]> git.proxmox.com Git - mirror_ubuntu-bionic-kernel.git/blob - arch/x86/kernel/entry_32.S
Merge branch 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/cooloney...
[mirror_ubuntu-bionic-kernel.git] / arch / x86 / kernel / entry_32.S
1 /*
2 *
3 * Copyright (C) 1991, 1992 Linus Torvalds
4 */
5
6 /*
7 * entry.S contains the system-call and fault low-level handling routines.
8 * This also contains the timer-interrupt handler, as well as all interrupts
9 * and faults that can result in a task-switch.
10 *
11 * NOTE: This code handles signal-recognition, which happens every time
12 * after a timer-interrupt and after each system call.
13 *
14 * I changed all the .align's to 4 (16 byte alignment), as that's faster
15 * on a 486.
16 *
17 * Stack layout in 'syscall_exit':
18 * ptrace needs to have all regs on the stack.
19 * if the order here is changed, it needs to be
20 * updated in fork.c:copy_process, signal.c:do_signal,
21 * ptrace.c and ptrace.h
22 *
23 * 0(%esp) - %ebx
24 * 4(%esp) - %ecx
25 * 8(%esp) - %edx
26 * C(%esp) - %esi
27 * 10(%esp) - %edi
28 * 14(%esp) - %ebp
29 * 18(%esp) - %eax
30 * 1C(%esp) - %ds
31 * 20(%esp) - %es
32 * 24(%esp) - %fs
33 * 28(%esp) - %gs saved iff !CONFIG_X86_32_LAZY_GS
34 * 2C(%esp) - orig_eax
35 * 30(%esp) - %eip
36 * 34(%esp) - %cs
37 * 38(%esp) - %eflags
38 * 3C(%esp) - %oldesp
39 * 40(%esp) - %oldss
40 *
41 * "current" is in register %ebx during any slow entries.
42 */
43
44 #include <linux/linkage.h>
45 #include <asm/thread_info.h>
46 #include <asm/irqflags.h>
47 #include <asm/errno.h>
48 #include <asm/segment.h>
49 #include <asm/smp.h>
50 #include <asm/page_types.h>
51 #include <asm/desc.h>
52 #include <asm/percpu.h>
53 #include <asm/dwarf2.h>
54 #include <asm/processor-flags.h>
55 #include <asm/ftrace.h>
56 #include <asm/irq_vectors.h>
57
58 /* Avoid __ASSEMBLER__'ifying <linux/audit.h> just for this. */
59 #include <linux/elf-em.h>
60 #define AUDIT_ARCH_I386 (EM_386|__AUDIT_ARCH_LE)
61 #define __AUDIT_ARCH_LE 0x40000000
62
63 #ifndef CONFIG_AUDITSYSCALL
64 #define sysenter_audit syscall_trace_entry
65 #define sysexit_audit syscall_exit_work
66 #endif
67
68 /*
69 * We use macros for low-level operations which need to be overridden
70 * for paravirtualization. The following will never clobber any registers:
71 * INTERRUPT_RETURN (aka. "iret")
72 * GET_CR0_INTO_EAX (aka. "movl %cr0, %eax")
73 * ENABLE_INTERRUPTS_SYSEXIT (aka "sti; sysexit").
74 *
75 * For DISABLE_INTERRUPTS/ENABLE_INTERRUPTS (aka "cli"/"sti"), you must
76 * specify what registers can be overwritten (CLBR_NONE, CLBR_EAX/EDX/ECX/ANY).
77 * Allowing a register to be clobbered can shrink the paravirt replacement
78 * enough to patch inline, increasing performance.
79 */
80
81 #define nr_syscalls ((syscall_table_size)/4)
82
83 #ifdef CONFIG_PREEMPT
84 #define preempt_stop(clobbers) DISABLE_INTERRUPTS(clobbers); TRACE_IRQS_OFF
85 #else
86 #define preempt_stop(clobbers)
87 #define resume_kernel restore_nocheck
88 #endif
89
90 .macro TRACE_IRQS_IRET
91 #ifdef CONFIG_TRACE_IRQFLAGS
92 testl $X86_EFLAGS_IF,PT_EFLAGS(%esp) # interrupts off?
93 jz 1f
94 TRACE_IRQS_ON
95 1:
96 #endif
97 .endm
98
99 #ifdef CONFIG_VM86
100 #define resume_userspace_sig check_userspace
101 #else
102 #define resume_userspace_sig resume_userspace
103 #endif
104
105 /*
106 * User gs save/restore
107 *
108 * %gs is used for userland TLS and kernel only uses it for stack
109 * canary which is required to be at %gs:20 by gcc. Read the comment
110 * at the top of stackprotector.h for more info.
111 *
112 * Local labels 98 and 99 are used.
113 */
114 #ifdef CONFIG_X86_32_LAZY_GS
115
116 /* unfortunately push/pop can't be no-op */
117 .macro PUSH_GS
118 pushl $0
119 CFI_ADJUST_CFA_OFFSET 4
120 .endm
121 .macro POP_GS pop=0
122 addl $(4 + \pop), %esp
123 CFI_ADJUST_CFA_OFFSET -(4 + \pop)
124 .endm
125 .macro POP_GS_EX
126 .endm
127
128 /* all the rest are no-op */
129 .macro PTGS_TO_GS
130 .endm
131 .macro PTGS_TO_GS_EX
132 .endm
133 .macro GS_TO_REG reg
134 .endm
135 .macro REG_TO_PTGS reg
136 .endm
137 .macro SET_KERNEL_GS reg
138 .endm
139
140 #else /* CONFIG_X86_32_LAZY_GS */
141
142 .macro PUSH_GS
143 pushl %gs
144 CFI_ADJUST_CFA_OFFSET 4
145 /*CFI_REL_OFFSET gs, 0*/
146 .endm
147
148 .macro POP_GS pop=0
149 98: popl %gs
150 CFI_ADJUST_CFA_OFFSET -4
151 /*CFI_RESTORE gs*/
152 .if \pop <> 0
153 add $\pop, %esp
154 CFI_ADJUST_CFA_OFFSET -\pop
155 .endif
156 .endm
157 .macro POP_GS_EX
158 .pushsection .fixup, "ax"
159 99: movl $0, (%esp)
160 jmp 98b
161 .section __ex_table, "a"
162 .align 4
163 .long 98b, 99b
164 .popsection
165 .endm
166
167 .macro PTGS_TO_GS
168 98: mov PT_GS(%esp), %gs
169 .endm
170 .macro PTGS_TO_GS_EX
171 .pushsection .fixup, "ax"
172 99: movl $0, PT_GS(%esp)
173 jmp 98b
174 .section __ex_table, "a"
175 .align 4
176 .long 98b, 99b
177 .popsection
178 .endm
179
180 .macro GS_TO_REG reg
181 movl %gs, \reg
182 /*CFI_REGISTER gs, \reg*/
183 .endm
184 .macro REG_TO_PTGS reg
185 movl \reg, PT_GS(%esp)
186 /*CFI_REL_OFFSET gs, PT_GS*/
187 .endm
188 .macro SET_KERNEL_GS reg
189 movl $(__KERNEL_STACK_CANARY), \reg
190 movl \reg, %gs
191 .endm
192
193 #endif /* CONFIG_X86_32_LAZY_GS */
194
195 .macro SAVE_ALL
196 cld
197 PUSH_GS
198 pushl %fs
199 CFI_ADJUST_CFA_OFFSET 4
200 /*CFI_REL_OFFSET fs, 0;*/
201 pushl %es
202 CFI_ADJUST_CFA_OFFSET 4
203 /*CFI_REL_OFFSET es, 0;*/
204 pushl %ds
205 CFI_ADJUST_CFA_OFFSET 4
206 /*CFI_REL_OFFSET ds, 0;*/
207 pushl %eax
208 CFI_ADJUST_CFA_OFFSET 4
209 CFI_REL_OFFSET eax, 0
210 pushl %ebp
211 CFI_ADJUST_CFA_OFFSET 4
212 CFI_REL_OFFSET ebp, 0
213 pushl %edi
214 CFI_ADJUST_CFA_OFFSET 4
215 CFI_REL_OFFSET edi, 0
216 pushl %esi
217 CFI_ADJUST_CFA_OFFSET 4
218 CFI_REL_OFFSET esi, 0
219 pushl %edx
220 CFI_ADJUST_CFA_OFFSET 4
221 CFI_REL_OFFSET edx, 0
222 pushl %ecx
223 CFI_ADJUST_CFA_OFFSET 4
224 CFI_REL_OFFSET ecx, 0
225 pushl %ebx
226 CFI_ADJUST_CFA_OFFSET 4
227 CFI_REL_OFFSET ebx, 0
228 movl $(__USER_DS), %edx
229 movl %edx, %ds
230 movl %edx, %es
231 movl $(__KERNEL_PERCPU), %edx
232 movl %edx, %fs
233 SET_KERNEL_GS %edx
234 .endm
235
236 .macro RESTORE_INT_REGS
237 popl %ebx
238 CFI_ADJUST_CFA_OFFSET -4
239 CFI_RESTORE ebx
240 popl %ecx
241 CFI_ADJUST_CFA_OFFSET -4
242 CFI_RESTORE ecx
243 popl %edx
244 CFI_ADJUST_CFA_OFFSET -4
245 CFI_RESTORE edx
246 popl %esi
247 CFI_ADJUST_CFA_OFFSET -4
248 CFI_RESTORE esi
249 popl %edi
250 CFI_ADJUST_CFA_OFFSET -4
251 CFI_RESTORE edi
252 popl %ebp
253 CFI_ADJUST_CFA_OFFSET -4
254 CFI_RESTORE ebp
255 popl %eax
256 CFI_ADJUST_CFA_OFFSET -4
257 CFI_RESTORE eax
258 .endm
259
260 .macro RESTORE_REGS pop=0
261 RESTORE_INT_REGS
262 1: popl %ds
263 CFI_ADJUST_CFA_OFFSET -4
264 /*CFI_RESTORE ds;*/
265 2: popl %es
266 CFI_ADJUST_CFA_OFFSET -4
267 /*CFI_RESTORE es;*/
268 3: popl %fs
269 CFI_ADJUST_CFA_OFFSET -4
270 /*CFI_RESTORE fs;*/
271 POP_GS \pop
272 .pushsection .fixup, "ax"
273 4: movl $0, (%esp)
274 jmp 1b
275 5: movl $0, (%esp)
276 jmp 2b
277 6: movl $0, (%esp)
278 jmp 3b
279 .section __ex_table, "a"
280 .align 4
281 .long 1b, 4b
282 .long 2b, 5b
283 .long 3b, 6b
284 .popsection
285 POP_GS_EX
286 .endm
287
288 .macro RING0_INT_FRAME
289 CFI_STARTPROC simple
290 CFI_SIGNAL_FRAME
291 CFI_DEF_CFA esp, 3*4
292 /*CFI_OFFSET cs, -2*4;*/
293 CFI_OFFSET eip, -3*4
294 .endm
295
296 .macro RING0_EC_FRAME
297 CFI_STARTPROC simple
298 CFI_SIGNAL_FRAME
299 CFI_DEF_CFA esp, 4*4
300 /*CFI_OFFSET cs, -2*4;*/
301 CFI_OFFSET eip, -3*4
302 .endm
303
304 .macro RING0_PTREGS_FRAME
305 CFI_STARTPROC simple
306 CFI_SIGNAL_FRAME
307 CFI_DEF_CFA esp, PT_OLDESP-PT_EBX
308 /*CFI_OFFSET cs, PT_CS-PT_OLDESP;*/
309 CFI_OFFSET eip, PT_EIP-PT_OLDESP
310 /*CFI_OFFSET es, PT_ES-PT_OLDESP;*/
311 /*CFI_OFFSET ds, PT_DS-PT_OLDESP;*/
312 CFI_OFFSET eax, PT_EAX-PT_OLDESP
313 CFI_OFFSET ebp, PT_EBP-PT_OLDESP
314 CFI_OFFSET edi, PT_EDI-PT_OLDESP
315 CFI_OFFSET esi, PT_ESI-PT_OLDESP
316 CFI_OFFSET edx, PT_EDX-PT_OLDESP
317 CFI_OFFSET ecx, PT_ECX-PT_OLDESP
318 CFI_OFFSET ebx, PT_EBX-PT_OLDESP
319 .endm
320
321 ENTRY(ret_from_fork)
322 CFI_STARTPROC
323 pushl %eax
324 CFI_ADJUST_CFA_OFFSET 4
325 call schedule_tail
326 GET_THREAD_INFO(%ebp)
327 popl %eax
328 CFI_ADJUST_CFA_OFFSET -4
329 pushl $0x0202 # Reset kernel eflags
330 CFI_ADJUST_CFA_OFFSET 4
331 popfl
332 CFI_ADJUST_CFA_OFFSET -4
333 jmp syscall_exit
334 CFI_ENDPROC
335 END(ret_from_fork)
336
337 /*
338 * Return to user mode is not as complex as all this looks,
339 * but we want the default path for a system call return to
340 * go as quickly as possible which is why some of this is
341 * less clear than it otherwise should be.
342 */
343
344 # userspace resumption stub bypassing syscall exit tracing
345 ALIGN
346 RING0_PTREGS_FRAME
347 ret_from_exception:
348 preempt_stop(CLBR_ANY)
349 ret_from_intr:
350 GET_THREAD_INFO(%ebp)
351 check_userspace:
352 movl PT_EFLAGS(%esp), %eax # mix EFLAGS and CS
353 movb PT_CS(%esp), %al
354 andl $(X86_EFLAGS_VM | SEGMENT_RPL_MASK), %eax
355 cmpl $USER_RPL, %eax
356 jb resume_kernel # not returning to v8086 or userspace
357
358 ENTRY(resume_userspace)
359 LOCKDEP_SYS_EXIT
360 DISABLE_INTERRUPTS(CLBR_ANY) # make sure we don't miss an interrupt
361 # setting need_resched or sigpending
362 # between sampling and the iret
363 TRACE_IRQS_OFF
364 movl TI_flags(%ebp), %ecx
365 andl $_TIF_WORK_MASK, %ecx # is there any work to be done on
366 # int/exception return?
367 jne work_pending
368 jmp restore_all
369 END(ret_from_exception)
370
371 #ifdef CONFIG_PREEMPT
372 ENTRY(resume_kernel)
373 DISABLE_INTERRUPTS(CLBR_ANY)
374 cmpl $0,TI_preempt_count(%ebp) # non-zero preempt_count ?
375 jnz restore_nocheck
376 need_resched:
377 movl TI_flags(%ebp), %ecx # need_resched set ?
378 testb $_TIF_NEED_RESCHED, %cl
379 jz restore_all
380 testl $X86_EFLAGS_IF,PT_EFLAGS(%esp) # interrupts off (exception path) ?
381 jz restore_all
382 call preempt_schedule_irq
383 jmp need_resched
384 END(resume_kernel)
385 #endif
386 CFI_ENDPROC
387
388 /* SYSENTER_RETURN points to after the "sysenter" instruction in
389 the vsyscall page. See vsyscall-sysentry.S, which defines the symbol. */
390
391 # sysenter call handler stub
392 ENTRY(ia32_sysenter_target)
393 CFI_STARTPROC simple
394 CFI_SIGNAL_FRAME
395 CFI_DEF_CFA esp, 0
396 CFI_REGISTER esp, ebp
397 movl TSS_sysenter_sp0(%esp),%esp
398 sysenter_past_esp:
399 /*
400 * Interrupts are disabled here, but we can't trace it until
401 * enough kernel state to call TRACE_IRQS_OFF can be called - but
402 * we immediately enable interrupts at that point anyway.
403 */
404 pushl $(__USER_DS)
405 CFI_ADJUST_CFA_OFFSET 4
406 /*CFI_REL_OFFSET ss, 0*/
407 pushl %ebp
408 CFI_ADJUST_CFA_OFFSET 4
409 CFI_REL_OFFSET esp, 0
410 pushfl
411 orl $X86_EFLAGS_IF, (%esp)
412 CFI_ADJUST_CFA_OFFSET 4
413 pushl $(__USER_CS)
414 CFI_ADJUST_CFA_OFFSET 4
415 /*CFI_REL_OFFSET cs, 0*/
416 /*
417 * Push current_thread_info()->sysenter_return to the stack.
418 * A tiny bit of offset fixup is necessary - 4*4 means the 4 words
419 * pushed above; +8 corresponds to copy_thread's esp0 setting.
420 */
421 pushl (TI_sysenter_return-THREAD_SIZE+8+4*4)(%esp)
422 CFI_ADJUST_CFA_OFFSET 4
423 CFI_REL_OFFSET eip, 0
424
425 pushl %eax
426 CFI_ADJUST_CFA_OFFSET 4
427 SAVE_ALL
428 ENABLE_INTERRUPTS(CLBR_NONE)
429
430 /*
431 * Load the potential sixth argument from user stack.
432 * Careful about security.
433 */
434 cmpl $__PAGE_OFFSET-3,%ebp
435 jae syscall_fault
436 1: movl (%ebp),%ebp
437 movl %ebp,PT_EBP(%esp)
438 .section __ex_table,"a"
439 .align 4
440 .long 1b,syscall_fault
441 .previous
442
443 GET_THREAD_INFO(%ebp)
444
445 /* Note, _TIF_SECCOMP is bit number 8, and so it needs testw and not testb */
446 testw $_TIF_WORK_SYSCALL_ENTRY,TI_flags(%ebp)
447 jnz sysenter_audit
448 sysenter_do_call:
449 cmpl $(nr_syscalls), %eax
450 jae syscall_badsys
451 call *sys_call_table(,%eax,4)
452 movl %eax,PT_EAX(%esp)
453 LOCKDEP_SYS_EXIT
454 DISABLE_INTERRUPTS(CLBR_ANY)
455 TRACE_IRQS_OFF
456 movl TI_flags(%ebp), %ecx
457 testw $_TIF_ALLWORK_MASK, %cx
458 jne sysexit_audit
459 sysenter_exit:
460 /* if something modifies registers it must also disable sysexit */
461 movl PT_EIP(%esp), %edx
462 movl PT_OLDESP(%esp), %ecx
463 xorl %ebp,%ebp
464 TRACE_IRQS_ON
465 1: mov PT_FS(%esp), %fs
466 PTGS_TO_GS
467 ENABLE_INTERRUPTS_SYSEXIT
468
469 #ifdef CONFIG_AUDITSYSCALL
470 sysenter_audit:
471 testw $(_TIF_WORK_SYSCALL_ENTRY & ~_TIF_SYSCALL_AUDIT),TI_flags(%ebp)
472 jnz syscall_trace_entry
473 addl $4,%esp
474 CFI_ADJUST_CFA_OFFSET -4
475 /* %esi already in 8(%esp) 6th arg: 4th syscall arg */
476 /* %edx already in 4(%esp) 5th arg: 3rd syscall arg */
477 /* %ecx already in 0(%esp) 4th arg: 2nd syscall arg */
478 movl %ebx,%ecx /* 3rd arg: 1st syscall arg */
479 movl %eax,%edx /* 2nd arg: syscall number */
480 movl $AUDIT_ARCH_I386,%eax /* 1st arg: audit arch */
481 call audit_syscall_entry
482 pushl %ebx
483 CFI_ADJUST_CFA_OFFSET 4
484 movl PT_EAX(%esp),%eax /* reload syscall number */
485 jmp sysenter_do_call
486
487 sysexit_audit:
488 testw $(_TIF_ALLWORK_MASK & ~_TIF_SYSCALL_AUDIT), %cx
489 jne syscall_exit_work
490 TRACE_IRQS_ON
491 ENABLE_INTERRUPTS(CLBR_ANY)
492 movl %eax,%edx /* second arg, syscall return value */
493 cmpl $0,%eax /* is it < 0? */
494 setl %al /* 1 if so, 0 if not */
495 movzbl %al,%eax /* zero-extend that */
496 inc %eax /* first arg, 0->1(AUDITSC_SUCCESS), 1->2(AUDITSC_FAILURE) */
497 call audit_syscall_exit
498 DISABLE_INTERRUPTS(CLBR_ANY)
499 TRACE_IRQS_OFF
500 movl TI_flags(%ebp), %ecx
501 testw $(_TIF_ALLWORK_MASK & ~_TIF_SYSCALL_AUDIT), %cx
502 jne syscall_exit_work
503 movl PT_EAX(%esp),%eax /* reload syscall return value */
504 jmp sysenter_exit
505 #endif
506
507 CFI_ENDPROC
508 .pushsection .fixup,"ax"
509 2: movl $0,PT_FS(%esp)
510 jmp 1b
511 .section __ex_table,"a"
512 .align 4
513 .long 1b,2b
514 .popsection
515 PTGS_TO_GS_EX
516 ENDPROC(ia32_sysenter_target)
517
518 # system call handler stub
519 ENTRY(system_call)
520 RING0_INT_FRAME # can't unwind into user space anyway
521 pushl %eax # save orig_eax
522 CFI_ADJUST_CFA_OFFSET 4
523 SAVE_ALL
524 GET_THREAD_INFO(%ebp)
525 # system call tracing in operation / emulation
526 /* Note, _TIF_SECCOMP is bit number 8, and so it needs testw and not testb */
527 testw $_TIF_WORK_SYSCALL_ENTRY,TI_flags(%ebp)
528 jnz syscall_trace_entry
529 cmpl $(nr_syscalls), %eax
530 jae syscall_badsys
531 syscall_call:
532 call *sys_call_table(,%eax,4)
533 movl %eax,PT_EAX(%esp) # store the return value
534 syscall_exit:
535 LOCKDEP_SYS_EXIT
536 DISABLE_INTERRUPTS(CLBR_ANY) # make sure we don't miss an interrupt
537 # setting need_resched or sigpending
538 # between sampling and the iret
539 TRACE_IRQS_OFF
540 movl TI_flags(%ebp), %ecx
541 testw $_TIF_ALLWORK_MASK, %cx # current->work
542 jne syscall_exit_work
543
544 restore_all:
545 movl PT_EFLAGS(%esp), %eax # mix EFLAGS, SS and CS
546 # Warning: PT_OLDSS(%esp) contains the wrong/random values if we
547 # are returning to the kernel.
548 # See comments in process.c:copy_thread() for details.
549 movb PT_OLDSS(%esp), %ah
550 movb PT_CS(%esp), %al
551 andl $(X86_EFLAGS_VM | (SEGMENT_TI_MASK << 8) | SEGMENT_RPL_MASK), %eax
552 cmpl $((SEGMENT_LDT << 8) | USER_RPL), %eax
553 CFI_REMEMBER_STATE
554 je ldt_ss # returning to user-space with LDT SS
555 restore_nocheck:
556 TRACE_IRQS_IRET
557 restore_nocheck_notrace:
558 RESTORE_REGS 4 # skip orig_eax/error_code
559 CFI_ADJUST_CFA_OFFSET -4
560 irq_return:
561 INTERRUPT_RETURN
562 .section .fixup,"ax"
563 ENTRY(iret_exc)
564 pushl $0 # no error code
565 pushl $do_iret_error
566 jmp error_code
567 .previous
568 .section __ex_table,"a"
569 .align 4
570 .long irq_return,iret_exc
571 .previous
572
573 CFI_RESTORE_STATE
574 ldt_ss:
575 larl PT_OLDSS(%esp), %eax
576 jnz restore_nocheck
577 testl $0x00400000, %eax # returning to 32bit stack?
578 jnz restore_nocheck # allright, normal return
579
580 #ifdef CONFIG_PARAVIRT
581 /*
582 * The kernel can't run on a non-flat stack if paravirt mode
583 * is active. Rather than try to fixup the high bits of
584 * ESP, bypass this code entirely. This may break DOSemu
585 * and/or Wine support in a paravirt VM, although the option
586 * is still available to implement the setting of the high
587 * 16-bits in the INTERRUPT_RETURN paravirt-op.
588 */
589 cmpl $0, pv_info+PARAVIRT_enabled
590 jne restore_nocheck
591 #endif
592
593 /* If returning to userspace with 16bit stack,
594 * try to fix the higher word of ESP, as the CPU
595 * won't restore it.
596 * This is an "official" bug of all the x86-compatible
597 * CPUs, which we can try to work around to make
598 * dosemu and wine happy. */
599 movl PT_OLDESP(%esp), %eax
600 movl %esp, %edx
601 call patch_espfix_desc
602 pushl $__ESPFIX_SS
603 CFI_ADJUST_CFA_OFFSET 4
604 pushl %eax
605 CFI_ADJUST_CFA_OFFSET 4
606 DISABLE_INTERRUPTS(CLBR_EAX)
607 TRACE_IRQS_OFF
608 lss (%esp), %esp
609 CFI_ADJUST_CFA_OFFSET -8
610 jmp restore_nocheck
611 CFI_ENDPROC
612 ENDPROC(system_call)
613
614 # perform work that needs to be done immediately before resumption
615 ALIGN
616 RING0_PTREGS_FRAME # can't unwind into user space anyway
617 work_pending:
618 testb $_TIF_NEED_RESCHED, %cl
619 jz work_notifysig
620 work_resched:
621 call schedule
622 LOCKDEP_SYS_EXIT
623 DISABLE_INTERRUPTS(CLBR_ANY) # make sure we don't miss an interrupt
624 # setting need_resched or sigpending
625 # between sampling and the iret
626 TRACE_IRQS_OFF
627 movl TI_flags(%ebp), %ecx
628 andl $_TIF_WORK_MASK, %ecx # is there any work to be done other
629 # than syscall tracing?
630 jz restore_all
631 testb $_TIF_NEED_RESCHED, %cl
632 jnz work_resched
633
634 work_notifysig: # deal with pending signals and
635 # notify-resume requests
636 #ifdef CONFIG_VM86
637 testl $X86_EFLAGS_VM, PT_EFLAGS(%esp)
638 movl %esp, %eax
639 jne work_notifysig_v86 # returning to kernel-space or
640 # vm86-space
641 xorl %edx, %edx
642 call do_notify_resume
643 jmp resume_userspace_sig
644
645 ALIGN
646 work_notifysig_v86:
647 pushl %ecx # save ti_flags for do_notify_resume
648 CFI_ADJUST_CFA_OFFSET 4
649 call save_v86_state # %eax contains pt_regs pointer
650 popl %ecx
651 CFI_ADJUST_CFA_OFFSET -4
652 movl %eax, %esp
653 #else
654 movl %esp, %eax
655 #endif
656 xorl %edx, %edx
657 call do_notify_resume
658 jmp resume_userspace_sig
659 END(work_pending)
660
661 # perform syscall exit tracing
662 ALIGN
663 syscall_trace_entry:
664 movl $-ENOSYS,PT_EAX(%esp)
665 movl %esp, %eax
666 call syscall_trace_enter
667 /* What it returned is what we'll actually use. */
668 cmpl $(nr_syscalls), %eax
669 jnae syscall_call
670 jmp syscall_exit
671 END(syscall_trace_entry)
672
673 # perform syscall exit tracing
674 ALIGN
675 syscall_exit_work:
676 testb $_TIF_WORK_SYSCALL_EXIT, %cl
677 jz work_pending
678 TRACE_IRQS_ON
679 ENABLE_INTERRUPTS(CLBR_ANY) # could let syscall_trace_leave() call
680 # schedule() instead
681 movl %esp, %eax
682 call syscall_trace_leave
683 jmp resume_userspace
684 END(syscall_exit_work)
685 CFI_ENDPROC
686
687 RING0_INT_FRAME # can't unwind into user space anyway
688 syscall_fault:
689 GET_THREAD_INFO(%ebp)
690 movl $-EFAULT,PT_EAX(%esp)
691 jmp resume_userspace
692 END(syscall_fault)
693
694 syscall_badsys:
695 movl $-ENOSYS,PT_EAX(%esp)
696 jmp resume_userspace
697 END(syscall_badsys)
698 CFI_ENDPROC
699
700 /*
701 * System calls that need a pt_regs pointer.
702 */
703 #define PTREGSCALL(name) \
704 ALIGN; \
705 ptregs_##name: \
706 leal 4(%esp),%eax; \
707 jmp sys_##name;
708
709 PTREGSCALL(iopl)
710 PTREGSCALL(fork)
711 PTREGSCALL(clone)
712 PTREGSCALL(vfork)
713 PTREGSCALL(execve)
714 PTREGSCALL(sigaltstack)
715 PTREGSCALL(sigreturn)
716 PTREGSCALL(rt_sigreturn)
717 PTREGSCALL(vm86)
718 PTREGSCALL(vm86old)
719
720 .macro FIXUP_ESPFIX_STACK
721 /* since we are on a wrong stack, we cant make it a C code :( */
722 PER_CPU(gdt_page, %ebx)
723 GET_DESC_BASE(GDT_ENTRY_ESPFIX_SS, %ebx, %eax, %ax, %al, %ah)
724 addl %esp, %eax
725 pushl $__KERNEL_DS
726 CFI_ADJUST_CFA_OFFSET 4
727 pushl %eax
728 CFI_ADJUST_CFA_OFFSET 4
729 lss (%esp), %esp
730 CFI_ADJUST_CFA_OFFSET -8
731 .endm
732 .macro UNWIND_ESPFIX_STACK
733 movl %ss, %eax
734 /* see if on espfix stack */
735 cmpw $__ESPFIX_SS, %ax
736 jne 27f
737 movl $__KERNEL_DS, %eax
738 movl %eax, %ds
739 movl %eax, %es
740 /* switch to normal stack */
741 FIXUP_ESPFIX_STACK
742 27:
743 .endm
744
745 /*
746 * Build the entry stubs and pointer table with some assembler magic.
747 * We pack 7 stubs into a single 32-byte chunk, which will fit in a
748 * single cache line on all modern x86 implementations.
749 */
750 .section .init.rodata,"a"
751 ENTRY(interrupt)
752 .text
753 .p2align 5
754 .p2align CONFIG_X86_L1_CACHE_SHIFT
755 ENTRY(irq_entries_start)
756 RING0_INT_FRAME
757 vector=FIRST_EXTERNAL_VECTOR
758 .rept (NR_VECTORS-FIRST_EXTERNAL_VECTOR+6)/7
759 .balign 32
760 .rept 7
761 .if vector < NR_VECTORS
762 .if vector <> FIRST_EXTERNAL_VECTOR
763 CFI_ADJUST_CFA_OFFSET -4
764 .endif
765 1: pushl $(~vector+0x80) /* Note: always in signed byte range */
766 CFI_ADJUST_CFA_OFFSET 4
767 .if ((vector-FIRST_EXTERNAL_VECTOR)%7) <> 6
768 jmp 2f
769 .endif
770 .previous
771 .long 1b
772 .text
773 vector=vector+1
774 .endif
775 .endr
776 2: jmp common_interrupt
777 .endr
778 END(irq_entries_start)
779
780 .previous
781 END(interrupt)
782 .previous
783
784 /*
785 * the CPU automatically disables interrupts when executing an IRQ vector,
786 * so IRQ-flags tracing has to follow that:
787 */
788 .p2align CONFIG_X86_L1_CACHE_SHIFT
789 common_interrupt:
790 addl $-0x80,(%esp) /* Adjust vector into the [-256,-1] range */
791 SAVE_ALL
792 TRACE_IRQS_OFF
793 movl %esp,%eax
794 call do_IRQ
795 jmp ret_from_intr
796 ENDPROC(common_interrupt)
797 CFI_ENDPROC
798
799 #define BUILD_INTERRUPT3(name, nr, fn) \
800 ENTRY(name) \
801 RING0_INT_FRAME; \
802 pushl $~(nr); \
803 CFI_ADJUST_CFA_OFFSET 4; \
804 SAVE_ALL; \
805 TRACE_IRQS_OFF \
806 movl %esp,%eax; \
807 call fn; \
808 jmp ret_from_intr; \
809 CFI_ENDPROC; \
810 ENDPROC(name)
811
812 #define BUILD_INTERRUPT(name, nr) BUILD_INTERRUPT3(name, nr, smp_##name)
813
814 /* The include is where all of the SMP etc. interrupts come from */
815 #include <asm/entry_arch.h>
816
817 ENTRY(coprocessor_error)
818 RING0_INT_FRAME
819 pushl $0
820 CFI_ADJUST_CFA_OFFSET 4
821 pushl $do_coprocessor_error
822 CFI_ADJUST_CFA_OFFSET 4
823 jmp error_code
824 CFI_ENDPROC
825 END(coprocessor_error)
826
827 ENTRY(simd_coprocessor_error)
828 RING0_INT_FRAME
829 pushl $0
830 CFI_ADJUST_CFA_OFFSET 4
831 pushl $do_simd_coprocessor_error
832 CFI_ADJUST_CFA_OFFSET 4
833 jmp error_code
834 CFI_ENDPROC
835 END(simd_coprocessor_error)
836
837 ENTRY(device_not_available)
838 RING0_INT_FRAME
839 pushl $-1 # mark this as an int
840 CFI_ADJUST_CFA_OFFSET 4
841 pushl $do_device_not_available
842 CFI_ADJUST_CFA_OFFSET 4
843 jmp error_code
844 CFI_ENDPROC
845 END(device_not_available)
846
847 #ifdef CONFIG_PARAVIRT
848 ENTRY(native_iret)
849 iret
850 .section __ex_table,"a"
851 .align 4
852 .long native_iret, iret_exc
853 .previous
854 END(native_iret)
855
856 ENTRY(native_irq_enable_sysexit)
857 sti
858 sysexit
859 END(native_irq_enable_sysexit)
860 #endif
861
862 ENTRY(overflow)
863 RING0_INT_FRAME
864 pushl $0
865 CFI_ADJUST_CFA_OFFSET 4
866 pushl $do_overflow
867 CFI_ADJUST_CFA_OFFSET 4
868 jmp error_code
869 CFI_ENDPROC
870 END(overflow)
871
872 ENTRY(bounds)
873 RING0_INT_FRAME
874 pushl $0
875 CFI_ADJUST_CFA_OFFSET 4
876 pushl $do_bounds
877 CFI_ADJUST_CFA_OFFSET 4
878 jmp error_code
879 CFI_ENDPROC
880 END(bounds)
881
882 ENTRY(invalid_op)
883 RING0_INT_FRAME
884 pushl $0
885 CFI_ADJUST_CFA_OFFSET 4
886 pushl $do_invalid_op
887 CFI_ADJUST_CFA_OFFSET 4
888 jmp error_code
889 CFI_ENDPROC
890 END(invalid_op)
891
892 ENTRY(coprocessor_segment_overrun)
893 RING0_INT_FRAME
894 pushl $0
895 CFI_ADJUST_CFA_OFFSET 4
896 pushl $do_coprocessor_segment_overrun
897 CFI_ADJUST_CFA_OFFSET 4
898 jmp error_code
899 CFI_ENDPROC
900 END(coprocessor_segment_overrun)
901
902 ENTRY(invalid_TSS)
903 RING0_EC_FRAME
904 pushl $do_invalid_TSS
905 CFI_ADJUST_CFA_OFFSET 4
906 jmp error_code
907 CFI_ENDPROC
908 END(invalid_TSS)
909
910 ENTRY(segment_not_present)
911 RING0_EC_FRAME
912 pushl $do_segment_not_present
913 CFI_ADJUST_CFA_OFFSET 4
914 jmp error_code
915 CFI_ENDPROC
916 END(segment_not_present)
917
918 ENTRY(stack_segment)
919 RING0_EC_FRAME
920 pushl $do_stack_segment
921 CFI_ADJUST_CFA_OFFSET 4
922 jmp error_code
923 CFI_ENDPROC
924 END(stack_segment)
925
926 ENTRY(alignment_check)
927 RING0_EC_FRAME
928 pushl $do_alignment_check
929 CFI_ADJUST_CFA_OFFSET 4
930 jmp error_code
931 CFI_ENDPROC
932 END(alignment_check)
933
934 ENTRY(divide_error)
935 RING0_INT_FRAME
936 pushl $0 # no error code
937 CFI_ADJUST_CFA_OFFSET 4
938 pushl $do_divide_error
939 CFI_ADJUST_CFA_OFFSET 4
940 jmp error_code
941 CFI_ENDPROC
942 END(divide_error)
943
944 #ifdef CONFIG_X86_MCE
945 ENTRY(machine_check)
946 RING0_INT_FRAME
947 pushl $0
948 CFI_ADJUST_CFA_OFFSET 4
949 pushl machine_check_vector
950 CFI_ADJUST_CFA_OFFSET 4
951 jmp error_code
952 CFI_ENDPROC
953 END(machine_check)
954 #endif
955
956 ENTRY(spurious_interrupt_bug)
957 RING0_INT_FRAME
958 pushl $0
959 CFI_ADJUST_CFA_OFFSET 4
960 pushl $do_spurious_interrupt_bug
961 CFI_ADJUST_CFA_OFFSET 4
962 jmp error_code
963 CFI_ENDPROC
964 END(spurious_interrupt_bug)
965
966 ENTRY(kernel_thread_helper)
967 pushl $0 # fake return address for unwinder
968 CFI_STARTPROC
969 movl %edx,%eax
970 push %edx
971 CFI_ADJUST_CFA_OFFSET 4
972 call *%ebx
973 push %eax
974 CFI_ADJUST_CFA_OFFSET 4
975 call do_exit
976 ud2 # padding for call trace
977 CFI_ENDPROC
978 ENDPROC(kernel_thread_helper)
979
980 #ifdef CONFIG_XEN
981 /* Xen doesn't set %esp to be precisely what the normal sysenter
982 entrypoint expects, so fix it up before using the normal path. */
983 ENTRY(xen_sysenter_target)
984 RING0_INT_FRAME
985 addl $5*4, %esp /* remove xen-provided frame */
986 CFI_ADJUST_CFA_OFFSET -5*4
987 jmp sysenter_past_esp
988 CFI_ENDPROC
989
990 ENTRY(xen_hypervisor_callback)
991 CFI_STARTPROC
992 pushl $0
993 CFI_ADJUST_CFA_OFFSET 4
994 SAVE_ALL
995 TRACE_IRQS_OFF
996
997 /* Check to see if we got the event in the critical
998 region in xen_iret_direct, after we've reenabled
999 events and checked for pending events. This simulates
1000 iret instruction's behaviour where it delivers a
1001 pending interrupt when enabling interrupts. */
1002 movl PT_EIP(%esp),%eax
1003 cmpl $xen_iret_start_crit,%eax
1004 jb 1f
1005 cmpl $xen_iret_end_crit,%eax
1006 jae 1f
1007
1008 jmp xen_iret_crit_fixup
1009
1010 ENTRY(xen_do_upcall)
1011 1: mov %esp, %eax
1012 call xen_evtchn_do_upcall
1013 jmp ret_from_intr
1014 CFI_ENDPROC
1015 ENDPROC(xen_hypervisor_callback)
1016
1017 # Hypervisor uses this for application faults while it executes.
1018 # We get here for two reasons:
1019 # 1. Fault while reloading DS, ES, FS or GS
1020 # 2. Fault while executing IRET
1021 # Category 1 we fix up by reattempting the load, and zeroing the segment
1022 # register if the load fails.
1023 # Category 2 we fix up by jumping to do_iret_error. We cannot use the
1024 # normal Linux return path in this case because if we use the IRET hypercall
1025 # to pop the stack frame we end up in an infinite loop of failsafe callbacks.
1026 # We distinguish between categories by maintaining a status value in EAX.
1027 ENTRY(xen_failsafe_callback)
1028 CFI_STARTPROC
1029 pushl %eax
1030 CFI_ADJUST_CFA_OFFSET 4
1031 movl $1,%eax
1032 1: mov 4(%esp),%ds
1033 2: mov 8(%esp),%es
1034 3: mov 12(%esp),%fs
1035 4: mov 16(%esp),%gs
1036 testl %eax,%eax
1037 popl %eax
1038 CFI_ADJUST_CFA_OFFSET -4
1039 lea 16(%esp),%esp
1040 CFI_ADJUST_CFA_OFFSET -16
1041 jz 5f
1042 addl $16,%esp
1043 jmp iret_exc # EAX != 0 => Category 2 (Bad IRET)
1044 5: pushl $0 # EAX == 0 => Category 1 (Bad segment)
1045 CFI_ADJUST_CFA_OFFSET 4
1046 SAVE_ALL
1047 jmp ret_from_exception
1048 CFI_ENDPROC
1049
1050 .section .fixup,"ax"
1051 6: xorl %eax,%eax
1052 movl %eax,4(%esp)
1053 jmp 1b
1054 7: xorl %eax,%eax
1055 movl %eax,8(%esp)
1056 jmp 2b
1057 8: xorl %eax,%eax
1058 movl %eax,12(%esp)
1059 jmp 3b
1060 9: xorl %eax,%eax
1061 movl %eax,16(%esp)
1062 jmp 4b
1063 .previous
1064 .section __ex_table,"a"
1065 .align 4
1066 .long 1b,6b
1067 .long 2b,7b
1068 .long 3b,8b
1069 .long 4b,9b
1070 .previous
1071 ENDPROC(xen_failsafe_callback)
1072
1073 #endif /* CONFIG_XEN */
1074
1075 #ifdef CONFIG_FUNCTION_TRACER
1076 #ifdef CONFIG_DYNAMIC_FTRACE
1077
1078 ENTRY(mcount)
1079 ret
1080 END(mcount)
1081
1082 ENTRY(ftrace_caller)
1083 cmpl $0, function_trace_stop
1084 jne ftrace_stub
1085
1086 pushl %eax
1087 pushl %ecx
1088 pushl %edx
1089 movl 0xc(%esp), %eax
1090 movl 0x4(%ebp), %edx
1091 subl $MCOUNT_INSN_SIZE, %eax
1092
1093 .globl ftrace_call
1094 ftrace_call:
1095 call ftrace_stub
1096
1097 popl %edx
1098 popl %ecx
1099 popl %eax
1100 #ifdef CONFIG_FUNCTION_GRAPH_TRACER
1101 .globl ftrace_graph_call
1102 ftrace_graph_call:
1103 jmp ftrace_stub
1104 #endif
1105
1106 .globl ftrace_stub
1107 ftrace_stub:
1108 ret
1109 END(ftrace_caller)
1110
1111 #else /* ! CONFIG_DYNAMIC_FTRACE */
1112
1113 ENTRY(mcount)
1114 cmpl $0, function_trace_stop
1115 jne ftrace_stub
1116
1117 cmpl $ftrace_stub, ftrace_trace_function
1118 jnz trace
1119 #ifdef CONFIG_FUNCTION_GRAPH_TRACER
1120 cmpl $ftrace_stub, ftrace_graph_return
1121 jnz ftrace_graph_caller
1122
1123 cmpl $ftrace_graph_entry_stub, ftrace_graph_entry
1124 jnz ftrace_graph_caller
1125 #endif
1126 .globl ftrace_stub
1127 ftrace_stub:
1128 ret
1129
1130 /* taken from glibc */
1131 trace:
1132 pushl %eax
1133 pushl %ecx
1134 pushl %edx
1135 movl 0xc(%esp), %eax
1136 movl 0x4(%ebp), %edx
1137 subl $MCOUNT_INSN_SIZE, %eax
1138
1139 call *ftrace_trace_function
1140
1141 popl %edx
1142 popl %ecx
1143 popl %eax
1144 jmp ftrace_stub
1145 END(mcount)
1146 #endif /* CONFIG_DYNAMIC_FTRACE */
1147 #endif /* CONFIG_FUNCTION_TRACER */
1148
1149 #ifdef CONFIG_FUNCTION_GRAPH_TRACER
1150 ENTRY(ftrace_graph_caller)
1151 cmpl $0, function_trace_stop
1152 jne ftrace_stub
1153
1154 pushl %eax
1155 pushl %ecx
1156 pushl %edx
1157 movl 0xc(%esp), %edx
1158 lea 0x4(%ebp), %eax
1159 subl $MCOUNT_INSN_SIZE, %edx
1160 call prepare_ftrace_return
1161 popl %edx
1162 popl %ecx
1163 popl %eax
1164 ret
1165 END(ftrace_graph_caller)
1166
1167 .globl return_to_handler
1168 return_to_handler:
1169 pushl $0
1170 pushl %eax
1171 pushl %ecx
1172 pushl %edx
1173 call ftrace_return_to_handler
1174 movl %eax, 0xc(%esp)
1175 popl %edx
1176 popl %ecx
1177 popl %eax
1178 ret
1179 #endif
1180
1181 .section .rodata,"a"
1182 #include "syscall_table_32.S"
1183
1184 syscall_table_size=(.-sys_call_table)
1185
1186 /*
1187 * Some functions should be protected against kprobes
1188 */
1189 .pushsection .kprobes.text, "ax"
1190
1191 ENTRY(page_fault)
1192 RING0_EC_FRAME
1193 pushl $do_page_fault
1194 CFI_ADJUST_CFA_OFFSET 4
1195 ALIGN
1196 error_code:
1197 /* the function address is in %gs's slot on the stack */
1198 pushl %fs
1199 CFI_ADJUST_CFA_OFFSET 4
1200 /*CFI_REL_OFFSET fs, 0*/
1201 pushl %es
1202 CFI_ADJUST_CFA_OFFSET 4
1203 /*CFI_REL_OFFSET es, 0*/
1204 pushl %ds
1205 CFI_ADJUST_CFA_OFFSET 4
1206 /*CFI_REL_OFFSET ds, 0*/
1207 pushl %eax
1208 CFI_ADJUST_CFA_OFFSET 4
1209 CFI_REL_OFFSET eax, 0
1210 pushl %ebp
1211 CFI_ADJUST_CFA_OFFSET 4
1212 CFI_REL_OFFSET ebp, 0
1213 pushl %edi
1214 CFI_ADJUST_CFA_OFFSET 4
1215 CFI_REL_OFFSET edi, 0
1216 pushl %esi
1217 CFI_ADJUST_CFA_OFFSET 4
1218 CFI_REL_OFFSET esi, 0
1219 pushl %edx
1220 CFI_ADJUST_CFA_OFFSET 4
1221 CFI_REL_OFFSET edx, 0
1222 pushl %ecx
1223 CFI_ADJUST_CFA_OFFSET 4
1224 CFI_REL_OFFSET ecx, 0
1225 pushl %ebx
1226 CFI_ADJUST_CFA_OFFSET 4
1227 CFI_REL_OFFSET ebx, 0
1228 cld
1229 movl $(__KERNEL_PERCPU), %ecx
1230 movl %ecx, %fs
1231 UNWIND_ESPFIX_STACK
1232 GS_TO_REG %ecx
1233 movl PT_GS(%esp), %edi # get the function address
1234 movl PT_ORIG_EAX(%esp), %edx # get the error code
1235 movl $-1, PT_ORIG_EAX(%esp) # no syscall to restart
1236 REG_TO_PTGS %ecx
1237 SET_KERNEL_GS %ecx
1238 movl $(__USER_DS), %ecx
1239 movl %ecx, %ds
1240 movl %ecx, %es
1241 TRACE_IRQS_OFF
1242 movl %esp,%eax # pt_regs pointer
1243 call *%edi
1244 jmp ret_from_exception
1245 CFI_ENDPROC
1246 END(page_fault)
1247
1248 /*
1249 * Debug traps and NMI can happen at the one SYSENTER instruction
1250 * that sets up the real kernel stack. Check here, since we can't
1251 * allow the wrong stack to be used.
1252 *
1253 * "TSS_sysenter_sp0+12" is because the NMI/debug handler will have
1254 * already pushed 3 words if it hits on the sysenter instruction:
1255 * eflags, cs and eip.
1256 *
1257 * We just load the right stack, and push the three (known) values
1258 * by hand onto the new stack - while updating the return eip past
1259 * the instruction that would have done it for sysenter.
1260 */
1261 .macro FIX_STACK offset ok label
1262 cmpw $__KERNEL_CS, 4(%esp)
1263 jne \ok
1264 \label:
1265 movl TSS_sysenter_sp0 + \offset(%esp), %esp
1266 CFI_DEF_CFA esp, 0
1267 CFI_UNDEFINED eip
1268 pushfl
1269 CFI_ADJUST_CFA_OFFSET 4
1270 pushl $__KERNEL_CS
1271 CFI_ADJUST_CFA_OFFSET 4
1272 pushl $sysenter_past_esp
1273 CFI_ADJUST_CFA_OFFSET 4
1274 CFI_REL_OFFSET eip, 0
1275 .endm
1276
1277 ENTRY(debug)
1278 RING0_INT_FRAME
1279 cmpl $ia32_sysenter_target,(%esp)
1280 jne debug_stack_correct
1281 FIX_STACK 12, debug_stack_correct, debug_esp_fix_insn
1282 debug_stack_correct:
1283 pushl $-1 # mark this as an int
1284 CFI_ADJUST_CFA_OFFSET 4
1285 SAVE_ALL
1286 TRACE_IRQS_OFF
1287 xorl %edx,%edx # error code 0
1288 movl %esp,%eax # pt_regs pointer
1289 call do_debug
1290 jmp ret_from_exception
1291 CFI_ENDPROC
1292 END(debug)
1293
1294 /*
1295 * NMI is doubly nasty. It can happen _while_ we're handling
1296 * a debug fault, and the debug fault hasn't yet been able to
1297 * clear up the stack. So we first check whether we got an
1298 * NMI on the sysenter entry path, but after that we need to
1299 * check whether we got an NMI on the debug path where the debug
1300 * fault happened on the sysenter path.
1301 */
1302 ENTRY(nmi)
1303 RING0_INT_FRAME
1304 pushl %eax
1305 CFI_ADJUST_CFA_OFFSET 4
1306 movl %ss, %eax
1307 cmpw $__ESPFIX_SS, %ax
1308 popl %eax
1309 CFI_ADJUST_CFA_OFFSET -4
1310 je nmi_espfix_stack
1311 cmpl $ia32_sysenter_target,(%esp)
1312 je nmi_stack_fixup
1313 pushl %eax
1314 CFI_ADJUST_CFA_OFFSET 4
1315 movl %esp,%eax
1316 /* Do not access memory above the end of our stack page,
1317 * it might not exist.
1318 */
1319 andl $(THREAD_SIZE-1),%eax
1320 cmpl $(THREAD_SIZE-20),%eax
1321 popl %eax
1322 CFI_ADJUST_CFA_OFFSET -4
1323 jae nmi_stack_correct
1324 cmpl $ia32_sysenter_target,12(%esp)
1325 je nmi_debug_stack_check
1326 nmi_stack_correct:
1327 /* We have a RING0_INT_FRAME here */
1328 pushl %eax
1329 CFI_ADJUST_CFA_OFFSET 4
1330 SAVE_ALL
1331 xorl %edx,%edx # zero error code
1332 movl %esp,%eax # pt_regs pointer
1333 call do_nmi
1334 jmp restore_nocheck_notrace
1335 CFI_ENDPROC
1336
1337 nmi_stack_fixup:
1338 RING0_INT_FRAME
1339 FIX_STACK 12, nmi_stack_correct, 1
1340 jmp nmi_stack_correct
1341
1342 nmi_debug_stack_check:
1343 /* We have a RING0_INT_FRAME here */
1344 cmpw $__KERNEL_CS,16(%esp)
1345 jne nmi_stack_correct
1346 cmpl $debug,(%esp)
1347 jb nmi_stack_correct
1348 cmpl $debug_esp_fix_insn,(%esp)
1349 ja nmi_stack_correct
1350 FIX_STACK 24, nmi_stack_correct, 1
1351 jmp nmi_stack_correct
1352
1353 nmi_espfix_stack:
1354 /* We have a RING0_INT_FRAME here.
1355 *
1356 * create the pointer to lss back
1357 */
1358 pushl %ss
1359 CFI_ADJUST_CFA_OFFSET 4
1360 pushl %esp
1361 CFI_ADJUST_CFA_OFFSET 4
1362 addl $4, (%esp)
1363 /* copy the iret frame of 12 bytes */
1364 .rept 3
1365 pushl 16(%esp)
1366 CFI_ADJUST_CFA_OFFSET 4
1367 .endr
1368 pushl %eax
1369 CFI_ADJUST_CFA_OFFSET 4
1370 SAVE_ALL
1371 FIXUP_ESPFIX_STACK # %eax == %esp
1372 xorl %edx,%edx # zero error code
1373 call do_nmi
1374 RESTORE_REGS
1375 lss 12+4(%esp), %esp # back to espfix stack
1376 CFI_ADJUST_CFA_OFFSET -24
1377 jmp irq_return
1378 CFI_ENDPROC
1379 END(nmi)
1380
1381 ENTRY(int3)
1382 RING0_INT_FRAME
1383 pushl $-1 # mark this as an int
1384 CFI_ADJUST_CFA_OFFSET 4
1385 SAVE_ALL
1386 TRACE_IRQS_OFF
1387 xorl %edx,%edx # zero error code
1388 movl %esp,%eax # pt_regs pointer
1389 call do_int3
1390 jmp ret_from_exception
1391 CFI_ENDPROC
1392 END(int3)
1393
1394 ENTRY(general_protection)
1395 RING0_EC_FRAME
1396 pushl $do_general_protection
1397 CFI_ADJUST_CFA_OFFSET 4
1398 jmp error_code
1399 CFI_ENDPROC
1400 END(general_protection)
1401
1402 /*
1403 * End of kprobes section
1404 */
1405 .popsection