X86: Add a check to catch Xen emulation of Hyper-V
[linux-2.6-block.git] / arch / x86 / kernel / entry_32.S
CommitLineData
1da177e4 1/*
1da177e4
LT
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 *
889f21ce 17 * Stack layout in 'syscall_exit':
1da177e4
LT
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
464d1a78 32 * 24(%esp) - %fs
ccbeed3a
TH
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
1da177e4
LT
40 *
41 * "current" is in register %ebx during any slow entries.
42 */
43
1da177e4 44#include <linux/linkage.h>
d7e7528b 45#include <linux/err.h>
1da177e4 46#include <asm/thread_info.h>
55f327fa 47#include <asm/irqflags.h>
1da177e4
LT
48#include <asm/errno.h>
49#include <asm/segment.h>
50#include <asm/smp.h>
0341c14d 51#include <asm/page_types.h>
be44d2aa 52#include <asm/percpu.h>
fe7cacc1 53#include <asm/dwarf2.h>
ab68ed98 54#include <asm/processor-flags.h>
395a59d0 55#include <asm/ftrace.h>
9b7dc567 56#include <asm/irq_vectors.h>
40d2e763 57#include <asm/cpufeature.h>
b4ca46e4 58#include <asm/alternative-asm.h>
6837a54d 59#include <asm/asm.h>
e59d1b0a 60#include <asm/smap.h>
1da177e4 61
af0575bb
RM
62/* Avoid __ASSEMBLER__'ifying <linux/audit.h> just for this. */
63#include <linux/elf-em.h>
64#define AUDIT_ARCH_I386 (EM_386|__AUDIT_ARCH_LE)
65#define __AUDIT_ARCH_LE 0x40000000
66
67#ifndef CONFIG_AUDITSYSCALL
68#define sysenter_audit syscall_trace_entry
69#define sysexit_audit syscall_exit_work
70#endif
71
ea714547
JO
72 .section .entry.text, "ax"
73
139ec7c4
RR
74/*
75 * We use macros for low-level operations which need to be overridden
76 * for paravirtualization. The following will never clobber any registers:
77 * INTERRUPT_RETURN (aka. "iret")
78 * GET_CR0_INTO_EAX (aka. "movl %cr0, %eax")
d75cd22f 79 * ENABLE_INTERRUPTS_SYSEXIT (aka "sti; sysexit").
139ec7c4
RR
80 *
81 * For DISABLE_INTERRUPTS/ENABLE_INTERRUPTS (aka "cli"/"sti"), you must
82 * specify what registers can be overwritten (CLBR_NONE, CLBR_EAX/EDX/ECX/ANY).
83 * Allowing a register to be clobbered can shrink the paravirt replacement
84 * enough to patch inline, increasing performance.
85 */
86
1da177e4 87#ifdef CONFIG_PREEMPT
139ec7c4 88#define preempt_stop(clobbers) DISABLE_INTERRUPTS(clobbers); TRACE_IRQS_OFF
1da177e4 89#else
139ec7c4 90#define preempt_stop(clobbers)
2e04bc76 91#define resume_kernel restore_all
1da177e4
LT
92#endif
93
55f327fa
IM
94.macro TRACE_IRQS_IRET
95#ifdef CONFIG_TRACE_IRQFLAGS
ab68ed98 96 testl $X86_EFLAGS_IF,PT_EFLAGS(%esp) # interrupts off?
55f327fa
IM
97 jz 1f
98 TRACE_IRQS_ON
991:
100#endif
101.endm
102
ccbeed3a
TH
103/*
104 * User gs save/restore
105 *
106 * %gs is used for userland TLS and kernel only uses it for stack
107 * canary which is required to be at %gs:20 by gcc. Read the comment
108 * at the top of stackprotector.h for more info.
109 *
110 * Local labels 98 and 99 are used.
111 */
112#ifdef CONFIG_X86_32_LAZY_GS
113
114 /* unfortunately push/pop can't be no-op */
115.macro PUSH_GS
df5d1874 116 pushl_cfi $0
ccbeed3a
TH
117.endm
118.macro POP_GS pop=0
119 addl $(4 + \pop), %esp
120 CFI_ADJUST_CFA_OFFSET -(4 + \pop)
121.endm
122.macro POP_GS_EX
123.endm
124
125 /* all the rest are no-op */
126.macro PTGS_TO_GS
127.endm
128.macro PTGS_TO_GS_EX
129.endm
130.macro GS_TO_REG reg
131.endm
132.macro REG_TO_PTGS reg
133.endm
134.macro SET_KERNEL_GS reg
135.endm
136
137#else /* CONFIG_X86_32_LAZY_GS */
138
139.macro PUSH_GS
df5d1874 140 pushl_cfi %gs
ccbeed3a
TH
141 /*CFI_REL_OFFSET gs, 0*/
142.endm
143
144.macro POP_GS pop=0
df5d1874 14598: popl_cfi %gs
ccbeed3a
TH
146 /*CFI_RESTORE gs*/
147 .if \pop <> 0
148 add $\pop, %esp
149 CFI_ADJUST_CFA_OFFSET -\pop
150 .endif
151.endm
152.macro POP_GS_EX
153.pushsection .fixup, "ax"
15499: movl $0, (%esp)
155 jmp 98b
ccbeed3a 156.popsection
6837a54d 157 _ASM_EXTABLE(98b,99b)
ccbeed3a
TH
158.endm
159
160.macro PTGS_TO_GS
16198: mov PT_GS(%esp), %gs
162.endm
163.macro PTGS_TO_GS_EX
164.pushsection .fixup, "ax"
16599: movl $0, PT_GS(%esp)
166 jmp 98b
ccbeed3a 167.popsection
6837a54d 168 _ASM_EXTABLE(98b,99b)
ccbeed3a
TH
169.endm
170
171.macro GS_TO_REG reg
172 movl %gs, \reg
173 /*CFI_REGISTER gs, \reg*/
174.endm
175.macro REG_TO_PTGS reg
176 movl \reg, PT_GS(%esp)
177 /*CFI_REL_OFFSET gs, PT_GS*/
178.endm
179.macro SET_KERNEL_GS reg
60a5317f 180 movl $(__KERNEL_STACK_CANARY), \reg
ccbeed3a
TH
181 movl \reg, %gs
182.endm
183
184#endif /* CONFIG_X86_32_LAZY_GS */
185
f0d96110
TH
186.macro SAVE_ALL
187 cld
ccbeed3a 188 PUSH_GS
df5d1874 189 pushl_cfi %fs
f0d96110 190 /*CFI_REL_OFFSET fs, 0;*/
df5d1874 191 pushl_cfi %es
f0d96110 192 /*CFI_REL_OFFSET es, 0;*/
df5d1874 193 pushl_cfi %ds
f0d96110 194 /*CFI_REL_OFFSET ds, 0;*/
df5d1874 195 pushl_cfi %eax
f0d96110 196 CFI_REL_OFFSET eax, 0
df5d1874 197 pushl_cfi %ebp
f0d96110 198 CFI_REL_OFFSET ebp, 0
df5d1874 199 pushl_cfi %edi
f0d96110 200 CFI_REL_OFFSET edi, 0
df5d1874 201 pushl_cfi %esi
f0d96110 202 CFI_REL_OFFSET esi, 0
df5d1874 203 pushl_cfi %edx
f0d96110 204 CFI_REL_OFFSET edx, 0
df5d1874 205 pushl_cfi %ecx
f0d96110 206 CFI_REL_OFFSET ecx, 0
df5d1874 207 pushl_cfi %ebx
f0d96110
TH
208 CFI_REL_OFFSET ebx, 0
209 movl $(__USER_DS), %edx
210 movl %edx, %ds
211 movl %edx, %es
212 movl $(__KERNEL_PERCPU), %edx
464d1a78 213 movl %edx, %fs
ccbeed3a 214 SET_KERNEL_GS %edx
f0d96110 215.endm
1da177e4 216
f0d96110 217.macro RESTORE_INT_REGS
df5d1874 218 popl_cfi %ebx
f0d96110 219 CFI_RESTORE ebx
df5d1874 220 popl_cfi %ecx
f0d96110 221 CFI_RESTORE ecx
df5d1874 222 popl_cfi %edx
f0d96110 223 CFI_RESTORE edx
df5d1874 224 popl_cfi %esi
f0d96110 225 CFI_RESTORE esi
df5d1874 226 popl_cfi %edi
f0d96110 227 CFI_RESTORE edi
df5d1874 228 popl_cfi %ebp
f0d96110 229 CFI_RESTORE ebp
df5d1874 230 popl_cfi %eax
fe7cacc1 231 CFI_RESTORE eax
f0d96110 232.endm
1da177e4 233
ccbeed3a 234.macro RESTORE_REGS pop=0
f0d96110 235 RESTORE_INT_REGS
df5d1874 2361: popl_cfi %ds
f0d96110 237 /*CFI_RESTORE ds;*/
df5d1874 2382: popl_cfi %es
f0d96110 239 /*CFI_RESTORE es;*/
df5d1874 2403: popl_cfi %fs
f0d96110 241 /*CFI_RESTORE fs;*/
ccbeed3a 242 POP_GS \pop
f0d96110
TH
243.pushsection .fixup, "ax"
2444: movl $0, (%esp)
245 jmp 1b
2465: movl $0, (%esp)
247 jmp 2b
2486: movl $0, (%esp)
249 jmp 3b
f95d47ca 250.popsection
6837a54d
PA
251 _ASM_EXTABLE(1b,4b)
252 _ASM_EXTABLE(2b,5b)
253 _ASM_EXTABLE(3b,6b)
ccbeed3a 254 POP_GS_EX
f0d96110 255.endm
1da177e4 256
f0d96110
TH
257.macro RING0_INT_FRAME
258 CFI_STARTPROC simple
259 CFI_SIGNAL_FRAME
260 CFI_DEF_CFA esp, 3*4
261 /*CFI_OFFSET cs, -2*4;*/
fe7cacc1 262 CFI_OFFSET eip, -3*4
f0d96110 263.endm
fe7cacc1 264
f0d96110
TH
265.macro RING0_EC_FRAME
266 CFI_STARTPROC simple
267 CFI_SIGNAL_FRAME
268 CFI_DEF_CFA esp, 4*4
269 /*CFI_OFFSET cs, -2*4;*/
fe7cacc1 270 CFI_OFFSET eip, -3*4
f0d96110 271.endm
fe7cacc1 272
f0d96110
TH
273.macro RING0_PTREGS_FRAME
274 CFI_STARTPROC simple
275 CFI_SIGNAL_FRAME
276 CFI_DEF_CFA esp, PT_OLDESP-PT_EBX
277 /*CFI_OFFSET cs, PT_CS-PT_OLDESP;*/
278 CFI_OFFSET eip, PT_EIP-PT_OLDESP
279 /*CFI_OFFSET es, PT_ES-PT_OLDESP;*/
280 /*CFI_OFFSET ds, PT_DS-PT_OLDESP;*/
281 CFI_OFFSET eax, PT_EAX-PT_OLDESP
282 CFI_OFFSET ebp, PT_EBP-PT_OLDESP
283 CFI_OFFSET edi, PT_EDI-PT_OLDESP
284 CFI_OFFSET esi, PT_ESI-PT_OLDESP
285 CFI_OFFSET edx, PT_EDX-PT_OLDESP
286 CFI_OFFSET ecx, PT_ECX-PT_OLDESP
eb5b7b9d 287 CFI_OFFSET ebx, PT_EBX-PT_OLDESP
f0d96110 288.endm
1da177e4
LT
289
290ENTRY(ret_from_fork)
fe7cacc1 291 CFI_STARTPROC
df5d1874 292 pushl_cfi %eax
1da177e4
LT
293 call schedule_tail
294 GET_THREAD_INFO(%ebp)
df5d1874
JB
295 popl_cfi %eax
296 pushl_cfi $0x0202 # Reset kernel eflags
297 popfl_cfi
1da177e4 298 jmp syscall_exit
fe7cacc1 299 CFI_ENDPROC
47a55cd7 300END(ret_from_fork)
1da177e4 301
22e2430d
AV
302ENTRY(ret_from_kernel_thread)
303 CFI_STARTPROC
304 pushl_cfi %eax
305 call schedule_tail
6783eaa2 306 GET_THREAD_INFO(%ebp)
22e2430d
AV
307 popl_cfi %eax
308 pushl_cfi $0x0202 # Reset kernel eflags
309 popfl_cfi
310 movl PT_EBP(%esp),%eax
311 call *PT_EBX(%esp)
312 movl $0,PT_EAX(%esp)
6783eaa2 313 jmp syscall_exit
22e2430d
AV
314 CFI_ENDPROC
315ENDPROC(ret_from_kernel_thread)
6783eaa2 316
a00e817f
MH
317/*
318 * Interrupt exit functions should be protected against kprobes
319 */
320 .pushsection .kprobes.text, "ax"
1da177e4
LT
321/*
322 * Return to user mode is not as complex as all this looks,
323 * but we want the default path for a system call return to
324 * go as quickly as possible which is why some of this is
325 * less clear than it otherwise should be.
326 */
327
328 # userspace resumption stub bypassing syscall exit tracing
329 ALIGN
fe7cacc1 330 RING0_PTREGS_FRAME
1da177e4 331ret_from_exception:
139ec7c4 332 preempt_stop(CLBR_ANY)
1da177e4
LT
333ret_from_intr:
334 GET_THREAD_INFO(%ebp)
29a2e283 335#ifdef CONFIG_VM86
eb5b7b9d
JF
336 movl PT_EFLAGS(%esp), %eax # mix EFLAGS and CS
337 movb PT_CS(%esp), %al
ab68ed98 338 andl $(X86_EFLAGS_VM | SEGMENT_RPL_MASK), %eax
29a2e283
DA
339#else
340 /*
6783eaa2 341 * We can be coming here from child spawned by kernel_thread().
29a2e283
DA
342 */
343 movl PT_CS(%esp), %eax
344 andl $SEGMENT_RPL_MASK, %eax
345#endif
78be3706
RR
346 cmpl $USER_RPL, %eax
347 jb resume_kernel # not returning to v8086 or userspace
f95d47ca 348
1da177e4 349ENTRY(resume_userspace)
c7e872e7 350 LOCKDEP_SYS_EXIT
139ec7c4 351 DISABLE_INTERRUPTS(CLBR_ANY) # make sure we don't miss an interrupt
1da177e4
LT
352 # setting need_resched or sigpending
353 # between sampling and the iret
e32e58a9 354 TRACE_IRQS_OFF
1da177e4
LT
355 movl TI_flags(%ebp), %ecx
356 andl $_TIF_WORK_MASK, %ecx # is there any work to be done on
357 # int/exception return?
358 jne work_pending
359 jmp restore_all
47a55cd7 360END(ret_from_exception)
1da177e4
LT
361
362#ifdef CONFIG_PREEMPT
363ENTRY(resume_kernel)
139ec7c4 364 DISABLE_INTERRUPTS(CLBR_ANY)
1da177e4 365 cmpl $0,TI_preempt_count(%ebp) # non-zero preempt_count ?
2e04bc76 366 jnz restore_all
1da177e4
LT
367need_resched:
368 movl TI_flags(%ebp), %ecx # need_resched set ?
369 testb $_TIF_NEED_RESCHED, %cl
370 jz restore_all
ab68ed98 371 testl $X86_EFLAGS_IF,PT_EFLAGS(%esp) # interrupts off (exception path) ?
1da177e4
LT
372 jz restore_all
373 call preempt_schedule_irq
374 jmp need_resched
47a55cd7 375END(resume_kernel)
1da177e4 376#endif
fe7cacc1 377 CFI_ENDPROC
a00e817f
MH
378/*
379 * End of kprobes section
380 */
381 .popsection
1da177e4
LT
382
383/* SYSENTER_RETURN points to after the "sysenter" instruction in
384 the vsyscall page. See vsyscall-sysentry.S, which defines the symbol. */
385
386 # sysenter call handler stub
0aa97fb2 387ENTRY(ia32_sysenter_target)
fe7cacc1 388 CFI_STARTPROC simple
adf14236 389 CFI_SIGNAL_FRAME
fe7cacc1
JB
390 CFI_DEF_CFA esp, 0
391 CFI_REGISTER esp, ebp
faca6227 392 movl TSS_sysenter_sp0(%esp),%esp
1da177e4 393sysenter_past_esp:
55f327fa 394 /*
d93c870b
JF
395 * Interrupts are disabled here, but we can't trace it until
396 * enough kernel state to call TRACE_IRQS_OFF can be called - but
397 * we immediately enable interrupts at that point anyway.
55f327fa 398 */
3234282f 399 pushl_cfi $__USER_DS
fe7cacc1 400 /*CFI_REL_OFFSET ss, 0*/
df5d1874 401 pushl_cfi %ebp
fe7cacc1 402 CFI_REL_OFFSET esp, 0
df5d1874 403 pushfl_cfi
d93c870b 404 orl $X86_EFLAGS_IF, (%esp)
3234282f 405 pushl_cfi $__USER_CS
fe7cacc1 406 /*CFI_REL_OFFSET cs, 0*/
e6e5494c
IM
407 /*
408 * Push current_thread_info()->sysenter_return to the stack.
409 * A tiny bit of offset fixup is necessary - 4*4 means the 4 words
410 * pushed above; +8 corresponds to copy_thread's esp0 setting.
411 */
7bf04be8 412 pushl_cfi ((TI_sysenter_return)-THREAD_SIZE+8+4*4)(%esp)
fe7cacc1 413 CFI_REL_OFFSET eip, 0
1da177e4 414
df5d1874 415 pushl_cfi %eax
d93c870b
JF
416 SAVE_ALL
417 ENABLE_INTERRUPTS(CLBR_NONE)
418
1da177e4
LT
419/*
420 * Load the potential sixth argument from user stack.
421 * Careful about security.
422 */
423 cmpl $__PAGE_OFFSET-3,%ebp
424 jae syscall_fault
e59d1b0a 425 ASM_STAC
1da177e4 4261: movl (%ebp),%ebp
e59d1b0a 427 ASM_CLAC
d93c870b 428 movl %ebp,PT_EBP(%esp)
6837a54d 429 _ASM_EXTABLE(1b,syscall_fault)
1da177e4 430
1da177e4
LT
431 GET_THREAD_INFO(%ebp)
432
88200bc2 433 testl $_TIF_WORK_SYSCALL_ENTRY,TI_flags(%ebp)
af0575bb
RM
434 jnz sysenter_audit
435sysenter_do_call:
303395ac 436 cmpl $(NR_syscalls), %eax
1da177e4
LT
437 jae syscall_badsys
438 call *sys_call_table(,%eax,4)
eb5b7b9d 439 movl %eax,PT_EAX(%esp)
c7e872e7 440 LOCKDEP_SYS_EXIT
42c24fa2 441 DISABLE_INTERRUPTS(CLBR_ANY)
55f327fa 442 TRACE_IRQS_OFF
1da177e4 443 movl TI_flags(%ebp), %ecx
88200bc2 444 testl $_TIF_ALLWORK_MASK, %ecx
af0575bb
RM
445 jne sysexit_audit
446sysenter_exit:
1da177e4 447/* if something modifies registers it must also disable sysexit */
eb5b7b9d
JF
448 movl PT_EIP(%esp), %edx
449 movl PT_OLDESP(%esp), %ecx
1da177e4 450 xorl %ebp,%ebp
55f327fa 451 TRACE_IRQS_ON
464d1a78 4521: mov PT_FS(%esp), %fs
ccbeed3a 453 PTGS_TO_GS
d75cd22f 454 ENABLE_INTERRUPTS_SYSEXIT
af0575bb
RM
455
456#ifdef CONFIG_AUDITSYSCALL
457sysenter_audit:
88200bc2 458 testl $(_TIF_WORK_SYSCALL_ENTRY & ~_TIF_SYSCALL_AUDIT),TI_flags(%ebp)
af0575bb
RM
459 jnz syscall_trace_entry
460 addl $4,%esp
461 CFI_ADJUST_CFA_OFFSET -4
462 /* %esi already in 8(%esp) 6th arg: 4th syscall arg */
463 /* %edx already in 4(%esp) 5th arg: 3rd syscall arg */
464 /* %ecx already in 0(%esp) 4th arg: 2nd syscall arg */
465 movl %ebx,%ecx /* 3rd arg: 1st syscall arg */
466 movl %eax,%edx /* 2nd arg: syscall number */
467 movl $AUDIT_ARCH_I386,%eax /* 1st arg: audit arch */
b05d8447 468 call __audit_syscall_entry
df5d1874 469 pushl_cfi %ebx
af0575bb
RM
470 movl PT_EAX(%esp),%eax /* reload syscall number */
471 jmp sysenter_do_call
472
473sysexit_audit:
88200bc2 474 testl $(_TIF_ALLWORK_MASK & ~_TIF_SYSCALL_AUDIT), %ecx
af0575bb
RM
475 jne syscall_exit_work
476 TRACE_IRQS_ON
477 ENABLE_INTERRUPTS(CLBR_ANY)
478 movl %eax,%edx /* second arg, syscall return value */
d7e7528b
EP
479 cmpl $-MAX_ERRNO,%eax /* is it an error ? */
480 setbe %al /* 1 if so, 0 if not */
af0575bb 481 movzbl %al,%eax /* zero-extend that */
d7e7528b 482 call __audit_syscall_exit
af0575bb
RM
483 DISABLE_INTERRUPTS(CLBR_ANY)
484 TRACE_IRQS_OFF
485 movl TI_flags(%ebp), %ecx
88200bc2 486 testl $(_TIF_ALLWORK_MASK & ~_TIF_SYSCALL_AUDIT), %ecx
af0575bb
RM
487 jne syscall_exit_work
488 movl PT_EAX(%esp),%eax /* reload syscall return value */
489 jmp sysenter_exit
490#endif
491
fe7cacc1 492 CFI_ENDPROC
f95d47ca 493.pushsection .fixup,"ax"
464d1a78 4942: movl $0,PT_FS(%esp)
f95d47ca 495 jmp 1b
f95d47ca 496.popsection
6837a54d 497 _ASM_EXTABLE(1b,2b)
ccbeed3a 498 PTGS_TO_GS_EX
0aa97fb2 499ENDPROC(ia32_sysenter_target)
1da177e4 500
a00e817f
MH
501/*
502 * syscall stub including irq exit should be protected against kprobes
503 */
504 .pushsection .kprobes.text, "ax"
1da177e4
LT
505 # system call handler stub
506ENTRY(system_call)
fe7cacc1 507 RING0_INT_FRAME # can't unwind into user space anyway
e59d1b0a 508 ASM_CLAC
df5d1874 509 pushl_cfi %eax # save orig_eax
1da177e4
LT
510 SAVE_ALL
511 GET_THREAD_INFO(%ebp)
ed75e8d5 512 # system call tracing in operation / emulation
88200bc2 513 testl $_TIF_WORK_SYSCALL_ENTRY,TI_flags(%ebp)
1da177e4 514 jnz syscall_trace_entry
303395ac 515 cmpl $(NR_syscalls), %eax
1da177e4
LT
516 jae syscall_badsys
517syscall_call:
518 call *sys_call_table(,%eax,4)
eb5b7b9d 519 movl %eax,PT_EAX(%esp) # store the return value
1da177e4 520syscall_exit:
c7e872e7 521 LOCKDEP_SYS_EXIT
139ec7c4 522 DISABLE_INTERRUPTS(CLBR_ANY) # make sure we don't miss an interrupt
1da177e4
LT
523 # setting need_resched or sigpending
524 # between sampling and the iret
55f327fa 525 TRACE_IRQS_OFF
1da177e4 526 movl TI_flags(%ebp), %ecx
88200bc2 527 testl $_TIF_ALLWORK_MASK, %ecx # current->work
1da177e4
LT
528 jne syscall_exit_work
529
530restore_all:
2e04bc76
AH
531 TRACE_IRQS_IRET
532restore_all_notrace:
eb5b7b9d
JF
533 movl PT_EFLAGS(%esp), %eax # mix EFLAGS, SS and CS
534 # Warning: PT_OLDSS(%esp) contains the wrong/random values if we
5df24082
SS
535 # are returning to the kernel.
536 # See comments in process.c:copy_thread() for details.
eb5b7b9d
JF
537 movb PT_OLDSS(%esp), %ah
538 movb PT_CS(%esp), %al
ab68ed98 539 andl $(X86_EFLAGS_VM | (SEGMENT_TI_MASK << 8) | SEGMENT_RPL_MASK), %eax
78be3706 540 cmpl $((SEGMENT_LDT << 8) | USER_RPL), %eax
fe7cacc1 541 CFI_REMEMBER_STATE
1da177e4
LT
542 je ldt_ss # returning to user-space with LDT SS
543restore_nocheck:
ccbeed3a 544 RESTORE_REGS 4 # skip orig_eax/error_code
f7f3d791 545irq_return:
3701d863 546 INTERRUPT_RETURN
1da177e4 547.section .fixup,"ax"
90e9f536 548ENTRY(iret_exc)
a879cbbb
LT
549 pushl $0 # no error code
550 pushl $do_iret_error
551 jmp error_code
1da177e4 552.previous
6837a54d 553 _ASM_EXTABLE(irq_return,iret_exc)
1da177e4 554
fe7cacc1 555 CFI_RESTORE_STATE
1da177e4 556ldt_ss:
eb5b7b9d 557 larl PT_OLDSS(%esp), %eax
1da177e4
LT
558 jnz restore_nocheck
559 testl $0x00400000, %eax # returning to 32bit stack?
560 jnz restore_nocheck # allright, normal return
d3561b7f
RR
561
562#ifdef CONFIG_PARAVIRT
563 /*
564 * The kernel can't run on a non-flat stack if paravirt mode
565 * is active. Rather than try to fixup the high bits of
566 * ESP, bypass this code entirely. This may break DOSemu
567 * and/or Wine support in a paravirt VM, although the option
568 * is still available to implement the setting of the high
569 * 16-bits in the INTERRUPT_RETURN paravirt-op.
570 */
93b1eab3 571 cmpl $0, pv_info+PARAVIRT_enabled
d3561b7f
RR
572 jne restore_nocheck
573#endif
574
dc4c2a0a
AH
575/*
576 * Setup and switch to ESPFIX stack
577 *
578 * We're returning to userspace with a 16 bit stack. The CPU will not
579 * restore the high word of ESP for us on executing iret... This is an
580 * "official" bug of all the x86-compatible CPUs, which we can work
581 * around to make dosemu and wine happy. We do this by preloading the
582 * high word of ESP with the high word of the userspace ESP while
583 * compensating for the offset by changing to the ESPFIX segment with
584 * a base address that matches for the difference.
585 */
72c511dd 586#define GDT_ESPFIX_SS PER_CPU_VAR(gdt_page) + (GDT_ENTRY_ESPFIX_SS * 8)
dc4c2a0a
AH
587 mov %esp, %edx /* load kernel esp */
588 mov PT_OLDESP(%esp), %eax /* load userspace esp */
589 mov %dx, %ax /* eax: new kernel esp */
590 sub %eax, %edx /* offset (low word is 0) */
dc4c2a0a 591 shr $16, %edx
72c511dd
BG
592 mov %dl, GDT_ESPFIX_SS + 4 /* bits 16..23 */
593 mov %dh, GDT_ESPFIX_SS + 7 /* bits 24..31 */
df5d1874
JB
594 pushl_cfi $__ESPFIX_SS
595 pushl_cfi %eax /* new kernel esp */
2e04bc76
AH
596 /* Disable interrupts, but do not irqtrace this section: we
597 * will soon execute iret and the tracer was already set to
598 * the irqstate after the iret */
139ec7c4 599 DISABLE_INTERRUPTS(CLBR_EAX)
dc4c2a0a 600 lss (%esp), %esp /* switch to espfix segment */
be44d2aa
SS
601 CFI_ADJUST_CFA_OFFSET -8
602 jmp restore_nocheck
fe7cacc1 603 CFI_ENDPROC
47a55cd7 604ENDPROC(system_call)
1da177e4
LT
605
606 # perform work that needs to be done immediately before resumption
607 ALIGN
fe7cacc1 608 RING0_PTREGS_FRAME # can't unwind into user space anyway
1da177e4
LT
609work_pending:
610 testb $_TIF_NEED_RESCHED, %cl
611 jz work_notifysig
612work_resched:
613 call schedule
c7e872e7 614 LOCKDEP_SYS_EXIT
139ec7c4 615 DISABLE_INTERRUPTS(CLBR_ANY) # make sure we don't miss an interrupt
1da177e4
LT
616 # setting need_resched or sigpending
617 # between sampling and the iret
55f327fa 618 TRACE_IRQS_OFF
1da177e4
LT
619 movl TI_flags(%ebp), %ecx
620 andl $_TIF_WORK_MASK, %ecx # is there any work to be done other
621 # than syscall tracing?
622 jz restore_all
623 testb $_TIF_NEED_RESCHED, %cl
624 jnz work_resched
625
626work_notifysig: # deal with pending signals and
627 # notify-resume requests
74b47a78 628#ifdef CONFIG_VM86
ab68ed98 629 testl $X86_EFLAGS_VM, PT_EFLAGS(%esp)
1da177e4
LT
630 movl %esp, %eax
631 jne work_notifysig_v86 # returning to kernel-space or
632 # vm86-space
969ae0bf
AV
6331:
634#else
635 movl %esp, %eax
636#endif
3596ff4e
SD
637 TRACE_IRQS_ON
638 ENABLE_INTERRUPTS(CLBR_NONE)
44fbbb3d
AV
639 movb PT_CS(%esp), %bl
640 andb $SEGMENT_RPL_MASK, %bl
641 cmpb $USER_RPL, %bl
642 jb resume_kernel
1da177e4
LT
643 xorl %edx, %edx
644 call do_notify_resume
44fbbb3d 645 jmp resume_userspace
1da177e4 646
969ae0bf 647#ifdef CONFIG_VM86
1da177e4
LT
648 ALIGN
649work_notifysig_v86:
df5d1874 650 pushl_cfi %ecx # save ti_flags for do_notify_resume
1da177e4 651 call save_v86_state # %eax contains pt_regs pointer
df5d1874 652 popl_cfi %ecx
1da177e4 653 movl %eax, %esp
969ae0bf 654 jmp 1b
74b47a78 655#endif
47a55cd7 656END(work_pending)
1da177e4
LT
657
658 # perform syscall exit tracing
659 ALIGN
660syscall_trace_entry:
eb5b7b9d 661 movl $-ENOSYS,PT_EAX(%esp)
1da177e4 662 movl %esp, %eax
d4d67150
RM
663 call syscall_trace_enter
664 /* What it returned is what we'll actually use. */
303395ac 665 cmpl $(NR_syscalls), %eax
1da177e4
LT
666 jnae syscall_call
667 jmp syscall_exit
47a55cd7 668END(syscall_trace_entry)
1da177e4
LT
669
670 # perform syscall exit tracing
671 ALIGN
672syscall_exit_work:
88200bc2 673 testl $_TIF_WORK_SYSCALL_EXIT, %ecx
1da177e4 674 jz work_pending
55f327fa 675 TRACE_IRQS_ON
d4d67150 676 ENABLE_INTERRUPTS(CLBR_ANY) # could let syscall_trace_leave() call
1da177e4
LT
677 # schedule() instead
678 movl %esp, %eax
d4d67150 679 call syscall_trace_leave
1da177e4 680 jmp resume_userspace
47a55cd7 681END(syscall_exit_work)
fe7cacc1 682 CFI_ENDPROC
1da177e4 683
fe7cacc1 684 RING0_INT_FRAME # can't unwind into user space anyway
1da177e4 685syscall_fault:
e59d1b0a 686 ASM_CLAC
1da177e4 687 GET_THREAD_INFO(%ebp)
eb5b7b9d 688 movl $-EFAULT,PT_EAX(%esp)
1da177e4 689 jmp resume_userspace
47a55cd7 690END(syscall_fault)
1da177e4 691
1da177e4 692syscall_badsys:
eb5b7b9d 693 movl $-ENOSYS,PT_EAX(%esp)
1da177e4 694 jmp resume_userspace
47a55cd7 695END(syscall_badsys)
fe7cacc1 696 CFI_ENDPROC
a00e817f
MH
697/*
698 * End of kprobes section
699 */
700 .popsection
1da177e4 701
253f29a4
BG
702/*
703 * System calls that need a pt_regs pointer.
704 */
e258e4e0 705#define PTREGSCALL0(name) \
303395ac 706ENTRY(ptregs_##name) ; \
253f29a4 707 leal 4(%esp),%eax; \
303395ac
PA
708 jmp sys_##name; \
709ENDPROC(ptregs_##name)
253f29a4 710
e258e4e0 711#define PTREGSCALL1(name) \
303395ac 712ENTRY(ptregs_##name) ; \
e258e4e0 713 leal 4(%esp),%edx; \
ce9119ad 714 movl (PT_EBX+4)(%esp),%eax; \
303395ac
PA
715 jmp sys_##name; \
716ENDPROC(ptregs_##name)
e258e4e0
BG
717
718#define PTREGSCALL2(name) \
303395ac 719ENTRY(ptregs_##name) ; \
e258e4e0 720 leal 4(%esp),%ecx; \
ce9119ad
PA
721 movl (PT_ECX+4)(%esp),%edx; \
722 movl (PT_EBX+4)(%esp),%eax; \
303395ac
PA
723 jmp sys_##name; \
724ENDPROC(ptregs_##name)
e258e4e0
BG
725
726#define PTREGSCALL3(name) \
303395ac 727ENTRY(ptregs_##name) ; \
a34107b5 728 CFI_STARTPROC; \
e258e4e0 729 leal 4(%esp),%eax; \
a34107b5 730 pushl_cfi %eax; \
e258e4e0
BG
731 movl PT_EDX(%eax),%ecx; \
732 movl PT_ECX(%eax),%edx; \
733 movl PT_EBX(%eax),%eax; \
734 call sys_##name; \
735 addl $4,%esp; \
a34107b5
JB
736 CFI_ADJUST_CFA_OFFSET -4; \
737 ret; \
738 CFI_ENDPROC; \
739ENDPROC(ptregs_##name)
e258e4e0 740
27f59559 741PTREGSCALL1(iopl)
e258e4e0
BG
742PTREGSCALL0(sigreturn)
743PTREGSCALL0(rt_sigreturn)
f1382f15
BG
744PTREGSCALL2(vm86)
745PTREGSCALL1(vm86old)
253f29a4 746
f0d96110 747.macro FIXUP_ESPFIX_STACK
dc4c2a0a
AH
748/*
749 * Switch back for ESPFIX stack to the normal zerobased stack
750 *
751 * We can't call C functions using the ESPFIX stack. This code reads
752 * the high word of the segment base from the GDT and swiches to the
753 * normal stack and adjusts ESP with the matching offset.
754 */
755 /* fixup the stack */
72c511dd
BG
756 mov GDT_ESPFIX_SS + 4, %al /* bits 16..23 */
757 mov GDT_ESPFIX_SS + 7, %ah /* bits 24..31 */
dc4c2a0a
AH
758 shl $16, %eax
759 addl %esp, %eax /* the adjusted stack pointer */
df5d1874
JB
760 pushl_cfi $__KERNEL_DS
761 pushl_cfi %eax
dc4c2a0a 762 lss (%esp), %esp /* switch to the normal stack segment */
f0d96110
TH
763 CFI_ADJUST_CFA_OFFSET -8
764.endm
765.macro UNWIND_ESPFIX_STACK
766 movl %ss, %eax
767 /* see if on espfix stack */
768 cmpw $__ESPFIX_SS, %ax
769 jne 27f
770 movl $__KERNEL_DS, %eax
771 movl %eax, %ds
772 movl %eax, %es
773 /* switch to normal stack */
774 FIXUP_ESPFIX_STACK
77527:
776.endm
1da177e4
LT
777
778/*
b7c6244f
PA
779 * Build the entry stubs and pointer table with some assembler magic.
780 * We pack 7 stubs into a single 32-byte chunk, which will fit in a
781 * single cache line on all modern x86 implementations.
1da177e4 782 */
4687518c 783.section .init.rodata,"a"
1da177e4 784ENTRY(interrupt)
ea714547 785.section .entry.text, "ax"
b7c6244f
PA
786 .p2align 5
787 .p2align CONFIG_X86_L1_CACHE_SHIFT
1da177e4 788ENTRY(irq_entries_start)
fe7cacc1 789 RING0_INT_FRAME
4687518c 790vector=FIRST_EXTERNAL_VECTOR
b7c6244f
PA
791.rept (NR_VECTORS-FIRST_EXTERNAL_VECTOR+6)/7
792 .balign 32
793 .rept 7
794 .if vector < NR_VECTORS
8665596e 795 .if vector <> FIRST_EXTERNAL_VECTOR
fe7cacc1 796 CFI_ADJUST_CFA_OFFSET -4
b7c6244f 797 .endif
df5d1874 7981: pushl_cfi $(~vector+0x80) /* Note: always in signed byte range */
8665596e 799 .if ((vector-FIRST_EXTERNAL_VECTOR)%7) <> 6
b7c6244f
PA
800 jmp 2f
801 .endif
802 .previous
1da177e4 803 .long 1b
ea714547 804 .section .entry.text, "ax"
1da177e4 805vector=vector+1
b7c6244f
PA
806 .endif
807 .endr
8082: jmp common_interrupt
1da177e4 809.endr
47a55cd7
JB
810END(irq_entries_start)
811
812.previous
813END(interrupt)
814.previous
1da177e4 815
55f327fa
IM
816/*
817 * the CPU automatically disables interrupts when executing an IRQ vector,
818 * so IRQ-flags tracing has to follow that:
819 */
b7c6244f 820 .p2align CONFIG_X86_L1_CACHE_SHIFT
1da177e4 821common_interrupt:
e59d1b0a 822 ASM_CLAC
b7c6244f 823 addl $-0x80,(%esp) /* Adjust vector into the [-256,-1] range */
1da177e4 824 SAVE_ALL
55f327fa 825 TRACE_IRQS_OFF
1da177e4
LT
826 movl %esp,%eax
827 call do_IRQ
828 jmp ret_from_intr
47a55cd7 829ENDPROC(common_interrupt)
fe7cacc1 830 CFI_ENDPROC
1da177e4 831
a00e817f
MH
832/*
833 * Irq entries should be protected against kprobes
834 */
835 .pushsection .kprobes.text, "ax"
02cf94c3 836#define BUILD_INTERRUPT3(name, nr, fn) \
1da177e4 837ENTRY(name) \
fe7cacc1 838 RING0_INT_FRAME; \
e59d1b0a 839 ASM_CLAC; \
df5d1874 840 pushl_cfi $~(nr); \
fe7cacc1 841 SAVE_ALL; \
55f327fa 842 TRACE_IRQS_OFF \
1da177e4 843 movl %esp,%eax; \
02cf94c3 844 call fn; \
55f327fa 845 jmp ret_from_intr; \
47a55cd7
JB
846 CFI_ENDPROC; \
847ENDPROC(name)
1da177e4 848
02cf94c3
TH
849#define BUILD_INTERRUPT(name, nr) BUILD_INTERRUPT3(name, nr, smp_##name)
850
1da177e4 851/* The include is where all of the SMP etc. interrupts come from */
1164dd00 852#include <asm/entry_arch.h>
1da177e4 853
1da177e4 854ENTRY(coprocessor_error)
fe7cacc1 855 RING0_INT_FRAME
e59d1b0a 856 ASM_CLAC
df5d1874
JB
857 pushl_cfi $0
858 pushl_cfi $do_coprocessor_error
1da177e4 859 jmp error_code
fe7cacc1 860 CFI_ENDPROC
47a55cd7 861END(coprocessor_error)
1da177e4
LT
862
863ENTRY(simd_coprocessor_error)
fe7cacc1 864 RING0_INT_FRAME
e59d1b0a 865 ASM_CLAC
df5d1874 866 pushl_cfi $0
40d2e763
BG
867#ifdef CONFIG_X86_INVD_BUG
868 /* AMD 486 bug: invd from userspace calls exception 19 instead of #GP */
df5d1874 869661: pushl_cfi $do_general_protection
40d2e763
BG
870662:
871.section .altinstructions,"a"
b4ca46e4 872 altinstruction_entry 661b, 663f, X86_FEATURE_XMM, 662b-661b, 664f-663f
40d2e763
BG
873.previous
874.section .altinstr_replacement,"ax"
875663: pushl $do_simd_coprocessor_error
876664:
877.previous
878#else
df5d1874 879 pushl_cfi $do_simd_coprocessor_error
40d2e763 880#endif
1da177e4 881 jmp error_code
fe7cacc1 882 CFI_ENDPROC
47a55cd7 883END(simd_coprocessor_error)
1da177e4
LT
884
885ENTRY(device_not_available)
fe7cacc1 886 RING0_INT_FRAME
e59d1b0a 887 ASM_CLAC
df5d1874
JB
888 pushl_cfi $-1 # mark this as an int
889 pushl_cfi $do_device_not_available
7643e9b9 890 jmp error_code
fe7cacc1 891 CFI_ENDPROC
47a55cd7 892END(device_not_available)
1da177e4 893
d3561b7f
RR
894#ifdef CONFIG_PARAVIRT
895ENTRY(native_iret)
3701d863 896 iret
6837a54d 897 _ASM_EXTABLE(native_iret, iret_exc)
47a55cd7 898END(native_iret)
d3561b7f 899
d75cd22f 900ENTRY(native_irq_enable_sysexit)
d3561b7f
RR
901 sti
902 sysexit
d75cd22f 903END(native_irq_enable_sysexit)
d3561b7f
RR
904#endif
905
1da177e4 906ENTRY(overflow)
fe7cacc1 907 RING0_INT_FRAME
e59d1b0a 908 ASM_CLAC
df5d1874
JB
909 pushl_cfi $0
910 pushl_cfi $do_overflow
1da177e4 911 jmp error_code
fe7cacc1 912 CFI_ENDPROC
47a55cd7 913END(overflow)
1da177e4
LT
914
915ENTRY(bounds)
fe7cacc1 916 RING0_INT_FRAME
e59d1b0a 917 ASM_CLAC
df5d1874
JB
918 pushl_cfi $0
919 pushl_cfi $do_bounds
1da177e4 920 jmp error_code
fe7cacc1 921 CFI_ENDPROC
47a55cd7 922END(bounds)
1da177e4
LT
923
924ENTRY(invalid_op)
fe7cacc1 925 RING0_INT_FRAME
e59d1b0a 926 ASM_CLAC
df5d1874
JB
927 pushl_cfi $0
928 pushl_cfi $do_invalid_op
1da177e4 929 jmp error_code
fe7cacc1 930 CFI_ENDPROC
47a55cd7 931END(invalid_op)
1da177e4
LT
932
933ENTRY(coprocessor_segment_overrun)
fe7cacc1 934 RING0_INT_FRAME
e59d1b0a 935 ASM_CLAC
df5d1874
JB
936 pushl_cfi $0
937 pushl_cfi $do_coprocessor_segment_overrun
1da177e4 938 jmp error_code
fe7cacc1 939 CFI_ENDPROC
47a55cd7 940END(coprocessor_segment_overrun)
1da177e4
LT
941
942ENTRY(invalid_TSS)
fe7cacc1 943 RING0_EC_FRAME
e59d1b0a 944 ASM_CLAC
df5d1874 945 pushl_cfi $do_invalid_TSS
1da177e4 946 jmp error_code
fe7cacc1 947 CFI_ENDPROC
47a55cd7 948END(invalid_TSS)
1da177e4
LT
949
950ENTRY(segment_not_present)
fe7cacc1 951 RING0_EC_FRAME
e59d1b0a 952 ASM_CLAC
df5d1874 953 pushl_cfi $do_segment_not_present
1da177e4 954 jmp error_code
fe7cacc1 955 CFI_ENDPROC
47a55cd7 956END(segment_not_present)
1da177e4
LT
957
958ENTRY(stack_segment)
fe7cacc1 959 RING0_EC_FRAME
e59d1b0a 960 ASM_CLAC
df5d1874 961 pushl_cfi $do_stack_segment
1da177e4 962 jmp error_code
fe7cacc1 963 CFI_ENDPROC
47a55cd7 964END(stack_segment)
1da177e4 965
1da177e4 966ENTRY(alignment_check)
fe7cacc1 967 RING0_EC_FRAME
e59d1b0a 968 ASM_CLAC
df5d1874 969 pushl_cfi $do_alignment_check
1da177e4 970 jmp error_code
fe7cacc1 971 CFI_ENDPROC
47a55cd7 972END(alignment_check)
1da177e4 973
d28c4393
P
974ENTRY(divide_error)
975 RING0_INT_FRAME
e59d1b0a 976 ASM_CLAC
df5d1874
JB
977 pushl_cfi $0 # no error code
978 pushl_cfi $do_divide_error
1da177e4 979 jmp error_code
fe7cacc1 980 CFI_ENDPROC
47a55cd7 981END(divide_error)
1da177e4
LT
982
983#ifdef CONFIG_X86_MCE
984ENTRY(machine_check)
fe7cacc1 985 RING0_INT_FRAME
e59d1b0a 986 ASM_CLAC
df5d1874
JB
987 pushl_cfi $0
988 pushl_cfi machine_check_vector
1da177e4 989 jmp error_code
fe7cacc1 990 CFI_ENDPROC
47a55cd7 991END(machine_check)
1da177e4
LT
992#endif
993
994ENTRY(spurious_interrupt_bug)
fe7cacc1 995 RING0_INT_FRAME
e59d1b0a 996 ASM_CLAC
df5d1874
JB
997 pushl_cfi $0
998 pushl_cfi $do_spurious_interrupt_bug
1da177e4 999 jmp error_code
fe7cacc1 1000 CFI_ENDPROC
47a55cd7 1001END(spurious_interrupt_bug)
a00e817f
MH
1002/*
1003 * End of kprobes section
1004 */
1005 .popsection
1da177e4 1006
5ead97c8 1007#ifdef CONFIG_XEN
e2a81baf
JF
1008/* Xen doesn't set %esp to be precisely what the normal sysenter
1009 entrypoint expects, so fix it up before using the normal path. */
1010ENTRY(xen_sysenter_target)
1011 RING0_INT_FRAME
1012 addl $5*4, %esp /* remove xen-provided frame */
2ddf9b7b 1013 CFI_ADJUST_CFA_OFFSET -5*4
e2a81baf 1014 jmp sysenter_past_esp
557d7d4e 1015 CFI_ENDPROC
e2a81baf 1016
5ead97c8
JF
1017ENTRY(xen_hypervisor_callback)
1018 CFI_STARTPROC
a349e23d 1019 pushl_cfi $-1 /* orig_ax = -1 => not a system call */
5ead97c8
JF
1020 SAVE_ALL
1021 TRACE_IRQS_OFF
9ec2b804
JF
1022
1023 /* Check to see if we got the event in the critical
1024 region in xen_iret_direct, after we've reenabled
1025 events and checked for pending events. This simulates
1026 iret instruction's behaviour where it delivers a
1027 pending interrupt when enabling interrupts. */
1028 movl PT_EIP(%esp),%eax
1029 cmpl $xen_iret_start_crit,%eax
1030 jb 1f
1031 cmpl $xen_iret_end_crit,%eax
1032 jae 1f
1033
0f2c8769 1034 jmp xen_iret_crit_fixup
e2a81baf 1035
e2a81baf 1036ENTRY(xen_do_upcall)
b77797fb 10371: mov %esp, %eax
5ead97c8
JF
1038 call xen_evtchn_do_upcall
1039 jmp ret_from_intr
1040 CFI_ENDPROC
1041ENDPROC(xen_hypervisor_callback)
1042
1043# Hypervisor uses this for application faults while it executes.
1044# We get here for two reasons:
1045# 1. Fault while reloading DS, ES, FS or GS
1046# 2. Fault while executing IRET
1047# Category 1 we fix up by reattempting the load, and zeroing the segment
1048# register if the load fails.
1049# Category 2 we fix up by jumping to do_iret_error. We cannot use the
1050# normal Linux return path in this case because if we use the IRET hypercall
1051# to pop the stack frame we end up in an infinite loop of failsafe callbacks.
1052# We distinguish between categories by maintaining a status value in EAX.
1053ENTRY(xen_failsafe_callback)
1054 CFI_STARTPROC
df5d1874 1055 pushl_cfi %eax
5ead97c8
JF
1056 movl $1,%eax
10571: mov 4(%esp),%ds
10582: mov 8(%esp),%es
10593: mov 12(%esp),%fs
10604: mov 16(%esp),%gs
a349e23d
DV
1061 /* EAX == 0 => Category 1 (Bad segment)
1062 EAX != 0 => Category 2 (Bad IRET) */
5ead97c8 1063 testl %eax,%eax
df5d1874 1064 popl_cfi %eax
5ead97c8
JF
1065 lea 16(%esp),%esp
1066 CFI_ADJUST_CFA_OFFSET -16
1067 jz 5f
a349e23d
DV
1068 jmp iret_exc
10695: pushl_cfi $-1 /* orig_ax = -1 => not a system call */
5ead97c8
JF
1070 SAVE_ALL
1071 jmp ret_from_exception
1072 CFI_ENDPROC
1073
1074.section .fixup,"ax"
10756: xorl %eax,%eax
1076 movl %eax,4(%esp)
1077 jmp 1b
10787: xorl %eax,%eax
1079 movl %eax,8(%esp)
1080 jmp 2b
10818: xorl %eax,%eax
1082 movl %eax,12(%esp)
1083 jmp 3b
10849: xorl %eax,%eax
1085 movl %eax,16(%esp)
1086 jmp 4b
1087.previous
6837a54d
PA
1088 _ASM_EXTABLE(1b,6b)
1089 _ASM_EXTABLE(2b,7b)
1090 _ASM_EXTABLE(3b,8b)
1091 _ASM_EXTABLE(4b,9b)
5ead97c8
JF
1092ENDPROC(xen_failsafe_callback)
1093
38e20b07
SY
1094BUILD_INTERRUPT3(xen_hvm_callback_vector, XEN_HVM_EVTCHN_CALLBACK,
1095 xen_evtchn_do_upcall)
1096
5ead97c8
JF
1097#endif /* CONFIG_XEN */
1098
606576ce 1099#ifdef CONFIG_FUNCTION_TRACER
d61f82d0
SR
1100#ifdef CONFIG_DYNAMIC_FTRACE
1101
1102ENTRY(mcount)
d61f82d0
SR
1103 ret
1104END(mcount)
1105
1106ENTRY(ftrace_caller)
60a7ecf4
SR
1107 cmpl $0, function_trace_stop
1108 jne ftrace_stub
1109
d61f82d0
SR
1110 pushl %eax
1111 pushl %ecx
1112 pushl %edx
08f6fba5
SR
1113 pushl $0 /* Pass NULL as regs pointer */
1114 movl 4*4(%esp), %eax
d61f82d0 1115 movl 0x4(%ebp), %edx
28fb5dfa 1116 leal function_trace_op, %ecx
395a59d0 1117 subl $MCOUNT_INSN_SIZE, %eax
d61f82d0
SR
1118
1119.globl ftrace_call
1120ftrace_call:
1121 call ftrace_stub
1122
08f6fba5 1123 addl $4,%esp /* skip NULL pointer */
d61f82d0
SR
1124 popl %edx
1125 popl %ecx
1126 popl %eax
4de72395 1127ftrace_ret:
5a45cfe1
SR
1128#ifdef CONFIG_FUNCTION_GRAPH_TRACER
1129.globl ftrace_graph_call
1130ftrace_graph_call:
1131 jmp ftrace_stub
1132#endif
d61f82d0
SR
1133
1134.globl ftrace_stub
1135ftrace_stub:
1136 ret
1137END(ftrace_caller)
1138
4de72395
SR
1139ENTRY(ftrace_regs_caller)
1140 pushf /* push flags before compare (in cs location) */
1141 cmpl $0, function_trace_stop
1142 jne ftrace_restore_flags
1143
1144 /*
1145 * i386 does not save SS and ESP when coming from kernel.
1146 * Instead, to get sp, &regs->sp is used (see ptrace.h).
1147 * Unfortunately, that means eflags must be at the same location
1148 * as the current return ip is. We move the return ip into the
1149 * ip location, and move flags into the return ip location.
1150 */
1151 pushl 4(%esp) /* save return ip into ip slot */
4de72395
SR
1152
1153 pushl $0 /* Load 0 into orig_ax */
1154 pushl %gs
1155 pushl %fs
1156 pushl %es
1157 pushl %ds
1158 pushl %eax
1159 pushl %ebp
1160 pushl %edi
1161 pushl %esi
1162 pushl %edx
1163 pushl %ecx
1164 pushl %ebx
1165
1166 movl 13*4(%esp), %eax /* Get the saved flags */
1167 movl %eax, 14*4(%esp) /* Move saved flags into regs->flags location */
1168 /* clobbering return ip */
1169 movl $__KERNEL_CS,13*4(%esp)
1170
1171 movl 12*4(%esp), %eax /* Load ip (1st parameter) */
a5e37863 1172 subl $MCOUNT_INSN_SIZE, %eax /* Adjust ip */
e4ea3f6b 1173 movl 0x4(%ebp), %edx /* Load parent ip (2nd parameter) */
4de72395 1174 leal function_trace_op, %ecx /* Save ftrace_pos in 3rd parameter */
e4ea3f6b 1175 pushl %esp /* Save pt_regs as 4th parameter */
4de72395
SR
1176
1177GLOBAL(ftrace_regs_call)
1178 call ftrace_stub
1179
1180 addl $4, %esp /* Skip pt_regs */
1181 movl 14*4(%esp), %eax /* Move flags back into cs */
1182 movl %eax, 13*4(%esp) /* Needed to keep addl from modifying flags */
1183 movl 12*4(%esp), %eax /* Get return ip from regs->ip */
4de72395
SR
1184 movl %eax, 14*4(%esp) /* Put return ip back for ret */
1185
1186 popl %ebx
1187 popl %ecx
1188 popl %edx
1189 popl %esi
1190 popl %edi
1191 popl %ebp
1192 popl %eax
1193 popl %ds
1194 popl %es
1195 popl %fs
1196 popl %gs
1197 addl $8, %esp /* Skip orig_ax and ip */
1198 popf /* Pop flags at end (no addl to corrupt flags) */
1199 jmp ftrace_ret
1200
1201ftrace_restore_flags:
1202 popf
1203 jmp ftrace_stub
d61f82d0
SR
1204#else /* ! CONFIG_DYNAMIC_FTRACE */
1205
16444a8a 1206ENTRY(mcount)
60a7ecf4
SR
1207 cmpl $0, function_trace_stop
1208 jne ftrace_stub
1209
16444a8a
ACM
1210 cmpl $ftrace_stub, ftrace_trace_function
1211 jnz trace
fb52607a 1212#ifdef CONFIG_FUNCTION_GRAPH_TRACER
c2324b69 1213 cmpl $ftrace_stub, ftrace_graph_return
fb52607a 1214 jnz ftrace_graph_caller
e49dc19c
SR
1215
1216 cmpl $ftrace_graph_entry_stub, ftrace_graph_entry
1217 jnz ftrace_graph_caller
caf4b323 1218#endif
16444a8a
ACM
1219.globl ftrace_stub
1220ftrace_stub:
1221 ret
1222
1223 /* taken from glibc */
1224trace:
1225 pushl %eax
1226 pushl %ecx
1227 pushl %edx
1228 movl 0xc(%esp), %eax
1229 movl 0x4(%ebp), %edx
395a59d0 1230 subl $MCOUNT_INSN_SIZE, %eax
16444a8a 1231
d61f82d0 1232 call *ftrace_trace_function
16444a8a
ACM
1233
1234 popl %edx
1235 popl %ecx
1236 popl %eax
16444a8a
ACM
1237 jmp ftrace_stub
1238END(mcount)
d61f82d0 1239#endif /* CONFIG_DYNAMIC_FTRACE */
606576ce 1240#endif /* CONFIG_FUNCTION_TRACER */
16444a8a 1241
fb52607a
FW
1242#ifdef CONFIG_FUNCTION_GRAPH_TRACER
1243ENTRY(ftrace_graph_caller)
caf4b323
FW
1244 pushl %eax
1245 pushl %ecx
1246 pushl %edx
1dc1c6ad 1247 movl 0xc(%esp), %edx
caf4b323 1248 lea 0x4(%ebp), %eax
71e308a2 1249 movl (%ebp), %ecx
bb4304c7 1250 subl $MCOUNT_INSN_SIZE, %edx
caf4b323 1251 call prepare_ftrace_return
caf4b323
FW
1252 popl %edx
1253 popl %ecx
1254 popl %eax
e7d3737e 1255 ret
fb52607a 1256END(ftrace_graph_caller)
caf4b323
FW
1257
1258.globl return_to_handler
1259return_to_handler:
caf4b323 1260 pushl %eax
caf4b323 1261 pushl %edx
71e308a2 1262 movl %ebp, %eax
caf4b323 1263 call ftrace_return_to_handler
194ec341 1264 movl %eax, %ecx
caf4b323 1265 popl %edx
caf4b323 1266 popl %eax
194ec341 1267 jmp *%ecx
e7d3737e 1268#endif
16444a8a 1269
d211af05
AH
1270/*
1271 * Some functions should be protected against kprobes
1272 */
1273 .pushsection .kprobes.text, "ax"
1274
1275ENTRY(page_fault)
1276 RING0_EC_FRAME
e59d1b0a 1277 ASM_CLAC
df5d1874 1278 pushl_cfi $do_page_fault
d211af05
AH
1279 ALIGN
1280error_code:
ccbeed3a 1281 /* the function address is in %gs's slot on the stack */
df5d1874 1282 pushl_cfi %fs
ccbeed3a 1283 /*CFI_REL_OFFSET fs, 0*/
df5d1874 1284 pushl_cfi %es
d211af05 1285 /*CFI_REL_OFFSET es, 0*/
df5d1874 1286 pushl_cfi %ds
d211af05 1287 /*CFI_REL_OFFSET ds, 0*/
df5d1874 1288 pushl_cfi %eax
d211af05 1289 CFI_REL_OFFSET eax, 0
df5d1874 1290 pushl_cfi %ebp
d211af05 1291 CFI_REL_OFFSET ebp, 0
df5d1874 1292 pushl_cfi %edi
d211af05 1293 CFI_REL_OFFSET edi, 0
df5d1874 1294 pushl_cfi %esi
d211af05 1295 CFI_REL_OFFSET esi, 0
df5d1874 1296 pushl_cfi %edx
d211af05 1297 CFI_REL_OFFSET edx, 0
df5d1874 1298 pushl_cfi %ecx
d211af05 1299 CFI_REL_OFFSET ecx, 0
df5d1874 1300 pushl_cfi %ebx
d211af05
AH
1301 CFI_REL_OFFSET ebx, 0
1302 cld
d211af05
AH
1303 movl $(__KERNEL_PERCPU), %ecx
1304 movl %ecx, %fs
1305 UNWIND_ESPFIX_STACK
ccbeed3a
TH
1306 GS_TO_REG %ecx
1307 movl PT_GS(%esp), %edi # get the function address
d211af05
AH
1308 movl PT_ORIG_EAX(%esp), %edx # get the error code
1309 movl $-1, PT_ORIG_EAX(%esp) # no syscall to restart
ccbeed3a
TH
1310 REG_TO_PTGS %ecx
1311 SET_KERNEL_GS %ecx
d211af05
AH
1312 movl $(__USER_DS), %ecx
1313 movl %ecx, %ds
1314 movl %ecx, %es
1315 TRACE_IRQS_OFF
1316 movl %esp,%eax # pt_regs pointer
1317 call *%edi
1318 jmp ret_from_exception
1319 CFI_ENDPROC
1320END(page_fault)
1321
1322/*
1323 * Debug traps and NMI can happen at the one SYSENTER instruction
1324 * that sets up the real kernel stack. Check here, since we can't
1325 * allow the wrong stack to be used.
1326 *
1327 * "TSS_sysenter_sp0+12" is because the NMI/debug handler will have
1328 * already pushed 3 words if it hits on the sysenter instruction:
1329 * eflags, cs and eip.
1330 *
1331 * We just load the right stack, and push the three (known) values
1332 * by hand onto the new stack - while updating the return eip past
1333 * the instruction that would have done it for sysenter.
1334 */
f0d96110
TH
1335.macro FIX_STACK offset ok label
1336 cmpw $__KERNEL_CS, 4(%esp)
1337 jne \ok
1338\label:
1339 movl TSS_sysenter_sp0 + \offset(%esp), %esp
1340 CFI_DEF_CFA esp, 0
1341 CFI_UNDEFINED eip
df5d1874
JB
1342 pushfl_cfi
1343 pushl_cfi $__KERNEL_CS
1344 pushl_cfi $sysenter_past_esp
d211af05 1345 CFI_REL_OFFSET eip, 0
f0d96110 1346.endm
d211af05
AH
1347
1348ENTRY(debug)
1349 RING0_INT_FRAME
e59d1b0a 1350 ASM_CLAC
d211af05
AH
1351 cmpl $ia32_sysenter_target,(%esp)
1352 jne debug_stack_correct
f0d96110 1353 FIX_STACK 12, debug_stack_correct, debug_esp_fix_insn
d211af05 1354debug_stack_correct:
df5d1874 1355 pushl_cfi $-1 # mark this as an int
d211af05
AH
1356 SAVE_ALL
1357 TRACE_IRQS_OFF
1358 xorl %edx,%edx # error code 0
1359 movl %esp,%eax # pt_regs pointer
1360 call do_debug
1361 jmp ret_from_exception
1362 CFI_ENDPROC
1363END(debug)
1364
1365/*
1366 * NMI is doubly nasty. It can happen _while_ we're handling
1367 * a debug fault, and the debug fault hasn't yet been able to
1368 * clear up the stack. So we first check whether we got an
1369 * NMI on the sysenter entry path, but after that we need to
1370 * check whether we got an NMI on the debug path where the debug
1371 * fault happened on the sysenter path.
1372 */
1373ENTRY(nmi)
1374 RING0_INT_FRAME
e59d1b0a 1375 ASM_CLAC
df5d1874 1376 pushl_cfi %eax
d211af05
AH
1377 movl %ss, %eax
1378 cmpw $__ESPFIX_SS, %ax
df5d1874 1379 popl_cfi %eax
d211af05
AH
1380 je nmi_espfix_stack
1381 cmpl $ia32_sysenter_target,(%esp)
1382 je nmi_stack_fixup
df5d1874 1383 pushl_cfi %eax
d211af05
AH
1384 movl %esp,%eax
1385 /* Do not access memory above the end of our stack page,
1386 * it might not exist.
1387 */
1388 andl $(THREAD_SIZE-1),%eax
1389 cmpl $(THREAD_SIZE-20),%eax
df5d1874 1390 popl_cfi %eax
d211af05
AH
1391 jae nmi_stack_correct
1392 cmpl $ia32_sysenter_target,12(%esp)
1393 je nmi_debug_stack_check
1394nmi_stack_correct:
1395 /* We have a RING0_INT_FRAME here */
df5d1874 1396 pushl_cfi %eax
d211af05 1397 SAVE_ALL
d211af05
AH
1398 xorl %edx,%edx # zero error code
1399 movl %esp,%eax # pt_regs pointer
1400 call do_nmi
2e04bc76 1401 jmp restore_all_notrace
d211af05
AH
1402 CFI_ENDPROC
1403
1404nmi_stack_fixup:
1405 RING0_INT_FRAME
f0d96110 1406 FIX_STACK 12, nmi_stack_correct, 1
d211af05
AH
1407 jmp nmi_stack_correct
1408
1409nmi_debug_stack_check:
1410 /* We have a RING0_INT_FRAME here */
1411 cmpw $__KERNEL_CS,16(%esp)
1412 jne nmi_stack_correct
1413 cmpl $debug,(%esp)
1414 jb nmi_stack_correct
1415 cmpl $debug_esp_fix_insn,(%esp)
1416 ja nmi_stack_correct
f0d96110 1417 FIX_STACK 24, nmi_stack_correct, 1
d211af05
AH
1418 jmp nmi_stack_correct
1419
1420nmi_espfix_stack:
1421 /* We have a RING0_INT_FRAME here.
1422 *
1423 * create the pointer to lss back
1424 */
df5d1874
JB
1425 pushl_cfi %ss
1426 pushl_cfi %esp
bda3a897 1427 addl $4, (%esp)
d211af05
AH
1428 /* copy the iret frame of 12 bytes */
1429 .rept 3
df5d1874 1430 pushl_cfi 16(%esp)
d211af05 1431 .endr
df5d1874 1432 pushl_cfi %eax
d211af05 1433 SAVE_ALL
d211af05
AH
1434 FIXUP_ESPFIX_STACK # %eax == %esp
1435 xorl %edx,%edx # zero error code
1436 call do_nmi
1437 RESTORE_REGS
1438 lss 12+4(%esp), %esp # back to espfix stack
1439 CFI_ADJUST_CFA_OFFSET -24
1440 jmp irq_return
1441 CFI_ENDPROC
1442END(nmi)
1443
1444ENTRY(int3)
1445 RING0_INT_FRAME
e59d1b0a 1446 ASM_CLAC
df5d1874 1447 pushl_cfi $-1 # mark this as an int
d211af05
AH
1448 SAVE_ALL
1449 TRACE_IRQS_OFF
1450 xorl %edx,%edx # zero error code
1451 movl %esp,%eax # pt_regs pointer
1452 call do_int3
1453 jmp ret_from_exception
1454 CFI_ENDPROC
1455END(int3)
1456
1457ENTRY(general_protection)
1458 RING0_EC_FRAME
df5d1874 1459 pushl_cfi $do_general_protection
d211af05
AH
1460 jmp error_code
1461 CFI_ENDPROC
1462END(general_protection)
1463
631bc487
GN
1464#ifdef CONFIG_KVM_GUEST
1465ENTRY(async_page_fault)
1466 RING0_EC_FRAME
e59d1b0a 1467 ASM_CLAC
60cf637a 1468 pushl_cfi $do_async_page_fault
631bc487
GN
1469 jmp error_code
1470 CFI_ENDPROC
2ae9d293 1471END(async_page_fault)
631bc487
GN
1472#endif
1473
d211af05
AH
1474/*
1475 * End of kprobes section
1476 */
1477 .popsection