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_files
exit_files
__exit_fs
exit_fs
__exit_sighand
exit_sighand
exit_mm
exit_notify
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 #undef 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 sem_exit (void);
  21 extern void kerneld_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->sig->action - 1;
  29 
  30         /* always generate signals for traced processes ??? */
  31         if (!(p->flags & PF_PTRACED)) {
  32                 /* don't bother with ignored signals (but SIGCHLD is special) */
  33                 if (sa->sa_handler == SIG_IGN && sig != SIGCHLD)
  34                         return 0;
  35                 /* some signals are ignored by default.. (but SIGCONT already did its deed) */
  36                 if ((sa->sa_handler == SIG_DFL) &&
  37                     (sig == SIGCONT || sig == SIGCHLD || sig == SIGWINCH))
  38                         return 0;
  39         }
  40         p->signal |= mask;
  41         if (p->state == TASK_INTERRUPTIBLE && (p->signal & ~p->blocked))
  42                 wake_up_process(p);
  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->euid != 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->sig)
  59                 return 0;
  60         if ((sig == SIGKILL) || (sig == SIGCONT)) {
  61                 if (p->state == TASK_STOPPED)
  62                         wake_up_process(p);
  63                 p->exit_code = 0;
  64                 p->signal &= ~( (1<<(SIGSTOP-1)) | (1<<(SIGTSTP-1)) |
  65                                 (1<<(SIGTTIN-1)) | (1<<(SIGTTOU-1)) );
  66         }
  67         if (sig == SIGSTOP || sig == SIGTSTP || sig == SIGTTIN || sig == SIGTTOU)
  68                 p->signal &= ~(1<<(SIGCONT-1));
  69         /* Actually generate the signal */
  70         generate(sig,p);
  71         return 0;
  72 }
  73 
  74 void notify_parent(struct task_struct * tsk)
     /*  */
  75 {
  76         if (tsk->p_pptr == task[smp_num_cpus])          /* Init */
  77                 tsk->exit_signal = SIGCHLD;
  78         send_sig(tsk->exit_signal, tsk->p_pptr, 1);
  79         wake_up_interruptible(&tsk->p_pptr->wait_chldexit);
  80 }
  81 
  82 void release(struct task_struct * p)
     /*  */
  83 {
  84         int i;
  85 
  86         if (!p)
  87                 return;
  88         if (p == current) {
  89                 printk("task releasing itself\n");
  90                 return;
  91         }
  92         for (i=1 ; i<NR_TASKS ; i++)
  93                 if (task[i] == p) {
  94                         nr_tasks--;
  95                         task[i] = NULL;
  96                         REMOVE_LINKS(p);
  97                         release_thread(p);
  98                         if (STACK_MAGIC != *(unsigned long *)p->kernel_stack_page)
  99                                 printk(KERN_ALERT "release: %s kernel stack corruption. Aiee\n", p->comm);
 100                         free_kernel_stack(p->kernel_stack_page);
 101                         current->cmin_flt += p->min_flt + p->cmin_flt;
 102                         current->cmaj_flt += p->maj_flt + p->cmaj_flt;
 103                         current->cnswap += p->nswap + p->cnswap;
 104                         kfree(p);
 105                         return;
 106                 }
 107         panic("trying to release non-existent task");
 108 }
 109 
 110 #ifdef DEBUG_PROC_TREE
 111 /*
 112  * Check to see if a task_struct pointer is present in the task[] array
 113  * Return 0 if found, and 1 if not found.
 114  */
 115 int bad_task_ptr(struct task_struct *p)
     /*  */
 116 {
 117         int     i;
 118 
 119         if (!p)
 120                 return 0;
 121         for (i=0 ; i<NR_TASKS ; i++)
 122                 if (task[i] == p)
 123                         return 0;
 124         return 1;
 125 }
 126         
 127 /*
 128  * This routine scans the pid tree and makes sure the rep invariant still
 129  * holds.  Used for debugging only, since it's very slow....
 130  *
 131  * It looks a lot scarier than it really is.... we're doing nothing more
 132  * than verifying the doubly-linked list found in p_ysptr and p_osptr, 
 133  * and checking it corresponds with the process tree defined by p_cptr and 
 134  * p_pptr;
 135  */
 136 void audit_ptree(void)
     /*  */
 137 {
 138         int     i;
 139 
 140         for (i=1 ; i<NR_TASKS ; i++) {
 141                 if (!task[i])
 142                         continue;
 143                 if (bad_task_ptr(task[i]->p_pptr))
 144                         printk("Warning, pid %d's parent link is bad\n",
 145                                 task[i]->pid);
 146                 if (bad_task_ptr(task[i]->p_cptr))
 147                         printk("Warning, pid %d's child link is bad\n",
 148                                 task[i]->pid);
 149                 if (bad_task_ptr(task[i]->p_ysptr))
 150                         printk("Warning, pid %d's ys link is bad\n",
 151                                 task[i]->pid);
 152                 if (bad_task_ptr(task[i]->p_osptr))
 153                         printk("Warning, pid %d's os link is bad\n",
 154                                 task[i]->pid);
 155                 if (task[i]->p_pptr == task[i])
 156                         printk("Warning, pid %d parent link points to self\n",
 157                                 task[i]->pid);
 158                 if (task[i]->p_cptr == task[i])
 159                         printk("Warning, pid %d child link points to self\n",
 160                                 task[i]->pid);
 161                 if (task[i]->p_ysptr == task[i])
 162                         printk("Warning, pid %d ys link points to self\n",
 163                                 task[i]->pid);
 164                 if (task[i]->p_osptr == task[i])
 165                         printk("Warning, pid %d os link points to self\n",
 166                                 task[i]->pid);
 167                 if (task[i]->p_osptr) {
 168                         if (task[i]->p_pptr != task[i]->p_osptr->p_pptr)
 169                                 printk(
 170                         "Warning, pid %d older 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_osptr->p_ysptr != task[i])
 174                                 printk(
 175                 "Warning, pid %d older sibling %d has mismatched ys link\n",
 176                                 task[i]->pid, task[i]->p_osptr->pid);
 177                 }
 178                 if (task[i]->p_ysptr) {
 179                         if (task[i]->p_pptr != task[i]->p_ysptr->p_pptr)
 180                                 printk(
 181                         "Warning, pid %d younger sibling %d parent is %d\n",
 182                                 task[i]->pid, task[i]->p_osptr->pid,
 183                                 task[i]->p_osptr->p_pptr->pid);
 184                         if (task[i]->p_ysptr->p_osptr != task[i])
 185                                 printk(
 186                 "Warning, pid %d younger sibling %d has mismatched os link\n",
 187                                 task[i]->pid, task[i]->p_ysptr->pid);
 188                 }
 189                 if (task[i]->p_cptr) {
 190                         if (task[i]->p_cptr->p_pptr != task[i])
 191                                 printk(
 192                         "Warning, pid %d youngest child %d has mismatched parent link\n",
 193                                 task[i]->pid, task[i]->p_cptr->pid);
 194                         if (task[i]->p_cptr->p_ysptr)
 195                                 printk(
 196                         "Warning, pid %d youngest child %d has non-NULL ys link\n",
 197                                 task[i]->pid, task[i]->p_cptr->pid);
 198                 }
 199         }
 200 }
 201 #endif /* DEBUG_PROC_TREE */
 202 
 203 /*
 204  * This checks not only the pgrp, but falls back on the pid if no
 205  * satisfactory pgrp is found. I dunno - gdb doesn't work correctly
 206  * without this...
 207  */
 208 int session_of_pgrp(int pgrp)
     /*  */
 209 {
 210         struct task_struct *p;
 211         int fallback;
 212 
 213         fallback = -1;
 214         for_each_task(p) {
 215                 if (p->session <= 0)
 216                         continue;
 217                 if (p->pgrp == pgrp)
 218                         return p->session;
 219                 if (p->pid == pgrp)
 220                         fallback = p->session;
 221         }
 222         return fallback;
 223 }
 224 
 225 /*
 226  * kill_pg() sends a signal to a process group: this is what the tty
 227  * control characters do (^C, ^Z etc)
 228  */
 229 int kill_pg(int pgrp, int sig, int priv)
     /*  */
 230 {
 231         struct task_struct *p;
 232         int err,retval = -ESRCH;
 233         int found = 0;
 234 
 235         if (sig<0 || sig>32 || pgrp<=0)
 236                 return -EINVAL;
 237         for_each_task(p) {
 238                 if (p->pgrp == pgrp) {
 239                         if ((err = send_sig(sig,p,priv)) != 0)
 240                                 retval = err;
 241                         else
 242                                 found++;
 243                 }
 244         }
 245         return(found ? 0 : retval);
 246 }
 247 
 248 /*
 249  * kill_sl() sends a signal to the session leader: this is used
 250  * to send SIGHUP to the controlling process of a terminal when
 251  * the connection is lost.
 252  */
 253 int kill_sl(int sess, int sig, int priv)
     /*  */
 254 {
 255         struct task_struct *p;
 256         int err,retval = -ESRCH;
 257         int found = 0;
 258 
 259         if (sig<0 || sig>32 || sess<=0)
 260                 return -EINVAL;
 261         for_each_task(p) {
 262                 if (p->session == sess && p->leader) {
 263                         if ((err = send_sig(sig,p,priv)) != 0)
 264                                 retval = err;
 265                         else
 266                                 found++;
 267                 }
 268         }
 269         return(found ? 0 : retval);
 270 }
 271 
 272 int kill_proc(int pid, int sig, int priv)
     /*  */
 273 {
 274         struct task_struct *p;
 275 
 276         if (sig<0 || sig>32)
 277                 return -EINVAL;
 278         for_each_task(p) {
 279                 if (p && p->pid == pid)
 280                         return send_sig(sig,p,priv);
 281         }
 282         return(-ESRCH);
 283 }
 284 
 285 /*
 286  * POSIX specifies that kill(-1,sig) is unspecified, but what we have
 287  * is probably wrong.  Should make it like BSD or SYSV.
 288  */
 289 asmlinkage int sys_kill(int pid,int sig)
     /*  */
 290 {
 291         int err, retval = 0, count = 0;
 292 
 293         if (!pid)
 294                 return(kill_pg(current->pgrp,sig,0));
 295         if (pid == -1) {
 296                 struct task_struct * p;
 297                 for_each_task(p) {
 298                         if (p->pid > 1 && p != current) {
 299                                 ++count;
 300                                 if ((err = send_sig(sig,p,0)) != -EPERM)
 301                                         retval = err;
 302                         }
 303                 }
 304                 return(count ? retval : -ESRCH);
 305         }
 306         if (pid < 0) 
 307                 return(kill_pg(-pid,sig,0));
 308         /* Normal kill */
 309         return(kill_proc(pid,sig,0));
 310 }
 311 
 312 /*
 313  * Determine if a process group is "orphaned", according to the POSIX
 314  * definition in 2.2.2.52.  Orphaned process groups are not to be affected
 315  * by terminal-generated stop signals.  Newly orphaned process groups are 
 316  * to receive a SIGHUP and a SIGCONT.
 317  * 
 318  * "I ask you, have you ever known what it is to be an orphan?"
 319  */
 320 int is_orphaned_pgrp(int pgrp)
     /*  */
 321 {
 322         struct task_struct *p;
 323 
 324         for_each_task(p) {
 325                 if ((p->pgrp != pgrp) || 
 326                     (p->state == TASK_ZOMBIE) ||
 327                     (p->p_pptr->pid == 1))
 328                         continue;
 329                 if ((p->p_pptr->pgrp != pgrp) &&
 330                     (p->p_pptr->session == p->session))
 331                         return 0;
 332         }
 333         return(1);      /* (sighing) "Often!" */
 334 }
 335 
 336 static inline int has_stopped_jobs(int pgrp)
     /*  */
 337 {
 338         struct task_struct * p;
 339 
 340         for_each_task(p) {
 341                 if (p->pgrp != pgrp)
 342                         continue;
 343                 if (p->state == TASK_STOPPED)
 344                         return(1);
 345         }
 346         return(0);
 347 }
 348 
 349 static inline void forget_original_parent(struct task_struct * father)
     /*  */
 350 {
 351         struct task_struct * p;
 352 
 353         for_each_task(p) {
 354                 if (p->p_opptr == father)
 355                         if (task[smp_num_cpus]) /* init */
 356                                 p->p_opptr = task[smp_num_cpus];
 357                         else
 358                                 p->p_opptr = task[0];
 359         }
 360 }
 361 
 362 static inline void __exit_files(struct task_struct *tsk)
     /*  */
 363 {
 364         struct files_struct * files = tsk->files;
 365 
 366         if (files) {
 367                 tsk->files = NULL;
 368                 if (!--files->count) {
 369                         int i;
 370                         for (i=0 ; i<NR_OPEN ; i++) {
 371                                 struct file * filp = files->fd[i];
 372                                 if (!filp)
 373                                         continue;
 374                                 files->fd[i] = NULL;
 375                                 close_fp(filp);
 376                         }
 377                         kfree(files);
 378                 }
 379         }
 380 }
 381 
 382 void exit_files(struct task_struct *tsk)
     /*  */
 383 {
 384   __exit_files(tsk);
 385 }
 386 
 387 static inline void __exit_fs(struct task_struct *tsk)
     /*  */
 388 {
 389         struct fs_struct * fs = tsk->fs;
 390 
 391         if (fs) {
 392                 tsk->fs = NULL;
 393                 if (!--fs->count) {
 394                         iput(fs->root);
 395                         iput(fs->pwd);
 396                         kfree(fs);
 397                 }
 398         }
 399 }
 400 
 401 void exit_fs(struct task_struct *tsk)
     /*  */
 402 {
 403   __exit_fs(tsk);
 404 }
 405 
 406 static inline void __exit_sighand(struct task_struct *tsk)
     /*  */
 407 {
 408         struct signal_struct * sig = tsk->sig;
 409 
 410         if (sig) {
 411                 tsk->sig = NULL;
 412                 if (!--sig->count) {
 413                         kfree(sig);
 414                 }
 415         }
 416 }
 417 
 418 void exit_sighand(struct task_struct *tsk)
     /*  */
 419 {
 420   __exit_sighand(tsk);
 421 }
 422 
 423 static inline void exit_mm(void)
     /*  */
 424 {
 425         struct mm_struct * mm = current->mm;
 426 
 427         current->swappable = 0;
 428         if (mm) {
 429                 if (!--mm->count) {
 430                         exit_mmap(mm);
 431                         free_page_tables(current);
 432                         kfree(mm);
 433                 }
 434                 current->mm = NULL;
 435         }
 436 }
 437 
 438 /* 
 439  * Send signals to all our closest relatives so that they know
 440  * to properly mourn us..
 441  */
 442 static void exit_notify(void)
     /*  */
 443 {
 444         struct task_struct * p;
 445 
 446         forget_original_parent(current);
 447         /* 
 448          * Check to see if any process groups have become orphaned
 449          * as a result of our exiting, and if they have any stopped
 450          * jobs, send them a SIGHUP and then a SIGCONT.  (POSIX 3.2.2.2)
 451          *
 452          * Case i: Our father is in a different pgrp than we are
 453          * and we were the only connection outside, so our pgrp
 454          * is about to become orphaned.
 455          */
 456         if ((current->p_pptr->pgrp != current->pgrp) &&
 457             (current->p_pptr->session == current->session) &&
 458             is_orphaned_pgrp(current->pgrp) &&
 459             has_stopped_jobs(current->pgrp)) {
 460                 kill_pg(current->pgrp,SIGHUP,1);
 461                 kill_pg(current->pgrp,SIGCONT,1);
 462         }
 463         /* Let father know we died */
 464         notify_parent(current);
 465         
 466         /*
 467          * This loop does two things:
 468          * 
 469          * A.  Make init inherit all the child processes
 470          * B.  Check to see if any process groups have become orphaned
 471          *      as a result of our exiting, and if they have any stopped
 472          *      jobs, send them a SIGHUP and then a SIGCONT.  (POSIX 3.2.2.2)
 473          */
 474         while ((p = current->p_cptr) != NULL) {
 475                 current->p_cptr = p->p_osptr;
 476                 p->p_ysptr = NULL;
 477                 p->flags &= ~(PF_PTRACED|PF_TRACESYS);
 478                 if (task[smp_num_cpus] && task[smp_num_cpus] != current) /* init */
 479                         p->p_pptr = task[smp_num_cpus];
 480                 else
 481                         p->p_pptr = task[0];
 482                 p->p_osptr = p->p_pptr->p_cptr;
 483                 p->p_osptr->p_ysptr = p;
 484                 p->p_pptr->p_cptr = p;
 485                 if (p->state == TASK_ZOMBIE)
 486                         notify_parent(p);
 487                 /*
 488                  * process group orphan check
 489                  * Case ii: Our child is in a different pgrp 
 490                  * than we are, and it was the only connection
 491                  * outside, so the child pgrp is now orphaned.
 492                  */
 493                 if ((p->pgrp != current->pgrp) &&
 494                     (p->session == current->session) &&
 495                     is_orphaned_pgrp(p->pgrp) &&
 496                     has_stopped_jobs(p->pgrp)) {
 497                         kill_pg(p->pgrp,SIGHUP,1);
 498                         kill_pg(p->pgrp,SIGCONT,1);
 499                 }
 500         }
 501         if (current->leader)
 502                 disassociate_ctty(1);
 503 }
 504 
 505 NORET_TYPE void do_exit(long code)
     /*  */
 506 {
 507         if (intr_count) {
 508                 printk("Aiee, killing interrupt handler\n");
 509                 intr_count = 0;
 510         }
 511 fake_volatile:
 512         current->flags |= PF_EXITING;
 513         del_timer(&current->real_timer);
 514         sem_exit();
 515         kerneld_exit();
 516         exit_mm();
 517         __exit_files(current);
 518         __exit_fs(current);
 519         __exit_sighand(current);
 520         exit_thread();
 521         current->state = TASK_ZOMBIE;
 522         current->exit_code = code;
 523         exit_notify();
 524 #ifdef DEBUG_PROC_TREE
 525         audit_ptree();
 526 #endif
 527         if (current->exec_domain && current->exec_domain->use_count)
 528                 (*current->exec_domain->use_count)--;
 529         if (current->binfmt && current->binfmt->use_count)
 530                 (*current->binfmt->use_count)--;
 531         schedule();
 532 /*
 533  * In order to get rid of the "volatile function does return" message
 534  * I did this little loop that confuses gcc to think do_exit really
 535  * is volatile. In fact it's schedule() that is volatile in some
 536  * circumstances: when current->state = ZOMBIE, schedule() never
 537  * returns.
 538  *
 539  * In fact the natural way to do all this is to have the label and the
 540  * goto right after each other, but I put the fake_volatile label at
 541  * the start of the function just in case something /really/ bad
 542  * happens, and the schedule returns. This way we can try again. I'm
 543  * not paranoid: it's just that everybody is out to get me.
 544  */
 545         goto fake_volatile;
 546 }
 547 
 548 asmlinkage int sys_exit(int error_code)
     /*  */
 549 {
 550         do_exit((error_code&0xff)<<8);
 551 }
 552 
 553 asmlinkage int sys_wait4(pid_t pid,unsigned int * stat_addr, int options, struct rusage * ru)
     /*  */
 554 {
 555         int flag, retval;
 556         struct wait_queue wait = { current, NULL };
 557         struct task_struct *p;
 558 
 559         if (stat_addr) {
 560                 flag = verify_area(VERIFY_WRITE, stat_addr, sizeof(*stat_addr));
 561                 if (flag)
 562                         return flag;
 563         }
 564         if (ru) {
 565                 flag = verify_area(VERIFY_WRITE, ru, sizeof(*ru));
 566                 if (flag)
 567                         return flag;
 568         }
 569         add_wait_queue(&current->wait_chldexit,&wait);
 570 repeat:
 571         flag=0;
 572         for (p = current->p_cptr ; p ; p = p->p_osptr) {
 573                 if (pid>0) {
 574                         if (p->pid != pid)
 575                                 continue;
 576                 } else if (!pid) {
 577                         if (p->pgrp != current->pgrp)
 578                                 continue;
 579                 } else if (pid != -1) {
 580                         if (p->pgrp != -pid)
 581                                 continue;
 582                 }
 583                 /* wait for cloned processes iff the __WCLONE flag is set */
 584                 if ((p->exit_signal != SIGCHLD) ^ ((options & __WCLONE) != 0))
 585                         continue;
 586                 flag = 1;
 587                 switch (p->state) {
 588                         case TASK_STOPPED:
 589                                 if (!p->exit_code)
 590                                         continue;
 591                                 if (!(options & WUNTRACED) && !(p->flags & PF_PTRACED))
 592                                         continue;
 593                                 if (ru != NULL)
 594                                         getrusage(p, RUSAGE_BOTH, ru);
 595                                 if (stat_addr)
 596                                         put_user((p->exit_code << 8) | 0x7f,
 597                                                 stat_addr);
 598                                 p->exit_code = 0;
 599                                 retval = p->pid;
 600                                 goto end_wait4;
 601                         case TASK_ZOMBIE:
 602                                 current->cutime += p->utime + p->cutime;
 603                                 current->cstime += p->stime + p->cstime;
 604                                 if (ru != NULL)
 605                                         getrusage(p, RUSAGE_BOTH, ru);
 606                                 if (stat_addr)
 607                                         put_user(p->exit_code, stat_addr);
 608                                 retval = p->pid;
 609                                 if (p->p_opptr != p->p_pptr) {
 610                                         REMOVE_LINKS(p);
 611                                         p->p_pptr = p->p_opptr;
 612                                         SET_LINKS(p);
 613                                         notify_parent(p);
 614                                 } else
 615                                         release(p);
 616 #ifdef DEBUG_PROC_TREE
 617                                 audit_ptree();
 618 #endif
 619                                 goto end_wait4;
 620                         default:
 621                                 continue;
 622                 }
 623         }
 624         if (flag) {
 625                 retval = 0;
 626                 if (options & WNOHANG)
 627                         goto end_wait4;
 628                 current->state=TASK_INTERRUPTIBLE;
 629                 schedule();
 630                 current->signal &= ~(1<<(SIGCHLD-1));
 631                 retval = -ERESTARTSYS;
 632                 if (current->signal & ~current->blocked)
 633                         goto end_wait4;
 634                 goto repeat;
 635         }
 636         retval = -ECHILD;
 637 end_wait4:
 638         remove_wait_queue(&current->wait_chldexit,&wait);
 639         return retval;
 640 }
 641 
 642 #ifndef __alpha__
 643 
 644 /*
 645  * sys_waitpid() remains for compatibility. waitpid() should be
 646  * implemented by calling sys_wait4() from libc.a.
 647  */
 648 asmlinkage int sys_waitpid(pid_t pid,unsigned int * stat_addr, int options)
     /*  */
 649 {
 650         return sys_wait4(pid, stat_addr, options, NULL);
 651 }
 652 
 653 #endif
/* */
root/kernel/exit.c

DEFINITIONS
