blob: 1a9094dab68255680c4da834ac06e708a4359438 [file] [log] [blame]
Linus Torvalds1da177e2005-04-16 15:20:36 -07001/*
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>
akpm@osdl.org35faa712005-04-16 15:24:54 -070031#include <linux/nmi.h>
Prasanna S Panchamukhi0f2fbdc2005-09-06 15:19:28 -070032#include <linux/kprobes.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070033
34#include <asm/system.h>
35#include <asm/uaccess.h>
36#include <asm/io.h>
37#include <asm/atomic.h>
38#include <asm/debugreg.h>
39#include <asm/desc.h>
40#include <asm/i387.h>
41#include <asm/kdebug.h>
42#include <asm/processor.h>
43
44#include <asm/smp.h>
45#include <asm/pgalloc.h>
46#include <asm/pda.h>
47#include <asm/proto.h>
48#include <asm/nmi.h>
49
Linus Torvalds1da177e2005-04-16 15:20:36 -070050extern struct gate_struct idt_table[256];
51
52asmlinkage void divide_error(void);
53asmlinkage void debug(void);
54asmlinkage void nmi(void);
55asmlinkage void int3(void);
56asmlinkage void overflow(void);
57asmlinkage void bounds(void);
58asmlinkage void invalid_op(void);
59asmlinkage void device_not_available(void);
60asmlinkage void double_fault(void);
61asmlinkage void coprocessor_segment_overrun(void);
62asmlinkage void invalid_TSS(void);
63asmlinkage void segment_not_present(void);
64asmlinkage void stack_segment(void);
65asmlinkage void general_protection(void);
66asmlinkage void page_fault(void);
67asmlinkage void coprocessor_error(void);
68asmlinkage void simd_coprocessor_error(void);
69asmlinkage void reserved(void);
70asmlinkage void alignment_check(void);
71asmlinkage void machine_check(void);
72asmlinkage void spurious_interrupt_bug(void);
73asmlinkage void call_debug(void);
74
75struct notifier_block *die_chain;
76static DEFINE_SPINLOCK(die_notifier_lock);
77
78int register_die_notifier(struct notifier_block *nb)
79{
80 int err = 0;
81 unsigned long flags;
82 spin_lock_irqsave(&die_notifier_lock, flags);
83 err = notifier_chain_register(&die_chain, nb);
84 spin_unlock_irqrestore(&die_notifier_lock, flags);
85 return err;
86}
87
88static inline void conditional_sti(struct pt_regs *regs)
89{
90 if (regs->eflags & X86_EFLAGS_IF)
91 local_irq_enable();
92}
93
94static int kstack_depth_to_print = 10;
95
96#ifdef CONFIG_KALLSYMS
97#include <linux/kallsyms.h>
98int printk_address(unsigned long address)
99{
100 unsigned long offset = 0, symsize;
101 const char *symname;
102 char *modname;
103 char *delim = ":";
104 char namebuf[128];
105
106 symname = kallsyms_lookup(address, &symsize, &offset, &modname, namebuf);
107 if (!symname)
108 return printk("[<%016lx>]", address);
109 if (!modname)
110 modname = delim = "";
111 return printk("<%016lx>{%s%s%s%s%+ld}",
112 address,delim,modname,delim,symname,offset);
113}
114#else
115int printk_address(unsigned long address)
116{
117 return printk("[<%016lx>]", address);
118}
119#endif
120
Andi Kleen0a658002005-04-16 15:25:17 -0700121static unsigned long *in_exception_stack(unsigned cpu, unsigned long stack,
122 unsigned *usedp, const char **idp)
123{
Jan Beulichb556b352006-01-11 22:43:00 +0100124 static char ids[][8] = {
Andi Kleen0a658002005-04-16 15:25:17 -0700125 [DEBUG_STACK - 1] = "#DB",
126 [NMI_STACK - 1] = "NMI",
127 [DOUBLEFAULT_STACK - 1] = "#DF",
128 [STACKFAULT_STACK - 1] = "#SS",
129 [MCE_STACK - 1] = "#MC",
Jan Beulichb556b352006-01-11 22:43:00 +0100130#if DEBUG_STKSZ > EXCEPTION_STKSZ
131 [N_EXCEPTION_STACKS ... N_EXCEPTION_STACKS + DEBUG_STKSZ / EXCEPTION_STKSZ - 2] = "#DB[?]"
132#endif
Andi Kleen0a658002005-04-16 15:25:17 -0700133 };
134 unsigned k;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700135
Andi Kleen0a658002005-04-16 15:25:17 -0700136 for (k = 0; k < N_EXCEPTION_STACKS; k++) {
137 unsigned long end;
138
Jan Beulichb556b352006-01-11 22:43:00 +0100139 switch (k + 1) {
140#if DEBUG_STKSZ > EXCEPTION_STKSZ
141 case DEBUG_STACK:
142 end = cpu_pda[cpu].debugstack + DEBUG_STKSZ;
143 break;
144#endif
145 default:
146 end = per_cpu(init_tss, cpu).ist[k];
147 break;
148 }
Andi Kleen0a658002005-04-16 15:25:17 -0700149 if (stack >= end)
150 continue;
151 if (stack >= end - EXCEPTION_STKSZ) {
152 if (*usedp & (1U << k))
153 break;
154 *usedp |= 1U << k;
155 *idp = ids[k];
156 return (unsigned long *)end;
157 }
Jan Beulichb556b352006-01-11 22:43:00 +0100158#if DEBUG_STKSZ > EXCEPTION_STKSZ
159 if (k == DEBUG_STACK - 1 && stack >= end - DEBUG_STKSZ) {
160 unsigned j = N_EXCEPTION_STACKS - 1;
161
162 do {
163 ++j;
164 end -= EXCEPTION_STKSZ;
165 ids[j][4] = '1' + (j - N_EXCEPTION_STACKS);
166 } while (stack < end - EXCEPTION_STKSZ);
167 if (*usedp & (1U << j))
168 break;
169 *usedp |= 1U << j;
170 *idp = ids[j];
171 return (unsigned long *)end;
172 }
173#endif
Linus Torvalds1da177e2005-04-16 15:20:36 -0700174 }
175 return NULL;
Andi Kleen0a658002005-04-16 15:25:17 -0700176}
Linus Torvalds1da177e2005-04-16 15:20:36 -0700177
178/*
179 * x86-64 can have upto three kernel stacks:
180 * process stack
181 * interrupt stack
Andi Kleen0a658002005-04-16 15:25:17 -0700182 * severe exception (double fault, nmi, stack fault, debug, mce) hardware stack
Linus Torvalds1da177e2005-04-16 15:20:36 -0700183 */
184
185void show_trace(unsigned long *stack)
186{
187 unsigned long addr;
Andi Kleen0a658002005-04-16 15:25:17 -0700188 const unsigned cpu = safe_smp_processor_id();
189 unsigned long *irqstack_end = (unsigned long *)cpu_pda[cpu].irqstackptr;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700190 int i;
Andi Kleen0a658002005-04-16 15:25:17 -0700191 unsigned used = 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700192
193 printk("\nCall Trace:");
Andi Kleen0a658002005-04-16 15:25:17 -0700194
195#define HANDLE_STACK(cond) \
196 do while (cond) { \
197 addr = *stack++; \
198 if (kernel_text_address(addr)) { \
199 /* \
200 * If the address is either in the text segment of the \
201 * kernel, or in the region which contains vmalloc'ed \
202 * memory, it *may* be the address of a calling \
203 * routine; if so, print it so that someone tracing \
204 * down the cause of the crash will be able to figure \
205 * out the call path that was taken. \
206 */ \
207 i += printk_address(addr); \
208 if (i > 50) { \
209 printk("\n "); \
210 i = 0; \
211 } \
212 else \
213 i += printk(" "); \
214 } \
215 } while (0)
216
217 for(i = 0; ; ) {
218 const char *id;
219 unsigned long *estack_end;
220 estack_end = in_exception_stack(cpu, (unsigned long)stack,
221 &used, &id);
222
223 if (estack_end) {
224 i += printk(" <%s> ", id);
225 HANDLE_STACK (stack < estack_end);
226 i += printk(" <EOE> ");
227 stack = (unsigned long *) estack_end[-2];
228 continue;
229 }
230 if (irqstack_end) {
231 unsigned long *irqstack;
232 irqstack = irqstack_end -
233 (IRQSTACKSIZE - 64) / sizeof(*irqstack);
234
235 if (stack >= irqstack && stack < irqstack_end) {
236 i += printk(" <IRQ> ");
237 HANDLE_STACK (stack < irqstack_end);
238 stack = (unsigned long *) (irqstack_end[-1]);
239 irqstack_end = NULL;
240 i += printk(" <EOI> ");
241 continue;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700242 }
243 }
Andi Kleen0a658002005-04-16 15:25:17 -0700244 break;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700245 }
Andi Kleen0a658002005-04-16 15:25:17 -0700246
247 HANDLE_STACK (((long) stack & (THREAD_SIZE-1)) != 0);
248#undef HANDLE_STACK
Linus Torvalds1da177e2005-04-16 15:20:36 -0700249 printk("\n");
250}
251
252void show_stack(struct task_struct *tsk, unsigned long * rsp)
253{
254 unsigned long *stack;
255 int i;
256 const int cpu = safe_smp_processor_id();
257 unsigned long *irqstack_end = (unsigned long *) (cpu_pda[cpu].irqstackptr);
258 unsigned long *irqstack = (unsigned long *) (cpu_pda[cpu].irqstackptr - IRQSTACKSIZE);
259
260 // debugging aid: "show_stack(NULL, NULL);" prints the
261 // back trace for this cpu.
262
263 if (rsp == NULL) {
264 if (tsk)
265 rsp = (unsigned long *)tsk->thread.rsp;
266 else
267 rsp = (unsigned long *)&rsp;
268 }
269
270 stack = rsp;
271 for(i=0; i < kstack_depth_to_print; i++) {
272 if (stack >= irqstack && stack <= irqstack_end) {
273 if (stack == irqstack_end) {
274 stack = (unsigned long *) (irqstack_end[-1]);
275 printk(" <EOI> ");
276 }
277 } else {
278 if (((long) stack & (THREAD_SIZE-1)) == 0)
279 break;
280 }
281 if (i && ((i % 4) == 0))
282 printk("\n ");
283 printk("%016lx ", *stack++);
akpm@osdl.org35faa712005-04-16 15:24:54 -0700284 touch_nmi_watchdog();
Linus Torvalds1da177e2005-04-16 15:20:36 -0700285 }
286 show_trace((unsigned long *)rsp);
287}
288
289/*
290 * The architecture-independent dump_stack generator
291 */
292void dump_stack(void)
293{
294 unsigned long dummy;
295 show_trace(&dummy);
296}
297
298EXPORT_SYMBOL(dump_stack);
299
300void show_registers(struct pt_regs *regs)
301{
302 int i;
Vincent Hanquez76381fe2005-06-23 00:08:46 -0700303 int in_kernel = !user_mode(regs);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700304 unsigned long rsp;
305 const int cpu = safe_smp_processor_id();
306 struct task_struct *cur = cpu_pda[cpu].pcurrent;
307
308 rsp = regs->rsp;
309
310 printk("CPU %d ", cpu);
311 __show_regs(regs);
312 printk("Process %s (pid: %d, threadinfo %p, task %p)\n",
313 cur->comm, cur->pid, cur->thread_info, cur);
314
315 /*
316 * When in-kernel, we also print out the stack and code at the
317 * time of the fault..
318 */
319 if (in_kernel) {
320
321 printk("Stack: ");
322 show_stack(NULL, (unsigned long*)rsp);
323
324 printk("\nCode: ");
325 if(regs->rip < PAGE_OFFSET)
326 goto bad;
327
328 for(i=0;i<20;i++)
329 {
330 unsigned char c;
331 if(__get_user(c, &((unsigned char*)regs->rip)[i])) {
332bad:
333 printk(" Bad RIP value.");
334 break;
335 }
336 printk("%02x ", c);
337 }
338 }
339 printk("\n");
340}
341
342void handle_BUG(struct pt_regs *regs)
343{
344 struct bug_frame f;
345 char tmp;
346
Vincent Hanquez76381fe2005-06-23 00:08:46 -0700347 if (user_mode(regs))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700348 return;
349 if (__copy_from_user(&f, (struct bug_frame *) regs->rip,
350 sizeof(struct bug_frame)))
351 return;
Jan Beulich049cdef2005-09-12 18:49:25 +0200352 if (f.filename >= 0 ||
Linus Torvalds1da177e2005-04-16 15:20:36 -0700353 f.ud2[0] != 0x0f || f.ud2[1] != 0x0b)
354 return;
Jan Beulich049cdef2005-09-12 18:49:25 +0200355 if (__get_user(tmp, (char *)(long)f.filename))
356 f.filename = (int)(long)"unmapped filename";
Linus Torvalds1da177e2005-04-16 15:20:36 -0700357 printk("----------- [cut here ] --------- [please bite here ] ---------\n");
Jan Beulich049cdef2005-09-12 18:49:25 +0200358 printk(KERN_ALERT "Kernel BUG at %.50s:%d\n", (char *)(long)f.filename, f.line);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700359}
360
Alexander Nyberg4f60fdf2005-05-25 12:31:28 -0700361#ifdef CONFIG_BUG
Linus Torvalds1da177e2005-04-16 15:20:36 -0700362void out_of_line_bug(void)
363{
364 BUG();
365}
Alexander Nyberg4f60fdf2005-05-25 12:31:28 -0700366#endif
Linus Torvalds1da177e2005-04-16 15:20:36 -0700367
368static DEFINE_SPINLOCK(die_lock);
369static int die_owner = -1;
370
Jan Beulich12091402005-09-12 18:49:24 +0200371unsigned long oops_begin(void)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700372{
Jan Beulich12091402005-09-12 18:49:24 +0200373 int cpu = safe_smp_processor_id();
374 unsigned long flags;
375
376 /* racy, but better than risking deadlock. */
377 local_irq_save(flags);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700378 if (!spin_trylock(&die_lock)) {
379 if (cpu == die_owner)
380 /* nested oops. should stop eventually */;
381 else
Jan Beulich12091402005-09-12 18:49:24 +0200382 spin_lock(&die_lock);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700383 }
Jan Beulich12091402005-09-12 18:49:24 +0200384 die_owner = cpu;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700385 console_verbose();
Jan Beulich12091402005-09-12 18:49:24 +0200386 bust_spinlocks(1);
387 return flags;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700388}
389
Jan Beulich12091402005-09-12 18:49:24 +0200390void oops_end(unsigned long flags)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700391{
392 die_owner = -1;
Jan Beulich12091402005-09-12 18:49:24 +0200393 bust_spinlocks(0);
394 spin_unlock_irqrestore(&die_lock, flags);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700395 if (panic_on_oops)
Jan Beulich12091402005-09-12 18:49:24 +0200396 panic("Oops");
397}
Linus Torvalds1da177e2005-04-16 15:20:36 -0700398
399void __die(const char * str, struct pt_regs * regs, long err)
400{
401 static int die_counter;
402 printk(KERN_EMERG "%s: %04lx [%u] ", str, err & 0xffff,++die_counter);
403#ifdef CONFIG_PREEMPT
404 printk("PREEMPT ");
405#endif
406#ifdef CONFIG_SMP
407 printk("SMP ");
408#endif
409#ifdef CONFIG_DEBUG_PAGEALLOC
410 printk("DEBUG_PAGEALLOC");
411#endif
412 printk("\n");
Jan Beulich6e3f3612006-01-11 22:42:14 +0100413 notify_die(DIE_OOPS, str, regs, err, current->thread.trap_no, SIGSEGV);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700414 show_registers(regs);
415 /* Executive summary in case the oops scrolled away */
416 printk(KERN_ALERT "RIP ");
417 printk_address(regs->rip);
418 printk(" RSP <%016lx>\n", regs->rsp);
419}
420
421void die(const char * str, struct pt_regs * regs, long err)
422{
Jan Beulich12091402005-09-12 18:49:24 +0200423 unsigned long flags = oops_begin();
424
Linus Torvalds1da177e2005-04-16 15:20:36 -0700425 handle_BUG(regs);
426 __die(str, regs, err);
Jan Beulich12091402005-09-12 18:49:24 +0200427 oops_end(flags);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700428 do_exit(SIGSEGV);
429}
Linus Torvalds1da177e2005-04-16 15:20:36 -0700430
431void die_nmi(char *str, struct pt_regs *regs)
432{
Jan Beulich12091402005-09-12 18:49:24 +0200433 unsigned long flags = oops_begin();
434
Linus Torvalds1da177e2005-04-16 15:20:36 -0700435 /*
436 * We are in trouble anyway, lets at least try
437 * to get a message out.
438 */
439 printk(str, safe_smp_processor_id());
440 show_registers(regs);
441 if (panic_on_timeout || panic_on_oops)
442 panic("nmi watchdog");
443 printk("console shuts up ...\n");
Jan Beulich12091402005-09-12 18:49:24 +0200444 oops_end(flags);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700445 do_exit(SIGSEGV);
446}
447
Prasanna S Panchamukhi0f2fbdc2005-09-06 15:19:28 -0700448static void __kprobes do_trap(int trapnr, int signr, char *str,
449 struct pt_regs * regs, long error_code,
450 siginfo_t *info)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700451{
Jan Beulich6e3f3612006-01-11 22:42:14 +0100452 struct task_struct *tsk = current;
453
Linus Torvalds1da177e2005-04-16 15:20:36 -0700454 conditional_sti(regs);
455
Jan Beulich6e3f3612006-01-11 22:42:14 +0100456 tsk->thread.error_code = error_code;
457 tsk->thread.trap_no = trapnr;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700458
Jan Beulich6e3f3612006-01-11 22:42:14 +0100459 if (user_mode(regs)) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700460 if (exception_trace && unhandled_signal(tsk, signr))
461 printk(KERN_INFO
462 "%s[%d] trap %s rip:%lx rsp:%lx error:%lx\n",
463 tsk->comm, tsk->pid, str,
464 regs->rip,regs->rsp,error_code);
465
Linus Torvalds1da177e2005-04-16 15:20:36 -0700466 if (info)
467 force_sig_info(signr, info, tsk);
468 else
469 force_sig(signr, tsk);
470 return;
471 }
472
473
474 /* kernel trap */
475 {
476 const struct exception_table_entry *fixup;
477 fixup = search_exception_tables(regs->rip);
478 if (fixup) {
479 regs->rip = fixup->fixup;
480 } else
481 die(str, regs, error_code);
482 return;
483 }
484}
485
486#define DO_ERROR(trapnr, signr, str, name) \
487asmlinkage void do_##name(struct pt_regs * regs, long error_code) \
488{ \
489 if (notify_die(DIE_TRAP, str, regs, error_code, trapnr, signr) \
490 == NOTIFY_STOP) \
491 return; \
492 do_trap(trapnr, signr, str, regs, error_code, NULL); \
493}
494
495#define DO_ERROR_INFO(trapnr, signr, str, name, sicode, siaddr) \
496asmlinkage void do_##name(struct pt_regs * regs, long error_code) \
497{ \
498 siginfo_t info; \
499 info.si_signo = signr; \
500 info.si_errno = 0; \
501 info.si_code = sicode; \
502 info.si_addr = (void __user *)siaddr; \
503 if (notify_die(DIE_TRAP, str, regs, error_code, trapnr, signr) \
504 == NOTIFY_STOP) \
505 return; \
506 do_trap(trapnr, signr, str, regs, error_code, &info); \
507}
508
509DO_ERROR_INFO( 0, SIGFPE, "divide error", divide_error, FPE_INTDIV, regs->rip)
510DO_ERROR( 4, SIGSEGV, "overflow", overflow)
511DO_ERROR( 5, SIGSEGV, "bounds", bounds)
512DO_ERROR_INFO( 6, SIGILL, "invalid operand", invalid_op, ILL_ILLOPN, regs->rip)
513DO_ERROR( 7, SIGSEGV, "device not available", device_not_available)
514DO_ERROR( 9, SIGFPE, "coprocessor segment overrun", coprocessor_segment_overrun)
515DO_ERROR(10, SIGSEGV, "invalid TSS", invalid_TSS)
516DO_ERROR(11, SIGBUS, "segment not present", segment_not_present)
517DO_ERROR_INFO(17, SIGBUS, "alignment check", alignment_check, BUS_ADRALN, 0)
518DO_ERROR(18, SIGSEGV, "reserved", reserved)
Andi Kleen6fefb0d2005-04-16 15:25:03 -0700519DO_ERROR(12, SIGBUS, "stack segment", stack_segment)
Jan Beulicheca37c12006-01-11 22:42:17 +0100520
521asmlinkage void do_double_fault(struct pt_regs * regs, long error_code)
522{
523 static const char str[] = "double fault";
524 struct task_struct *tsk = current;
525
526 /* Return not checked because double check cannot be ignored */
527 notify_die(DIE_TRAP, str, regs, error_code, 8, SIGSEGV);
528
529 tsk->thread.error_code = error_code;
530 tsk->thread.trap_no = 8;
531
532 /* This is always a kernel trap and never fixable (and thus must
533 never return). */
534 for (;;)
535 die(str, regs, error_code);
536}
Linus Torvalds1da177e2005-04-16 15:20:36 -0700537
Prasanna S Panchamukhi0f2fbdc2005-09-06 15:19:28 -0700538asmlinkage void __kprobes do_general_protection(struct pt_regs * regs,
539 long error_code)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700540{
Jan Beulich6e3f3612006-01-11 22:42:14 +0100541 struct task_struct *tsk = current;
542
Linus Torvalds1da177e2005-04-16 15:20:36 -0700543 conditional_sti(regs);
544
Jan Beulich6e3f3612006-01-11 22:42:14 +0100545 tsk->thread.error_code = error_code;
546 tsk->thread.trap_no = 13;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700547
Jan Beulich6e3f3612006-01-11 22:42:14 +0100548 if (user_mode(regs)) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700549 if (exception_trace && unhandled_signal(tsk, SIGSEGV))
550 printk(KERN_INFO
551 "%s[%d] general protection rip:%lx rsp:%lx error:%lx\n",
552 tsk->comm, tsk->pid,
553 regs->rip,regs->rsp,error_code);
554
Linus Torvalds1da177e2005-04-16 15:20:36 -0700555 force_sig(SIGSEGV, tsk);
556 return;
557 }
558
559 /* kernel gp */
560 {
561 const struct exception_table_entry *fixup;
562 fixup = search_exception_tables(regs->rip);
563 if (fixup) {
564 regs->rip = fixup->fixup;
565 return;
566 }
567 if (notify_die(DIE_GPF, "general protection fault", regs,
568 error_code, 13, SIGSEGV) == NOTIFY_STOP)
569 return;
570 die("general protection fault", regs, error_code);
571 }
572}
573
574static void mem_parity_error(unsigned char reason, struct pt_regs * regs)
575{
576 printk("Uhhuh. NMI received. Dazed and confused, but trying to continue\n");
577 printk("You probably have a hardware problem with your RAM chips\n");
578
579 /* Clear and disable the memory parity error line. */
580 reason = (reason & 0xf) | 4;
581 outb(reason, 0x61);
582}
583
584static void io_check_error(unsigned char reason, struct pt_regs * regs)
585{
586 printk("NMI: IOCK error (debug interrupt?)\n");
587 show_registers(regs);
588
589 /* Re-enable the IOCK line, wait for a few seconds */
590 reason = (reason & 0xf) | 8;
591 outb(reason, 0x61);
592 mdelay(2000);
593 reason &= ~8;
594 outb(reason, 0x61);
595}
596
597static void unknown_nmi_error(unsigned char reason, struct pt_regs * regs)
598{ printk("Uhhuh. NMI received for unknown reason %02x.\n", reason);
599 printk("Dazed and confused, but trying to continue\n");
600 printk("Do you have a strange power saving mode enabled?\n");
601}
602
Andi Kleen6fefb0d2005-04-16 15:25:03 -0700603/* Runs on IST stack. This code must keep interrupts off all the time.
604 Nested NMIs are prevented by the CPU. */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700605asmlinkage void default_do_nmi(struct pt_regs *regs)
606{
607 unsigned char reason = 0;
Ashok Raj76e4f662005-06-25 14:55:00 -0700608 int cpu;
609
610 cpu = smp_processor_id();
Linus Torvalds1da177e2005-04-16 15:20:36 -0700611
612 /* Only the BSP gets external NMIs from the system. */
Ashok Raj76e4f662005-06-25 14:55:00 -0700613 if (!cpu)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700614 reason = get_nmi_reason();
615
616 if (!(reason & 0xc0)) {
Jan Beulich6e3f3612006-01-11 22:42:14 +0100617 if (notify_die(DIE_NMI_IPI, "nmi_ipi", regs, reason, 2, SIGINT)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700618 == NOTIFY_STOP)
619 return;
620#ifdef CONFIG_X86_LOCAL_APIC
621 /*
622 * Ok, so this is none of the documented NMI sources,
623 * so it must be the NMI watchdog.
624 */
625 if (nmi_watchdog > 0) {
626 nmi_watchdog_tick(regs,reason);
627 return;
628 }
629#endif
630 unknown_nmi_error(reason, regs);
631 return;
632 }
Jan Beulich6e3f3612006-01-11 22:42:14 +0100633 if (notify_die(DIE_NMI, "nmi", regs, reason, 2, SIGINT) == NOTIFY_STOP)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700634 return;
635
636 /* AK: following checks seem to be broken on modern chipsets. FIXME */
637
638 if (reason & 0x80)
639 mem_parity_error(reason, regs);
640 if (reason & 0x40)
641 io_check_error(reason, regs);
642}
643
Jan Beulichb556b352006-01-11 22:43:00 +0100644/* runs on IST stack. */
Prasanna S Panchamukhi0f2fbdc2005-09-06 15:19:28 -0700645asmlinkage void __kprobes do_int3(struct pt_regs * regs, long error_code)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700646{
647 if (notify_die(DIE_INT3, "int3", regs, error_code, 3, SIGTRAP) == NOTIFY_STOP) {
648 return;
649 }
650 do_trap(3, SIGTRAP, "int3", regs, error_code, NULL);
651 return;
652}
653
Andi Kleen6fefb0d2005-04-16 15:25:03 -0700654/* Help handler running on IST stack to switch back to user stack
655 for scheduling or signal handling. The actual stack switch is done in
656 entry.S */
657asmlinkage struct pt_regs *sync_regs(struct pt_regs *eregs)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700658{
Andi Kleen6fefb0d2005-04-16 15:25:03 -0700659 struct pt_regs *regs = eregs;
660 /* Did already sync */
661 if (eregs == (struct pt_regs *)eregs->rsp)
662 ;
663 /* Exception from user space */
Vincent Hanquez76381fe2005-06-23 00:08:46 -0700664 else if (user_mode(eregs))
Andi Kleen6fefb0d2005-04-16 15:25:03 -0700665 regs = ((struct pt_regs *)current->thread.rsp0) - 1;
666 /* Exception from kernel and interrupts are enabled. Move to
667 kernel process stack. */
668 else if (eregs->eflags & X86_EFLAGS_IF)
669 regs = (struct pt_regs *)(eregs->rsp -= sizeof(struct pt_regs));
670 if (eregs != regs)
671 *regs = *eregs;
672 return regs;
673}
674
675/* runs on IST stack. */
Prasanna S Panchamukhi0f2fbdc2005-09-06 15:19:28 -0700676asmlinkage void __kprobes do_debug(struct pt_regs * regs,
677 unsigned long error_code)
Andi Kleen6fefb0d2005-04-16 15:25:03 -0700678{
Linus Torvalds1da177e2005-04-16 15:20:36 -0700679 unsigned long condition;
680 struct task_struct *tsk = current;
681 siginfo_t info;
682
Vincent Hanqueze9129e52005-06-23 00:08:46 -0700683 get_debugreg(condition, 6);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700684
685 if (notify_die(DIE_DEBUG, "debug", regs, condition, error_code,
Andi Kleendaeeafe2005-04-16 15:25:13 -0700686 SIGTRAP) == NOTIFY_STOP)
Andi Kleen6fefb0d2005-04-16 15:25:03 -0700687 return;
Andi Kleendaeeafe2005-04-16 15:25:13 -0700688
Linus Torvalds1da177e2005-04-16 15:20:36 -0700689 conditional_sti(regs);
690
691 /* Mask out spurious debug traps due to lazy DR7 setting */
692 if (condition & (DR_TRAP0|DR_TRAP1|DR_TRAP2|DR_TRAP3)) {
693 if (!tsk->thread.debugreg7) {
694 goto clear_dr7;
695 }
696 }
697
698 tsk->thread.debugreg6 = condition;
699
700 /* Mask out spurious TF errors due to lazy TF clearing */
Andi Kleendaeeafe2005-04-16 15:25:13 -0700701 if (condition & DR_STEP) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700702 /*
703 * The TF error should be masked out only if the current
704 * process is not traced and if the TRAP flag has been set
705 * previously by a tracing process (condition detected by
706 * the PT_DTRACE flag); remember that the i386 TRAP flag
707 * can be modified by the process itself in user mode,
708 * allowing programs to debug themselves without the ptrace()
709 * interface.
710 */
Vincent Hanquez76381fe2005-06-23 00:08:46 -0700711 if (!user_mode(regs))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700712 goto clear_TF_reenable;
Andi Kleenbe61bff2005-04-16 15:24:57 -0700713 /*
714 * Was the TF flag set by a debugger? If so, clear it now,
715 * so that register information is correct.
716 */
717 if (tsk->ptrace & PT_DTRACE) {
718 regs->eflags &= ~TF_MASK;
719 tsk->ptrace &= ~PT_DTRACE;
720 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700721 }
722
723 /* Ok, finally something we can handle */
724 tsk->thread.trap_no = 1;
725 tsk->thread.error_code = error_code;
726 info.si_signo = SIGTRAP;
727 info.si_errno = 0;
728 info.si_code = TRAP_BRKPT;
Vincent Hanquez76381fe2005-06-23 00:08:46 -0700729 if (!user_mode(regs))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700730 goto clear_dr7;
731
732 info.si_addr = (void __user *)regs->rip;
733 force_sig_info(SIGTRAP, &info, tsk);
734clear_dr7:
Vincent Hanqueze9129e52005-06-23 00:08:46 -0700735 set_debugreg(0UL, 7);
Andi Kleen6fefb0d2005-04-16 15:25:03 -0700736 return;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700737
738clear_TF_reenable:
739 set_tsk_thread_flag(tsk, TIF_SINGLESTEP);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700740 regs->eflags &= ~TF_MASK;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700741}
742
Jan Beulich6e3f3612006-01-11 22:42:14 +0100743static int kernel_math_error(struct pt_regs *regs, const char *str, int trapnr)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700744{
745 const struct exception_table_entry *fixup;
746 fixup = search_exception_tables(regs->rip);
747 if (fixup) {
748 regs->rip = fixup->fixup;
749 return 1;
750 }
Jan Beulich6e3f3612006-01-11 22:42:14 +0100751 notify_die(DIE_GPF, str, regs, 0, trapnr, SIGFPE);
Andi Kleen3a848f62005-04-16 15:25:06 -0700752 /* Illegal floating point operation in the kernel */
Jan Beulich6e3f3612006-01-11 22:42:14 +0100753 current->thread.trap_no = trapnr;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700754 die(str, regs, 0);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700755 return 0;
756}
757
758/*
759 * Note that we play around with the 'TS' bit in an attempt to get
760 * the correct behaviour even in the presence of the asynchronous
761 * IRQ13 behaviour
762 */
763asmlinkage void do_coprocessor_error(struct pt_regs *regs)
764{
765 void __user *rip = (void __user *)(regs->rip);
766 struct task_struct * task;
767 siginfo_t info;
768 unsigned short cwd, swd;
769
770 conditional_sti(regs);
Vincent Hanquez76381fe2005-06-23 00:08:46 -0700771 if (!user_mode(regs) &&
Jan Beulich6e3f3612006-01-11 22:42:14 +0100772 kernel_math_error(regs, "kernel x87 math error", 16))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700773 return;
774
775 /*
776 * Save the info for the exception handler and clear the error.
777 */
778 task = current;
779 save_init_fpu(task);
780 task->thread.trap_no = 16;
781 task->thread.error_code = 0;
782 info.si_signo = SIGFPE;
783 info.si_errno = 0;
784 info.si_code = __SI_FAULT;
785 info.si_addr = rip;
786 /*
787 * (~cwd & swd) will mask out exceptions that are not set to unmasked
788 * status. 0x3f is the exception bits in these regs, 0x200 is the
789 * C1 reg you need in case of a stack fault, 0x040 is the stack
790 * fault bit. We should only be taking one exception at a time,
791 * so if this combination doesn't produce any single exception,
792 * then we have a bad program that isn't synchronizing its FPU usage
793 * and it will suffer the consequences since we won't be able to
794 * fully reproduce the context of the exception
795 */
796 cwd = get_fpu_cwd(task);
797 swd = get_fpu_swd(task);
Chuck Ebbertff347b22005-09-12 18:49:25 +0200798 switch (swd & ~cwd & 0x3f) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700799 case 0x000:
800 default:
801 break;
802 case 0x001: /* Invalid Op */
Chuck Ebbertff347b22005-09-12 18:49:25 +0200803 /*
804 * swd & 0x240 == 0x040: Stack Underflow
805 * swd & 0x240 == 0x240: Stack Overflow
806 * User must clear the SF bit (0x40) if set
807 */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700808 info.si_code = FPE_FLTINV;
809 break;
810 case 0x002: /* Denormalize */
811 case 0x010: /* Underflow */
812 info.si_code = FPE_FLTUND;
813 break;
814 case 0x004: /* Zero Divide */
815 info.si_code = FPE_FLTDIV;
816 break;
817 case 0x008: /* Overflow */
818 info.si_code = FPE_FLTOVF;
819 break;
820 case 0x020: /* Precision */
821 info.si_code = FPE_FLTRES;
822 break;
823 }
824 force_sig_info(SIGFPE, &info, task);
825}
826
827asmlinkage void bad_intr(void)
828{
829 printk("bad interrupt");
830}
831
832asmlinkage void do_simd_coprocessor_error(struct pt_regs *regs)
833{
834 void __user *rip = (void __user *)(regs->rip);
835 struct task_struct * task;
836 siginfo_t info;
837 unsigned short mxcsr;
838
839 conditional_sti(regs);
Vincent Hanquez76381fe2005-06-23 00:08:46 -0700840 if (!user_mode(regs) &&
Jan Beulich6e3f3612006-01-11 22:42:14 +0100841 kernel_math_error(regs, "kernel simd math error", 19))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700842 return;
843
844 /*
845 * Save the info for the exception handler and clear the error.
846 */
847 task = current;
848 save_init_fpu(task);
849 task->thread.trap_no = 19;
850 task->thread.error_code = 0;
851 info.si_signo = SIGFPE;
852 info.si_errno = 0;
853 info.si_code = __SI_FAULT;
854 info.si_addr = rip;
855 /*
856 * The SIMD FPU exceptions are handled a little differently, as there
857 * is only a single status/control register. Thus, to determine which
858 * unmasked exception was caught we must mask the exception mask bits
859 * at 0x1f80, and then use these to mask the exception bits at 0x3f.
860 */
861 mxcsr = get_fpu_mxcsr(task);
862 switch (~((mxcsr & 0x1f80) >> 7) & (mxcsr & 0x3f)) {
863 case 0x000:
864 default:
865 break;
866 case 0x001: /* Invalid Op */
867 info.si_code = FPE_FLTINV;
868 break;
869 case 0x002: /* Denormalize */
870 case 0x010: /* Underflow */
871 info.si_code = FPE_FLTUND;
872 break;
873 case 0x004: /* Zero Divide */
874 info.si_code = FPE_FLTDIV;
875 break;
876 case 0x008: /* Overflow */
877 info.si_code = FPE_FLTOVF;
878 break;
879 case 0x020: /* Precision */
880 info.si_code = FPE_FLTRES;
881 break;
882 }
883 force_sig_info(SIGFPE, &info, task);
884}
885
886asmlinkage void do_spurious_interrupt_bug(struct pt_regs * regs)
887{
888}
889
890asmlinkage void __attribute__((weak)) smp_thermal_interrupt(void)
891{
892}
893
Jacob Shin89b831e2005-11-05 17:25:53 +0100894asmlinkage void __attribute__((weak)) mce_threshold_interrupt(void)
895{
896}
897
Linus Torvalds1da177e2005-04-16 15:20:36 -0700898/*
899 * 'math_state_restore()' saves the current math information in the
900 * old math state array, and gets the new ones from the current task
901 *
902 * Careful.. There are problems with IBM-designed IRQ13 behaviour.
903 * Don't touch unless you *really* know how it works.
904 */
905asmlinkage void math_state_restore(void)
906{
907 struct task_struct *me = current;
908 clts(); /* Allow maths ops (or we recurse) */
909
910 if (!used_math())
911 init_fpu(me);
912 restore_fpu_checking(&me->thread.i387.fxsave);
913 me->thread_info->status |= TS_USEDFPU;
914}
915
916void do_call_debug(struct pt_regs *regs)
917{
918 notify_die(DIE_CALL, "debug call", regs, 0, 255, SIGINT);
919}
920
921void __init trap_init(void)
922{
923 set_intr_gate(0,&divide_error);
924 set_intr_gate_ist(1,&debug,DEBUG_STACK);
925 set_intr_gate_ist(2,&nmi,NMI_STACK);
Jan Beulichb556b352006-01-11 22:43:00 +0100926 set_system_gate_ist(3,&int3,DEBUG_STACK); /* int3 can be called from all */
Jan Beulich0a521582006-01-11 22:42:08 +0100927 set_system_gate(4,&overflow); /* int4 can be called from all */
928 set_intr_gate(5,&bounds);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700929 set_intr_gate(6,&invalid_op);
930 set_intr_gate(7,&device_not_available);
931 set_intr_gate_ist(8,&double_fault, DOUBLEFAULT_STACK);
932 set_intr_gate(9,&coprocessor_segment_overrun);
933 set_intr_gate(10,&invalid_TSS);
934 set_intr_gate(11,&segment_not_present);
935 set_intr_gate_ist(12,&stack_segment,STACKFAULT_STACK);
936 set_intr_gate(13,&general_protection);
937 set_intr_gate(14,&page_fault);
938 set_intr_gate(15,&spurious_interrupt_bug);
939 set_intr_gate(16,&coprocessor_error);
940 set_intr_gate(17,&alignment_check);
941#ifdef CONFIG_X86_MCE
942 set_intr_gate_ist(18,&machine_check, MCE_STACK);
943#endif
944 set_intr_gate(19,&simd_coprocessor_error);
945
946#ifdef CONFIG_IA32_EMULATION
947 set_system_gate(IA32_SYSCALL_VECTOR, ia32_syscall);
948#endif
949
950 set_intr_gate(KDB_VECTOR, call_debug);
951
952 /*
953 * Should be a barrier for any external CPU state.
954 */
955 cpu_init();
956}
957
958
959/* Actual parsing is done early in setup.c. */
960static int __init oops_dummy(char *s)
961{
962 panic_on_oops = 1;
963 return -1;
964}
965__setup("oops=", oops_dummy);
966
967static int __init kstack_setup(char *s)
968{
969 kstack_depth_to_print = simple_strtoul(s,NULL,0);
970 return 0;
971}
972__setup("kstack=", kstack_setup);
973