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 active on any
  31                  * other CPU
  32                  */
  33                 for (;;)  {
  34                         int has_cpu;
  35                         spin_lock(&scheduler_lock);
  36                         has_cpu = p->has_cpu;
  37                         spin_unlock(&scheduler_lock);
  38                         if (!has_cpu)
  39                                 break;
  40                         do {
  41                                 barrier();
  42                         } while (p->has_cpu);
  43                 }
  44 #endif
  45                 free_uid(p);
  46                 nr_tasks--;
  47                 add_free_taskslot(p->tarray_ptr);
  48
  49                 write_lock_irq(&tasklist_lock);
  50                 unhash_pid(p);
  51                 REMOVE_LINKS(p);
  52                 write_unlock_irq(&tasklist_lock);
  53
  54                 release_thread(p);
  55                 current->cmin_flt += p->min_flt + p->cmin_flt;
  56                 current->cmaj_flt += p->maj_flt + p->cmaj_flt;
  57                 current->cnswap += p->nswap + p->cnswap;
  58                 free_task_struct(p);
  59         } else {
  60                 printk("task releasing itself\n");
  61         }
  62 }
  63
  64 /*
  65  * This checks not only the pgrp, but falls back on the pid if no
  66  * satisfactory pgrp is found. I dunno - gdb doesn't work correctly
  67  * without this...
  68  */
  69 int session_of_pgrp(int pgrp)
  70 {
  71         struct task_struct *p;
  72         int fallback;
  73
  74         fallback = -1;
  75         read_lock(&tasklist_lock);
  76         for_each_task(p) {
  77                 if (p->session <= 0)
  78                         continue;
  79                 if (p->pgrp == pgrp) {
  80                         fallback = p->session;
  81                         break;
  82                 }
  83                 if (p->pid == pgrp)
  84                         fallback = p->session;
  85         }
  86         read_unlock(&tasklist_lock);
  87         return fallback;
  88 }
  89
  90 /*
  91  * Determine if a process group is "orphaned", according to the POSIX
  92  * definition in 2.2.2.52.  Orphaned process groups are not to be affected
  93  * by terminal-generated stop signals.  Newly orphaned process groups are
  94  * to receive a SIGHUP and a SIGCONT.
  95  *
  96  * "I ask you, have you ever known what it is to be an orphan?"
  97  */
  98 static int will_become_orphaned_pgrp(int pgrp, struct task_struct * ignored_task)
  99 {
 100         struct task_struct *p;
 101
 102         read_lock(&tasklist_lock);
 103         for_each_task(p) {
 104                 if ((p == ignored_task) || (p->pgrp != pgrp) ||
 105                     (p->state == TASK_ZOMBIE) ||
 106                     (p->p_pptr->pid == 1))
 107                         continue;
 108                 if ((p->p_pptr->pgrp != pgrp) &&
 109                     (p->p_pptr->session == p->session)) {
 110                         read_unlock(&tasklist_lock);
 111                         return 0;
 112                 }
 113         }
 114         read_unlock(&tasklist_lock);
 115         return 1;       /* (sighing) "Often!" */
 116 }
 117
 118 int is_orphaned_pgrp(int pgrp)
 119 {
 120         return will_become_orphaned_pgrp(pgrp, 0);
 121 }
 122
 123 static inline int has_stopped_jobs(int pgrp)
 124 {
 125         int retval = 0;
 126         struct task_struct * p;
 127
 128         read_lock(&tasklist_lock);
 129         for_each_task(p) {
 130                 if (p->pgrp != pgrp)
 131                         continue;
 132                 if (p->state != TASK_STOPPED)
 133                         continue;
 134                 retval = 1;
 135                 break;
 136         }
 137         read_unlock(&tasklist_lock);
 138         return retval;
 139 }
 140
 141 static inline void forget_original_parent(struct task_struct * father)
 142 {
 143         struct task_struct * p;
 144
 145         read_lock(&tasklist_lock);
 146         for_each_task(p) {
 147                 if (p->p_opptr == father) {
 148                         p->exit_signal = SIGCHLD;
 149                         p->p_opptr = child_reaper; /* init */
 150                         if (p->pdeath_signal) send_sig(p->pdeath_signal, p, 0);
 151                 }
 152         }
 153         read_unlock(&tasklist_lock);
 154 }
 155
 156 static inline void close_files(struct files_struct * files)
 157 {
 158         int i, j;
 159
 160         j = 0;
 161         for (;;) {
 162                 unsigned long set = files->open_fds.fds_bits[j];
 163                 i = j * __NFDBITS;
 164                 j++;
 165                 if (i >= files->max_fds)
 166                         break;
 167                 while (set) {
 168                         if (set & 1) {
 169                                 struct file * file = files->fd[i];
 170                                 if (file) {
 171                                         files->fd[i] = NULL;
 172                                         filp_close(file, files);
 173                                 }
 174                         }
 175                         i++;
 176                         set >>= 1;
 177                 }
 178         }
 179 }
 180
 181 extern kmem_cache_t *files_cachep;
 182
 183 static inline void __exit_files(struct task_struct *tsk)
 184 {
 185         struct files_struct * files = tsk->files;
 186
 187         if (files) {
 188                 tsk->files = NULL;
 189                 if (atomic_dec_and_test(&files->count)) {
 190                         close_files(files);
 191                         /*
 192                          * Free the fd array as appropriate ...
 193                          */
 194                         if (NR_OPEN * sizeof(struct file *) == PAGE_SIZE)
 195                                 free_page((unsigned long) files->fd);
 196                         else
 197                                 kfree(files->fd);
 198                         kmem_cache_free(files_cachep, files);
 199                 }
 200         }
 201 }
 202
 203 void exit_files(struct task_struct *tsk)
 204 {
 205         __exit_files(tsk);
 206 }
 207
 208 static inline void __exit_fs(struct task_struct *tsk)
 209 {
 210         struct fs_struct * fs = tsk->fs;
 211
 212         if (fs) {
 213                 tsk->fs = NULL;
 214                 if (atomic_dec_and_test(&fs->count)) {
 215                         dput(fs->root);
 216                         dput(fs->pwd);
 217                         kfree(fs);
 218                 }
 219         }
 220 }
 221
 222 void exit_fs(struct task_struct *tsk)
 223 {
 224         __exit_fs(tsk);
 225 }
 226
 227 static inline void __exit_sighand(struct task_struct *tsk)
 228 {
 229         struct signal_struct * sig = tsk->sig;
 230
 231         if (sig) {
 232                 unsigned long flags;
 233
 234                 spin_lock_irqsave(&tsk->sigmask_lock, flags);
 235                 tsk->sig = NULL;
 236                 spin_unlock_irqrestore(&tsk->sigmask_lock, flags);
 237                 if (atomic_dec_and_test(&sig->count))
 238                         kfree(sig);
 239         }
 240
 241         flush_signals(tsk);
 242 }
 243
 244 void exit_sighand(struct task_struct *tsk)
 245 {
 246         __exit_sighand(tsk);
 247 }
 248
 249 static inline void __exit_mm(struct task_struct * tsk)
 250 {
 251         struct mm_struct * mm = tsk->mm;
 252
 253         /* Set us up to use the kernel mm state */
 254         if (mm != &init_mm) {
 255                 flush_cache_mm(mm);
 256                 flush_tlb_mm(mm);
 257                 destroy_context(mm);
 258                 tsk->mm = &init_mm;
 259                 tsk->swappable = 0;
 260                 SET_PAGE_DIR(tsk, swapper_pg_dir);
 261                 mm_release();
 262                 mmput(mm);
 263         }
 264 }
 265
 266 void exit_mm(struct task_struct *tsk)
 267 {
 268         __exit_mm(tsk);
 269 }
 270
 271 /*
 272  * Send signals to all our closest relatives so that they know
 273  * to properly mourn us..
 274  */
 275 static void exit_notify(void)
 276 {
 277         struct task_struct * p;
 278
 279         forget_original_parent(current);
 280         /*
 281          * Check to see if any process groups have become orphaned
 282          * as a result of our exiting, and if they have any stopped
 283          * jobs, send them a SIGHUP and then a SIGCONT.  (POSIX 3.2.2.2)
 284          *
 285          * Case i: Our father is in a different pgrp than we are
 286          * and we were the only connection outside, so our pgrp
 287          * is about to become orphaned.
 288          */
 289         if ((current->p_pptr->pgrp != current->pgrp) &&
 290             (current->p_pptr->session == current->session) &&
 291             will_become_orphaned_pgrp(current->pgrp, current) &&
 292             has_stopped_jobs(current->pgrp)) {
 293                 kill_pg(current->pgrp,SIGHUP,1);
 294                 kill_pg(current->pgrp,SIGCONT,1);
 295         }
 296
 297         /* Let father know we died */
 298         notify_parent(current, current->exit_signal);
 299
 300         /*
 301          * This loop does two things:
 302          *
 303          * A.  Make init inherit all the child processes
 304          * B.  Check to see if any process groups have become orphaned
 305          *      as a result of our exiting, and if they have any stopped
 306          *      jobs, send them a SIGHUP and then a SIGCONT.  (POSIX 3.2.2.2)
 307          */
 308
 309         write_lock_irq(&tasklist_lock);
 310         while (current->p_cptr != NULL) {
 311                 p = current->p_cptr;
 312                 current->p_cptr = p->p_osptr;
 313                 p->p_ysptr = NULL;
 314                 p->flags &= ~(PF_PTRACED|PF_TRACESYS);
 315
 316                 p->p_pptr = p->p_opptr;
 317                 p->p_osptr = p->p_pptr->p_cptr;
 318                 if (p->p_osptr)
 319                         p->p_osptr->p_ysptr = p;
 320                 p->p_pptr->p_cptr = p;
 321                 if (p->state == TASK_ZOMBIE)
 322                         notify_parent(p, p->exit_signal);
 323                 /*
 324                  * process group orphan check
 325                  * Case ii: Our child is in a different pgrp
 326                  * than we are, and it was the only connection
 327                  * outside, so the child pgrp is now orphaned.
 328                  */
 329                 if ((p->pgrp != current->pgrp) &&
 330                     (p->session == current->session)) {
 331                         int pgrp = p->pgrp;
 332
 333                         write_unlock_irq(&tasklist_lock);
 334                         if (is_orphaned_pgrp(pgrp) && has_stopped_jobs(pgrp)) {
 335                                 kill_pg(pgrp,SIGHUP,1);
 336                                 kill_pg(pgrp,SIGCONT,1);
 337                         }
 338                         write_lock_irq(&tasklist_lock);
 339                 }
 340         }
 341         write_unlock_irq(&tasklist_lock);
 342
 343         if (current->leader)
 344                 disassociate_ctty(1);
 345 }
 346
 347 NORET_TYPE void do_exit(long code)
 348 {
 349         struct task_struct *tsk = current;
 350
 351         if (in_interrupt())
 352                 printk("Aiee, killing interrupt handler\n");
 353         if (!tsk->pid)
 354                 panic("Attempted to kill the idle task!");
 355         tsk->flags |= PF_EXITING;
 356         start_bh_atomic();
 357         del_timer(&tsk->real_timer);
 358         end_bh_atomic();
 359
 360         lock_kernel();
 361 fake_volatile:
 362 #ifdef CONFIG_BSD_PROCESS_ACCT
 363         acct_process(code);
 364 #endif
 365         sem_exit();
 366         __exit_mm(tsk);
 367 #if CONFIG_AP1000
 368         exit_msc(tsk);
 369 #endif
 370         __exit_files(tsk);
 371         __exit_fs(tsk);
 372         __exit_sighand(tsk);
 373         exit_thread();
 374         tsk->state = TASK_ZOMBIE;
 375         tsk->exit_code = code;
 376         exit_notify();
 377 #ifdef DEBUG_PROC_TREE
 378         audit_ptree();
 379 #endif
 380         if (tsk->exec_domain && tsk->exec_domain->module)
 381                 __MOD_DEC_USE_COUNT(tsk->exec_domain->module);
 382         if (tsk->binfmt && tsk->binfmt->module)
 383                 __MOD_DEC_USE_COUNT(tsk->binfmt->module);
 384         schedule();
 385 /*
 386  * In order to get rid of the "volatile function does return" message
 387  * I did this little loop that confuses gcc to think do_exit really
 388  * is volatile. In fact it's schedule() that is volatile in some
 389  * circumstances: when current->state = ZOMBIE, schedule() never
 390  * returns.
 391  *
 392  * In fact the natural way to do all this is to have the label and the
 393  * goto right after each other, but I put the fake_volatile label at
 394  * the start of the function just in case something /really/ bad
 395  * happens, and the schedule returns. This way we can try again. I'm
 396  * not paranoid: it's just that everybody is out to get me.
 397  */
 398         goto fake_volatile;
 399 }
 400
 401 asmlinkage int sys_exit(int error_code)
 402 {
 403         do_exit((error_code&0xff)<<8);
 404 }
 405
 406 asmlinkage int sys_wait4(pid_t pid,unsigned int * stat_addr, int options, struct rusage * ru)
 407 {
 408         int flag, retval;
 409         struct wait_queue wait = { current, NULL };
 410         struct task_struct *p;
 411
 412         if (options & ~(WNOHANG|WUNTRACED|__WCLONE))
 413                 return -EINVAL;
 414
 415         add_wait_queue(&current->wait_chldexit,&wait);
 416 repeat:
 417         flag = 0;
 418         read_lock(&tasklist_lock);
 419         for (p = current->p_cptr ; p ; p = p->p_osptr) {
 420                 if (pid>0) {
 421                         if (p->pid != pid)
 422                                 continue;
 423                 } else if (!pid) {
 424                         if (p->pgrp != current->pgrp)
 425                                 continue;
 426                 } else if (pid != -1) {
 427                         if (p->pgrp != -pid)
 428                                 continue;
 429                 }
 430                 /* wait for cloned processes iff the __WCLONE flag is set */
 431                 if ((p->exit_signal != SIGCHLD) ^ ((options & __WCLONE) != 0))
 432                         continue;
 433                 flag = 1;
 434                 switch (p->state) {
 435                         case TASK_STOPPED:
 436                                 if (!p->exit_code)
 437                                         continue;
 438                                 if (!(options & WUNTRACED) && !(p->flags & PF_PTRACED))
 439                                         continue;
 440                                 read_unlock(&tasklist_lock);
 441                                 retval = ru ? getrusage(p, RUSAGE_BOTH, ru) : 0;
 442                                 if (!retval && stat_addr)
 443                                         retval = put_user((p->exit_code << 8) | 0x7f, stat_addr);
 444                                 if (!retval) {
 445                                         p->exit_code = 0;
 446                                         retval = p->pid;
 447                                 }
 448                                 goto end_wait4;
 449                         case TASK_ZOMBIE:
 450                                 current->times.tms_cutime += p->times.tms_utime + p->times.tms_cutime;
 451                                 current->times.tms_cstime += p->times.tms_stime + p->times.tms_cstime;
 452                                 read_unlock(&tasklist_lock);
 453                                 retval = ru ? getrusage(p, RUSAGE_BOTH, ru) : 0;
 454                                 if (!retval && stat_addr)
 455                                         retval = put_user(p->exit_code, stat_addr);
 456                                 if (retval)
 457                                         goto end_wait4;
 458                                 retval = p->pid;
 459                                 if (p->p_opptr != p->p_pptr) {
 460                                         write_lock_irq(&tasklist_lock);
 461                                         REMOVE_LINKS(p);
 462                                         p->p_pptr = p->p_opptr;
 463                                         SET_LINKS(p);
 464                                         write_unlock_irq(&tasklist_lock);
 465                                         notify_parent(p, SIGCHLD);
 466                                 } else
 467                                         release(p);
 468 #ifdef DEBUG_PROC_TREE
 469                                 audit_ptree();
 470 #endif
 471                                 goto end_wait4;
 472                         default:
 473                                 continue;
 474                 }
 475         }
 476         read_unlock(&tasklist_lock);
 477         if (flag) {
 478                 retval = 0;
 479                 if (options & WNOHANG)
 480                         goto end_wait4;
 481                 retval = -ERESTARTSYS;
 482                 if (signal_pending(current))
 483                         goto end_wait4;
 484                 current->state=TASK_INTERRUPTIBLE;
 485                 schedule();
 486                 goto repeat;
 487         }
 488         retval = -ECHILD;
 489 end_wait4:
 490         remove_wait_queue(&current->wait_chldexit,&wait);
 491         return retval;
 492 }
 493
 494 #ifndef __alpha__
 495
 496 /*
 497  * sys_waitpid() remains for compatibility. waitpid() should be
 498  * implemented by calling sys_wait4() from libc.a.
 499  */
 500 asmlinkage int sys_waitpid(pid_t pid,unsigned int * stat_addr, int options)
 501 {
 502         return sys_wait4(pid, stat_addr, options, NULL);
 503 }
 504
 505 #endif