《linux 内核完全剖析》 exit.c 代码分析笔记

 

  exit.c 代码分析笔记

 release

          释放进程的函数release() 主要根据指定进程的任务数据结构指针，在任务数组中删除指定的进程指针，释放相关内存页，并立刻让内核重新调度进程的运行。

void release(struct task_struct * p) //释放p指向的进程
{
int i;

if (!p) //常规检测p是否为0
return;
if (p == current) { //不能把自己给释放了
printk("task releasing itself\n\r");
return;
}
for (i=1 ; i<NR_TASKS ; i++) //扫描所有的进程
if (task[i]==p) { //找出p进程
task[i]=NULL; //把p置空
/* Update links */
if (p->p_osptr) //调整链表
p->p_osptr->p_ysptr = p->p_ysptr;
if (p->p_ysptr)
p->p_ysptr->p_osptr = p->p_osptr;
else
p->p_pptr->p_cptr = p->p_osptr;
free_page((long)p); //释放p进程占用的内存页
schedule();//任务调度
return;
}
panic("trying to release non-existent task");
}

bad_task_ptr 和audit_ptree

#ifdef DEBUG_PROC_TREE //下面这部分代码是调试用的
/*
* Check to see if a task_struct pointer is present in the task[] array
* Return 0 if found, and 1 if not found.
*/
int bad_task_ptr(struct task_struct *p)
{
int     i;

if (!p)
return 0;
for (i=0 ; i<NR_TASKS ; i++)
if (task[i] == p)
return 0;
return 1; //如果p不在task数组里面，那这个指针就有问题！
}

/*
* This routine scans the pid tree and make sure the rep invarient still
* holds.  Used for debugging only, since it's very slow....
*
* It looks a lot scarier than it really is.... we're doing ænothing more
* than verifying the doubly-linked list foundæin p_ysptr and p_osptr,
* and checking it corresponds with the process tree defined by p_cptr and
* p_pptr;
*/
void audit_ptree() //其实不难，就是打印进程树的相关信息，方便调试用
{
int    i;

for (i=1 ; i<NR_TASKS ; i++) {
if (!task[i])
continue;
if (bad_task_ptr(task[i]->p_pptr))
printk("Warning, pid %d's parent link is bad\n",
task[i]->pid);
if (bad_task_ptr(task[i]->p_cptr))
printk("Warning, pid %d's child link is bad\n",
task[i]->pid);
if (bad_task_ptr(task[i]->p_ysptr))
printk("Warning, pid %d's ys link is bad\n",
task[i]->pid);
if (bad_task_ptr(task[i]->p_osptr))
printk("Warning, pid %d's os link is bad\n",
task[i]->pid);
if (task[i]->p_pptr == task[i])
printk("Warning, pid %d parent link points to self\n");
if (task[i]->p_cptr == task[i])
printk("Warning, pid %d child link points to self\n");
if (task[i]->p_ysptr == task[i])
printk("Warning, pid %d ys link points to self\n");
if (task[i]->p_osptr == task[i])
printk("Warning, pid %d os link points to self\n");
if (task[i]->p_osptr) {
if (task[i]->p_pptr != task[i]->p_osptr->p_pptr)
printk(
"Warning, pid %d older sibling %d parent is %d\n",
task[i]->pid, task[i]->p_osptr->pid,
task[i]->p_osptr->p_pptr->pid);
if (task[i]->p_osptr->p_ysptr != task[i])
printk(
"Warning, pid %d older sibling %d has mismatched ys link\n",
task[i]->pid, task[i]->p_osptr->pid);
}
if (task[i]->p_ysptr) {
if (task[i]->p_pptr != task[i]->p_ysptr->p_pptr)
printk(
"Warning, pid %d younger sibling %d parent is %d\n",
task[i]->pid, task[i]->p_osptr->pid,
task[i]->p_osptr->p_pptr->pid);
if (task[i]->p_ysptr->p_osptr != task[i])
printk(
"Warning, pid %d younger sibling %d has mismatched os link\n",
task[i]->pid, task[i]->p_ysptr->pid);
}
if (task[i]->p_cptr) {
if (task[i]->p_cptr->p_pptr != task[i])
printk(
"Warning, pid %d youngest child %d has mismatched parent link\n",
task[i]->pid, task[i]->p_cptr->pid);
if (task[i]->p_cptr->p_ysptr)
printk(
"Warning, pid %d youngest child %d has non-NULL ys link\n",
task[i]->pid, task[i]->p_cptr->pid);
}
}
}
#endif /* DEBUG_PROC_TREE */

send_sig

static inline int send_sig(long sig,struct task_struct * p,int priv) //给进程p，发送信号。priv 是强制发送信号的标识
{
if (!p) //常规的p非空检测
return -EINVAL;
if (!priv && (current->euid!=p->euid) && !suser())
//如果不具有超级用户权限，又不是当前session里面的进程，且没有强制发送信号，进入if，return
return -EPERM;
if ((sig == SIGKILL) || (sig == SIGCONT)) { //如果要发送的信号是SIGKILL或者SIGCONT
if (p->state == TASK_STOPPED) //如果当前进程处于stop状态，则将其置于TASK_RUNNING状态
p->state = TASK_RUNNING;
p->exit_code = 0;
p->signal &= ~( (1<<(SIGSTOP-1)) | (1<<(SIGTSTP-1)) |
(1<<(SIGTTIN-1)) | (1<<(SIGTTOU-1)) ); //消除SIGSTOP SIGTSTP SIGTTIN SIGTTOU
}
/* If the signal will be ignored, don't even post it */
if ((int) p->sigaction[sig-1].sa_handler == 1) //如果handler是 ignore 就不要送信号鸟。。
return 0;
/* Depends on order SIGSTOP, SIGTSTP, SIGTTIN, SIGTTOU */
if ((sig >= SIGSTOP) && (sig <= SIGTTOU))
//如果信号含有 SIGSTOP, SIGTSTP, SIGTTIN, SIGTTOU其中任何信号，那么就消除SIGCONT信号
p->signal &= ~(1<<(SIGCONT-1));
/* Actually deliver the signal */
p->signal |= (1<<(sig-1)); //最后这里才把信号写入信号变量
return 0;
}

session_of_pgrp

int session_of_pgrp(int pgrp) //获取process group的session ID ，没有找到返回-1
{
struct task_struct **p;

for (p = &LAST_TASK ; p > &FIRST_TASK ; --p)
if ((*p)->pgrp == pgrp)
return((*p)->session);
return -1;
}

kill_pg

int kill_pg(int pgrp, int sig, int priv) //给指定的进程组发送信号
{
struct task_struct **p;
int err,retval = -ESRCH; //指定的进程不存在
int found = 0;

if (sig<1 || sig>32 || pgrp<=0)
return -EINVAL;
for (p = &LAST_TASK ; p > &FIRST_TASK ; --p)
if ((*p)->pgrp == pgrp) { //对每一个属于process group的进程发送信号sig
if (sig && (err = send_sig(sig,*p,priv)))
retval = err;
else
found++;
}
return(found ? 0 : retval);
}

kill_proc

int kill_proc(int pid, int sig, int priv) //给指定进程发送信号
{
struct task_struct **p;

if (sig<1 || sig>32)
return -EINVAL;
for (p = &LAST_TASK ; p > &FIRST_TASK ; --p)
if ((*p)->pid == pid)
return(sig ? send_sig(sig,*p,priv) : 0);
return(-ESRCH);
}

sys_kill

        系统调用sys_kill 用于项进程发送任何指定的信号，根据参数pid不同的数值，该系统调用会向不同的进或进程组发送信号。

/*
* POSIX specifies that kill(-1,sig) is unspecified, but what we have
* is probably wrong.  Should make it like BSD or SYSV.
*/
int sys_kill(int pid,int sig)
{
struct task_struct **p = NR_TASKS + task;
int err, retval = 0;

if (!pid) //如果是进程0 init ，用最高权限发送信号，将信号发送给当前进程所处进程组的所有进程
return(kill_pg(current->pid,sig,0));
if (pid == -1) { //如果 pid是-1，把信号发送到除了init进程外的所有进程！
while (--p > &FIRST_TASK)
if (err = send_sig(sig,*p,0))
retval = err;
return(retval);
}
if (pid < 0) //如果pid<0，发送到|pid| 所处进程组的所有进程
return(kill_pg(-pid,sig,0));
/* Normal kill */
return(kill_proc(pid,sig,0)); //否则，发送到进程pid
}

is_orphaned_pgrp

关于孤儿进程组的探讨

/*
* Determine if a process group is "orphaned", according to the POSIX
* definition in 2.2.2.52.  Orphaned process groups are not to be affected
* by terminal-generated stop signals.  Newly orphaned process groups are
* to receive a SIGHUP and a SIGCONT.
*
* "I ask you, have you ever known what it is to be an orphan?"
*/
int is_orphaned_pgrp(int pgrp) //判断是否为一个孤儿进程组
{
struct task_struct **p;

for (p = &LAST_TASK ; p > &FIRST_TASK ; --p) {
if (!(*p) || //如果进程不存在，下一个
((*p)->pgrp != pgrp) ||  // 如果进程所处的进程组不是pgrp，下一个
((*p)->state == TASK_ZOMBIE) || // 进程的状态是zombie，下一个
((*p)->p_pptr->pid == 1)) //进程parent是init ,下一个
continue;
if (((*p)->p_pptr->pgrp != pgrp) && //如果父进程所在的组不在pgrp，但是父进程所在的session存在
((*p)->p_pptr->session == (*p)->session))
return 0;
}
return(1);    /* (sighing) "Often!" */
}

has_stopped_jobs

static int has_stopped_jobs(int pgrp) //判断进程组内是否有进程处于TASK_STOPPED状态
{
struct task_struct ** p;

for (p = &LAST_TASK ; p > &FIRST_TASK ; --p) {
if ((*p)->pgrp != pgrp)
continue;
if ((*p)->state == TASK_STOPPED)
return(1);
}
return(0);
}

do_exit

volatile void do_exit(long code)
{
struct task_struct *p;
int i;

free_page_tables(get_base(current->ldt[1]),get_limit(0x0f));
free_page_tables(get_base(current->ldt[2]),get_limit(0x17));
/*这里采用的两个参数0x0f即1111B，因此他指向的段优先级为3，存储在LDT表中，索引为1。也就是当前任务段的代码段描述符。

另外一个参数0x17即10111B，因此他指向的段优先级为3，存储在LDT表中，索引为2.也就是当前任务的数及堆栈段描述符。

这样，通过get_limit(0x0f)和get_limit(0x17)就得到了当前任务代码段和数据堆栈段的长度。

因此，
free_page_tables(get_base(current->ldt[1]),get_limit(0x0f));
free_page_tables(get_base(current->ldt[2]),get_limit(0x17));就释放了当前任务的代码段和数据堆栈段。*/
for (i=0 ; i<NR_OPEN ; i++)
if (current->filp[i]) //filp文件结构指针表，filp[i]就是文件i被打开的ID
sys_close(i); //关闭进程当前打开的文件
iput(current->pwd); //和fs inode有关系
current->pwd = NULL; //当前进程的工作目录置为NULL
iput(current->root);//和fs inode有关系
current->root = NULL;//当前进程的根目录置为NULL
iput(current->executable);//和fs inode有关系
current->executable = NULL;//当前进程的执行程序文件的i节点置为NULL
iput(current->library);//和fs inode有关系
current->library = NULL;//当前进程的库文件置为NULL
current->state = TASK_ZOMBIE; //状态置为TASK_ZOMBIE
current->exit_code = code; //设置进程退出码
/*
* Check to see if any process groups have become orphaned
* as a result of our exiting, and if they have any stopped
* jobs, send them a SIGUP and then a SIGCONT.  (POSIX 3.2.2.2)
*
* Case i: Our father is in a different pgrp than we are
* and we were the only connection outside, so our pgrp
* is about to become orphaned.
*/
if ((current->p_pptr->pgrp != current->pgrp) &&
(current->p_pptr->session == current->session) &&
is_orphaned_pgrp(current->pgrp) &&
has_stopped_jobs(current->pgrp)) { //如果当前进程组是个orphaned group
kill_pg(current->pgrp,SIGHUP,1);
//用于只是这些进程和当前session 断开联系，我这里不是很明白为什么要发送SIGHUP 和 SIGCONT
kill_pg(current->pgrp,SIGCONT,1);
}
/* Let father know we died */
current->p_pptr->signal |= (1<<(SIGCHLD-1)); // 告诉parent process 当前进程要挂掉鸟

/*
* This loop does two things:
*
* A.  Make init inherit all the child processes
* B.  Check to see if any process groups have become orphaned
*    as a result of our exiting, and if they have any stopped
*    jons, send them a SIGUP and then a SIGCONT.  (POSIX 3.2.2.2)
*/
if (p = current->p_cptr) { //如果当前进程的child process 非空，进入if
while (1) {
p->p_pptr = task[1]; //让init进程领养current 进程的child 进程
if (p->state == TASK_ZOMBIE) //p进程沦为zombie
task[1]->signal |= (1<<(SIGCHLD-1)); //向init进程发送SIGCHLD信号
/*
* process group orphan check
* Case ii: Our child is in a different pgrp
* than we are, and it was the only connection
* outside, so the child pgrp is now orphaned.
*/
if ((p->pgrp != current->pgrp) &&
(p->session == current->session) &&
is_orphaned_pgrp(p->pgrp) &&
has_stopped_jobs(p->pgrp)) {
//判断p进程所在进程组是否是orphaned group，如果是。发送SIGHUP和SIGCONT信号
kill_pg(p->pgrp,SIGHUP,1);
kill_pg(p->pgrp,SIGCONT,1);
}
if (p->p_osptr) { //如果p_osptr非空，则把p->p_osptr赋值给p
p = p->p_osptr;
continue; //下一轮循环，知道p_osptr为NULL
}
/*
* This is it; link everything into init's children
* and leave
*/
p->p_osptr = task[1]->p_cptr;
task[1]->p_cptr->p_ysptr = p;
task[1]->p_cptr = current->p_cptr; //这三句把这个child process —— p 更新为 init 进程的 最年轻的child process
current->p_cptr = 0; //处理完所有的child process ，把 当前进程的p_cptr置为NULL
break; //处理完了跳出while
}
}
if (current->leader) { //如果当前进程是session leader
struct task_struct **p;
struct tty_struct *tty;

if (current->tty >= 0) { //切断当前进程和terminal 的联系
tty = TTY_TABLE(current->tty); //tty 还没看布吉岛。。。
if (tty->pgrp>0)
kill_pg(tty->pgrp, SIGHUP, 1);
tty->pgrp = 0;
tty->session = 0;
}
for (p = &LAST_TASK ; p > &FIRST_TASK ; --p) //把所有和当前进程同一个session的进程的tty都置为-1
if ((*p)->session == current->session)
(*p)->tty = -1;
}
if (last_task_used_math == current)
//数字协处理器木有看。。。布吉岛，
//但是语句的意思是上次用过协处理器的进程是当前进程的话酒吧last_task_used_math 置为NULL
last_task_used_math = NULL;
#ifdef DEBUG_PROC_TREE
audit_ptree();
#endif
schedule(); //进程调度
}

sys_exit

int sys_exit(int error_code) //系统调用sys_exit函数，通过调用do_exit 函数来实现
{
do_exit((error_code&0xff)<<8);
//error_code是用户程序提供的退出状态信息，只有低字节有效
//把error_code 左移八位是空出第八位供wait()和waitpid()函数使用
}

sys_waitpid

             用于挂起当前进程，知道pid指定的进程退出或者收到要求该进程终止的信号，或者是需要调用一个signal handler。如果pid 所指的进程早已退出(zombie)则本调用立刻返回。子进程的所有资源将被释放。

int sys_waitpid(pid_t pid,unsigned long * stat_addr, int options)

//其实我一般都木有怎么用stat_addr 这边选项，一般我都NULL了 <—_<—
//waitpid居然放在。。。exit.c 里，木有别的，只有惊奇
{
int flag;
struct task_struct *p;
unsigned long oldblocked;

verify_area(stat_addr,4); //验证stat_addr 地址处4byte 可写
repeat:
flag=0;
for (p = current->p_cptr ; p ; p = p->p_osptr) {
if (pid>0) {
if (p->pid != pid) //找到pid描述的child进程的指针
continue;
} else if (!pid) { //如果pid等于0， 找到pid 所在的进程组
if (p->pgrp != current->pgrp)
continue;
} else if (pid != -1) {   //如果pid < -1 ，找到|pid|所在的进程组
if (p->pgrp != -pid)
continue;
}
// 如果以上判断田间都没进if去，pid = -1 ，这里不是很明白，看完fork之后回头再update
switch (p->state) {
case TASK_STOPPED: //如果进程p处于TASK_STOPPED状态
if (!(options & WUNTRACED) ||  //如果option的WUNTRACED 没有置位或者 exit_code 等于0 ,对下一个older child进行检测
!p->exit_code)
continue;
put_fs_long((p->exit_code << 8) | 0x7f,
stat_addr);//把第一个参数写入第二个参数指向的地址
p->exit_code = 0; //正常退出
return p->pid; //返回结束child 的pid
case TASK_ZOMBIE:
current->cutime += p->utime;
current->cstime += p->stime; //更新时间参数
flag = p->pid;
put_fs_long(p->exit_code, stat_addr); //把p->exit_code 写入stat_addr
release(p); //释放进程p占用的内存页
#ifdef DEBUG_PROC_TREE
audit_ptree();
#endif
return flag;
default:
flag=1;
continue;
}
}
if (flag) { //存在不处于TASK_STOPPED 和TASK_ZOMBIE状态的进程
if (options & WNOHANG) //如果WNOHANG(表示若没有子进程处于退出或者终止状态就返回)在options中置位，就立刻返回0；
return 0; //一般我是不会用WNOHANG的。。。坦白的说我都没用过
current->state=TASK_INTERRUPTIBLE; //更新进程状态，置为TASK_INTERRUPTIBLE
oldblocked = current->blocked; //当前进程阻塞的信号储存于oldblocked中
current->blocked &= ~(1<<(SIGCHLD-1)); //消除当前进程中的SIGCHLD信号
schedule();//进程调度
current->blocked = oldblocked; //回复原来的信号图
if (current->signal & ~(current->blocked | (1<<(SIGCHLD-1))))//如果收到了除了SIGCHLD之外的其他信号，return 挂掉
return -ERESTARTSYS;
else
goto repeat; //否则goto到repeat 继续重复，直到return 出去。。。。。原来是这样waitpid的啊！
}
return -ECHILD; //没有找到pid对应的child process，返回错误码
}