traps.c 21.9 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 88 89 90 91
static inline void conditional_sti(struct pt_regs *regs)
{
	if (regs->flags & X86_EFLAGS_IF)
		local_irq_enable();
}

92 93 94 95 96 97 98
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
99 100 101 102 103 104
static inline void conditional_cli(struct pt_regs *regs)
{
	if (regs->flags & X86_EFLAGS_IF)
		local_irq_disable();
}

105 106 107 108 109 110 111
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
112
static void __kprobes
113
do_trap(int trapnr, int signr, char *str, struct pt_regs *regs,
I
Ingo Molnar 已提交
114
	long error_code, siginfo_t *info)
L
Linus Torvalds 已提交
115
{
116 117
	struct task_struct *tsk = current;

118
#ifdef CONFIG_X86_32
119
	if (regs->flags & X86_VM_MASK) {
120 121 122 123 124
		/*
		 * 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
125 126 127
			goto vm86_trap;
		goto trap_signal;
	}
128
#endif
L
Linus Torvalds 已提交
129

130
	if (!user_mode(regs))
L
Linus Torvalds 已提交
131 132
		goto kernel_trap;

133
#ifdef CONFIG_X86_32
I
Ingo Molnar 已提交
134
trap_signal:
135
#endif
I
Ingo Molnar 已提交
136 137 138 139 140 141 142 143 144 145 146
	/*
	 * 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;
147

148 149 150 151 152 153 154 155 156 157 158 159
#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
160 161 162 163 164
	if (info)
		force_sig_info(signr, info, tsk);
	else
		force_sig(signr, tsk);
	return;
L
Linus Torvalds 已提交
165

I
Ingo Molnar 已提交
166 167 168 169 170
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
171
	}
I
Ingo Molnar 已提交
172
	return;
L
Linus Torvalds 已提交
173

174
#ifdef CONFIG_X86_32
I
Ingo Molnar 已提交
175 176 177 178 179
vm86_trap:
	if (handle_vm86_trap((struct kernel_vm86_regs *) regs,
						error_code, trapnr))
		goto trap_signal;
	return;
180
#endif
L
Linus Torvalds 已提交
181 182
}

I
Ingo Molnar 已提交
183
#define DO_ERROR(trapnr, signr, str, name)				\
184
dotraplinkage void do_##name(struct pt_regs *regs, long error_code)	\
I
Ingo Molnar 已提交
185 186
{									\
	if (notify_die(DIE_TRAP, str, regs, error_code, trapnr, signr)	\
187
							== NOTIFY_STOP)	\
I
Ingo Molnar 已提交
188
		return;							\
189
	conditional_sti(regs);						\
190
	do_trap(trapnr, signr, str, regs, error_code, NULL);		\
L
Linus Torvalds 已提交
191 192
}

193
#define DO_ERROR_INFO(trapnr, signr, str, name, sicode, siaddr)		\
194
dotraplinkage void do_##name(struct pt_regs *regs, long error_code)	\
I
Ingo Molnar 已提交
195 196 197 198 199 200 201
{									\
	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)	\
202
							== NOTIFY_STOP)	\
I
Ingo Molnar 已提交
203
		return;							\
204
	conditional_sti(regs);						\
205
	do_trap(trapnr, signr, str, regs, error_code, &info);		\
L
Linus Torvalds 已提交
206 207
}

208 209 210 211
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)
212
DO_ERROR(9, SIGFPE, "coprocessor segment overrun", coprocessor_segment_overrun)
213
DO_ERROR(10, SIGSEGV, "invalid TSS", invalid_TSS)
214
DO_ERROR(11, SIGBUS, "segment not present", segment_not_present)
215
#ifdef CONFIG_X86_32
216
DO_ERROR(12, SIGBUS, "stack segment", stack_segment)
217
#endif
218
DO_ERROR_INFO(17, SIGBUS, "alignment check", alignment_check, BUS_ADRALN, 0)
L
Linus Torvalds 已提交
219

220 221 222 223 224 225 226 227 228 229 230 231 232 233 234 235 236 237 238 239 240 241 242
#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;

243 244 245 246
	/*
	 * This is always a kernel trap and never fixable (and thus must
	 * never return).
	 */
247 248 249 250 251
	for (;;)
		die(str, regs, error_code);
}
#endif

252
dotraplinkage void __kprobes
253
do_general_protection(struct pt_regs *regs, long error_code)
L
Linus Torvalds 已提交
254
{
255
	struct task_struct *tsk;
I
Ingo Molnar 已提交
256

257 258
	conditional_sti(regs);

259
#ifdef CONFIG_X86_32
260
	if (regs->flags & X86_VM_MASK)
L
Linus Torvalds 已提交
261
		goto gp_in_vm86;
262
#endif
L
Linus Torvalds 已提交
263

264
	tsk = current;
265
	if (!user_mode(regs))
L
Linus Torvalds 已提交
266 267
		goto gp_in_kernel;

268 269
	tsk->thread.error_code = error_code;
	tsk->thread.trap_no = 13;
I
Ingo Molnar 已提交
270

271 272
	if (show_unhandled_signals && unhandled_signal(tsk, SIGSEGV) &&
			printk_ratelimit()) {
273
		printk(KERN_INFO
274 275 276
			"%s[%d] general protection ip:%lx sp:%lx error:%lx",
			tsk->comm, task_pid_nr(tsk),
			regs->ip, regs->sp, error_code);
277 278 279
		print_vma_addr(" in ", regs->ip);
		printk("\n");
	}
280

281
	force_sig(SIGSEGV, tsk);
L
Linus Torvalds 已提交
282 283
	return;

284
#ifdef CONFIG_X86_32
L
Linus Torvalds 已提交
285 286 287 288
gp_in_vm86:
	local_irq_enable();
	handle_vm86_fault((struct kernel_vm86_regs *) regs, error_code);
	return;
289
#endif
L
Linus Torvalds 已提交
290 291

gp_in_kernel:
292 293 294 295 296 297
	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
298
				error_code, 13, SIGSEGV) == NOTIFY_STOP)
299 300
		return;
	die("general protection fault", regs, error_code);
L
Linus Torvalds 已提交
301 302
}

303
static notrace __kprobes void
I
Ingo Molnar 已提交
304
mem_parity_error(unsigned char reason, struct pt_regs *regs)
L
Linus Torvalds 已提交
305
{
I
Ingo Molnar 已提交
306 307 308 309 310 311
	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
312 313

#if defined(CONFIG_EDAC)
I
Ingo Molnar 已提交
314
	if (edac_handler_set()) {
D
Dave Jiang 已提交
315 316 317 318 319
		edac_atomic_assert_error();
		return;
	}
#endif

320
	if (panic_on_unrecovered_nmi)
I
Ingo Molnar 已提交
321
		panic("NMI: Not continuing");
L
Linus Torvalds 已提交
322

323
	printk(KERN_EMERG "Dazed and confused, but trying to continue\n");
L
Linus Torvalds 已提交
324 325

	/* Clear and disable the memory parity error line. */
326 327
	reason = (reason & 0xf) | 4;
	outb(reason, 0x61);
L
Linus Torvalds 已提交
328 329
}

330
static notrace __kprobes void
I
Ingo Molnar 已提交
331
io_check_error(unsigned char reason, struct pt_regs *regs)
L
Linus Torvalds 已提交
332 333 334
{
	unsigned long i;

335
	printk(KERN_EMERG "NMI: IOCK error (debug interrupt?)\n");
L
Linus Torvalds 已提交
336 337
	show_registers(regs);

338 339 340
	if (panic_on_io_nmi)
		panic("NMI IOCK error: Not continuing");

L
Linus Torvalds 已提交
341 342 343
	/* Re-enable the IOCK line, wait for a few seconds */
	reason = (reason & 0xf) | 8;
	outb(reason, 0x61);
I
Ingo Molnar 已提交
344

L
Linus Torvalds 已提交
345
	i = 2000;
I
Ingo Molnar 已提交
346 347 348
	while (--i)
		udelay(1000);

L
Linus Torvalds 已提交
349 350 351 352
	reason &= ~8;
	outb(reason, 0x61);
}

353
static notrace __kprobes void
I
Ingo Molnar 已提交
354
unknown_nmi_error(unsigned char reason, struct pt_regs *regs)
L
Linus Torvalds 已提交
355
{
356 357
	if (notify_die(DIE_NMIUNKNOWN, "nmi", regs, reason, 2, SIGINT) ==
			NOTIFY_STOP)
J
Jason Wessel 已提交
358
		return;
L
Linus Torvalds 已提交
359
#ifdef CONFIG_MCA
I
Ingo Molnar 已提交
360 361 362 363 364
	/*
	 * Might actually be able to figure out what the guilty party
	 * is:
	 */
	if (MCA_bus) {
L
Linus Torvalds 已提交
365 366 367 368
		mca_handle_nmi();
		return;
	}
#endif
I
Ingo Molnar 已提交
369 370 371 372
	printk(KERN_EMERG
		"Uhhuh. NMI received for unknown reason %02x on CPU %d.\n",
			reason, smp_processor_id());

373
	printk(KERN_EMERG "Do you have a strange power saving mode enabled?\n");
374
	if (panic_on_unrecovered_nmi)
I
Ingo Molnar 已提交
375
		panic("NMI: Not continuing");
376

377
	printk(KERN_EMERG "Dazed and confused, but trying to continue\n");
L
Linus Torvalds 已提交
378 379
}

380
static notrace __kprobes void default_do_nmi(struct pt_regs *regs)
L
Linus Torvalds 已提交
381 382
{
	unsigned char reason = 0;
383 384 385
	int cpu;

	cpu = smp_processor_id();
L
Linus Torvalds 已提交
386

387 388
	/* Only the BSP gets external NMIs from the system. */
	if (!cpu)
L
Linus Torvalds 已提交
389
		reason = get_nmi_reason();
I
Ingo Molnar 已提交
390

L
Linus Torvalds 已提交
391
	if (!(reason & 0xc0)) {
392
		if (notify_die(DIE_NMI_IPI, "nmi_ipi", regs, reason, 2, SIGINT)
393
								== NOTIFY_STOP)
L
Linus Torvalds 已提交
394 395 396 397 398 399
			return;
#ifdef CONFIG_X86_LOCAL_APIC
		/*
		 * Ok, so this is none of the documented NMI sources,
		 * so it must be the NMI watchdog.
		 */
400
		if (nmi_watchdog_tick(regs, reason))
L
Linus Torvalds 已提交
401
			return;
402
		if (!do_nmi_callback(regs, cpu))
403
			unknown_nmi_error(reason, regs);
I
Ingo Molnar 已提交
404 405 406
#else
		unknown_nmi_error(reason, regs);
#endif
407

L
Linus Torvalds 已提交
408 409
		return;
	}
410
	if (notify_die(DIE_NMI, "nmi", regs, reason, 2, SIGINT) == NOTIFY_STOP)
L
Linus Torvalds 已提交
411
		return;
412 413

	/* AK: following checks seem to be broken on modern chipsets. FIXME */
L
Linus Torvalds 已提交
414 415 416 417
	if (reason & 0x80)
		mem_parity_error(reason, regs);
	if (reason & 0x40)
		io_check_error(reason, regs);
418
#ifdef CONFIG_X86_32
L
Linus Torvalds 已提交
419 420
	/*
	 * Reassert NMI in case it became active meanwhile
I
Ingo Molnar 已提交
421
	 * as it's edge-triggered:
L
Linus Torvalds 已提交
422 423
	 */
	reassert_nmi();
424
#endif
L
Linus Torvalds 已提交
425 426
}

427 428
dotraplinkage notrace __kprobes void
do_nmi(struct pt_regs *regs, long error_code)
L
Linus Torvalds 已提交
429 430 431
{
	nmi_enter();

432
	inc_irq_stat(__nmi_count);
L
Linus Torvalds 已提交
433

434 435
	if (!ignore_nmis)
		default_do_nmi(regs);
L
Linus Torvalds 已提交
436 437 438 439

	nmi_exit();
}

440 441 442 443 444 445 446 447 448 449 450 451
void stop_nmi(void)
{
	acpi_nmi_disable();
	ignore_nmis++;
}

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

452
/* May run on IST stack. */
453
dotraplinkage void __kprobes do_int3(struct pt_regs *regs, long error_code)
L
Linus Torvalds 已提交
454
{
455 456 457 458 459
#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 */
460
#ifdef CONFIG_KPROBES
L
Linus Torvalds 已提交
461 462
	if (notify_die(DIE_INT3, "int3", regs, error_code, 3, SIGTRAP)
			== NOTIFY_STOP)
463
		return;
464 465 466 467 468
#else
	if (notify_die(DIE_TRAP, "int3", regs, error_code, 3, SIGTRAP)
			== NOTIFY_STOP)
		return;
#endif
I
Ingo Molnar 已提交
469

470
	preempt_conditional_sti(regs);
471
	do_trap(3, SIGTRAP, "int3", regs, error_code, NULL);
472
	preempt_conditional_cli(regs);
L
Linus Torvalds 已提交
473 474
}

475
#ifdef CONFIG_X86_64
476 477 478 479 480
/*
 * 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
 */
481 482 483 484 485 486 487 488 489
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);
490 491 492 493
	/*
	 * Exception from kernel and interrupts are enabled. Move to
	 * kernel process stack.
	 */
494 495 496 497 498 499 500 501
	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
502 503 504 505 506 507 508 509 510 511
/*
 * 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
512
 *
L
Linus Torvalds 已提交
513 514 515 516 517 518 519 520 521 522
 * 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)
523 524
 *
 * May run on IST stack.
L
Linus Torvalds 已提交
525
 */
526
dotraplinkage void __kprobes do_debug(struct pt_regs *regs, long error_code)
L
Linus Torvalds 已提交
527 528
{
	struct task_struct *tsk = current;
529
	unsigned long dr6;
530
	int si_code;
L
Linus Torvalds 已提交
531

532
	get_debugreg(dr6, 6);
L
Linus Torvalds 已提交
533

534 535 536
	/* Filter out all the reserved bits which are preset to 1 */
	dr6 &= ~DR6_RESERVED;

V
Vegard Nossum 已提交
537
	/* Catch kmemcheck conditions first of all! */
538
	if ((dr6 & DR_STEP) && kmemcheck_trap(regs))
V
Vegard Nossum 已提交
539 540
		return;

541 542
	/* DR6 may or may not be cleared by the CPU */
	set_debugreg(0, 6);
543

P
Peter Zijlstra 已提交
544 545 546 547 548
	/*
	 * The processor cleared BTF, so don't mark that we need it set.
	 */
	clear_tsk_thread_flag(tsk, TIF_BLOCKSTEP);

549 550 551
	/* Store the virtualized DR6 value */
	tsk->thread.debugreg6 = dr6;

552 553
	if (notify_die(DIE_DEBUG, "debug", regs, PTR_ERR(&dr6), error_code,
							SIGTRAP) == NOTIFY_STOP)
L
Linus Torvalds 已提交
554
		return;
555

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

559 560 561 562
	if (regs->flags & X86_VM_MASK) {
		handle_vm86_trap((struct kernel_vm86_regs *) regs,
				error_code, 1);
		return;
L
Linus Torvalds 已提交
563 564 565
	}

	/*
566 567 568 569 570
	 * 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
571
	 */
572 573 574 575
	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
576
	}
577 578 579
	si_code = get_si_code(tsk->thread.debugreg6);
	if (tsk->thread.debugreg6 & (DR_STEP | DR_TRAP_BITS))
		send_sigtrap(tsk, regs, error_code, si_code);
580
	preempt_conditional_cli(regs);
L
Linus Torvalds 已提交
581 582 583 584 585 586 587 588 589

	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
 */
590
void math_error(struct pt_regs *regs, int error_code, int trapnr)
L
Linus Torvalds 已提交
591
{
592
	struct task_struct *task = current;
L
Linus Torvalds 已提交
593
	siginfo_t info;
594
	unsigned short err;
595 596 597 598 599 600 601 602 603 604 605 606 607 608 609
	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
610 611 612 613 614

	/*
	 * Save the info for the exception handler and clear the error.
	 */
	save_init_fpu(task);
615 616
	task->thread.trap_no = trapnr;
	task->thread.error_code = error_code;
L
Linus Torvalds 已提交
617 618
	info.si_signo = SIGFPE;
	info.si_errno = 0;
619 620 621 622 623 624 625 626 627 628 629 630 631 632 633
	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);
634

635 636 637 638 639 640 641 642 643 644 645
		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;
	}
646 647

	if (err & 0x001) {	/* Invalid op */
I
Ingo Molnar 已提交
648 649 650 651 652 653
		/*
		 * swd & 0x240 == 0x040: Stack Underflow
		 * swd & 0x240 == 0x240: Stack Overflow
		 * User must clear the SF bit (0x40) if set
		 */
		info.si_code = FPE_FLTINV;
654
	} else if (err & 0x004) { /* Divide by Zero */
I
Ingo Molnar 已提交
655
		info.si_code = FPE_FLTDIV;
656
	} else if (err & 0x008) { /* Overflow */
I
Ingo Molnar 已提交
657
		info.si_code = FPE_FLTOVF;
658 659 660
	} else if (err & 0x012) { /* Denormal, Underflow */
		info.si_code = FPE_FLTUND;
	} else if (err & 0x020) { /* Precision */
I
Ingo Molnar 已提交
661
		info.si_code = FPE_FLTRES;
662
	} else {
663 664 665 666
		/*
		 * If we're using IRQ 13, or supposedly even some trap 16
		 * implementations, it's possible we get a spurious trap...
		 */
667
		return;		/* Spurious trap, no error */
L
Linus Torvalds 已提交
668 669 670 671
	}
	force_sig_info(SIGFPE, &info, task);
}

672
dotraplinkage void do_coprocessor_error(struct pt_regs *regs, long error_code)
L
Linus Torvalds 已提交
673
{
674
#ifdef CONFIG_X86_32
L
Linus Torvalds 已提交
675
	ignore_fpu_irq = 1;
676 677
#endif

678
	math_error(regs, error_code, 16);
L
Linus Torvalds 已提交
679 680
}

681 682
dotraplinkage void
do_simd_coprocessor_error(struct pt_regs *regs, long error_code)
L
Linus Torvalds 已提交
683
{
684
	math_error(regs, error_code, 19);
L
Linus Torvalds 已提交
685 686
}

687 688
dotraplinkage void
do_spurious_interrupt_bug(struct pt_regs *regs, long error_code)
L
Linus Torvalds 已提交
689
{
690
	conditional_sti(regs);
L
Linus Torvalds 已提交
691 692
#if 0
	/* No need to warn about this any longer. */
I
Ingo Molnar 已提交
693
	printk(KERN_INFO "Ignoring P6 Local APIC Spurious Interrupt Bug...\n");
L
Linus Torvalds 已提交
694 695 696
#endif
}

697
asmlinkage void __attribute__((weak)) smp_thermal_interrupt(void)
L
Linus Torvalds 已提交
698 699
{
}
700

701
asmlinkage void __attribute__((weak)) smp_threshold_interrupt(void)
702 703 704
{
}

705 706 707 708 709
/*
 * __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)
710
{
711 712 713 714 715 716 717 718 719 720 721 722 723 724
	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++;
725
}
L
Linus Torvalds 已提交
726 727

/*
I
Ingo Molnar 已提交
728
 * 'math_state_restore()' saves the current math information in the
L
Linus Torvalds 已提交
729 730 731 732 733 734 735 736
 * 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).
 */
737
asmlinkage void math_state_restore(void)
L
Linus Torvalds 已提交
738 739 740 741
{
	struct thread_info *thread = current_thread_info();
	struct task_struct *tsk = thread->task;

742 743 744 745 746 747 748 749 750 751 752 753 754 755 756
	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
757
	clts();				/* Allow maths ops (or we recurse) */
758

759
	__math_state_restore();
L
Linus Torvalds 已提交
760
}
761
EXPORT_SYMBOL_GPL(math_state_restore);
L
Linus Torvalds 已提交
762 763

#ifndef CONFIG_MATH_EMULATION
764
void math_emulate(struct math_emu_info *info)
L
Linus Torvalds 已提交
765
{
I
Ingo Molnar 已提交
766 767 768 769
	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
770 771 772 773
	schedule();
}
#endif /* CONFIG_MATH_EMULATION */

774
dotraplinkage void __kprobes
775
do_device_not_available(struct pt_regs *regs, long error_code)
776
{
777
#ifdef CONFIG_X86_32
778
	if (read_cr0() & X86_CR0_EM) {
779 780
		struct math_emu_info info = { };

781
		conditional_sti(regs);
782

783
		info.regs = regs;
784
		math_emulate(&info);
785 786 787 788
	} else {
		math_state_restore(); /* interrupts still off */
		conditional_sti(regs);
	}
789 790 791
#else
	math_state_restore();
#endif
792 793
}

794
#ifdef CONFIG_X86_32
795
dotraplinkage void do_iret_error(struct pt_regs *regs, long error_code)
796 797 798 799 800 801 802
{
	siginfo_t info;
	local_irq_enable();

	info.si_signo = SIGILL;
	info.si_errno = 0;
	info.si_code = ILL_BADSTK;
803
	info.si_addr = NULL;
804 805 806
	if (notify_die(DIE_TRAP, "iret exception",
			regs, error_code, 32, SIGILL) == NOTIFY_STOP)
		return;
807
	do_trap(32, SIGILL, "iret exception", regs, error_code, &info);
808
}
809
#endif
810

L
Linus Torvalds 已提交
811 812
void __init trap_init(void)
{
813 814
	int i;

L
Linus Torvalds 已提交
815
#ifdef CONFIG_EISA
I
Ingo Molnar 已提交
816
	void __iomem *p = early_ioremap(0x0FFFD9, 4);
I
Ingo Molnar 已提交
817 818

	if (readl(p) == 'E' + ('I'<<8) + ('S'<<16) + ('A'<<24))
L
Linus Torvalds 已提交
819
		EISA_bus = 1;
I
Ingo Molnar 已提交
820
	early_iounmap(p, 4);
L
Linus Torvalds 已提交
821 822
#endif

823
	set_intr_gate(0, &divide_error);
824 825 826 827 828 829
	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);
830
	set_intr_gate(5, &bounds);
831
	set_intr_gate(6, &invalid_op);
832
	set_intr_gate(7, &device_not_available);
833
#ifdef CONFIG_X86_32
834
	set_task_gate(8, GDT_ENTRY_DOUBLEFAULT_TSS);
835 836 837
#else
	set_intr_gate_ist(8, &double_fault, DOUBLEFAULT_STACK);
#endif
838
	set_intr_gate(9, &coprocessor_segment_overrun);
839
	set_intr_gate(10, &invalid_TSS);
840
	set_intr_gate(11, &segment_not_present);
841
	set_intr_gate_ist(12, &stack_segment, STACKFAULT_STACK);
842
	set_intr_gate(13, &general_protection);
I
Ingo Molnar 已提交
843
	set_intr_gate(14, &page_fault);
844
	set_intr_gate(15, &spurious_interrupt_bug);
845
	set_intr_gate(16, &coprocessor_error);
846
	set_intr_gate(17, &alignment_check);
L
Linus Torvalds 已提交
847
#ifdef CONFIG_X86_MCE
848
	set_intr_gate_ist(18, &machine_check, MCE_STACK);
L
Linus Torvalds 已提交
849
#endif
850
	set_intr_gate(19, &simd_coprocessor_error);
L
Linus Torvalds 已提交
851

852 853 854 855
	/* Reserve all the builtin and the syscall vector: */
	for (i = 0; i < FIRST_EXTERNAL_VECTOR; i++)
		set_bit(i, used_vectors);

856 857
#ifdef CONFIG_IA32_EMULATION
	set_system_intr_gate(IA32_SYSCALL_VECTOR, ia32_syscall);
858
	set_bit(IA32_SYSCALL_VECTOR, used_vectors);
859 860 861
#endif

#ifdef CONFIG_X86_32
862 863 864 865 866 867
	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
868 869
		printk(KERN_INFO
			"Enabling unmasked SIMD FPU exception support... ");
870 871 872 873
		set_in_cr4(X86_CR4_OSXMMEXCPT);
		printk("done.\n");
	}

874
	set_system_trap_gate(SYSCALL_VECTOR, &system_call);
875
	set_bit(SYSCALL_VECTOR, used_vectors);
876
#endif
877

L
Linus Torvalds 已提交
878
	/*
I
Ingo Molnar 已提交
879
	 * Should be a barrier for any external CPU state:
L
Linus Torvalds 已提交
880 881 882
	 */
	cpu_init();

883
	x86_init.irqs.trap_init();
L
Linus Torvalds 已提交
884
}