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>
53 #include <asm/paravirt.h>
57 #ifndef CONFIG_PREEMPT
58 #define retint_kernel retint_restore_args
61 #ifdef CONFIG_PARAVIRT
62 ENTRY(native_irq_enable_syscall_ret)
63 movq %gs:pda_oldrsp,%rsp
66 #endif /* CONFIG_PARAVIRT */
69 .macro TRACE_IRQS_IRETQ offset=ARGOFFSET
70 #ifdef CONFIG_TRACE_IRQFLAGS
71 bt $9,EFLAGS-\offset(%rsp) /* interrupts off? */
79 * C code is not supposed to know about undefined top of stack. Every time
80 * a C function with an pt_regs argument is called from the SYSCALL based
81 * fast path FIXUP_TOP_OF_STACK is needed.
82 * RESTORE_TOP_OF_STACK syncs the syscall state after any possible ptregs
86 /* %rsp:at FRAMEEND */
87 .macro FIXUP_TOP_OF_STACK tmp
88 movq %gs:pda_oldrsp,\tmp
90 movq $__USER_DS,SS(%rsp)
91 movq $__USER_CS,CS(%rsp)
93 movq R11(%rsp),\tmp /* get eflags */
94 movq \tmp,EFLAGS(%rsp)
97 .macro RESTORE_TOP_OF_STACK tmp,offset=0
98 movq RSP-\offset(%rsp),\tmp
99 movq \tmp,%gs:pda_oldrsp
100 movq EFLAGS-\offset(%rsp),\tmp
101 movq \tmp,R11-\offset(%rsp)
104 .macro FAKE_STACK_FRAME child_rip
105 /* push in order ss, rsp, eflags, cs, rip */
108 CFI_ADJUST_CFA_OFFSET 8
109 /*CFI_REL_OFFSET ss,0*/
111 CFI_ADJUST_CFA_OFFSET 8
113 pushq $(1<<9) /* eflags - interrupts on */
114 CFI_ADJUST_CFA_OFFSET 8
115 /*CFI_REL_OFFSET rflags,0*/
116 pushq $__KERNEL_CS /* cs */
117 CFI_ADJUST_CFA_OFFSET 8
118 /*CFI_REL_OFFSET cs,0*/
119 pushq \child_rip /* rip */
120 CFI_ADJUST_CFA_OFFSET 8
122 pushq %rax /* orig rax */
123 CFI_ADJUST_CFA_OFFSET 8
126 .macro UNFAKE_STACK_FRAME
128 CFI_ADJUST_CFA_OFFSET -(6*8)
131 .macro CFI_DEFAULT_STACK start=1
137 CFI_DEF_CFA_OFFSET SS+8
139 CFI_REL_OFFSET r15,R15
140 CFI_REL_OFFSET r14,R14
141 CFI_REL_OFFSET r13,R13
142 CFI_REL_OFFSET r12,R12
143 CFI_REL_OFFSET rbp,RBP
144 CFI_REL_OFFSET rbx,RBX
145 CFI_REL_OFFSET r11,R11
146 CFI_REL_OFFSET r10,R10
149 CFI_REL_OFFSET rax,RAX
150 CFI_REL_OFFSET rcx,RCX
151 CFI_REL_OFFSET rdx,RDX
152 CFI_REL_OFFSET rsi,RSI
153 CFI_REL_OFFSET rdi,RDI
154 CFI_REL_OFFSET rip,RIP
155 /*CFI_REL_OFFSET cs,CS*/
156 /*CFI_REL_OFFSET rflags,EFLAGS*/
157 CFI_REL_OFFSET rsp,RSP
158 /*CFI_REL_OFFSET ss,SS*/
161 * A newly forked process directly context switches into this.
166 push kernel_eflags(%rip)
167 CFI_ADJUST_CFA_OFFSET 4
168 popf # reset kernel eflags
169 CFI_ADJUST_CFA_OFFSET -4
171 GET_THREAD_INFO(%rcx)
172 testl $(_TIF_SYSCALL_TRACE|_TIF_SYSCALL_AUDIT),threadinfo_flags(%rcx)
176 testl $3,CS-ARGOFFSET(%rsp) # from kernel_thread?
177 je int_ret_from_sys_call
178 testl $_TIF_IA32,threadinfo_flags(%rcx)
179 jnz int_ret_from_sys_call
180 RESTORE_TOP_OF_STACK %rdi,ARGOFFSET
181 jmp ret_from_sys_call
184 call syscall_trace_leave
185 GET_THREAD_INFO(%rcx)
191 * System call entry. Upto 6 arguments in registers are supported.
193 * SYSCALL does not save anything on the stack and does not change the
199 * rax system call number
201 * rcx return address for syscall/sysret, C arg3
204 * r10 arg3 (--> moved to rcx for C)
207 * r11 eflags for syscall/sysret, temporary for C
208 * r12-r15,rbp,rbx saved by C code, not touched.
210 * Interrupts are off on entry.
211 * Only called from user space.
213 * XXX if we had a free scratch register we could save the RSP into the stack frame
214 * and report it properly in ps. Unfortunately we haven't.
216 * When user can change the frames always force IRET. That is because
217 * it deals with uncanonical addresses better. SYSRET has trouble
218 * with them due to bugs in both AMD and Intel CPUs.
224 CFI_DEF_CFA rsp,PDA_STACKOFFSET
226 /*CFI_REGISTER rflags,r11*/
229 * A hypervisor implementation might want to use a label
230 * after the swapgs, so that it can do the swapgs
231 * for the guest and jump here on syscall.
233 ENTRY(system_call_after_swapgs)
235 movq %rsp,%gs:pda_oldrsp
236 movq %gs:pda_kernelstack,%rsp
238 * No need to follow this irqs off/on section - it's straight
241 ENABLE_INTERRUPTS(CLBR_NONE)
243 movq %rax,ORIG_RAX-ARGOFFSET(%rsp)
244 movq %rcx,RIP-ARGOFFSET(%rsp)
245 CFI_REL_OFFSET rip,RIP-ARGOFFSET
246 GET_THREAD_INFO(%rcx)
247 testl $(_TIF_SYSCALL_TRACE|_TIF_SYSCALL_AUDIT|_TIF_SECCOMP),threadinfo_flags(%rcx)
249 cmpq $__NR_syscall_max,%rax
252 call *sys_call_table(,%rax,8) # XXX: rip relative
253 movq %rax,RAX-ARGOFFSET(%rsp)
255 * Syscall return path ending with SYSRET (fast path)
256 * Has incomplete stack frame and undefined top of stack.
259 movl $_TIF_ALLWORK_MASK,%edi
263 GET_THREAD_INFO(%rcx)
264 DISABLE_INTERRUPTS(CLBR_NONE)
266 movl threadinfo_flags(%rcx),%edx
271 * sysretq will re-enable interrupts:
274 movq RIP-ARGOFFSET(%rsp),%rcx
276 RESTORE_ARGS 0,-ARG_SKIP,1
277 /*CFI_REGISTER rflags,r11*/
278 ENABLE_INTERRUPTS_SYSCALL_RET
281 /* Handle reschedules */
282 /* edx: work, edi: workmask */
284 bt $TIF_NEED_RESCHED,%edx
287 ENABLE_INTERRUPTS(CLBR_NONE)
289 CFI_ADJUST_CFA_OFFSET 8
292 CFI_ADJUST_CFA_OFFSET -8
295 /* Handle a signal */
298 ENABLE_INTERRUPTS(CLBR_NONE)
299 testl $_TIF_DO_NOTIFY_MASK,%edx
302 /* Really a signal */
303 /* edx: work flags (arg3) */
304 leaq do_notify_resume(%rip),%rax
305 leaq -ARGOFFSET(%rsp),%rdi # &pt_regs -> arg1
306 xorl %esi,%esi # oldset -> arg2
307 call ptregscall_common
308 1: movl $_TIF_NEED_RESCHED,%edi
309 /* Use IRET because user could have changed frame. This
310 works because ptregscall_common has called FIXUP_TOP_OF_STACK. */
311 DISABLE_INTERRUPTS(CLBR_NONE)
316 movq $-ENOSYS,RAX-ARGOFFSET(%rsp)
317 jmp ret_from_sys_call
319 /* Do syscall tracing */
322 movq $-ENOSYS,RAX(%rsp) /* ptrace can change this for a bad syscall */
323 FIXUP_TOP_OF_STACK %rdi
325 call syscall_trace_enter
326 LOAD_ARGS ARGOFFSET /* reload args from stack in case ptrace changed it */
328 cmpq $__NR_syscall_max,%rax
329 ja int_ret_from_sys_call /* RAX(%rsp) set to -ENOSYS above */
330 movq %r10,%rcx /* fixup for C */
331 call *sys_call_table(,%rax,8)
332 movq %rax,RAX-ARGOFFSET(%rsp)
333 /* Use IRET because user could have changed frame */
336 * Syscall return path ending with IRET.
337 * Has correct top of stack, but partial stack frame.
339 .globl int_ret_from_sys_call
340 int_ret_from_sys_call:
341 DISABLE_INTERRUPTS(CLBR_NONE)
343 testl $3,CS-ARGOFFSET(%rsp)
344 je retint_restore_args
345 movl $_TIF_ALLWORK_MASK,%edi
346 /* 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
363 ENABLE_INTERRUPTS(CLBR_NONE)
365 CFI_ADJUST_CFA_OFFSET 8
368 CFI_ADJUST_CFA_OFFSET -8
369 DISABLE_INTERRUPTS(CLBR_NONE)
373 /* handle signals and tracing -- both require a full stack frame */
376 ENABLE_INTERRUPTS(CLBR_NONE)
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
391 testl $_TIF_DO_NOTIFY_MASK,%edx
393 movq %rsp,%rdi # &ptregs -> arg1
394 xorl %esi,%esi # oldset -> arg2
395 call do_notify_resume
396 1: movl $_TIF_NEED_RESCHED,%edi
399 DISABLE_INTERRUPTS(CLBR_NONE)
406 * Certain special system calls that need to save a complete full stack frame.
409 .macro PTREGSCALL label,func,arg
412 leaq \func(%rip),%rax
413 leaq -ARGOFFSET+8(%rsp),\arg /* 8 for return address */
414 jmp ptregscall_common
420 PTREGSCALL stub_clone, sys_clone, %r8
421 PTREGSCALL stub_fork, sys_fork, %rdi
422 PTREGSCALL stub_vfork, sys_vfork, %rdi
423 PTREGSCALL stub_rt_sigsuspend, sys_rt_sigsuspend, %rdx
424 PTREGSCALL stub_sigaltstack, sys_sigaltstack, %rdx
425 PTREGSCALL stub_iopl, sys_iopl, %rsi
427 ENTRY(ptregscall_common)
429 CFI_ADJUST_CFA_OFFSET -8
430 CFI_REGISTER rip, r11
433 CFI_REGISTER rip, r15
434 FIXUP_TOP_OF_STACK %r11
436 RESTORE_TOP_OF_STACK %r11
438 CFI_REGISTER rip, r11
441 CFI_ADJUST_CFA_OFFSET 8
442 CFI_REL_OFFSET rip, 0
445 END(ptregscall_common)
450 CFI_ADJUST_CFA_OFFSET -8
451 CFI_REGISTER rip, r11
453 FIXUP_TOP_OF_STACK %r11
456 RESTORE_TOP_OF_STACK %r11
459 jmp int_ret_from_sys_call
464 * sigreturn is special because it needs to restore all registers on return.
465 * This cannot be done with SYSRET, so use the IRET return path instead.
467 ENTRY(stub_rt_sigreturn)
470 CFI_ADJUST_CFA_OFFSET -8
473 FIXUP_TOP_OF_STACK %r11
474 call sys_rt_sigreturn
475 movq %rax,RAX(%rsp) # fixme, this could be done at the higher layer
477 jmp int_ret_from_sys_call
479 END(stub_rt_sigreturn)
482 * 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 */
542 DISABLE_INTERRUPTS(CLBR_NONE)
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
562 movl threadinfo_flags(%rcx),%edx
567 retint_swapgs: /* return to user-space */
569 * The iretq could re-enable interrupts:
571 DISABLE_INTERRUPTS(CLBR_ANY)
576 retint_restore_args: /* return to kernel space */
577 DISABLE_INTERRUPTS(CLBR_ANY)
579 * The iretq could re-enable interrupts:
588 .section __ex_table, "a"
589 .quad irq_return, bad_iret
592 #ifdef CONFIG_PARAVIRT
596 .section __ex_table,"a"
597 .quad native_iret, bad_iret
604 * The iret traps when the %cs or %ss being restored is bogus.
605 * We've lost the original trap vector and error code.
606 * #GPF is the most likely one to get for an invalid selector.
607 * So pretend we completed the iret and took the #GPF in user mode.
609 * We are now running with the kernel GS after exception recovery.
610 * But error_entry expects us to have user GS to match the user %cs,
616 jmp general_protection
620 /* edi: workmask, edx: work */
623 bt $TIF_NEED_RESCHED,%edx
626 ENABLE_INTERRUPTS(CLBR_NONE)
628 CFI_ADJUST_CFA_OFFSET 8
631 CFI_ADJUST_CFA_OFFSET -8
632 GET_THREAD_INFO(%rcx)
633 DISABLE_INTERRUPTS(CLBR_NONE)
638 testl $_TIF_DO_NOTIFY_MASK,%edx
641 ENABLE_INTERRUPTS(CLBR_NONE)
643 movq $-1,ORIG_RAX(%rsp)
644 xorl %esi,%esi # oldset
645 movq %rsp,%rdi # &pt_regs
646 call do_notify_resume
648 DISABLE_INTERRUPTS(CLBR_NONE)
650 movl $_TIF_NEED_RESCHED,%edi
651 GET_THREAD_INFO(%rcx)
654 #ifdef CONFIG_PREEMPT
655 /* Returning to kernel space. Check if we need preemption */
656 /* rcx: threadinfo. interrupts off. */
658 cmpl $0,threadinfo_preempt_count(%rcx)
659 jnz retint_restore_args
660 bt $TIF_NEED_RESCHED,threadinfo_flags(%rcx)
661 jnc retint_restore_args
662 bt $9,EFLAGS-ARGOFFSET(%rsp) /* interrupts off? */
663 jnc retint_restore_args
664 call preempt_schedule_irq
669 END(common_interrupt)
674 .macro apicinterrupt num,func
677 CFI_ADJUST_CFA_OFFSET 8
683 ENTRY(thermal_interrupt)
684 apicinterrupt THERMAL_APIC_VECTOR,smp_thermal_interrupt
685 END(thermal_interrupt)
687 ENTRY(threshold_interrupt)
688 apicinterrupt THRESHOLD_APIC_VECTOR,mce_threshold_interrupt
689 END(threshold_interrupt)
692 ENTRY(reschedule_interrupt)
693 apicinterrupt RESCHEDULE_VECTOR,smp_reschedule_interrupt
694 END(reschedule_interrupt)
696 .macro INVALIDATE_ENTRY num
697 ENTRY(invalidate_interrupt\num)
698 apicinterrupt INVALIDATE_TLB_VECTOR_START+\num,smp_invalidate_interrupt
699 END(invalidate_interrupt\num)
711 ENTRY(call_function_interrupt)
712 apicinterrupt CALL_FUNCTION_VECTOR,smp_call_function_interrupt
713 END(call_function_interrupt)
714 ENTRY(irq_move_cleanup_interrupt)
715 apicinterrupt IRQ_MOVE_CLEANUP_VECTOR,smp_irq_move_cleanup_interrupt
716 END(irq_move_cleanup_interrupt)
719 ENTRY(apic_timer_interrupt)
720 apicinterrupt LOCAL_TIMER_VECTOR,smp_apic_timer_interrupt
721 END(apic_timer_interrupt)
723 ENTRY(error_interrupt)
724 apicinterrupt ERROR_APIC_VECTOR,smp_error_interrupt
727 ENTRY(spurious_interrupt)
728 apicinterrupt SPURIOUS_APIC_VECTOR,smp_spurious_interrupt
729 END(spurious_interrupt)
732 * Exception entry points.
736 pushq $0 /* push error code/oldrax */
737 CFI_ADJUST_CFA_OFFSET 8
738 pushq %rax /* push real oldrax to the rdi slot */
739 CFI_ADJUST_CFA_OFFSET 8
746 .macro errorentry sym
749 CFI_ADJUST_CFA_OFFSET 8
756 /* error code is on the stack already */
757 /* handle NMI like exceptions that can happen everywhere */
758 .macro paranoidentry sym, ist=0, irqtrace=1
762 movl $MSR_GS_BASE,%ecx
770 movq %gs:pda_data_offset, %rbp
773 movq ORIG_RAX(%rsp),%rsi
774 movq $-1,ORIG_RAX(%rsp)
776 subq $EXCEPTION_STKSZ, per_cpu__init_tss + TSS_ist + (\ist - 1) * 8(%rbp)
780 addq $EXCEPTION_STKSZ, per_cpu__init_tss + TSS_ist + (\ist - 1) * 8(%rbp)
782 DISABLE_INTERRUPTS(CLBR_NONE)
789 * "Paranoid" exit path from exception stack.
790 * Paranoid because this is used by NMIs and cannot take
791 * any kernel state for granted.
792 * We don't do kernel preemption checks here, because only
793 * NMI should be common and it does not enable IRQs and
794 * cannot get reschedule ticks.
796 * "trace" is 0 for the NMI handler only, because irq-tracing
797 * is fundamentally NMI-unsafe. (we cannot change the soft and
798 * hard flags at once, atomically)
800 .macro paranoidexit trace=1
801 /* ebx: no swapgs flag */
803 testl %ebx,%ebx /* swapgs needed? */
804 jnz paranoid_restore\trace
806 jnz paranoid_userspace\trace
807 paranoid_swapgs\trace:
812 paranoid_restore\trace:
815 paranoid_userspace\trace:
816 GET_THREAD_INFO(%rcx)
817 movl threadinfo_flags(%rcx),%ebx
818 andl $_TIF_WORK_MASK,%ebx
819 jz paranoid_swapgs\trace
820 movq %rsp,%rdi /* &pt_regs */
822 movq %rax,%rsp /* switch stack for scheduling */
823 testl $_TIF_NEED_RESCHED,%ebx
824 jnz paranoid_schedule\trace
825 movl %ebx,%edx /* arg3: thread flags */
829 ENABLE_INTERRUPTS(CLBR_NONE)
830 xorl %esi,%esi /* arg2: oldset */
831 movq %rsp,%rdi /* arg1: &pt_regs */
832 call do_notify_resume
833 DISABLE_INTERRUPTS(CLBR_NONE)
837 jmp paranoid_userspace\trace
838 paranoid_schedule\trace:
842 ENABLE_INTERRUPTS(CLBR_ANY)
844 DISABLE_INTERRUPTS(CLBR_ANY)
848 jmp paranoid_userspace\trace
853 * Exception entry point. This expects an error code/orig_rax on the stack
854 * and the exception handler in %rax.
856 KPROBE_ENTRY(error_entry)
859 /* rdi slot contains rax, oldrax contains error code */
862 CFI_ADJUST_CFA_OFFSET (14*8)
864 CFI_REL_OFFSET rsi,RSI
865 movq 14*8(%rsp),%rsi /* load rax from rdi slot */
868 CFI_REL_OFFSET rdx,RDX
870 CFI_REL_OFFSET rcx,RCX
871 movq %rsi,10*8(%rsp) /* store rax */
872 CFI_REL_OFFSET rax,RAX
878 CFI_REL_OFFSET r10,R10
880 CFI_REL_OFFSET r11,R11
882 CFI_REL_OFFSET rbx,RBX
884 CFI_REL_OFFSET rbp,RBP
886 CFI_REL_OFFSET r12,R12
888 CFI_REL_OFFSET r13,R13
890 CFI_REL_OFFSET r14,R14
892 CFI_REL_OFFSET r15,R15
900 CFI_REL_OFFSET rdi,RDI
902 movq ORIG_RAX(%rsp),%rsi /* get error code */
903 movq $-1,ORIG_RAX(%rsp)
905 /* ebx: no swapgs flag (1: don't need swapgs, 0: need it) */
909 DISABLE_INTERRUPTS(CLBR_NONE)
911 GET_THREAD_INFO(%rcx)
915 movl threadinfo_flags(%rcx),%edx
916 movl $_TIF_WORK_MASK,%edi
924 /* There are two places in the kernel that can potentially fault with
925 usergs. Handle them here. The exception handlers after
926 iret run with kernel gs again, so don't set the user space flag.
927 B stepping K8s sometimes report an truncated RIP for IRET
928 exceptions returning to compat mode. Check for these here too. */
929 leaq irq_return(%rip),%rbp
932 movl %ebp,%ebp /* zero extend */
935 cmpq $gs_change,RIP(%rsp)
938 KPROBE_END(error_entry)
940 /* Reload gs selector with exception handling */
941 /* edi: new selector */
945 CFI_ADJUST_CFA_OFFSET 8
946 DISABLE_INTERRUPTS(CLBR_ANY | ~(CLBR_RDI))
950 2: mfence /* workaround */
953 CFI_ADJUST_CFA_OFFSET -8
956 ENDPROC(load_gs_index)
958 .section __ex_table,"a"
960 .quad gs_change,bad_gs
963 /* running with kernelgs */
965 SWAPGS /* switch back to user gs */
972 * Create a kernel thread.
974 * C extern interface:
975 * extern long kernel_thread(int (*fn)(void *), void * arg, unsigned long flags)
977 * asm input arguments:
978 * rdi: fn, rsi: arg, rdx: flags
982 FAKE_STACK_FRAME $child_rip
985 # rdi: flags, rsi: usp, rdx: will be &pt_regs
987 orq kernel_thread_flags(%rip),%rdi
1000 * It isn't worth to check for reschedule here,
1001 * so internally to the x86_64 port you can rely on kernel_thread()
1002 * not to reschedule the child before returning, this avoids the need
1003 * of hacks for example to fork off the per-CPU idle tasks.
1004 * [Hopefully no generic code relies on the reschedule -AK]
1010 ENDPROC(kernel_thread)
1013 pushq $0 # fake return address
1016 * Here we are in the child and the registers are set as they were
1017 * at kernel_thread() invocation in the parent.
1029 * execve(). This function needs to use IRET, not SYSRET, to set up all state properly.
1031 * C extern interface:
1032 * extern long execve(char *name, char **argv, char **envp)
1034 * asm input arguments:
1035 * rdi: name, rsi: argv, rdx: envp
1037 * We want to fallback into:
1038 * extern long sys_execve(char *name, char **argv,char **envp, struct pt_regs *regs)
1040 * do_sys_execve asm fallback arguments:
1041 * rdi: name, rsi: argv, rdx: envp, rcx: fake frame on the stack
1043 ENTRY(kernel_execve)
1049 movq %rax, RAX(%rsp)
1052 je int_ret_from_sys_call
1057 ENDPROC(kernel_execve)
1059 KPROBE_ENTRY(page_fault)
1060 errorentry do_page_fault
1061 KPROBE_END(page_fault)
1063 ENTRY(coprocessor_error)
1064 zeroentry do_coprocessor_error
1065 END(coprocessor_error)
1067 ENTRY(simd_coprocessor_error)
1068 zeroentry do_simd_coprocessor_error
1069 END(simd_coprocessor_error)
1071 ENTRY(device_not_available)
1072 zeroentry math_state_restore
1073 END(device_not_available)
1075 /* runs on exception stack */
1079 CFI_ADJUST_CFA_OFFSET 8
1080 paranoidentry do_debug, DEBUG_STACK
1084 /* runs on exception stack */
1088 CFI_ADJUST_CFA_OFFSET 8
1089 paranoidentry do_nmi, 0, 0
1090 #ifdef CONFIG_TRACE_IRQFLAGS
1101 CFI_ADJUST_CFA_OFFSET 8
1102 paranoidentry do_int3, DEBUG_STACK
1108 zeroentry do_overflow
1116 zeroentry do_invalid_op
1119 ENTRY(coprocessor_segment_overrun)
1120 zeroentry do_coprocessor_segment_overrun
1121 END(coprocessor_segment_overrun)
1124 zeroentry do_reserved
1127 /* runs on exception stack */
1130 paranoidentry do_double_fault
1136 errorentry do_invalid_TSS
1139 ENTRY(segment_not_present)
1140 errorentry do_segment_not_present
1141 END(segment_not_present)
1143 /* runs on exception stack */
1144 ENTRY(stack_segment)
1146 paranoidentry do_stack_segment
1151 KPROBE_ENTRY(general_protection)
1152 errorentry do_general_protection
1153 KPROBE_END(general_protection)
1155 ENTRY(alignment_check)
1156 errorentry do_alignment_check
1157 END(alignment_check)
1160 zeroentry do_divide_error
1163 ENTRY(spurious_interrupt_bug)
1164 zeroentry do_spurious_interrupt_bug
1165 END(spurious_interrupt_bug)
1167 #ifdef CONFIG_X86_MCE
1168 /* runs on exception stack */
1169 ENTRY(machine_check)
1172 CFI_ADJUST_CFA_OFFSET 8
1173 paranoidentry do_machine_check
1179 /* Call softirq on interrupt stack. Interrupts are off. */
1183 CFI_ADJUST_CFA_OFFSET 8
1184 CFI_REL_OFFSET rbp,0
1186 CFI_DEF_CFA_REGISTER rbp
1187 incl %gs:pda_irqcount
1188 cmove %gs:pda_irqstackptr,%rsp
1189 push %rbp # backlink for old unwinder
1192 CFI_DEF_CFA_REGISTER rsp
1193 CFI_ADJUST_CFA_OFFSET -8
1194 decl %gs:pda_irqcount
1197 ENDPROC(call_softirq)
1199 KPROBE_ENTRY(ignore_sysret)
1204 ENDPROC(ignore_sysret)