linux SYSCTL的原理分析

• 1./proc/sys      support

需要打开内核选项CONFIG_PROC_SYSCTL

fs/proc/root.c

void __init proc_root_init(void)

{

    ...

    proc_sys_init();    //注册/proc/sys

}

#ifdef CONFIG_PROC_SYSCTL

extern int proc_sys_init(void);

extern void sysctl_head_put(struct ctl_table_header *);

#else

static inline void proc_sys_init(void) { }

static inline void sysctl_head_put(struct ctl_table_header *head) { }

#endif

proc-$(CONFIG_PROC_SYSCTL)    += proc_sysctl.o

fs/proc/proc_sysctl.c

int __init proc_sys_init(void)

{

    struct proc_dir_entry *proc_sys_root;

    proc_sys_root = proc_mkdir(" sys ", NULL);

    proc_sys_root->proc_iops = &proc_sys_dir_operations;

    proc_sys_root->proc_fops = &proc_sys_dir_file_operations;

    proc_sys_root->nlink = 0;

    return sysctl_init ();

}

• 2.实现分析：

kernel/sysctl.c

#ifdef   CONFIG_SYSCTL

int __init sysctl_init(void)

{

    struct ctl_table_header *hdr;

    hdr = register_sysctl_table ( sysctl_base_table );

    kmemleak_not_leak(hdr);

    return 0;

}

#endif /* CONFIG_SYSCTL */

因此可以看出，CONFIG_SYSCTL实际上是在/proc/sys目录下多加一些入口

• 3.内核需要打开

CONFIG_SYSCTL  配置选项

• 4.sysctl命令

以前sysctl命令是通过sysctl系统调用实现的，不过新实现的都是通过直接读写proc文件系统实现的

sysctl系统调用的实现：

kernel/sysctl_binary.c

SYSCALL_DEFINE1(sysctl, struct __sysctl_args __user *, args)

{

    do_sysctl（）；

}

COMPAT_SYSCALL_DEFINE1(sysctl, struct compat_sysctl_args __user *, args)

{

    do_sysctl（）；

}

static ssize_t do_sysctl (int __user *args_name, int nlen,

    void __user *oldval, size_t oldlen, void __user *newval, size_t newlen)

{

....

    return binary_sysctl (name, nlen, oldval, oldlen, newval, newlen);

}

#ifdef CONFIG_SYSCTL_SYSCALL

binary_sysctl()

{

    pathname = sysctl_getname (name, nlen, &table);

    mnt = task_active_pid_ns(current)-> proc_mnt ;

    file = file_open_root (mnt->mnt_root, mnt, pathname, flags);

........

}

#else /* CONFIG_SYSCTL_SYSCALL */

static ssize_t binary_sysctl(const int *name, int nlen,

    void __user *oldval, size_t oldlen, void __user *newval, size_t newlen)

{

    return -ENOSYS;

}

#endif /* CONFIG_SYSCTL_SYSCALL */

static const struct bin_table *get_sysctl(const int *name, int nlen, char *path)

{

    const struct bin_table *table = & bin_root_table [0];

    int ctl_name;

    /* The binary sysctl tables have a small maximum depth so

     * there is no danger of overflowing our path as it PATH_MAX

     * bytes long.

     */

    memcpy(path, " sys /", 4);

    path += 4;

......

}

因此sysctl系统调用需要CONFIG_SYSCTL_SYSCALL打开，这个系统调用只是把传进来的name根据bin_root_table表转化为/proc/sys下面的路径，然后读写这些文件实现的，而

sysctl_init（）则是在/proc/sys下面加入很多入口

可以看出 sysctl系统调用也只是转化了一下路径而已，最终还是得操作/proc/sys下面的文件。

kernel/sysctl_binary.c中可以看到支持的配置目录：

static const struct bin_table bin_root_table [] = {

    { CTL_DIR,    CTL_KERN,    "kernel",     bin_kern_table },      //proc/sys/kernel

    { CTL_DIR,    CTL_VM,        "vm",        bin_vm_table },

    { CTL_DIR,    CTL_NET,    "net",        bin_net_table },             //proc/sys/net

    /* CTL_PROC not used */

    { CTL_DIR,    CTL_FS,        "fs",        bin_fs_table },

    /* CTL_DEBUG "debug" no longer used */

    { CTL_DIR,    CTL_DEV,    "dev",        bin_dev_table },

    { CTL_DIR,    CTL_BUS,    "bus",        bin_bus_table },

    { CTL_DIR,    CTL_ABI,    "abi" },

    /* CTL_CPU not used */

    /* CTL_ARLAN "arlan" no longer used */

    { CTL_DIR,    CTL_S390DBF,    "s390dbf",    bin_s390dbf_table },

    { CTL_DIR,    CTL_SUNRPC,    "sunrpc",    bin_sunrpc_table },

    { CTL_DIR,    CTL_PM,        "pm",        bin_pm_table },

    {}

};

static const struct bin_table bin_kern_table [] = {

    { CTL_STR,    KERN_OSTYPE,            "ostype" },

    { CTL_STR,    KERN_OSRELEASE,            "osrelease" },

    /* KERN_OSREV not used */

    { CTL_STR,    KERN_VERSION,            "version" },

    /* KERN_SECUREMASK not used */

    /* KERN_PROF not used */

    { CTL_STR,    KERN_NODENAME,            "hostname" },

    { CTL_STR,    KERN_DOMAINNAME,        "domainname" },

    { CTL_INT,    KERN_PANIC,            "panic" },

    { CTL_INT,    KERN_REALROOTDEV,        "real-root-dev" },

    { CTL_STR,    KERN_SPARC_REBOOT,        "reboot-cmd" },

    { CTL_INT,    KERN_CTLALTDEL,            " ctrl-alt-del " },

    { CTL_INT,    KERN_PRINTK,            "printk" },

    /* KERN_NAMETRANS not used */

    /* KERN_PPC_HTABRECLAIM not used */

    /* KERN_PPC_ZEROPAGED not used */

    { CTL_INT,    KERN_PPC_POWERSAVE_NAP,        "powersave-nap" },

    { CTL_STR,    KERN_MODPROBE,            "modprobe" },

    { CTL_INT,    KERN_SG_BIG_BUFF,        "sg-big-buff" },

    { CTL_INT,    KERN_ACCT,            "acct" },

    /* KERN_PPC_L2CR "l2cr" no longer used */

    /* KERN_RTSIGNR not used */

    /* KERN_RTSIGMAX not used */

    { CTL_ULONG,    KERN_SHMMAX,            "shmmax" },

    { CTL_INT,    KERN_MSGMAX,            "msgmax" },

    { CTL_INT,    KERN_MSGMNB,            "msgmnb" },

    /* KERN_MSGPOOL not used*/

    { CTL_INT,    KERN_SYSRQ,            "sysrq" },

    { CTL_INT,    KERN_MAX_THREADS,        "threads-max" },

    { CTL_DIR,    KERN_RANDOM,            "random",    bin_random_table },

    { CTL_ULONG,    KERN_SHMALL,            "shmall" },

    { CTL_INT,    KERN_MSGMNI,            "msgmni" },

    { CTL_INT,    KERN_SEM,            "sem" },

    { CTL_INT,    KERN_SPARC_STOP_A,        "stop-a" },

    { CTL_INT,    KERN_SHMMNI,            "shmmni" },

    { CTL_INT,    KERN_OVERFLOWUID,        "overflowuid" },

    { CTL_INT,    KERN_OVERFLOWGID,        "overflowgid" },

    { CTL_STR,    KERN_HOTPLUG,            "hotplug", },

    { CTL_INT,    KERN_IEEE_EMULATION_WARNINGS,    "ieee_emulation_warnings" },

    { CTL_INT,    KERN_S390_USER_DEBUG_LOGGING,    "userprocess_debug" },

    { CTL_INT,    KERN_CORE_USES_PID,        "core_uses_pid" },

    /* KERN_TAINTED "tainted" no longer used */

    { CTL_INT,    KERN_CADPID,            "cad_pid" },

    { CTL_INT,    KERN_PIDMAX,            "pid_max" },

    { CTL_STR,    KERN_CORE_PATTERN,        "core_pattern" },

    { CTL_INT,    KERN_PANIC_ON_OOPS,        "panic_on_oops" },

    { CTL_INT,    KERN_HPPA_PWRSW,        "soft-power" },

    { CTL_INT,    KERN_HPPA_UNALIGNED,        "unaligned-trap" },

    { CTL_INT,    KERN_PRINTK_RATELIMIT,        "printk_ratelimit" },

    { CTL_INT,    KERN_PRINTK_RATELIMIT_BURST,    "printk_ratelimit_burst" },

    { CTL_DIR,    KERN_PTY,            "pty",        bin_pty_table },

    { CTL_INT,    KERN_NGROUPS_MAX,        "ngroups_max" },

    { CTL_INT,    KERN_SPARC_SCONS_PWROFF,    "scons-poweroff" },

    /* KERN_HZ_TIMER "hz_timer" no longer used */

    { CTL_INT,    KERN_UNKNOWN_NMI_PANIC,        "unknown_nmi_panic" },

    { CTL_INT,    KERN_BOOTLOADER_TYPE,        "bootloader_type" },

    { CTL_INT,    KERN_RANDOMIZE,            "randomize_va_space" },

    { CTL_INT,    KERN_SPIN_RETRY,        "spin_retry" },

    /* KERN_ACPI_VIDEO_FLAGS "acpi_video_flags" no longer used */

    { CTL_INT,    KERN_IA64_UNALIGNED,        "ignore-unaligned-usertrap" },

    { CTL_INT,    KERN_COMPAT_LOG,        "compat-log" },

    { CTL_INT,    KERN_MAX_LOCK_DEPTH,        "max_lock_depth" },

    { CTL_INT,    KERN_PANIC_ON_NMI,        "panic_on_unrecovered_nmi" },

    { CTL_INT,    KERN_PANIC_ON_WARN,        "panic_on_warn" },

    {}

}