traps.c 22.0 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
 */
I
Ingo Molnar 已提交
12 13 14 15 16 17
#include <linux/interrupt.h>
#include <linux/kallsyms.h>
#include <linux/spinlock.h>
#include <linux/kprobes.h>
#include <linux/uaccess.h>
#include <linux/kdebug.h>
18
#include <linux/kgdb.h>
L
Linus Torvalds 已提交
19
#include <linux/kernel.h>
I
Ingo Molnar 已提交
20 21
#include <linux/module.h>
#include <linux/ptrace.h>
L
Linus Torvalds 已提交
22
#include <linux/string.h>
I
Ingo Molnar 已提交
23
#include <linux/delay.h>
L
Linus Torvalds 已提交
24
#include <linux/errno.h>
I
Ingo Molnar 已提交
25 26
#include <linux/kexec.h>
#include <linux/sched.h>
L
Linus Torvalds 已提交
27 28
#include <linux/timer.h>
#include <linux/init.h>
J
Jeremy Fitzhardinge 已提交
29
#include <linux/bug.h>
I
Ingo Molnar 已提交
30 31
#include <linux/nmi.h>
#include <linux/mm.h>
32 33
#include <linux/smp.h>
#include <linux/io.h>
L
Linus Torvalds 已提交
34 35 36 37 38 39 40 41 42 43

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

#ifdef CONFIG_MCA
#include <linux/mca.h>
#endif

D
Dave Jiang 已提交
44 45 46 47
#if defined(CONFIG_EDAC)
#include <linux/edac.h>
#endif

V
Vegard Nossum 已提交
48
#include <asm/kmemcheck.h>
I
Ingo Molnar 已提交
49
#include <asm/stacktrace.h>
L
Linus Torvalds 已提交
50 51
#include <asm/processor.h>
#include <asm/debugreg.h>
I
Ingo Molnar 已提交
52 53
#include <asm/atomic.h>
#include <asm/system.h>
54
#include <asm/traps.h>
L
Linus Torvalds 已提交
55 56
#include <asm/desc.h>
#include <asm/i387.h>
H
Hidetoshi Seto 已提交
57
#include <asm/mce.h>
58

59
#include <asm/mach_traps.h>
60

61
#ifdef CONFIG_X86_64
62
#include <asm/x86_init.h>
63 64 65
#include <asm/pgalloc.h>
#include <asm/proto.h>
#else
66
#include <asm/processor-flags.h>
I
Ingo Molnar 已提交
67
#include <asm/setup.h>
L
Linus Torvalds 已提交
68 69 70 71

asmlinkage int system_call(void);

/* Do we ignore FPU interrupts ? */
I
Ingo Molnar 已提交
72
char ignore_fpu_irq;
L
Linus Torvalds 已提交
73 74 75

/*
 * The IDT has to be page-aligned to simplify the Pentium
76
 * F0 0F bug workaround.
L
Linus Torvalds 已提交
77
 */
78
gate_desc idt_table[NR_VECTORS] __page_aligned_data = { { { { 0, 0 } } }, };
79
#endif
L
Linus Torvalds 已提交
80

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

84
static int ignore_nmis;
85

86 87
int unknown_nmi_panic;

88 89 90 91 92 93
static inline void conditional_sti(struct pt_regs *regs)
{
	if (regs->flags & X86_EFLAGS_IF)
		local_irq_enable();
}

94 95 96 97 98 99 100
static inline void preempt_conditional_sti(struct pt_regs *regs)
{
	inc_preempt_count();
	if (regs->flags & X86_EFLAGS_IF)
		local_irq_enable();
}

T
Thomas Gleixner 已提交
101 102 103 104 105 106
static inline void conditional_cli(struct pt_regs *regs)
{
	if (regs->flags & X86_EFLAGS_IF)
		local_irq_disable();
}

107 108 109 110 111 112 113
static inline void preempt_conditional_cli(struct pt_regs *regs)
{
	if (regs->flags & X86_EFLAGS_IF)
		local_irq_disable();
	dec_preempt_count();
}

I
Ingo Molnar 已提交
114
static void __kprobes
115
do_trap(int trapnr, int signr, char *str, struct pt_regs *regs,
I
Ingo Molnar 已提交
116
	long error_code, siginfo_t *info)
L
Linus Torvalds 已提交
117
{
118 119
	struct task_struct *tsk = current;

120
#ifdef CONFIG_X86_32
121
	if (regs->flags & X86_VM_MASK) {
122 123 124 125 126
		/*
		 * traps 0, 1, 3, 4, and 5 should be forwarded to vm86.
		 * On nmi (interrupt 2), do_trap should not be called.
		 */
		if (trapnr < 6)
L
Linus Torvalds 已提交
127 128 129
			goto vm86_trap;
		goto trap_signal;
	}
130
#endif
L
Linus Torvalds 已提交
131

132
	if (!user_mode(regs))
L
Linus Torvalds 已提交
133 134
		goto kernel_trap;

135
#ifdef CONFIG_X86_32
I
Ingo Molnar 已提交
136
trap_signal:
137
#endif
I
Ingo Molnar 已提交
138 139 140 141 142 143 144 145 146 147 148
	/*
	 * We want error_code and trap_no set for userspace faults and
	 * 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;
	tsk->thread.trap_no = trapnr;
149

150 151 152 153 154 155 156 157 158 159 160 161
#ifdef CONFIG_X86_64
	if (show_unhandled_signals && unhandled_signal(tsk, signr) &&
	    printk_ratelimit()) {
		printk(KERN_INFO
		       "%s[%d] trap %s ip:%lx sp:%lx error:%lx",
		       tsk->comm, tsk->pid, str,
		       regs->ip, regs->sp, error_code);
		print_vma_addr(" in ", regs->ip);
		printk("\n");
	}
#endif

I
Ingo Molnar 已提交
162 163 164 165 166
	if (info)
		force_sig_info(signr, info, tsk);
	else
		force_sig(signr, tsk);
	return;
L
Linus Torvalds 已提交
167

I
Ingo Molnar 已提交
168 169 170 171 172
kernel_trap:
	if (!fixup_exception(regs)) {
		tsk->thread.error_code = error_code;
		tsk->thread.trap_no = trapnr;
		die(str, regs, error_code);
L
Linus Torvalds 已提交
173
	}
I
Ingo Molnar 已提交
174
	return;
L
Linus Torvalds 已提交
175

176
#ifdef CONFIG_X86_32
I
Ingo Molnar 已提交
177 178 179 180 181
vm86_trap:
	if (handle_vm86_trap((struct kernel_vm86_regs *) regs,
						error_code, trapnr))
		goto trap_signal;
	return;
182
#endif
L
Linus Torvalds 已提交
183 184
}

I
Ingo Molnar 已提交
185
#define DO_ERROR(trapnr, signr, str, name)				\
186
dotraplinkage void do_##name(struct pt_regs *regs, long error_code)	\
I
Ingo Molnar 已提交
187 188
{									\
	if (notify_die(DIE_TRAP, str, regs, error_code, trapnr, signr)	\
189
							== NOTIFY_STOP)	\
I
Ingo Molnar 已提交
190
		return;							\
191
	conditional_sti(regs);						\
192
	do_trap(trapnr, signr, str, regs, error_code, NULL);		\
L
Linus Torvalds 已提交
193 194
}

195
#define DO_ERROR_INFO(trapnr, signr, str, name, sicode, siaddr)		\
196
dotraplinkage void do_##name(struct pt_regs *regs, long error_code)	\
I
Ingo Molnar 已提交
197 198 199 200 201 202 203
{									\
	siginfo_t info;							\
	info.si_signo = signr;						\
	info.si_errno = 0;						\
	info.si_code = sicode;						\
	info.si_addr = (void __user *)siaddr;				\
	if (notify_die(DIE_TRAP, str, regs, error_code, trapnr, signr)	\
204
							== NOTIFY_STOP)	\
I
Ingo Molnar 已提交
205
		return;							\
206
	conditional_sti(regs);						\
207
	do_trap(trapnr, signr, str, regs, error_code, &info);		\
L
Linus Torvalds 已提交
208 209
}

210 211 212 213
DO_ERROR_INFO(0, SIGFPE, "divide error", divide_error, FPE_INTDIV, regs->ip)
DO_ERROR(4, SIGSEGV, "overflow", overflow)
DO_ERROR(5, SIGSEGV, "bounds", bounds)
DO_ERROR_INFO(6, SIGILL, "invalid opcode", invalid_op, ILL_ILLOPN, regs->ip)
214
DO_ERROR(9, SIGFPE, "coprocessor segment overrun", coprocessor_segment_overrun)
215
DO_ERROR(10, SIGSEGV, "invalid TSS", invalid_TSS)
216
DO_ERROR(11, SIGBUS, "segment not present", segment_not_present)
217
#ifdef CONFIG_X86_32
218
DO_ERROR(12, SIGBUS, "stack segment", stack_segment)
219
#endif
220
DO_ERROR_INFO(17, SIGBUS, "alignment check", alignment_check, BUS_ADRALN, 0)
L
Linus Torvalds 已提交
221

222 223 224 225 226 227 228 229 230 231 232 233 234 235 236 237 238 239 240 241 242 243 244
#ifdef CONFIG_X86_64
/* Runs on IST stack */
dotraplinkage void do_stack_segment(struct pt_regs *regs, long error_code)
{
	if (notify_die(DIE_TRAP, "stack segment", regs, error_code,
			12, SIGBUS) == NOTIFY_STOP)
		return;
	preempt_conditional_sti(regs);
	do_trap(12, SIGBUS, "stack segment", regs, error_code, NULL);
	preempt_conditional_cli(regs);
}

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

	/* Return not checked because double check cannot be ignored */
	notify_die(DIE_TRAP, str, regs, error_code, 8, SIGSEGV);

	tsk->thread.error_code = error_code;
	tsk->thread.trap_no = 8;

245 246 247 248
	/*
	 * This is always a kernel trap and never fixable (and thus must
	 * never return).
	 */
249 250 251 252 253
	for (;;)
		die(str, regs, error_code);
}
#endif

254
dotraplinkage void __kprobes
255
do_general_protection(struct pt_regs *regs, long error_code)
L
Linus Torvalds 已提交
256
{
257
	struct task_struct *tsk;
I
Ingo Molnar 已提交
258

259 260
	conditional_sti(regs);

261
#ifdef CONFIG_X86_32
262
	if (regs->flags & X86_VM_MASK)
L
Linus Torvalds 已提交
263
		goto gp_in_vm86;
264
#endif
L
Linus Torvalds 已提交
265

266
	tsk = current;
267
	if (!user_mode(regs))
L
Linus Torvalds 已提交
268 269
		goto gp_in_kernel;

270 271
	tsk->thread.error_code = error_code;
	tsk->thread.trap_no = 13;
I
Ingo Molnar 已提交
272

273 274
	if (show_unhandled_signals && unhandled_signal(tsk, SIGSEGV) &&
			printk_ratelimit()) {
275
		printk(KERN_INFO
276 277 278
			"%s[%d] general protection ip:%lx sp:%lx error:%lx",
			tsk->comm, task_pid_nr(tsk),
			regs->ip, regs->sp, error_code);
279 280 281
		print_vma_addr(" in ", regs->ip);
		printk("\n");
	}
282

283
	force_sig(SIGSEGV, tsk);
L
Linus Torvalds 已提交
284 285
	return;

286
#ifdef CONFIG_X86_32
L
Linus Torvalds 已提交
287 288 289 290
gp_in_vm86:
	local_irq_enable();
	handle_vm86_fault((struct kernel_vm86_regs *) regs, error_code);
	return;
291
#endif
L
Linus Torvalds 已提交
292 293

gp_in_kernel:
294 295 296 297 298 299
	if (fixup_exception(regs))
		return;

	tsk->thread.error_code = error_code;
	tsk->thread.trap_no = 13;
	if (notify_die(DIE_GPF, "general protection fault", regs,
L
Linus Torvalds 已提交
300
				error_code, 13, SIGSEGV) == NOTIFY_STOP)
301 302
		return;
	die("general protection fault", regs, error_code);
L
Linus Torvalds 已提交
303 304
}

305 306 307 308 309 310 311
static int __init setup_unknown_nmi_panic(char *str)
{
	unknown_nmi_panic = 1;
	return 1;
}
__setup("unknown_nmi_panic", setup_unknown_nmi_panic);

312
static notrace __kprobes void
313
pci_serr_error(unsigned char reason, struct pt_regs *regs)
L
Linus Torvalds 已提交
314
{
315 316
	pr_emerg("NMI: PCI system error (SERR) for reason %02x on CPU %d.\n",
		 reason, smp_processor_id());
D
Dave Jiang 已提交
317

318 319 320 321
	/*
	 * On some machines, PCI SERR line is used to report memory
	 * errors. EDAC makes use of it.
	 */
D
Dave Jiang 已提交
322
#if defined(CONFIG_EDAC)
I
Ingo Molnar 已提交
323
	if (edac_handler_set()) {
D
Dave Jiang 已提交
324 325 326 327 328
		edac_atomic_assert_error();
		return;
	}
#endif

329
	if (panic_on_unrecovered_nmi)
I
Ingo Molnar 已提交
330
		panic("NMI: Not continuing");
L
Linus Torvalds 已提交
331

332
	pr_emerg("Dazed and confused, but trying to continue\n");
L
Linus Torvalds 已提交
333

334 335 336
	/* Clear and disable the PCI SERR error line. */
	reason = (reason & NMI_REASON_CLEAR_MASK) | NMI_REASON_CLEAR_SERR;
	outb(reason, NMI_REASON_PORT);
L
Linus Torvalds 已提交
337 338
}

339
static notrace __kprobes void
I
Ingo Molnar 已提交
340
io_check_error(unsigned char reason, struct pt_regs *regs)
L
Linus Torvalds 已提交
341 342 343
{
	unsigned long i;

344 345 346
	pr_emerg(
	"NMI: IOCK error (debug interrupt?) for reason %02x on CPU %d.\n",
		 reason, smp_processor_id());
L
Linus Torvalds 已提交
347 348
	show_registers(regs);

349 350 351
	if (panic_on_io_nmi)
		panic("NMI IOCK error: Not continuing");

L
Linus Torvalds 已提交
352
	/* Re-enable the IOCK line, wait for a few seconds */
353 354
	reason = (reason & NMI_REASON_CLEAR_MASK) | NMI_REASON_CLEAR_IOCHK;
	outb(reason, NMI_REASON_PORT);
I
Ingo Molnar 已提交
355

356 357 358 359 360
	i = 20000;
	while (--i) {
		touch_nmi_watchdog();
		udelay(100);
	}
I
Ingo Molnar 已提交
361

362 363
	reason &= ~NMI_REASON_CLEAR_IOCHK;
	outb(reason, NMI_REASON_PORT);
L
Linus Torvalds 已提交
364 365
}

366
static notrace __kprobes void
I
Ingo Molnar 已提交
367
unknown_nmi_error(unsigned char reason, struct pt_regs *regs)
L
Linus Torvalds 已提交
368
{
369 370
	if (notify_die(DIE_NMIUNKNOWN, "nmi", regs, reason, 2, SIGINT) ==
			NOTIFY_STOP)
J
Jason Wessel 已提交
371
		return;
L
Linus Torvalds 已提交
372
#ifdef CONFIG_MCA
I
Ingo Molnar 已提交
373 374 375 376 377
	/*
	 * Might actually be able to figure out what the guilty party
	 * is:
	 */
	if (MCA_bus) {
L
Linus Torvalds 已提交
378 379 380 381
		mca_handle_nmi();
		return;
	}
#endif
382 383
	pr_emerg("Uhhuh. NMI received for unknown reason %02x on CPU %d.\n",
		 reason, smp_processor_id());
I
Ingo Molnar 已提交
384

385
	pr_emerg("Do you have a strange power saving mode enabled?\n");
386
	if (unknown_nmi_panic || panic_on_unrecovered_nmi)
I
Ingo Molnar 已提交
387
		panic("NMI: Not continuing");
388

389
	pr_emerg("Dazed and confused, but trying to continue\n");
L
Linus Torvalds 已提交
390 391
}

392
static notrace __kprobes void default_do_nmi(struct pt_regs *regs)
L
Linus Torvalds 已提交
393 394
{
	unsigned char reason = 0;
395 396 397
	int cpu;

	cpu = smp_processor_id();
L
Linus Torvalds 已提交
398

399 400
	/* Only the BSP gets external NMIs from the system. */
	if (!cpu)
L
Linus Torvalds 已提交
401
		reason = get_nmi_reason();
I
Ingo Molnar 已提交
402

403
	if (!(reason & NMI_REASON_MASK)) {
404
		if (notify_die(DIE_NMI_IPI, "nmi_ipi", regs, reason, 2, SIGINT)
405
								== NOTIFY_STOP)
L
Linus Torvalds 已提交
406
			return;
407

L
Linus Torvalds 已提交
408
#ifdef CONFIG_X86_LOCAL_APIC
409 410 411
		if (notify_die(DIE_NMI, "nmi", regs, reason, 2, SIGINT)
							== NOTIFY_STOP)
			return;
I
Ingo Molnar 已提交
412
#endif
413
		unknown_nmi_error(reason, regs);
414

L
Linus Torvalds 已提交
415 416
		return;
	}
417
	if (notify_die(DIE_NMI, "nmi", regs, reason, 2, SIGINT) == NOTIFY_STOP)
L
Linus Torvalds 已提交
418
		return;
419 420

	/* AK: following checks seem to be broken on modern chipsets. FIXME */
421 422 423
	if (reason & NMI_REASON_SERR)
		pci_serr_error(reason, regs);
	if (reason & NMI_REASON_IOCHK)
L
Linus Torvalds 已提交
424
		io_check_error(reason, regs);
425
#ifdef CONFIG_X86_32
L
Linus Torvalds 已提交
426 427
	/*
	 * Reassert NMI in case it became active meanwhile
I
Ingo Molnar 已提交
428
	 * as it's edge-triggered:
L
Linus Torvalds 已提交
429 430
	 */
	reassert_nmi();
431
#endif
L
Linus Torvalds 已提交
432 433
}

434 435
dotraplinkage notrace __kprobes void
do_nmi(struct pt_regs *regs, long error_code)
L
Linus Torvalds 已提交
436 437 438
{
	nmi_enter();

439
	inc_irq_stat(__nmi_count);
L
Linus Torvalds 已提交
440

441 442
	if (!ignore_nmis)
		default_do_nmi(regs);
L
Linus Torvalds 已提交
443 444 445 446

	nmi_exit();
}

447 448 449 450 451 452 453 454 455 456
void stop_nmi(void)
{
	ignore_nmis++;
}

void restart_nmi(void)
{
	ignore_nmis--;
}

457
/* May run on IST stack. */
458
dotraplinkage void __kprobes do_int3(struct pt_regs *regs, long error_code)
L
Linus Torvalds 已提交
459
{
460 461 462 463 464
#ifdef CONFIG_KGDB_LOW_LEVEL_TRAP
	if (kgdb_ll_trap(DIE_INT3, "int3", regs, error_code, 3, SIGTRAP)
			== NOTIFY_STOP)
		return;
#endif /* CONFIG_KGDB_LOW_LEVEL_TRAP */
465
#ifdef CONFIG_KPROBES
L
Linus Torvalds 已提交
466 467
	if (notify_die(DIE_INT3, "int3", regs, error_code, 3, SIGTRAP)
			== NOTIFY_STOP)
468
		return;
469 470 471 472 473
#else
	if (notify_die(DIE_TRAP, "int3", regs, error_code, 3, SIGTRAP)
			== NOTIFY_STOP)
		return;
#endif
I
Ingo Molnar 已提交
474

475
	preempt_conditional_sti(regs);
476
	do_trap(3, SIGTRAP, "int3", regs, error_code, NULL);
477
	preempt_conditional_cli(regs);
L
Linus Torvalds 已提交
478 479
}

480
#ifdef CONFIG_X86_64
481 482 483 484 485
/*
 * 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
 */
486 487 488 489 490 491 492 493 494
asmlinkage __kprobes struct pt_regs *sync_regs(struct pt_regs *eregs)
{
	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);
495 496 497 498
	/*
	 * Exception from kernel and interrupts are enabled. Move to
	 * kernel process stack.
	 */
499 500 501 502 503 504 505 506
	else if (eregs->flags & X86_EFLAGS_IF)
		regs = (struct pt_regs *)(eregs->sp -= sizeof(struct pt_regs));
	if (eregs != regs)
		*regs = *eregs;
	return regs;
}
#endif

L
Linus Torvalds 已提交
507 508 509 510 511 512 513 514 515 516
/*
 * 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
517
 *
L
Linus Torvalds 已提交
518 519 520 521 522 523 524 525 526 527
 * 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)
528 529
 *
 * May run on IST stack.
L
Linus Torvalds 已提交
530
 */
531
dotraplinkage void __kprobes do_debug(struct pt_regs *regs, long error_code)
L
Linus Torvalds 已提交
532 533
{
	struct task_struct *tsk = current;
534
	int user_icebp = 0;
535
	unsigned long dr6;
536
	int si_code;
L
Linus Torvalds 已提交
537

538
	get_debugreg(dr6, 6);
L
Linus Torvalds 已提交
539

540 541 542
	/* Filter out all the reserved bits which are preset to 1 */
	dr6 &= ~DR6_RESERVED;

543 544 545 546 547 548 549 550
	/*
	 * 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
551
	/* Catch kmemcheck conditions first of all! */
552
	if ((dr6 & DR_STEP) && kmemcheck_trap(regs))
V
Vegard Nossum 已提交
553 554
		return;

555 556
	/* DR6 may or may not be cleared by the CPU */
	set_debugreg(0, 6);
557

P
Peter Zijlstra 已提交
558 559 560 561 562
	/*
	 * The processor cleared BTF, so don't mark that we need it set.
	 */
	clear_tsk_thread_flag(tsk, TIF_BLOCKSTEP);

563 564 565
	/* Store the virtualized DR6 value */
	tsk->thread.debugreg6 = dr6;

566 567
	if (notify_die(DIE_DEBUG, "debug", regs, PTR_ERR(&dr6), error_code,
							SIGTRAP) == NOTIFY_STOP)
L
Linus Torvalds 已提交
568
		return;
569

L
Linus Torvalds 已提交
570
	/* It's safe to allow irq's after DR6 has been saved */
571
	preempt_conditional_sti(regs);
L
Linus Torvalds 已提交
572

573 574 575
	if (regs->flags & X86_VM_MASK) {
		handle_vm86_trap((struct kernel_vm86_regs *) regs,
				error_code, 1);
576
		preempt_conditional_cli(regs);
577
		return;
L
Linus Torvalds 已提交
578 579 580
	}

	/*
581 582 583 584 585
	 * 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
586
	 */
587 588 589 590
	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
591
	}
592
	si_code = get_si_code(tsk->thread.debugreg6);
593
	if (tsk->thread.debugreg6 & (DR_STEP | DR_TRAP_BITS) || user_icebp)
594
		send_sigtrap(tsk, regs, error_code, si_code);
595
	preempt_conditional_cli(regs);
L
Linus Torvalds 已提交
596 597 598 599 600 601 602 603 604

	return;
}

/*
 * 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
 */
605
void math_error(struct pt_regs *regs, int error_code, int trapnr)
L
Linus Torvalds 已提交
606
{
607
	struct task_struct *task = current;
L
Linus Torvalds 已提交
608
	siginfo_t info;
609
	unsigned short err;
610 611 612 613 614 615 616 617 618 619 620 621 622 623 624
	char *str = (trapnr == 16) ? "fpu exception" : "simd exception";

	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;
			task->thread.trap_no = trapnr;
			die(str, regs, error_code);
		}
		return;
	}
L
Linus Torvalds 已提交
625 626 627 628 629

	/*
	 * Save the info for the exception handler and clear the error.
	 */
	save_init_fpu(task);
630 631
	task->thread.trap_no = trapnr;
	task->thread.error_code = error_code;
L
Linus Torvalds 已提交
632 633
	info.si_signo = SIGFPE;
	info.si_errno = 0;
634 635 636 637 638 639 640 641 642 643 644 645 646 647 648
	info.si_addr = (void __user *)regs->ip;
	if (trapnr == 16) {
		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);
649

650 651 652 653 654 655 656 657 658 659 660
		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;
	}
661 662

	if (err & 0x001) {	/* Invalid op */
I
Ingo Molnar 已提交
663 664 665 666 667 668
		/*
		 * swd & 0x240 == 0x040: Stack Underflow
		 * swd & 0x240 == 0x240: Stack Overflow
		 * User must clear the SF bit (0x40) if set
		 */
		info.si_code = FPE_FLTINV;
669
	} else if (err & 0x004) { /* Divide by Zero */
I
Ingo Molnar 已提交
670
		info.si_code = FPE_FLTDIV;
671
	} else if (err & 0x008) { /* Overflow */
I
Ingo Molnar 已提交
672
		info.si_code = FPE_FLTOVF;
673 674 675
	} else if (err & 0x012) { /* Denormal, Underflow */
		info.si_code = FPE_FLTUND;
	} else if (err & 0x020) { /* Precision */
I
Ingo Molnar 已提交
676
		info.si_code = FPE_FLTRES;
677
	} else {
678 679 680 681
		/*
		 * If we're using IRQ 13, or supposedly even some trap 16
		 * implementations, it's possible we get a spurious trap...
		 */
682
		return;		/* Spurious trap, no error */
L
Linus Torvalds 已提交
683 684 685 686
	}
	force_sig_info(SIGFPE, &info, task);
}

687
dotraplinkage void do_coprocessor_error(struct pt_regs *regs, long error_code)
L
Linus Torvalds 已提交
688
{
689
#ifdef CONFIG_X86_32
L
Linus Torvalds 已提交
690
	ignore_fpu_irq = 1;
691 692
#endif

693
	math_error(regs, error_code, 16);
L
Linus Torvalds 已提交
694 695
}

696 697
dotraplinkage void
do_simd_coprocessor_error(struct pt_regs *regs, long error_code)
L
Linus Torvalds 已提交
698
{
699
	math_error(regs, error_code, 19);
L
Linus Torvalds 已提交
700 701
}

702 703
dotraplinkage void
do_spurious_interrupt_bug(struct pt_regs *regs, long error_code)
L
Linus Torvalds 已提交
704
{
705
	conditional_sti(regs);
L
Linus Torvalds 已提交
706 707
#if 0
	/* No need to warn about this any longer. */
I
Ingo Molnar 已提交
708
	printk(KERN_INFO "Ignoring P6 Local APIC Spurious Interrupt Bug...\n");
L
Linus Torvalds 已提交
709 710 711
#endif
}

712
asmlinkage void __attribute__((weak)) smp_thermal_interrupt(void)
L
Linus Torvalds 已提交
713 714
{
}
715

716
asmlinkage void __attribute__((weak)) smp_threshold_interrupt(void)
717 718 719
{
}

720 721 722 723 724
/*
 * __math_state_restore assumes that cr0.TS is already clear and the
 * fpu state is all ready for use.  Used during context switch.
 */
void __math_state_restore(void)
725
{
726 727 728 729 730 731 732 733 734 735 736 737 738 739
	struct thread_info *thread = current_thread_info();
	struct task_struct *tsk = thread->task;

	/*
	 * Paranoid restore. send a SIGSEGV if we fail to restore the state.
	 */
	if (unlikely(restore_fpu_checking(tsk))) {
		stts();
		force_sig(SIGSEGV, tsk);
		return;
	}

	thread->status |= TS_USEDFPU;	/* So we fnsave on switch_to() */
	tsk->fpu_counter++;
740
}
L
Linus Torvalds 已提交
741 742

/*
I
Ingo Molnar 已提交
743
 * 'math_state_restore()' saves the current math information in the
L
Linus Torvalds 已提交
744 745 746 747 748 749 750 751
 * 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.
 *
 * Must be called with kernel preemption disabled (in this case,
 * local interrupts are disabled at the call-site in entry.S).
 */
752
asmlinkage void math_state_restore(void)
L
Linus Torvalds 已提交
753 754 755 756
{
	struct thread_info *thread = current_thread_info();
	struct task_struct *tsk = thread->task;

757 758 759 760 761 762 763 764 765 766 767 768 769 770 771
	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();
	}

I
Ingo Molnar 已提交
772
	clts();				/* Allow maths ops (or we recurse) */
773

774
	__math_state_restore();
L
Linus Torvalds 已提交
775
}
776
EXPORT_SYMBOL_GPL(math_state_restore);
L
Linus Torvalds 已提交
777

778
dotraplinkage void __kprobes
779
do_device_not_available(struct pt_regs *regs, long error_code)
780
{
781
#ifdef CONFIG_MATH_EMULATION
782
	if (read_cr0() & X86_CR0_EM) {
783 784
		struct math_emu_info info = { };

785
		conditional_sti(regs);
786

787
		info.regs = regs;
788
		math_emulate(&info);
789
		return;
790
	}
791 792 793 794
#endif
	math_state_restore(); /* interrupts still off */
#ifdef CONFIG_X86_32
	conditional_sti(regs);
795
#endif
796 797
}

798
#ifdef CONFIG_X86_32
799
dotraplinkage void do_iret_error(struct pt_regs *regs, long error_code)
800 801 802 803 804 805 806
{
	siginfo_t info;
	local_irq_enable();

	info.si_signo = SIGILL;
	info.si_errno = 0;
	info.si_code = ILL_BADSTK;
807
	info.si_addr = NULL;
808 809 810
	if (notify_die(DIE_TRAP, "iret exception",
			regs, error_code, 32, SIGILL) == NOTIFY_STOP)
		return;
811
	do_trap(32, SIGILL, "iret exception", regs, error_code, &info);
812
}
813
#endif
814

815 816 817 818 819 820 821 822 823 824
/* Set of traps needed for early debugging. */
void __init early_trap_init(void)
{
	set_intr_gate_ist(1, &debug, DEBUG_STACK);
	/* int3 can be called from all */
	set_system_intr_gate_ist(3, &int3, DEBUG_STACK);
	set_intr_gate(14, &page_fault);
	load_idt(&idt_descr);
}

L
Linus Torvalds 已提交
825 826
void __init trap_init(void)
{
827 828
	int i;

L
Linus Torvalds 已提交
829
#ifdef CONFIG_EISA
I
Ingo Molnar 已提交
830
	void __iomem *p = early_ioremap(0x0FFFD9, 4);
I
Ingo Molnar 已提交
831 832

	if (readl(p) == 'E' + ('I'<<8) + ('S'<<16) + ('A'<<24))
L
Linus Torvalds 已提交
833
		EISA_bus = 1;
I
Ingo Molnar 已提交
834
	early_iounmap(p, 4);
L
Linus Torvalds 已提交
835 836
#endif

837
	set_intr_gate(0, &divide_error);
838 839 840
	set_intr_gate_ist(2, &nmi, NMI_STACK);
	/* int4 can be called from all */
	set_system_intr_gate(4, &overflow);
841
	set_intr_gate(5, &bounds);
842
	set_intr_gate(6, &invalid_op);
843
	set_intr_gate(7, &device_not_available);
844
#ifdef CONFIG_X86_32
845
	set_task_gate(8, GDT_ENTRY_DOUBLEFAULT_TSS);
846 847 848
#else
	set_intr_gate_ist(8, &double_fault, DOUBLEFAULT_STACK);
#endif
849
	set_intr_gate(9, &coprocessor_segment_overrun);
850
	set_intr_gate(10, &invalid_TSS);
851
	set_intr_gate(11, &segment_not_present);
852
	set_intr_gate_ist(12, &stack_segment, STACKFAULT_STACK);
853
	set_intr_gate(13, &general_protection);
854
	set_intr_gate(15, &spurious_interrupt_bug);
855
	set_intr_gate(16, &coprocessor_error);
856
	set_intr_gate(17, &alignment_check);
L
Linus Torvalds 已提交
857
#ifdef CONFIG_X86_MCE
858
	set_intr_gate_ist(18, &machine_check, MCE_STACK);
L
Linus Torvalds 已提交
859
#endif
860
	set_intr_gate(19, &simd_coprocessor_error);
L
Linus Torvalds 已提交
861

862 863 864 865
	/* Reserve all the builtin and the syscall vector: */
	for (i = 0; i < FIRST_EXTERNAL_VECTOR; i++)
		set_bit(i, used_vectors);

866 867
#ifdef CONFIG_IA32_EMULATION
	set_system_intr_gate(IA32_SYSCALL_VECTOR, ia32_syscall);
868
	set_bit(IA32_SYSCALL_VECTOR, used_vectors);
869 870 871
#endif

#ifdef CONFIG_X86_32
872
	set_system_trap_gate(SYSCALL_VECTOR, &system_call);
873
	set_bit(SYSCALL_VECTOR, used_vectors);
874
#endif
875

L
Linus Torvalds 已提交
876
	/*
I
Ingo Molnar 已提交
877
	 * Should be a barrier for any external CPU state:
L
Linus Torvalds 已提交
878 879 880
	 */
	cpu_init();

881
	x86_init.irqs.trap_init();
L
Linus Torvalds 已提交
882
}