/* ptrace.c */
/* By Ross Biro 1/23/92 */
/* edited by Linus Torvalds */

#include <linux/head.h>
#include <linux/kernel.h>
#include <linux/sched.h>
#include <linux/mm.h>
#include <linux/errno.h>
#include <linux/ptrace.h>
#include <linux/user.h>

#include <asm/segment.h>
#include <asm/system.h>
#include <linux/debugreg.h>

/*
 * does not yet catch signals sent when the child dies.
 * in exit.c or in signal.c.
 */
//exit.c和signal.c当子进程退出时任然捕获不到信号
/* determines which flags the user has access to. */
/* 1 = access 0 = no access */
#define FLAG_MASK 0x00044dd5

/* set's the trap flag. */
#define TRAP_FLAG 0x100

/*
 * this is the number to subtract from the top of the stack. To find
 * the local frame.
 */
#define MAGICNUMBER 68

/* change a pid into a task struct. */
//获取指定pid的任务
static inline struct task_struct * get_task(int pid)
{
    int i;
    //遍历所有任务进程,查找指定进程号的进程
    for (i = 1; i < NR_TASKS; i++) {
        if (task[i] != NULL && (task[i]->pid == pid))
            return task[i];
    }
    return NULL;
}

/*
 * this routine will get a word off of the processes priviledged stack.
 * the offset is how far from the base addr as stored in the TSS.  
 * this routine assumes that all the priviledged stacks are in our
 * data space.
 */   
 //获取栈的偏移处的内容
static inline int get_stack_long(struct task_struct *task, int offset)
{
    unsigned char *stack;
    //获取到栈基址
    stack = (unsigned char *)task->tss.esp0;
    //计算偏移
    stack += offset;
    //取出该位置的内容
    return (*((int *)stack));
}

/*
 * this routine will put a word on the processes priviledged stack.
 * the offset is how far from the base addr as stored in the TSS.  
 * this routine assumes that all the priviledged stacks are in our
 * data space.
 */
 //设置栈指定偏移处的内容
static inline int put_stack_long(struct task_struct *task, int offset,
    unsigned long data)
{
    unsigned char * stack;
    //栈基址
    stack = (unsigned char *) task->tss.esp0;
    //偏移位置
    stack += offset;
    //设置内容
    *(unsigned long *) stack = data;
    return 0;
}

/*
 * This routine gets a long from any process space by following the page
 * tables. NOTE! You should check that the long isn't on a page boundary,
 * and that it is in the task area before calling this: this routine does
 * no checking.
 *
 * NOTE2! This uses "tsk->tss.cr3" even though we know it's currently always
 * zero. This routine shouldn't have to change when we make a better mm.
 */
 //获取指定任务的内存控制任意位置处的数据
static unsigned long get_long(struct task_struct * tsk,
    unsigned long addr)
{
    unsigned long page;

repeat:
    page = *PAGE_DIR_OFFSET(tsk->tss.cr3,addr);
    if (page & PAGE_PRESENT) {
        page &= PAGE_MASK;
        page += PAGE_PTR(addr);
        page = *((unsigned long *) page);
    }
    if (!(page & PAGE_PRESENT)) {
        do_no_page(0,addr,tsk,0);
        goto repeat;
    }
/* this is a hack for non-kernel-mapped video buffers and similar */
    if (page >= high_memory)
        return 0;
    page &= PAGE_MASK;
    page += addr & ~PAGE_MASK;
    return *(unsigned long *) page;
}

/*
 * This routine puts a long into any process space by following the page
 * tables. NOTE! You should check that the long isn't on a page boundary,
 * and that it is in the task area before calling this: this routine does
 * no checking.
 *
 * Now keeps R/W state of page so that a text page stays readonly
 * even if a debugger scribbles breakpoints into it.  -M.U-
 */
static void put_long(struct task_struct * tsk, unsigned long addr,
    unsigned long data)
{
    unsigned long page, pte = 0;
    int readonly = 0;

repeat:
    page = *PAGE_DIR_OFFSET(tsk->tss.cr3,addr);
    if (page & PAGE_PRESENT) {
        page &= PAGE_MASK;
        page += PAGE_PTR(addr);
        pte = page;
        page = *((unsigned long *) page);
    }
    if (!(page & PAGE_PRESENT)) {
        do_no_page(0 /* PAGE_RW */ ,addr,tsk,0);
        goto repeat;
    }
    if (!(page & PAGE_RW)) {
        if(!(page & PAGE_COW))
            readonly = 1;
        do_wp_page(PAGE_RW | PAGE_PRESENT,addr,tsk,0);
        goto repeat;
    }
/* this is a hack for non-kernel-mapped video buffers and similar */
    if (page >= high_memory)
        return;
/* we're bypassing pagetables, so we have to set the dirty bit ourselves */
    *(unsigned long *) pte |= (PAGE_DIRTY|PAGE_COW);
    page &= PAGE_MASK;
    page += addr & ~PAGE_MASK;
    *(unsigned long *) page = data;
    if(readonly) {
        *(unsigned long *) pte &=~ (PAGE_RW|PAGE_COW);
        invalidate();
    }
}

/*
 * This routine checks the page boundaries, and that the offset is
 * within the task area. It then calls get_long() to read a long.
 */
static int read_long(struct task_struct * tsk, unsigned long addr,
    unsigned long * result)
{
    unsigned long low,high;

if (addr > TASK_SIZE-sizeof(long))
        return -EIO;
    if ((addr & ~PAGE_MASK) > PAGE_SIZE-sizeof(long)) {
        low = get_long(tsk,addr & ~(sizeof(long)-1));
        high = get_long(tsk,(addr+sizeof(long)) & ~(sizeof(long)-1));
        switch (addr & (sizeof(long)-1)) {
            case 1:
                low >>= 8;
                low |= high << 24;
                break;
            case 2:
                low >>= 16;
                low |= high << 16;
                break;
            case 3:
                low >>= 24;
                low |= high << 8;
                break;
        }
        *result = low;
    } else
        *result = get_long(tsk,addr);
    return 0;
}

/*
 * This routine checks the page boundaries, and that the offset is
 * within the task area. It then calls put_long() to write a long.
 */
static int write_long(struct task_struct * tsk, unsigned long addr,
    unsigned long data)
{
    unsigned long low,high;

if (addr > TASK_SIZE-sizeof(long))
        return -EIO;
    if ((addr & ~PAGE_MASK) > PAGE_SIZE-sizeof(long)) {
        low = get_long(tsk,addr & ~(sizeof(long)-1));
        high = get_long(tsk,(addr+sizeof(long)) & ~(sizeof(long)-1));
        switch (addr & (sizeof(long)-1)) {
            case 0: /* shouldn't happen, but safety first */
                low = data;
                break;
            case 1:
                low &= 0x000000ff;
                low |= data << 8;
                high &= ~0xff;
                high |= data >> 24;
                break;
            case 2:
                low &= 0x0000ffff;
                low |= data << 16;
                high &= ~0xffff;
                high |= data >> 16;
                break;
            case 3:
                low &= 0x00ffffff;
                low |= data << 24;
                high &= ~0xffffff;
                high |= data >> 8;
                break;
        }
        put_long(tsk,addr & ~(sizeof(long)-1),low);
        put_long(tsk,(addr+sizeof(long)) & ~(sizeof(long)-1),high);
    } else
        put_long(tsk,addr,data);
    return 0;
}

asmlinkage int sys_ptrace(long request, long pid, long addr, long data)
{
    struct task_struct *child;
    struct user * dummy;
    int i;

dummy = NULL;

if (request == PTRACE_TRACEME) {
        /* are we already being traced? */
        if (current->flags & PF_PTRACED)
            return -EPERM;
        /* set the ptrace bit in the proccess flags. */
        current->flags |= PF_PTRACED;
        return 0;
    }
    if (pid == 1)        /* you may not mess with init */
        return -EPERM;
    if (!(child = get_task(pid)))
        return -ESRCH;
    if (request == PTRACE_ATTACH) {
        if (child == current)
            return -EPERM;
        if ((!child->dumpable || (current->uid != child->euid) ||
             (current->gid != child->egid)) && !suser())
            return -EPERM;
        /* the same process cannot be attached many times */
        if (child->flags & PF_PTRACED)
            return -EPERM;
        child->flags |= PF_PTRACED;
        if (child->p_pptr != current) {
            REMOVE_LINKS(child);
            child->p_pptr = current;
            SET_LINKS(child);
        }
        send_sig(SIGSTOP, child, 1);
        return 0;
    }
    if (!(child->flags & PF_PTRACED))
        return -ESRCH;
    if (child->state != TASK_STOPPED) {
        if (request != PTRACE_KILL)
            return -ESRCH;
    }
    if (child->p_pptr != current)
        return -ESRCH;

switch (request) {
    /* when I and D space are seperate, these will need to be fixed. */
        case PTRACE_PEEKTEXT: /* read word at location addr. */
        case PTRACE_PEEKDATA: {
            unsigned long tmp;
            int res;

res = read_long(child, addr, &tmp);
            if (res < 0)
                return res;
            res = verify_area(VERIFY_WRITE, (void *) data, sizeof(long));
            if (!res)
                put_fs_long(tmp,(unsigned long *) data);
            return res;
        }

/* read the word at location addr in the USER area. */
        case PTRACE_PEEKUSR: {
            unsigned long tmp;
            int res;

if ((addr & 3) || addr < 0 ||
                addr > sizeof(struct user) - 3)
                return -EIO;

res = verify_area(VERIFY_WRITE, (void *) data, sizeof(long));
            if (res)
                return res;
            tmp = 0;  /* Default return condition */
            if(addr < 17*sizeof(long)) {
              addr = addr >> 2; /* temporary hack. */

tmp = get_stack_long(child, sizeof(long)*addr - MAGICNUMBER);
              if (addr == DS || addr == ES ||
                  addr == FS || addr == GS ||
                  addr == CS || addr == SS)
                tmp &= 0xffff;
            };
            if(addr >= (long) &dummy->u_debugreg[0] &&
               addr <= (long) &dummy->u_debugreg[7]){
                addr -= (long) &dummy->u_debugreg[0];
                addr = addr >> 2;
                tmp = child->debugreg[addr];
            };
            put_fs_long(tmp,(unsigned long *) data);
            return 0;
        }

/* when I and D space are seperate, this will have to be fixed. */
        case PTRACE_POKETEXT: /* write the word at location addr. */
        case PTRACE_POKEDATA:
            return write_long(child,addr,data);

case PTRACE_POKEUSR: /* write the word at location addr in the USER area */
            if ((addr & 3) || addr < 0 ||
                addr > sizeof(struct user) - 3)
                return -EIO;

addr = addr >> 2; /* temproary hack. */

if (addr == ORIG_EAX)
                return -EIO;
            if (addr == DS || addr == ES ||
                addr == FS || addr == GS ||
                addr == CS || addr == SS) {
                    data &= 0xffff;
                    if (data && (data & 3) != 3)
                    return -EIO;
            }
            if (addr == EFL) {   /* flags. */
                data &= FLAG_MASK;
                data |= get_stack_long(child, EFL*sizeof(long)-MAGICNUMBER)  & ~FLAG_MASK;
            }
          /* Do not allow the user to set the debug register for kernel
             address space */
          if(addr < 17){
              if (put_stack_long(child, sizeof(long)*addr-MAGICNUMBER, data))
                return -EIO;
            return 0;
            };

/* We need to be very careful here.  We implicitly
             want to modify a portion of the task_struct, and we
             have to be selective about what portions we allow someone
             to modify. */

addr = addr << 2;  /* Convert back again */
          if(addr >= (long) &dummy->u_debugreg[0] &&
             addr <= (long) &dummy->u_debugreg[7]){

if(addr == (long) &dummy->u_debugreg[4]) return -EIO;
              if(addr == (long) &dummy->u_debugreg[5]) return -EIO;
              if(addr < (long) &dummy->u_debugreg[4] &&
                 ((unsigned long) data) >= 0xbffffffd) return -EIO;
              
              if(addr == (long) &dummy->u_debugreg[7]) {
                  data &= ~DR_CONTROL_RESERVED;
                  for(i=0; i<4; i++)
                      if ((0x5f54 >> ((data >> (16 + 4*i)) & 0xf)) & 1)
                          return -EIO;
              };

addr -= (long) &dummy->u_debugreg;
              addr = addr >> 2;
              child->debugreg[addr] = data;
              return 0;
          };
          return -EIO;

case PTRACE_SYSCALL: /* continue and stop at next (return from) syscall */
        case PTRACE_CONT: { /* restart after signal. */
            long tmp;

if ((unsigned long) data > NSIG)
                return -EIO;
            if (request == PTRACE_SYSCALL)
                child->flags |= PF_TRACESYS;
            else
                child->flags &= ~PF_TRACESYS;
            child->exit_code = data;
            child->state = TASK_RUNNING;
    /* make sure the single step bit is not set. */
            tmp = get_stack_long(child, sizeof(long)*EFL-MAGICNUMBER) & ~TRAP_FLAG;
            put_stack_long(child, sizeof(long)*EFL-MAGICNUMBER,tmp);
            return 0;
        }

/*
 * make the child exit.  Best I can do is send it a sigkill.
 * perhaps it should be put in the status that it want's to
 * exit.
 */
        case PTRACE_KILL: {
            long tmp;

child->state = TASK_RUNNING;
            child->exit_code = SIGKILL;
    /* make sure the single step bit is not set. */
            tmp = get_stack_long(child, sizeof(long)*EFL-MAGICNUMBER) & ~TRAP_FLAG;
            put_stack_long(child, sizeof(long)*EFL-MAGICNUMBER,tmp);
            return 0;
        }

case PTRACE_SINGLESTEP: {  /* set the trap flag. */
            long tmp;

if ((unsigned long) data > NSIG)
                return -EIO;
            child->flags &= ~PF_TRACESYS;
            tmp = get_stack_long(child, sizeof(long)*EFL-MAGICNUMBER) | TRAP_FLAG;
            put_stack_long(child, sizeof(long)*EFL-MAGICNUMBER,tmp);
            child->state = TASK_RUNNING;
            child->exit_code = data;
    /* give it a chance to run. */
            return 0;
        }

case PTRACE_DETACH: { /* detach a process that was attached. */
            long tmp;

if ((unsigned long) data > NSIG)
                return -EIO;
            child->flags &= ~(PF_PTRACED|PF_TRACESYS);
            child->state = TASK_RUNNING;
            child->exit_code = data;
            REMOVE_LINKS(child);
            child->p_pptr = child->p_opptr;
            SET_LINKS(child);
            /* make sure the single step bit is not set. */
            tmp = get_stack_long(child, sizeof(long)*EFL-MAGICNUMBER) & ~TRAP_FLAG;
            put_stack_long(child, sizeof(long)*EFL-MAGICNUMBER,tmp);
            return 0;
        }

default:
            return -EIO;
    }
}

//系统调用
asmlinkage void syscall_trace(void)
{
    if ((current->flags & (PF_PTRACED|PF_TRACESYS))
            != (PF_PTRACED|PF_TRACESYS))
        return;
    current->exit_code = SIGTRAP;
    current->state = TASK_STOPPED;
    notify_parent(current);
    schedule();
    /*
     * this isn't the same as continuing with a signal, but it will do
     * for normal use.  strace only continues with a signal if the
     * stopping signal is not SIGTRAP.  -brl
     */
    if (current->exit_code)
        current->signal |= (1 << (current->exit_code - 1));
    current->exit_code = 0;
}

kernel/ptrace.c的更多相关文章

  1. Playing with ptrace, Part II

    Playing with ptrace, Part II Issue From Issue # December Dec , By Pradeep Padala inSysAdmin In Part ...

  2. linux下 玩转ptrace

    译者序:在开发Hust Online Judge的过程中,查阅了不少资料,关于调试器技术的资料在网上是很少,即便是UNIX编程巨著<UNIX环境高级编程>中,相关内容也不多,直到我在 ht ...

  3. 玩转ptrace(转)

    下面是转帖的内容,写的很详细.但是不同的linux发行版中头文件的路径和名称并不相同.如在某些发行版中<linux/user.h>就不存在,其中定义的变量出现在<asm/ptrace ...

  4. 由于ptrace.h文件导致的内核编译出错的解决方法

    arch/x86/kernel/ptrace.c:1472:17: error: conflicting types for 'syscall_trace_enter'  In file includ ...

  5. CVE-2019-13272 Linux kernel 权限许可和访问控制问题漏洞

    漏洞简介: Linuxkernel是美国Linux基金会发布的开源操作系统Linux所使用的内核. Linuxkernel5.1.17之前版本中存在安全漏洞,该漏洞源于kernel/ptrace.c文 ...

  6. ptrace函数深入分析

    ptrace函数:进程跟踪. 形式:#include<sys/ptrace.h> Int ptrace(int request,int pid,int addr,int data); 概述 ...

  7. LInux_System_Call_INT_80h

    Int 80h Website (Copy from Linux-System-Call) List of Linux/i386 system calls Copyright (C) 1999-200 ...

  8. Process Kill Technology && Process Protection Against In Linux

    目录 . 引言 . Kill Process By Kill Command && SIGNAL . Kill Process By Resource Limits . Kill Pr ...

  9. systemtap 列出所有linux 内核模块与相关函数0

    diskiohttp://blog.163.com/digoal%40126/blog/static/16387704020131015105532435/ [root@localhost linux ...

随机推荐

  1. Visor 隐藏应用之一 CSS3 生成器

    Visor 有很多隐藏功能,这些功能并不是系统的核心功能,只是为了测试一些技术,做了些对初学前端的工程师还算有用的功能.今天就介绍一下CSS3 生成器. 先介绍一下Note 组件: Note是基本图形 ...

  2. HDU 4067 hdoj 4067 Random Maze 最小费用流

    给出n个点,m条边,入口s和出口t,对于每条边有两个值a,b,如果保留这条边需要花费:否则,移除这条边需要花费b. 题目要求用最小费用构造一个有向图满足以下条件: 1.只有一个入口和出口 2.所有路都 ...

  3. HTML+CSS--继续学习

    为网页中的文字设置字体为宋体. body{font-family:"宋体";} 文字以斜体样式在浏览器中显示: p a{font-style:italic;} 设置文字以粗体样式显 ...

  4. js动画之获取元素属性

    首先我们要介绍一些知识 offsetWidth element.offsetWidth = width + padding + border; width 我们也知道element.style.wid ...

  5. DotSpatial 删除图层要素

    //添加图层后,定义图层,并获取图层 //遍历要素,并进行删除 FeatureSet fs = null; fs = (FeatureSet) map1.Layers[0].DataSet; //要素 ...

  6. tomcat+nginx简单实现负载均衡

    1.环境准备 在前面的博客中我已经安装好nginx和一台tomcat了.现在就在加一台tomcat tomcat1:  /apps/tomcat/tomcat1/apache-tomcat-7.0.6 ...

  7. matlab初学之strcat、num2str

    文章出处: http://blog.sina.com.cn/s/blog_6fb8aa0d01019id5.html http://wenda.so.com/q/1439143662729624 ht ...

  8. CIImage实现滤镜效果

    Core Image also provides autoadjustment methods that analyze an image for common deficiencies and re ...

  9. mySQL基本操作学习笔记(一)

                                                                                                        ...

  10. docker push 实现过程

    这一篇文章分析一下docker push的过程:docker push是将本地的镜像上传到registry service的过程: 根据前几篇文章,可以知道客户端的命令是在api/client/pus ...