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                 tsk->sig = NULL;
 243                 if (atomic_dec_and_test(&sig->count))
 244                         kfree(sig);
 245         }
 246
 247         flush_signals(tsk);
 248 }
 249
 250 void exit_sighand(struct task_struct *tsk)
 251 {
 252         __exit_sighand(tsk);
 253 }
 254
 255 static inline void __exit_mm(struct task_struct * tsk)
 256 {
 257         struct mm_struct * mm = tsk->mm;
 258
 259         /* Set us up to use the kernel mm state */
 260         if (mm != &init_mm) {
 261                 flush_cache_mm(mm);
 262                 flush_tlb_mm(mm);
 263                 destroy_context(mm);
 264                 tsk->mm = &init_mm;
 265                 tsk->swappable = 0;
 266                 SET_PAGE_DIR(tsk, swapper_pg_dir);
 267                 mmput(mm);
 268         }
 269 }
 270
 271 void exit_mm(struct task_struct *tsk)
 272 {
 273         __exit_mm(tsk);
 274 }
 275
 276 /*
 277  * Send signals to all our closest relatives so that they know
 278  * to properly mourn us..
 279  */
 280 static void exit_notify(void)
 281 {
 282         struct task_struct * p;
 283
 284         forget_original_parent(current);
 285         /*
 286          * 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          * Case i: Our father is in a different pgrp than we are
 291          * and we were the only connection outside, so our pgrp
 292          * is about to become orphaned.
 293          */
 294         if ((current->p_pptr->pgrp != current->pgrp) &&
 295             (current->p_pptr->session == current->session) &&
 296             will_become_orphaned_pgrp(current->pgrp, current) &&
 297             has_stopped_jobs(current->pgrp)) {
 298                 kill_pg(current->pgrp,SIGHUP,1);
 299                 kill_pg(current->pgrp,SIGCONT,1);
 300         }
 301         /* Let father know we died */
 302         notify_parent(current, current->exit_signal);
 303
 304         /*
 305          * This loop does two things:
 306          *
 307          * A.  Make init inherit all the child processes
 308          * B.  Check to see if any process groups have become orphaned
 309          *      as a result of our exiting, and if they have any stopped
 310          *      jobs, send them a SIGHUP and then a SIGCONT.  (POSIX 3.2.2.2)
 311          */
 312
 313         write_lock_irq(&tasklist_lock);
 314         while (current->p_cptr != NULL) {
 315                 p = current->p_cptr;
 316                 current->p_cptr = p->p_osptr;
 317                 p->p_ysptr = NULL;
 318                 p->flags &= ~(PF_PTRACED|PF_TRACESYS);
 319
 320                 p->p_pptr = p->p_opptr;
 321                 p->p_osptr = p->p_pptr->p_cptr;
 322                 if (p->p_osptr)
 323                         p->p_osptr->p_ysptr = p;
 324                 p->p_pptr->p_cptr = p;
 325                 if (p->state == TASK_ZOMBIE)
 326                         notify_parent(p, p->exit_signal);
 327                 /*
 328                  * process group orphan check
 329                  * Case ii: Our child is in a different pgrp
 330                  * than we are, and it was the only connection
 331                  * outside, so the child pgrp is now orphaned.
 332                  */
 333                 if ((p->pgrp != current->pgrp) &&
 334                     (p->session == current->session)) {
 335                         int pgrp = p->pgrp;
 336
 337                         write_unlock_irq(&tasklist_lock);
 338                         if (is_orphaned_pgrp(pgrp) && has_stopped_jobs(pgrp)) {
 339                                 kill_pg(pgrp,SIGHUP,1);
 340                                 kill_pg(pgrp,SIGCONT,1);
 341                         }
 342                         write_lock_irq(&tasklist_lock);
 343                 }
 344         }
 345         write_unlock_irq(&tasklist_lock);
 346
 347         if (current->leader)
 348                 disassociate_ctty(1);
 349 }
 350
 351 NORET_TYPE void do_exit(long code)
 352 {
 353         struct task_struct *tsk = current;
 354
 355         if (in_interrupt())
 356                 printk("Aiee, killing interrupt handler\n");
 357         if (!tsk->pid)
 358                 panic("Attempted to kill the idle task!");
 359         tsk->flags |= PF_EXITING;
 360         del_timer(&tsk->real_timer);
 361
 362         lock_kernel();
 363 fake_volatile:
 364 #ifdef CONFIG_BSD_PROCESS_ACCT
 365         acct_process(code);
 366 #endif
 367         sem_exit();
 368         __exit_mm(tsk);
 369 #if CONFIG_AP1000
 370         exit_msc(tsk);
 371 #endif
 372         __exit_files(tsk);
 373         __exit_fs(tsk);
 374         __exit_sighand(tsk);
 375         exit_thread();
 376         tsk->state = TASK_ZOMBIE;
 377         tsk->exit_code = code;
 378         exit_notify();
 379 #ifdef DEBUG_PROC_TREE
 380         audit_ptree();
 381 #endif
 382         if (tsk->exec_domain && tsk->exec_domain->module)
 383                 __MOD_DEC_USE_COUNT(tsk->exec_domain->module);
 384         if (tsk->binfmt && tsk->binfmt->module)
 385                 __MOD_DEC_USE_COUNT(tsk->binfmt->module);
 386         schedule();
 387 /*
 388  * In order to get rid of the "volatile function does return" message
 389  * I did this little loop that confuses gcc to think do_exit really
 390  * is volatile. In fact it's schedule() that is volatile in some
 391  * circumstances: when current->state = ZOMBIE, schedule() never
 392  * returns.
 393  *
 394  * In fact the natural way to do all this is to have the label and the
 395  * goto right after each other, but I put the fake_volatile label at
 396  * the start of the function just in case something /really/ bad
 397  * happens, and the schedule returns. This way we can try again. I'm
 398  * not paranoid: it's just that everybody is out to get me.
 399  */
 400         goto fake_volatile;
 401 }
 402
 403 asmlinkage int sys_exit(int error_code)
 404 {
 405         do_exit((error_code&0xff)<<8);
 406 }
 407
 408 asmlinkage int sys_wait4(pid_t pid,unsigned int * stat_addr, int options, struct rusage * ru)
 409 {
 410         int flag, retval;
 411         struct wait_queue wait = { current, NULL };
 412         struct task_struct *p;
 413
 414         if (stat_addr) {
 415                 if(verify_area(VERIFY_WRITE, stat_addr, sizeof(*stat_addr)))
 416                         return -EFAULT;
 417         }
 418         if (ru) {
 419                 if(verify_area(VERIFY_WRITE, ru, sizeof(*ru)))
 420                         return -EFAULT;
 421         }
 422
 423         if (options & ~(WNOHANG|WUNTRACED|__WCLONE))
 424                 return -EINVAL;
 425
 426         add_wait_queue(&current->wait_chldexit,&wait);
 427 repeat:
 428         flag = 0;
 429         read_lock(&tasklist_lock);
 430         for (p = current->p_cptr ; p ; p = p->p_osptr) {
 431                 if (pid>0) {
 432                         if (p->pid != pid)
 433                                 continue;
 434                 } else if (!pid) {
 435                         if (p->pgrp != current->pgrp)
 436                                 continue;
 437                 } else if (pid != -1) {
 438                         if (p->pgrp != -pid)
 439                                 continue;
 440                 }
 441                 /* wait for cloned processes iff the __WCLONE flag is set */
 442                 if ((p->exit_signal != SIGCHLD) ^ ((options & __WCLONE) != 0))
 443                         continue;
 444                 flag = 1;
 445                 switch (p->state) {
 446                         case TASK_STOPPED:
 447                                 if (!p->exit_code)
 448                                         continue;
 449                                 if (!(options & WUNTRACED) && !(p->flags & PF_PTRACED))
 450                                         continue;
 451                                 read_unlock(&tasklist_lock);
 452                                 if (ru != NULL)
 453                                         getrusage(p, RUSAGE_BOTH, ru);
 454                                 if (stat_addr)
 455                                         __put_user((p->exit_code << 8) | 0x7f, stat_addr);
 456                                 p->exit_code = 0;
 457                                 retval = p->pid;
 458                                 goto end_wait4;
 459                         case TASK_ZOMBIE:
 460                                 current->times.tms_cutime += p->times.tms_utime + p->times.tms_cutime;
 461                                 current->times.tms_cstime += p->times.tms_stime + p->times.tms_cstime;
 462                                 read_unlock(&tasklist_lock);
 463                                 if (ru != NULL)
 464                                         getrusage(p, RUSAGE_BOTH, ru);
 465                                 if (stat_addr)
 466                                         __put_user(p->exit_code, stat_addr);
 467                                 retval = p->pid;
 468                                 if (p->p_opptr != p->p_pptr) {
 469                                         write_lock_irq(&tasklist_lock);
 470                                         REMOVE_LINKS(p);
 471                                         p->p_pptr = p->p_opptr;
 472                                         SET_LINKS(p);
 473                                         write_unlock_irq(&tasklist_lock);
 474                                         notify_parent(p, SIGCHLD);
 475                                 } else
 476                                         release(p);
 477 #ifdef DEBUG_PROC_TREE
 478                                 audit_ptree();
 479 #endif
 480                                 goto end_wait4;
 481                         default:
 482                                 continue;
 483                 }
 484         }
 485         read_unlock(&tasklist_lock);
 486         if (flag) {
 487                 retval = 0;
 488                 if (options & WNOHANG)
 489                         goto end_wait4;
 490                 retval = -ERESTARTSYS;
 491                 if (signal_pending(current))
 492                         goto end_wait4;
 493                 current->state=TASK_INTERRUPTIBLE;
 494                 schedule();
 495                 goto repeat;
 496         }
 497         retval = -ECHILD;
 498 end_wait4:
 499         remove_wait_queue(&current->wait_chldexit,&wait);
 500         return retval;
 501 }
 502
 503 #ifndef __alpha__
 504
 505 /*
 506  * sys_waitpid() remains for compatibility. waitpid() should be
 507  * implemented by calling sys_wait4() from libc.a.
 508  */
 509 asmlinkage int sys_waitpid(pid_t pid,unsigned int * stat_addr, int options)
 510 {
 511         return sys_wait4(pid, stat_addr, options, NULL);
 512 }
 513
 514 #endif