traps.c 21.5 KB
Newer Older
L
Linus Torvalds 已提交
1 2
/*
 *  Copyright (C) 1991, 1992  Linus Torvalds
3
 *  Copyright (C) 2000, 2001, 2002 Andi Kleen, SuSE Labs
L
Linus Torvalds 已提交
4 5 6 7 8 9
 *
 *  Pentium III FXSR, SSE support
 *	Gareth Hughes <gareth@valinux.com>, May 2000
 */

/*
10
 * Handle hardware traps and faults.
L
Linus Torvalds 已提交
11
 */
12 13 14

#define pr_fmt(fmt) KBUILD_MODNAME ": " fmt

15
#include <linux/context_tracking.h>
I
Ingo Molnar 已提交
16 17 18 19 20 21
#include <linux/interrupt.h>
#include <linux/kallsyms.h>
#include <linux/spinlock.h>
#include <linux/kprobes.h>
#include <linux/uaccess.h>
#include <linux/kdebug.h>
22
#include <linux/kgdb.h>
L
Linus Torvalds 已提交
23
#include <linux/kernel.h>
I
Ingo Molnar 已提交
24 25
#include <linux/module.h>
#include <linux/ptrace.h>
26
#include <linux/uprobes.h>
L
Linus Torvalds 已提交
27
#include <linux/string.h>
I
Ingo Molnar 已提交
28
#include <linux/delay.h>
L
Linus Torvalds 已提交
29
#include <linux/errno.h>
I
Ingo Molnar 已提交
30 31
#include <linux/kexec.h>
#include <linux/sched.h>
L
Linus Torvalds 已提交
32 33
#include <linux/timer.h>
#include <linux/init.h>
J
Jeremy Fitzhardinge 已提交
34
#include <linux/bug.h>
I
Ingo Molnar 已提交
35 36
#include <linux/nmi.h>
#include <linux/mm.h>
37 38
#include <linux/smp.h>
#include <linux/io.h>
L
Linus Torvalds 已提交
39 40 41 42 43 44

#ifdef CONFIG_EISA
#include <linux/ioport.h>
#include <linux/eisa.h>
#endif

D
Dave Jiang 已提交
45 46 47 48
#if defined(CONFIG_EDAC)
#include <linux/edac.h>
#endif

V
Vegard Nossum 已提交
49
#include <asm/kmemcheck.h>
I
Ingo Molnar 已提交
50
#include <asm/stacktrace.h>
L
Linus Torvalds 已提交
51 52
#include <asm/processor.h>
#include <asm/debugreg.h>
A
Arun Sharma 已提交
53
#include <linux/atomic.h>
54
#include <asm/ftrace.h>
55
#include <asm/traps.h>
L
Linus Torvalds 已提交
56 57
#include <asm/desc.h>
#include <asm/i387.h>
58
#include <asm/fpu-internal.h>
H
Hidetoshi Seto 已提交
59
#include <asm/mce.h>
60
#include <asm/fixmap.h>
61
#include <asm/mach_traps.h>
62
#include <asm/alternative.h>
63

64
#ifdef CONFIG_X86_64
65
#include <asm/x86_init.h>
66 67
#include <asm/pgalloc.h>
#include <asm/proto.h>
K
Kees Cook 已提交
68 69 70

/* No need to be aligned, but done to keep all IDTs defined the same way. */
gate_desc debug_idt_table[NR_VECTORS] __page_aligned_bss;
71
#else
72
#include <asm/processor-flags.h>
I
Ingo Molnar 已提交
73
#include <asm/setup.h>
L
Linus Torvalds 已提交
74 75

asmlinkage int system_call(void);
76
#endif
L
Linus Torvalds 已提交
77

K
Kees Cook 已提交
78 79 80
/* Must be page-aligned because the real IDT is used in a fixmap. */
gate_desc idt_table[NR_VECTORS] __page_aligned_bss;

81 82 83
DECLARE_BITMAP(used_vectors, NR_VECTORS);
EXPORT_SYMBOL_GPL(used_vectors);

84 85 86 87 88 89
static inline void conditional_sti(struct pt_regs *regs)
{
	if (regs->flags & X86_EFLAGS_IF)
		local_irq_enable();
}

90 91
static inline void preempt_conditional_sti(struct pt_regs *regs)
{
92
	preempt_count_inc();
93 94 95 96
	if (regs->flags & X86_EFLAGS_IF)
		local_irq_enable();
}

T
Thomas Gleixner 已提交
97 98 99 100 101 102
static inline void conditional_cli(struct pt_regs *regs)
{
	if (regs->flags & X86_EFLAGS_IF)
		local_irq_disable();
}

103 104 105 106
static inline void preempt_conditional_cli(struct pt_regs *regs)
{
	if (regs->flags & X86_EFLAGS_IF)
		local_irq_disable();
107
	preempt_count_dec();
108 109
}

110
static nokprobe_inline int
111 112
do_trap_no_signal(struct task_struct *tsk, int trapnr, char *str,
		  struct pt_regs *regs,	long error_code)
L
Linus Torvalds 已提交
113
{
114
#ifdef CONFIG_X86_32
115
	if (regs->flags & X86_VM_MASK) {
116
		/*
117
		 * Traps 0, 1, 3, 4, and 5 should be forwarded to vm86.
118 119
		 * On nmi (interrupt 2), do_trap should not be called.
		 */
120 121 122 123 124 125
		if (trapnr < X86_TRAP_UD) {
			if (!handle_vm86_trap((struct kernel_vm86_regs *) regs,
						error_code, trapnr))
				return 0;
		}
		return -1;
L
Linus Torvalds 已提交
126
	}
127
#endif
128 129 130 131 132 133 134 135
	if (!user_mode(regs)) {
		if (!fixup_exception(regs)) {
			tsk->thread.error_code = error_code;
			tsk->thread.trap_nr = trapnr;
			die(str, regs, error_code);
		}
		return 0;
	}
L
Linus Torvalds 已提交
136

137 138
	return -1;
}
L
Linus Torvalds 已提交
139

140 141
static siginfo_t *fill_trap_info(struct pt_regs *regs, int signr, int trapnr,
				siginfo_t *info)
142 143 144 145 146
{
	unsigned long siaddr;
	int sicode;

	switch (trapnr) {
147 148 149
	default:
		return SEND_SIG_PRIV;

150 151
	case X86_TRAP_DE:
		sicode = FPE_INTDIV;
152
		siaddr = uprobe_get_trap_addr(regs);
153 154 155
		break;
	case X86_TRAP_UD:
		sicode = ILL_ILLOPN;
156
		siaddr = uprobe_get_trap_addr(regs);
157 158 159 160 161 162 163 164 165 166 167
		break;
	case X86_TRAP_AC:
		sicode = BUS_ADRALN;
		siaddr = 0;
		break;
	}

	info->si_signo = signr;
	info->si_errno = 0;
	info->si_code = sicode;
	info->si_addr = (void __user *)siaddr;
168
	return info;
169 170
}

171
static void
172 173 174 175 176 177 178 179
do_trap(int trapnr, int signr, char *str, struct pt_regs *regs,
	long error_code, siginfo_t *info)
{
	struct task_struct *tsk = current;


	if (!do_trap_no_signal(tsk, trapnr, str, regs, error_code))
		return;
I
Ingo Molnar 已提交
180
	/*
181
	 * We want error_code and trap_nr set for userspace faults and
I
Ingo Molnar 已提交
182 183 184 185 186 187 188 189
	 * kernelspace faults which result in die(), but not
	 * kernelspace faults which are fixed up.  die() gives the
	 * process no chance to handle the signal and notice the
	 * kernel fault information, so that won't result in polluting
	 * the information about previously queued, but not yet
	 * delivered, faults.  See also do_general_protection below.
	 */
	tsk->thread.error_code = error_code;
190
	tsk->thread.trap_nr = trapnr;
191

192 193 194
#ifdef CONFIG_X86_64
	if (show_unhandled_signals && unhandled_signal(tsk, signr) &&
	    printk_ratelimit()) {
195 196 197
		pr_info("%s[%d] trap %s ip:%lx sp:%lx error:%lx",
			tsk->comm, tsk->pid, str,
			regs->ip, regs->sp, error_code);
198
		print_vma_addr(" in ", regs->ip);
199
		pr_cont("\n");
200 201 202
	}
#endif

203
	force_sig_info(signr, info ?: SEND_SIG_PRIV, tsk);
L
Linus Torvalds 已提交
204
}
205
NOKPROBE_SYMBOL(do_trap);
L
Linus Torvalds 已提交
206

207
static void do_error_trap(struct pt_regs *regs, long error_code, char *str,
208
			  unsigned long trapnr, int signr)
209 210
{
	enum ctx_state prev_state = exception_enter();
211
	siginfo_t info;
212 213 214 215

	if (notify_die(DIE_TRAP, str, regs, error_code, trapnr, signr) !=
			NOTIFY_STOP) {
		conditional_sti(regs);
216 217
		do_trap(trapnr, signr, str, regs, error_code,
			fill_trap_info(regs, signr, trapnr, &info));
218 219 220 221 222
	}

	exception_exit(prev_state);
}

I
Ingo Molnar 已提交
223
#define DO_ERROR(trapnr, signr, str, name)				\
224
dotraplinkage void do_##name(struct pt_regs *regs, long error_code)	\
I
Ingo Molnar 已提交
225
{									\
226
	do_error_trap(regs, error_code, str, trapnr, signr);		\
L
Linus Torvalds 已提交
227 228
}

O
Oleg Nesterov 已提交
229 230 231 232 233 234 235
DO_ERROR(X86_TRAP_DE,     SIGFPE,  "divide error",		divide_error)
DO_ERROR(X86_TRAP_OF,     SIGSEGV, "overflow",			overflow)
DO_ERROR(X86_TRAP_BR,     SIGSEGV, "bounds",			bounds)
DO_ERROR(X86_TRAP_UD,     SIGILL,  "invalid opcode",		invalid_op)
DO_ERROR(X86_TRAP_OLD_MF, SIGFPE,  "coprocessor segment overrun",coprocessor_segment_overrun)
DO_ERROR(X86_TRAP_TS,     SIGSEGV, "invalid TSS",		invalid_TSS)
DO_ERROR(X86_TRAP_NP,     SIGBUS,  "segment not present",	segment_not_present)
236
#ifdef CONFIG_X86_32
O
Oleg Nesterov 已提交
237
DO_ERROR(X86_TRAP_SS,     SIGBUS,  "stack segment",		stack_segment)
238
#endif
O
Oleg Nesterov 已提交
239
DO_ERROR(X86_TRAP_AC,     SIGBUS,  "alignment check",		alignment_check)
L
Linus Torvalds 已提交
240

241 242 243 244
#ifdef CONFIG_X86_64
/* Runs on IST stack */
dotraplinkage void do_stack_segment(struct pt_regs *regs, long error_code)
{
245 246 247
	enum ctx_state prev_state;

	prev_state = exception_enter();
248
	if (notify_die(DIE_TRAP, "stack segment", regs, error_code,
249 250 251 252 253
		       X86_TRAP_SS, SIGBUS) != NOTIFY_STOP) {
		preempt_conditional_sti(regs);
		do_trap(X86_TRAP_SS, SIGBUS, "stack segment", regs, error_code, NULL);
		preempt_conditional_cli(regs);
	}
254
	exception_exit(prev_state);
255 256 257 258 259 260 261
}

dotraplinkage void do_double_fault(struct pt_regs *regs, long error_code)
{
	static const char str[] = "double fault";
	struct task_struct *tsk = current;

262
	exception_enter();
263
	/* Return not checked because double check cannot be ignored */
264
	notify_die(DIE_TRAP, str, regs, error_code, X86_TRAP_DF, SIGSEGV);
265 266

	tsk->thread.error_code = error_code;
267
	tsk->thread.trap_nr = X86_TRAP_DF;
268

269 270 271
#ifdef CONFIG_DOUBLEFAULT
	df_debug(regs, error_code);
#endif
272 273 274 275
	/*
	 * This is always a kernel trap and never fixable (and thus must
	 * never return).
	 */
276 277 278 279 280
	for (;;)
		die(str, regs, error_code);
}
#endif

281
dotraplinkage void
282
do_general_protection(struct pt_regs *regs, long error_code)
L
Linus Torvalds 已提交
283
{
284
	struct task_struct *tsk;
285
	enum ctx_state prev_state;
I
Ingo Molnar 已提交
286

287
	prev_state = exception_enter();
288 289
	conditional_sti(regs);

290
#ifdef CONFIG_X86_32
291 292 293
	if (regs->flags & X86_VM_MASK) {
		local_irq_enable();
		handle_vm86_fault((struct kernel_vm86_regs *) regs, error_code);
294
		goto exit;
295
	}
296
#endif
L
Linus Torvalds 已提交
297

298
	tsk = current;
299 300
	if (!user_mode(regs)) {
		if (fixup_exception(regs))
301
			goto exit;
302 303 304

		tsk->thread.error_code = error_code;
		tsk->thread.trap_nr = X86_TRAP_GP;
305 306
		if (notify_die(DIE_GPF, "general protection fault", regs, error_code,
			       X86_TRAP_GP, SIGSEGV) != NOTIFY_STOP)
307
			die("general protection fault", regs, error_code);
308
		goto exit;
309
	}
L
Linus Torvalds 已提交
310

311
	tsk->thread.error_code = error_code;
312
	tsk->thread.trap_nr = X86_TRAP_GP;
I
Ingo Molnar 已提交
313

314 315
	if (show_unhandled_signals && unhandled_signal(tsk, SIGSEGV) &&
			printk_ratelimit()) {
316
		pr_info("%s[%d] general protection ip:%lx sp:%lx error:%lx",
317 318
			tsk->comm, task_pid_nr(tsk),
			regs->ip, regs->sp, error_code);
319
		print_vma_addr(" in ", regs->ip);
320
		pr_cont("\n");
321
	}
322

323
	force_sig_info(SIGSEGV, SEND_SIG_PRIV, tsk);
324
exit:
325
	exception_exit(prev_state);
L
Linus Torvalds 已提交
326
}
327
NOKPROBE_SYMBOL(do_general_protection);
L
Linus Torvalds 已提交
328

329
/* May run on IST stack. */
330
dotraplinkage void notrace do_int3(struct pt_regs *regs, long error_code)
L
Linus Torvalds 已提交
331
{
332 333
	enum ctx_state prev_state;

334
#ifdef CONFIG_DYNAMIC_FTRACE
335 336 337 338 339 340
	/*
	 * ftrace must be first, everything else may cause a recursive crash.
	 * See note by declaration of modifying_ftrace_code in ftrace.c
	 */
	if (unlikely(atomic_read(&modifying_ftrace_code)) &&
	    ftrace_int3_handler(regs))
341 342
		return;
#endif
343 344 345
	if (poke_int3_handler(regs))
		return;

346
	prev_state = exception_enter();
347
#ifdef CONFIG_KGDB_LOW_LEVEL_TRAP
348 349
	if (kgdb_ll_trap(DIE_INT3, "int3", regs, error_code, X86_TRAP_BP,
				SIGTRAP) == NOTIFY_STOP)
350
		goto exit;
351
#endif /* CONFIG_KGDB_LOW_LEVEL_TRAP */
352

353 354
#ifdef CONFIG_KPROBES
	if (kprobe_int3_handler(regs))
355
		goto exit;
356 357
#endif

358 359
	if (notify_die(DIE_INT3, "int3", regs, error_code, X86_TRAP_BP,
			SIGTRAP) == NOTIFY_STOP)
360
		goto exit;
I
Ingo Molnar 已提交
361

362 363 364 365 366
	/*
	 * Let others (NMI) know that the debug stack is in use
	 * as we may switch to the interrupt stack.
	 */
	debug_stack_usage_inc();
367
	preempt_conditional_sti(regs);
368
	do_trap(X86_TRAP_BP, SIGTRAP, "int3", regs, error_code, NULL);
369
	preempt_conditional_cli(regs);
370
	debug_stack_usage_dec();
371
exit:
372
	exception_exit(prev_state);
L
Linus Torvalds 已提交
373
}
374
NOKPROBE_SYMBOL(do_int3);
L
Linus Torvalds 已提交
375

376
#ifdef CONFIG_X86_64
377 378 379 380 381
/*
 * Help handler running on IST stack to switch back to user stack
 * for scheduling or signal handling. The actual stack switch is done in
 * entry.S
 */
382
asmlinkage __visible struct pt_regs *sync_regs(struct pt_regs *eregs)
383 384 385 386 387 388 389 390
{
	struct pt_regs *regs = eregs;
	/* Did already sync */
	if (eregs == (struct pt_regs *)eregs->sp)
		;
	/* Exception from user space */
	else if (user_mode(eregs))
		regs = task_pt_regs(current);
391 392 393 394
	/*
	 * Exception from kernel and interrupts are enabled. Move to
	 * kernel process stack.
	 */
395 396 397 398 399 400
	else if (eregs->flags & X86_EFLAGS_IF)
		regs = (struct pt_regs *)(eregs->sp -= sizeof(struct pt_regs));
	if (eregs != regs)
		*regs = *eregs;
	return regs;
}
401
NOKPROBE_SYMBOL(sync_regs);
402 403
#endif

L
Linus Torvalds 已提交
404 405 406 407 408 409 410 411 412 413
/*
 * Our handling of the processor debug registers is non-trivial.
 * We do not clear them on entry and exit from the kernel. Therefore
 * it is possible to get a watchpoint trap here from inside the kernel.
 * However, the code in ./ptrace.c has ensured that the user can
 * only set watchpoints on userspace addresses. Therefore the in-kernel
 * watchpoint trap can only occur in code which is reading/writing
 * from user space. Such code must not hold kernel locks (since it
 * can equally take a page fault), therefore it is safe to call
 * force_sig_info even though that claims and releases locks.
I
Ingo Molnar 已提交
414
 *
L
Linus Torvalds 已提交
415 416 417 418 419 420 421 422 423 424
 * Code in ./signal.c ensures that the debug control register
 * is restored before we deliver any signal, and therefore that
 * user code runs with the correct debug control register even though
 * we clear it here.
 *
 * Being careful here means that we don't have to be as careful in a
 * lot of more complicated places (task switching can be a bit lazy
 * about restoring all the debug state, and ptrace doesn't have to
 * find every occurrence of the TF bit that could be saved away even
 * by user code)
425 426
 *
 * May run on IST stack.
L
Linus Torvalds 已提交
427
 */
428
dotraplinkage void do_debug(struct pt_regs *regs, long error_code)
L
Linus Torvalds 已提交
429 430
{
	struct task_struct *tsk = current;
431
	enum ctx_state prev_state;
432
	int user_icebp = 0;
433
	unsigned long dr6;
434
	int si_code;
L
Linus Torvalds 已提交
435

436 437
	prev_state = exception_enter();

438
	get_debugreg(dr6, 6);
L
Linus Torvalds 已提交
439

440 441 442
	/* Filter out all the reserved bits which are preset to 1 */
	dr6 &= ~DR6_RESERVED;

443 444 445 446 447 448 449 450
	/*
	 * If dr6 has no reason to give us about the origin of this trap,
	 * then it's very likely the result of an icebp/int01 trap.
	 * User wants a sigtrap for that.
	 */
	if (!dr6 && user_mode(regs))
		user_icebp = 1;

V
Vegard Nossum 已提交
451
	/* Catch kmemcheck conditions first of all! */
452
	if ((dr6 & DR_STEP) && kmemcheck_trap(regs))
453
		goto exit;
V
Vegard Nossum 已提交
454

455 456
	/* DR6 may or may not be cleared by the CPU */
	set_debugreg(0, 6);
457

P
Peter Zijlstra 已提交
458 459 460 461 462
	/*
	 * The processor cleared BTF, so don't mark that we need it set.
	 */
	clear_tsk_thread_flag(tsk, TIF_BLOCKSTEP);

463 464 465
	/* Store the virtualized DR6 value */
	tsk->thread.debugreg6 = dr6;

466 467 468 469 470
#ifdef CONFIG_KPROBES
	if (kprobe_debug_handler(regs))
		goto exit;
#endif

471
	if (notify_die(DIE_DEBUG, "debug", regs, (long)&dr6, error_code,
472
							SIGTRAP) == NOTIFY_STOP)
473
		goto exit;
474

475 476 477 478 479 480
	/*
	 * Let others (NMI) know that the debug stack is in use
	 * as we may switch to the interrupt stack.
	 */
	debug_stack_usage_inc();

L
Linus Torvalds 已提交
481
	/* It's safe to allow irq's after DR6 has been saved */
482
	preempt_conditional_sti(regs);
L
Linus Torvalds 已提交
483

484
	if (regs->flags & X86_VM_MASK) {
485 486
		handle_vm86_trap((struct kernel_vm86_regs *) regs, error_code,
					X86_TRAP_DB);
487
		preempt_conditional_cli(regs);
488
		debug_stack_usage_dec();
489
		goto exit;
L
Linus Torvalds 已提交
490 491 492
	}

	/*
493 494 495 496 497
	 * Single-stepping through system calls: ignore any exceptions in
	 * kernel space, but re-enable TF when returning to user mode.
	 *
	 * We already checked v86 mode above, so we can check for kernel mode
	 * by just checking the CPL of CS.
L
Linus Torvalds 已提交
498
	 */
499 500 501 502
	if ((dr6 & DR_STEP) && !user_mode(regs)) {
		tsk->thread.debugreg6 &= ~DR_STEP;
		set_tsk_thread_flag(tsk, TIF_SINGLESTEP);
		regs->flags &= ~X86_EFLAGS_TF;
L
Linus Torvalds 已提交
503
	}
504
	si_code = get_si_code(tsk->thread.debugreg6);
505
	if (tsk->thread.debugreg6 & (DR_STEP | DR_TRAP_BITS) || user_icebp)
506
		send_sigtrap(tsk, regs, error_code, si_code);
507
	preempt_conditional_cli(regs);
508
	debug_stack_usage_dec();
L
Linus Torvalds 已提交
509

510
exit:
511
	exception_exit(prev_state);
L
Linus Torvalds 已提交
512
}
513
NOKPROBE_SYMBOL(do_debug);
L
Linus Torvalds 已提交
514 515 516 517 518 519

/*
 * Note that we play around with the 'TS' bit in an attempt to get
 * the correct behaviour even in the presence of the asynchronous
 * IRQ13 behaviour
 */
520
static void math_error(struct pt_regs *regs, int error_code, int trapnr)
L
Linus Torvalds 已提交
521
{
522
	struct task_struct *task = current;
L
Linus Torvalds 已提交
523
	siginfo_t info;
524
	unsigned short err;
525 526
	char *str = (trapnr == X86_TRAP_MF) ? "fpu exception" :
						"simd exception";
527 528 529 530 531 532 533 534 535

	if (notify_die(DIE_TRAP, str, regs, error_code, trapnr, SIGFPE) == NOTIFY_STOP)
		return;
	conditional_sti(regs);

	if (!user_mode_vm(regs))
	{
		if (!fixup_exception(regs)) {
			task->thread.error_code = error_code;
536
			task->thread.trap_nr = trapnr;
537 538 539 540
			die(str, regs, error_code);
		}
		return;
	}
L
Linus Torvalds 已提交
541 542 543 544 545

	/*
	 * Save the info for the exception handler and clear the error.
	 */
	save_init_fpu(task);
546
	task->thread.trap_nr = trapnr;
547
	task->thread.error_code = error_code;
L
Linus Torvalds 已提交
548 549
	info.si_signo = SIGFPE;
	info.si_errno = 0;
550
	info.si_addr = (void __user *)uprobe_get_trap_addr(regs);
551
	if (trapnr == X86_TRAP_MF) {
552 553 554 555 556 557 558 559 560 561 562 563 564
		unsigned short cwd, swd;
		/*
		 * (~cwd & swd) will mask out exceptions that are not set to unmasked
		 * status.  0x3f is the exception bits in these regs, 0x200 is the
		 * C1 reg you need in case of a stack fault, 0x040 is the stack
		 * fault bit.  We should only be taking one exception at a time,
		 * so if this combination doesn't produce any single exception,
		 * then we have a bad program that isn't synchronizing its FPU usage
		 * and it will suffer the consequences since we won't be able to
		 * fully reproduce the context of the exception
		 */
		cwd = get_fpu_cwd(task);
		swd = get_fpu_swd(task);
565

566 567 568 569 570 571 572 573 574 575 576
		err = swd & ~cwd;
	} else {
		/*
		 * The SIMD FPU exceptions are handled a little differently, as there
		 * is only a single status/control register.  Thus, to determine which
		 * unmasked exception was caught we must mask the exception mask bits
		 * at 0x1f80, and then use these to mask the exception bits at 0x3f.
		 */
		unsigned short mxcsr = get_fpu_mxcsr(task);
		err = ~(mxcsr >> 7) & mxcsr;
	}
577 578

	if (err & 0x001) {	/* Invalid op */
I
Ingo Molnar 已提交
579 580 581 582 583 584
		/*
		 * swd & 0x240 == 0x040: Stack Underflow
		 * swd & 0x240 == 0x240: Stack Overflow
		 * User must clear the SF bit (0x40) if set
		 */
		info.si_code = FPE_FLTINV;
585
	} else if (err & 0x004) { /* Divide by Zero */
I
Ingo Molnar 已提交
586
		info.si_code = FPE_FLTDIV;
587
	} else if (err & 0x008) { /* Overflow */
I
Ingo Molnar 已提交
588
		info.si_code = FPE_FLTOVF;
589 590 591
	} else if (err & 0x012) { /* Denormal, Underflow */
		info.si_code = FPE_FLTUND;
	} else if (err & 0x020) { /* Precision */
I
Ingo Molnar 已提交
592
		info.si_code = FPE_FLTRES;
593
	} else {
594
		/*
595 596 597
		 * If we're using IRQ 13, or supposedly even some trap
		 * X86_TRAP_MF implementations, it's possible
		 * we get a spurious trap, which is not an error.
598
		 */
599
		return;
L
Linus Torvalds 已提交
600 601 602 603
	}
	force_sig_info(SIGFPE, &info, task);
}

604
dotraplinkage void do_coprocessor_error(struct pt_regs *regs, long error_code)
L
Linus Torvalds 已提交
605
{
606 607 608
	enum ctx_state prev_state;

	prev_state = exception_enter();
609
	math_error(regs, error_code, X86_TRAP_MF);
610
	exception_exit(prev_state);
L
Linus Torvalds 已提交
611 612
}

613 614
dotraplinkage void
do_simd_coprocessor_error(struct pt_regs *regs, long error_code)
L
Linus Torvalds 已提交
615
{
616 617 618
	enum ctx_state prev_state;

	prev_state = exception_enter();
619
	math_error(regs, error_code, X86_TRAP_XF);
620
	exception_exit(prev_state);
L
Linus Torvalds 已提交
621 622
}

623 624
dotraplinkage void
do_spurious_interrupt_bug(struct pt_regs *regs, long error_code)
L
Linus Torvalds 已提交
625
{
626
	conditional_sti(regs);
L
Linus Torvalds 已提交
627 628
#if 0
	/* No need to warn about this any longer. */
629
	pr_info("Ignoring P6 Local APIC Spurious Interrupt Bug...\n");
L
Linus Torvalds 已提交
630 631 632
#endif
}

633
asmlinkage __visible void __attribute__((weak)) smp_thermal_interrupt(void)
L
Linus Torvalds 已提交
634 635
{
}
636

637
asmlinkage __visible void __attribute__((weak)) smp_threshold_interrupt(void)
638 639 640
{
}

L
Linus Torvalds 已提交
641
/*
I
Ingo Molnar 已提交
642
 * 'math_state_restore()' saves the current math information in the
L
Linus Torvalds 已提交
643 644 645 646 647
 * old math state array, and gets the new ones from the current task
 *
 * Careful.. There are problems with IBM-designed IRQ13 behaviour.
 * Don't touch unless you *really* know how it works.
 *
648 649
 * Must be called with kernel preemption disabled (eg with local
 * local interrupts as in the case of do_device_not_available).
L
Linus Torvalds 已提交
650
 */
651
void math_state_restore(void)
L
Linus Torvalds 已提交
652
{
653
	struct task_struct *tsk = current;
L
Linus Torvalds 已提交
654

655 656 657 658 659 660 661 662 663 664 665 666 667 668 669
	if (!tsk_used_math(tsk)) {
		local_irq_enable();
		/*
		 * does a slab alloc which can sleep
		 */
		if (init_fpu(tsk)) {
			/*
			 * ran out of memory!
			 */
			do_group_exit(SIGKILL);
			return;
		}
		local_irq_disable();
	}

670
	__thread_fpu_begin(tsk);
671

672 673 674 675
	/*
	 * Paranoid restore. send a SIGSEGV if we fail to restore the state.
	 */
	if (unlikely(restore_fpu_checking(tsk))) {
676
		drop_init_fpu(tsk);
677
		force_sig_info(SIGSEGV, SEND_SIG_PRIV, tsk);
678 679
		return;
	}
680

681
	tsk->thread.fpu_counter++;
L
Linus Torvalds 已提交
682
}
683
EXPORT_SYMBOL_GPL(math_state_restore);
L
Linus Torvalds 已提交
684

685
dotraplinkage void
686
do_device_not_available(struct pt_regs *regs, long error_code)
687
{
688 689 690
	enum ctx_state prev_state;

	prev_state = exception_enter();
691
	BUG_ON(use_eager_fpu());
692

693
#ifdef CONFIG_MATH_EMULATION
694
	if (read_cr0() & X86_CR0_EM) {
695 696
		struct math_emu_info info = { };

697
		conditional_sti(regs);
698

699
		info.regs = regs;
700
		math_emulate(&info);
701
		exception_exit(prev_state);
702
		return;
703
	}
704 705 706 707
#endif
	math_state_restore(); /* interrupts still off */
#ifdef CONFIG_X86_32
	conditional_sti(regs);
708
#endif
709
	exception_exit(prev_state);
710
}
711
NOKPROBE_SYMBOL(do_device_not_available);
712

713
#ifdef CONFIG_X86_32
714
dotraplinkage void do_iret_error(struct pt_regs *regs, long error_code)
715 716
{
	siginfo_t info;
717
	enum ctx_state prev_state;
718

719
	prev_state = exception_enter();
720 721 722 723 724
	local_irq_enable();

	info.si_signo = SIGILL;
	info.si_errno = 0;
	info.si_code = ILL_BADSTK;
725
	info.si_addr = NULL;
726
	if (notify_die(DIE_TRAP, "iret exception", regs, error_code,
727 728 729 730
			X86_TRAP_IRET, SIGILL) != NOTIFY_STOP) {
		do_trap(X86_TRAP_IRET, SIGILL, "iret exception", regs, error_code,
			&info);
	}
731
	exception_exit(prev_state);
732
}
733
#endif
734

735 736 737
/* Set of traps needed for early debugging. */
void __init early_trap_init(void)
{
738
	set_intr_gate_ist(X86_TRAP_DB, &debug, DEBUG_STACK);
739
	/* int3 can be called from all */
740
	set_system_intr_gate_ist(X86_TRAP_BP, &int3, DEBUG_STACK);
741
#ifdef CONFIG_X86_32
742
	set_intr_gate(X86_TRAP_PF, page_fault);
743
#endif
744 745 746
	load_idt(&idt_descr);
}

747 748 749
void __init early_trap_pf_init(void)
{
#ifdef CONFIG_X86_64
750
	set_intr_gate(X86_TRAP_PF, page_fault);
751 752 753
#endif
}

L
Linus Torvalds 已提交
754 755
void __init trap_init(void)
{
756 757
	int i;

L
Linus Torvalds 已提交
758
#ifdef CONFIG_EISA
I
Ingo Molnar 已提交
759
	void __iomem *p = early_ioremap(0x0FFFD9, 4);
I
Ingo Molnar 已提交
760 761

	if (readl(p) == 'E' + ('I'<<8) + ('S'<<16) + ('A'<<24))
L
Linus Torvalds 已提交
762
		EISA_bus = 1;
I
Ingo Molnar 已提交
763
	early_iounmap(p, 4);
L
Linus Torvalds 已提交
764 765
#endif

766
	set_intr_gate(X86_TRAP_DE, divide_error);
767
	set_intr_gate_ist(X86_TRAP_NMI, &nmi, NMI_STACK);
768
	/* int4 can be called from all */
769
	set_system_intr_gate(X86_TRAP_OF, &overflow);
770 771 772
	set_intr_gate(X86_TRAP_BR, bounds);
	set_intr_gate(X86_TRAP_UD, invalid_op);
	set_intr_gate(X86_TRAP_NM, device_not_available);
773
#ifdef CONFIG_X86_32
774
	set_task_gate(X86_TRAP_DF, GDT_ENTRY_DOUBLEFAULT_TSS);
775
#else
776
	set_intr_gate_ist(X86_TRAP_DF, &double_fault, DOUBLEFAULT_STACK);
777
#endif
778 779 780
	set_intr_gate(X86_TRAP_OLD_MF, coprocessor_segment_overrun);
	set_intr_gate(X86_TRAP_TS, invalid_TSS);
	set_intr_gate(X86_TRAP_NP, segment_not_present);
781
	set_intr_gate_ist(X86_TRAP_SS, &stack_segment, STACKFAULT_STACK);
782 783 784 785
	set_intr_gate(X86_TRAP_GP, general_protection);
	set_intr_gate(X86_TRAP_SPURIOUS, spurious_interrupt_bug);
	set_intr_gate(X86_TRAP_MF, coprocessor_error);
	set_intr_gate(X86_TRAP_AC, alignment_check);
L
Linus Torvalds 已提交
786
#ifdef CONFIG_X86_MCE
787
	set_intr_gate_ist(X86_TRAP_MC, &machine_check, MCE_STACK);
L
Linus Torvalds 已提交
788
#endif
789
	set_intr_gate(X86_TRAP_XF, simd_coprocessor_error);
L
Linus Torvalds 已提交
790

791 792 793 794
	/* Reserve all the builtin and the syscall vector: */
	for (i = 0; i < FIRST_EXTERNAL_VECTOR; i++)
		set_bit(i, used_vectors);

795 796
#ifdef CONFIG_IA32_EMULATION
	set_system_intr_gate(IA32_SYSCALL_VECTOR, ia32_syscall);
797
	set_bit(IA32_SYSCALL_VECTOR, used_vectors);
798 799 800
#endif

#ifdef CONFIG_X86_32
801
	set_system_trap_gate(SYSCALL_VECTOR, &system_call);
802
	set_bit(SYSCALL_VECTOR, used_vectors);
803
#endif
804

805 806 807 808 809 810 811 812
	/*
	 * Set the IDT descriptor to a fixed read-only location, so that the
	 * "sidt" instruction will not leak the location of the kernel, and
	 * to defend the IDT against arbitrary memory write vulnerabilities.
	 * It will be reloaded in cpu_init() */
	__set_fixmap(FIX_RO_IDT, __pa_symbol(idt_table), PAGE_KERNEL_RO);
	idt_descr.address = fix_to_virt(FIX_RO_IDT);

L
Linus Torvalds 已提交
813
	/*
I
Ingo Molnar 已提交
814
	 * Should be a barrier for any external CPU state:
L
Linus Torvalds 已提交
815 816 817
	 */
	cpu_init();

818
	x86_init.irqs.trap_init();
819 820

#ifdef CONFIG_X86_64
821
	memcpy(&debug_idt_table, &idt_table, IDT_ENTRIES * 16);
822 823
	set_nmi_gate(X86_TRAP_DB, &debug);
	set_nmi_gate(X86_TRAP_BP, &int3);
824
#endif
L
Linus Torvalds 已提交
825
}