ktap代码分析笔记 (1) —— ktap.c

ktap.c源码如下（省去版权信息）：

#include <linux/version.h>
#if LINUX_VERSION_CODE < KERNEL_VERSION(3, 1, 0)
#error "Currently ktap don't support kernel older than 3.1"
#endif

#if !CONFIG_EVENT_TRACING
#error "Please enable CONFIG_EVENT_TRACING before compile ktap"
#endif

#if !CONFIG_PERF_EVENTS
#error "Please enable CONFIG_PERF_EVENTS before compile ktap"
#endif

#include <linux/module.h>
#include <linux/errno.h>
#include <linux/file.h>
#include <linux/slab.h>
#include <linux/fcntl.h>
#include <linux/sched.h>
#include <linux/poll.h>
#include <linux/anon_inodes.h>
#include <linux/debugfs.h>
#include <linux/vmalloc.h>
#include "../include/ktap_types.h"
#include "ktap.h"
#include "kp_bcread.h"
#include "kp_vm.h"

/* common helper function */
long gettimeofday_ns(void)
{
    struct timespec now;

    getnstimeofday(&now);
    return now.tv_sec * NSEC_PER_SEC + now.tv_nsec;
}

static int load_trunk(ktap_option_t *parm, unsigned long **buff)
{
    int ret;
    unsigned long *vmstart;

    vmstart = vmalloc(parm->trunk_len);
    if (!vmstart)
        return -ENOMEM;

    ret = copy_from_user(vmstart, (void __user *)parm->trunk,
                 parm->trunk_len);
    if (ret < 0) {
        vfree(vmstart);
        return -EFAULT;
    }

    *buff = vmstart;
    return 0;
}

static struct dentry *kp_dir_dentry;

/* Ktap Main Entry */
static int ktap_main(struct file *file, ktap_option_t *parm)
{
    unsigned long *buff = NULL;
    ktap_state_t *ks;
    ktap_proto_t *pt;
    long start_time, delta_time;
    int ret;

    start_time = gettimeofday_ns();

    ks = kp_vm_new_state(parm, kp_dir_dentry);
    if (unlikely(!ks))
        return -ENOEXEC;

    file->private_data = ks;

    ret = load_trunk(parm, &buff);
    if (ret) {
        kp_error(ks, "cannot load file\n");
        goto out;
    }

    pt = kp_bcread(ks, (unsigned char *)buff, parm->trunk_len);

    vfree(buff);

    if (pt) {
        /* validate byte code */
        if (kp_vm_validate_code(ks, pt, ks->stack))
            goto out;

        delta_time = (gettimeofday_ns() - start_time) / NSEC_PER_USEC;
        kp_verbose_printf(ks, "booting time: %d (us)\n", delta_time);

        /* enter vm */
        kp_vm_call_proto(ks, pt);
    }

 out:
    kp_vm_exit(ks);
    return ret;
}


static void print_version(void)
{
}

static long ktap_ioctl(struct file *file, unsigned int cmd, unsigned long arg)
{
    ktap_option_t parm;
    int ret;

    switch (cmd) {
    case KTAP_CMD_IOC_VERSION:
        print_version();
        return 0;
    case KTAP_CMD_IOC_RUN:
        /*
         * must be root to run ktap script (at least for now)
         *
         * TODO: check perf_paranoid sysctl and allow non-root user
         * to use ktap for tracing process(like uprobe) ?
         */
        if (!capable(CAP_SYS_ADMIN))
            return -EACCES;

        ret = copy_from_user(&parm, (void __user *)arg,
                     sizeof(ktap_option_t));
        if (ret < 0)
            return -EFAULT;

        return ktap_main(file, &parm);
    default:
        return -EINVAL;
    };

        return 0;
}

static const struct file_operations ktap_fops = {
    .llseek                 = no_llseek,
    .unlocked_ioctl         = ktap_ioctl,
};

static long ktapvm_ioctl(struct file *file, unsigned int cmd, unsigned long arg)
{
    int new_fd, err;
    struct file *new_file;

    new_fd = get_unused_fd_flags(0);
    if (new_fd < 0)
        return new_fd;

    new_file = anon_inode_getfile("[ktap]", &ktap_fops, NULL, O_RDWR);
    if (IS_ERR(new_file)) {
        err = PTR_ERR(new_file);
        put_unused_fd(new_fd);
        return err;
    }

    file->private_data = NULL;
    fd_install(new_fd, new_file);
    return new_fd;
}

static const struct file_operations ktapvm_fops = {
    .owner  = THIS_MODULE,
    .unlocked_ioctl         = ktapvm_ioctl,
};

int (*kp_ftrace_profile_set_filter)(struct perf_event *event, int event_id,
                    const char *filter_str);

struct syscall_metadata **syscalls_metadata;

/*TODO: kill this function in future */
static int __init init_dummy_kernel_functions(void)
{
    unsigned long *addr;

    /*
     * ktap need symbol ftrace_profile_set_filter to set event filter, 
     * export it in future. 
     */
#ifdef CONFIG_PPC64
    kp_ftrace_profile_set_filter =
        (void *)kallsyms_lookup_name(".ftrace_profile_set_filter");
#else
    kp_ftrace_profile_set_filter =
        (void *)kallsyms_lookup_name("ftrace_profile_set_filter");
#endif
    if (!kp_ftrace_profile_set_filter) {
        pr_err("ktap: cannot lookup ftrace_profile_set_filter "
            "in kallsyms\n");
        return -1;
    }

    /* use syscalls_metadata for syscall event handling */
    addr = (void *)kallsyms_lookup_name("syscalls_metadata");
    if (!addr) {
        pr_err("ktap: cannot lookup syscalls_metadata in kallsyms\n");
        return -1;
    }

    syscalls_metadata = (struct syscall_metadata **)*addr;
    return 0;
}

static int __init init_ktap(void)
{
    struct dentry *ktapvm_dentry;

    if (init_dummy_kernel_functions())
        return -1;

    kp_dir_dentry = debugfs_create_dir("ktap", NULL);
    if (!kp_dir_dentry) {
        pr_err("ktap: debugfs_create_dir failed\n");
        return -1;
    }

    ktapvm_dentry = debugfs_create_file("ktapvm", 0444, kp_dir_dentry, NULL,
                        &ktapvm_fops);

    if (!ktapvm_dentry) {
        pr_err("ktapvm: cannot create ktapvm file\n");
        debugfs_remove_recursive(kp_dir_dentry);
        return -1;
    }

    return 0;
}

static void __exit exit_ktap(void)
{
    debugfs_remove_recursive(kp_dir_dentry);
}

module_init(init_ktap);
module_exit(exit_ktap);

MODULE_AUTHOR("Jovi Zhangwei <jovi.zhangwei@gmail.com>");
MODULE_DESCRIPTION("ktap");
MODULE_LICENSE("GPL");

int kp_max_loop_count = 100000;
module_param_named(max_loop_count, kp_max_loop_count, int, S_IRUGO | S_IWUSR);
MODULE_PARM_DESC(max_loop_count, "max loop execution count"); 

（1）

#include <linux/version.h>
#if LINUX_VERSION_CODE < KERNEL_VERSION(3, 1, 0)
#error "Currently ktap don't support kernel older than 3.1"
#endif

#if !CONFIG_EVENT_TRACING
#error "Please enable CONFIG_EVENT_TRACING before compile ktap"
#endif

#if !CONFIG_PERF_EVENTS
#error "Please enable CONFIG_PERF_EVENTS before compile ktap"
#endif

ktap目前只能在3.1.0以上的kernel版本上运行，并且要有perf的支持。

（2）

module_init(init_ktap);
module_exit(exit_ktap);

MODULE_AUTHOR("Jovi Zhangwei <jovi.zhangwei@gmail.com>");
MODULE_DESCRIPTION("ktap");
MODULE_LICENSE("GPL");

int kp_max_loop_count = 100000;
module_param_named(max_loop_count, kp_max_loop_count, int, S_IRUGO | S_IWUSR);
MODULE_PARM_DESC(max_loop_count, "max loop execution count"); 

以上定义了一些ktapvm.ko这个内核模块的一些信息。

（3）

static int __init init_dummy_kernel_functions(void)
{
    unsigned long *addr;

    /*
     * ktap need symbol ftrace_profile_set_filter to set event filter, 
     * export it in future. 
     */
#ifdef CONFIG_PPC64
    kp_ftrace_profile_set_filter =
        (void *)kallsyms_lookup_name(".ftrace_profile_set_filter");
#else
    kp_ftrace_profile_set_filter =
        (void *)kallsyms_lookup_name("ftrace_profile_set_filter");
#endif
    if (!kp_ftrace_profile_set_filter) {
        pr_err("ktap: cannot lookup ftrace_profile_set_filter "
            "in kallsyms\n");
        return -1;
    }

    /* use syscalls_metadata for syscall event handling */
    addr = (void *)kallsyms_lookup_name("syscalls_metadata");
    if (!addr) {
        pr_err("ktap: cannot lookup syscalls_metadata in kallsyms\n");
        return -1;
    }

    syscalls_metadata = (struct syscall_metadata **)*addr;
    return 0;
}

static int __init init_ktap(void)
{
    struct dentry *ktapvm_dentry;

    if (init_dummy_kernel_functions())
        return -1;

    kp_dir_dentry = debugfs_create_dir("ktap", NULL);
    if (!kp_dir_dentry) {
        pr_err("ktap: debugfs_create_dir failed\n");
        return -1;
    }

    ktapvm_dentry = debugfs_create_file("ktapvm", 0444, kp_dir_dentry, NULL,
                        &ktapvm_fops);

    if (!ktapvm_dentry) {
        pr_err("ktapvm: cannot create ktapvm file\n");
        debugfs_remove_recursive(kp_dir_dentry);
        return -1;
    }

    return 0;
}

static void __exit exit_ktap(void)
{
    debugfs_remove_recursive(kp_dir_dentry);
}

module_init(init_ktap);
module_exit(exit_ktap);

ktapvm.ko的加载函数init_ktap会创建/sys/kernel/debug/ktap/这个文件夹，并在下面创建ktapvm文件。kallsyms_lookup_name函数则用来查找指定symbol的内存地址，而syscalls_metadata在目前代码中并没有使用。卸载模块函数exit_ktap则会删除/sys/kernel/debug/ktap/这个文件夹以及文件夹下的所有文件。

（4）

static const struct file_operations ktapvm_fops = {
    .owner  = THIS_MODULE,
    .unlocked_ioctl         = ktapvm_ioctl,
};

以上代码定义了ktapvm.ko的核心函数：ktapvm_ioctl，对这个模块的所有的操作都是通过ktapvm_ioctl函数完成。

static long ktapvm_ioctl(struct file *file, unsigned int cmd, unsigned long arg)
{
    int new_fd, err;
    struct file *new_file;

    new_fd = get_unused_fd_flags(0);
    if (new_fd < 0)
        return new_fd;

    new_file = anon_inode_getfile("[ktap]", &ktap_fops, NULL, O_RDWR);
    if (IS_ERR(new_file)) {
        err = PTR_ERR(new_file);
        put_unused_fd(new_fd);
        return err;
    }

    file->private_data = NULL;
    fd_install(new_fd, new_file);
    return new_fd;
}

ktapvm模块只提供了一个ioctl函数：ktapvm_ioctl，而这个函数的作用就是生成一个[ktap]的文件描述符。

（6）

static long ktap_ioctl(struct file *file, unsigned int cmd, unsigned long arg)
{
    ktap_option_t parm;
    int ret;

    switch (cmd) {
    case KTAP_CMD_IOC_VERSION:
        print_version();
        return 0;
    case KTAP_CMD_IOC_RUN:
        /*
         * must be root to run ktap script (at least for now)
         *
         * TODO: check perf_paranoid sysctl and allow non-root user
         * to use ktap for tracing process(like uprobe) ?
         */
        if (!capable(CAP_SYS_ADMIN))
            return -EACCES;

        ret = copy_from_user(&parm, (void __user *)arg,
                     sizeof(ktap_option_t));
        if (ret < 0)
            return -EFAULT;

        return ktap_main(file, &parm);
    default:
        return -EINVAL;
    };

        return 0;
}

static const struct file_operations ktap_fops = {
    .llseek                 = no_llseek,
    .unlocked_ioctl         = ktap_ioctl,
};

[ktap]文件的ioctl操作函数是ktap_ioctl，它的核心功能就是从user-space空间拷贝ktap_option_t，然后调用ktap_main函数执行ktap命令。

（7）

/* common helper function */
long gettimeofday_ns(void)
{
    struct timespec now;

    getnstimeofday(&now);
    return now.tv_sec * NSEC_PER_SEC + now.tv_nsec;
}

static int load_trunk(ktap_option_t *parm, unsigned long **buff)
{
    int ret;
    unsigned long *vmstart;

    vmstart = vmalloc(parm->trunk_len);
    if (!vmstart)
        return -ENOMEM;

    ret = copy_from_user(vmstart, (void __user *)parm->trunk,
                 parm->trunk_len);
    if (ret < 0) {
        vfree(vmstart);
        return -EFAULT;
    }

    *buff = vmstart;
    return 0;
}

gettimeofday_ns返回当前的纳秒值，而load_trunk则从user-space空间拷贝ktap命令。

（8）

static struct dentry *kp_dir_dentry;

/* Ktap Main Entry */
static int ktap_main(struct file *file, ktap_option_t *parm)
{
    unsigned long *buff = NULL;
    ktap_state_t *ks;
    ktap_proto_t *pt;
    long start_time, delta_time;
    int ret;

    start_time = gettimeofday_ns();

    ks = kp_vm_new_state(parm, kp_dir_dentry);
    if (unlikely(!ks))
        return -ENOEXEC;

    file->private_data = ks;

    ret = load_trunk(parm, &buff);
    if (ret) {
        kp_error(ks, "cannot load file\n");
        goto out;
    }

    pt = kp_bcread(ks, (unsigned char *)buff, parm->trunk_len);

    vfree(buff);

    if (pt) {
        /* validate byte code */
        if (kp_vm_validate_code(ks, pt, ks->stack))
            goto out;

        delta_time = (gettimeofday_ns() - start_time) / NSEC_PER_USEC;
        kp_verbose_printf(ks, "booting time: %d (us)\n", delta_time);

        /* enter vm */
        kp_vm_call_proto(ks, pt);
    }

 out:
    kp_vm_exit(ks);
    return ret;
}

ktap_main是运行ktap命令的函数，关于其中调用其它函数的细节，会在其它部分讨论，此处不再赘述。