2 * linux/arch/x86_64/entry.S
4 * Copyright (C) 1991, 1992 Linus Torvalds
5 * Copyright (C) 2000, 2001, 2002 Andi Kleen SuSE Labs
6 * Copyright (C) 2000 Pavel Machek <pavel@suse.cz>
10 * entry.S contains the system-call and fault low-level handling routines.
12 * NOTE: This code handles signal-recognition, which happens every time
13 * after an interrupt and after each system call.
15 * Normal syscalls and interrupts don't save a full stack frame, this is
16 * only done for syscall tracing, signals or fork/exec et.al.
18 * A note on terminology:
19 * - top of stack: Architecture defined interrupt frame from SS to RIP
20 * at the top of the kernel process stack.
21 * - partial stack frame: partially saved registers upto R11.
22 * - full stack frame: Like partial stack frame, but all register saved.
25 * - CFI macros are used to generate dwarf2 unwind information for better
26 * backtraces. They don't change any code.
27 * - SAVE_ALL/RESTORE_ALL - Save/restore all registers
28 * - SAVE_ARGS/RESTORE_ARGS - Save/restore registers that C functions modify.
29 * There are unfortunately lots of special cases where some registers
30 * not touched. The macro is a big mess that should be cleaned up.
31 * - SAVE_REST/RESTORE_REST - Handle the registers not saved by SAVE_ARGS.
32 * Gives a full stack frame.
33 * - ENTRY/END Define functions in the symbol table.
34 * - FIXUP_TOP_OF_STACK/RESTORE_TOP_OF_STACK - Fix up the hardware stack
35 * frame that is otherwise undefined after a SYSCALL
36 * - TRACE_IRQ_* - Trace hard interrupt state for lock debugging.
37 * - errorentry/paranoidentry/zeroentry - Define exception entry points.
40 #include <linux/linkage.h>
41 #include <asm/segment.h>
42 #include <asm/cache.h>
43 #include <asm/errno.h>
44 #include <asm/dwarf2.h>
45 #include <asm/calling.h>
46 #include <asm/asm-offsets.h>
48 #include <asm/unistd.h>
49 #include <asm/thread_info.h>
50 #include <asm/hw_irq.h>
52 #include <asm/irqflags.h>
56 #ifndef CONFIG_PREEMPT
57 #define retint_kernel retint_restore_args
61 .macro TRACE_IRQS_IRETQ offset=ARGOFFSET
62 #ifdef CONFIG_TRACE_IRQFLAGS
63 bt $9,EFLAGS-\offset(%rsp) /* interrupts off? */
71 * C code is not supposed to know about undefined top of stack. Every time
72 * a C function with an pt_regs argument is called from the SYSCALL based
73 * fast path FIXUP_TOP_OF_STACK is needed.
74 * RESTORE_TOP_OF_STACK syncs the syscall state after any possible ptregs
78 /* %rsp:at FRAMEEND */
79 .macro FIXUP_TOP_OF_STACK tmp
80 movq %gs:pda_oldrsp,\tmp
82 movq $__USER_DS,SS(%rsp)
83 movq $__USER_CS,CS(%rsp)
85 movq R11(%rsp),\tmp /* get eflags */
86 movq \tmp,EFLAGS(%rsp)
89 .macro RESTORE_TOP_OF_STACK tmp,offset=0
90 movq RSP-\offset(%rsp),\tmp
91 movq \tmp,%gs:pda_oldrsp
92 movq EFLAGS-\offset(%rsp),\tmp
93 movq \tmp,R11-\offset(%rsp)
96 .macro FAKE_STACK_FRAME child_rip
97 /* push in order ss, rsp, eflags, cs, rip */
100 CFI_ADJUST_CFA_OFFSET 8
101 /*CFI_REL_OFFSET ss,0*/
103 CFI_ADJUST_CFA_OFFSET 8
105 pushq $(1<<9) /* eflags - interrupts on */
106 CFI_ADJUST_CFA_OFFSET 8
107 /*CFI_REL_OFFSET rflags,0*/
108 pushq $__KERNEL_CS /* cs */
109 CFI_ADJUST_CFA_OFFSET 8
110 /*CFI_REL_OFFSET cs,0*/
111 pushq \child_rip /* rip */
112 CFI_ADJUST_CFA_OFFSET 8
114 pushq %rax /* orig rax */
115 CFI_ADJUST_CFA_OFFSET 8
118 .macro UNFAKE_STACK_FRAME
120 CFI_ADJUST_CFA_OFFSET -(6*8)
123 .macro CFI_DEFAULT_STACK start=1
128 CFI_DEF_CFA_OFFSET SS+8
130 CFI_REL_OFFSET r15,R15
131 CFI_REL_OFFSET r14,R14
132 CFI_REL_OFFSET r13,R13
133 CFI_REL_OFFSET r12,R12
134 CFI_REL_OFFSET rbp,RBP
135 CFI_REL_OFFSET rbx,RBX
136 CFI_REL_OFFSET r11,R11
137 CFI_REL_OFFSET r10,R10
140 CFI_REL_OFFSET rax,RAX
141 CFI_REL_OFFSET rcx,RCX
142 CFI_REL_OFFSET rdx,RDX
143 CFI_REL_OFFSET rsi,RSI
144 CFI_REL_OFFSET rdi,RDI
145 CFI_REL_OFFSET rip,RIP
146 /*CFI_REL_OFFSET cs,CS*/
147 /*CFI_REL_OFFSET rflags,EFLAGS*/
148 CFI_REL_OFFSET rsp,RSP
149 /*CFI_REL_OFFSET ss,SS*/
152 * A newly forked process directly context switches into this.
158 GET_THREAD_INFO(%rcx)
159 testl $(_TIF_SYSCALL_TRACE|_TIF_SYSCALL_AUDIT),threadinfo_flags(%rcx)
163 testl $3,CS-ARGOFFSET(%rsp) # from kernel_thread?
164 je int_ret_from_sys_call
165 testl $_TIF_IA32,threadinfo_flags(%rcx)
166 jnz int_ret_from_sys_call
167 RESTORE_TOP_OF_STACK %rdi,ARGOFFSET
168 jmp ret_from_sys_call
171 call syscall_trace_leave
172 GET_THREAD_INFO(%rcx)
178 * System call entry. Upto 6 arguments in registers are supported.
180 * SYSCALL does not save anything on the stack and does not change the
186 * rax system call number
188 * rcx return address for syscall/sysret, C arg3
191 * r10 arg3 (--> moved to rcx for C)
194 * r11 eflags for syscall/sysret, temporary for C
195 * r12-r15,rbp,rbx saved by C code, not touched.
197 * Interrupts are off on entry.
198 * Only called from user space.
200 * XXX if we had a free scratch register we could save the RSP into the stack frame
201 * and report it properly in ps. Unfortunately we haven't.
203 * When user can change the frames always force IRET. That is because
204 * it deals with uncanonical addresses better. SYSRET has trouble
205 * with them due to bugs in both AMD and Intel CPUs.
210 CFI_DEF_CFA rsp,PDA_STACKOFFSET
212 /*CFI_REGISTER rflags,r11*/
214 movq %rsp,%gs:pda_oldrsp
215 movq %gs:pda_kernelstack,%rsp
217 * No need to follow this irqs off/on section - it's straight
222 movq %rax,ORIG_RAX-ARGOFFSET(%rsp)
223 movq %rcx,RIP-ARGOFFSET(%rsp)
224 CFI_REL_OFFSET rip,RIP-ARGOFFSET
225 GET_THREAD_INFO(%rcx)
226 testl $(_TIF_SYSCALL_TRACE|_TIF_SYSCALL_AUDIT|_TIF_SECCOMP),threadinfo_flags(%rcx)
229 cmpq $__NR_syscall_max,%rax
232 call *sys_call_table(,%rax,8) # XXX: rip relative
233 movq %rax,RAX-ARGOFFSET(%rsp)
235 * Syscall return path ending with SYSRET (fast path)
236 * Has incomplete stack frame and undefined top of stack.
238 .globl ret_from_sys_call
240 movl $_TIF_ALLWORK_MASK,%edi
243 GET_THREAD_INFO(%rcx)
246 movl threadinfo_flags(%rcx),%edx
251 * sysretq will re-enable interrupts:
254 movq RIP-ARGOFFSET(%rsp),%rcx
256 RESTORE_ARGS 0,-ARG_SKIP,1
257 /*CFI_REGISTER rflags,r11*/
258 movq %gs:pda_oldrsp,%rsp
262 /* Handle reschedules */
263 /* edx: work, edi: workmask */
266 bt $TIF_NEED_RESCHED,%edx
271 CFI_ADJUST_CFA_OFFSET 8
274 CFI_ADJUST_CFA_OFFSET -8
277 /* Handle a signal */
281 testl $(_TIF_SIGPENDING|_TIF_NOTIFY_RESUME|_TIF_SINGLESTEP),%edx
284 /* Really a signal */
285 /* edx: work flags (arg3) */
286 leaq do_notify_resume(%rip),%rax
287 leaq -ARGOFFSET(%rsp),%rdi # &pt_regs -> arg1
288 xorl %esi,%esi # oldset -> arg2
289 call ptregscall_common
290 1: movl $_TIF_NEED_RESCHED,%edi
291 /* Use IRET because user could have changed frame. This
292 works because ptregscall_common has called FIXUP_TOP_OF_STACK. */
298 movq $-ENOSYS,RAX-ARGOFFSET(%rsp)
299 jmp ret_from_sys_call
301 /* Do syscall tracing */
305 movq $-ENOSYS,RAX(%rsp)
306 FIXUP_TOP_OF_STACK %rdi
308 call syscall_trace_enter
309 LOAD_ARGS ARGOFFSET /* reload args from stack in case ptrace changed it */
311 cmpq $__NR_syscall_max,%rax
313 movq %r10,%rcx /* fixup for C */
314 call *sys_call_table(,%rax,8)
315 1: movq %rax,RAX-ARGOFFSET(%rsp)
316 /* Use IRET because user could have changed frame */
317 jmp int_ret_from_sys_call
322 * Syscall return path ending with IRET.
323 * Has correct top of stack, but partial stack frame.
325 ENTRY(int_ret_from_sys_call)
327 CFI_DEF_CFA rsp,SS+8-ARGOFFSET
328 /*CFI_REL_OFFSET ss,SS-ARGOFFSET*/
329 CFI_REL_OFFSET rsp,RSP-ARGOFFSET
330 /*CFI_REL_OFFSET rflags,EFLAGS-ARGOFFSET*/
331 /*CFI_REL_OFFSET cs,CS-ARGOFFSET*/
332 CFI_REL_OFFSET rip,RIP-ARGOFFSET
333 CFI_REL_OFFSET rdx,RDX-ARGOFFSET
334 CFI_REL_OFFSET rcx,RCX-ARGOFFSET
335 CFI_REL_OFFSET rax,RAX-ARGOFFSET
336 CFI_REL_OFFSET rdi,RDI-ARGOFFSET
337 CFI_REL_OFFSET rsi,RSI-ARGOFFSET
338 CFI_REL_OFFSET r8,R8-ARGOFFSET
339 CFI_REL_OFFSET r9,R9-ARGOFFSET
340 CFI_REL_OFFSET r10,R10-ARGOFFSET
341 CFI_REL_OFFSET r11,R11-ARGOFFSET
344 testl $3,CS-ARGOFFSET(%rsp)
345 je retint_restore_args
346 movl $_TIF_ALLWORK_MASK,%edi
347 /* edi: mask to check */
349 GET_THREAD_INFO(%rcx)
350 movl threadinfo_flags(%rcx),%edx
353 andl $~TS_COMPAT,threadinfo_status(%rcx)
356 /* Either reschedule or signal or syscall exit tracking needed. */
357 /* First do a reschedule test. */
358 /* edx: work, edi: workmask */
360 bt $TIF_NEED_RESCHED,%edx
365 CFI_ADJUST_CFA_OFFSET 8
368 CFI_ADJUST_CFA_OFFSET -8
373 /* handle signals and tracing -- both require a full stack frame */
378 /* Check for syscall exit trace */
379 testl $(_TIF_SYSCALL_TRACE|_TIF_SYSCALL_AUDIT|_TIF_SINGLESTEP),%edx
382 CFI_ADJUST_CFA_OFFSET 8
383 leaq 8(%rsp),%rdi # &ptregs -> arg1
384 call syscall_trace_leave
386 CFI_ADJUST_CFA_OFFSET -8
387 andl $~(_TIF_SYSCALL_TRACE|_TIF_SYSCALL_AUDIT|_TIF_SINGLESTEP),%edi
393 testl $(_TIF_NOTIFY_RESUME|_TIF_SIGPENDING|_TIF_SINGLESTEP),%edx
395 movq %rsp,%rdi # &ptregs -> arg1
396 xorl %esi,%esi # oldset -> arg2
397 call do_notify_resume
398 1: movl $_TIF_NEED_RESCHED,%edi
405 END(int_ret_from_sys_call)
408 * Certain special system calls that need to save a complete full stack frame.
411 .macro PTREGSCALL label,func,arg
414 leaq \func(%rip),%rax
415 leaq -ARGOFFSET+8(%rsp),\arg /* 8 for return address */
416 jmp ptregscall_common
422 PTREGSCALL stub_clone, sys_clone, %r8
423 PTREGSCALL stub_fork, sys_fork, %rdi
424 PTREGSCALL stub_vfork, sys_vfork, %rdi
425 PTREGSCALL stub_rt_sigsuspend, sys_rt_sigsuspend, %rdx
426 PTREGSCALL stub_sigaltstack, sys_sigaltstack, %rdx
427 PTREGSCALL stub_iopl, sys_iopl, %rsi
429 ENTRY(ptregscall_common)
431 CFI_ADJUST_CFA_OFFSET -8
432 CFI_REGISTER rip, r11
435 CFI_REGISTER rip, r15
436 FIXUP_TOP_OF_STACK %r11
438 RESTORE_TOP_OF_STACK %r11
440 CFI_REGISTER rip, r11
443 CFI_ADJUST_CFA_OFFSET 8
444 CFI_REL_OFFSET rip, 0
447 END(ptregscall_common)
452 CFI_ADJUST_CFA_OFFSET -8
453 CFI_REGISTER rip, r11
455 FIXUP_TOP_OF_STACK %r11
457 RESTORE_TOP_OF_STACK %r11
460 jmp int_ret_from_sys_call
465 * sigreturn is special because it needs to restore all registers on return.
466 * This cannot be done with SYSRET, so use the IRET return path instead.
468 ENTRY(stub_rt_sigreturn)
471 CFI_ADJUST_CFA_OFFSET -8
474 FIXUP_TOP_OF_STACK %r11
475 call sys_rt_sigreturn
476 movq %rax,RAX(%rsp) # fixme, this could be done at the higher layer
478 jmp int_ret_from_sys_call
480 END(stub_rt_sigreturn)
483 * initial frame state for interrupts and exceptions
487 CFI_DEF_CFA rsp,SS+8-\ref
488 /*CFI_REL_OFFSET ss,SS-\ref*/
489 CFI_REL_OFFSET rsp,RSP-\ref
490 /*CFI_REL_OFFSET rflags,EFLAGS-\ref*/
491 /*CFI_REL_OFFSET cs,CS-\ref*/
492 CFI_REL_OFFSET rip,RIP-\ref
495 /* initial frame state for interrupts (and exceptions without error code) */
496 #define INTR_FRAME _frame RIP
497 /* initial frame state for exceptions with error code (and interrupts with
498 vector already pushed) */
499 #define XCPT_FRAME _frame ORIG_RAX
502 * Interrupt entry/exit.
504 * Interrupt entry points save only callee clobbered registers in fast path.
506 * Entry runs with interrupts off.
509 /* 0(%rsp): interrupt number */
510 .macro interrupt func
513 leaq -ARGOFFSET(%rsp),%rdi # arg1 for handler
515 CFI_ADJUST_CFA_OFFSET 8
516 CFI_REL_OFFSET rbp, 0
518 CFI_DEF_CFA_REGISTER rbp
522 /* irqcount is used to check if a CPU is already on an interrupt
523 stack or not. While this is essentially redundant with preempt_count
524 it is a little cheaper to use a separate counter in the PDA
525 (short of moving irq_enter into assembly, which would be too
527 1: incl %gs:pda_irqcount
528 cmoveq %gs:pda_irqstackptr,%rsp
529 push %rbp # backlink for old unwinder
531 * We entered an interrupt context - irqs are off:
537 ENTRY(common_interrupt)
540 /* 0(%rsp): oldrsp-ARGOFFSET */
544 decl %gs:pda_irqcount
546 CFI_DEF_CFA_REGISTER rsp
547 CFI_ADJUST_CFA_OFFSET -8
549 GET_THREAD_INFO(%rcx)
550 testl $3,CS-ARGOFFSET(%rsp)
553 /* Interrupt came from user space */
555 * Has a correct top of stack, but a partial stack frame
556 * %rcx: thread info. Interrupts off.
558 retint_with_reschedule:
559 movl $_TIF_WORK_MASK,%edi
561 movl threadinfo_flags(%rcx),%edx
567 * The iretq could re-enable interrupts:
577 * The iretq could re-enable interrupts:
585 .section __ex_table,"a"
586 .quad iret_label,bad_iret
589 /* force a signal here? this matches i386 behaviour */
590 /* running with kernel gs */
592 movq $11,%rdi /* SIGSEGV */
598 /* edi: workmask, edx: work */
601 bt $TIF_NEED_RESCHED,%edx
606 CFI_ADJUST_CFA_OFFSET 8
609 CFI_ADJUST_CFA_OFFSET -8
610 GET_THREAD_INFO(%rcx)
616 testl $(_TIF_SIGPENDING|_TIF_NOTIFY_RESUME|_TIF_SINGLESTEP),%edx
621 movq $-1,ORIG_RAX(%rsp)
622 xorl %esi,%esi # oldset
623 movq %rsp,%rdi # &pt_regs
624 call do_notify_resume
628 movl $_TIF_NEED_RESCHED,%edi
629 GET_THREAD_INFO(%rcx)
632 #ifdef CONFIG_PREEMPT
633 /* Returning to kernel space. Check if we need preemption */
634 /* rcx: threadinfo. interrupts off. */
636 cmpl $0,threadinfo_preempt_count(%rcx)
637 jnz retint_restore_args
638 bt $TIF_NEED_RESCHED,threadinfo_flags(%rcx)
639 jnc retint_restore_args
640 bt $9,EFLAGS-ARGOFFSET(%rsp) /* interrupts off? */
641 jnc retint_restore_args
642 call preempt_schedule_irq
647 END(common_interrupt)
652 .macro apicinterrupt num,func
655 CFI_ADJUST_CFA_OFFSET 8
661 ENTRY(thermal_interrupt)
662 apicinterrupt THERMAL_APIC_VECTOR,smp_thermal_interrupt
663 END(thermal_interrupt)
665 ENTRY(threshold_interrupt)
666 apicinterrupt THRESHOLD_APIC_VECTOR,mce_threshold_interrupt
667 END(threshold_interrupt)
670 ENTRY(reschedule_interrupt)
671 apicinterrupt RESCHEDULE_VECTOR,smp_reschedule_interrupt
672 END(reschedule_interrupt)
674 .macro INVALIDATE_ENTRY num
675 ENTRY(invalidate_interrupt\num)
676 apicinterrupt INVALIDATE_TLB_VECTOR_START+\num,smp_invalidate_interrupt
677 END(invalidate_interrupt\num)
689 ENTRY(call_function_interrupt)
690 apicinterrupt CALL_FUNCTION_VECTOR,smp_call_function_interrupt
691 END(call_function_interrupt)
694 ENTRY(apic_timer_interrupt)
695 apicinterrupt LOCAL_TIMER_VECTOR,smp_apic_timer_interrupt
696 END(apic_timer_interrupt)
698 ENTRY(error_interrupt)
699 apicinterrupt ERROR_APIC_VECTOR,smp_error_interrupt
702 ENTRY(spurious_interrupt)
703 apicinterrupt SPURIOUS_APIC_VECTOR,smp_spurious_interrupt
704 END(spurious_interrupt)
707 * Exception entry points.
711 pushq $0 /* push error code/oldrax */
712 CFI_ADJUST_CFA_OFFSET 8
713 pushq %rax /* push real oldrax to the rdi slot */
714 CFI_ADJUST_CFA_OFFSET 8
720 .macro errorentry sym
723 CFI_ADJUST_CFA_OFFSET 8
729 /* error code is on the stack already */
730 /* handle NMI like exceptions that can happen everywhere */
731 .macro paranoidentry sym, ist=0, irqtrace=1
735 movl $MSR_GS_BASE,%ecx
743 movq %gs:pda_data_offset, %rbp
746 movq ORIG_RAX(%rsp),%rsi
747 movq $-1,ORIG_RAX(%rsp)
749 subq $EXCEPTION_STKSZ, per_cpu__init_tss + TSS_ist + (\ist - 1) * 8(%rbp)
753 addq $EXCEPTION_STKSZ, per_cpu__init_tss + TSS_ist + (\ist - 1) * 8(%rbp)
762 * "Paranoid" exit path from exception stack.
763 * Paranoid because this is used by NMIs and cannot take
764 * any kernel state for granted.
765 * We don't do kernel preemption checks here, because only
766 * NMI should be common and it does not enable IRQs and
767 * cannot get reschedule ticks.
769 * "trace" is 0 for the NMI handler only, because irq-tracing
770 * is fundamentally NMI-unsafe. (we cannot change the soft and
771 * hard flags at once, atomically)
773 .macro paranoidexit trace=1
774 /* ebx: no swapgs flag */
776 testl %ebx,%ebx /* swapgs needed? */
777 jnz paranoid_restore\trace
779 jnz paranoid_userspace\trace
780 paranoid_swapgs\trace:
785 paranoid_restore\trace:
788 paranoid_userspace\trace:
789 GET_THREAD_INFO(%rcx)
790 movl threadinfo_flags(%rcx),%ebx
791 andl $_TIF_WORK_MASK,%ebx
792 jz paranoid_swapgs\trace
793 movq %rsp,%rdi /* &pt_regs */
795 movq %rax,%rsp /* switch stack for scheduling */
796 testl $_TIF_NEED_RESCHED,%ebx
797 jnz paranoid_schedule\trace
798 movl %ebx,%edx /* arg3: thread flags */
803 xorl %esi,%esi /* arg2: oldset */
804 movq %rsp,%rdi /* arg1: &pt_regs */
805 call do_notify_resume
810 jmp paranoid_userspace\trace
811 paranoid_schedule\trace:
821 jmp paranoid_userspace\trace
826 * Exception entry point. This expects an error code/orig_rax on the stack
827 * and the exception handler in %rax.
829 KPROBE_ENTRY(error_entry)
831 /* rdi slot contains rax, oldrax contains error code */
834 CFI_ADJUST_CFA_OFFSET (14*8)
836 CFI_REL_OFFSET rsi,RSI
837 movq 14*8(%rsp),%rsi /* load rax from rdi slot */
839 CFI_REL_OFFSET rdx,RDX
841 CFI_REL_OFFSET rcx,RCX
842 movq %rsi,10*8(%rsp) /* store rax */
843 CFI_REL_OFFSET rax,RAX
849 CFI_REL_OFFSET r10,R10
851 CFI_REL_OFFSET r11,R11
853 CFI_REL_OFFSET rbx,RBX
855 CFI_REL_OFFSET rbp,RBP
857 CFI_REL_OFFSET r12,R12
859 CFI_REL_OFFSET r13,R13
861 CFI_REL_OFFSET r14,R14
863 CFI_REL_OFFSET r15,R15
872 movq ORIG_RAX(%rsp),%rsi /* get error code */
873 movq $-1,ORIG_RAX(%rsp)
875 /* ebx: no swapgs flag (1: don't need swapgs, 0: need it) */
881 GET_THREAD_INFO(%rcx)
884 movl threadinfo_flags(%rcx),%edx
885 movl $_TIF_WORK_MASK,%edi
889 * The iret might restore flags:
899 /* There are two places in the kernel that can potentially fault with
900 usergs. Handle them here. The exception handlers after
901 iret run with kernel gs again, so don't set the user space flag.
902 B stepping K8s sometimes report an truncated RIP for IRET
903 exceptions returning to compat mode. Check for these here too. */
904 leaq iret_label(%rip),%rbp
907 movl %ebp,%ebp /* zero extend */
910 cmpq $gs_change,RIP(%rsp)
913 KPROBE_END(error_entry)
915 /* Reload gs selector with exception handling */
916 /* edi: new selector */
920 CFI_ADJUST_CFA_OFFSET 8
925 2: mfence /* workaround */
928 CFI_ADJUST_CFA_OFFSET -8
931 ENDPROC(load_gs_index)
933 .section __ex_table,"a"
935 .quad gs_change,bad_gs
938 /* running with kernelgs */
940 swapgs /* switch back to user gs */
947 * Create a kernel thread.
949 * C extern interface:
950 * extern long kernel_thread(int (*fn)(void *), void * arg, unsigned long flags)
952 * asm input arguments:
953 * rdi: fn, rsi: arg, rdx: flags
957 FAKE_STACK_FRAME $child_rip
960 # rdi: flags, rsi: usp, rdx: will be &pt_regs
962 orq kernel_thread_flags(%rip),%rdi
975 * It isn't worth to check for reschedule here,
976 * so internally to the x86_64 port you can rely on kernel_thread()
977 * not to reschedule the child before returning, this avoids the need
978 * of hacks for example to fork off the per-CPU idle tasks.
979 * [Hopefully no generic code relies on the reschedule -AK]
985 ENDPROC(kernel_thread)
988 pushq $0 # fake return address
991 * Here we are in the child and the registers are set as they were
992 * at kernel_thread() invocation in the parent.
1004 * execve(). This function needs to use IRET, not SYSRET, to set up all state properly.
1006 * C extern interface:
1007 * extern long execve(char *name, char **argv, char **envp)
1009 * asm input arguments:
1010 * rdi: name, rsi: argv, rdx: envp
1012 * We want to fallback into:
1013 * extern long sys_execve(char *name, char **argv,char **envp, struct pt_regs regs)
1015 * do_sys_execve asm fallback arguments:
1016 * rdi: name, rsi: argv, rdx: envp, fake frame on the stack
1023 movq %rax, RAX(%rsp)
1026 je int_ret_from_sys_call
1033 KPROBE_ENTRY(page_fault)
1034 errorentry do_page_fault
1035 KPROBE_END(page_fault)
1037 ENTRY(coprocessor_error)
1038 zeroentry do_coprocessor_error
1039 END(coprocessor_error)
1041 ENTRY(simd_coprocessor_error)
1042 zeroentry do_simd_coprocessor_error
1043 END(simd_coprocessor_error)
1045 ENTRY(device_not_available)
1046 zeroentry math_state_restore
1047 END(device_not_available)
1049 /* runs on exception stack */
1053 CFI_ADJUST_CFA_OFFSET 8
1054 paranoidentry do_debug, DEBUG_STACK
1058 /* runs on exception stack */
1062 CFI_ADJUST_CFA_OFFSET 8
1063 paranoidentry do_nmi, 0, 0
1064 #ifdef CONFIG_TRACE_IRQFLAGS
1075 CFI_ADJUST_CFA_OFFSET 8
1076 paranoidentry do_int3, DEBUG_STACK
1082 zeroentry do_overflow
1090 zeroentry do_invalid_op
1093 ENTRY(coprocessor_segment_overrun)
1094 zeroentry do_coprocessor_segment_overrun
1095 END(coprocessor_segment_overrun)
1098 zeroentry do_reserved
1101 /* runs on exception stack */
1104 paranoidentry do_double_fault
1110 errorentry do_invalid_TSS
1113 ENTRY(segment_not_present)
1114 errorentry do_segment_not_present
1115 END(segment_not_present)
1117 /* runs on exception stack */
1118 ENTRY(stack_segment)
1120 paranoidentry do_stack_segment
1125 KPROBE_ENTRY(general_protection)
1126 errorentry do_general_protection
1127 KPROBE_END(general_protection)
1129 ENTRY(alignment_check)
1130 errorentry do_alignment_check
1131 END(alignment_check)
1134 zeroentry do_divide_error
1137 ENTRY(spurious_interrupt_bug)
1138 zeroentry do_spurious_interrupt_bug
1139 END(spurious_interrupt_bug)
1141 #ifdef CONFIG_X86_MCE
1142 /* runs on exception stack */
1143 ENTRY(machine_check)
1146 CFI_ADJUST_CFA_OFFSET 8
1147 paranoidentry do_machine_check
1153 /* Call softirq on interrupt stack. Interrupts are off. */
1157 CFI_ADJUST_CFA_OFFSET 8
1158 CFI_REL_OFFSET rbp,0
1160 CFI_DEF_CFA_REGISTER rbp
1161 incl %gs:pda_irqcount
1162 cmove %gs:pda_irqstackptr,%rsp
1163 push %rbp # backlink for old unwinder
1166 CFI_DEF_CFA_REGISTER rsp
1167 CFI_ADJUST_CFA_OFFSET -8
1168 decl %gs:pda_irqcount
1171 ENDPROC(call_softirq)
1173 #ifdef CONFIG_STACK_UNWIND
1174 ENTRY(arch_unwind_init_running)
1176 movq %r15, R15(%rdi)
1177 movq %r14, R14(%rdi)
1179 movq %r13, R13(%rdi)
1180 movq %r12, R12(%rdi)
1182 movq %rbp, RBP(%rdi)
1183 movq %rbx, RBX(%rdi)
1185 movq %rax, R11(%rdi)
1186 movq %rax, R10(%rdi)
1189 movq %rax, RAX(%rdi)
1190 movq %rax, RCX(%rdi)
1191 movq %rax, RDX(%rdi)
1192 movq %rax, RSI(%rdi)
1193 movq %rax, RDI(%rdi)
1194 movq %rax, ORIG_RAX(%rdi)
1195 movq %rcx, RIP(%rdi)
1197 movq $__KERNEL_CS, CS(%rdi)
1198 movq %rax, EFLAGS(%rdi)
1199 movq %rcx, RSP(%rdi)
1200 movq $__KERNEL_DS, SS(%rdi)
1203 ENDPROC(arch_unwind_init_running)