2 * Copyright (C) 1991, 1992 Linus Torvalds
3 * Copyright (C) 2000, 2001, 2002 Andi Kleen, SuSE Labs
5 * Pentium III FXSR, SSE support
6 * Gareth Hughes <gareth@valinux.com>, May 2000
10 * Handle hardware traps and faults.
13 #define pr_fmt(fmt) KBUILD_MODNAME ": " fmt
15 #include <linux/context_tracking.h>
16 #include <linux/interrupt.h>
17 #include <linux/kallsyms.h>
18 #include <linux/spinlock.h>
19 #include <linux/kprobes.h>
20 #include <linux/uaccess.h>
21 #include <linux/kdebug.h>
22 #include <linux/kgdb.h>
23 #include <linux/kernel.h>
24 #include <linux/module.h>
25 #include <linux/ptrace.h>
26 #include <linux/uprobes.h>
27 #include <linux/string.h>
28 #include <linux/delay.h>
29 #include <linux/errno.h>
30 #include <linux/kexec.h>
31 #include <linux/sched.h>
32 #include <linux/timer.h>
33 #include <linux/init.h>
34 #include <linux/bug.h>
35 #include <linux/nmi.h>
37 #include <linux/smp.h>
41 #include <linux/ioport.h>
42 #include <linux/eisa.h>
45 #if defined(CONFIG_EDAC)
46 #include <linux/edac.h>
49 #include <asm/kmemcheck.h>
50 #include <asm/stacktrace.h>
51 #include <asm/processor.h>
52 #include <asm/debugreg.h>
53 #include <linux/atomic.h>
54 #include <asm/ftrace.h>
55 #include <asm/traps.h>
58 #include <asm/fpu-internal.h>
60 #include <asm/fixmap.h>
61 #include <asm/mach_traps.h>
62 #include <asm/alternative.h>
66 #include <asm/x86_init.h>
67 #include <asm/pgalloc.h>
68 #include <asm/proto.h>
70 /* No need to be aligned, but done to keep all IDTs defined the same way. */
71 gate_desc debug_idt_table
[NR_VECTORS
] __page_aligned_bss
;
73 #include <asm/processor-flags.h>
74 #include <asm/setup.h>
76 asmlinkage
int system_call(void);
79 /* Must be page-aligned because the real IDT is used in a fixmap. */
80 gate_desc idt_table
[NR_VECTORS
] __page_aligned_bss
;
82 DECLARE_BITMAP(used_vectors
, NR_VECTORS
);
83 EXPORT_SYMBOL_GPL(used_vectors
);
85 static inline void conditional_sti(struct pt_regs
*regs
)
87 if (regs
->flags
& X86_EFLAGS_IF
)
91 static inline void preempt_conditional_sti(struct pt_regs
*regs
)
94 if (regs
->flags
& X86_EFLAGS_IF
)
98 static inline void conditional_cli(struct pt_regs
*regs
)
100 if (regs
->flags
& X86_EFLAGS_IF
)
104 static inline void preempt_conditional_cli(struct pt_regs
*regs
)
106 if (regs
->flags
& X86_EFLAGS_IF
)
111 static nokprobe_inline
int
112 do_trap_no_signal(struct task_struct
*tsk
, int trapnr
, char *str
,
113 struct pt_regs
*regs
, long error_code
)
116 if (regs
->flags
& X86_VM_MASK
) {
118 * Traps 0, 1, 3, 4, and 5 should be forwarded to vm86.
119 * On nmi (interrupt 2), do_trap should not be called.
121 if (trapnr
< X86_TRAP_UD
) {
122 if (!handle_vm86_trap((struct kernel_vm86_regs
*) regs
,
129 if (!user_mode(regs
)) {
130 if (!fixup_exception(regs
)) {
131 tsk
->thread
.error_code
= error_code
;
132 tsk
->thread
.trap_nr
= trapnr
;
133 die(str
, regs
, error_code
);
141 static siginfo_t
*fill_trap_info(struct pt_regs
*regs
, int signr
, int trapnr
,
144 unsigned long siaddr
;
149 return SEND_SIG_PRIV
;
153 siaddr
= uprobe_get_trap_addr(regs
);
157 siaddr
= uprobe_get_trap_addr(regs
);
165 info
->si_signo
= signr
;
167 info
->si_code
= sicode
;
168 info
->si_addr
= (void __user
*)siaddr
;
173 do_trap(int trapnr
, int signr
, char *str
, struct pt_regs
*regs
,
174 long error_code
, siginfo_t
*info
)
176 struct task_struct
*tsk
= current
;
179 if (!do_trap_no_signal(tsk
, trapnr
, str
, regs
, error_code
))
182 * We want error_code and trap_nr set for userspace faults and
183 * kernelspace faults which result in die(), but not
184 * kernelspace faults which are fixed up. die() gives the
185 * process no chance to handle the signal and notice the
186 * kernel fault information, so that won't result in polluting
187 * the information about previously queued, but not yet
188 * delivered, faults. See also do_general_protection below.
190 tsk
->thread
.error_code
= error_code
;
191 tsk
->thread
.trap_nr
= trapnr
;
194 if (show_unhandled_signals
&& unhandled_signal(tsk
, signr
) &&
195 printk_ratelimit()) {
196 pr_info("%s[%d] trap %s ip:%lx sp:%lx error:%lx",
197 tsk
->comm
, tsk
->pid
, str
,
198 regs
->ip
, regs
->sp
, error_code
);
199 print_vma_addr(" in ", regs
->ip
);
204 force_sig_info(signr
, info
?: SEND_SIG_PRIV
, tsk
);
206 NOKPROBE_SYMBOL(do_trap
);
208 static void do_error_trap(struct pt_regs
*regs
, long error_code
, char *str
,
209 unsigned long trapnr
, int signr
)
211 enum ctx_state prev_state
= exception_enter();
214 if (notify_die(DIE_TRAP
, str
, regs
, error_code
, trapnr
, signr
) !=
216 conditional_sti(regs
);
217 do_trap(trapnr
, signr
, str
, regs
, error_code
,
218 fill_trap_info(regs
, signr
, trapnr
, &info
));
221 exception_exit(prev_state
);
224 #define DO_ERROR(trapnr, signr, str, name) \
225 dotraplinkage void do_##name(struct pt_regs *regs, long error_code) \
227 do_error_trap(regs, error_code, str, trapnr, signr); \
230 DO_ERROR(X86_TRAP_DE
, SIGFPE
, "divide error", divide_error
)
231 DO_ERROR(X86_TRAP_OF
, SIGSEGV
, "overflow", overflow
)
232 DO_ERROR(X86_TRAP_UD
, SIGILL
, "invalid opcode", invalid_op
)
233 DO_ERROR(X86_TRAP_OLD_MF
, SIGFPE
, "coprocessor segment overrun",coprocessor_segment_overrun
)
234 DO_ERROR(X86_TRAP_TS
, SIGSEGV
, "invalid TSS", invalid_TSS
)
235 DO_ERROR(X86_TRAP_NP
, SIGBUS
, "segment not present", segment_not_present
)
236 DO_ERROR(X86_TRAP_SS
, SIGBUS
, "stack segment", stack_segment
)
237 DO_ERROR(X86_TRAP_AC
, SIGBUS
, "alignment check", alignment_check
)
240 /* Runs on IST stack */
241 dotraplinkage
void do_double_fault(struct pt_regs
*regs
, long error_code
)
243 static const char str
[] = "double fault";
244 struct task_struct
*tsk
= current
;
246 #ifdef CONFIG_X86_ESPFIX64
247 extern unsigned char native_irq_return_iret
[];
250 * If IRET takes a non-IST fault on the espfix64 stack, then we
251 * end up promoting it to a doublefault. In that case, modify
252 * the stack to make it look like we just entered the #GP
253 * handler from user space, similar to bad_iret.
255 if (((long)regs
->sp
>> PGDIR_SHIFT
) == ESPFIX_PGD_ENTRY
&&
256 regs
->cs
== __KERNEL_CS
&&
257 regs
->ip
== (unsigned long)native_irq_return_iret
)
259 struct pt_regs
*normal_regs
= task_pt_regs(current
);
261 /* Fake a #GP(0) from userspace. */
262 memmove(&normal_regs
->ip
, (void *)regs
->sp
, 5*8);
263 normal_regs
->orig_ax
= 0; /* Missing (lost) #GP error code */
264 regs
->ip
= (unsigned long)general_protection
;
265 regs
->sp
= (unsigned long)&normal_regs
->orig_ax
;
271 /* Return not checked because double check cannot be ignored */
272 notify_die(DIE_TRAP
, str
, regs
, error_code
, X86_TRAP_DF
, SIGSEGV
);
274 tsk
->thread
.error_code
= error_code
;
275 tsk
->thread
.trap_nr
= X86_TRAP_DF
;
277 #ifdef CONFIG_DOUBLEFAULT
278 df_debug(regs
, error_code
);
281 * This is always a kernel trap and never fixable (and thus must
285 die(str
, regs
, error_code
);
289 dotraplinkage
void do_bounds(struct pt_regs
*regs
, long error_code
)
291 struct task_struct
*tsk
= current
;
292 struct xsave_struct
*xsave_buf
;
293 enum ctx_state prev_state
;
294 struct bndcsr
*bndcsr
;
297 prev_state
= exception_enter();
298 if (notify_die(DIE_TRAP
, "bounds", regs
, error_code
,
299 X86_TRAP_BR
, SIGSEGV
) == NOTIFY_STOP
)
301 conditional_sti(regs
);
303 if (!user_mode(regs
))
304 die("bounds", regs
, error_code
);
306 if (!cpu_feature_enabled(X86_FEATURE_MPX
)) {
307 /* The exception is not from Intel MPX */
312 * We need to look at BNDSTATUS to resolve this exception.
313 * It is not directly accessible, though, so we need to
314 * do an xsave and then pull it out of the xsave buffer.
316 fpu_save_init(&tsk
->thread
.fpu
);
317 xsave_buf
= &(tsk
->thread
.fpu
.state
->xsave
);
318 bndcsr
= get_xsave_addr(xsave_buf
, XSTATE_BNDCSR
);
323 * The error code field of the BNDSTATUS register communicates status
324 * information of a bound range exception #BR or operation involving
327 switch (bndcsr
->bndstatus
& MPX_BNDSTA_ERROR_CODE
) {
328 case 2: /* Bound directory has invalid entry. */
329 if (mpx_handle_bd_fault(xsave_buf
))
331 break; /* Success, it was handled */
332 case 1: /* Bound violation. */
333 info
= mpx_generate_siginfo(regs
, xsave_buf
);
336 * We failed to decode the MPX instruction. Act as if
337 * the exception was not caused by MPX.
342 * Success, we decoded the instruction and retrieved
343 * an 'info' containing the address being accessed
344 * which caused the exception. This information
345 * allows and application to possibly handle the
346 * #BR exception itself.
348 do_trap(X86_TRAP_BR
, SIGSEGV
, "bounds", regs
, error_code
, info
);
351 case 0: /* No exception caused by Intel MPX operations. */
354 die("bounds", regs
, error_code
);
358 exception_exit(prev_state
);
362 * This path out is for all the cases where we could not
363 * handle the exception in some way (like allocating a
364 * table or telling userspace about it. We will also end
365 * up here if the kernel has MPX turned off at compile
368 do_trap(X86_TRAP_BR
, SIGSEGV
, "bounds", regs
, error_code
, NULL
);
369 exception_exit(prev_state
);
373 do_general_protection(struct pt_regs
*regs
, long error_code
)
375 struct task_struct
*tsk
;
376 enum ctx_state prev_state
;
378 prev_state
= exception_enter();
379 conditional_sti(regs
);
382 if (regs
->flags
& X86_VM_MASK
) {
384 handle_vm86_fault((struct kernel_vm86_regs
*) regs
, error_code
);
390 if (!user_mode(regs
)) {
391 if (fixup_exception(regs
))
394 tsk
->thread
.error_code
= error_code
;
395 tsk
->thread
.trap_nr
= X86_TRAP_GP
;
396 if (notify_die(DIE_GPF
, "general protection fault", regs
, error_code
,
397 X86_TRAP_GP
, SIGSEGV
) != NOTIFY_STOP
)
398 die("general protection fault", regs
, error_code
);
402 tsk
->thread
.error_code
= error_code
;
403 tsk
->thread
.trap_nr
= X86_TRAP_GP
;
405 if (show_unhandled_signals
&& unhandled_signal(tsk
, SIGSEGV
) &&
406 printk_ratelimit()) {
407 pr_info("%s[%d] general protection ip:%lx sp:%lx error:%lx",
408 tsk
->comm
, task_pid_nr(tsk
),
409 regs
->ip
, regs
->sp
, error_code
);
410 print_vma_addr(" in ", regs
->ip
);
414 force_sig_info(SIGSEGV
, SEND_SIG_PRIV
, tsk
);
416 exception_exit(prev_state
);
418 NOKPROBE_SYMBOL(do_general_protection
);
420 /* May run on IST stack. */
421 dotraplinkage
void notrace
do_int3(struct pt_regs
*regs
, long error_code
)
423 enum ctx_state prev_state
;
425 #ifdef CONFIG_DYNAMIC_FTRACE
427 * ftrace must be first, everything else may cause a recursive crash.
428 * See note by declaration of modifying_ftrace_code in ftrace.c
430 if (unlikely(atomic_read(&modifying_ftrace_code
)) &&
431 ftrace_int3_handler(regs
))
434 if (poke_int3_handler(regs
))
437 prev_state
= exception_enter();
438 #ifdef CONFIG_KGDB_LOW_LEVEL_TRAP
439 if (kgdb_ll_trap(DIE_INT3
, "int3", regs
, error_code
, X86_TRAP_BP
,
440 SIGTRAP
) == NOTIFY_STOP
)
442 #endif /* CONFIG_KGDB_LOW_LEVEL_TRAP */
444 #ifdef CONFIG_KPROBES
445 if (kprobe_int3_handler(regs
))
449 if (notify_die(DIE_INT3
, "int3", regs
, error_code
, X86_TRAP_BP
,
450 SIGTRAP
) == NOTIFY_STOP
)
454 * Let others (NMI) know that the debug stack is in use
455 * as we may switch to the interrupt stack.
457 debug_stack_usage_inc();
458 preempt_conditional_sti(regs
);
459 do_trap(X86_TRAP_BP
, SIGTRAP
, "int3", regs
, error_code
, NULL
);
460 preempt_conditional_cli(regs
);
461 debug_stack_usage_dec();
463 exception_exit(prev_state
);
465 NOKPROBE_SYMBOL(do_int3
);
469 * Help handler running on IST stack to switch off the IST stack if the
470 * interrupted code was in user mode. The actual stack switch is done in
473 asmlinkage __visible notrace
struct pt_regs
*sync_regs(struct pt_regs
*eregs
)
475 struct pt_regs
*regs
= task_pt_regs(current
);
479 NOKPROBE_SYMBOL(sync_regs
);
481 struct bad_iret_stack
{
482 void *error_entry_ret
;
486 asmlinkage __visible notrace
487 struct bad_iret_stack
*fixup_bad_iret(struct bad_iret_stack
*s
)
490 * This is called from entry_64.S early in handling a fault
491 * caused by a bad iret to user mode. To handle the fault
492 * correctly, we want move our stack frame to task_pt_regs
493 * and we want to pretend that the exception came from the
496 struct bad_iret_stack
*new_stack
=
497 container_of(task_pt_regs(current
),
498 struct bad_iret_stack
, regs
);
500 /* Copy the IRET target to the new stack. */
501 memmove(&new_stack
->regs
.ip
, (void *)s
->regs
.sp
, 5*8);
503 /* Copy the remainder of the stack from the current stack. */
504 memmove(new_stack
, s
, offsetof(struct bad_iret_stack
, regs
.ip
));
506 BUG_ON(!user_mode_vm(&new_stack
->regs
));
509 NOKPROBE_SYMBOL(fixup_bad_iret
);
513 * Our handling of the processor debug registers is non-trivial.
514 * We do not clear them on entry and exit from the kernel. Therefore
515 * it is possible to get a watchpoint trap here from inside the kernel.
516 * However, the code in ./ptrace.c has ensured that the user can
517 * only set watchpoints on userspace addresses. Therefore the in-kernel
518 * watchpoint trap can only occur in code which is reading/writing
519 * from user space. Such code must not hold kernel locks (since it
520 * can equally take a page fault), therefore it is safe to call
521 * force_sig_info even though that claims and releases locks.
523 * Code in ./signal.c ensures that the debug control register
524 * is restored before we deliver any signal, and therefore that
525 * user code runs with the correct debug control register even though
528 * Being careful here means that we don't have to be as careful in a
529 * lot of more complicated places (task switching can be a bit lazy
530 * about restoring all the debug state, and ptrace doesn't have to
531 * find every occurrence of the TF bit that could be saved away even
534 * May run on IST stack.
536 dotraplinkage
void do_debug(struct pt_regs
*regs
, long error_code
)
538 struct task_struct
*tsk
= current
;
539 enum ctx_state prev_state
;
544 prev_state
= exception_enter();
546 get_debugreg(dr6
, 6);
548 /* Filter out all the reserved bits which are preset to 1 */
549 dr6
&= ~DR6_RESERVED
;
552 * If dr6 has no reason to give us about the origin of this trap,
553 * then it's very likely the result of an icebp/int01 trap.
554 * User wants a sigtrap for that.
556 if (!dr6
&& user_mode(regs
))
559 /* Catch kmemcheck conditions first of all! */
560 if ((dr6
& DR_STEP
) && kmemcheck_trap(regs
))
563 /* DR6 may or may not be cleared by the CPU */
567 * The processor cleared BTF, so don't mark that we need it set.
569 clear_tsk_thread_flag(tsk
, TIF_BLOCKSTEP
);
571 /* Store the virtualized DR6 value */
572 tsk
->thread
.debugreg6
= dr6
;
574 #ifdef CONFIG_KPROBES
575 if (kprobe_debug_handler(regs
))
579 if (notify_die(DIE_DEBUG
, "debug", regs
, (long)&dr6
, error_code
,
580 SIGTRAP
) == NOTIFY_STOP
)
584 * Let others (NMI) know that the debug stack is in use
585 * as we may switch to the interrupt stack.
587 debug_stack_usage_inc();
589 /* It's safe to allow irq's after DR6 has been saved */
590 preempt_conditional_sti(regs
);
592 if (regs
->flags
& X86_VM_MASK
) {
593 handle_vm86_trap((struct kernel_vm86_regs
*) regs
, error_code
,
595 preempt_conditional_cli(regs
);
596 debug_stack_usage_dec();
601 * Single-stepping through system calls: ignore any exceptions in
602 * kernel space, but re-enable TF when returning to user mode.
604 * We already checked v86 mode above, so we can check for kernel mode
605 * by just checking the CPL of CS.
607 if ((dr6
& DR_STEP
) && !user_mode(regs
)) {
608 tsk
->thread
.debugreg6
&= ~DR_STEP
;
609 set_tsk_thread_flag(tsk
, TIF_SINGLESTEP
);
610 regs
->flags
&= ~X86_EFLAGS_TF
;
612 si_code
= get_si_code(tsk
->thread
.debugreg6
);
613 if (tsk
->thread
.debugreg6
& (DR_STEP
| DR_TRAP_BITS
) || user_icebp
)
614 send_sigtrap(tsk
, regs
, error_code
, si_code
);
615 preempt_conditional_cli(regs
);
616 debug_stack_usage_dec();
619 exception_exit(prev_state
);
621 NOKPROBE_SYMBOL(do_debug
);
624 * Note that we play around with the 'TS' bit in an attempt to get
625 * the correct behaviour even in the presence of the asynchronous
628 static void math_error(struct pt_regs
*regs
, int error_code
, int trapnr
)
630 struct task_struct
*task
= current
;
633 char *str
= (trapnr
== X86_TRAP_MF
) ? "fpu exception" :
636 if (notify_die(DIE_TRAP
, str
, regs
, error_code
, trapnr
, SIGFPE
) == NOTIFY_STOP
)
638 conditional_sti(regs
);
640 if (!user_mode_vm(regs
))
642 if (!fixup_exception(regs
)) {
643 task
->thread
.error_code
= error_code
;
644 task
->thread
.trap_nr
= trapnr
;
645 die(str
, regs
, error_code
);
651 * Save the info for the exception handler and clear the error.
654 task
->thread
.trap_nr
= trapnr
;
655 task
->thread
.error_code
= error_code
;
656 info
.si_signo
= SIGFPE
;
658 info
.si_addr
= (void __user
*)uprobe_get_trap_addr(regs
);
659 if (trapnr
== X86_TRAP_MF
) {
660 unsigned short cwd
, swd
;
662 * (~cwd & swd) will mask out exceptions that are not set to unmasked
663 * status. 0x3f is the exception bits in these regs, 0x200 is the
664 * C1 reg you need in case of a stack fault, 0x040 is the stack
665 * fault bit. We should only be taking one exception at a time,
666 * so if this combination doesn't produce any single exception,
667 * then we have a bad program that isn't synchronizing its FPU usage
668 * and it will suffer the consequences since we won't be able to
669 * fully reproduce the context of the exception
671 cwd
= get_fpu_cwd(task
);
672 swd
= get_fpu_swd(task
);
677 * The SIMD FPU exceptions are handled a little differently, as there
678 * is only a single status/control register. Thus, to determine which
679 * unmasked exception was caught we must mask the exception mask bits
680 * at 0x1f80, and then use these to mask the exception bits at 0x3f.
682 unsigned short mxcsr
= get_fpu_mxcsr(task
);
683 err
= ~(mxcsr
>> 7) & mxcsr
;
686 if (err
& 0x001) { /* Invalid op */
688 * swd & 0x240 == 0x040: Stack Underflow
689 * swd & 0x240 == 0x240: Stack Overflow
690 * User must clear the SF bit (0x40) if set
692 info
.si_code
= FPE_FLTINV
;
693 } else if (err
& 0x004) { /* Divide by Zero */
694 info
.si_code
= FPE_FLTDIV
;
695 } else if (err
& 0x008) { /* Overflow */
696 info
.si_code
= FPE_FLTOVF
;
697 } else if (err
& 0x012) { /* Denormal, Underflow */
698 info
.si_code
= FPE_FLTUND
;
699 } else if (err
& 0x020) { /* Precision */
700 info
.si_code
= FPE_FLTRES
;
703 * If we're using IRQ 13, or supposedly even some trap
704 * X86_TRAP_MF implementations, it's possible
705 * we get a spurious trap, which is not an error.
709 force_sig_info(SIGFPE
, &info
, task
);
712 dotraplinkage
void do_coprocessor_error(struct pt_regs
*regs
, long error_code
)
714 enum ctx_state prev_state
;
716 prev_state
= exception_enter();
717 math_error(regs
, error_code
, X86_TRAP_MF
);
718 exception_exit(prev_state
);
722 do_simd_coprocessor_error(struct pt_regs
*regs
, long error_code
)
724 enum ctx_state prev_state
;
726 prev_state
= exception_enter();
727 math_error(regs
, error_code
, X86_TRAP_XF
);
728 exception_exit(prev_state
);
732 do_spurious_interrupt_bug(struct pt_regs
*regs
, long error_code
)
734 conditional_sti(regs
);
736 /* No need to warn about this any longer. */
737 pr_info("Ignoring P6 Local APIC Spurious Interrupt Bug...\n");
741 asmlinkage __visible
void __attribute__((weak
)) smp_thermal_interrupt(void)
745 asmlinkage __visible
void __attribute__((weak
)) smp_threshold_interrupt(void)
750 * 'math_state_restore()' saves the current math information in the
751 * old math state array, and gets the new ones from the current task
753 * Careful.. There are problems with IBM-designed IRQ13 behaviour.
754 * Don't touch unless you *really* know how it works.
756 * Must be called with kernel preemption disabled (eg with local
757 * local interrupts as in the case of do_device_not_available).
759 void math_state_restore(void)
761 struct task_struct
*tsk
= current
;
763 if (!tsk_used_math(tsk
)) {
766 * does a slab alloc which can sleep
772 do_group_exit(SIGKILL
);
778 __thread_fpu_begin(tsk
);
781 * Paranoid restore. send a SIGSEGV if we fail to restore the state.
783 if (unlikely(restore_fpu_checking(tsk
))) {
785 force_sig_info(SIGSEGV
, SEND_SIG_PRIV
, tsk
);
789 tsk
->thread
.fpu_counter
++;
791 EXPORT_SYMBOL_GPL(math_state_restore
);
794 do_device_not_available(struct pt_regs
*regs
, long error_code
)
796 enum ctx_state prev_state
;
798 prev_state
= exception_enter();
799 BUG_ON(use_eager_fpu());
801 #ifdef CONFIG_MATH_EMULATION
802 if (read_cr0() & X86_CR0_EM
) {
803 struct math_emu_info info
= { };
805 conditional_sti(regs
);
809 exception_exit(prev_state
);
813 math_state_restore(); /* interrupts still off */
815 conditional_sti(regs
);
817 exception_exit(prev_state
);
819 NOKPROBE_SYMBOL(do_device_not_available
);
822 dotraplinkage
void do_iret_error(struct pt_regs
*regs
, long error_code
)
825 enum ctx_state prev_state
;
827 prev_state
= exception_enter();
830 info
.si_signo
= SIGILL
;
832 info
.si_code
= ILL_BADSTK
;
834 if (notify_die(DIE_TRAP
, "iret exception", regs
, error_code
,
835 X86_TRAP_IRET
, SIGILL
) != NOTIFY_STOP
) {
836 do_trap(X86_TRAP_IRET
, SIGILL
, "iret exception", regs
, error_code
,
839 exception_exit(prev_state
);
843 /* Set of traps needed for early debugging. */
844 void __init
early_trap_init(void)
846 set_intr_gate_ist(X86_TRAP_DB
, &debug
, DEBUG_STACK
);
847 /* int3 can be called from all */
848 set_system_intr_gate_ist(X86_TRAP_BP
, &int3
, DEBUG_STACK
);
850 set_intr_gate(X86_TRAP_PF
, page_fault
);
852 load_idt(&idt_descr
);
855 void __init
early_trap_pf_init(void)
858 set_intr_gate(X86_TRAP_PF
, page_fault
);
862 void __init
trap_init(void)
867 void __iomem
*p
= early_ioremap(0x0FFFD9, 4);
869 if (readl(p
) == 'E' + ('I'<<8) + ('S'<<16) + ('A'<<24))
874 set_intr_gate(X86_TRAP_DE
, divide_error
);
875 set_intr_gate_ist(X86_TRAP_NMI
, &nmi
, NMI_STACK
);
876 /* int4 can be called from all */
877 set_system_intr_gate(X86_TRAP_OF
, &overflow
);
878 set_intr_gate(X86_TRAP_BR
, bounds
);
879 set_intr_gate(X86_TRAP_UD
, invalid_op
);
880 set_intr_gate(X86_TRAP_NM
, device_not_available
);
882 set_task_gate(X86_TRAP_DF
, GDT_ENTRY_DOUBLEFAULT_TSS
);
884 set_intr_gate_ist(X86_TRAP_DF
, &double_fault
, DOUBLEFAULT_STACK
);
886 set_intr_gate(X86_TRAP_OLD_MF
, coprocessor_segment_overrun
);
887 set_intr_gate(X86_TRAP_TS
, invalid_TSS
);
888 set_intr_gate(X86_TRAP_NP
, segment_not_present
);
889 set_intr_gate(X86_TRAP_SS
, stack_segment
);
890 set_intr_gate(X86_TRAP_GP
, general_protection
);
891 set_intr_gate(X86_TRAP_SPURIOUS
, spurious_interrupt_bug
);
892 set_intr_gate(X86_TRAP_MF
, coprocessor_error
);
893 set_intr_gate(X86_TRAP_AC
, alignment_check
);
894 #ifdef CONFIG_X86_MCE
895 set_intr_gate_ist(X86_TRAP_MC
, &machine_check
, MCE_STACK
);
897 set_intr_gate(X86_TRAP_XF
, simd_coprocessor_error
);
899 /* Reserve all the builtin and the syscall vector: */
900 for (i
= 0; i
< FIRST_EXTERNAL_VECTOR
; i
++)
901 set_bit(i
, used_vectors
);
903 #ifdef CONFIG_IA32_EMULATION
904 set_system_intr_gate(IA32_SYSCALL_VECTOR
, ia32_syscall
);
905 set_bit(IA32_SYSCALL_VECTOR
, used_vectors
);
909 set_system_trap_gate(SYSCALL_VECTOR
, &system_call
);
910 set_bit(SYSCALL_VECTOR
, used_vectors
);
914 * Set the IDT descriptor to a fixed read-only location, so that the
915 * "sidt" instruction will not leak the location of the kernel, and
916 * to defend the IDT against arbitrary memory write vulnerabilities.
917 * It will be reloaded in cpu_init() */
918 __set_fixmap(FIX_RO_IDT
, __pa_symbol(idt_table
), PAGE_KERNEL_RO
);
919 idt_descr
.address
= fix_to_virt(FIX_RO_IDT
);
922 * Should be a barrier for any external CPU state:
926 x86_init
.irqs
.trap_init();
929 memcpy(&debug_idt_table
, &idt_table
, IDT_ENTRIES
* 16);
930 set_nmi_gate(X86_TRAP_DB
, &debug
);
931 set_nmi_gate(X86_TRAP_BP
, &int3
);