_to_print = val;
 323         return 0;
 324 }
 325 early_param("kstack", kstack_setup);
 326
 327 static int __init code_bytes_setup(char *s)
 328 {
 329         ssize_t ret;
 330         unsigned long val;
 331
 332         if (!s)
 333                 return -EINVAL;
 334
 335         ret = kstrtoul(s, 0, &val);
 336         if (ret)
 337                 return ret;
 338
 339         code_bytes = val;
 340         if (code_bytes > 8192)
 341                 code_bytes = 8192;
 342
 343         return 1;
 344 }
 345 __setup("code_bytes=", code_bytes_setup);