2 * linux/arch/i386/entry.S
4 * Copyright (C) 1991, 1992 Linus Torvalds
8 * entry.S contains the system-call and fault low-level handling routines.
9 * This also contains the timer-interrupt handler, as well as all interrupts
10 * and faults that can result in a task-switch.
12 * NOTE: This code handles signal-recognition, which happens every time
13 * after a timer-interrupt and after each system call.
15 * I changed all the .align's to 4 (16 byte alignment), as that's faster
18 * Stack layout in 'ret_from_system_call':
19 * ptrace needs to have all regs on the stack.
20 * if the order here is changed, it needs to be
21 * updated in fork.c:copy_process, signal.c:do_signal,
22 * ptrace.c and ptrace.h
40 * "current" is in register %ebx during any slow entries.
43 #include <linux/config.h>
44 #include <linux/linkage.h>
45 #include <asm/thread_info.h>
46 #include <asm/errno.h>
47 #include <asm/segment.h>
51 #include "irq_vectors.h"
53 #define nr_syscalls ((syscall_table_size)/4)
79 #define preempt_stop cli
82 #define resume_kernel restore_nocheck
96 movl $(__USER_DS), %edx; \
100 #define RESTORE_INT_REGS \
109 #define RESTORE_REGS \
113 .section .fixup,"ax"; \
119 .section __ex_table,"a";\
129 GET_THREAD_INFO(%ebp)
134 * Return to user mode is not as complex as all this looks,
135 * but we want the default path for a system call return to
136 * go as quickly as possible which is why some of this is
137 * less clear than it otherwise should be.
140 # userspace resumption stub bypassing syscall exit tracing
145 GET_THREAD_INFO(%ebp)
146 movl EFLAGS(%esp), %eax # mix EFLAGS and CS
148 testl $(VM_MASK | 3), %eax
150 ENTRY(resume_userspace)
151 cli # make sure we don't miss an interrupt
152 # setting need_resched or sigpending
153 # between sampling and the iret
154 movl TI_flags(%ebp), %ecx
155 andl $_TIF_WORK_MASK, %ecx # is there any work to be done on
156 # int/exception return?
160 #ifdef CONFIG_PREEMPT
163 cmpl $0,TI_preempt_count(%ebp) # non-zero preempt_count ?
166 movl TI_flags(%ebp), %ecx # need_resched set ?
167 testb $_TIF_NEED_RESCHED, %cl
169 testl $IF_MASK,EFLAGS(%esp) # interrupts off (exception path) ?
171 call preempt_schedule_irq
175 /* SYSENTER_RETURN points to after the "sysenter" instruction in
176 the vsyscall page. See vsyscall-sysentry.S, which defines the symbol. */
178 # sysenter call handler stub
179 ENTRY(sysenter_entry)
180 movl TSS_sysenter_esp0(%esp),%esp
187 pushl $SYSENTER_RETURN
190 * Load the potential sixth argument from user stack.
191 * Careful about security.
193 cmpl $__PAGE_OFFSET-3,%ebp
196 .section __ex_table,"a"
198 .long 1b,syscall_fault
203 GET_THREAD_INFO(%ebp)
205 /* Note, _TIF_SECCOMP is bit number 8, and so it needs testw and not testb */
206 testw $(_TIF_SYSCALL_TRACE|_TIF_SYSCALL_AUDIT|_TIF_SECCOMP),TI_flags(%ebp)
207 jnz syscall_trace_entry
208 cmpl $(nr_syscalls), %eax
210 call *sys_call_table(,%eax,4)
213 movl TI_flags(%ebp), %ecx
214 testw $_TIF_ALLWORK_MASK, %cx
215 jne syscall_exit_work
216 /* if something modifies registers it must also disable sysexit */
218 movl OLDESP(%esp), %ecx
224 # system call handler stub
226 pushl %eax # save orig_eax
228 GET_THREAD_INFO(%ebp)
229 # system call tracing in operation
230 /* Note, _TIF_SECCOMP is bit number 8, and so it needs testw and not testb */
231 testw $(_TIF_SYSCALL_TRACE|_TIF_SYSCALL_AUDIT|_TIF_SECCOMP),TI_flags(%ebp)
232 jnz syscall_trace_entry
233 cmpl $(nr_syscalls), %eax
236 call *sys_call_table(,%eax,4)
237 movl %eax,EAX(%esp) # store the return value
239 cli # make sure we don't miss an interrupt
240 # setting need_resched or sigpending
241 # between sampling and the iret
242 movl TI_flags(%ebp), %ecx
243 testw $_TIF_ALLWORK_MASK, %cx # current->work
244 jne syscall_exit_work
247 movl EFLAGS(%esp), %eax # mix EFLAGS, SS and CS
248 # Warning: OLDSS(%esp) contains the wrong/random values if we
249 # are returning to the kernel.
250 # See comments in process.c:copy_thread() for details.
251 movb OLDSS(%esp), %ah
253 andl $(VM_MASK | (4 << 8) | 3), %eax
254 cmpl $((4 << 8) | 3), %eax
255 je ldt_ss # returning to user-space with LDT SS
263 movl $__USER_DS, %edx
269 .section __ex_table,"a"
275 larl OLDSS(%esp), %eax
277 testl $0x00400000, %eax # returning to 32bit stack?
278 jnz restore_nocheck # allright, normal return
279 /* If returning to userspace with 16bit stack,
280 * try to fix the higher word of ESP, as the CPU
282 * This is an "official" bug of all the x86-compatible
283 * CPUs, which we can try to work around to make
284 * dosemu and wine happy. */
285 subl $8, %esp # reserve space for switch16 pointer
288 /* Set up the 16bit stack frame with switch32 pointer on top,
289 * and a switch16 pointer on top of the current frame. */
290 call setup_x86_bogus_stack
292 lss 20+4(%esp), %esp # switch to 16bit stack
294 .section __ex_table,"a"
299 # perform work that needs to be done immediately before resumption
302 testb $_TIF_NEED_RESCHED, %cl
306 cli # make sure we don't miss an interrupt
307 # setting need_resched or sigpending
308 # between sampling and the iret
309 movl TI_flags(%ebp), %ecx
310 andl $_TIF_WORK_MASK, %ecx # is there any work to be done other
311 # than syscall tracing?
313 testb $_TIF_NEED_RESCHED, %cl
316 work_notifysig: # deal with pending signals and
317 # notify-resume requests
318 testl $VM_MASK, EFLAGS(%esp)
320 jne work_notifysig_v86 # returning to kernel-space or
323 call do_notify_resume
328 pushl %ecx # save ti_flags for do_notify_resume
329 call save_v86_state # %eax contains pt_regs pointer
333 call do_notify_resume
336 # perform syscall exit tracing
339 movl $-ENOSYS,EAX(%esp)
342 call do_syscall_trace
343 movl ORIG_EAX(%esp), %eax
344 cmpl $(nr_syscalls), %eax
348 # perform syscall exit tracing
351 testb $(_TIF_SYSCALL_TRACE|_TIF_SYSCALL_AUDIT|_TIF_SINGLESTEP), %cl
353 sti # could let do_syscall_trace() call
357 call do_syscall_trace
362 pushl %eax # save orig_eax
364 GET_THREAD_INFO(%ebp)
365 movl $-EFAULT,EAX(%esp)
370 movl $-ENOSYS,EAX(%esp)
373 #define FIXUP_ESPFIX_STACK \
375 /* switch to 32bit stack using the pointer on top of 16bit stack */ \
376 lss %ss:CPU_16BIT_STACK_SIZE-8, %esp; \
377 /* copy data from 16bit stack to 32bit stack */ \
378 call fixup_x86_bogus_stack; \
379 /* put ESP to the proper location */ \
381 #define UNWIND_ESPFIX_STACK \
384 /* see if on 16bit stack */ \
385 cmpw $__ESPFIX_SS, %ax; \
387 movl $__KERNEL_DS, %edx; \
390 /* switch to 32bit stack */ \
395 * Build the entry stubs and pointer table with
396 * some assembler magic.
403 ENTRY(irq_entries_start)
421 #define BUILD_INTERRUPT(name, nr) \
429 /* The include is where all of the SMP etc. interrupts come from */
430 #include "entry_arch.h"
433 pushl $0 # no error code
434 pushl $do_divide_error
451 movl ES(%esp), %edi # get the function address
452 movl ORIG_EAX(%esp), %edx # get the error code
453 movl %eax, ORIG_EAX(%esp)
455 movl $(__USER_DS), %ecx
458 movl %esp,%eax # pt_regs pointer
460 jmp ret_from_exception
462 ENTRY(coprocessor_error)
464 pushl $do_coprocessor_error
467 ENTRY(simd_coprocessor_error)
469 pushl $do_simd_coprocessor_error
472 ENTRY(device_not_available)
473 pushl $-1 # mark this as an int
476 testl $0x4, %eax # EM (math emulation bit)
477 jne device_not_available_emulate
479 call math_state_restore
480 jmp ret_from_exception
481 device_not_available_emulate:
482 pushl $0 # temporary storage for ORIG_EIP
485 jmp ret_from_exception
488 * Debug traps and NMI can happen at the one SYSENTER instruction
489 * that sets up the real kernel stack. Check here, since we can't
490 * allow the wrong stack to be used.
492 * "TSS_sysenter_esp0+12" is because the NMI/debug handler will have
493 * already pushed 3 words if it hits on the sysenter instruction:
494 * eflags, cs and eip.
496 * We just load the right stack, and push the three (known) values
497 * by hand onto the new stack - while updating the return eip past
498 * the instruction that would have done it for sysenter.
500 #define FIX_STACK(offset, ok, label) \
501 cmpw $__KERNEL_CS,4(%esp); \
504 movl TSS_sysenter_esp0+offset(%esp),%esp; \
506 pushl $__KERNEL_CS; \
507 pushl $sysenter_past_esp
510 cmpl $sysenter_entry,(%esp)
511 jne debug_stack_correct
512 FIX_STACK(12, debug_stack_correct, debug_esp_fix_insn)
514 pushl $-1 # mark this as an int
516 xorl %edx,%edx # error code 0
517 movl %esp,%eax # pt_regs pointer
521 jmp ret_from_exception
524 * NMI is doubly nasty. It can happen _while_ we're handling
525 * a debug fault, and the debug fault hasn't yet been able to
526 * clear up the stack. So we first check whether we got an
527 * NMI on the sysenter entry path, but after that we need to
528 * check whether we got an NMI on the debug path where the debug
529 * fault happened on the sysenter path.
534 cmpw $__ESPFIX_SS, %ax
537 cmpl $sysenter_entry,(%esp)
541 /* Do not access memory above the end of our stack page,
542 * it might not exist.
544 andl $(THREAD_SIZE-1),%eax
545 cmpl $(THREAD_SIZE-20),%eax
547 jae nmi_stack_correct
548 cmpl $sysenter_entry,12(%esp)
549 je nmi_debug_stack_check
553 xorl %edx,%edx # zero error code
554 movl %esp,%eax # pt_regs pointer
559 FIX_STACK(12,nmi_stack_correct, 1)
560 jmp nmi_stack_correct
561 nmi_debug_stack_check:
562 cmpw $__KERNEL_CS,16(%esp)
563 jne nmi_stack_correct
564 cmpl $debug - 1,(%esp)
565 jle nmi_stack_correct
566 cmpl $debug_esp_fix_insn,(%esp)
567 jle nmi_debug_stack_fixup
568 nmi_debug_stack_fixup:
569 FIX_STACK(24,nmi_stack_correct, 1)
570 jmp nmi_stack_correct
573 /* create the pointer to lss back */
578 /* copy the iret frame of 12 bytes */
584 FIXUP_ESPFIX_STACK # %eax == %esp
585 xorl %edx,%edx # zero error code
588 lss 12+4(%esp), %esp # back to 16bit stack
590 .section __ex_table,"a"
596 pushl $-1 # mark this as an int
598 xorl %edx,%edx # zero error code
599 movl %esp,%eax # pt_regs pointer
603 jmp ret_from_exception
620 ENTRY(coprocessor_segment_overrun)
622 pushl $do_coprocessor_segment_overrun
626 pushl $do_invalid_TSS
629 ENTRY(segment_not_present)
630 pushl $do_segment_not_present
634 pushl $do_stack_segment
637 ENTRY(general_protection)
638 pushl $do_general_protection
641 ENTRY(alignment_check)
642 pushl $do_alignment_check
649 #ifdef CONFIG_X86_MCE
652 pushl machine_check_vector
656 ENTRY(spurious_interrupt_bug)
658 pushl $do_spurious_interrupt_bug
662 ENTRY(sys_call_table)
663 .long sys_restart_syscall /* 0 - old "setup()" system call, used for restarting */
668 .long sys_open /* 5 */
673 .long sys_unlink /* 10 */
678 .long sys_chmod /* 15 */
680 .long sys_ni_syscall /* old break syscall holder */
683 .long sys_getpid /* 20 */
688 .long sys_stime /* 25 */
693 .long sys_utime /* 30 */
694 .long sys_ni_syscall /* old stty syscall holder */
695 .long sys_ni_syscall /* old gtty syscall holder */
698 .long sys_ni_syscall /* 35 - old ftime syscall holder */
703 .long sys_rmdir /* 40 */
707 .long sys_ni_syscall /* old prof syscall holder */
708 .long sys_brk /* 45 */
713 .long sys_getegid16 /* 50 */
715 .long sys_umount /* recycled never used phys() */
716 .long sys_ni_syscall /* old lock syscall holder */
718 .long sys_fcntl /* 55 */
719 .long sys_ni_syscall /* old mpx syscall holder */
721 .long sys_ni_syscall /* old ulimit syscall holder */
723 .long sys_umask /* 60 */
728 .long sys_getpgrp /* 65 */
733 .long sys_setreuid16 /* 70 */
737 .long sys_sethostname
738 .long sys_setrlimit /* 75 */
739 .long sys_old_getrlimit
741 .long sys_gettimeofday
742 .long sys_settimeofday
743 .long sys_getgroups16 /* 80 */
744 .long sys_setgroups16
748 .long sys_readlink /* 85 */
753 .long old_mmap /* 90 */
758 .long sys_fchown16 /* 95 */
759 .long sys_getpriority
760 .long sys_setpriority
761 .long sys_ni_syscall /* old profil syscall holder */
763 .long sys_fstatfs /* 100 */
768 .long sys_getitimer /* 105 */
773 .long sys_iopl /* 110 */
775 .long sys_ni_syscall /* old "idle" system call */
778 .long sys_swapoff /* 115 */
783 .long sys_clone /* 120 */
784 .long sys_setdomainname
788 .long sys_mprotect /* 125 */
789 .long sys_sigprocmask
790 .long sys_ni_syscall /* old "create_module" */
791 .long sys_init_module
792 .long sys_delete_module
793 .long sys_ni_syscall /* 130: old "get_kernel_syms" */
798 .long sys_sysfs /* 135 */
799 .long sys_personality
800 .long sys_ni_syscall /* reserved for afs_syscall */
803 .long sys_llseek /* 140 */
808 .long sys_readv /* 145 */
813 .long sys_mlock /* 150 */
817 .long sys_sched_setparam
818 .long sys_sched_getparam /* 155 */
819 .long sys_sched_setscheduler
820 .long sys_sched_getscheduler
821 .long sys_sched_yield
822 .long sys_sched_get_priority_max
823 .long sys_sched_get_priority_min /* 160 */
824 .long sys_sched_rr_get_interval
827 .long sys_setresuid16
828 .long sys_getresuid16 /* 165 */
830 .long sys_ni_syscall /* Old sys_query_module */
833 .long sys_setresgid16 /* 170 */
834 .long sys_getresgid16
836 .long sys_rt_sigreturn
837 .long sys_rt_sigaction
838 .long sys_rt_sigprocmask /* 175 */
839 .long sys_rt_sigpending
840 .long sys_rt_sigtimedwait
841 .long sys_rt_sigqueueinfo
842 .long sys_rt_sigsuspend
843 .long sys_pread64 /* 180 */
848 .long sys_capset /* 185 */
849 .long sys_sigaltstack
851 .long sys_ni_syscall /* reserved for streams1 */
852 .long sys_ni_syscall /* reserved for streams2 */
853 .long sys_vfork /* 190 */
857 .long sys_ftruncate64
858 .long sys_stat64 /* 195 */
863 .long sys_getgid /* 200 */
868 .long sys_getgroups /* 205 */
873 .long sys_setresgid /* 210 */
878 .long sys_setfsuid /* 215 */
883 .long sys_getdents64 /* 220 */
885 .long sys_ni_syscall /* reserved for TUX */
888 .long sys_readahead /* 225 */
893 .long sys_lgetxattr /* 230 */
898 .long sys_removexattr /* 235 */
899 .long sys_lremovexattr
900 .long sys_fremovexattr
903 .long sys_futex /* 240 */
904 .long sys_sched_setaffinity
905 .long sys_sched_getaffinity
906 .long sys_set_thread_area
907 .long sys_get_thread_area
908 .long sys_io_setup /* 245 */
910 .long sys_io_getevents
913 .long sys_fadvise64 /* 250 */
916 .long sys_lookup_dcookie
917 .long sys_epoll_create
918 .long sys_epoll_ctl /* 255 */
920 .long sys_remap_file_pages
921 .long sys_set_tid_address
922 .long sys_timer_create
923 .long sys_timer_settime /* 260 */
924 .long sys_timer_gettime
925 .long sys_timer_getoverrun
926 .long sys_timer_delete
927 .long sys_clock_settime
928 .long sys_clock_gettime /* 265 */
929 .long sys_clock_getres
930 .long sys_clock_nanosleep
933 .long sys_tgkill /* 270 */
935 .long sys_fadvise64_64
936 .long sys_ni_syscall /* sys_vserver */
938 .long sys_get_mempolicy
939 .long sys_set_mempolicy
942 .long sys_mq_timedsend
943 .long sys_mq_timedreceive /* 280 */
945 .long sys_mq_getsetattr
946 .long sys_ni_syscall /* reserved for kexec */
948 .long sys_ni_syscall /* 285 */ /* available */
950 .long sys_request_key
953 syscall_table_size=(.-sys_call_table)