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