traps.c 23.8 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 18
#include <linux/interrupt.h>
#include <linux/kallsyms.h>
#include <linux/spinlock.h>
#include <linux/kprobes.h>
#include <linux/uaccess.h>
#include <linux/utsname.h>
#include <linux/kdebug.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 62 63 64
#ifdef CONFIG_X86_64
#include <asm/pgalloc.h>
#include <asm/proto.h>
#else
65
#include <asm/processor-flags.h>
I
Ingo Molnar 已提交
66
#include <asm/setup.h>
67
#include <asm/traps.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 76 77 78

/*
 * The IDT has to be page-aligned to simplify the Pentium
 * F0 0F bug workaround.. We have a special link segment
 * for this.
 */
79
gate_desc idt_table[NR_VECTORS]
80
	__attribute__((__section__(".data.idt"))) = { { { { 0, 0 } } }, };
81
#endif
L
Linus Torvalds 已提交
82

83 84 85
DECLARE_BITMAP(used_vectors, NR_VECTORS);
EXPORT_SYMBOL_GPL(used_vectors);

86
static int ignore_nmis;
87

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();
}

114
#ifdef CONFIG_X86_32
I
Ingo Molnar 已提交
115 116
static inline void
die_if_kernel(const char *str, struct pt_regs *regs, long err)
L
Linus Torvalds 已提交
117
{
118
	if (!user_mode_vm(regs))
L
Linus Torvalds 已提交
119 120
		die(str, regs, err);
}
121
#endif
122

I
Ingo Molnar 已提交
123
static void __kprobes
124
do_trap(int trapnr, int signr, char *str, struct pt_regs *regs,
I
Ingo Molnar 已提交
125
	long error_code, siginfo_t *info)
L
Linus Torvalds 已提交
126
{
127 128
	struct task_struct *tsk = current;

129
#ifdef CONFIG_X86_32
130
	if (regs->flags & X86_VM_MASK) {
131 132 133 134 135
		/*
		 * 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
136 137 138
			goto vm86_trap;
		goto trap_signal;
	}
139
#endif
L
Linus Torvalds 已提交
140

141
	if (!user_mode(regs))
L
Linus Torvalds 已提交
142 143
		goto kernel_trap;

144
#ifdef CONFIG_X86_32
I
Ingo Molnar 已提交
145
trap_signal:
146
#endif
I
Ingo Molnar 已提交
147 148 149 150 151 152 153 154 155 156 157
	/*
	 * 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;
158

159 160 161 162 163 164 165 166 167 168 169 170
#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
171 172 173 174 175
	if (info)
		force_sig_info(signr, info, tsk);
	else
		force_sig(signr, tsk);
	return;
L
Linus Torvalds 已提交
176

I
Ingo Molnar 已提交
177 178 179 180 181
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
182
	}
I
Ingo Molnar 已提交
183
	return;
L
Linus Torvalds 已提交
184

185
#ifdef CONFIG_X86_32
I
Ingo Molnar 已提交
186 187 188 189 190
vm86_trap:
	if (handle_vm86_trap((struct kernel_vm86_regs *) regs,
						error_code, trapnr))
		goto trap_signal;
	return;
191
#endif
L
Linus Torvalds 已提交
192 193
}

I
Ingo Molnar 已提交
194
#define DO_ERROR(trapnr, signr, str, name)				\
195
dotraplinkage void do_##name(struct pt_regs *regs, long error_code)	\
I
Ingo Molnar 已提交
196 197
{									\
	if (notify_die(DIE_TRAP, str, regs, error_code, trapnr, signr)	\
198
							== NOTIFY_STOP)	\
I
Ingo Molnar 已提交
199
		return;							\
200
	conditional_sti(regs);						\
201
	do_trap(trapnr, signr, str, regs, error_code, NULL);		\
L
Linus Torvalds 已提交
202 203
}

204
#define DO_ERROR_INFO(trapnr, signr, str, name, sicode, siaddr)		\
205
dotraplinkage void do_##name(struct pt_regs *regs, long error_code)	\
I
Ingo Molnar 已提交
206 207 208 209 210 211 212
{									\
	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)	\
213
							== NOTIFY_STOP)	\
I
Ingo Molnar 已提交
214
		return;							\
215
	conditional_sti(regs);						\
216
	do_trap(trapnr, signr, str, regs, error_code, &info);		\
L
Linus Torvalds 已提交
217 218
}

219 220 221 222
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)
223
DO_ERROR(9, SIGFPE, "coprocessor segment overrun", coprocessor_segment_overrun)
224
DO_ERROR(10, SIGSEGV, "invalid TSS", invalid_TSS)
225
DO_ERROR(11, SIGBUS, "segment not present", segment_not_present)
226
#ifdef CONFIG_X86_32
227
DO_ERROR(12, SIGBUS, "stack segment", stack_segment)
228
#endif
229
DO_ERROR_INFO(17, SIGBUS, "alignment check", alignment_check, BUS_ADRALN, 0)
L
Linus Torvalds 已提交
230

231 232 233 234 235 236 237 238 239 240 241 242 243 244 245 246 247 248 249 250 251 252 253
#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;

254 255 256 257
	/*
	 * This is always a kernel trap and never fixable (and thus must
	 * never return).
	 */
258 259 260 261 262
	for (;;)
		die(str, regs, error_code);
}
#endif

263
dotraplinkage void __kprobes
264
do_general_protection(struct pt_regs *regs, long error_code)
L
Linus Torvalds 已提交
265
{
266
	struct task_struct *tsk;
I
Ingo Molnar 已提交
267

268 269
	conditional_sti(regs);

270
#ifdef CONFIG_X86_32
271
	if (regs->flags & X86_VM_MASK)
L
Linus Torvalds 已提交
272
		goto gp_in_vm86;
273
#endif
L
Linus Torvalds 已提交
274

275
	tsk = current;
276
	if (!user_mode(regs))
L
Linus Torvalds 已提交
277 278
		goto gp_in_kernel;

279 280
	tsk->thread.error_code = error_code;
	tsk->thread.trap_no = 13;
I
Ingo Molnar 已提交
281

282 283
	if (show_unhandled_signals && unhandled_signal(tsk, SIGSEGV) &&
			printk_ratelimit()) {
284
		printk(KERN_INFO
285 286 287
			"%s[%d] general protection ip:%lx sp:%lx error:%lx",
			tsk->comm, task_pid_nr(tsk),
			regs->ip, regs->sp, error_code);
288 289 290
		print_vma_addr(" in ", regs->ip);
		printk("\n");
	}
291

292
	force_sig(SIGSEGV, tsk);
L
Linus Torvalds 已提交
293 294
	return;

295
#ifdef CONFIG_X86_32
L
Linus Torvalds 已提交
296 297 298 299
gp_in_vm86:
	local_irq_enable();
	handle_vm86_fault((struct kernel_vm86_regs *) regs, error_code);
	return;
300
#endif
L
Linus Torvalds 已提交
301 302

gp_in_kernel:
303 304 305 306 307 308
	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
309
				error_code, 13, SIGSEGV) == NOTIFY_STOP)
310 311
		return;
	die("general protection fault", regs, error_code);
L
Linus Torvalds 已提交
312 313
}

314
static notrace __kprobes void
I
Ingo Molnar 已提交
315
mem_parity_error(unsigned char reason, struct pt_regs *regs)
L
Linus Torvalds 已提交
316
{
I
Ingo Molnar 已提交
317 318 319 320 321 322
	printk(KERN_EMERG
		"Uhhuh. NMI received for unknown reason %02x on CPU %d.\n",
			reason, smp_processor_id());

	printk(KERN_EMERG
		"You have some hardware problem, likely on the PCI bus.\n");
D
Dave Jiang 已提交
323 324

#if defined(CONFIG_EDAC)
I
Ingo Molnar 已提交
325
	if (edac_handler_set()) {
D
Dave Jiang 已提交
326 327 328 329 330
		edac_atomic_assert_error();
		return;
	}
#endif

331
	if (panic_on_unrecovered_nmi)
I
Ingo Molnar 已提交
332
		panic("NMI: Not continuing");
L
Linus Torvalds 已提交
333

334
	printk(KERN_EMERG "Dazed and confused, but trying to continue\n");
L
Linus Torvalds 已提交
335 336

	/* Clear and disable the memory parity error line. */
337 338
	reason = (reason & 0xf) | 4;
	outb(reason, 0x61);
L
Linus Torvalds 已提交
339 340
}

341
static notrace __kprobes void
I
Ingo Molnar 已提交
342
io_check_error(unsigned char reason, struct pt_regs *regs)
L
Linus Torvalds 已提交
343 344 345
{
	unsigned long i;

346
	printk(KERN_EMERG "NMI: IOCK error (debug interrupt?)\n");
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
352 353 354
	/* Re-enable the IOCK line, wait for a few seconds */
	reason = (reason & 0xf) | 8;
	outb(reason, 0x61);
I
Ingo Molnar 已提交
355

L
Linus Torvalds 已提交
356
	i = 2000;
I
Ingo Molnar 已提交
357 358 359
	while (--i)
		udelay(1000);

L
Linus Torvalds 已提交
360 361 362 363
	reason &= ~8;
	outb(reason, 0x61);
}

364
static notrace __kprobes void
I
Ingo Molnar 已提交
365
unknown_nmi_error(unsigned char reason, struct pt_regs *regs)
L
Linus Torvalds 已提交
366
{
367 368
	if (notify_die(DIE_NMIUNKNOWN, "nmi", regs, reason, 2, SIGINT) ==
			NOTIFY_STOP)
J
Jason Wessel 已提交
369
		return;
L
Linus Torvalds 已提交
370
#ifdef CONFIG_MCA
I
Ingo Molnar 已提交
371 372 373 374 375
	/*
	 * Might actually be able to figure out what the guilty party
	 * is:
	 */
	if (MCA_bus) {
L
Linus Torvalds 已提交
376 377 378 379
		mca_handle_nmi();
		return;
	}
#endif
I
Ingo Molnar 已提交
380 381 382 383
	printk(KERN_EMERG
		"Uhhuh. NMI received for unknown reason %02x on CPU %d.\n",
			reason, smp_processor_id());

384
	printk(KERN_EMERG "Do you have a strange power saving mode enabled?\n");
385
	if (panic_on_unrecovered_nmi)
I
Ingo Molnar 已提交
386
		panic("NMI: Not continuing");
387

388
	printk(KERN_EMERG "Dazed and confused, but trying to continue\n");
L
Linus Torvalds 已提交
389 390
}

391
static notrace __kprobes void default_do_nmi(struct pt_regs *regs)
L
Linus Torvalds 已提交
392 393
{
	unsigned char reason = 0;
394 395 396
	int cpu;

	cpu = smp_processor_id();
L
Linus Torvalds 已提交
397

398 399
	/* Only the BSP gets external NMIs from the system. */
	if (!cpu)
L
Linus Torvalds 已提交
400
		reason = get_nmi_reason();
I
Ingo Molnar 已提交
401

L
Linus Torvalds 已提交
402
	if (!(reason & 0xc0)) {
403
		if (notify_die(DIE_NMI_IPI, "nmi_ipi", regs, reason, 2, SIGINT)
404
								== NOTIFY_STOP)
L
Linus Torvalds 已提交
405 406 407 408 409 410
			return;
#ifdef CONFIG_X86_LOCAL_APIC
		/*
		 * Ok, so this is none of the documented NMI sources,
		 * so it must be the NMI watchdog.
		 */
411
		if (nmi_watchdog_tick(regs, reason))
L
Linus Torvalds 已提交
412
			return;
413
		if (!do_nmi_callback(regs, cpu))
414
			unknown_nmi_error(reason, regs);
I
Ingo Molnar 已提交
415 416 417
#else
		unknown_nmi_error(reason, regs);
#endif
418

L
Linus Torvalds 已提交
419 420
		return;
	}
421
	if (notify_die(DIE_NMI, "nmi", regs, reason, 2, SIGINT) == NOTIFY_STOP)
L
Linus Torvalds 已提交
422
		return;
423 424

	/* AK: following checks seem to be broken on modern chipsets. FIXME */
L
Linus Torvalds 已提交
425 426 427 428
	if (reason & 0x80)
		mem_parity_error(reason, regs);
	if (reason & 0x40)
		io_check_error(reason, regs);
429
#ifdef CONFIG_X86_32
L
Linus Torvalds 已提交
430 431
	/*
	 * Reassert NMI in case it became active meanwhile
I
Ingo Molnar 已提交
432
	 * as it's edge-triggered:
L
Linus Torvalds 已提交
433 434
	 */
	reassert_nmi();
435
#endif
L
Linus Torvalds 已提交
436 437
}

438 439
dotraplinkage notrace __kprobes void
do_nmi(struct pt_regs *regs, long error_code)
L
Linus Torvalds 已提交
440 441 442
{
	nmi_enter();

443
	inc_irq_stat(__nmi_count);
L
Linus Torvalds 已提交
444

445 446
	if (!ignore_nmis)
		default_do_nmi(regs);
L
Linus Torvalds 已提交
447 448 449 450

	nmi_exit();
}

451 452 453 454 455 456 457 458 459 460 461 462
void stop_nmi(void)
{
	acpi_nmi_disable();
	ignore_nmis++;
}

void restart_nmi(void)
{
	ignore_nmis--;
	acpi_nmi_enable();
}

463
/* May run on IST stack. */
464
dotraplinkage void __kprobes do_int3(struct pt_regs *regs, long error_code)
L
Linus Torvalds 已提交
465
{
466
#ifdef CONFIG_KPROBES
L
Linus Torvalds 已提交
467 468
	if (notify_die(DIE_INT3, "int3", regs, error_code, 3, SIGTRAP)
			== NOTIFY_STOP)
469
		return;
470 471 472 473 474
#else
	if (notify_die(DIE_TRAP, "int3", regs, error_code, 3, SIGTRAP)
			== NOTIFY_STOP)
		return;
#endif
I
Ingo Molnar 已提交
475

476
	preempt_conditional_sti(regs);
477
	do_trap(3, SIGTRAP, "int3", regs, error_code, NULL);
478
	preempt_conditional_cli(regs);
L
Linus Torvalds 已提交
479 480
}

481
#ifdef CONFIG_X86_64
482 483 484 485 486
/*
 * 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
 */
487 488 489 490 491 492 493 494 495
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);
496 497 498 499
	/*
	 * Exception from kernel and interrupts are enabled. Move to
	 * kernel process stack.
	 */
500 501 502 503 504 505 506 507
	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
508 509 510 511 512 513 514 515 516 517
/*
 * 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
518
 *
L
Linus Torvalds 已提交
519 520 521 522 523 524 525 526 527 528
 * 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)
529 530
 *
 * May run on IST stack.
L
Linus Torvalds 已提交
531
 */
532
dotraplinkage void __kprobes do_debug(struct pt_regs *regs, long error_code)
L
Linus Torvalds 已提交
533 534
{
	struct task_struct *tsk = current;
535
	unsigned long condition;
536
	int si_code;
L
Linus Torvalds 已提交
537

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

V
Vegard Nossum 已提交
540 541 542 543
	/* Catch kmemcheck conditions first of all! */
	if (condition & DR_STEP && kmemcheck_trap(regs))
		return;

544 545 546 547 548 549
	/*
	 * The processor cleared BTF, so don't mark that we need it set.
	 */
	clear_tsk_thread_flag(tsk, TIF_DEBUGCTLMSR);
	tsk->thread.debugctlmsr = 0;

L
Linus Torvalds 已提交
550
	if (notify_die(DIE_DEBUG, "debug", regs, condition, error_code,
551
						SIGTRAP) == NOTIFY_STOP)
L
Linus Torvalds 已提交
552
		return;
553

L
Linus Torvalds 已提交
554
	/* It's safe to allow irq's after DR6 has been saved */
555
	preempt_conditional_sti(regs);
L
Linus Torvalds 已提交
556 557 558

	/* Mask out spurious debug traps due to lazy DR7 setting */
	if (condition & (DR_TRAP0|DR_TRAP1|DR_TRAP2|DR_TRAP3)) {
559
		if (!tsk->thread.debugreg7)
L
Linus Torvalds 已提交
560 561 562
			goto clear_dr7;
	}

563
#ifdef CONFIG_X86_32
564
	if (regs->flags & X86_VM_MASK)
L
Linus Torvalds 已提交
565
		goto debug_vm86;
566
#endif
L
Linus Torvalds 已提交
567 568

	/* Save debug status register where ptrace can see it */
569
	tsk->thread.debugreg6 = condition;
L
Linus Torvalds 已提交
570 571 572 573 574 575

	/*
	 * Single-stepping through TF: make sure we ignore any events in
	 * kernel space (but re-enable TF when returning to user mode).
	 */
	if (condition & DR_STEP) {
576
		if (!user_mode(regs))
L
Linus Torvalds 已提交
577 578 579
			goto clear_TF_reenable;
	}

580
	si_code = get_si_code(condition);
L
Linus Torvalds 已提交
581
	/* Ok, finally something we can handle */
582
	send_sigtrap(tsk, regs, error_code, si_code);
L
Linus Torvalds 已提交
583

I
Ingo Molnar 已提交
584 585
	/*
	 * Disable additional traps. They'll be re-enabled when
L
Linus Torvalds 已提交
586 587 588
	 * the signal is delivered.
	 */
clear_dr7:
589
	set_debugreg(0, 7);
590
	preempt_conditional_cli(regs);
L
Linus Torvalds 已提交
591 592
	return;

593
#ifdef CONFIG_X86_32
L
Linus Torvalds 已提交
594
debug_vm86:
T
Thomas Gleixner 已提交
595 596
	/* reenable preemption: handle_vm86_trap() might sleep */
	dec_preempt_count();
L
Linus Torvalds 已提交
597
	handle_vm86_trap((struct kernel_vm86_regs *) regs, error_code, 1);
T
Thomas Gleixner 已提交
598
	conditional_cli(regs);
L
Linus Torvalds 已提交
599
	return;
600
#endif
L
Linus Torvalds 已提交
601 602 603

clear_TF_reenable:
	set_tsk_thread_flag(tsk, TIF_SINGLESTEP);
604
	regs->flags &= ~X86_EFLAGS_TF;
605
	preempt_conditional_cli(regs);
L
Linus Torvalds 已提交
606 607 608
	return;
}

609 610 611 612 613 614 615 616 617 618 619 620 621 622
#ifdef CONFIG_X86_64
static int kernel_math_error(struct pt_regs *regs, const char *str, int trapnr)
{
	if (fixup_exception(regs))
		return 1;

	notify_die(DIE_GPF, str, regs, 0, trapnr, SIGFPE);
	/* Illegal floating point operation in the kernel */
	current->thread.trap_no = trapnr;
	die(str, regs, 0);
	return 0;
}
#endif

L
Linus Torvalds 已提交
623 624 625 626 627
/*
 * 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
 */
628
void math_error(void __user *ip)
L
Linus Torvalds 已提交
629
{
I
Ingo Molnar 已提交
630
	struct task_struct *task;
L
Linus Torvalds 已提交
631
	siginfo_t info;
632
	unsigned short cwd, swd, err;
L
Linus Torvalds 已提交
633 634 635 636 637 638 639 640 641 642

	/*
	 * Save the info for the exception handler and clear the error.
	 */
	task = current;
	save_init_fpu(task);
	task->thread.trap_no = 16;
	task->thread.error_code = 0;
	info.si_signo = SIGFPE;
	info.si_errno = 0;
643
	info.si_addr = ip;
L
Linus Torvalds 已提交
644 645 646 647 648 649
	/*
	 * (~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,
650
	 * then we have a bad program that isn't synchronizing its FPU usage
L
Linus Torvalds 已提交
651 652 653 654 655
	 * 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);
656

657
	err = swd & ~cwd;
658 659

	if (err & 0x001) {	/* Invalid op */
I
Ingo Molnar 已提交
660 661 662 663 664 665
		/*
		 * swd & 0x240 == 0x040: Stack Underflow
		 * swd & 0x240 == 0x240: Stack Overflow
		 * User must clear the SF bit (0x40) if set
		 */
		info.si_code = FPE_FLTINV;
666
	} else if (err & 0x004) { /* Divide by Zero */
I
Ingo Molnar 已提交
667
		info.si_code = FPE_FLTDIV;
668
	} else if (err & 0x008) { /* Overflow */
I
Ingo Molnar 已提交
669
		info.si_code = FPE_FLTOVF;
670 671 672
	} else if (err & 0x012) { /* Denormal, Underflow */
		info.si_code = FPE_FLTUND;
	} else if (err & 0x020) { /* Precision */
I
Ingo Molnar 已提交
673
		info.si_code = FPE_FLTRES;
674
	} else {
675 676 677 678
		/*
		 * If we're using IRQ 13, or supposedly even some trap 16
		 * implementations, it's possible we get a spurious trap...
		 */
679
		return;		/* Spurious trap, no error */
L
Linus Torvalds 已提交
680 681 682 683
	}
	force_sig_info(SIGFPE, &info, task);
}

684
dotraplinkage void do_coprocessor_error(struct pt_regs *regs, long error_code)
L
Linus Torvalds 已提交
685
{
686
	conditional_sti(regs);
687 688

#ifdef CONFIG_X86_32
L
Linus Torvalds 已提交
689
	ignore_fpu_irq = 1;
690 691 692 693 694 695
#else
	if (!user_mode(regs) &&
	    kernel_math_error(regs, "kernel x87 math error", 16))
		return;
#endif

696
	math_error((void __user *)regs->ip);
L
Linus Torvalds 已提交
697 698
}

699
static void simd_math_error(void __user *ip)
L
Linus Torvalds 已提交
700
{
I
Ingo Molnar 已提交
701 702
	struct task_struct *task;
	siginfo_t info;
703
	unsigned short mxcsr;
L
Linus Torvalds 已提交
704 705 706 707 708 709 710 711 712 713 714

	/*
	 * Save the info for the exception handler and clear the error.
	 */
	task = current;
	save_init_fpu(task);
	task->thread.trap_no = 19;
	task->thread.error_code = 0;
	info.si_signo = SIGFPE;
	info.si_errno = 0;
	info.si_code = __SI_FAULT;
715
	info.si_addr = ip;
L
Linus Torvalds 已提交
716 717 718 719 720 721 722 723
	/*
	 * 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.
	 */
	mxcsr = get_fpu_mxcsr(task);
	switch (~((mxcsr & 0x1f80) >> 7) & (mxcsr & 0x3f)) {
I
Ingo Molnar 已提交
724 725 726 727 728 729 730 731 732 733 734 735 736 737 738 739 740 741 742
	case 0x000:
	default:
		break;
	case 0x001: /* Invalid Op */
		info.si_code = FPE_FLTINV;
		break;
	case 0x002: /* Denormalize */
	case 0x010: /* Underflow */
		info.si_code = FPE_FLTUND;
		break;
	case 0x004: /* Zero Divide */
		info.si_code = FPE_FLTDIV;
		break;
	case 0x008: /* Overflow */
		info.si_code = FPE_FLTOVF;
		break;
	case 0x020: /* Precision */
		info.si_code = FPE_FLTRES;
		break;
L
Linus Torvalds 已提交
743 744 745 746
	}
	force_sig_info(SIGFPE, &info, task);
}

747 748
dotraplinkage void
do_simd_coprocessor_error(struct pt_regs *regs, long error_code)
L
Linus Torvalds 已提交
749
{
750 751
	conditional_sti(regs);

752
#ifdef CONFIG_X86_32
L
Linus Torvalds 已提交
753 754 755
	if (cpu_has_xmm) {
		/* Handle SIMD FPU exceptions on PIII+ processors. */
		ignore_fpu_irq = 1;
756
		simd_math_error((void __user *)regs->ip);
I
Ingo Molnar 已提交
757 758 759 760 761 762
		return;
	}
	/*
	 * Handle strange cache flush from user space exception
	 * in all other cases.  This is undocumented behaviour.
	 */
763
	if (regs->flags & X86_VM_MASK) {
I
Ingo Molnar 已提交
764 765
		handle_vm86_fault((struct kernel_vm86_regs *)regs, error_code);
		return;
L
Linus Torvalds 已提交
766
	}
I
Ingo Molnar 已提交
767 768 769 770
	current->thread.trap_no = 19;
	current->thread.error_code = error_code;
	die_if_kernel("cache flush denied", regs, error_code);
	force_sig(SIGSEGV, current);
771 772 773 774 775 776
#else
	if (!user_mode(regs) &&
			kernel_math_error(regs, "kernel simd math error", 19))
		return;
	simd_math_error((void __user *)regs->ip);
#endif
L
Linus Torvalds 已提交
777 778
}

779 780
dotraplinkage void
do_spurious_interrupt_bug(struct pt_regs *regs, long error_code)
L
Linus Torvalds 已提交
781
{
782
	conditional_sti(regs);
L
Linus Torvalds 已提交
783 784
#if 0
	/* No need to warn about this any longer. */
I
Ingo Molnar 已提交
785
	printk(KERN_INFO "Ignoring P6 Local APIC Spurious Interrupt Bug...\n");
L
Linus Torvalds 已提交
786 787 788
#endif
}

789 790 791 792
asmlinkage void __attribute__((weak)) smp_thermal_interrupt(void)
{
}

793
asmlinkage void __attribute__((weak)) smp_threshold_interrupt(void)
794 795
{
}
L
Linus Torvalds 已提交
796

797 798 799 800 801 802 803 804 805 806 807 808 809 810 811 812 813 814 815 816 817 818
/*
 * __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)
{
	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++;
}

L
Linus Torvalds 已提交
819
/*
I
Ingo Molnar 已提交
820
 * 'math_state_restore()' saves the current math information in the
L
Linus Torvalds 已提交
821 822 823 824 825 826 827 828
 * 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).
 */
829
asmlinkage void math_state_restore(void)
L
Linus Torvalds 已提交
830 831 832 833
{
	struct thread_info *thread = current_thread_info();
	struct task_struct *tsk = thread->task;

834 835 836 837 838 839 840 841 842 843 844 845 846 847 848
	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
849
	clts();				/* Allow maths ops (or we recurse) */
850

851
	__math_state_restore();
L
Linus Torvalds 已提交
852
}
853
EXPORT_SYMBOL_GPL(math_state_restore);
L
Linus Torvalds 已提交
854 855

#ifndef CONFIG_MATH_EMULATION
856
void math_emulate(struct math_emu_info *info)
L
Linus Torvalds 已提交
857
{
I
Ingo Molnar 已提交
858 859 860 861
	printk(KERN_EMERG
		"math-emulation not enabled and no coprocessor found.\n");
	printk(KERN_EMERG "killing %s.\n", current->comm);
	force_sig(SIGFPE, current);
L
Linus Torvalds 已提交
862 863 864 865
	schedule();
}
#endif /* CONFIG_MATH_EMULATION */

866 867
dotraplinkage void __kprobes
do_device_not_available(struct pt_regs *regs, long error_code)
868
{
869
#ifdef CONFIG_X86_32
870
	if (read_cr0() & X86_CR0_EM) {
871 872
		struct math_emu_info info = { };

873
		conditional_sti(regs);
874

875
		info.regs = regs;
876
		math_emulate(&info);
877 878
	} else {
		math_state_restore(); /* interrupts still off */
879
		conditional_sti(regs);
880
	}
881 882 883
#else
	math_state_restore();
#endif
884 885
}

886
#ifdef CONFIG_X86_32
887
dotraplinkage void do_iret_error(struct pt_regs *regs, long error_code)
888 889 890 891 892 893 894
{
	siginfo_t info;
	local_irq_enable();

	info.si_signo = SIGILL;
	info.si_errno = 0;
	info.si_code = ILL_BADSTK;
895
	info.si_addr = NULL;
896 897 898
	if (notify_die(DIE_TRAP, "iret exception",
			regs, error_code, 32, SIGILL) == NOTIFY_STOP)
		return;
899
	do_trap(32, SIGILL, "iret exception", regs, error_code, &info);
900
}
901
#endif
902

L
Linus Torvalds 已提交
903 904
void __init trap_init(void)
{
905 906
	int i;

L
Linus Torvalds 已提交
907
#ifdef CONFIG_EISA
I
Ingo Molnar 已提交
908
	void __iomem *p = early_ioremap(0x0FFFD9, 4);
I
Ingo Molnar 已提交
909 910

	if (readl(p) == 'E' + ('I'<<8) + ('S'<<16) + ('A'<<24))
L
Linus Torvalds 已提交
911
		EISA_bus = 1;
I
Ingo Molnar 已提交
912
	early_iounmap(p, 4);
L
Linus Torvalds 已提交
913 914
#endif

915
	set_intr_gate(0, &divide_error);
916 917 918 919 920 921
	set_intr_gate_ist(1, &debug, DEBUG_STACK);
	set_intr_gate_ist(2, &nmi, NMI_STACK);
	/* int3 can be called from all */
	set_system_intr_gate_ist(3, &int3, DEBUG_STACK);
	/* int4 can be called from all */
	set_system_intr_gate(4, &overflow);
922
	set_intr_gate(5, &bounds);
923
	set_intr_gate(6, &invalid_op);
924
	set_intr_gate(7, &device_not_available);
925
#ifdef CONFIG_X86_32
926
	set_task_gate(8, GDT_ENTRY_DOUBLEFAULT_TSS);
927 928 929
#else
	set_intr_gate_ist(8, &double_fault, DOUBLEFAULT_STACK);
#endif
930
	set_intr_gate(9, &coprocessor_segment_overrun);
931
	set_intr_gate(10, &invalid_TSS);
932
	set_intr_gate(11, &segment_not_present);
933
	set_intr_gate_ist(12, &stack_segment, STACKFAULT_STACK);
934
	set_intr_gate(13, &general_protection);
I
Ingo Molnar 已提交
935
	set_intr_gate(14, &page_fault);
936
	set_intr_gate(15, &spurious_interrupt_bug);
937
	set_intr_gate(16, &coprocessor_error);
938
	set_intr_gate(17, &alignment_check);
L
Linus Torvalds 已提交
939
#ifdef CONFIG_X86_MCE
940
	set_intr_gate_ist(18, &machine_check, MCE_STACK);
L
Linus Torvalds 已提交
941
#endif
942
	set_intr_gate(19, &simd_coprocessor_error);
L
Linus Torvalds 已提交
943

944 945 946 947
	/* Reserve all the builtin and the syscall vector: */
	for (i = 0; i < FIRST_EXTERNAL_VECTOR; i++)
		set_bit(i, used_vectors);

948 949
#ifdef CONFIG_IA32_EMULATION
	set_system_intr_gate(IA32_SYSCALL_VECTOR, ia32_syscall);
950
	set_bit(IA32_SYSCALL_VECTOR, used_vectors);
951 952 953
#endif

#ifdef CONFIG_X86_32
954 955 956 957 958 959
	if (cpu_has_fxsr) {
		printk(KERN_INFO "Enabling fast FPU save and restore... ");
		set_in_cr4(X86_CR4_OSFXSR);
		printk("done.\n");
	}
	if (cpu_has_xmm) {
I
Ingo Molnar 已提交
960 961
		printk(KERN_INFO
			"Enabling unmasked SIMD FPU exception support... ");
962 963 964 965
		set_in_cr4(X86_CR4_OSXMMEXCPT);
		printk("done.\n");
	}

966
	set_system_trap_gate(SYSCALL_VECTOR, &system_call);
967
	set_bit(SYSCALL_VECTOR, used_vectors);
968
#endif
969

L
Linus Torvalds 已提交
970
	/*
I
Ingo Molnar 已提交
971
	 * Should be a barrier for any external CPU state:
L
Linus Torvalds 已提交
972 973 974
	 */
	cpu_init();

975
#ifdef CONFIG_X86_32
I
Ingo Molnar 已提交
976
	x86_quirk_trap_init();
977
#endif
L
Linus Torvalds 已提交
978
}