kernel/exit.c

   1 /*
   2  *  linux/kernel/exit.c
   3  *
   4  *  Copyright (C) 1991, 1992  Linus Torvalds
   5  */
   6
   7 #include <linux/config.h>
   8 #include <linux/malloc.h>
   9 #include <linux/interrupt.h>
  10 #include <linux/smp_lock.h>
  11 #include <linux/module.h>
  12 #ifdef CONFIG_BSD_PROCESS_ACCT
  13 #include <linux/acct.h>
  14 #endif
  15
  16 #include <asm/uaccess.h>
  17 #include <asm/pgtable.h>
  18 #include <asm/mmu_context.h>
  19
  20 extern void sem_exit (void);
  21 extern struct task_struct *child_reaper;
  22
  23 int getrusage(struct task_struct *, int, struct rusage *);
  24
  25 void release(struct task_struct * p)
  26 {
  27         if (p != current) {
  28 #ifdef __SMP__
  29                 int has_cpu;
  30
  31                 /*
  32                  * Wait to make sure the process isn't on the
  33                  * runqueue (active on some other CPU still)
  34                  */
  35                 do {
  36                         spin_lock_irq(&runqueue_lock);
  37                         has_cpu = p->has_cpu;
  38                         spin_unlock_irq(&runqueue_lock);
  39                 } while (has_cpu);
  40 #endif
  41                 free_uid(p);
  42                 unhash_process(p);
  43
  44                 release_thread(p);
  45                 current->cmin_flt += p->min_flt + p->cmin_flt;
  46                 current->cmaj_flt += p->maj_flt + p->cmaj_flt;
  47                 current->cnswap += p->nswap + p->cnswap;
  48                 free_task_struct(p);
  49         } else {
  50                 printk("task releasing itself\n");
  51         }
  52 }
  53
  54 /*
  55  * This checks not only the pgrp, but falls back on the pid if no
  56  * satisfactory pgrp is found. I dunno - gdb doesn't work correctly
  57  * without this...
  58  */
  59 int session_of_pgrp(int pgrp)
  60 {
  61         struct task_struct *p;
  62         int fallback;
  63
  64         fallback = -1;
  65         read_lock(&tasklist_lock);
  66         for_each_task(p) {
  67                 if (p->session <= 0)
  68                         continue;
  69                 if (p->pgrp == pgrp) {
  70                         fallback = p->session;
  71                         break;
  72                 }
  73                 if (p->pid == pgrp)
  74                         fallback = p->session;
  75         }
  76         read_unlock(&tasklist_lock);
  77         return fallback;
  78 }
  79
  80 /*
  81  * Determine if a process group is "orphaned", according to the POSIX
  82  * definition in 2.2.2.52.  Orphaned process groups are not to be affected
  83  * by terminal-generated stop signals.  Newly orphaned process groups are
  84  * to receive a SIGHUP and a SIGCONT.
  85  *
  86  * "I ask you, have you ever known what it is to be an orphan?"
  87  */
  88 static int will_become_orphaned_pgrp(int pgrp, struct task_struct * ignored_task)
  89 {
  90         struct task_struct *p;
  91
  92         read_lock(&tasklist_lock);
  93         for_each_task(p) {
  94                 if ((p == ignored_task) || (p->pgrp != pgrp) ||
  95                     (p->state == TASK_ZOMBIE) ||
  96                     (p->p_pptr->pid == 1))
  97                         continue;
  98                 if ((p->p_pptr->pgrp != pgrp) &&
  99                     (p->p_pptr->session == p->session)) {
 100                         read_unlock(&tasklist_lock);
 101                         return 0;
 102                 }
 103         }
 104         read_unlock(&tasklist_lock);
 105         return 1;       /* (sighing) "Often!" */
 106 }
 107
 108 int is_orphaned_pgrp(int pgrp)
 109 {
 110         return will_become_orphaned_pgrp(pgrp, 0);
 111 }
 112
 113 static inline int has_stopped_jobs(int pgrp)
 114 {
 115         int retval = 0;
 116         struct task_struct * p;
 117
 118         read_lock(&tasklist_lock);
 119         for_each_task(p) {
 120                 if (p->pgrp != pgrp)
 121                         continue;
 122                 if (p->state != TASK_STOPPED)
 123                         continue;
 124                 retval = 1;
 125                 break;
 126         }
 127         read_unlock(&tasklist_lock);
 128         return retval;
 129 }
 130
 131 static inline void forget_original_parent(struct task_struct * father)
 132 {
 133         struct task_struct * p;
 134
 135         read_lock(&tasklist_lock);
 136         for_each_task(p) {
 137                 if (p->p_opptr == father) {
 138                         /* We dont want people slaying init */
 139                         p->exit_signal = SIGCHLD;
 140                         p->self_exec_id++;
 141                         p->p_opptr = child_reaper; /* init */
 142                         if (p->pdeath_signal) send_sig(p->pdeath_signal, p, 0);
 143                 }
 144         }
 145         read_unlock(&tasklist_lock);
 146 }
 147
 148 static inline void close_files(struct files_struct * files)
 149 {
 150         int i, j;
 151
 152         j = 0;
 153         for (;;) {
 154                 unsigned long set;
 155                 i = j * __NFDBITS;
 156                 if (i >= files->max_fdset || i >= files->max_fds)
 157                         break;
 158                 set = files->open_fds->fds_bits[j++];
 159                 while (set) {
 160                         if (set & 1) {
 161                                 struct file * file = xchg(&files->fd[i], NULL);
 162                                 if (file)
 163                                         filp_close(file, files);
 164                         }
 165                         i++;
 166                         set >>= 1;
 167                 }
 168         }
 169 }
 170
 171 extern kmem_cache_t *files_cachep;
 172
 173 static inline void __exit_files(struct task_struct *tsk)
 174 {
 175         struct files_struct * files = xchg(&tsk->files, NULL);
 176
 177         if (files) {
 178                 if (atomic_dec_and_test(&files->count)) {
 179                         close_files(files);
 180                         /*
 181                          * Free the fd and fdset arrays if we expanded them.
 182                          */
 183                         if (files->fd != &files->fd_array[0])
 184                                 free_fd_array(files->fd, files->max_fds);
 185                         if (files->max_fdset > __FD_SETSIZE) {
 186                                 free_fdset(files->open_fds, files->max_fdset);
 187                                 free_fdset(files->close_on_exec, files->max_fdset);
 188                         }
 189                         kmem_cache_free(files_cachep, files);
 190                 }
 191         }
 192 }
 193
 194 void exit_files(struct task_struct *tsk)
 195 {
 196         __exit_files(tsk);
 197 }
 198
 199 static inline void __exit_fs(struct task_struct *tsk)
 200 {
 201         struct fs_struct * fs = tsk->fs;
 202
 203         if (fs) {
 204                 tsk->fs = NULL;
 205                 if (atomic_dec_and_test(&fs->count)) {
 206                         dput(fs->root);
 207                         dput(fs->pwd);
 208                         kfree(fs);
 209                 }
 210         }
 211 }
 212
 213 void exit_fs(struct task_struct *tsk)
 214 {
 215         __exit_fs(tsk);
 216 }
 217
 218 static inline void __exit_sighand(struct task_struct *tsk)
 219 {
 220         struct signal_struct * sig = tsk->sig;
 221
 222         if (sig) {
 223                 unsigned long flags;
 224
 225                 spin_lock_irqsave(&tsk->sigmask_lock, flags);
 226                 tsk->sig = NULL;
 227                 spin_unlock_irqrestore(&tsk->sigmask_lock, flags);
 228                 if (atomic_dec_and_test(&sig->count))
 229                         kfree(sig);
 230         }
 231
 232         flush_signals(tsk);
 233 }
 234
 235 void exit_sighand(struct task_struct *tsk)
 236 {
 237         __exit_sighand(tsk);
 238 }
 239
 240 /*
 241  * We can use these to temporarily drop into
 242  * "lazy TLB" mode and back.
 243  */
 244 struct mm_struct * start_lazy_tlb(void)
 245 {
 246         struct mm_struct *mm = current->mm;
 247         current->mm = NULL;
 248         /* active_mm is still 'mm' */
 249         atomic_inc(&mm->mm_count);
 250         enter_lazy_tlb(mm, current, smp_processor_id());
 251         return mm;
 252 }
 253
 254 void end_lazy_tlb(struct mm_struct *mm)
 255 {
 256         struct mm_struct *active_mm = current->active_mm;
 257
 258         current->mm = mm;
 259         if (mm != active_mm) {
 260                 current->active_mm = mm;
 261                 activate_mm(active_mm, mm);
 262         }
 263         mmdrop(active_mm);
 264 }
 265
 266 /*
 267  * Turn us into a lazy TLB process if we
 268  * aren't already..
 269  */
 270 static inline void __exit_mm(struct task_struct * tsk)
 271 {
 272         struct mm_struct * mm = tsk->mm;
 273
 274         if (mm) {
 275                 atomic_inc(&mm->mm_count);
 276                 mm_release();
 277                 if (mm != tsk->active_mm) BUG();
 278                 tsk->mm = NULL;
 279                 enter_lazy_tlb(mm, current, smp_processor_id());
 280                 mmput(mm);
 281         }
 282 }
 283
 284 void exit_mm(struct task_struct *tsk)
 285 {
 286         __exit_mm(tsk);
 287 }
 288
 289 /*
 290  * Send signals to all our closest relatives so that they know
 291  * to properly mourn us..
 292  */
 293 static void exit_notify(void)
 294 {
 295         struct task_struct * p, *t;
 296
 297         forget_original_parent(current);
 298         /*
 299          * Check to see if any process groups have become orphaned
 300          * as a result of our exiting, and if they have any stopped
 301          * jobs, send them a SIGHUP and then a SIGCONT.  (POSIX 3.2.2.2)
 302          *
 303          * Case i: Our father is in a different pgrp than we are
 304          * and we were the only connection outside, so our pgrp
 305          * is about to become orphaned.
 306          */
 307
 308         t = current->p_pptr;
 309
 310         if ((t->pgrp != current->pgrp) &&
 311             (t->session == current->session) &&
 312             will_become_orphaned_pgrp(current->pgrp, current) &&
 313             has_stopped_jobs(current->pgrp)) {
 314                 kill_pg(current->pgrp,SIGHUP,1);
 315                 kill_pg(current->pgrp,SIGCONT,1);
 316         }
 317
 318         /* Let father know we died
 319          *
 320          * Thread signals are configurable, but you aren't going to use
 321          * that to send signals to arbitary processes.
 322          * That stops right now.
 323          *
 324          * If the parent exec id doesn't match the exec id we saved
 325          * when we started then we know the parent has changed security
 326          * domain.
 327          *
 328          * If our self_exec id doesn't match our parent_exec_id then
 329          * we have changed execution domain as these two values started
 330          * the same after a fork.
 331          *
 332          */
 333
 334         if(current->exit_signal != SIGCHLD &&
 335             ( current->parent_exec_id != t->self_exec_id  ||
 336               current->self_exec_id != current->parent_exec_id)
 337             && !capable(CAP_KILL))
 338                 current->exit_signal = SIGCHLD;
 339
 340         notify_parent(current, current->exit_signal);
 341
 342         /*
 343          * This loop does two things:
 344          *
 345          * A.  Make init inherit all the child processes
 346          * B.  Check to see if any process groups have become orphaned
 347          *      as a result of our exiting, and if they have any stopped
 348          *      jobs, send them a SIGHUP and then a SIGCONT.  (POSIX 3.2.2.2)
 349          */
 350
 351         write_lock_irq(&tasklist_lock);
 352         while (current->p_cptr != NULL) {
 353                 p = current->p_cptr;
 354                 current->p_cptr = p->p_osptr;
 355                 p->p_ysptr = NULL;
 356                 p->flags &= ~(PF_PTRACED|PF_TRACESYS);
 357
 358                 p->p_pptr = p->p_opptr;
 359                 p->p_osptr = p->p_pptr->p_cptr;
 360                 if (p->p_osptr)
 361                         p->p_osptr->p_ysptr = p;
 362                 p->p_pptr->p_cptr = p;
 363                 if (p->state == TASK_ZOMBIE)
 364                         notify_parent(p, p->exit_signal);
 365                 /*
 366                  * process group orphan check
 367                  * Case ii: Our child is in a different pgrp
 368                  * than we are, and it was the only connection
 369                  * outside, so the child pgrp is now orphaned.
 370                  */
 371                 if ((p->pgrp != current->pgrp) &&
 372                     (p->session == current->session)) {
 373                         int pgrp = p->pgrp;
 374
 375                         write_unlock_irq(&tasklist_lock);
 376                         if (is_orphaned_pgrp(pgrp) && has_stopped_jobs(pgrp)) {
 377                                 kill_pg(pgrp,SIGHUP,1);
 378                                 kill_pg(pgrp,SIGCONT,1);
 379                         }
 380                         write_lock_irq(&tasklist_lock);
 381                 }
 382         }
 383         write_unlock_irq(&tasklist_lock);
 384
 385         if (current->leader)
 386                 disassociate_ctty(1);
 387 }
 388
 389 NORET_TYPE void do_exit(long code)
 390 {
 391         struct task_struct *tsk = current;
 392
 393         if (in_interrupt())
 394                 printk("Aiee, killing interrupt handler\n");
 395         if (!tsk->pid)
 396                 panic("Attempted to kill the idle task!");
 397         tsk->flags |= PF_EXITING;
 398         del_timer_sync(&tsk->real_timer);
 399
 400         lock_kernel();
 401 fake_volatile:
 402 #ifdef CONFIG_BSD_PROCESS_ACCT
 403         acct_process(code);
 404 #endif
 405         task_lock(tsk);
 406         sem_exit();
 407         __exit_mm(tsk);
 408         __exit_files(tsk);
 409         __exit_fs(tsk);
 410         __exit_sighand(tsk);
 411         exit_thread();
 412         tsk->state = TASK_ZOMBIE;
 413         tsk->exit_code = code;
 414         exit_notify();
 415         task_unlock(tsk);
 416         if (tsk->exec_domain && tsk->exec_domain->module)
 417                 __MOD_DEC_USE_COUNT(tsk->exec_domain->module);
 418         if (tsk->binfmt && tsk->binfmt->module)
 419                 __MOD_DEC_USE_COUNT(tsk->binfmt->module);
 420         schedule();
 421 /*
 422  * In order to get rid of the "volatile function does return" message
 423  * I did this little loop that confuses gcc to think do_exit really
 424  * is volatile. In fact it's schedule() that is volatile in some
 425  * circumstances: when current->state = ZOMBIE, schedule() never
 426  * returns.
 427  *
 428  * In fact the natural way to do all this is to have the label and the
 429  * goto right after each other, but I put the fake_volatile label at
 430  * the start of the function just in case something /really/ bad
 431  * happens, and the schedule returns. This way we can try again. I'm
 432  * not paranoid: it's just that everybody is out to get me.
 433  */
 434         goto fake_volatile;
 435 }
 436
 437 asmlinkage long sys_exit(int error_code)
 438 {
 439         do_exit((error_code&0xff)<<8);
 440 }
 441
 442 asmlinkage long sys_wait4(pid_t pid,unsigned int * stat_addr, int options, struct rusage * ru)
 443 {
 444         int flag, retval;
 445         DECLARE_WAITQUEUE(wait, current);
 446         struct task_struct *p;
 447
 448         if (options & ~(WNOHANG|WUNTRACED|__WCLONE|__WALL))
 449                 return -EINVAL;
 450
 451         add_wait_queue(&current->wait_chldexit,&wait);
 452 repeat:
 453         flag = 0;
 454         current->state = TASK_INTERRUPTIBLE;
 455         read_lock(&tasklist_lock);
 456         for (p = current->p_cptr ; p ; p = p->p_osptr) {
 457                 if (pid>0) {
 458                         if (p->pid != pid)
 459                                 continue;
 460                 } else if (!pid) {
 461                         if (p->pgrp != current->pgrp)
 462                                 continue;
 463                 } else if (pid != -1) {
 464                         if (p->pgrp != -pid)
 465                                 continue;
 466                 }
 467                 /* Wait for all children (clone and not) if __WALL is set;
 468                  * otherwise, wait for clone children *only* if __WCLONE is
 469                  * set; otherwise, wait for non-clone children *only*.  (Note:
 470                  * A "clone" child here is one that reports to its parent
 471                  * using a signal other than SIGCHLD.) */
 472                 if (((p->exit_signal != SIGCHLD) ^ ((options & __WCLONE) != 0))
 473                     && !(options & __WALL))
 474                         continue;
 475                 flag = 1;
 476                 switch (p->state) {
 477                         case TASK_STOPPED:
 478                                 if (!p->exit_code)
 479                                         continue;
 480                                 if (!(options & WUNTRACED) && !(p->flags & PF_PTRACED))
 481                                         continue;
 482                                 read_unlock(&tasklist_lock);
 483                                 retval = ru ? getrusage(p, RUSAGE_BOTH, ru) : 0;
 484                                 if (!retval && stat_addr)
 485                                         retval = put_user((p->exit_code << 8) | 0x7f, stat_addr);
 486                                 if (!retval) {
 487                                         p->exit_code = 0;
 488                                         retval = p->pid;
 489                                 }
 490                                 goto end_wait4;
 491                         case TASK_ZOMBIE:
 492                                 current->times.tms_cutime += p->times.tms_utime + p->times.tms_cutime;
 493                                 current->times.tms_cstime += p->times.tms_stime + p->times.tms_cstime;
 494                                 read_unlock(&tasklist_lock);
 495                                 retval = ru ? getrusage(p, RUSAGE_BOTH, ru) : 0;
 496                                 if (!retval && stat_addr)
 497                                         retval = put_user(p->exit_code, stat_addr);
 498                                 if (retval)
 499                                         goto end_wait4;
 500                                 retval = p->pid;
 501                                 if (p->p_opptr != p->p_pptr) {
 502                                         write_lock_irq(&tasklist_lock);
 503                                         REMOVE_LINKS(p);
 504                                         p->p_pptr = p->p_opptr;
 505                                         SET_LINKS(p);
 506                                         write_unlock_irq(&tasklist_lock);
 507                                         notify_parent(p, SIGCHLD);
 508                                 } else
 509                                         release(p);
 510                                 goto end_wait4;
 511                         default:
 512                                 continue;
 513                 }
 514         }
 515         read_unlock(&tasklist_lock);
 516         if (flag) {
 517                 retval = 0;
 518                 if (options & WNOHANG)
 519                         goto end_wait4;
 520                 retval = -ERESTARTSYS;
 521                 if (signal_pending(current))
 522                         goto end_wait4;
 523                 schedule();
 524                 goto repeat;
 525         }
 526         retval = -ECHILD;
 527 end_wait4:
 528         current->state = TASK_RUNNING;
 529         remove_wait_queue(&current->wait_chldexit,&wait);
 530         return retval;
 531 }
 532
 533 #if !defined(__alpha__) && !defined(__ia64__)
 534
 535 /*
 536  * sys_waitpid() remains for compatibility. waitpid() should be
 537  * implemented by calling sys_wait4() from libc.a.
 538  */
 539 asmlinkage long sys_waitpid(pid_t pid,unsigned int * stat_addr, int options)
 540 {
 541         return sys_wait4(pid, stat_addr, options, NULL);
 542 }
 543
 544 #endif