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/wait.h>
   9 #include <linux/errno.h>
  10 #include <linux/signal.h>
  11 #include <linux/sched.h>
  12 #include <linux/kernel.h>
  13 #include <linux/resource.h>
  14 #include <linux/mm.h>
  15 #include <linux/tty.h>
  16 #include <linux/malloc.h>
  17 #include <linux/slab.h>
  18 #include <linux/interrupt.h>
  19 #include <linux/smp.h>
  20 #include <linux/smp_lock.h>
  21 #include <linux/module.h>
  22 #include <linux/slab.h>
  23 #ifdef CONFIG_BSD_PROCESS_ACCT
  24 #include <linux/acct.h>
  25 #endif
  26
  27 #include <asm/uaccess.h>
  28 #include <asm/pgtable.h>
  29 #include <asm/mmu_context.h>
  30
  31 extern void sem_exit (void);
  32
  33 int getrusage(struct task_struct *, int, struct rusage *);
  34
  35 static void release(struct task_struct * p)
  36 {
  37         if (p != current) {
  38 #ifdef __SMP__
  39                 /*
  40                  * Wait to make sure the process isn't active on any
  41                  * other CPU
  42                  */
  43                 for (;;)  {
  44                         int has_cpu;
  45                         spin_lock(&scheduler_lock);
  46                         has_cpu = p->has_cpu;
  47                         spin_unlock(&scheduler_lock);
  48                         if (!has_cpu)
  49                                 break;
  50                         do {
  51                                 barrier();
  52                         } while (p->has_cpu);
  53                 }
  54 #endif
  55                 free_uid(p);
  56                 nr_tasks--;
  57                 add_free_taskslot(p->tarray_ptr);
  58
  59                 write_lock_irq(&tasklist_lock);
  60                 unhash_pid(p);
  61                 REMOVE_LINKS(p);
  62                 write_unlock_irq(&tasklist_lock);
  63
  64                 release_thread(p);
  65                 current->cmin_flt += p->min_flt + p->cmin_flt;
  66                 current->cmaj_flt += p->maj_flt + p->cmaj_flt;
  67                 current->cnswap += p->nswap + p->cnswap;
  68                 free_task_struct(p);
  69         } else {
  70                 printk("task releasing itself\n");
  71         }
  72 }
  73
  74 /*
  75  * This checks not only the pgrp, but falls back on the pid if no
  76  * satisfactory pgrp is found. I dunno - gdb doesn't work correctly
  77  * without this...
  78  */
  79 int session_of_pgrp(int pgrp)
  80 {
  81         struct task_struct *p;
  82         int fallback;
  83
  84         fallback = -1;
  85         read_lock(&tasklist_lock);
  86         for_each_task(p) {
  87                 if (p->session <= 0)
  88                         continue;
  89                 if (p->pgrp == pgrp) {
  90                         fallback = p->session;
  91                         break;
  92                 }
  93                 if (p->pid == pgrp)
  94                         fallback = p->session;
  95         }
  96         read_unlock(&tasklist_lock);
  97         return fallback;
  98 }
  99
 100 /*
 101  * Determine if a process group is "orphaned", according to the POSIX
 102  * definition in 2.2.2.52.  Orphaned process groups are not to be affected
 103  * by terminal-generated stop signals.  Newly orphaned process groups are
 104  * to receive a SIGHUP and a SIGCONT.
 105  *
 106  * "I ask you, have you ever known what it is to be an orphan?"
 107  */
 108 static int will_become_orphaned_pgrp(int pgrp, struct task_struct * ignored_task)
 109 {
 110         struct task_struct *p;
 111
 112         read_lock(&tasklist_lock);
 113         for_each_task(p) {
 114                 if ((p == ignored_task) || (p->pgrp != pgrp) ||
 115                     (p->state == TASK_ZOMBIE) ||
 116                     (p->p_pptr->pid == 1))
 117                         continue;
 118                 if ((p->p_pptr->pgrp != pgrp) &&
 119                     (p->p_pptr->session == p->session)) {
 120                         read_unlock(&tasklist_lock);
 121                         return 0;
 122                 }
 123         }
 124         read_unlock(&tasklist_lock);
 125         return 1;       /* (sighing) "Often!" */
 126 }
 127
 128 int is_orphaned_pgrp(int pgrp)
 129 {
 130         return will_become_orphaned_pgrp(pgrp, 0);
 131 }
 132
 133 static inline int has_stopped_jobs(int pgrp)
 134 {
 135         int retval = 0;
 136         struct task_struct * p;
 137
 138         read_lock(&tasklist_lock);
 139         for_each_task(p) {
 140                 if (p->pgrp != pgrp)
 141                         continue;
 142                 if (p->state != TASK_STOPPED)
 143                         continue;
 144                 retval = 1;
 145                 break;
 146         }
 147         read_unlock(&tasklist_lock);
 148         return retval;
 149 }
 150
 151 static inline void forget_original_parent(struct task_struct * father)
 152 {
 153         struct task_struct * p;
 154
 155         read_lock(&tasklist_lock);
 156         for_each_task(p) {
 157                 if (p->p_opptr == father) {
 158                         p->exit_signal = SIGCHLD;
 159                         p->p_opptr = task[smp_num_cpus] ? : task[0]; /* init */
 160                         if (p->pdeath_signal) send_sig(p->pdeath_signal, p, 0);
 161                 }
 162         }
 163         read_unlock(&tasklist_lock);
 164 }
 165
 166 static inline void close_files(struct files_struct * files)
 167 {
 168         int i, j;
 169
 170         j = 0;
 171         for (;;) {
 172                 unsigned long set = files->open_fds.fds_bits[j];
 173                 i = j * __NFDBITS;
 174                 j++;
 175                 if (i >= files->max_fds)
 176                         break;
 177                 while (set) {
 178                         if (set & 1) {
 179                                 struct file * file = files->fd[i];
 180                                 if (file) {
 181                                         files->fd[i] = NULL;
 182                                         close_fp(file, files);
 183                                 }
 184                         }
 185                         i++;
 186                         set >>= 1;
 187                 }
 188         }
 189 }
 190
 191 extern kmem_cache_t *files_cachep;
 192
 193 static inline void __exit_files(struct task_struct *tsk)
 194 {
 195         struct files_struct * files = tsk->files;
 196
 197         if (files) {
 198                 tsk->files = NULL;
 199                 if (atomic_dec_and_test(&files->count)) {
 200                         close_files(files);
 201                         /*
 202                          * Free the fd array as appropriate ...
 203                          */
 204                         if (NR_OPEN * sizeof(struct file *) == PAGE_SIZE)
 205                                 free_page((unsigned long) files->fd);
 206                         else
 207                                 kfree(files->fd);
 208                         kmem_cache_free(files_cachep, files);
 209                 }
 210         }
 211 }
 212
 213 void exit_files(struct task_struct *tsk)
 214 {
 215         __exit_files(tsk);
 216 }
 217
 218 static inline void __exit_fs(struct task_struct *tsk)
 219 {
 220         struct fs_struct * fs = tsk->fs;
 221
 222         if (fs) {
 223                 tsk->fs = NULL;
 224                 if (atomic_dec_and_test(&fs->count)) {
 225                         dput(fs->root);
 226                         dput(fs->pwd);
 227                         kfree(fs);
 228                 }
 229         }
 230 }
 231
 232 void exit_fs(struct task_struct *tsk)
 233 {
 234         __exit_fs(tsk);
 235 }
 236
 237 static inline void __exit_sighand(struct task_struct *tsk)
 238 {
 239         struct signal_struct * sig = tsk->sig;
 240
 241         if (sig) {
 242                 unsigned long flags;
 243
 244                 spin_lock_irqsave(&tsk->sigmask_lock, flags);
 245                 tsk->sig = NULL;
 246                 spin_unlock_irqrestore(&tsk->sigmask_lock, flags);
 247                 if (atomic_dec_and_test(&sig->count))
 248                         kfree(sig);
 249         }
 250
 251         flush_signals(tsk);
 252 }
 253
 254 void exit_sighand(struct task_struct *tsk)
 255 {
 256         __exit_sighand(tsk);
 257 }
 258
 259 static inline void __exit_mm(struct task_struct * tsk)
 260 {
 261         struct mm_struct * mm = tsk->mm;
 262
 263         /* Set us up to use the kernel mm state */
 264         if (mm != &init_mm) {
 265                 flush_cache_mm(mm);
 266                 flush_tlb_mm(mm);
 267                 destroy_context(mm);
 268                 tsk->mm = &init_mm;
 269                 tsk->swappable = 0;
 270                 SET_PAGE_DIR(tsk, swapper_pg_dir);
 271                 mmput(mm);
 272         }
 273 }
 274
 275 void exit_mm(struct task_struct *tsk)
 276 {
 277         __exit_mm(tsk);
 278 }
 279
 280 /*
 281  * Send signals to all our closest relatives so that they know
 282  * to properly mourn us..
 283  */
 284 static void exit_notify(void)
 285 {
 286         struct task_struct * p;
 287
 288         forget_original_parent(current);
 289         /*
 290          * Check to see if any process groups have become orphaned
 291          * as a result of our exiting, and if they have any stopped
 292          * jobs, send them a SIGHUP and then a SIGCONT.  (POSIX 3.2.2.2)
 293          *
 294          * Case i: Our father is in a different pgrp than we are
 295          * and we were the only connection outside, so our pgrp
 296          * is about to become orphaned.
 297          */
 298         if ((current->p_pptr->pgrp != current->pgrp) &&
 299             (current->p_pptr->session == current->session) &&
 300             will_become_orphaned_pgrp(current->pgrp, current) &&
 301             has_stopped_jobs(current->pgrp)) {
 302                 kill_pg(current->pgrp,SIGHUP,1);
 303                 kill_pg(current->pgrp,SIGCONT,1);
 304         }
 305         /* Let father know we died */
 306         notify_parent(current, current->exit_signal);
 307
 308         /*
 309          * This loop does two things:
 310          *
 311          * A.  Make init inherit all the child processes
 312          * B.  Check to see if any process groups have become orphaned
 313          *      as a result of our exiting, and if they have any stopped
 314          *      jobs, send them a SIGHUP and then a SIGCONT.  (POSIX 3.2.2.2)
 315          */
 316
 317         write_lock_irq(&tasklist_lock);
 318         while (current->p_cptr != NULL) {
 319                 p = current->p_cptr;
 320                 current->p_cptr = p->p_osptr;
 321                 p->p_ysptr = NULL;
 322                 p->flags &= ~(PF_PTRACED|PF_TRACESYS);
 323
 324                 p->p_pptr = p->p_opptr;
 325                 p->p_osptr = p->p_pptr->p_cptr;
 326                 if (p->p_osptr)
 327                         p->p_osptr->p_ysptr = p;
 328                 p->p_pptr->p_cptr = p;
 329                 if (p->state == TASK_ZOMBIE)
 330                         notify_parent(p, p->exit_signal);
 331                 /*
 332                  * process group orphan check
 333                  * Case ii: Our child is in a different pgrp
 334                  * than we are, and it was the only connection
 335                  * outside, so the child pgrp is now orphaned.
 336                  */
 337                 if ((p->pgrp != current->pgrp) &&
 338                     (p->session == current->session)) {
 339                         int pgrp = p->pgrp;
 340
 341                         write_unlock_irq(&tasklist_lock);
 342                         if (is_orphaned_pgrp(pgrp) && has_stopped_jobs(pgrp)) {
 343                                 kill_pg(pgrp,SIGHUP,1);
 344                                 kill_pg(pgrp,SIGCONT,1);
 345                         }
 346                         write_lock_irq(&tasklist_lock);
 347                 }
 348         }
 349         write_unlock_irq(&tasklist_lock);
 350
 351         if (current->leader)
 352                 disassociate_ctty(1);
 353 }
 354
 355 NORET_TYPE void do_exit(long code)
 356 {
 357         struct task_struct *tsk = current;
 358
 359         if (in_interrupt())
 360                 printk("Aiee, killing interrupt handler\n");
 361         if (!tsk->pid)
 362                 panic("Attempted to kill the idle task!");
 363         tsk->flags |= PF_EXITING;
 364         del_timer(&tsk->real_timer);
 365
 366         lock_kernel();
 367 fake_volatile:
 368 #ifdef CONFIG_BSD_PROCESS_ACCT
 369         acct_process(code);
 370 #endif
 371         sem_exit();
 372         __exit_mm(tsk);
 373 #if CONFIG_AP1000
 374         exit_msc(tsk);
 375 #endif
 376         __exit_files(tsk);
 377         __exit_fs(tsk);
 378         __exit_sighand(tsk);
 379         exit_thread();
 380         tsk->state = TASK_ZOMBIE;
 381         tsk->exit_code = code;
 382         exit_notify();
 383 #ifdef DEBUG_PROC_TREE
 384         audit_ptree();
 385 #endif
 386         if (tsk->exec_domain && tsk->exec_domain->module)
 387                 __MOD_DEC_USE_COUNT(tsk->exec_domain->module);
 388         if (tsk->binfmt && tsk->binfmt->module)
 389                 __MOD_DEC_USE_COUNT(tsk->binfmt->module);
 390         schedule();
 391 /*
 392  * In order to get rid of the "volatile function does return" message
 393  * I did this little loop that confuses gcc to think do_exit really
 394  * is volatile. In fact it's schedule() that is volatile in some
 395  * circumstances: when current->state = ZOMBIE, schedule() never
 396  * returns.
 397  *
 398  * In fact the natural way to do all this is to have the label and the
 399  * goto right after each other, but I put the fake_volatile label at
 400  * the start of the function just in case something /really/ bad
 401  * happens, and the schedule returns. This way we can try again. I'm
 402  * not paranoid: it's just that everybody is out to get me.
 403  */
 404         goto fake_volatile;
 405 }
 406
 407 asmlinkage int sys_exit(int error_code)
 408 {
 409         do_exit((error_code&0xff)<<8);
 410 }
 411
 412 asmlinkage int sys_wait4(pid_t pid,unsigned int * stat_addr, int options, struct rusage * ru)
 413 {
 414         int flag, retval;
 415         struct wait_queue wait = { current, NULL };
 416         struct task_struct *p;
 417
 418         if (stat_addr) {
 419                 if(verify_area(VERIFY_WRITE, stat_addr, sizeof(*stat_addr)))
 420                         return -EFAULT;
 421         }
 422         if (ru) {
 423                 if(verify_area(VERIFY_WRITE, ru, sizeof(*ru)))
 424                         return -EFAULT;
 425         }
 426
 427         if (options & ~(WNOHANG|WUNTRACED|__WCLONE))
 428                 return -EINVAL;
 429
 430         add_wait_queue(&current->wait_chldexit,&wait);
 431 repeat:
 432         flag = 0;
 433         read_lock(&tasklist_lock);
 434         for (p = current->p_cptr ; p ; p = p->p_osptr) {
 435                 if (pid>0) {
 436                         if (p->pid != pid)
 437                                 continue;
 438                 } else if (!pid) {
 439                         if (p->pgrp != current->pgrp)
 440                                 continue;
 441                 } else if (pid != -1) {
 442                         if (p->pgrp != -pid)
 443                                 continue;
 444                 }
 445                 /* wait for cloned processes iff the __WCLONE flag is set */
 446                 if ((p->exit_signal != SIGCHLD) ^ ((options & __WCLONE) != 0))
 447                         continue;
 448                 flag = 1;
 449                 switch (p->state) {
 450                         case TASK_STOPPED:
 451                                 if (!p->exit_code)
 452                                         continue;
 453                                 if (!(options & WUNTRACED) && !(p->flags & PF_PTRACED))
 454                                         continue;
 455                                 read_unlock(&tasklist_lock);
 456                                 if (ru != NULL)
 457                                         getrusage(p, RUSAGE_BOTH, ru);
 458                                 if (stat_addr)
 459                                         __put_user((p->exit_code << 8) | 0x7f, stat_addr);
 460                                 p->exit_code = 0;
 461                                 retval = p->pid;
 462                                 goto end_wait4;
 463                         case TASK_ZOMBIE:
 464                                 current->times.tms_cutime += p->times.tms_utime + p->times.tms_cutime;
 465                                 current->times.tms_cstime += p->times.tms_stime + p->times.tms_cstime;
 466                                 read_unlock(&tasklist_lock);
 467                                 if (ru != NULL)
 468                                         getrusage(p, RUSAGE_BOTH, ru);
 469                                 if (stat_addr)
 470                                         __put_user(p->exit_code, stat_addr);
 471                                 retval = p->pid;
 472                                 if (p->p_opptr != p->p_pptr) {
 473                                         write_lock_irq(&tasklist_lock);
 474                                         REMOVE_LINKS(p);
 475                                         p->p_pptr = p->p_opptr;
 476                                         SET_LINKS(p);
 477                                         write_unlock_irq(&tasklist_lock);
 478                                         notify_parent(p, SIGCHLD);
 479                                 } else
 480                                         release(p);
 481 #ifdef DEBUG_PROC_TREE
 482                                 audit_ptree();
 483 #endif
 484                                 goto end_wait4;
 485                         default:
 486                                 continue;
 487                 }
 488         }
 489         read_unlock(&tasklist_lock);
 490         if (flag) {
 491                 retval = 0;
 492                 if (options & WNOHANG)
 493                         goto end_wait4;
 494                 retval = -ERESTARTSYS;
 495                 if (signal_pending(current))
 496                         goto end_wait4;
 497                 current->state=TASK_INTERRUPTIBLE;
 498                 schedule();
 499                 goto repeat;
 500         }
 501         retval = -ECHILD;
 502 end_wait4:
 503         remove_wait_queue(&current->wait_chldexit,&wait);
 504         return retval;
 505 }
 506
 507 #ifndef __alpha__
 508
 509 /*
 510  * sys_waitpid() remains for compatibility. waitpid() should be
 511  * implemented by calling sys_wait4() from libc.a.
 512  */
 513 asmlinkage int sys_waitpid(pid_t pid,unsigned int * stat_addr, int options)
 514 {
 515         return sys_wait4(pid, stat_addr, options, NULL);
 516 }
 517
 518 #endif