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
exit_mm
exit_files
exit_fs
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 #include <linux/malloc.h>
  18 
  19 #include <asm/segment.h>
  20 extern void shm_exit (void);
  21 extern void sem_exit (void);
  22 
  23 int getrusage(struct task_struct *, int, struct rusage *);
  24 
  25 static int generate(unsigned long sig, struct task_struct * p)
     /*  */
  26 {
  27         unsigned long mask = 1 << (sig-1);
  28         struct sigaction * sa = sig + p->sigaction - 1;
  29 
  30         /* always generate signals for traced processes ??? */
  31         if (p->flags & PF_PTRACED) {
  32                 p->signal |= mask;
  33                 return 1;
  34         }
  35         /* don't bother with ignored signals (but SIGCHLD is special) */
  36         if (sa->sa_handler == SIG_IGN && sig != SIGCHLD)
  37                 return 0;
  38         /* some signals are ignored by default.. (but SIGCONT already did its deed) */
  39         if ((sa->sa_handler == SIG_DFL) &&
  40             (sig == SIGCONT || sig == SIGCHLD || sig == SIGWINCH))
  41                 return 0;
  42         p->signal |= mask;
  43         return 1;
  44 }
  45 
  46 int send_sig(unsigned long sig,struct task_struct * p,int priv)
     /*  */
  47 {
  48         if (!p || sig > 32)
  49                 return -EINVAL;
  50         if (!priv && ((sig != SIGCONT) || (current->session != p->session)) &&
  51             (current->euid != p->euid) && (current->uid != p->uid) && !suser())
  52                 return -EPERM;
  53         if (!sig)
  54                 return 0;
  55         /*
  56          * Forget it if the process is already zombie'd.
  57          */
  58         if (p->state == TASK_ZOMBIE)
  59                 return 0;
  60         if ((sig == SIGKILL) || (sig == SIGCONT)) {
  61                 if (p->state == TASK_STOPPED)
  62                         p->state = TASK_RUNNING;
  63                 p->exit_code = 0;
  64                 p->signal &= ~( (1<<(SIGSTOP-1)) | (1<<(SIGTSTP-1)) |
  65                                 (1<<(SIGTTIN-1)) | (1<<(SIGTTOU-1)) );
  66         }
  67         /* Depends on order SIGSTOP, SIGTSTP, SIGTTIN, SIGTTOU */
  68         if ((sig >= SIGSTOP) && (sig <= SIGTTOU)) 
  69                 p->signal &= ~(1<<(SIGCONT-1));
  70         /* Actually generate the signal */
  71         generate(sig,p);
  72         return 0;
  73 }
  74 
  75 void notify_parent(struct task_struct * tsk)
     /*  */
  76 {
  77         if (tsk->p_pptr == task[1])
  78                 tsk->exit_signal = SIGCHLD;
  79         send_sig(tsk->exit_signal, tsk->p_pptr, 1);
  80         wake_up_interruptible(&tsk->p_pptr->wait_chldexit);
  81 }
  82 
  83 void release(struct task_struct * p)
     /*  */
  84 {
  85         int i;
  86 
  87         if (!p)
  88                 return;
  89         if (p == current) {
  90                 printk("task releasing itself\n");
  91                 return;
  92         }
  93         for (i=1 ; i<NR_TASKS ; i++)
  94                 if (task[i] == p) {
  95                         task[i] = NULL;
  96                         REMOVE_LINKS(p);
  97                         if (STACK_MAGIC != *(unsigned long *)p->kernel_stack_page)
  98                                 printk(KERN_ALERT "release: %s kernel stack corruption. Aiee\n", p->comm);
  99                         free_page(p->kernel_stack_page);
 100                         free_page((long) p);
 101                         return;
 102                 }
 103         panic("trying to release non-existent task");
 104 }
 105 
 106 #ifdef DEBUG_PROC_TREE
 107 /*
 108  * Check to see if a task_struct pointer is present in the task[] array
 109  * Return 0 if found, and 1 if not found.
 110  */
 111 int bad_task_ptr(struct task_struct *p)
     /*  */
 112 {
 113         int     i;
 114 
 115         if (!p)
 116                 return 0;
 117         for (i=0 ; i<NR_TASKS ; i++)
 118                 if (task[i] == p)
 119                         return 0;
 120         return 1;
 121 }
 122         
 123 /*
 124  * This routine scans the pid tree and makes sure the rep invariant still
 125  * holds.  Used for debugging only, since it's very slow....
 126  *
 127  * It looks a lot scarier than it really is.... we're doing nothing more
 128  * than verifying the doubly-linked list found in p_ysptr and p_osptr, 
 129  * and checking it corresponds with the process tree defined by p_cptr and 
 130  * p_pptr;
 131  */
 132 void audit_ptree(void)
     /*  */
 133 {
 134         int     i;
 135 
 136         for (i=1 ; i<NR_TASKS ; i++) {
 137                 if (!task[i])
 138                         continue;
 139                 if (bad_task_ptr(task[i]->p_pptr))
 140                         printk("Warning, pid %d's parent link is bad\n",
 141                                 task[i]->pid);
 142                 if (bad_task_ptr(task[i]->p_cptr))
 143                         printk("Warning, pid %d's child link is bad\n",
 144                                 task[i]->pid);
 145                 if (bad_task_ptr(task[i]->p_ysptr))
 146                         printk("Warning, pid %d's ys link is bad\n",
 147                                 task[i]->pid);
 148                 if (bad_task_ptr(task[i]->p_osptr))
 149                         printk("Warning, pid %d's os link is bad\n",
 150                                 task[i]->pid);
 151                 if (task[i]->p_pptr == task[i])
 152                         printk("Warning, pid %d parent link points to self\n",
 153                                 task[i]->pid);
 154                 if (task[i]->p_cptr == task[i])
 155                         printk("Warning, pid %d child link points to self\n",
 156                                 task[i]->pid);
 157                 if (task[i]->p_ysptr == task[i])
 158                         printk("Warning, pid %d ys link points to self\n",
 159                                 task[i]->pid);
 160                 if (task[i]->p_osptr == task[i])
 161                         printk("Warning, pid %d os link points to self\n",
 162                                 task[i]->pid);
 163                 if (task[i]->p_osptr) {
 164                         if (task[i]->p_pptr != task[i]->p_osptr->p_pptr)
 165                                 printk(
 166                         "Warning, pid %d older sibling %d parent is %d\n",
 167                                 task[i]->pid, task[i]->p_osptr->pid,
 168                                 task[i]->p_osptr->p_pptr->pid);
 169                         if (task[i]->p_osptr->p_ysptr != task[i])
 170                                 printk(
 171                 "Warning, pid %d older sibling %d has mismatched ys link\n",
 172                                 task[i]->pid, task[i]->p_osptr->pid);
 173                 }
 174                 if (task[i]->p_ysptr) {
 175                         if (task[i]->p_pptr != task[i]->p_ysptr->p_pptr)
 176                                 printk(
 177                         "Warning, pid %d younger sibling %d parent is %d\n",
 178                                 task[i]->pid, task[i]->p_osptr->pid,
 179                                 task[i]->p_osptr->p_pptr->pid);
 180                         if (task[i]->p_ysptr->p_osptr != task[i])
 181                                 printk(
 182                 "Warning, pid %d younger sibling %d has mismatched os link\n",
 183                                 task[i]->pid, task[i]->p_ysptr->pid);
 184                 }
 185                 if (task[i]->p_cptr) {
 186                         if (task[i]->p_cptr->p_pptr != task[i])
 187                                 printk(
 188                         "Warning, pid %d youngest child %d has mismatched parent link\n",
 189                                 task[i]->pid, task[i]->p_cptr->pid);
 190                         if (task[i]->p_cptr->p_ysptr)
 191                                 printk(
 192                         "Warning, pid %d youngest child %d has non-NULL ys link\n",
 193                                 task[i]->pid, task[i]->p_cptr->pid);
 194                 }
 195         }
 196 }
 197 #endif /* DEBUG_PROC_TREE */
 198 
 199 /*
 200  * This checks not only the pgrp, but falls back on the pid if no
 201  * satisfactory pgrp is found. I dunno - gdb doesn't work correctly
 202  * without this...
 203  */
 204 int session_of_pgrp(int pgrp)
     /*  */
 205 {
 206         struct task_struct *p;
 207         int fallback;
 208 
 209         fallback = -1;
 210         for_each_task(p) {
 211                 if (p->session <= 0)
 212                         continue;
 213                 if (p->pgrp == pgrp)
 214                         return p->session;
 215                 if (p->pid == pgrp)
 216                         fallback = p->session;
 217         }
 218         return fallback;
 219 }
 220 
 221 /*
 222  * kill_pg() sends a signal to a process group: this is what the tty
 223  * control characters do (^C, ^Z etc)
 224  */
 225 int kill_pg(int pgrp, int sig, int priv)
     /*  */
 226 {
 227         struct task_struct *p;
 228         int err,retval = -ESRCH;
 229         int found = 0;
 230 
 231         if (sig<0 || sig>32 || pgrp<=0)
 232                 return -EINVAL;
 233         for_each_task(p) {
 234                 if (p->pgrp == pgrp) {
 235                         if ((err = send_sig(sig,p,priv)) != 0)
 236                                 retval = err;
 237                         else
 238                                 found++;
 239                 }
 240         }
 241         return(found ? 0 : retval);
 242 }
 243 
 244 /*
 245  * kill_sl() sends a signal to the session leader: this is used
 246  * to send SIGHUP to the controlling process of a terminal when
 247  * the connection is lost.
 248  */
 249 int kill_sl(int sess, int sig, int priv)
     /*  */
 250 {
 251         struct task_struct *p;
 252         int err,retval = -ESRCH;
 253         int found = 0;
 254 
 255         if (sig<0 || sig>32 || sess<=0)
 256                 return -EINVAL;
 257         for_each_task(p) {
 258                 if (p->session == sess && p->leader) {
 259                         if ((err = send_sig(sig,p,priv)) != 0)
 260                                 retval = err;
 261                         else
 262                                 found++;
 263                 }
 264         }
 265         return(found ? 0 : retval);
 266 }
 267 
 268 int kill_proc(int pid, int sig, int priv)
     /*  */
 269 {
 270         struct task_struct *p;
 271 
 272         if (sig<0 || sig>32)
 273                 return -EINVAL;
 274         for_each_task(p) {
 275                 if (p && p->pid == pid)
 276                         return send_sig(sig,p,priv);
 277         }
 278         return(-ESRCH);
 279 }
 280 
 281 /*
 282  * POSIX specifies that kill(-1,sig) is unspecified, but what we have
 283  * is probably wrong.  Should make it like BSD or SYSV.
 284  */
 285 asmlinkage int sys_kill(int pid,int sig)
     /*  */
 286 {
 287         int err, retval = 0, count = 0;
 288 
 289         if (!pid)
 290                 return(kill_pg(current->pgrp,sig,0));
 291         if (pid == -1) {
 292                 struct task_struct * p;
 293                 for_each_task(p) {
 294                         if (p->pid > 1 && p != current) {
 295                                 ++count;
 296                                 if ((err = send_sig(sig,p,0)) != -EPERM)
 297                                         retval = err;
 298                         }
 299                 }
 300                 return(count ? retval : -ESRCH);
 301         }
 302         if (pid < 0) 
 303                 return(kill_pg(-pid,sig,0));
 304         /* Normal kill */
 305         return(kill_proc(pid,sig,0));
 306 }
 307 
 308 /*
 309  * Determine if a process group is "orphaned", according to the POSIX
 310  * definition in 2.2.2.52.  Orphaned process groups are not to be affected
 311  * by terminal-generated stop signals.  Newly orphaned process groups are 
 312  * to receive a SIGHUP and a SIGCONT.
 313  * 
 314  * "I ask you, have you ever known what it is to be an orphan?"
 315  */
 316 int is_orphaned_pgrp(int pgrp)
     /*  */
 317 {
 318         struct task_struct *p;
 319 
 320         for_each_task(p) {
 321                 if ((p->pgrp != pgrp) || 
 322                     (p->state == TASK_ZOMBIE) ||
 323                     (p->p_pptr->pid == 1))
 324                         continue;
 325                 if ((p->p_pptr->pgrp != pgrp) &&
 326                     (p->p_pptr->session == p->session))
 327                         return 0;
 328         }
 329         return(1);      /* (sighing) "Often!" */
 330 }
 331 
 332 static int has_stopped_jobs(int pgrp)
     /*  */
 333 {
 334         struct task_struct * p;
 335 
 336         for_each_task(p) {
 337                 if (p->pgrp != pgrp)
 338                         continue;
 339                 if (p->state == TASK_STOPPED)
 340                         return(1);
 341         }
 342         return(0);
 343 }
 344 
 345 static void forget_original_parent(struct task_struct * father)
     /*  */
 346 {
 347         struct task_struct * p;
 348 
 349         for_each_task(p) {
 350                 if (p->p_opptr == father)
 351                         if (task[1])
 352                                 p->p_opptr = task[1];
 353                         else
 354                                 p->p_opptr = task[0];
 355         }
 356 }
 357 
 358 static void exit_mm(void)
     /*  */
 359 {
 360         struct vm_area_struct * mpnt;
 361 
 362         mpnt = current->mm->mmap;
 363         current->mm->mmap = NULL;
 364         while (mpnt) {
 365                 struct vm_area_struct * next = mpnt->vm_next;
 366                 if (mpnt->vm_ops && mpnt->vm_ops->close)
 367                         mpnt->vm_ops->close(mpnt);
 368                 if (mpnt->vm_inode)
 369                         iput(mpnt->vm_inode);
 370                 kfree(mpnt);
 371                 mpnt = next;
 372         }
 373 
 374         /* forget local segments */
 375         __asm__ __volatile__("mov %w0,%%fs ; mov %w0,%%gs ; lldt %w0"
 376                 : /* no outputs */
 377                 : "r" (0));
 378         current->tss.ldt = 0;
 379         if (current->ldt) {
 380                 void * ldt = current->ldt;
 381                 current->ldt = NULL;
 382                 vfree(ldt);
 383         }
 384 
 385         free_page_tables(current);
 386 }
 387 
 388 static void exit_files(void)
     /*  */
 389 {
 390         int i;
 391 
 392         for (i=0 ; i<NR_OPEN ; i++)
 393                 if (current->files->fd[i])
 394                         sys_close(i);
 395 }
 396 
 397 static void exit_fs(void)
     /*  */
 398 {
 399         iput(current->fs->pwd);
 400         current->fs->pwd = NULL;
 401         iput(current->fs->root);
 402         current->fs->root = NULL;
 403 }
 404 
 405 NORET_TYPE void do_exit(long code)
     /*  */
 406 {
 407         struct task_struct *p;
 408 
 409         if (intr_count) {
 410                 printk("Aiee, killing interrupt handler\n");
 411                 intr_count = 0;
 412         }
 413 fake_volatile:
 414         if (current->semun)
 415                 sem_exit();
 416         if (current->shm)
 417                 shm_exit();
 418         exit_mm();
 419         exit_files();
 420         exit_fs();
 421         forget_original_parent(current);
 422         /* 
 423          * Check to see if any process groups have become orphaned
 424          * as a result of our exiting, and if they have any stopped
 425          * jobs, send them a SIGUP and then a SIGCONT.  (POSIX 3.2.2.2)
 426          *
 427          * Case i: Our father is in a different pgrp than we are
 428          * and we were the only connection outside, so our pgrp
 429          * is about to become orphaned.
 430          */
 431         if ((current->p_pptr->pgrp != current->pgrp) &&
 432             (current->p_pptr->session == current->session) &&
 433             is_orphaned_pgrp(current->pgrp) &&
 434             has_stopped_jobs(current->pgrp)) {
 435                 kill_pg(current->pgrp,SIGHUP,1);
 436                 kill_pg(current->pgrp,SIGCONT,1);
 437         }
 438         /* Let father know we died */
 439         notify_parent(current);
 440         
 441         /*
 442          * This loop does two things:
 443          * 
 444          * A.  Make init inherit all the child processes
 445          * B.  Check to see if any process groups have become orphaned
 446          *      as a result of our exiting, and if they have any stopped
 447          *      jobs, send them a SIGHUP and then a SIGCONT.  (POSIX 3.2.2.2)
 448          */
 449         while ((p = current->p_cptr) != NULL) {
 450                 current->p_cptr = p->p_osptr;
 451                 p->p_ysptr = NULL;
 452                 p->flags &= ~(PF_PTRACED|PF_TRACESYS);
 453                 if (task[1] && task[1] != current)
 454                         p->p_pptr = task[1];
 455                 else
 456                         p->p_pptr = task[0];
 457                 p->p_osptr = p->p_pptr->p_cptr;
 458                 p->p_osptr->p_ysptr = p;
 459                 p->p_pptr->p_cptr = p;
 460                 if (p->state == TASK_ZOMBIE)
 461                         notify_parent(p);
 462                 /*
 463                  * process group orphan check
 464                  * Case ii: Our child is in a different pgrp 
 465                  * than we are, and it was the only connection
 466                  * outside, so the child pgrp is now orphaned.
 467                  */
 468                 if ((p->pgrp != current->pgrp) &&
 469                     (p->session == current->session) &&
 470                     is_orphaned_pgrp(p->pgrp) &&
 471                     has_stopped_jobs(p->pgrp)) {
 472                         kill_pg(p->pgrp,SIGHUP,1);
 473                         kill_pg(p->pgrp,SIGCONT,1);
 474                 }
 475         }
 476         if (current->leader)
 477                 disassociate_ctty(1);
 478         if (last_task_used_math == current)
 479                 last_task_used_math = NULL;
 480         current->state = TASK_ZOMBIE;
 481         current->exit_code = code;
 482         current->mm->rss = 0;
 483 #ifdef DEBUG_PROC_TREE
 484         audit_ptree();
 485 #endif
 486         if (current->exec_domain && current->exec_domain->use_count)
 487                 (*current->exec_domain->use_count)--;
 488         if (current->binfmt && current->binfmt->use_count)
 489                 (*current->binfmt->use_count)--;
 490         schedule();
 491 /*
 492  * In order to get rid of the "volatile function does return" message
 493  * I did this little loop that confuses gcc to think do_exit really
 494  * is volatile. In fact it's schedule() that is volatile in some
 495  * circumstances: when current->state = ZOMBIE, schedule() never
 496  * returns.
 497  *
 498  * In fact the natural way to do all this is to have the label and the
 499  * goto right after each other, but I put the fake_volatile label at
 500  * the start of the function just in case something /really/ bad
 501  * happens, and the schedule returns. This way we can try again. I'm
 502  * not paranoid: it's just that everybody is out to get me.
 503  */
 504         goto fake_volatile;
 505 }
 506 
 507 asmlinkage int sys_exit(int error_code)
     /*  */
 508 {
 509         do_exit((error_code&0xff)<<8);
 510 }
 511 
 512 asmlinkage int sys_wait4(pid_t pid,unsigned long * stat_addr, int options, struct rusage * ru)
     /*  */
 513 {
 514         int flag, retval;
 515         struct wait_queue wait = { current, NULL };
 516         struct task_struct *p;
 517 
 518         if (stat_addr) {
 519                 flag = verify_area(VERIFY_WRITE, stat_addr, 4);
 520                 if (flag)
 521                         return flag;
 522         }
 523         add_wait_queue(&current->wait_chldexit,&wait);
 524 repeat:
 525         flag=0;
 526         for (p = current->p_cptr ; p ; p = p->p_osptr) {
 527                 if (pid>0) {
 528                         if (p->pid != pid)
 529                                 continue;
 530                 } else if (!pid) {
 531                         if (p->pgrp != current->pgrp)
 532                                 continue;
 533                 } else if (pid != -1) {
 534                         if (p->pgrp != -pid)
 535                                 continue;
 536                 }
 537                 /* wait for cloned processes iff the __WCLONE flag is set */
 538                 if ((p->exit_signal != SIGCHLD) ^ ((options & __WCLONE) != 0))
 539                         continue;
 540                 flag = 1;
 541                 switch (p->state) {
 542                         case TASK_STOPPED:
 543                                 if (!p->exit_code)
 544                                         continue;
 545                                 if (!(options & WUNTRACED) && !(p->flags & PF_PTRACED))
 546                                         continue;
 547                                 if (stat_addr)
 548                                         put_fs_long((p->exit_code << 8) | 0x7f,
 549                                                 stat_addr);
 550                                 p->exit_code = 0;
 551                                 if (ru != NULL)
 552                                         getrusage(p, RUSAGE_BOTH, ru);
 553                                 retval = p->pid;
 554                                 goto end_wait4;
 555                         case TASK_ZOMBIE:
 556                                 current->cutime += p->utime + p->cutime;
 557                                 current->cstime += p->stime + p->cstime;
 558                                 current->mm->cmin_flt += p->mm->min_flt + p->mm->cmin_flt;
 559                                 current->mm->cmaj_flt += p->mm->maj_flt + p->mm->cmaj_flt;
 560                                 if (ru != NULL)
 561                                         getrusage(p, RUSAGE_BOTH, ru);
 562                                 flag = p->pid;
 563                                 if (stat_addr)
 564                                         put_fs_long(p->exit_code, stat_addr);
 565                                 if (p->p_opptr != p->p_pptr) {
 566                                         REMOVE_LINKS(p);
 567                                         p->p_pptr = p->p_opptr;
 568                                         SET_LINKS(p);
 569                                         notify_parent(p);
 570                                 } else
 571                                         release(p);
 572 #ifdef DEBUG_PROC_TREE
 573                                 audit_ptree();
 574 #endif
 575                                 retval = flag;
 576                                 goto end_wait4;
 577                         default:
 578                                 continue;
 579                 }
 580         }
 581         if (flag) {
 582                 retval = 0;
 583                 if (options & WNOHANG)
 584                         goto end_wait4;
 585                 current->state=TASK_INTERRUPTIBLE;
 586                 schedule();
 587                 current->signal &= ~(1<<(SIGCHLD-1));
 588                 retval = -ERESTARTSYS;
 589                 if (current->signal & ~current->blocked)
 590                         goto end_wait4;
 591                 goto repeat;
 592         }
 593         retval = -ECHILD;
 594 end_wait4:
 595         remove_wait_queue(&current->wait_chldexit,&wait);
 596         return retval;
 597 }
 598 
 599 /*
 600  * sys_waitpid() remains for compatibility. waitpid() should be
 601  * implemented by calling sys_wait4() from libc.a.
 602  */
 603 asmlinkage int sys_waitpid(pid_t pid,unsigned long * stat_addr, int options)
     /*  */
 604 {
 605         return sys_wait4(pid, stat_addr, options, NULL);
 606 }
/* */
root/kernel/exit.c

DEFINITIONS
