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 static void release(struct task_struct * p)
  26 {
  27         if (p != current) {
  28 #ifdef __SMP__
  29                 /*
  30                  * Wait to make sure the process isn't on the
  31                  * runqueue (active on some other CPU still)
  32                  */
  33                 do {
  34                         barrier();
  35                 } while (p->has_cpu);
  36 #endif
  37                 free_uid(p);
  38                 unhash_process(p);
  39
  40                 release_thread(p);
  41                 current->cmin_flt += p->min_flt + p->cmin_flt;
  42                 current->cmaj_flt += p->maj_flt + p->cmaj_flt;
  43                 current->cnswap += p->nswap + p->cnswap;
  44                 free_task_struct(p);
  45         } else {
  46                 printk("task releasing itself\n");
  47         }
  48 }
  49
  50 /*
  51  * This checks not only the pgrp, but falls back on the pid if no
  52  * satisfactory pgrp is found. I dunno - gdb doesn't work correctly
  53  * without this...
  54  */
  55 int session_of_pgrp(int pgrp)
  56 {
  57         struct task_struct *p;
  58         int fallback;
  59
  60         fallback = -1;
  61         read_lock(&tasklist_lock);
  62         for_each_task(p) {
  63                 if (p->session <= 0)
  64                         continue;
  65                 if (p->pgrp == pgrp) {
  66                         fallback = p->session;
  67                         break;
  68                 }
  69                 if (p->pid == pgrp)
  70                         fallback = p->session;
  71         }
  72         read_unlock(&tasklist_lock);
  73         return fallback;
  74 }
  75
  76 /*
  77  * Determine if a process group is "orphaned", according to the POSIX
  78  * definition in 2.2.2.52.  Orphaned process groups are not to be affected
  79  * by terminal-generated stop signals.  Newly orphaned process groups are
  80  * to receive a SIGHUP and a SIGCONT.
  81  *
  82  * "I ask you, have you ever known what it is to be an orphan?"
  83  */
  84 static int will_become_orphaned_pgrp(int pgrp, struct task_struct * ignored_task)
  85 {
  86         struct task_struct *p;
  87
  88         read_lock(&tasklist_lock);
  89         for_each_task(p) {
  90                 if ((p == ignored_task) || (p->pgrp != pgrp) ||
  91                     (p->state == TASK_ZOMBIE) ||
  92                     (p->p_pptr->pid == 1))
  93                         continue;
  94                 if ((p->p_pptr->pgrp != pgrp) &&
  95                     (p->p_pptr->session == p->session)) {
  96                         read_unlock(&tasklist_lock);
  97                         return 0;
  98                 }
  99         }
 100         read_unlock(&tasklist_lock);
 101         return 1;       /* (sighing) "Often!" */
 102 }
 103
 104 int is_orphaned_pgrp(int pgrp)
 105 {
 106         return will_become_orphaned_pgrp(pgrp, 0);
 107 }
 108
 109 static inline int has_stopped_jobs(int pgrp)
 110 {
 111         int retval = 0;
 112         struct task_struct * p;
 113
 114         read_lock(&tasklist_lock);
 115         for_each_task(p) {
 116                 if (p->pgrp != pgrp)
 117                         continue;
 118                 if (p->state != TASK_STOPPED)
 119                         continue;
 120                 retval = 1;
 121                 break;
 122         }
 123         read_unlock(&tasklist_lock);
 124         return retval;
 125 }
 126
 127 static inline void forget_original_parent(struct task_struct * father)
 128 {
 129         struct task_struct * p;
 130
 131         read_lock(&tasklist_lock);
 132         for_each_task(p) {
 133                 if (p->p_opptr == father) {
 134                         p->exit_signal = SIGCHLD;
 135                         p->p_opptr = child_reaper; /* init */
 136                         if (p->pdeath_signal) send_sig(p->pdeath_signal, p, 0);
 137                 }
 138         }
 139         read_unlock(&tasklist_lock);
 140 }
 141
 142 static inline void close_files(struct files_struct * files)
 143 {
 144         int i, j;
 145
 146         j = 0;
 147         for (;;) {
 148                 unsigned long set = files->open_fds.fds_bits[j];
 149                 i = j * __NFDBITS;
 150                 j++;
 151                 if (i >= files->max_fds)
 152                         break;
 153                 while (set) {
 154                         if (set & 1) {
 155                                 struct file * file = xchg(&files->fd[i], NULL);
 156                                 if (file)
 157                                         filp_close(file, files);
 158                         }
 159                         i++;
 160                         set >>= 1;
 161                 }
 162         }
 163 }
 164
 165 extern kmem_cache_t *files_cachep;
 166
 167 static inline void __exit_files(struct task_struct *tsk)
 168 {
 169         struct files_struct * files = xchg(&tsk->files, NULL);
 170
 171         if (files) {
 172                 if (atomic_dec_and_test(&files->count)) {
 173                         close_files(files);
 174                         /*
 175                          * Free the fd array as appropriate ...
 176                          */
 177                         if (NR_OPEN * sizeof(struct file *) == PAGE_SIZE)
 178                                 free_page((unsigned long) files->fd);
 179                         else
 180                                 kfree(files->fd);
 181                         kmem_cache_free(files_cachep, files);
 182                 }
 183         }
 184 }
 185
 186 void exit_files(struct task_struct *tsk)
 187 {
 188         __exit_files(tsk);
 189 }
 190
 191 static inline void __exit_fs(struct task_struct *tsk)
 192 {
 193         struct fs_struct * fs = tsk->fs;
 194
 195         if (fs) {
 196                 tsk->fs = NULL;
 197                 if (atomic_dec_and_test(&fs->count)) {
 198                         dput(fs->root);
 199                         dput(fs->pwd);
 200                         kfree(fs);
 201                 }
 202         }
 203 }
 204
 205 void exit_fs(struct task_struct *tsk)
 206 {
 207         __exit_fs(tsk);
 208 }
 209
 210 static inline void __exit_sighand(struct task_struct *tsk)
 211 {
 212         struct signal_struct * sig = tsk->sig;
 213
 214         if (sig) {
 215                 unsigned long flags;
 216
 217                 spin_lock_irqsave(&tsk->sigmask_lock, flags);
 218                 tsk->sig = NULL;
 219                 spin_unlock_irqrestore(&tsk->sigmask_lock, flags);
 220                 if (atomic_dec_and_test(&sig->count))
 221                         kfree(sig);
 222         }
 223
 224         flush_signals(tsk);
 225 }
 226
 227 void exit_sighand(struct task_struct *tsk)
 228 {
 229         __exit_sighand(tsk);
 230 }
 231
 232 /*
 233  * Turn us into a lazy TLB process if we
 234  * aren't already..
 235  */
 236 static inline void __exit_mm(struct task_struct * tsk)
 237 {
 238         struct mm_struct * mm = tsk->mm;
 239
 240         if (mm) {
 241                 mm_release();
 242                 atomic_inc(&mm->mm_count);
 243                 tsk->mm = NULL;
 244                 mmput(mm);
 245         }
 246 }
 247
 248 void exit_mm(struct task_struct *tsk)
 249 {
 250         __exit_mm(tsk);
 251 }
 252
 253 /*
 254  * Send signals to all our closest relatives so that they know
 255  * to properly mourn us..
 256  */
 257 static void exit_notify(void)
 258 {
 259         struct task_struct * p;
 260
 261         forget_original_parent(current);
 262         /*
 263          * Check to see if any process groups have become orphaned
 264          * as a result of our exiting, and if they have any stopped
 265          * jobs, send them a SIGHUP and then a SIGCONT.  (POSIX 3.2.2.2)
 266          *
 267          * Case i: Our father is in a different pgrp than we are
 268          * and we were the only connection outside, so our pgrp
 269          * is about to become orphaned.
 270          */
 271         if ((current->p_pptr->pgrp != current->pgrp) &&
 272             (current->p_pptr->session == current->session) &&
 273             will_become_orphaned_pgrp(current->pgrp, current) &&
 274             has_stopped_jobs(current->pgrp)) {
 275                 kill_pg(current->pgrp,SIGHUP,1);
 276                 kill_pg(current->pgrp,SIGCONT,1);
 277         }
 278
 279         /* Let father know we died */
 280         notify_parent(current, current->exit_signal);
 281
 282         /*
 283          * This loop does two things:
 284          *
 285          * A.  Make init inherit all the child processes
 286          * B.  Check to see if any process groups have become orphaned
 287          *      as a result of our exiting, and if they have any stopped
 288          *      jobs, send them a SIGHUP and then a SIGCONT.  (POSIX 3.2.2.2)
 289          */
 290
 291         write_lock_irq(&tasklist_lock);
 292         while (current->p_cptr != NULL) {
 293                 p = current->p_cptr;
 294                 current->p_cptr = p->p_osptr;
 295                 p->p_ysptr = NULL;
 296                 p->flags &= ~(PF_PTRACED|PF_TRACESYS);
 297
 298                 p->p_pptr = p->p_opptr;
 299                 p->p_osptr = p->p_pptr->p_cptr;
 300                 if (p->p_osptr)
 301                         p->p_osptr->p_ysptr = p;
 302                 p->p_pptr->p_cptr = p;
 303                 if (p->state == TASK_ZOMBIE)
 304                         notify_parent(p, p->exit_signal);
 305                 /*
 306                  * process group orphan check
 307                  * Case ii: Our child is in a different pgrp
 308                  * than we are, and it was the only connection
 309                  * outside, so the child pgrp is now orphaned.
 310                  */
 311                 if ((p->pgrp != current->pgrp) &&
 312                     (p->session == current->session)) {
 313                         int pgrp = p->pgrp;
 314
 315                         write_unlock_irq(&tasklist_lock);
 316                         if (is_orphaned_pgrp(pgrp) && has_stopped_jobs(pgrp)) {
 317                                 kill_pg(pgrp,SIGHUP,1);
 318                                 kill_pg(pgrp,SIGCONT,1);
 319                         }
 320                         write_lock_irq(&tasklist_lock);
 321                 }
 322         }
 323         write_unlock_irq(&tasklist_lock);
 324
 325         if (current->leader)
 326                 disassociate_ctty(1);
 327 }
 328
 329 NORET_TYPE void do_exit(long code)
 330 {
 331         struct task_struct *tsk = current;
 332
 333         if (in_interrupt())
 334                 printk("Aiee, killing interrupt handler\n");
 335         if (!tsk->pid)
 336                 panic("Attempted to kill the idle task!");
 337         tsk->flags |= PF_EXITING;
 338         start_bh_atomic();
 339         del_timer(&tsk->real_timer);
 340         end_bh_atomic();
 341
 342         lock_kernel();
 343 fake_volatile:
 344 #ifdef CONFIG_BSD_PROCESS_ACCT
 345         acct_process(code);
 346 #endif
 347         sem_exit();
 348         __exit_mm(tsk);
 349 #if CONFIG_AP1000
 350         exit_msc(tsk);
 351 #endif
 352         __exit_files(tsk);
 353         __exit_fs(tsk);
 354         __exit_sighand(tsk);
 355         exit_thread();
 356         tsk->state = TASK_ZOMBIE;
 357         tsk->exit_code = code;
 358         exit_notify();
 359 #ifdef DEBUG_PROC_TREE
 360         audit_ptree();
 361 #endif
 362         if (tsk->exec_domain && tsk->exec_domain->module)
 363                 __MOD_DEC_USE_COUNT(tsk->exec_domain->module);
 364         if (tsk->binfmt && tsk->binfmt->module)
 365                 __MOD_DEC_USE_COUNT(tsk->binfmt->module);
 366         schedule();
 367 /*
 368  * In order to get rid of the "volatile function does return" message
 369  * I did this little loop that confuses gcc to think do_exit really
 370  * is volatile. In fact it's schedule() that is volatile in some
 371  * circumstances: when current->state = ZOMBIE, schedule() never
 372  * returns.
 373  *
 374  * In fact the natural way to do all this is to have the label and the
 375  * goto right after each other, but I put the fake_volatile label at
 376  * the start of the function just in case something /really/ bad
 377  * happens, and the schedule returns. This way we can try again. I'm
 378  * not paranoid: it's just that everybody is out to get me.
 379  */
 380         goto fake_volatile;
 381 }
 382
 383 asmlinkage int sys_exit(int error_code)
 384 {
 385         do_exit((error_code&0xff)<<8);
 386 }
 387
 388 asmlinkage int sys_wait4(pid_t pid,unsigned int * stat_addr, int options, struct rusage * ru)
 389 {
 390         int flag, retval;
 391         DECLARE_WAITQUEUE(wait, current);
 392         struct task_struct *p;
 393
 394         if (options & ~(WNOHANG|WUNTRACED|__WCLONE))
 395                 return -EINVAL;
 396
 397         add_wait_queue(&current->wait_chldexit,&wait);
 398 repeat:
 399         flag = 0;
 400         read_lock(&tasklist_lock);
 401         for (p = current->p_cptr ; p ; p = p->p_osptr) {
 402                 if (pid>0) {
 403                         if (p->pid != pid)
 404                                 continue;
 405                 } else if (!pid) {
 406                         if (p->pgrp != current->pgrp)
 407                                 continue;
 408                 } else if (pid != -1) {
 409                         if (p->pgrp != -pid)
 410                                 continue;
 411                 }
 412                 /* wait for cloned processes iff the __WCLONE flag is set */
 413                 if ((p->exit_signal != SIGCHLD) ^ ((options & __WCLONE) != 0))
 414                         continue;
 415                 flag = 1;
 416                 switch (p->state) {
 417                         case TASK_STOPPED:
 418                                 if (!p->exit_code)
 419                                         continue;
 420                                 if (!(options & WUNTRACED) && !(p->flags & PF_PTRACED))
 421                                         continue;
 422                                 read_unlock(&tasklist_lock);
 423                                 retval = ru ? getrusage(p, RUSAGE_BOTH, ru) : 0;
 424                                 if (!retval && stat_addr)
 425                                         retval = put_user((p->exit_code << 8) | 0x7f, stat_addr);
 426                                 if (!retval) {
 427                                         p->exit_code = 0;
 428                                         retval = p->pid;
 429                                 }
 430                                 goto end_wait4;
 431                         case TASK_ZOMBIE:
 432                                 current->times.tms_cutime += p->times.tms_utime + p->times.tms_cutime;
 433                                 current->times.tms_cstime += p->times.tms_stime + p->times.tms_cstime;
 434                                 read_unlock(&tasklist_lock);
 435                                 retval = ru ? getrusage(p, RUSAGE_BOTH, ru) : 0;
 436                                 if (!retval && stat_addr)
 437                                         retval = put_user(p->exit_code, stat_addr);
 438                                 if (retval)
 439                                         goto end_wait4;
 440                                 retval = p->pid;
 441                                 if (p->p_opptr != p->p_pptr) {
 442                                         write_lock_irq(&tasklist_lock);
 443                                         REMOVE_LINKS(p);
 444                                         p->p_pptr = p->p_opptr;
 445                                         SET_LINKS(p);
 446                                         write_unlock_irq(&tasklist_lock);
 447                                         notify_parent(p, SIGCHLD);
 448                                 } else
 449                                         release(p);
 450 #ifdef DEBUG_PROC_TREE
 451                                 audit_ptree();
 452 #endif
 453                                 goto end_wait4;
 454                         default:
 455                                 continue;
 456                 }
 457         }
 458         read_unlock(&tasklist_lock);
 459         if (flag) {
 460                 retval = 0;
 461                 if (options & WNOHANG)
 462                         goto end_wait4;
 463                 retval = -ERESTARTSYS;
 464                 if (signal_pending(current))
 465                         goto end_wait4;
 466                 current->state=TASK_INTERRUPTIBLE;
 467                 schedule();
 468                 goto repeat;
 469         }
 470         retval = -ECHILD;
 471 end_wait4:
 472         remove_wait_queue(&current->wait_chldexit,&wait);
 473         return retval;
 474 }
 475
 476 #ifndef __alpha__
 477
 478 /*
 479  * sys_waitpid() remains for compatibility. waitpid() should be
 480  * implemented by calling sys_wait4() from libc.a.
 481  */
 482 asmlinkage int sys_waitpid(pid_t pid,unsigned int * stat_addr, int options)
 483 {
 484         return sys_wait4(pid, stat_addr, options, NULL);
 485 }
 486
 487 #endif