linux系统调用用户态到内核态流程 |
您所在的位置：网站首页 › 简述用户如何利用系统调用,以及系统工作过程 › linux系统调用用户态到内核态流程 |
linux系统调用用户态到内核态流程
|
Linux系统调用流程分析 一直很好奇系统调用的详细过程，只简单的了解系统调用是用户与内核交互的借口，看了几天的内容，不知道从何下手开始写，今天略有头绪，做以简单总结。 1， 什么是系统调用。 系统调用：系统调用是操作系统为用户提供的一系列API；系统调用将用户的请求发给内核，内核执行完以后，将结果返回给用户； 2， 一个系统调用的流程。
如上图，是系统调用write()函数的调用过程，不难看出，在用户开始调用系统调用时，触发了软中断Int0X80。到这肯定有人疑问，为什么中断号是0X80，因为代码中把系统调用跟0X80绑定在一起。 中断向量：一个内存单元，存放中断处理程序入口地址和程序运行时所需处理机状态字。 
这里不再深入讲解向量表的生成，只需要知道每个中断都有对应的中断号并对应处理程序。 linux的中断向量表： 
a.0X80中断号：
下面是_system_call的汇编代码(asm.S)： _system_call: cmp eax,nr_system_calls-1 ;// 调用号如果超出范围的话就在eax 中置-1 并退出。 ja bad_sys_call push ds ;// 保存原段寄存器值。 push es push fs push edx ;// ebx,ecx,edx 中放着系统调用相应的C 语言函数的调用参数。 push ecx ;// push %ebx,%ecx,%edx as parameters push ebx ;// to the system call mov edx,10h ;// set up ds,es to kernel space mov ds,dx ;// ds,es 指向内核数据段(全局描述符表中数据段描述符)。 mov es,dx mov edx,17h ;// fs points to local data space mov fs,dx ;// fs 指向局部数据段(局部描述符表中数据段描述符)。 ;// 下面这句操作数的含义是：调用地址 = _sys_call_table + %eax * 4。参见列表后的说明。 ;// 对应的C 程序中的sys_call_table 在include/linux/sys.h 中，其中定义了一个包括72 个 ;// 系统调用C 处理函数的地址数组表。 call [_sys_call_table+eax*4] push eax ;// 把系统调用号入栈。 mov eax,_current ;// 取当前任务（进程）数据结构地址??eax。 ;// 下面97-100 行查看当前任务的运行状态。如果不在就绪状态(state 不等于0)就去执行调度程序。 ;// 如果该任务在就绪状态但counter[??]值等于0，则也去执行调度程序。 cmp dword ptr [state+eax],0 ;// state jne reschedule cmp dword ptr [counter+eax],0 ;// counter je rescheduleb.现场保护与恢复 因为用户态和内核态不共用一个堆栈，所以用户态调用系统调用时，得先保存用户态的信息到寄存器，然后切入到内核态，将寄存器中的信息拷贝到内核栈开始执行，执行结束后，还原之前用户态的状态让用户态继续往下执行。 用户态程序通过ebx,ecx,edx等寄存器，传递函数参数给系统调用，ebx,ecx,edx等寄存器通过SAVE_ALL宏压到内核栈。 系统调用结束之后，RESTORE_ALL宏把SAVE_ALL宏压入的寄存器反向弹出，以用于恢复现场。 
c.调用系统调用表 syscall_handler_t *sys_call_table[] = { [ __NR_restart_syscall ] = (syscall_handler_t *) sys_restart_syscall, [ __NR_exit ] = (syscall_handler_t *) sys_exit, [ __NR_fork ] = (syscall_handler_t *) sys_fork, [ __NR_read ] = (syscall_handler_t *) sys_read, [ __NR_write ] = (syscall_handler_t *) sys_write, /* These three are declared differently in asm/unistd.h */ [ __NR_open ] = (syscall_handler_t *) sys_open, [ __NR_close ] = (syscall_handler_t *) sys_close, [ __NR_creat ] = (syscall_handler_t *) sys_creat, [ __NR_link ] = (syscall_handler_t *) sys_link, [ __NR_unlink ] = (syscall_handler_t *) sys_unlink, [ __NR_execve ] = (syscall_handler_t *) sys_execve, /* declared differently in kern_util.h */ [ __NR_chdir ] = (syscall_handler_t *) sys_chdir, [ __NR_time ] = um_time, [ __NR_mknod ] = (syscall_handler_t *) sys_mknod, [ __NR_chmod ] = (syscall_handler_t *) sys_chmod, [ __NR_lchown ] = (syscall_handler_t *) sys_lchown16, [ __NR_lseek ] = (syscall_handler_t *) sys_lseek, [ __NR_getpid ] = (syscall_handler_t *) sys_getpid, [ __NR_mount ] = um_mount, [ __NR_setuid ] = (syscall_handler_t *) sys_setuid16, [ __NR_getuid ] = (syscall_handler_t *) sys_getuid16, [ __NR_ptrace ] = (syscall_handler_t *) sys_ptrace, [ __NR_alarm ] = (syscall_handler_t *) sys_alarm, [ __NR_pause ] = (syscall_handler_t *) sys_pause, [ __NR_utime ] = (syscall_handler_t *) sys_utime, [ __NR_access ] = (syscall_handler_t *) sys_access, [ __NR_sync ] = (syscall_handler_t *) sys_sync, [ __NR_kill ] = (syscall_handler_t *) sys_kill, [ __NR_rename ] = (syscall_handler_t *) sys_rename, [ __NR_mkdir ] = (syscall_handler_t *) sys_mkdir, [ __NR_rmdir ] = (syscall_handler_t *) sys_rmdir, /* Declared differently in asm/unistd.h */ [ __NR_dup ] = (syscall_handler_t *) sys_dup, [ __NR_pipe ] = (syscall_handler_t *) sys_pipe, [ __NR_times ] = (syscall_handler_t *) sys_times, [ __NR_brk ] = (syscall_handler_t *) sys_brk, [ __NR_setgid ] = (syscall_handler_t *) sys_setgid16, [ __NR_getgid ] = (syscall_handler_t *) sys_getgid16, [ __NR_geteuid ] = (syscall_handler_t *) sys_geteuid16, [ __NR_getegid ] = (syscall_handler_t *) sys_getegid16, [ __NR_acct ] = (syscall_handler_t *) sys_acct, [ __NR_umount2 ] = (syscall_handler_t *) sys_umount, [ __NR_ioctl ] = (syscall_handler_t *) sys_ioctl, [ __NR_fcntl ] = (syscall_handler_t *) sys_fcntl, [ __NR_setpgid ] = (syscall_handler_t *) sys_setpgid, [ __NR_umask ] = (syscall_handler_t *) sys_umask, [ __NR_chroot ] = (syscall_handler_t *) sys_chroot, [ __NR_ustat ] = (syscall_handler_t *) sys_ustat, [ __NR_dup2 ] = (syscall_handler_t *) sys_dup2, [ __NR_getppid ] = (syscall_handler_t *) sys_getppid, [ __NR_getpgrp ] = (syscall_handler_t *) sys_getpgrp, [ __NR_setsid ] = (syscall_handler_t *) sys_setsid, [ __NR_setreuid ] = (syscall_handler_t *) sys_setreuid16, [ __NR_setregid ] = (syscall_handler_t *) sys_setregid16, [ __NR_sethostname ] = (syscall_handler_t *) sys_sethostname, [ __NR_setrlimit ] = (syscall_handler_t *) sys_setrlimit, [ __NR_getrlimit ] = (syscall_handler_t *) sys_old_getrlimit, [ __NR_getrusage ] = (syscall_handler_t *) sys_getrusage, [ __NR_gettimeofday ] = (syscall_handler_t *) sys_gettimeofday, [ __NR_settimeofday ] = (syscall_handler_t *) sys_settimeofday, [ __NR_getgroups ] = (syscall_handler_t *) sys_getgroups16, [ __NR_setgroups ] = (syscall_handler_t *) sys_setgroups16, [ __NR_symlink ] = (syscall_handler_t *) sys_symlink, [ __NR_readlink ] = (syscall_handler_t *) sys_readlink, [ __NR_uselib ] = (syscall_handler_t *) sys_uselib, [ __NR_swapon ] = (syscall_handler_t *) sys_swapon, [ __NR_reboot ] = (syscall_handler_t *) sys_reboot, [ __NR_munmap ] = (syscall_handler_t *) sys_munmap, [ __NR_truncate ] = (syscall_handler_t *) sys_truncate, [ __NR_ftruncate ] = (syscall_handler_t *) sys_ftruncate, [ __NR_fchmod ] = (syscall_handler_t *) sys_fchmod, [ __NR_fchown ] = (syscall_handler_t *) sys_fchown16, [ __NR_getpriority ] = (syscall_handler_t *) sys_getpriority, [ __NR_setpriority ] = (syscall_handler_t *) sys_setpriority, [ __NR_statfs ] = (syscall_handler_t *) sys_statfs, [ __NR_fstatfs ] = (syscall_handler_t *) sys_fstatfs, [ __NR_ioperm ] = (syscall_handler_t *) sys_ni_syscall, [ __NR_syslog ] = (syscall_handler_t *) sys_syslog, [ __NR_setitimer ] = (syscall_handler_t *) sys_setitimer, [ __NR_getitimer ] = (syscall_handler_t *) sys_getitimer, [ __NR_stat ] = (syscall_handler_t *) sys_newstat, [ __NR_lstat ] = (syscall_handler_t *) sys_newlstat, [ __NR_fstat ] = (syscall_handler_t *) sys_newfstat, [ __NR_vhangup ] = (syscall_handler_t *) sys_vhangup, [ __NR_wait4 ] = (syscall_handler_t *) sys_wait4, [ __NR_swapoff ] = (syscall_handler_t *) sys_swapoff, [ __NR_sysinfo ] = (syscall_handler_t *) sys_sysinfo, [ __NR_fsync ] = (syscall_handler_t *) sys_fsync, [ __NR_clone ] = (syscall_handler_t *) sys_clone, [ __NR_setdomainname ] = (syscall_handler_t *) sys_setdomainname, [ __NR_uname ] = (syscall_handler_t *) sys_newuname, [ __NR_adjtimex ] = (syscall_handler_t *) sys_adjtimex, [ __NR_mprotect ] = (syscall_handler_t *) sys_mprotect, [ __NR_create_module ] = (syscall_handler_t *) sys_ni_syscall, [ __NR_init_module ] = (syscall_handler_t *) sys_init_module, [ __NR_delete_module ] = (syscall_handler_t *) sys_delete_module, [ __NR_get_kernel_syms ] = (syscall_handler_t *) sys_ni_syscall, [ __NR_quotactl ] = (syscall_handler_t *) sys_quotactl, [ __NR_getpgid ] = (syscall_handler_t *) sys_getpgid, [ __NR_fchdir ] = (syscall_handler_t *) sys_fchdir, [ __NR_sysfs ] = (syscall_handler_t *) sys_sysfs, [ __NR_personality ] = (syscall_handler_t *) sys_personality, [ __NR_afs_syscall ] = (syscall_handler_t *) sys_ni_syscall, [ __NR_setfsuid ] = (syscall_handler_t *) sys_setfsuid16, [ __NR_setfsgid ] = (syscall_handler_t *) sys_setfsgid16, [ __NR_getdents ] = (syscall_handler_t *) sys_getdents, [ __NR_flock ] = (syscall_handler_t *) sys_flock, [ __NR_msync ] = (syscall_handler_t *) sys_msync, [ __NR_readv ] = (syscall_handler_t *) sys_readv, [ __NR_writev ] = (syscall_handler_t *) sys_writev, [ __NR_getsid ] = (syscall_handler_t *) sys_getsid, [ __NR_fdatasync ] = (syscall_handler_t *) sys_fdatasync, [ __NR__sysctl ] = (syscall_handler_t *) sys_sysctl, [ __NR_mlock ] = (syscall_handler_t *) sys_mlock, [ __NR_munlock ] = (syscall_handler_t *) sys_munlock, [ __NR_mlockall ] = (syscall_handler_t *) sys_mlockall, [ __NR_munlockall ] = (syscall_handler_t *) sys_munlockall, [ __NR_sched_setparam ] = (syscall_handler_t *) sys_sched_setparam, [ __NR_sched_getparam ] = (syscall_handler_t *) sys_sched_getparam, [ __NR_sched_setscheduler ] = (syscall_handler_t *) sys_sched_setscheduler, [ __NR_sched_getscheduler ] = (syscall_handler_t *) sys_sched_getscheduler, [ __NR_sched_yield ] = (syscall_handler_t *) yield, [ __NR_sched_get_priority_max ] = (syscall_handler_t *) sys_sched_get_priority_max, [ __NR_sched_get_priority_min ] = (syscall_handler_t *) sys_sched_get_priority_min, [ __NR_sched_rr_get_interval ] = (syscall_handler_t *) sys_sched_rr_get_interval, [ __NR_nanosleep ] = (syscall_handler_t *) sys_nanosleep, [ __NR_mremap ] = (syscall_handler_t *) sys_mremap, [ __NR_setresuid ] = (syscall_handler_t *) sys_setresuid16, [ __NR_getresuid ] = (syscall_handler_t *) sys_getresuid16, [ __NR_query_module ] = (syscall_handler_t *) sys_ni_syscall, [ __NR_poll ] = (syscall_handler_t *) sys_poll, [ __NR_nfsservctl ] = (syscall_handler_t *) NFSSERVCTL, [ __NR_setresgid ] = (syscall_handler_t *) sys_setresgid16, [ __NR_getresgid ] = (syscall_handler_t *) sys_getresgid16, [ __NR_prctl ] = (syscall_handler_t *) sys_prctl, [ __NR_rt_sigreturn ] = (syscall_handler_t *) sys_rt_sigreturn, [ __NR_rt_sigaction ] = (syscall_handler_t *) sys_rt_sigaction, [ __NR_rt_sigprocmask ] = (syscall_handler_t *) sys_rt_sigprocmask, [ __NR_rt_sigpending ] = (syscall_handler_t *) sys_rt_sigpending, [ __NR_rt_sigtimedwait ] = (syscall_handler_t *) sys_rt_sigtimedwait, [ __NR_rt_sigqueueinfo ] = (syscall_handler_t *) sys_rt_sigqueueinfo, [ __NR_rt_sigsuspend ] = (syscall_handler_t *) sys_rt_sigsuspend, [ __NR_pread64 ] = (syscall_handler_t *) sys_pread64, [ __NR_pwrite64 ] = (syscall_handler_t *) sys_pwrite64, [ __NR_chown ] = (syscall_handler_t *) sys_chown16, [ __NR_getcwd ] = (syscall_handler_t *) sys_getcwd, [ __NR_capget ] = (syscall_handler_t *) sys_capget, [ __NR_capset ] = (syscall_handler_t *) sys_capset, [ __NR_sigaltstack ] = (syscall_handler_t *) sys_sigaltstack, [ __NR_sendfile ] = (syscall_handler_t *) sys_sendfile, [ __NR_getpmsg ] = (syscall_handler_t *) sys_ni_syscall, [ __NR_putpmsg ] = (syscall_handler_t *) sys_ni_syscall, [ __NR_vfork ] = (syscall_handler_t *) sys_vfork, [ __NR_getdents64 ] = (syscall_handler_t *) sys_getdents64, [ __NR_gettid ] = (syscall_handler_t *) sys_gettid, [ __NR_readahead ] = (syscall_handler_t *) sys_readahead, [ __NR_setxattr ] = (syscall_handler_t *) sys_setxattr, [ __NR_lsetxattr ] = (syscall_handler_t *) sys_lsetxattr, [ __NR_fsetxattr ] = (syscall_handler_t *) sys_fsetxattr, [ __NR_getxattr ] = (syscall_handler_t *) sys_getxattr, [ __NR_lgetxattr ] = (syscall_handler_t *) sys_lgetxattr, [ __NR_fgetxattr ] = (syscall_handler_t *) sys_fgetxattr, [ __NR_listxattr ] = (syscall_handler_t *) sys_listxattr, [ __NR_llistxattr ] = (syscall_handler_t *) sys_llistxattr, [ __NR_flistxattr ] = (syscall_handler_t *) sys_flistxattr, [ __NR_removexattr ] = (syscall_handler_t *) sys_removexattr, [ __NR_lremovexattr ] = (syscall_handler_t *) sys_lremovexattr, [ __NR_fremovexattr ] = (syscall_handler_t *) sys_fremovexattr, [ __NR_tkill ] = (syscall_handler_t *) sys_tkill, [ __NR_futex ] = (syscall_handler_t *) sys_futex, [ __NR_sched_setaffinity ] = (syscall_handler_t *) sys_sched_setaffinity, [ __NR_sched_getaffinity ] = (syscall_handler_t *) sys_sched_getaffinity, [ __NR_io_setup ] = (syscall_handler_t *) sys_io_setup, [ __NR_io_destroy ] = (syscall_handler_t *) sys_io_destroy, [ __NR_io_getevents ] = (syscall_handler_t *) sys_io_getevents, [ __NR_io_submit ] = (syscall_handler_t *) sys_io_submit, [ __NR_io_cancel ] = (syscall_handler_t *) sys_io_cancel, [ __NR_exit_group ] = (syscall_handler_t *) sys_exit_group, [ __NR_lookup_dcookie ] = (syscall_handler_t *) sys_lookup_dcookie, [ __NR_epoll_create ] = (syscall_handler_t *) sys_epoll_create, [ __NR_epoll_ctl ] = (syscall_handler_t *) sys_epoll_ctl, [ __NR_epoll_wait ] = (syscall_handler_t *) sys_epoll_wait, [ __NR_set_tid_address ] = (syscall_handler_t *) sys_set_tid_address, [ __NR_timer_create ] = (syscall_handler_t *) sys_timer_create, [ __NR_timer_settime ] = (syscall_handler_t *) sys_timer_settime, [ __NR_timer_gettime ] = (syscall_handler_t *) sys_timer_gettime, [ __NR_timer_getoverrun ] = (syscall_handler_t *) sys_timer_getoverrun, [ __NR_timer_delete ] = (syscall_handler_t *) sys_timer_delete, [ __NR_clock_settime ] = (syscall_handler_t *) sys_clock_settime, [ __NR_clock_gettime ] = (syscall_handler_t *) sys_clock_gettime, [ __NR_clock_getres ] = (syscall_handler_t *) sys_clock_getres, [ __NR_clock_nanosleep ] = (syscall_handler_t *) sys_clock_nanosleep, [ __NR_statfs64 ] = (syscall_handler_t *) sys_statfs64, [ __NR_fstatfs64 ] = (syscall_handler_t *) sys_fstatfs64, [ __NR_tgkill ] = (syscall_handler_t *) sys_tgkill, [ __NR_utimes ] = (syscall_handler_t *) sys_utimes, [ __NR_fadvise64_64 ] = (syscall_handler_t *) sys_fadvise64_64, [ __NR_vserver ] = (syscall_handler_t *) sys_vserver, [ __NR_mbind ] = (syscall_handler_t *) sys_mbind, [ __NR_get_mempolicy ] = (syscall_handler_t *) sys_get_mempolicy, [ __NR_set_mempolicy ] = (syscall_handler_t *) sys_set_mempolicy, [ __NR_mq_open ] = (syscall_handler_t *) sys_mq_open, [ __NR_mq_unlink ] = (syscall_handler_t *) sys_mq_unlink, [ __NR_mq_timedsend ] = (syscall_handler_t *) sys_mq_timedsend, [ __NR_mq_timedreceive ] = (syscall_handler_t *) sys_mq_timedreceive, [ __NR_mq_notify ] = (syscall_handler_t *) sys_mq_notify, [ __NR_mq_getsetattr ] = (syscall_handler_t *) sys_mq_getsetattr, [ __NR_sys_kexec_load ] = (syscall_handler_t *) sys_ni_syscall, [ __NR_waitid ] = (syscall_handler_t *) sys_waitid, [ 285 ] = (syscall_handler_t *) sys_ni_syscall, [ __NR_add_key ] = (syscall_handler_t *) sys_add_key, [ __NR_request_key ] = (syscall_handler_t *) sys_request_key, [ __NR_keyctl ] = (syscall_handler_t *) sys_keyctl, ARCH_SYSCALLS [ LAST_SYSCALL + 1 ... NR_syscalls ] = (syscall_handler_t *) sys_ni_syscall };从这里会跳转到具体的系统调用。 每个系统调用对应一个调用号（unistd.h） #ifndef _ALPHA_UNISTD_H #define _ALPHA_UNISTD_H #define __NR_osf_syscall 0 /* not implemented */ #define __NR_exit 1 #define __NR_fork 2 #define __NR_read 3 #define __NR_write 4 #define __NR_osf_old_open 5 /* not implemented */ #define __NR_close 6 #define __NR_osf_wait4 7 #define __NR_osf_old_creat 8 /* not implemented */ #define __NR_link 9 #define __NR_unlink 10 #define __NR_osf_execve 11 /* not implemented */ #define __NR_chdir 12 #define __NR_fchdir 13 #define __NR_mknod 14 #define __NR_chmod 15 #define __NR_chown 16 #define __NR_brk 17 #define __NR_osf_getfsstat 18 /* not implemented */ #define __NR_lseek 19 #define __NR_getxpid 20 #define __NR_osf_mount 21 #define __NR_umount 22 #define __NR_setuid 23 #define __NR_getxuid 24 #define __NR_exec_with_loader 25 /* not implemented */ #define __NR_ptrace 26 #define __NR_osf_nrecvmsg 27 /* not implemented */ #define __NR_osf_nsendmsg 28 /* not implemented */ #define __NR_osf_nrecvfrom 29 /* not implemented */ #define __NR_osf_naccept 30 /* not implemented */ #define __NR_osf_ngetpeername 31 /* not implemented */ #define __NR_osf_ngetsockname 32 /* not implemented */ #define __NR_access 33 #define __NR_osf_chflags 34 /* not implemented */ #define __NR_osf_fchflags 35 /* not implemented */ #define __NR_sync 36 #define __NR_kill 37 #define __NR_osf_old_stat 38 /* not implemented */ #define __NR_setpgid 39 #define __NR_osf_old_lstat 40 /* not implemented */ #define __NR_dup 41 #define __NR_pipe 42 #define __NR_osf_set_program_attributes 43 #define __NR_osf_profil 44 /* not implemented */ #define __NR_open 45 #define __NR_osf_old_sigaction 46 /* not implemented */ #define __NR_getxgid 47 #define __NR_osf_sigprocmask 48 #define __NR_osf_getlogin 49 /* not implemented */ #define __NR_osf_setlogin 50 /* not implemented */ #define __NR_acct 51 #define __NR_sigpending 52 #define __NR_ioctl 54 #define __NR_osf_reboot 55 /* not implemented */ #define __NR_osf_revoke 56 /* not implemented */ #define __NR_symlink 57 #define __NR_readlink 58 #define __NR_execve 59 #define __NR_umask 60 #define __NR_chroot 61 #define __NR_osf_old_fstat 62 /* not implemented */ #define __NR_getpgrp 63 #define __NR_getpagesize 64 #define __NR_osf_mremap 65 /* not implemented */ #define __NR_vfork 66 #define __NR_stat 67 #define __NR_lstat 68 #define __NR_osf_sbrk 69 /* not implemented */ #define __NR_osf_sstk 70 /* not implemented */ #define __NR_mmap 71 /* OSF/1 mmap is superset of Linux */ #define __NR_osf_old_vadvise 72 /* not implemented */ #define __NR_munmap 73 #define __NR_mprotect 74 #define __NR_madvise 75 #define __NR_vhangup 76 #define __NR_osf_kmodcall 77 /* not implemented */ #define __NR_osf_mincore 78 /* not implemented */ #define __NR_getgroups 79 #define __NR_setgroups 80 #define __NR_osf_old_getpgrp 81 /* not implemented */ #define __NR_setpgrp 82 /* BSD alias for setpgid */ #define __NR_osf_setitimer 83 #define __NR_osf_old_wait 84 /* not implemented */ #define __NR_osf_table 85 /* not implemented */ #define __NR_osf_getitimer 86 #define __NR_gethostname 87 #define __NR_sethostname 88 #define __NR_getdtablesize 89 #define __NR_dup2 90 #define __NR_fstat 91 #define __NR_fcntl 92 #define __NR_osf_select 93 #define __NR_poll 94 #define __NR_fsync 95 #define __NR_setpriority 96 #define __NR_socket 97 #define __NR_connect 98 #define __NR_accept 99 #define __NR_getpriority 100 ..........
|
【本文地址】
今日新闻 |
推荐新闻 |