kernel/exit.c

/* */
This source file includes following definitions.
generate
send_sig
notify_parent
release
bad_task_ptr
audit_ptree
session_of_pgrp
kill_pg
kill_sl
kill_proc
sys_kill
is_orphaned_pgrp
has_stopped_jobs
forget_original_parent
do_exit
sys_exit
sys_wait4
sys_waitpid
   1 /*
   2  *  linux/kernel/exit.c
   3  *
   4  *  Copyright (C) 1991, 1992  Linus Torvalds
   5  */
   6 
   7 #define DEBUG_PROC_TREE
   8 
   9 #include <linux/wait.h>
  10 #include <linux/errno.h>
  11 #include <linux/signal.h>
  12 #include <linux/sched.h>
  13 #include <linux/kernel.h>
  14 #include <linux/resource.h>
  15 #include <linux/mm.h>
  16 #include <linux/tty.h>
  17 
  18 #include <asm/segment.h>
  19 extern void shm_exit (void);
  20 extern void sem_exit (void);
  21 
  22 int getrusage(struct task_struct *, int, struct rusage *);
  23 
  24 static int generate(unsigned long sig, struct task_struct * p)
     /*  */
  25 {
  26         unsigned long mask = 1 << (sig-1);
  27         struct sigaction * sa = sig + p->sigaction - 1;
  28 
  29         /* always generate signals for traced processes ??? */
  30         if (p->flags & PF_PTRACED) {
  31                 p->signal |= mask;
  32                 return 1;
  33         }
  34         /* don't bother with ignored signals (but SIGCHLD is special) */
  35         if (sa->sa_handler == SIG_IGN && sig != SIGCHLD)
  36                 return 0;
  37         /* some signals are ignored by default.. (but SIGCONT already did its deed) */
  38         if ((sa->sa_handler == SIG_DFL) &&
  39             (sig == SIGCONT || sig == SIGCHLD || sig == SIGWINCH))
  40                 return 0;
  41         p->signal |= mask;
  42         return 1;
  43 }
  44 
  45 int send_sig(unsigned long sig,struct task_struct * p,int priv)
     /*  */
  46 {
  47         if (!p || sig > 32)
  48                 return -EINVAL;
  49         if (!priv && ((sig != SIGCONT) || (current->session != p->session)) &&
  50             (current->euid != p->euid) && (current->uid != p->uid) && !suser())
  51                 return -EPERM;
  52         if (!sig)
  53                 return 0;
  54         if ((sig == SIGKILL) || (sig == SIGCONT)) {
  55                 if (p->state == TASK_STOPPED)
  56                         p->state = TASK_RUNNING;
  57                 p->exit_code = 0;
  58                 p->signal &= ~( (1<<(SIGSTOP-1)) | (1<<(SIGTSTP-1)) |
  59                                 (1<<(SIGTTIN-1)) | (1<<(SIGTTOU-1)) );
  60         }
  61         /* Depends on order SIGSTOP, SIGTSTP, SIGTTIN, SIGTTOU */
  62         if ((sig >= SIGSTOP) && (sig <= SIGTTOU)) 
  63                 p->signal &= ~(1<<(SIGCONT-1));
  64         /* Actually generate the signal */
  65         if (!generate(sig,p))
  66                 return 0;
  67         return 0;
  68 }
  69 
  70 void notify_parent(struct task_struct * tsk)
     /*  */
  71 {
  72         if (tsk->p_pptr == task[1])
  73                 tsk->exit_signal = SIGCHLD;
  74         send_sig(tsk->exit_signal, tsk->p_pptr, 1);
  75         wake_up_interruptible(&tsk->p_pptr->wait_chldexit);
  76 }
  77 
  78 void release(struct task_struct * p)
     /*  */
  79 {
  80         int i;
  81 
  82         if (!p)
  83                 return;
  84         if (p == current) {
  85                 printk("task releasing itself\n");
  86                 return;
  87         }
  88         for (i=1 ; i<NR_TASKS ; i++)
  89                 if (task[i] == p) {
  90                         task[i] = NULL;
  91                         REMOVE_LINKS(p);
  92                         free_page(p->kernel_stack_page);
  93                         free_page((long) p);
  94                         return;
  95                 }
  96         panic("trying to release non-existent task");
  97 }
  98 
  99 #ifdef DEBUG_PROC_TREE
 100 /*
 101  * Check to see if a task_struct pointer is present in the task[] array
 102  * Return 0 if found, and 1 if not found.
 103  */
 104 int bad_task_ptr(struct task_struct *p)
     /*  */
 105 {
 106         int     i;
 107 
 108         if (!p)
 109                 return 0;
 110         for (i=0 ; i<NR_TASKS ; i++)
 111                 if (task[i] == p)
 112                         return 0;
 113         return 1;
 114 }
 115         
 116 /*
 117  * This routine scans the pid tree and make sure the rep invarient still
 118  * holds.  Used for debugging only, since it's very slow....
 119  *
 120  * It looks a lot scarier than it really is.... we're doing nothing more
 121  * than verifying the doubly-linked list found in p_ysptr and p_osptr, 
 122  * and checking it corresponds with the process tree defined by p_cptr and 
 123  * p_pptr;
 124  */
 125 void audit_ptree(void)
     /*  */
 126 {
 127         int     i;
 128 
 129         for (i=1 ; i<NR_TASKS ; i++) {
 130                 if (!task[i])
 131                         continue;
 132                 if (bad_task_ptr(task[i]->p_pptr))
 133                         printk("Warning, pid %d's parent link is bad\n",
 134                                 task[i]->pid);
 135                 if (bad_task_ptr(task[i]->p_cptr))
 136                         printk("Warning, pid %d's child link is bad\n",
 137                                 task[i]->pid);
 138                 if (bad_task_ptr(task[i]->p_ysptr))
 139                         printk("Warning, pid %d's ys link is bad\n",
 140                                 task[i]->pid);
 141                 if (bad_task_ptr(task[i]->p_osptr))
 142                         printk("Warning, pid %d's os link is bad\n",
 143                                 task[i]->pid);
 144                 if (task[i]->p_pptr == task[i])
 145                         printk("Warning, pid %d parent link points to self\n",
 146                                 task[i]->pid);
 147                 if (task[i]->p_cptr == task[i])
 148                         printk("Warning, pid %d child link points to self\n",
 149                                 task[i]->pid);
 150                 if (task[i]->p_ysptr == task[i])
 151                         printk("Warning, pid %d ys link points to self\n",
 152                                 task[i]->pid);
 153                 if (task[i]->p_osptr == task[i])
 154                         printk("Warning, pid %d os link points to self\n",
 155                                 task[i]->pid);
 156                 if (task[i]->p_osptr) {
 157                         if (task[i]->p_pptr != task[i]->p_osptr->p_pptr)
 158                                 printk(
 159                         "Warning, pid %d older sibling %d parent is %d\n",
 160                                 task[i]->pid, task[i]->p_osptr->pid,
 161                                 task[i]->p_osptr->p_pptr->pid);
 162                         if (task[i]->p_osptr->p_ysptr != task[i])
 163                                 printk(
 164                 "Warning, pid %d older sibling %d has mismatched ys link\n",
 165                                 task[i]->pid, task[i]->p_osptr->pid);
 166                 }
 167                 if (task[i]->p_ysptr) {
 168                         if (task[i]->p_pptr != task[i]->p_ysptr->p_pptr)
 169                                 printk(
 170                         "Warning, pid %d younger sibling %d parent is %d\n",
 171                                 task[i]->pid, task[i]->p_osptr->pid,
 172                                 task[i]->p_osptr->p_pptr->pid);
 173                         if (task[i]->p_ysptr->p_osptr != task[i])
 174                                 printk(
 175                 "Warning, pid %d younger sibling %d has mismatched os link\n",
 176                                 task[i]->pid, task[i]->p_ysptr->pid);
 177                 }
 178                 if (task[i]->p_cptr) {
 179                         if (task[i]->p_cptr->p_pptr != task[i])
 180                                 printk(
 181                         "Warning, pid %d youngest child %d has mismatched parent link\n",
 182                                 task[i]->pid, task[i]->p_cptr->pid);
 183                         if (task[i]->p_cptr->p_ysptr)
 184                                 printk(
 185                         "Warning, pid %d youngest child %d has non-NULL ys link\n",
 186                                 task[i]->pid, task[i]->p_cptr->pid);
 187                 }
 188         }
 189 }
 190 #endif /* DEBUG_PROC_TREE */
 191 
 192 /*
 193  * This checks not only the pgrp, but falls back on the pid if no
 194  * satisfactory prgp is found. I dunno - gdb doesn't work correctly
 195  * without this...
 196  */
 197 int session_of_pgrp(int pgrp)
     /*  */
 198 {
 199         struct task_struct **p;
 200         int fallback;
 201 
 202         fallback = -1;
 203         for (p = &LAST_TASK ; p > &FIRST_TASK ; --p) {
 204                 if (!*p || (*p)->session <= 0)
 205                         continue;
 206                 if ((*p)->pgrp == pgrp)
 207                         return (*p)->session;
 208                 if ((*p)->pid == pgrp)
 209                         fallback = (*p)->session;
 210         }
 211         return fallback;
 212 }
 213 
 214 /*
 215  * kill_pg() sends a signal to a process group: this is what the tty
 216  * control characters do (^C, ^Z etc)
 217  */
 218 int kill_pg(int pgrp, int sig, int priv)
     /*  */
 219 {
 220         struct task_struct **p;
 221         int err,retval = -ESRCH;
 222         int found = 0;
 223 
 224         if (sig<0 || sig>32 || pgrp<=0)
 225                 return -EINVAL;
 226         for (p = &LAST_TASK ; p > &FIRST_TASK ; --p)
 227                 if (*p && (*p)->pgrp == pgrp) {
 228                         if ((err = send_sig(sig,*p,priv)) != 0)
 229                                 retval = err;
 230                         else
 231                                 found++;
 232                 }
 233         return(found ? 0 : retval);
 234 }
 235 
 236 /*
 237  * kill_sl() sends a signal to the session leader: this is used
 238  * to send SIGHUP to the controlling process of a terminal when
 239  * the connection is lost.
 240  */
 241 int kill_sl(int sess, int sig, int priv)
     /*  */
 242 {
 243         struct task_struct **p;
 244         int err,retval = -ESRCH;
 245         int found = 0;
 246 
 247         if (sig<0 || sig>32 || sess<=0)
 248                 return -EINVAL;
 249         for (p = &LAST_TASK ; p > &FIRST_TASK ; --p)
 250                 if (*p && (*p)->session == sess && (*p)->leader) {
 251                         if ((err = send_sig(sig,*p,priv)) != 0)
 252                                 retval = err;
 253                         else
 254                                 found++;
 255                 }
 256         return(found ? 0 : retval);
 257 }
 258 
 259 int kill_proc(int pid, int sig, int priv)
     /*  */
 260 {
 261         struct task_struct **p;
 262 
 263         if (sig<0 || sig>32)
 264                 return -EINVAL;
 265         for (p = &LAST_TASK ; p > &FIRST_TASK ; --p)
 266                 if (*p && (*p)->pid == pid)
 267                         return send_sig(sig,*p,priv);
 268         return(-ESRCH);
 269 }
 270 
 271 /*
 272  * POSIX specifies that kill(-1,sig) is unspecified, but what we have
 273  * is probably wrong.  Should make it like BSD or SYSV.
 274  */
 275 extern "C" int sys_kill(int pid,int sig)
     /*  */
 276 {
 277         struct task_struct **p = NR_TASKS + task;
 278         int err, retval = 0, count = 0;
 279 
 280         if (!pid)
 281                 return(kill_pg(current->pgrp,sig,0));
 282         if (pid == -1) {
 283                 while (--p > &FIRST_TASK)
 284                         if (*p && (*p)->pid > 1 && *p != current) {
 285                                 ++count;
 286                                 if ((err = send_sig(sig,*p,0)) != -EPERM)
 287                                         retval = err;
 288                         }
 289                 return(count ? retval : -ESRCH);
 290         }
 291         if (pid < 0) 
 292                 return(kill_pg(-pid,sig,0));
 293         /* Normal kill */
 294         return(kill_proc(pid,sig,0));
 295 }
 296 
 297 /*
 298  * Determine if a process group is "orphaned", according to the POSIX
 299  * definition in 2.2.2.52.  Orphaned process groups are not to be affected
 300  * by terminal-generated stop signals.  Newly orphaned process groups are 
 301  * to receive a SIGHUP and a SIGCONT.
 302  * 
 303  * "I ask you, have you ever known what it is to be an orphan?"
 304  */
 305 int is_orphaned_pgrp(int pgrp)
     /*  */
 306 {
 307         struct task_struct **p;
 308 
 309         for (p = &LAST_TASK ; p > &FIRST_TASK ; --p) {
 310                 if (!(*p) ||
 311                     ((*p)->pgrp != pgrp) || 
 312                     ((*p)->state == TASK_ZOMBIE) ||
 313                     ((*p)->p_pptr->pid == 1))
 314                         continue;
 315                 if (((*p)->p_pptr->pgrp != pgrp) &&
 316                     ((*p)->p_pptr->session == (*p)->session))
 317                         return 0;
 318         }
 319         return(1);      /* (sighing) "Often!" */
 320 }
 321 
 322 static int has_stopped_jobs(int pgrp)
     /*  */
 323 {
 324         struct task_struct ** p;
 325 
 326         for (p = &LAST_TASK ; p > &FIRST_TASK ; --p) {
 327                 if (!*p || (*p)->pgrp != pgrp)
 328                         continue;
 329                 if ((*p)->state == TASK_STOPPED)
 330                         return(1);
 331         }
 332         return(0);
 333 }
 334 
 335 static void forget_original_parent(struct task_struct * father)
     /*  */
 336 {
 337         struct task_struct ** p;
 338 
 339         for (p = &LAST_TASK ; p > &FIRST_TASK ; --p)
 340                 if (*p && (*p)->p_opptr == father)
 341                         if (task[1])
 342                                 (*p)->p_opptr = task[1];
 343                         else
 344                                 (*p)->p_opptr = task[0];
 345 }
 346 
 347 volatile void do_exit(long code)
     /*  */
 348 {
 349         struct task_struct *p;
 350         int i;
 351 
 352 fake_volatile:
 353         if (current->semun)
 354                 sem_exit();
 355         if (current->shm)
 356                 shm_exit();
 357         free_page_tables(current);
 358         for (i=0 ; i<NR_OPEN ; i++)
 359                 if (current->filp[i])
 360                         sys_close(i);
 361         forget_original_parent(current);
 362         iput(current->pwd);
 363         current->pwd = NULL;
 364         iput(current->root);
 365         current->root = NULL;
 366         iput(current->executable);
 367         current->executable = NULL;
 368         /* Release all of the old mmap stuff. */
 369         
 370         {
 371                 struct vm_area_struct * mpnt, *mpnt1;
 372                 mpnt = current->mmap;
 373                 current->mmap = NULL;
 374                 while (mpnt) {
 375                         mpnt1 = mpnt->vm_next;
 376                         if (mpnt->vm_ops->close)
 377                                 mpnt->vm_ops->close(mpnt);
 378                         kfree(mpnt);
 379                         mpnt = mpnt1;
 380                 }
 381         }
 382 
 383         current->state = TASK_ZOMBIE;
 384         current->exit_code = code;
 385         current->rss = 0;
 386         /* 
 387          * Check to see if any process groups have become orphaned
 388          * as a result of our exiting, and if they have any stopped
 389          * jobs, send them a SIGUP and then a SIGCONT.  (POSIX 3.2.2.2)
 390          *
 391          * Case i: Our father is in a different pgrp than we are
 392          * and we were the only connection outside, so our pgrp
 393          * is about to become orphaned.
 394          */
 395         if ((current->p_pptr->pgrp != current->pgrp) &&
 396             (current->p_pptr->session == current->session) &&
 397             is_orphaned_pgrp(current->pgrp) &&
 398             has_stopped_jobs(current->pgrp)) {
 399                 kill_pg(current->pgrp,SIGHUP,1);
 400                 kill_pg(current->pgrp,SIGCONT,1);
 401         }
 402         /* Let father know we died */
 403         notify_parent(current);
 404         
 405         /*
 406          * This loop does two things:
 407          * 
 408          * A.  Make init inherit all the child processes
 409          * B.  Check to see if any process groups have become orphaned
 410          *      as a result of our exiting, and if they have any stopped
 411          *      jobs, send them a SIGHUP and then a SIGCONT.  (POSIX 3.2.2.2)
 412          */
 413         while ((p = current->p_cptr) != NULL) {
 414                 current->p_cptr = p->p_osptr;
 415                 p->p_ysptr = NULL;
 416                 p->flags &= ~(PF_PTRACED|PF_TRACESYS);
 417                 if (task[1])
 418                         p->p_pptr = task[1];
 419                 else
 420                         p->p_pptr = task[0];
 421                 p->p_osptr = p->p_pptr->p_cptr;
 422                 p->p_osptr->p_ysptr = p;
 423                 p->p_pptr->p_cptr = p;
 424                 if (p->state == TASK_ZOMBIE)
 425                         notify_parent(p);
 426                 /*
 427                  * process group orphan check
 428                  * Case ii: Our child is in a different pgrp 
 429                  * than we are, and it was the only connection
 430                  * outside, so the child pgrp is now orphaned.
 431                  */
 432                 if ((p->pgrp != current->pgrp) &&
 433                     (p->session == current->session) &&
 434                     is_orphaned_pgrp(p->pgrp) &&
 435                     has_stopped_jobs(p->pgrp)) {
 436                         kill_pg(p->pgrp,SIGHUP,1);
 437                         kill_pg(p->pgrp,SIGCONT,1);
 438                 }
 439         }
 440         if (current->leader) {
 441                 struct task_struct **p;
 442                 struct tty_struct *tty;
 443 
 444                 if (current->tty >= 0) {
 445                         tty = TTY_TABLE(current->tty);
 446                         if (tty) {
 447                                 if (tty->pgrp > 0)
 448                                         kill_pg(tty->pgrp, SIGHUP, 1);
 449                                 tty->pgrp = -1;
 450                                 tty->session = 0;
 451                         }
 452                 }
 453                 for (p = &LAST_TASK ; p > &FIRST_TASK ; --p)
 454                         if (*p && (*p)->session == current->session)
 455                                 (*p)->tty = -1;
 456         }
 457         if (last_task_used_math == current)
 458                 last_task_used_math = NULL;
 459 #ifdef DEBUG_PROC_TREE
 460         audit_ptree();
 461 #endif
 462         schedule();
 463 /*
 464  * In order to get rid of the "volatile function does return" message
 465  * I did this little loop that confuses gcc to think do_exit really
 466  * is volatile. In fact it's schedule() that is volatile in some
 467  * circumstances: when current->state = ZOMBIE, schedule() never
 468  * returns.
 469  *
 470  * In fact the natural way to do all this is to have the label and the
 471  * goto right after each other, but I put the fake_volatile label at
 472  * the start of the function just in case something /really/ bad
 473  * happens, and the schedule returns. This way we can try again. I'm
 474  * not paranoid: it's just that everybody is out to get me.
 475  */
 476         goto fake_volatile;
 477 }
 478 
 479 extern "C" int sys_exit(int error_code)
     /*  */
 480 {
 481         do_exit((error_code&0xff)<<8);
 482 }
 483 
 484 extern "C" int sys_wait4(pid_t pid,unsigned long * stat_addr, int options, struct rusage * ru)
     /*  */
 485 {
 486         int flag, retval;
 487         struct wait_queue wait = { current, NULL };
 488         struct task_struct *p;
 489 
 490         if (stat_addr) {
 491                 flag = verify_area(VERIFY_WRITE, stat_addr, 4);
 492                 if (flag)
 493                         return flag;
 494         }
 495         add_wait_queue(&current->wait_chldexit,&wait);
 496 repeat:
 497         flag=0;
 498         for (p = current->p_cptr ; p ; p = p->p_osptr) {
 499                 if (pid>0) {
 500                         if (p->pid != pid)
 501                                 continue;
 502                 } else if (!pid) {
 503                         if (p->pgrp != current->pgrp)
 504                                 continue;
 505                 } else if (pid != -1) {
 506                         if (p->pgrp != -pid)
 507                                 continue;
 508                 }
 509                 /* wait for cloned processes iff the __WCLONE flag is set */
 510                 if ((p->exit_signal != SIGCHLD) ^ ((options & __WCLONE) != 0))
 511                         continue;
 512                 flag = 1;
 513                 switch (p->state) {
 514                         case TASK_STOPPED:
 515                                 if (!p->exit_code)
 516                                         continue;
 517                                 if (!(options & WUNTRACED) && !(p->flags & PF_PTRACED))
 518                                         continue;
 519                                 if (stat_addr)
 520                                         put_fs_long((p->exit_code << 8) | 0x7f,
 521                                                 stat_addr);
 522                                 p->exit_code = 0;
 523                                 if (ru != NULL)
 524                                         getrusage(p, RUSAGE_BOTH, ru);
 525                                 retval = p->pid;
 526                                 goto end_wait4;
 527                         case TASK_ZOMBIE:
 528                                 current->cutime += p->utime + p->cutime;
 529                                 current->cstime += p->stime + p->cstime;
 530                                 current->cmin_flt += p->min_flt + p->cmin_flt;
 531                                 current->cmaj_flt += p->maj_flt + p->cmaj_flt;
 532                                 if (ru != NULL)
 533                                         getrusage(p, RUSAGE_BOTH, ru);
 534                                 flag = p->pid;
 535                                 if (stat_addr)
 536                                         put_fs_long(p->exit_code, stat_addr);
 537                                 if (p->p_opptr != p->p_pptr) {
 538                                         REMOVE_LINKS(p);
 539                                         p->p_pptr = p->p_opptr;
 540                                         SET_LINKS(p);
 541                                         notify_parent(p);
 542                                 } else
 543                                         release(p);
 544 #ifdef DEBUG_PROC_TREE
 545                                 audit_ptree();
 546 #endif
 547                                 retval = flag;
 548                                 goto end_wait4;
 549                         default:
 550                                 continue;
 551                 }
 552         }
 553         if (flag) {
 554                 retval = 0;
 555                 if (options & WNOHANG)
 556                         goto end_wait4;
 557                 current->state=TASK_INTERRUPTIBLE;
 558                 schedule();
 559                 current->signal &= ~(1<<(SIGCHLD-1));
 560                 retval = -ERESTARTSYS;
 561                 if (current->signal & ~current->blocked)
 562                         goto end_wait4;
 563                 goto repeat;
 564         }
 565         retval = -ECHILD;
 566 end_wait4:
 567         remove_wait_queue(&current->wait_chldexit,&wait);
 568         return retval;
 569 }
 570 
 571 /*
 572  * sys_waitpid() remains for compatibility. waitpid() should be
 573  * implemented by calling sys_wait4() from libc.a.
 574  */
 575 extern "C" int sys_waitpid(pid_t pid,unsigned long * stat_addr, int options)
     /*  */
 576 {
 577         return sys_wait4(pid, stat_addr, options, NULL);
 578 }
/* */
root/kernel/exit.c

DEFINITIONS
