]>
Commit | Line | Data |
---|---|---|
1da177e4 LT |
1 | /* |
2 | * linux/arch/x86-64/traps.c | |
3 | * | |
4 | * Copyright (C) 1991, 1992 Linus Torvalds | |
5 | * Copyright (C) 2000, 2001, 2002 Andi Kleen, SuSE Labs | |
6 | * | |
7 | * Pentium III FXSR, SSE support | |
8 | * Gareth Hughes <gareth@valinux.com>, May 2000 | |
9 | * | |
10 | * $Id: traps.c,v 1.36 2002/03/24 11:09:10 ak Exp $ | |
11 | */ | |
12 | ||
13 | /* | |
14 | * 'Traps.c' handles hardware traps and faults after we have saved some | |
15 | * state in 'entry.S'. | |
16 | */ | |
17 | #include <linux/config.h> | |
18 | #include <linux/sched.h> | |
19 | #include <linux/kernel.h> | |
20 | #include <linux/string.h> | |
21 | #include <linux/errno.h> | |
22 | #include <linux/ptrace.h> | |
23 | #include <linux/timer.h> | |
24 | #include <linux/mm.h> | |
25 | #include <linux/init.h> | |
26 | #include <linux/delay.h> | |
27 | #include <linux/spinlock.h> | |
28 | #include <linux/interrupt.h> | |
29 | #include <linux/module.h> | |
30 | #include <linux/moduleparam.h> | |
35faa714 | 31 | #include <linux/nmi.h> |
0f2fbdcb | 32 | #include <linux/kprobes.h> |
8bcc5280 | 33 | #include <linux/kexec.h> |
1da177e4 LT |
34 | |
35 | #include <asm/system.h> | |
36 | #include <asm/uaccess.h> | |
37 | #include <asm/io.h> | |
38 | #include <asm/atomic.h> | |
39 | #include <asm/debugreg.h> | |
40 | #include <asm/desc.h> | |
41 | #include <asm/i387.h> | |
42 | #include <asm/kdebug.h> | |
43 | #include <asm/processor.h> | |
44 | ||
45 | #include <asm/smp.h> | |
46 | #include <asm/pgalloc.h> | |
47 | #include <asm/pda.h> | |
48 | #include <asm/proto.h> | |
49 | #include <asm/nmi.h> | |
50 | ||
1da177e4 LT |
51 | asmlinkage void divide_error(void); |
52 | asmlinkage void debug(void); | |
53 | asmlinkage void nmi(void); | |
54 | asmlinkage void int3(void); | |
55 | asmlinkage void overflow(void); | |
56 | asmlinkage void bounds(void); | |
57 | asmlinkage void invalid_op(void); | |
58 | asmlinkage void device_not_available(void); | |
59 | asmlinkage void double_fault(void); | |
60 | asmlinkage void coprocessor_segment_overrun(void); | |
61 | asmlinkage void invalid_TSS(void); | |
62 | asmlinkage void segment_not_present(void); | |
63 | asmlinkage void stack_segment(void); | |
64 | asmlinkage void general_protection(void); | |
65 | asmlinkage void page_fault(void); | |
66 | asmlinkage void coprocessor_error(void); | |
67 | asmlinkage void simd_coprocessor_error(void); | |
68 | asmlinkage void reserved(void); | |
69 | asmlinkage void alignment_check(void); | |
70 | asmlinkage void machine_check(void); | |
71 | asmlinkage void spurious_interrupt_bug(void); | |
1da177e4 | 72 | |
e041c683 | 73 | ATOMIC_NOTIFIER_HEAD(die_chain); |
1da177e4 LT |
74 | |
75 | int register_die_notifier(struct notifier_block *nb) | |
76 | { | |
8c914cb7 | 77 | vmalloc_sync_all(); |
e041c683 AS |
78 | return atomic_notifier_chain_register(&die_chain, nb); |
79 | } | |
80 | EXPORT_SYMBOL(register_die_notifier); | |
81 | ||
82 | int unregister_die_notifier(struct notifier_block *nb) | |
83 | { | |
84 | return atomic_notifier_chain_unregister(&die_chain, nb); | |
1da177e4 | 85 | } |
e041c683 | 86 | EXPORT_SYMBOL(unregister_die_notifier); |
1da177e4 LT |
87 | |
88 | static inline void conditional_sti(struct pt_regs *regs) | |
89 | { | |
90 | if (regs->eflags & X86_EFLAGS_IF) | |
91 | local_irq_enable(); | |
92 | } | |
93 | ||
a65d17c9 JB |
94 | static inline void preempt_conditional_sti(struct pt_regs *regs) |
95 | { | |
96 | preempt_disable(); | |
97 | if (regs->eflags & X86_EFLAGS_IF) | |
98 | local_irq_enable(); | |
99 | } | |
100 | ||
101 | static inline void preempt_conditional_cli(struct pt_regs *regs) | |
102 | { | |
103 | if (regs->eflags & X86_EFLAGS_IF) | |
104 | local_irq_disable(); | |
105 | preempt_enable_no_resched(); | |
106 | } | |
107 | ||
1da177e4 LT |
108 | static int kstack_depth_to_print = 10; |
109 | ||
110 | #ifdef CONFIG_KALLSYMS | |
111 | #include <linux/kallsyms.h> | |
112 | int printk_address(unsigned long address) | |
113 | { | |
114 | unsigned long offset = 0, symsize; | |
115 | const char *symname; | |
116 | char *modname; | |
117 | char *delim = ":"; | |
118 | char namebuf[128]; | |
119 | ||
120 | symname = kallsyms_lookup(address, &symsize, &offset, &modname, namebuf); | |
121 | if (!symname) | |
122 | return printk("[<%016lx>]", address); | |
123 | if (!modname) | |
124 | modname = delim = ""; | |
125 | return printk("<%016lx>{%s%s%s%s%+ld}", | |
2b692a87 | 126 | address, delim, modname, delim, symname, offset); |
1da177e4 LT |
127 | } |
128 | #else | |
129 | int printk_address(unsigned long address) | |
130 | { | |
131 | return printk("[<%016lx>]", address); | |
132 | } | |
133 | #endif | |
134 | ||
0a658002 AK |
135 | static unsigned long *in_exception_stack(unsigned cpu, unsigned long stack, |
136 | unsigned *usedp, const char **idp) | |
137 | { | |
b556b35e | 138 | static char ids[][8] = { |
0a658002 AK |
139 | [DEBUG_STACK - 1] = "#DB", |
140 | [NMI_STACK - 1] = "NMI", | |
141 | [DOUBLEFAULT_STACK - 1] = "#DF", | |
142 | [STACKFAULT_STACK - 1] = "#SS", | |
143 | [MCE_STACK - 1] = "#MC", | |
b556b35e JB |
144 | #if DEBUG_STKSZ > EXCEPTION_STKSZ |
145 | [N_EXCEPTION_STACKS ... N_EXCEPTION_STACKS + DEBUG_STKSZ / EXCEPTION_STKSZ - 2] = "#DB[?]" | |
146 | #endif | |
0a658002 AK |
147 | }; |
148 | unsigned k; | |
1da177e4 | 149 | |
0a658002 AK |
150 | for (k = 0; k < N_EXCEPTION_STACKS; k++) { |
151 | unsigned long end; | |
152 | ||
b556b35e JB |
153 | switch (k + 1) { |
154 | #if DEBUG_STKSZ > EXCEPTION_STKSZ | |
155 | case DEBUG_STACK: | |
df79efde | 156 | end = cpu_pda(cpu)->debugstack + DEBUG_STKSZ; |
b556b35e JB |
157 | break; |
158 | #endif | |
159 | default: | |
160 | end = per_cpu(init_tss, cpu).ist[k]; | |
161 | break; | |
162 | } | |
0a658002 AK |
163 | if (stack >= end) |
164 | continue; | |
165 | if (stack >= end - EXCEPTION_STKSZ) { | |
166 | if (*usedp & (1U << k)) | |
167 | break; | |
168 | *usedp |= 1U << k; | |
169 | *idp = ids[k]; | |
170 | return (unsigned long *)end; | |
171 | } | |
b556b35e JB |
172 | #if DEBUG_STKSZ > EXCEPTION_STKSZ |
173 | if (k == DEBUG_STACK - 1 && stack >= end - DEBUG_STKSZ) { | |
174 | unsigned j = N_EXCEPTION_STACKS - 1; | |
175 | ||
176 | do { | |
177 | ++j; | |
178 | end -= EXCEPTION_STKSZ; | |
179 | ids[j][4] = '1' + (j - N_EXCEPTION_STACKS); | |
180 | } while (stack < end - EXCEPTION_STKSZ); | |
181 | if (*usedp & (1U << j)) | |
182 | break; | |
183 | *usedp |= 1U << j; | |
184 | *idp = ids[j]; | |
185 | return (unsigned long *)end; | |
186 | } | |
187 | #endif | |
1da177e4 LT |
188 | } |
189 | return NULL; | |
0a658002 | 190 | } |
1da177e4 LT |
191 | |
192 | /* | |
193 | * x86-64 can have upto three kernel stacks: | |
194 | * process stack | |
195 | * interrupt stack | |
0a658002 | 196 | * severe exception (double fault, nmi, stack fault, debug, mce) hardware stack |
1da177e4 LT |
197 | */ |
198 | ||
199 | void show_trace(unsigned long *stack) | |
200 | { | |
0a658002 | 201 | const unsigned cpu = safe_smp_processor_id(); |
df79efde | 202 | unsigned long *irqstack_end = (unsigned long *)cpu_pda(cpu)->irqstackptr; |
1da177e4 | 203 | int i; |
0a658002 | 204 | unsigned used = 0; |
1da177e4 LT |
205 | |
206 | printk("\nCall Trace:"); | |
0a658002 AK |
207 | |
208 | #define HANDLE_STACK(cond) \ | |
209 | do while (cond) { \ | |
1b2f6304 | 210 | unsigned long addr = *stack++; \ |
0a658002 | 211 | if (kernel_text_address(addr)) { \ |
1b2f6304 JB |
212 | if (i > 50) { \ |
213 | printk("\n "); \ | |
214 | i = 0; \ | |
215 | } \ | |
216 | else \ | |
217 | i += printk(" "); \ | |
0a658002 AK |
218 | /* \ |
219 | * If the address is either in the text segment of the \ | |
220 | * kernel, or in the region which contains vmalloc'ed \ | |
221 | * memory, it *may* be the address of a calling \ | |
222 | * routine; if so, print it so that someone tracing \ | |
223 | * down the cause of the crash will be able to figure \ | |
224 | * out the call path that was taken. \ | |
225 | */ \ | |
226 | i += printk_address(addr); \ | |
0a658002 AK |
227 | } \ |
228 | } while (0) | |
229 | ||
1b2f6304 | 230 | for(i = 11; ; ) { |
0a658002 AK |
231 | const char *id; |
232 | unsigned long *estack_end; | |
233 | estack_end = in_exception_stack(cpu, (unsigned long)stack, | |
234 | &used, &id); | |
235 | ||
236 | if (estack_end) { | |
1b2f6304 | 237 | i += printk(" <%s>", id); |
0a658002 | 238 | HANDLE_STACK (stack < estack_end); |
1b2f6304 | 239 | i += printk(" <EOE>"); |
0a658002 AK |
240 | stack = (unsigned long *) estack_end[-2]; |
241 | continue; | |
1da177e4 | 242 | } |
0a658002 AK |
243 | if (irqstack_end) { |
244 | unsigned long *irqstack; | |
245 | irqstack = irqstack_end - | |
246 | (IRQSTACKSIZE - 64) / sizeof(*irqstack); | |
247 | ||
248 | if (stack >= irqstack && stack < irqstack_end) { | |
1b2f6304 | 249 | i += printk(" <IRQ>"); |
0a658002 AK |
250 | HANDLE_STACK (stack < irqstack_end); |
251 | stack = (unsigned long *) (irqstack_end[-1]); | |
252 | irqstack_end = NULL; | |
1b2f6304 | 253 | i += printk(" <EOI>"); |
0a658002 | 254 | continue; |
1da177e4 | 255 | } |
1da177e4 | 256 | } |
0a658002 | 257 | break; |
1da177e4 | 258 | } |
0a658002 AK |
259 | |
260 | HANDLE_STACK (((long) stack & (THREAD_SIZE-1)) != 0); | |
261 | #undef HANDLE_STACK | |
1da177e4 LT |
262 | printk("\n"); |
263 | } | |
264 | ||
265 | void show_stack(struct task_struct *tsk, unsigned long * rsp) | |
266 | { | |
267 | unsigned long *stack; | |
268 | int i; | |
269 | const int cpu = safe_smp_processor_id(); | |
df79efde RT |
270 | unsigned long *irqstack_end = (unsigned long *) (cpu_pda(cpu)->irqstackptr); |
271 | unsigned long *irqstack = (unsigned long *) (cpu_pda(cpu)->irqstackptr - IRQSTACKSIZE); | |
1da177e4 LT |
272 | |
273 | // debugging aid: "show_stack(NULL, NULL);" prints the | |
274 | // back trace for this cpu. | |
275 | ||
276 | if (rsp == NULL) { | |
277 | if (tsk) | |
278 | rsp = (unsigned long *)tsk->thread.rsp; | |
279 | else | |
280 | rsp = (unsigned long *)&rsp; | |
281 | } | |
282 | ||
283 | stack = rsp; | |
284 | for(i=0; i < kstack_depth_to_print; i++) { | |
285 | if (stack >= irqstack && stack <= irqstack_end) { | |
286 | if (stack == irqstack_end) { | |
287 | stack = (unsigned long *) (irqstack_end[-1]); | |
288 | printk(" <EOI> "); | |
289 | } | |
290 | } else { | |
291 | if (((long) stack & (THREAD_SIZE-1)) == 0) | |
292 | break; | |
293 | } | |
294 | if (i && ((i % 4) == 0)) | |
295 | printk("\n "); | |
296 | printk("%016lx ", *stack++); | |
35faa714 | 297 | touch_nmi_watchdog(); |
1da177e4 LT |
298 | } |
299 | show_trace((unsigned long *)rsp); | |
300 | } | |
301 | ||
302 | /* | |
303 | * The architecture-independent dump_stack generator | |
304 | */ | |
305 | void dump_stack(void) | |
306 | { | |
307 | unsigned long dummy; | |
308 | show_trace(&dummy); | |
309 | } | |
310 | ||
311 | EXPORT_SYMBOL(dump_stack); | |
312 | ||
313 | void show_registers(struct pt_regs *regs) | |
314 | { | |
315 | int i; | |
76381fee | 316 | int in_kernel = !user_mode(regs); |
1da177e4 LT |
317 | unsigned long rsp; |
318 | const int cpu = safe_smp_processor_id(); | |
df79efde | 319 | struct task_struct *cur = cpu_pda(cpu)->pcurrent; |
1da177e4 LT |
320 | |
321 | rsp = regs->rsp; | |
322 | ||
323 | printk("CPU %d ", cpu); | |
324 | __show_regs(regs); | |
325 | printk("Process %s (pid: %d, threadinfo %p, task %p)\n", | |
e4f17c43 | 326 | cur->comm, cur->pid, task_thread_info(cur), cur); |
1da177e4 LT |
327 | |
328 | /* | |
329 | * When in-kernel, we also print out the stack and code at the | |
330 | * time of the fault.. | |
331 | */ | |
332 | if (in_kernel) { | |
333 | ||
334 | printk("Stack: "); | |
335 | show_stack(NULL, (unsigned long*)rsp); | |
336 | ||
337 | printk("\nCode: "); | |
2b692a87 | 338 | if (regs->rip < PAGE_OFFSET) |
1da177e4 LT |
339 | goto bad; |
340 | ||
2b692a87 | 341 | for (i=0; i<20; i++) { |
1da177e4 | 342 | unsigned char c; |
2b692a87 | 343 | if (__get_user(c, &((unsigned char*)regs->rip)[i])) { |
1da177e4 LT |
344 | bad: |
345 | printk(" Bad RIP value."); | |
346 | break; | |
347 | } | |
348 | printk("%02x ", c); | |
349 | } | |
350 | } | |
351 | printk("\n"); | |
352 | } | |
353 | ||
354 | void handle_BUG(struct pt_regs *regs) | |
355 | { | |
356 | struct bug_frame f; | |
5f1d189f JB |
357 | long len; |
358 | const char *prefix = ""; | |
1da177e4 | 359 | |
76381fee | 360 | if (user_mode(regs)) |
1da177e4 | 361 | return; |
77a75333 | 362 | if (__copy_from_user(&f, (const void __user *) regs->rip, |
1da177e4 LT |
363 | sizeof(struct bug_frame))) |
364 | return; | |
049cdefe | 365 | if (f.filename >= 0 || |
1da177e4 LT |
366 | f.ud2[0] != 0x0f || f.ud2[1] != 0x0b) |
367 | return; | |
5f1d189f JB |
368 | len = __strnlen_user((char *)(long)f.filename, PATH_MAX) - 1; |
369 | if (len < 0 || len >= PATH_MAX) | |
049cdefe | 370 | f.filename = (int)(long)"unmapped filename"; |
5f1d189f JB |
371 | else if (len > 50) { |
372 | f.filename += len - 50; | |
373 | prefix = "..."; | |
374 | } | |
1da177e4 | 375 | printk("----------- [cut here ] --------- [please bite here ] ---------\n"); |
5f1d189f | 376 | printk(KERN_ALERT "Kernel BUG at %s%.50s:%d\n", prefix, (char *)(long)f.filename, f.line); |
1da177e4 LT |
377 | } |
378 | ||
4f60fdf6 | 379 | #ifdef CONFIG_BUG |
1da177e4 LT |
380 | void out_of_line_bug(void) |
381 | { | |
382 | BUG(); | |
383 | } | |
4f60fdf6 | 384 | #endif |
1da177e4 LT |
385 | |
386 | static DEFINE_SPINLOCK(die_lock); | |
387 | static int die_owner = -1; | |
388 | ||
eddb6fb9 | 389 | unsigned __kprobes long oops_begin(void) |
1da177e4 | 390 | { |
1209140c JB |
391 | int cpu = safe_smp_processor_id(); |
392 | unsigned long flags; | |
393 | ||
394 | /* racy, but better than risking deadlock. */ | |
395 | local_irq_save(flags); | |
1da177e4 LT |
396 | if (!spin_trylock(&die_lock)) { |
397 | if (cpu == die_owner) | |
398 | /* nested oops. should stop eventually */; | |
399 | else | |
1209140c | 400 | spin_lock(&die_lock); |
1da177e4 | 401 | } |
1209140c | 402 | die_owner = cpu; |
1da177e4 | 403 | console_verbose(); |
1209140c JB |
404 | bust_spinlocks(1); |
405 | return flags; | |
1da177e4 LT |
406 | } |
407 | ||
eddb6fb9 | 408 | void __kprobes oops_end(unsigned long flags) |
1da177e4 LT |
409 | { |
410 | die_owner = -1; | |
1209140c JB |
411 | bust_spinlocks(0); |
412 | spin_unlock_irqrestore(&die_lock, flags); | |
1da177e4 | 413 | if (panic_on_oops) |
1209140c JB |
414 | panic("Oops"); |
415 | } | |
1da177e4 | 416 | |
eddb6fb9 | 417 | void __kprobes __die(const char * str, struct pt_regs * regs, long err) |
1da177e4 LT |
418 | { |
419 | static int die_counter; | |
420 | printk(KERN_EMERG "%s: %04lx [%u] ", str, err & 0xffff,++die_counter); | |
421 | #ifdef CONFIG_PREEMPT | |
422 | printk("PREEMPT "); | |
423 | #endif | |
424 | #ifdef CONFIG_SMP | |
425 | printk("SMP "); | |
426 | #endif | |
427 | #ifdef CONFIG_DEBUG_PAGEALLOC | |
428 | printk("DEBUG_PAGEALLOC"); | |
429 | #endif | |
430 | printk("\n"); | |
6e3f3617 | 431 | notify_die(DIE_OOPS, str, regs, err, current->thread.trap_no, SIGSEGV); |
1da177e4 LT |
432 | show_registers(regs); |
433 | /* Executive summary in case the oops scrolled away */ | |
434 | printk(KERN_ALERT "RIP "); | |
435 | printk_address(regs->rip); | |
436 | printk(" RSP <%016lx>\n", regs->rsp); | |
8bcc5280 VG |
437 | if (kexec_should_crash(current)) |
438 | crash_kexec(regs); | |
1da177e4 LT |
439 | } |
440 | ||
441 | void die(const char * str, struct pt_regs * regs, long err) | |
442 | { | |
1209140c JB |
443 | unsigned long flags = oops_begin(); |
444 | ||
1da177e4 LT |
445 | handle_BUG(regs); |
446 | __die(str, regs, err); | |
1209140c | 447 | oops_end(flags); |
1da177e4 LT |
448 | do_exit(SIGSEGV); |
449 | } | |
1da177e4 | 450 | |
eddb6fb9 | 451 | void __kprobes die_nmi(char *str, struct pt_regs *regs) |
1da177e4 | 452 | { |
1209140c JB |
453 | unsigned long flags = oops_begin(); |
454 | ||
1da177e4 LT |
455 | /* |
456 | * We are in trouble anyway, lets at least try | |
457 | * to get a message out. | |
458 | */ | |
459 | printk(str, safe_smp_processor_id()); | |
460 | show_registers(regs); | |
8bcc5280 VG |
461 | if (kexec_should_crash(current)) |
462 | crash_kexec(regs); | |
1da177e4 LT |
463 | if (panic_on_timeout || panic_on_oops) |
464 | panic("nmi watchdog"); | |
465 | printk("console shuts up ...\n"); | |
1209140c | 466 | oops_end(flags); |
1da177e4 LT |
467 | do_exit(SIGSEGV); |
468 | } | |
469 | ||
0f2fbdcb PP |
470 | static void __kprobes do_trap(int trapnr, int signr, char *str, |
471 | struct pt_regs * regs, long error_code, | |
472 | siginfo_t *info) | |
1da177e4 | 473 | { |
6e3f3617 JB |
474 | struct task_struct *tsk = current; |
475 | ||
1da177e4 LT |
476 | conditional_sti(regs); |
477 | ||
6e3f3617 JB |
478 | tsk->thread.error_code = error_code; |
479 | tsk->thread.trap_no = trapnr; | |
1da177e4 | 480 | |
6e3f3617 | 481 | if (user_mode(regs)) { |
1da177e4 LT |
482 | if (exception_trace && unhandled_signal(tsk, signr)) |
483 | printk(KERN_INFO | |
484 | "%s[%d] trap %s rip:%lx rsp:%lx error:%lx\n", | |
485 | tsk->comm, tsk->pid, str, | |
2b692a87 | 486 | regs->rip, regs->rsp, error_code); |
1da177e4 | 487 | |
1da177e4 LT |
488 | if (info) |
489 | force_sig_info(signr, info, tsk); | |
490 | else | |
491 | force_sig(signr, tsk); | |
492 | return; | |
493 | } | |
494 | ||
495 | ||
496 | /* kernel trap */ | |
497 | { | |
498 | const struct exception_table_entry *fixup; | |
499 | fixup = search_exception_tables(regs->rip); | |
2b692a87 | 500 | if (fixup) |
1da177e4 | 501 | regs->rip = fixup->fixup; |
2b692a87 | 502 | else |
1da177e4 LT |
503 | die(str, regs, error_code); |
504 | return; | |
505 | } | |
506 | } | |
507 | ||
508 | #define DO_ERROR(trapnr, signr, str, name) \ | |
509 | asmlinkage void do_##name(struct pt_regs * regs, long error_code) \ | |
510 | { \ | |
511 | if (notify_die(DIE_TRAP, str, regs, error_code, trapnr, signr) \ | |
512 | == NOTIFY_STOP) \ | |
513 | return; \ | |
514 | do_trap(trapnr, signr, str, regs, error_code, NULL); \ | |
515 | } | |
516 | ||
517 | #define DO_ERROR_INFO(trapnr, signr, str, name, sicode, siaddr) \ | |
518 | asmlinkage void do_##name(struct pt_regs * regs, long error_code) \ | |
519 | { \ | |
520 | siginfo_t info; \ | |
521 | info.si_signo = signr; \ | |
522 | info.si_errno = 0; \ | |
523 | info.si_code = sicode; \ | |
524 | info.si_addr = (void __user *)siaddr; \ | |
525 | if (notify_die(DIE_TRAP, str, regs, error_code, trapnr, signr) \ | |
526 | == NOTIFY_STOP) \ | |
527 | return; \ | |
528 | do_trap(trapnr, signr, str, regs, error_code, &info); \ | |
529 | } | |
530 | ||
531 | DO_ERROR_INFO( 0, SIGFPE, "divide error", divide_error, FPE_INTDIV, regs->rip) | |
532 | DO_ERROR( 4, SIGSEGV, "overflow", overflow) | |
533 | DO_ERROR( 5, SIGSEGV, "bounds", bounds) | |
100c0e36 | 534 | DO_ERROR_INFO( 6, SIGILL, "invalid opcode", invalid_op, ILL_ILLOPN, regs->rip) |
1da177e4 LT |
535 | DO_ERROR( 7, SIGSEGV, "device not available", device_not_available) |
536 | DO_ERROR( 9, SIGFPE, "coprocessor segment overrun", coprocessor_segment_overrun) | |
537 | DO_ERROR(10, SIGSEGV, "invalid TSS", invalid_TSS) | |
538 | DO_ERROR(11, SIGBUS, "segment not present", segment_not_present) | |
539 | DO_ERROR_INFO(17, SIGBUS, "alignment check", alignment_check, BUS_ADRALN, 0) | |
540 | DO_ERROR(18, SIGSEGV, "reserved", reserved) | |
6fefb0d1 | 541 | DO_ERROR(12, SIGBUS, "stack segment", stack_segment) |
eca37c18 JB |
542 | |
543 | asmlinkage void do_double_fault(struct pt_regs * regs, long error_code) | |
544 | { | |
545 | static const char str[] = "double fault"; | |
546 | struct task_struct *tsk = current; | |
547 | ||
548 | /* Return not checked because double check cannot be ignored */ | |
549 | notify_die(DIE_TRAP, str, regs, error_code, 8, SIGSEGV); | |
550 | ||
551 | tsk->thread.error_code = error_code; | |
552 | tsk->thread.trap_no = 8; | |
553 | ||
554 | /* This is always a kernel trap and never fixable (and thus must | |
555 | never return). */ | |
556 | for (;;) | |
557 | die(str, regs, error_code); | |
558 | } | |
1da177e4 | 559 | |
0f2fbdcb PP |
560 | asmlinkage void __kprobes do_general_protection(struct pt_regs * regs, |
561 | long error_code) | |
1da177e4 | 562 | { |
6e3f3617 JB |
563 | struct task_struct *tsk = current; |
564 | ||
1da177e4 LT |
565 | conditional_sti(regs); |
566 | ||
6e3f3617 JB |
567 | tsk->thread.error_code = error_code; |
568 | tsk->thread.trap_no = 13; | |
1da177e4 | 569 | |
6e3f3617 | 570 | if (user_mode(regs)) { |
1da177e4 LT |
571 | if (exception_trace && unhandled_signal(tsk, SIGSEGV)) |
572 | printk(KERN_INFO | |
573 | "%s[%d] general protection rip:%lx rsp:%lx error:%lx\n", | |
574 | tsk->comm, tsk->pid, | |
2b692a87 | 575 | regs->rip, regs->rsp, error_code); |
1da177e4 | 576 | |
1da177e4 LT |
577 | force_sig(SIGSEGV, tsk); |
578 | return; | |
579 | } | |
580 | ||
581 | /* kernel gp */ | |
582 | { | |
583 | const struct exception_table_entry *fixup; | |
584 | fixup = search_exception_tables(regs->rip); | |
585 | if (fixup) { | |
586 | regs->rip = fixup->fixup; | |
587 | return; | |
588 | } | |
589 | if (notify_die(DIE_GPF, "general protection fault", regs, | |
590 | error_code, 13, SIGSEGV) == NOTIFY_STOP) | |
591 | return; | |
592 | die("general protection fault", regs, error_code); | |
593 | } | |
594 | } | |
595 | ||
eddb6fb9 AK |
596 | static __kprobes void |
597 | mem_parity_error(unsigned char reason, struct pt_regs * regs) | |
1da177e4 LT |
598 | { |
599 | printk("Uhhuh. NMI received. Dazed and confused, but trying to continue\n"); | |
600 | printk("You probably have a hardware problem with your RAM chips\n"); | |
601 | ||
602 | /* Clear and disable the memory parity error line. */ | |
603 | reason = (reason & 0xf) | 4; | |
604 | outb(reason, 0x61); | |
605 | } | |
606 | ||
eddb6fb9 AK |
607 | static __kprobes void |
608 | io_check_error(unsigned char reason, struct pt_regs * regs) | |
1da177e4 LT |
609 | { |
610 | printk("NMI: IOCK error (debug interrupt?)\n"); | |
611 | show_registers(regs); | |
612 | ||
613 | /* Re-enable the IOCK line, wait for a few seconds */ | |
614 | reason = (reason & 0xf) | 8; | |
615 | outb(reason, 0x61); | |
616 | mdelay(2000); | |
617 | reason &= ~8; | |
618 | outb(reason, 0x61); | |
619 | } | |
620 | ||
eddb6fb9 AK |
621 | static __kprobes void |
622 | unknown_nmi_error(unsigned char reason, struct pt_regs * regs) | |
1da177e4 LT |
623 | { printk("Uhhuh. NMI received for unknown reason %02x.\n", reason); |
624 | printk("Dazed and confused, but trying to continue\n"); | |
625 | printk("Do you have a strange power saving mode enabled?\n"); | |
626 | } | |
627 | ||
6fefb0d1 AK |
628 | /* Runs on IST stack. This code must keep interrupts off all the time. |
629 | Nested NMIs are prevented by the CPU. */ | |
eddb6fb9 | 630 | asmlinkage __kprobes void default_do_nmi(struct pt_regs *regs) |
1da177e4 LT |
631 | { |
632 | unsigned char reason = 0; | |
76e4f660 AR |
633 | int cpu; |
634 | ||
635 | cpu = smp_processor_id(); | |
1da177e4 LT |
636 | |
637 | /* Only the BSP gets external NMIs from the system. */ | |
76e4f660 | 638 | if (!cpu) |
1da177e4 LT |
639 | reason = get_nmi_reason(); |
640 | ||
641 | if (!(reason & 0xc0)) { | |
6e3f3617 | 642 | if (notify_die(DIE_NMI_IPI, "nmi_ipi", regs, reason, 2, SIGINT) |
1da177e4 LT |
643 | == NOTIFY_STOP) |
644 | return; | |
645 | #ifdef CONFIG_X86_LOCAL_APIC | |
646 | /* | |
647 | * Ok, so this is none of the documented NMI sources, | |
648 | * so it must be the NMI watchdog. | |
649 | */ | |
650 | if (nmi_watchdog > 0) { | |
651 | nmi_watchdog_tick(regs,reason); | |
652 | return; | |
653 | } | |
654 | #endif | |
655 | unknown_nmi_error(reason, regs); | |
656 | return; | |
657 | } | |
6e3f3617 | 658 | if (notify_die(DIE_NMI, "nmi", regs, reason, 2, SIGINT) == NOTIFY_STOP) |
1da177e4 LT |
659 | return; |
660 | ||
661 | /* AK: following checks seem to be broken on modern chipsets. FIXME */ | |
662 | ||
663 | if (reason & 0x80) | |
664 | mem_parity_error(reason, regs); | |
665 | if (reason & 0x40) | |
666 | io_check_error(reason, regs); | |
667 | } | |
668 | ||
b556b35e | 669 | /* runs on IST stack. */ |
0f2fbdcb | 670 | asmlinkage void __kprobes do_int3(struct pt_regs * regs, long error_code) |
1da177e4 LT |
671 | { |
672 | if (notify_die(DIE_INT3, "int3", regs, error_code, 3, SIGTRAP) == NOTIFY_STOP) { | |
673 | return; | |
674 | } | |
675 | do_trap(3, SIGTRAP, "int3", regs, error_code, NULL); | |
676 | return; | |
677 | } | |
678 | ||
6fefb0d1 AK |
679 | /* Help handler running on IST stack to switch back to user stack |
680 | for scheduling or signal handling. The actual stack switch is done in | |
681 | entry.S */ | |
eddb6fb9 | 682 | asmlinkage __kprobes struct pt_regs *sync_regs(struct pt_regs *eregs) |
6fefb0d1 AK |
683 | { |
684 | struct pt_regs *regs = eregs; | |
685 | /* Did already sync */ | |
686 | if (eregs == (struct pt_regs *)eregs->rsp) | |
687 | ; | |
688 | /* Exception from user space */ | |
76381fee | 689 | else if (user_mode(eregs)) |
bb049232 | 690 | regs = task_pt_regs(current); |
6fefb0d1 AK |
691 | /* Exception from kernel and interrupts are enabled. Move to |
692 | kernel process stack. */ | |
693 | else if (eregs->eflags & X86_EFLAGS_IF) | |
694 | regs = (struct pt_regs *)(eregs->rsp -= sizeof(struct pt_regs)); | |
695 | if (eregs != regs) | |
696 | *regs = *eregs; | |
697 | return regs; | |
698 | } | |
699 | ||
1da177e4 | 700 | /* runs on IST stack. */ |
0f2fbdcb PP |
701 | asmlinkage void __kprobes do_debug(struct pt_regs * regs, |
702 | unsigned long error_code) | |
1da177e4 | 703 | { |
1da177e4 LT |
704 | unsigned long condition; |
705 | struct task_struct *tsk = current; | |
706 | siginfo_t info; | |
707 | ||
e9129e56 | 708 | get_debugreg(condition, 6); |
1da177e4 LT |
709 | |
710 | if (notify_die(DIE_DEBUG, "debug", regs, condition, error_code, | |
daeeafec | 711 | SIGTRAP) == NOTIFY_STOP) |
6fefb0d1 | 712 | return; |
daeeafec | 713 | |
a65d17c9 | 714 | preempt_conditional_sti(regs); |
1da177e4 LT |
715 | |
716 | /* Mask out spurious debug traps due to lazy DR7 setting */ | |
717 | if (condition & (DR_TRAP0|DR_TRAP1|DR_TRAP2|DR_TRAP3)) { | |
718 | if (!tsk->thread.debugreg7) { | |
719 | goto clear_dr7; | |
720 | } | |
721 | } | |
722 | ||
723 | tsk->thread.debugreg6 = condition; | |
724 | ||
725 | /* Mask out spurious TF errors due to lazy TF clearing */ | |
daeeafec | 726 | if (condition & DR_STEP) { |
1da177e4 LT |
727 | /* |
728 | * The TF error should be masked out only if the current | |
729 | * process is not traced and if the TRAP flag has been set | |
730 | * previously by a tracing process (condition detected by | |
731 | * the PT_DTRACE flag); remember that the i386 TRAP flag | |
732 | * can be modified by the process itself in user mode, | |
733 | * allowing programs to debug themselves without the ptrace() | |
734 | * interface. | |
735 | */ | |
76381fee | 736 | if (!user_mode(regs)) |
1da177e4 | 737 | goto clear_TF_reenable; |
be61bff7 AK |
738 | /* |
739 | * Was the TF flag set by a debugger? If so, clear it now, | |
740 | * so that register information is correct. | |
741 | */ | |
742 | if (tsk->ptrace & PT_DTRACE) { | |
743 | regs->eflags &= ~TF_MASK; | |
744 | tsk->ptrace &= ~PT_DTRACE; | |
745 | } | |
1da177e4 LT |
746 | } |
747 | ||
748 | /* Ok, finally something we can handle */ | |
749 | tsk->thread.trap_no = 1; | |
750 | tsk->thread.error_code = error_code; | |
751 | info.si_signo = SIGTRAP; | |
752 | info.si_errno = 0; | |
753 | info.si_code = TRAP_BRKPT; | |
01b8faae JB |
754 | info.si_addr = user_mode(regs) ? (void __user *)regs->rip : NULL; |
755 | force_sig_info(SIGTRAP, &info, tsk); | |
1da177e4 | 756 | |
1da177e4 | 757 | clear_dr7: |
e9129e56 | 758 | set_debugreg(0UL, 7); |
a65d17c9 | 759 | preempt_conditional_cli(regs); |
6fefb0d1 | 760 | return; |
1da177e4 LT |
761 | |
762 | clear_TF_reenable: | |
763 | set_tsk_thread_flag(tsk, TIF_SINGLESTEP); | |
1da177e4 | 764 | regs->eflags &= ~TF_MASK; |
a65d17c9 | 765 | preempt_conditional_cli(regs); |
1da177e4 LT |
766 | } |
767 | ||
6e3f3617 | 768 | static int kernel_math_error(struct pt_regs *regs, const char *str, int trapnr) |
1da177e4 LT |
769 | { |
770 | const struct exception_table_entry *fixup; | |
771 | fixup = search_exception_tables(regs->rip); | |
772 | if (fixup) { | |
773 | regs->rip = fixup->fixup; | |
774 | return 1; | |
775 | } | |
6e3f3617 | 776 | notify_die(DIE_GPF, str, regs, 0, trapnr, SIGFPE); |
3a848f63 | 777 | /* Illegal floating point operation in the kernel */ |
6e3f3617 | 778 | current->thread.trap_no = trapnr; |
1da177e4 | 779 | die(str, regs, 0); |
1da177e4 LT |
780 | return 0; |
781 | } | |
782 | ||
783 | /* | |
784 | * Note that we play around with the 'TS' bit in an attempt to get | |
785 | * the correct behaviour even in the presence of the asynchronous | |
786 | * IRQ13 behaviour | |
787 | */ | |
788 | asmlinkage void do_coprocessor_error(struct pt_regs *regs) | |
789 | { | |
790 | void __user *rip = (void __user *)(regs->rip); | |
791 | struct task_struct * task; | |
792 | siginfo_t info; | |
793 | unsigned short cwd, swd; | |
794 | ||
795 | conditional_sti(regs); | |
76381fee | 796 | if (!user_mode(regs) && |
6e3f3617 | 797 | kernel_math_error(regs, "kernel x87 math error", 16)) |
1da177e4 LT |
798 | return; |
799 | ||
800 | /* | |
801 | * Save the info for the exception handler and clear the error. | |
802 | */ | |
803 | task = current; | |
804 | save_init_fpu(task); | |
805 | task->thread.trap_no = 16; | |
806 | task->thread.error_code = 0; | |
807 | info.si_signo = SIGFPE; | |
808 | info.si_errno = 0; | |
809 | info.si_code = __SI_FAULT; | |
810 | info.si_addr = rip; | |
811 | /* | |
812 | * (~cwd & swd) will mask out exceptions that are not set to unmasked | |
813 | * status. 0x3f is the exception bits in these regs, 0x200 is the | |
814 | * C1 reg you need in case of a stack fault, 0x040 is the stack | |
815 | * fault bit. We should only be taking one exception at a time, | |
816 | * so if this combination doesn't produce any single exception, | |
817 | * then we have a bad program that isn't synchronizing its FPU usage | |
818 | * and it will suffer the consequences since we won't be able to | |
819 | * fully reproduce the context of the exception | |
820 | */ | |
821 | cwd = get_fpu_cwd(task); | |
822 | swd = get_fpu_swd(task); | |
ff347b22 | 823 | switch (swd & ~cwd & 0x3f) { |
1da177e4 LT |
824 | case 0x000: |
825 | default: | |
826 | break; | |
827 | case 0x001: /* Invalid Op */ | |
ff347b22 CE |
828 | /* |
829 | * swd & 0x240 == 0x040: Stack Underflow | |
830 | * swd & 0x240 == 0x240: Stack Overflow | |
831 | * User must clear the SF bit (0x40) if set | |
832 | */ | |
1da177e4 LT |
833 | info.si_code = FPE_FLTINV; |
834 | break; | |
835 | case 0x002: /* Denormalize */ | |
836 | case 0x010: /* Underflow */ | |
837 | info.si_code = FPE_FLTUND; | |
838 | break; | |
839 | case 0x004: /* Zero Divide */ | |
840 | info.si_code = FPE_FLTDIV; | |
841 | break; | |
842 | case 0x008: /* Overflow */ | |
843 | info.si_code = FPE_FLTOVF; | |
844 | break; | |
845 | case 0x020: /* Precision */ | |
846 | info.si_code = FPE_FLTRES; | |
847 | break; | |
848 | } | |
849 | force_sig_info(SIGFPE, &info, task); | |
850 | } | |
851 | ||
852 | asmlinkage void bad_intr(void) | |
853 | { | |
854 | printk("bad interrupt"); | |
855 | } | |
856 | ||
857 | asmlinkage void do_simd_coprocessor_error(struct pt_regs *regs) | |
858 | { | |
859 | void __user *rip = (void __user *)(regs->rip); | |
860 | struct task_struct * task; | |
861 | siginfo_t info; | |
862 | unsigned short mxcsr; | |
863 | ||
864 | conditional_sti(regs); | |
76381fee | 865 | if (!user_mode(regs) && |
6e3f3617 | 866 | kernel_math_error(regs, "kernel simd math error", 19)) |
1da177e4 LT |
867 | return; |
868 | ||
869 | /* | |
870 | * Save the info for the exception handler and clear the error. | |
871 | */ | |
872 | task = current; | |
873 | save_init_fpu(task); | |
874 | task->thread.trap_no = 19; | |
875 | task->thread.error_code = 0; | |
876 | info.si_signo = SIGFPE; | |
877 | info.si_errno = 0; | |
878 | info.si_code = __SI_FAULT; | |
879 | info.si_addr = rip; | |
880 | /* | |
881 | * The SIMD FPU exceptions are handled a little differently, as there | |
882 | * is only a single status/control register. Thus, to determine which | |
883 | * unmasked exception was caught we must mask the exception mask bits | |
884 | * at 0x1f80, and then use these to mask the exception bits at 0x3f. | |
885 | */ | |
886 | mxcsr = get_fpu_mxcsr(task); | |
887 | switch (~((mxcsr & 0x1f80) >> 7) & (mxcsr & 0x3f)) { | |
888 | case 0x000: | |
889 | default: | |
890 | break; | |
891 | case 0x001: /* Invalid Op */ | |
892 | info.si_code = FPE_FLTINV; | |
893 | break; | |
894 | case 0x002: /* Denormalize */ | |
895 | case 0x010: /* Underflow */ | |
896 | info.si_code = FPE_FLTUND; | |
897 | break; | |
898 | case 0x004: /* Zero Divide */ | |
899 | info.si_code = FPE_FLTDIV; | |
900 | break; | |
901 | case 0x008: /* Overflow */ | |
902 | info.si_code = FPE_FLTOVF; | |
903 | break; | |
904 | case 0x020: /* Precision */ | |
905 | info.si_code = FPE_FLTRES; | |
906 | break; | |
907 | } | |
908 | force_sig_info(SIGFPE, &info, task); | |
909 | } | |
910 | ||
911 | asmlinkage void do_spurious_interrupt_bug(struct pt_regs * regs) | |
912 | { | |
913 | } | |
914 | ||
915 | asmlinkage void __attribute__((weak)) smp_thermal_interrupt(void) | |
89b831ef JS |
916 | { |
917 | } | |
918 | ||
919 | asmlinkage void __attribute__((weak)) mce_threshold_interrupt(void) | |
1da177e4 LT |
920 | { |
921 | } | |
922 | ||
923 | /* | |
924 | * 'math_state_restore()' saves the current math information in the | |
925 | * old math state array, and gets the new ones from the current task | |
926 | * | |
927 | * Careful.. There are problems with IBM-designed IRQ13 behaviour. | |
928 | * Don't touch unless you *really* know how it works. | |
929 | */ | |
930 | asmlinkage void math_state_restore(void) | |
931 | { | |
932 | struct task_struct *me = current; | |
933 | clts(); /* Allow maths ops (or we recurse) */ | |
934 | ||
935 | if (!used_math()) | |
936 | init_fpu(me); | |
937 | restore_fpu_checking(&me->thread.i387.fxsave); | |
e4f17c43 | 938 | task_thread_info(me)->status |= TS_USEDFPU; |
1da177e4 LT |
939 | } |
940 | ||
1da177e4 LT |
941 | void __init trap_init(void) |
942 | { | |
943 | set_intr_gate(0,÷_error); | |
944 | set_intr_gate_ist(1,&debug,DEBUG_STACK); | |
945 | set_intr_gate_ist(2,&nmi,NMI_STACK); | |
b556b35e | 946 | set_system_gate_ist(3,&int3,DEBUG_STACK); /* int3 can be called from all */ |
0a521588 JB |
947 | set_system_gate(4,&overflow); /* int4 can be called from all */ |
948 | set_intr_gate(5,&bounds); | |
1da177e4 LT |
949 | set_intr_gate(6,&invalid_op); |
950 | set_intr_gate(7,&device_not_available); | |
951 | set_intr_gate_ist(8,&double_fault, DOUBLEFAULT_STACK); | |
952 | set_intr_gate(9,&coprocessor_segment_overrun); | |
953 | set_intr_gate(10,&invalid_TSS); | |
954 | set_intr_gate(11,&segment_not_present); | |
955 | set_intr_gate_ist(12,&stack_segment,STACKFAULT_STACK); | |
956 | set_intr_gate(13,&general_protection); | |
957 | set_intr_gate(14,&page_fault); | |
958 | set_intr_gate(15,&spurious_interrupt_bug); | |
959 | set_intr_gate(16,&coprocessor_error); | |
960 | set_intr_gate(17,&alignment_check); | |
961 | #ifdef CONFIG_X86_MCE | |
962 | set_intr_gate_ist(18,&machine_check, MCE_STACK); | |
963 | #endif | |
964 | set_intr_gate(19,&simd_coprocessor_error); | |
965 | ||
966 | #ifdef CONFIG_IA32_EMULATION | |
967 | set_system_gate(IA32_SYSCALL_VECTOR, ia32_syscall); | |
968 | #endif | |
969 | ||
1da177e4 LT |
970 | /* |
971 | * Should be a barrier for any external CPU state. | |
972 | */ | |
973 | cpu_init(); | |
974 | } | |
975 | ||
976 | ||
977 | /* Actual parsing is done early in setup.c. */ | |
978 | static int __init oops_dummy(char *s) | |
979 | { | |
980 | panic_on_oops = 1; | |
9b41046c | 981 | return 1; |
1da177e4 LT |
982 | } |
983 | __setup("oops=", oops_dummy); | |
984 | ||
985 | static int __init kstack_setup(char *s) | |
986 | { | |
987 | kstack_depth_to_print = simple_strtoul(s,NULL,0); | |
9b41046c | 988 | return 1; |
1da177e4 LT |
989 | } |
990 | __setup("kstack=", kstack_setup); | |
991 |