fs/exec.c

   1 /*
   2  *  linux/fs/exec.c
   3  *
   4  *  Copyright (C) 1991, 1992  Linus Torvalds
   5  */
   6
   7 /*
   8  * #!-checking implemented by tytso.
   9  */
  10 /*
  11  * Demand-loading implemented 01.12.91 - no need to read anything but
  12  * the header into memory. The inode of the executable is put into
  13  * "current->executable", and page faults do the actual loading. Clean.
  14  *
  15  * Once more I can proudly say that linux stood up to being changed: it
  16  * was less than 2 hours work to get demand-loading completely implemented.
  17  *
  18  * Demand loading changed July 1993 by Eric Youngdale.   Use mmap instead,
  19  * current->executable is only used by the procfs.  This allows a dispatch
  20  * table to check for several different types  of binary formats.  We keep
  21  * trying until we recognize the file or we run out of supported binary
  22  * formats.
  23  */
  24
  25 #include <linux/config.h>
  26 #include <linux/slab.h>
  27 #include <linux/file.h>
  28 #include <linux/mman.h>
  29 #include <linux/a.out.h>
  30 #include <linux/stat.h>
  31 #include <linux/fcntl.h>
  32 #include <linux/smp_lock.h>
  33 #include <linux/init.h>
  34 #include <linux/pagemap.h>
  35 #include <linux/highmem.h>
  36 #include <linux/spinlock.h>
  37 #include <linux/personality.h>
  38 #include <linux/binfmts.h>
  39 #include <linux/swap.h>
  40 #include <linux/utsname.h>
  41 #include <linux/module.h>
  42 #include <linux/namei.h>
  43 #include <linux/proc_fs.h>
  44 #include <linux/ptrace.h>
  45 #include <linux/mount.h>
  46 #include <linux/security.h>
  47 #include <linux/syscalls.h>
  48 #include <linux/rmap.h>
  49
  50 #include <asm/uaccess.h>
  51 #include <asm/mmu_context.h>
  52
  53 #ifdef CONFIG_KMOD
  54 #include <linux/kmod.h>
  55 #endif
  56
  57 int core_uses_pid;
  58 char core_pattern[65] = "core";
  59 /* The maximal length of core_pattern is also specified in sysctl.c */
  60
  61 static struct linux_binfmt *formats;
  62 static rwlock_t binfmt_lock = RW_LOCK_UNLOCKED;
  63
  64 int register_binfmt(struct linux_binfmt * fmt)
  65 {
  66         struct linux_binfmt ** tmp = &formats;
  67
  68         if (!fmt)
  69                 return -EINVAL;
  70         if (fmt->next)
  71                 return -EBUSY;
  72         write_lock(&binfmt_lock);
  73         while (*tmp) {
  74                 if (fmt == *tmp) {
  75                         write_unlock(&binfmt_lock);
  76                         return -EBUSY;
  77                 }
  78                 tmp = &(*tmp)->next;
  79         }
  80         fmt->next = formats;
  81         formats = fmt;
  82         write_unlock(&binfmt_lock);
  83         return 0;
  84 }
  85
  86 EXPORT_SYMBOL(register_binfmt);
  87
  88 int unregister_binfmt(struct linux_binfmt * fmt)
  89 {
  90         struct linux_binfmt ** tmp = &formats;
  91
  92         write_lock(&binfmt_lock);
  93         while (*tmp) {
  94                 if (fmt == *tmp) {
  95                         *tmp = fmt->next;
  96                         write_unlock(&binfmt_lock);
  97                         return 0;
  98                 }
  99                 tmp = &(*tmp)->next;
 100         }
 101         write_unlock(&binfmt_lock);
 102         return -EINVAL;
 103 }
 104
 105 EXPORT_SYMBOL(unregister_binfmt);
 106
 107 static inline void put_binfmt(struct linux_binfmt * fmt)
 108 {
 109         module_put(fmt->module);
 110 }
 111
 112 /*
 113  * Note that a shared library must be both readable and executable due to
 114  * security reasons.
 115  *
 116  * Also note that we take the address to load from from the file itself.
 117  */
 118 asmlinkage long sys_uselib(const char __user * library)
 119 {
 120         struct file * file;
 121         struct nameidata nd;
 122         int error;
 123
 124         nd.intent.open.flags = FMODE_READ;
 125         error = __user_walk(library, LOOKUP_FOLLOW|LOOKUP_OPEN, &nd);
 126         if (error)
 127                 goto out;
 128
 129         error = -EINVAL;
 130         if (!S_ISREG(nd.dentry->d_inode->i_mode))
 131                 goto exit;
 132
 133         error = permission(nd.dentry->d_inode, MAY_READ | MAY_EXEC, &nd);
 134         if (error)
 135                 goto exit;
 136
 137         file = dentry_open(nd.dentry, nd.mnt, O_RDONLY);
 138         error = PTR_ERR(file);
 139         if (IS_ERR(file))
 140                 goto out;
 141
 142         error = -ENOEXEC;
 143         if(file->f_op) {
 144                 struct linux_binfmt * fmt;
 145
 146                 read_lock(&binfmt_lock);
 147                 for (fmt = formats ; fmt ; fmt = fmt->next) {
 148                         if (!fmt->load_shlib)
 149                                 continue;
 150                         if (!try_module_get(fmt->module))
 151                                 continue;
 152                         read_unlock(&binfmt_lock);
 153                         error = fmt->load_shlib(file);
 154                         read_lock(&binfmt_lock);
 155                         put_binfmt(fmt);
 156                         if (error != -ENOEXEC)
 157                                 break;
 158                 }
 159                 read_unlock(&binfmt_lock);
 160         }
 161         fput(file);
 162 out:
 163         return error;
 164 exit:
 165         path_release(&nd);
 166         goto out;
 167 }
 168
 169 /*
 170  * count() counts the number of strings in array ARGV.
 171  */
 172 static int count(char __user * __user * argv, int max)
 173 {
 174         int i = 0;
 175
 176         if (argv != NULL) {
 177                 for (;;) {
 178                         char __user * p;
 179
 180                         if (get_user(p, argv))
 181                                 return -EFAULT;
 182                         if (!p)
 183                                 break;
 184                         argv++;
 185                         if(++i > max)
 186                                 return -E2BIG;
 187                 }
 188         }
 189         return i;
 190 }
 191
 192 /*
 193  * 'copy_strings()' copies argument/environment strings from user
 194  * memory to free pages in kernel mem. These are in a format ready
 195  * to be put directly into the top of new user memory.
 196  */
 197 int copy_strings(int argc,char __user * __user * argv, struct linux_binprm *bprm)
 198 {
 199         struct page *kmapped_page = NULL;
 200         char *kaddr = NULL;
 201         int ret;
 202
 203         while (argc-- > 0) {
 204                 char __user *str;
 205                 int len;
 206                 unsigned long pos;
 207
 208                 if (get_user(str, argv+argc) ||
 209                                 !(len = strnlen_user(str, bprm->p))) {
 210                         ret = -EFAULT;
 211                         goto out;
 212                 }
 213
 214                 if (bprm->p < len)  {
 215                         ret = -E2BIG;
 216                         goto out;
 217                 }
 218
 219                 bprm->p -= len;
 220                 /* XXX: add architecture specific overflow check here. */
 221                 pos = bprm->p;
 222
 223                 while (len > 0) {
 224                         int i, new, err;
 225                         int offset, bytes_to_copy;
 226                         struct page *page;
 227
 228                         offset = pos % PAGE_SIZE;
 229                         i = pos/PAGE_SIZE;
 230                         page = bprm->page[i];
 231                         new = 0;
 232                         if (!page) {
 233                                 page = alloc_page(GFP_HIGHUSER);
 234                                 bprm->page[i] = page;
 235                                 if (!page) {
 236                                         ret = -ENOMEM;
 237                                         goto out;
 238                                 }
 239                                 new = 1;
 240                         }
 241
 242                         if (page != kmapped_page) {
 243                                 if (kmapped_page)
 244                                         kunmap(kmapped_page);
 245                                 kmapped_page = page;
 246                                 kaddr = kmap(kmapped_page);
 247                         }
 248                         if (new && offset)
 249                                 memset(kaddr, 0, offset);
 250                         bytes_to_copy = PAGE_SIZE - offset;
 251                         if (bytes_to_copy > len) {
 252                                 bytes_to_copy = len;
 253                                 if (new)
 254                                         memset(kaddr+offset+len, 0,
 255                                                 PAGE_SIZE-offset-len);
 256                         }
 257                         err = copy_from_user(kaddr+offset, str, bytes_to_copy);
 258                         if (err) {
 259                                 ret = -EFAULT;
 260                                 goto out;
 261                         }
 262
 263                         pos += bytes_to_copy;
 264                         str += bytes_to_copy;
 265                         len -= bytes_to_copy;
 266                 }
 267         }
 268         ret = 0;
 269 out:
 270         if (kmapped_page)
 271                 kunmap(kmapped_page);
 272         return ret;
 273 }
 274
 275 /*
 276  * Like copy_strings, but get argv and its values from kernel memory.
 277  */
 278 int copy_strings_kernel(int argc,char ** argv, struct linux_binprm *bprm)
 279 {
 280         int r;
 281         mm_segment_t oldfs = get_fs();
 282         set_fs(KERNEL_DS);
 283         r = copy_strings(argc, (char __user * __user *)argv, bprm);
 284         set_fs(oldfs);
 285         return r;
 286 }
 287
 288 EXPORT_SYMBOL(copy_strings_kernel);
 289
 290 #ifdef CONFIG_MMU
 291 /*
 292  * This routine is used to map in a page into an address space: needed by
 293  * execve() for the initial stack and environment pages.
 294  *
 295  * vma->vm_mm->mmap_sem is held for writing.
 296  */
 297 void install_arg_page(struct vm_area_struct *vma,
 298                         struct page *page, unsigned long address)
 299 {
 300         struct mm_struct *mm = vma->vm_mm;
 301         pgd_t * pgd;
 302         pmd_t * pmd;
 303         pte_t * pte;
 304
 305         if (unlikely(anon_vma_prepare(vma)))
 306                 goto out_sig;
 307
 308         flush_dcache_page(page);
 309         pgd = pgd_offset(mm, address);
 310
 311         spin_lock(&mm->page_table_lock);
 312         pmd = pmd_alloc(mm, pgd, address);
 313         if (!pmd)
 314                 goto out;
 315         pte = pte_alloc_map(mm, pmd, address);
 316         if (!pte)
 317                 goto out;
 318         if (!pte_none(*pte)) {
 319                 pte_unmap(pte);
 320                 goto out;
 321         }
 322         mm->rss++;
 323         lru_cache_add_active(page);
 324         set_pte(pte, pte_mkdirty(pte_mkwrite(mk_pte(
 325                                         page, vma->vm_page_prot))));
 326         page_add_anon_rmap(page, vma, address);
 327         pte_unmap(pte);
 328         spin_unlock(&mm->page_table_lock);
 329
 330         /* no need for flush_tlb */
 331         return;
 332 out:
 333         spin_unlock(&mm->page_table_lock);
 334 out_sig:
 335         __free_page(page);
 336         force_sig(SIGKILL, current);
 337 }
 338
 339 int setup_arg_pages(struct linux_binprm *bprm, int executable_stack)
 340 {
 341         unsigned long stack_base;
 342         struct vm_area_struct *mpnt;
 343         struct mm_struct *mm = current->mm;
 344         int i;
 345         long arg_size;
 346
 347 #ifdef CONFIG_STACK_GROWSUP
 348         /* Move the argument and environment strings to the bottom of the
 349          * stack space.
 350          */
 351         int offset, j;
 352         char *to, *from;
 353
 354         /* Start by shifting all the pages down */
 355         i = 0;
 356         for (j = 0; j < MAX_ARG_PAGES; j++) {
 357                 struct page *page = bprm->page[j];
 358                 if (!page)
 359                         continue;
 360                 bprm->page[i++] = page;
 361         }
 362
 363         /* Now move them within their pages */
 364         offset = bprm->p % PAGE_SIZE;
 365         to = kmap(bprm->page[0]);
 366         for (j = 1; j < i; j++) {
 367                 memmove(to, to + offset, PAGE_SIZE - offset);
 368                 from = kmap(bprm->page[j]);
 369                 memcpy(to + PAGE_SIZE - offset, from, offset);
 370                 kunmap(bprm->page[j - 1]);
 371                 to = from;
 372         }
 373         memmove(to, to + offset, PAGE_SIZE - offset);
 374         kunmap(bprm->page[j - 1]);
 375
 376         /* Adjust bprm->p to point to the end of the strings. */
 377         bprm->p = PAGE_SIZE * i - offset;
 378
 379         /* Limit stack size to 1GB */
 380         stack_base = current->rlim[RLIMIT_STACK].rlim_max;
 381         if (stack_base > (1 << 30))
 382                 stack_base = 1 << 30;
 383         stack_base = PAGE_ALIGN(STACK_TOP - stack_base);
 384
 385         mm->arg_start = stack_base;
 386         arg_size = i << PAGE_SHIFT;
 387
 388         /* zero pages that were copied above */
 389         while (i < MAX_ARG_PAGES)
 390                 bprm->page[i++] = NULL;
 391 #else
 392         stack_base = STACK_TOP - MAX_ARG_PAGES * PAGE_SIZE;
 393         mm->arg_start = bprm->p + stack_base;
 394         arg_size = STACK_TOP - (PAGE_MASK & (unsigned long) mm->arg_start);
 395 #endif
 396
 397         bprm->p += stack_base;
 398         if (bprm->loader)
 399                 bprm->loader += stack_base;
 400         bprm->exec += stack_base;
 401
 402         mpnt = kmem_cache_alloc(vm_area_cachep, SLAB_KERNEL);
 403         if (!mpnt)
 404                 return -ENOMEM;
 405
 406         if (security_vm_enough_memory(arg_size >> PAGE_SHIFT)) {
 407                 kmem_cache_free(vm_area_cachep, mpnt);
 408                 return -ENOMEM;
 409         }
 410
 411         memset(mpnt, 0, sizeof(*mpnt));
 412
 413         down_write(&mm->mmap_sem);
 414         {
 415                 mpnt->vm_mm = mm;
 416 #ifdef CONFIG_STACK_GROWSUP
 417                 mpnt->vm_start = stack_base;
 418                 mpnt->vm_end = PAGE_MASK &
 419                         (PAGE_SIZE - 1 + (unsigned long) bprm->p);
 420 #else
 421                 mpnt->vm_start = PAGE_MASK & (unsigned long) bprm->p;
 422                 mpnt->vm_end = STACK_TOP;
 423 #endif
 424                 /* Adjust stack execute permissions; explicitly enable
 425                  * for EXSTACK_ENABLE_X, disable for EXSTACK_DISABLE_X
 426                  * and leave alone (arch default) otherwise. */
 427                 if (unlikely(executable_stack == EXSTACK_ENABLE_X))
 428                         mpnt->vm_flags = VM_STACK_FLAGS |  VM_EXEC;
 429                 else if (executable_stack == EXSTACK_DISABLE_X)
 430                         mpnt->vm_flags = VM_STACK_FLAGS & ~VM_EXEC;
 431                 else
 432                         mpnt->vm_flags = VM_STACK_FLAGS;
 433                 mpnt->vm_flags |= mm->def_flags;
 434                 mpnt->vm_page_prot = protection_map[mpnt->vm_flags & 0x7];
 435                 insert_vm_struct(mm, mpnt);
 436                 mm->stack_vm = mm->total_vm = vma_pages(mpnt);
 437         }
 438
 439         for (i = 0 ; i < MAX_ARG_PAGES ; i++) {
 440                 struct page *page = bprm->page[i];
 441                 if (page) {
 442                         bprm->page[i] = NULL;
 443                         install_arg_page(mpnt, page, stack_base);
 444                 }
 445                 stack_base += PAGE_SIZE;
 446         }
 447         up_write(&mm->mmap_sem);
 448
 449         return 0;
 450 }
 451
 452 EXPORT_SYMBOL(setup_arg_pages);
 453
 454 #define free_arg_pages(bprm) do { } while (0)
 455
 456 #else
 457
 458 static inline void free_arg_pages(struct linux_binprm *bprm)
 459 {
 460         int i;
 461
 462         for (i = 0; i < MAX_ARG_PAGES; i++) {
 463                 if (bprm->page[i])
 464                         __free_page(bprm->page[i]);
 465                 bprm->page[i] = NULL;
 466         }
 467 }
 468
 469 #endif /* CONFIG_MMU */
 470
 471 struct file *open_exec(const char *name)
 472 {
 473         struct nameidata nd;
 474         int err;
 475         struct file *file;
 476
 477         nd.intent.open.flags = FMODE_READ;
 478         err = path_lookup(name, LOOKUP_FOLLOW|LOOKUP_OPEN, &nd);
 479         file = ERR_PTR(err);
 480
 481         if (!err) {
 482                 struct inode *inode = nd.dentry->d_inode;
 483                 file = ERR_PTR(-EACCES);
 484                 if (!(nd.mnt->mnt_flags & MNT_NOEXEC) &&
 485                     S_ISREG(inode->i_mode)) {
 486                         int err = permission(inode, MAY_EXEC, &nd);
 487                         if (!err && !(inode->i_mode & 0111))
 488                                 err = -EACCES;
 489                         file = ERR_PTR(err);
 490                         if (!err) {
 491                                 file = dentry_open(nd.dentry, nd.mnt, O_RDONLY);
 492                                 if (!IS_ERR(file)) {
 493                                         err = deny_write_access(file);
 494                                         if (err) {
 495                                                 fput(file);
 496                                                 file = ERR_PTR(err);
 497                                         }
 498                                 }
 499 out:
 500                                 return file;
 501                         }
 502                 }
 503                 path_release(&nd);
 504         }
 505         goto out;
 506 }
 507
 508 EXPORT_SYMBOL(open_exec);
 509
 510 int kernel_read(struct file *file, unsigned long offset,
 511         char *addr, unsigned long count)
 512 {
 513         mm_segment_t old_fs;
 514         loff_t pos = offset;
 515         int result;
 516
 517         old_fs = get_fs();
 518         set_fs(get_ds());
 519         /* The cast to a user pointer is valid due to the set_fs() */
 520         result = vfs_read(file, (void __user *)addr, count, &pos);
 521         set_fs(old_fs);
 522         return result;
 523 }
 524
 525 EXPORT_SYMBOL(kernel_read);
 526
 527 static int exec_mmap(struct mm_struct *mm)
 528 {
 529         struct task_struct *tsk;
 530         struct mm_struct * old_mm, *active_mm;
 531
 532         /* Add it to the list of mm's */
 533         spin_lock(&mmlist_lock);
 534         list_add(&mm->mmlist, &init_mm.mmlist);
 535         mmlist_nr++;
 536         spin_unlock(&mmlist_lock);
 537
 538         /* Notify parent that we're no longer interested in the old VM */
 539         tsk = current;
 540         old_mm = current->mm;
 541         mm_release(tsk, old_mm);
 542
 543         task_lock(tsk);
 544         active_mm = tsk->active_mm;
 545         tsk->mm = mm;
 546         tsk->active_mm = mm;
 547         activate_mm(active_mm, mm);
 548         task_unlock(tsk);
 549         arch_pick_mmap_layout(mm);
 550         if (old_mm) {
 551                 if (active_mm != old_mm) BUG();
 552                 mmput(old_mm);
 553                 return 0;
 554         }
 555         mmdrop(active_mm);
 556         return 0;
 557 }
 558
 559 /*
 560  * This function makes sure the current process has its own signal table,
 561  * so that flush_signal_handlers can later reset the handlers without
 562  * disturbing other processes.  (Other processes might share the signal
 563  * table via the CLONE_SIGHAND option to clone().)
 564  */
 565 static inline int de_thread(struct task_struct *tsk)
 566 {
 567         struct signal_struct *newsig, *oldsig = tsk->signal;
 568         struct sighand_struct *newsighand, *oldsighand = tsk->sighand;
 569         spinlock_t *lock = &oldsighand->siglock;
 570         int count;
 571
 572         /*
 573          * If we don't share sighandlers, then we aren't sharing anything
 574          * and we can just re-use it all.
 575          */
 576         if (atomic_read(&oldsighand->count) <= 1)
 577                 return 0;
 578
 579         newsighand = kmem_cache_alloc(sighand_cachep, GFP_KERNEL);
 580         if (!newsighand)
 581                 return -ENOMEM;
 582
 583         spin_lock_init(&newsighand->siglock);
 584         atomic_set(&newsighand->count, 1);
 585         memcpy(newsighand->action, oldsighand->action, sizeof(newsighand->action));
 586
 587         /*
 588          * See if we need to allocate a new signal structure
 589          */
 590         newsig = NULL;
 591         if (atomic_read(&oldsig->count) > 1) {
 592                 newsig = kmem_cache_alloc(signal_cachep, GFP_KERNEL);
 593                 if (!newsig) {
 594                         kmem_cache_free(sighand_cachep, newsighand);
 595                         return -ENOMEM;
 596                 }
 597                 atomic_set(&newsig->count, 1);
 598                 newsig->group_exit = 0;
 599                 newsig->group_exit_code = 0;
 600                 newsig->group_exit_task = NULL;
 601                 newsig->group_stop_count = 0;
 602                 newsig->curr_target = NULL;
 603                 init_sigpending(&newsig->shared_pending);
 604                 INIT_LIST_HEAD(&newsig->posix_timers);
 605
 606                 newsig->tty = oldsig->tty;
 607                 newsig->pgrp = oldsig->pgrp;
 608                 newsig->session = oldsig->session;
 609                 newsig->leader = oldsig->leader;
 610                 newsig->tty_old_pgrp = oldsig->tty_old_pgrp;
 611         }
 612
 613         if (thread_group_empty(current))
 614                 goto no_thread_group;
 615
 616         /*
 617          * Kill all other threads in the thread group.
 618          * We must hold tasklist_lock to call zap_other_threads.
 619          */
 620         read_lock(&tasklist_lock);
 621         spin_lock_irq(lock);
 622         if (oldsig->group_exit) {
 623                 /*
 624                  * Another group action in progress, just
 625                  * return so that the signal is processed.
 626                  */
 627                 spin_unlock_irq(lock);
 628                 read_unlock(&tasklist_lock);
 629                 kmem_cache_free(sighand_cachep, newsighand);
 630                 if (newsig)
 631                         kmem_cache_free(signal_cachep, newsig);
 632                 return -EAGAIN;
 633         }
 634         oldsig->group_exit = 1;
 635         zap_other_threads(current);
 636         read_unlock(&tasklist_lock);
 637
 638         /*
 639          * Account for the thread group leader hanging around:
 640          */
 641         count = 2;
 642         if (current->pid == current->tgid)
 643                 count = 1;
 644         while (atomic_read(&oldsig->count) > count) {
 645                 oldsig->group_exit_task = current;
 646                 oldsig->notify_count = count;
 647                 __set_current_state(TASK_UNINTERRUPTIBLE);
 648                 spin_unlock_irq(lock);
 649                 schedule();
 650                 spin_lock_irq(lock);
 651         }
 652         spin_unlock_irq(lock);
 653
 654         /*
 655          * At this point all other threads have exited, all we have to
 656          * do is to wait for the thread group leader to become inactive,
 657          * and to assume its PID:
 658          */
 659         if (current->pid != current->tgid) {
 660                 struct task_struct *leader = current->group_leader, *parent;
 661                 struct dentry *proc_dentry1, *proc_dentry2;
 662                 unsigned long state, ptrace;
 663
 664                 /*
 665                  * Wait for the thread group leader to be a zombie.
 666                  * It should already be zombie at this point, most
 667                  * of the time.
 668                  */
 669                 while (leader->state != TASK_ZOMBIE)
 670                         yield();
 671
 672                 spin_lock(&leader->proc_lock);
 673                 spin_lock(&current->proc_lock);
 674                 proc_dentry1 = proc_pid_unhash(current);
 675                 proc_dentry2 = proc_pid_unhash(leader);
 676                 write_lock_irq(&tasklist_lock);
 677
 678                 if (leader->tgid != current->tgid)
 679                         BUG();
 680                 if (current->pid == current->tgid)
 681                         BUG();
 682                 /*
 683                  * An exec() starts a new thread group with the
 684                  * TGID of the previous thread group. Rehash the
 685                  * two threads with a switched PID, and release
 686                  * the former thread group leader:
 687                  */
 688                 ptrace = leader->ptrace;
 689                 parent = leader->parent;
 690
 691                 ptrace_unlink(current);
 692                 ptrace_unlink(leader);
 693                 remove_parent(current);
 694                 remove_parent(leader);
 695
 696                 switch_exec_pids(leader, current);
 697
 698                 current->parent = current->real_parent = leader->real_parent;
 699                 leader->parent = leader->real_parent = child_reaper;
 700                 current->group_leader = current;
 701                 leader->group_leader = leader;
 702
 703                 add_parent(current, current->parent);
 704                 add_parent(leader, leader->parent);
 705                 if (ptrace) {
 706                         current->ptrace = ptrace;
 707                         __ptrace_link(current, parent);
 708                 }
 709
 710                 list_del(&current->tasks);
 711                 list_add_tail(&current->tasks, &init_task.tasks);
 712                 current->exit_signal = SIGCHLD;
 713                 state = leader->state;
 714
 715                 write_unlock_irq(&tasklist_lock);
 716                 spin_unlock(&leader->proc_lock);
 717                 spin_unlock(&current->proc_lock);
 718                 proc_pid_flush(proc_dentry1);
 719                 proc_pid_flush(proc_dentry2);
 720
 721                 if (state != TASK_ZOMBIE)
 722                         BUG();
 723                 release_task(leader);
 724         }
 725
 726 no_thread_group:
 727
 728         write_lock_irq(&tasklist_lock);
 729         spin_lock(&oldsighand->siglock);
 730         spin_lock(&newsighand->siglock);
 731
 732         if (current == oldsig->curr_target)
 733                 oldsig->curr_target = next_thread(current);
 734         if (newsig)
 735                 current->signal = newsig;
 736         current->sighand = newsighand;
 737         init_sigpending(&current->pending);
 738         recalc_sigpending();
 739
 740         spin_unlock(&newsighand->siglock);
 741         spin_unlock(&oldsighand->siglock);
 742         write_unlock_irq(&tasklist_lock);
 743
 744         if (newsig && atomic_dec_and_test(&oldsig->count)) {
 745                 exit_itimers(oldsig);
 746                 kmem_cache_free(signal_cachep, oldsig);
 747         }
 748
 749         if (atomic_dec_and_test(&oldsighand->count))
 750                 kmem_cache_free(sighand_cachep, oldsighand);
 751
 752         if (!thread_group_empty(current))
 753                 BUG();
 754         if (current->tgid != current->pid)
 755                 BUG();
 756         return 0;
 757 }
 758
 759 /*
 760  * These functions flushes out all traces of the currently running executable
 761  * so that a new one can be started
 762  */
 763
 764 static inline void flush_old_files(struct files_struct * files)
 765 {
 766         long j = -1;
 767
 768         spin_lock(&files->file_lock);
 769         for (;;) {
 770                 unsigned long set, i;
 771
 772                 j++;
 773                 i = j * __NFDBITS;
 774                 if (i >= files->max_fds || i >= files->max_fdset)
 775                         break;
 776                 set = files->close_on_exec->fds_bits[j];
 777                 if (!set)
 778                         continue;
 779                 files->close_on_exec->fds_bits[j] = 0;
 780                 spin_unlock(&files->file_lock);
 781                 for ( ; set ; i++,set >>= 1) {
 782                         if (set & 1) {
 783                                 sys_close(i);
 784                         }
 785                 }
 786                 spin_lock(&files->file_lock);
 787
 788         }
 789         spin_unlock(&files->file_lock);
 790 }
 791
 792 void get_task_comm(char *buf, struct task_struct *tsk)
 793 {
 794         /* buf must be at least sizeof(tsk->comm) in size */
 795         task_lock(tsk);
 796         memcpy(buf, tsk->comm, sizeof(tsk->comm));
 797         task_unlock(tsk);
 798 }
 799
 800 void set_task_comm(struct task_struct *tsk, char *buf)
 801 {
 802         task_lock(tsk);
 803         strlcpy(tsk->comm, buf, sizeof(tsk->comm));
 804         task_unlock(tsk);
 805 }
 806
 807 int flush_old_exec(struct linux_binprm * bprm)
 808 {
 809         char * name;
 810         int i, ch, retval;
 811         struct files_struct *files;
 812         char tcomm[sizeof(current->comm)];
 813
 814         /*
 815          * Make sure we have a private signal table and that
 816          * we are unassociated from the previous thread group.
 817          */
 818         retval = de_thread(current);
 819         if (retval)
 820                 goto out;
 821
 822         /*
 823          * Make sure we have private file handles. Ask the
 824          * fork helper to do the work for us and the exit
 825          * helper to do the cleanup of the old one.
 826          */
 827         files = current->files;         /* refcounted so safe to hold */
 828         retval = unshare_files();
 829         if (retval)
 830                 goto out;
 831         /*
 832          * Release all of the old mmap stuff
 833          */
 834         retval = exec_mmap(bprm->mm);
 835         if (retval)
 836                 goto mmap_failed;
 837
 838         bprm->mm = NULL;                /* We're using it now */
 839
 840         /* This is the point of no return */
 841         steal_locks(files);
 842         put_files_struct(files);
 843
 844         current->sas_ss_sp = current->sas_ss_size = 0;
 845
 846         if (current->euid == current->uid && current->egid == current->gid)
 847                 current->mm->dumpable = 1;
 848         name = bprm->filename;
 849         for (i=0; (ch = *(name++)) != '\0';) {
 850                 if (ch == '/')
 851                         i = 0;
 852                 else
 853                         if (i < (sizeof(tcomm) - 1))
 854                                 tcomm[i++] = ch;
 855         }
 856         tcomm[i] = '\0';
 857         set_task_comm(current, tcomm);
 858
 859         flush_thread();
 860
 861         if (bprm->e_uid != current->euid || bprm->e_gid != current->egid ||
 862             permission(bprm->file->f_dentry->d_inode,MAY_READ, NULL) ||
 863             (bprm->interp_flags & BINPRM_FLAGS_ENFORCE_NONDUMP))
 864                 current->mm->dumpable = 0;
 865
 866         /* An exec changes our domain. We are no longer part of the thread
 867            group */
 868
 869         current->self_exec_id++;
 870
 871         flush_signal_handlers(current, 0);
 872         flush_old_files(current->files);
 873
 874         return 0;
 875
 876 mmap_failed:
 877         put_files_struct(current->files);
 878         current->files = files;
 879 out:
 880         return retval;
 881 }
 882
 883 EXPORT_SYMBOL(flush_old_exec);
 884
 885 /*
 886  * Fill the binprm structure from the inode.
 887  * Check permissions, then read the first 128 (BINPRM_BUF_SIZE) bytes
 888  */
 889 int prepare_binprm(struct linux_binprm *bprm)
 890 {
 891         int mode;
 892         struct inode * inode = bprm->file->f_dentry->d_inode;
 893         int retval;
 894
 895         mode = inode->i_mode;
 896         /*
 897          * Check execute perms again - if the caller has CAP_DAC_OVERRIDE,
 898          * vfs_permission lets a non-executable through
 899          */
 900         if (!(mode & 0111))     /* with at least _one_ execute bit set */
 901                 return -EACCES;
 902         if (bprm->file->f_op == NULL)
 903                 return -EACCES;
 904
 905         bprm->e_uid = current->euid;
 906         bprm->e_gid = current->egid;
 907
 908         if(!(bprm->file->f_vfsmnt->mnt_flags & MNT_NOSUID)) {
 909                 /* Set-uid? */
 910                 if (mode & S_ISUID) {
 911                         current->personality &= ~PER_CLEAR_ON_SETID;
 912                         bprm->e_uid = inode->i_uid;
 913                 }
 914
 915                 /* Set-gid? */
 916                 /*
 917                  * If setgid is set but no group execute bit then this
 918                  * is a candidate for mandatory locking, not a setgid
 919                  * executable.
 920                  */
 921                 if ((mode & (S_ISGID | S_IXGRP)) == (S_ISGID | S_IXGRP)) {
 922                         current->personality &= ~PER_CLEAR_ON_SETID;
 923                         bprm->e_gid = inode->i_gid;
 924                 }
 925         }
 926
 927         /* fill in binprm security blob */
 928         retval = security_bprm_set(bprm);
 929         if (retval)
 930                 return retval;
 931
 932         memset(bprm->buf,0,BINPRM_BUF_SIZE);
 933         return kernel_read(bprm->file,0,bprm->buf,BINPRM_BUF_SIZE);
 934 }
 935
 936 EXPORT_SYMBOL(prepare_binprm);
 937
 938 static inline int unsafe_exec(struct task_struct *p)
 939 {
 940         int unsafe = 0;
 941         if (p->ptrace & PT_PTRACED) {
 942                 if (p->ptrace & PT_PTRACE_CAP)
 943                         unsafe |= LSM_UNSAFE_PTRACE_CAP;
 944                 else
 945                         unsafe |= LSM_UNSAFE_PTRACE;
 946         }
 947         if (atomic_read(&p->fs->count) > 1 ||
 948             atomic_read(&p->files->count) > 1 ||
 949             atomic_read(&p->sighand->count) > 1)
 950                 unsafe |= LSM_UNSAFE_SHARE;
 951
 952         return unsafe;
 953 }
 954
 955 void compute_creds(struct linux_binprm *bprm)
 956 {
 957         int unsafe;
 958         task_lock(current);
 959         unsafe = unsafe_exec(current);
 960         security_bprm_apply_creds(bprm, unsafe);
 961         task_unlock(current);
 962 }
 963
 964 EXPORT_SYMBOL(compute_creds);
 965
 966 void remove_arg_zero(struct linux_binprm *bprm)
 967 {
 968         if (bprm->argc) {
 969                 unsigned long offset;
 970                 char * kaddr;
 971                 struct page *page;
 972
 973                 offset = bprm->p % PAGE_SIZE;
 974                 goto inside;
 975
 976                 while (bprm->p++, *(kaddr+offset++)) {
 977                         if (offset != PAGE_SIZE)
 978                                 continue;
 979                         offset = 0;
 980                         kunmap_atomic(kaddr, KM_USER0);
 981 inside:
 982                         page = bprm->page[bprm->p/PAGE_SIZE];
 983                         kaddr = kmap_atomic(page, KM_USER0);
 984                 }
 985                 kunmap_atomic(kaddr, KM_USER0);
 986                 bprm->argc--;
 987         }
 988 }
 989
 990 EXPORT_SYMBOL(remove_arg_zero);
 991
 992 /*
 993  * cycle the list of binary formats handler, until one recognizes the image
 994  */
 995 int search_binary_handler(struct linux_binprm *bprm,struct pt_regs *regs)
 996 {
 997         int try,retval=0;
 998         struct linux_binfmt *fmt;
 999 #ifdef __alpha__
1000         /* handle /sbin/loader.. */
1001         {
1002             struct exec * eh = (struct exec *) bprm->buf;
1003
1004             if (!bprm->loader && eh->fh.f_magic == 0x183 &&
1005                 (eh->fh.f_flags & 0x3000) == 0x3000)
1006             {
1007                 struct file * file;
1008                 unsigned long loader;
1009
1010                 allow_write_access(bprm->file);
1011                 fput(bprm->file);
1012                 bprm->file = NULL;
1013
1014                 loader = PAGE_SIZE*MAX_ARG_PAGES-sizeof(void *);
1015
1016                 file = open_exec("/sbin/loader");
1017                 retval = PTR_ERR(file);
1018                 if (IS_ERR(file))
1019                         return retval;
1020
1021                 /* Remember if the application is TASO.  */
1022                 bprm->sh_bang = eh->ah.entry < 0x100000000UL;
1023
1024                 bprm->file = file;
1025                 bprm->loader = loader;
1026                 retval = prepare_binprm(bprm);
1027                 if (retval<0)
1028                         return retval;
1029                 /* should call search_binary_handler recursively here,
1030                    but it does not matter */
1031             }
1032         }
1033 #endif
1034         retval = security_bprm_check(bprm);
1035         if (retval)
1036                 return retval;
1037
1038         /* kernel module loader fixup */
1039         /* so we don't try to load run modprobe in kernel space. */
1040         set_fs(USER_DS);
1041         for (try=0; try<2; try++) {
1042                 read_lock(&binfmt_lock);
1043                 for (fmt = formats ; fmt ; fmt = fmt->next) {
1044                         int (*fn)(struct linux_binprm *, struct pt_regs *) = fmt->load_binary;
1045                         if (!fn)
1046                                 continue;
1047                         if (!try_module_get(fmt->module))
1048                                 continue;
1049                         read_unlock(&binfmt_lock);
1050                         retval = fn(bprm, regs);
1051                         if (retval >= 0) {
1052                                 put_binfmt(fmt);
1053                                 allow_write_access(bprm->file);
1054                                 if (bprm->file)
1055                                         fput(bprm->file);
1056                                 bprm->file = NULL;
1057                                 current->did_exec = 1;
1058                                 return retval;
1059                         }
1060                         read_lock(&binfmt_lock);
1061                         put_binfmt(fmt);
1062                         if (retval != -ENOEXEC || bprm->mm == NULL)
1063                                 break;
1064                         if (!bprm->file) {
1065                                 read_unlock(&binfmt_lock);
1066                                 return retval;
1067                         }
1068                 }
1069                 read_unlock(&binfmt_lock);
1070                 if (retval != -ENOEXEC || bprm->mm == NULL) {
1071                         break;
1072 #ifdef CONFIG_KMOD
1073                 }else{
1074 #define printable(c) (((c)=='\t') || ((c)=='\n') || (0x20<=(c) && (c)<=0x7e))
1075                         if (printable(bprm->buf[0]) &&
1076                             printable(bprm->buf[1]) &&
1077                             printable(bprm->buf[2]) &&
1078                             printable(bprm->buf[3]))
1079                                 break; /* -ENOEXEC */
1080                         request_module("binfmt-%04x", *(unsigned short *)(&bprm->buf[2]));
1081 #endif
1082                 }
1083         }
1084         return retval;
1085 }
1086
1087 EXPORT_SYMBOL(search_binary_handler);
1088
1089 /*
1090  * sys_execve() executes a new program.
1091  */
1092 int do_execve(char * filename,
1093         char __user *__user *argv,
1094         char __user *__user *envp,
1095         struct pt_regs * regs)
1096 {
1097         struct linux_binprm *bprm;
1098         struct file *file;
1099         int retval;
1100         int i;
1101
1102         file = open_exec(filename);
1103
1104         retval = PTR_ERR(file);
1105         if (IS_ERR(file))
1106                 return retval;
1107
1108         sched_exec();
1109
1110         retval = -ENOMEM;
1111         bprm = kmalloc(sizeof(*bprm), GFP_KERNEL);
1112         if (!bprm)
1113                 goto out_ret;
1114         memset(bprm, 0, sizeof(*bprm));
1115
1116         bprm->p = PAGE_SIZE*MAX_ARG_PAGES-sizeof(void *);
1117
1118         bprm->file = file;
1119         bprm->filename = filename;
1120         bprm->interp = filename;
1121         bprm->mm = mm_alloc();
1122         if (!bprm->mm)
1123                 goto out_file;
1124
1125         retval = init_new_context(current, bprm->mm);
1126         if (retval < 0)
1127                 goto out_mm;
1128
1129         bprm->argc = count(argv, bprm->p / sizeof(void *));
1130         if ((retval = bprm->argc) < 0)
1131                 goto out_mm;
1132
1133         bprm->envc = count(envp, bprm->p / sizeof(void *));
1134         if ((retval = bprm->envc) < 0)
1135                 goto out_mm;
1136
1137         retval = security_bprm_alloc(bprm);
1138         if (retval)
1139                 goto out;
1140
1141         retval = prepare_binprm(bprm);
1142         if (retval < 0)
1143                 goto out;
1144
1145         retval = copy_strings_kernel(1, &bprm->filename, bprm);
1146         if (retval < 0)
1147                 goto out;
1148
1149         bprm->exec = bprm->p;
1150         retval = copy_strings(bprm->envc, envp, bprm);
1151         if (retval < 0)
1152                 goto out;
1153
1154         retval = copy_strings(bprm->argc, argv, bprm);
1155         if (retval < 0)
1156                 goto out;
1157
1158         retval = search_binary_handler(bprm,regs);
1159         if (retval >= 0) {
1160                 free_arg_pages(bprm);
1161
1162                 /* execve success */
1163                 security_bprm_free(bprm);
1164                 kfree(bprm);
1165                 return retval;
1166         }
1167
1168 out:
1169         /* Something went wrong, return the inode and free the argument pages*/
1170         for (i = 0 ; i < MAX_ARG_PAGES ; i++) {
1171                 struct page * page = bprm->page[i];
1172                 if (page)
1173                         __free_page(page);
1174         }
1175
1176         if (bprm->security)
1177                 security_bprm_free(bprm);
1178
1179 out_mm:
1180         if (bprm->mm)
1181                 mmdrop(bprm->mm);
1182
1183 out_file:
1184         if (bprm->file) {
1185                 allow_write_access(bprm->file);
1186                 fput(bprm->file);
1187         }
1188         kfree(bprm);
1189
1190 out_ret:
1191         return retval;
1192 }
1193
1194 EXPORT_SYMBOL(do_execve);
1195
1196 int set_binfmt(struct linux_binfmt *new)
1197 {
1198         struct linux_binfmt *old = current->binfmt;
1199
1200         if (new) {
1201                 if (!try_module_get(new->module))
1202                         return -1;
1203         }
1204         current->binfmt = new;
1205         if (old)
1206                 module_put(old->module);
1207         return 0;
1208 }
1209
1210 EXPORT_SYMBOL(set_binfmt);
1211
1212 #define CORENAME_MAX_SIZE 64
1213
1214 /* format_corename will inspect the pattern parameter, and output a
1215  * name into corename, which must have space for at least
1216  * CORENAME_MAX_SIZE bytes plus one byte for the zero terminator.
1217  */
1218 static void format_corename(char *corename, const char *pattern, long signr)
1219 {
1220         const char *pat_ptr = pattern;
1221         char *out_ptr = corename;
1222         char *const out_end = corename + CORENAME_MAX_SIZE;
1223         int rc;
1224         int pid_in_pattern = 0;
1225
1226         /* Repeat as long as we have more pattern to process and more output
1227            space */
1228         while (*pat_ptr) {
1229                 if (*pat_ptr != '%') {
1230                         if (out_ptr == out_end)
1231                                 goto out;
1232                         *out_ptr++ = *pat_ptr++;
1233                 } else {
1234                         switch (*++pat_ptr) {
1235                         case 0:
1236                                 goto out;
1237                         /* Double percent, output one percent */
1238                         case '%':
1239                                 if (out_ptr == out_end)
1240                                         goto out;
1241                                 *out_ptr++ = '%';
1242                                 break;
1243                         /* pid */
1244                         case 'p':
1245                                 pid_in_pattern = 1;
1246                                 rc = snprintf(out_ptr, out_end - out_ptr,
1247                                               "%d", current->tgid);
1248                                 if (rc > out_end - out_ptr)
1249                                         goto out;
1250                                 out_ptr += rc;
1251                                 break;
1252                         /* uid */
1253                         case 'u':
1254                                 rc = snprintf(out_ptr, out_end - out_ptr,
1255                                               "%d", current->uid);
1256                                 if (rc > out_end - out_ptr)
1257                                         goto out;
1258                                 out_ptr += rc;
1259                                 break;
1260                         /* gid */
1261                         case 'g':
1262                                 rc = snprintf(out_ptr, out_end - out_ptr,
1263                                               "%d", current->gid);
1264                                 if (rc > out_end - out_ptr)
1265                                         goto out;
1266                                 out_ptr += rc;
1267                                 break;
1268                         /* signal that caused the coredump */
1269                         case 's':
1270                                 rc = snprintf(out_ptr, out_end - out_ptr,
1271                                               "%ld", signr);
1272                                 if (rc > out_end - out_ptr)
1273                                         goto out;
1274                                 out_ptr += rc;
1275                                 break;
1276                         /* UNIX time of coredump */
1277                         case 't': {
1278                                 struct timeval tv;
1279                                 do_gettimeofday(&tv);
1280                                 rc = snprintf(out_ptr, out_end - out_ptr,
1281                                               "%lu", tv.tv_sec);
1282                                 if (rc > out_end - out_ptr)
1283                                         goto out;
1284                                 out_ptr += rc;
1285                                 break;
1286                         }
1287                         /* hostname */
1288                         case 'h':
1289                                 down_read(&uts_sem);
1290                                 rc = snprintf(out_ptr, out_end - out_ptr,
1291                                               "%s", system_utsname.nodename);
1292                                 up_read(&uts_sem);
1293                                 if (rc > out_end - out_ptr)
1294                                         goto out;
1295                                 out_ptr += rc;
1296                                 break;
1297                         /* executable */
1298                         case 'e':
1299                                 rc = snprintf(out_ptr, out_end - out_ptr,
1300                                               "%s", current->comm);
1301                                 if (rc > out_end - out_ptr)
1302                                         goto out;
1303                                 out_ptr += rc;
1304                                 break;
1305                         default:
1306                                 break;
1307                         }
1308                         ++pat_ptr;
1309                 }
1310         }
1311         /* Backward compatibility with core_uses_pid:
1312          *
1313          * If core_pattern does not include a %p (as is the default)
1314          * and core_uses_pid is set, then .%pid will be appended to
1315          * the filename */
1316         if (!pid_in_pattern
1317             && (core_uses_pid || atomic_read(&current->mm->mm_users) != 1)) {
1318                 rc = snprintf(out_ptr, out_end - out_ptr,
1319                               ".%d", current->tgid);
1320                 if (rc > out_end - out_ptr)
1321                         goto out;
1322                 out_ptr += rc;
1323         }
1324       out:
1325         *out_ptr = 0;
1326 }
1327
1328 static void zap_threads (struct mm_struct *mm)
1329 {
1330         struct task_struct *g, *p;
1331         struct task_struct *tsk = current;
1332         struct completion *vfork_done = tsk->vfork_done;
1333
1334         /*
1335          * Make sure nobody is waiting for us to release the VM,
1336          * otherwise we can deadlock when we wait on each other
1337          */
1338         if (vfork_done) {
1339                 tsk->vfork_done = NULL;
1340                 complete(vfork_done);
1341         }
1342
1343         read_lock(&tasklist_lock);
1344         do_each_thread(g,p)
1345                 if (mm == p->mm && p != tsk) {
1346                         force_sig_specific(SIGKILL, p);
1347                         mm->core_waiters++;
1348                 }
1349         while_each_thread(g,p);
1350
1351         read_unlock(&tasklist_lock);
1352 }
1353
1354 static void coredump_wait(struct mm_struct *mm)
1355 {
1356         DECLARE_COMPLETION(startup_done);
1357
1358         mm->core_waiters++; /* let other threads block */
1359         mm->core_startup_done = &startup_done;
1360
1361         /* give other threads a chance to run: */
1362         yield();
1363
1364         zap_threads(mm);
1365         if (--mm->core_waiters) {
1366                 up_write(&mm->mmap_sem);
1367                 wait_for_completion(&startup_done);
1368         } else
1369                 up_write(&mm->mmap_sem);
1370         BUG_ON(mm->core_waiters);
1371 }
1372
1373 int do_coredump(long signr, int exit_code, struct pt_regs * regs)
1374 {
1375         char corename[CORENAME_MAX_SIZE + 1];
1376         struct mm_struct *mm = current->mm;
1377         struct linux_binfmt * binfmt;
1378         struct inode * inode;
1379         struct file * file;
1380         int retval = 0;
1381
1382         binfmt = current->binfmt;
1383         if (!binfmt || !binfmt->core_dump)
1384                 goto fail;
1385         down_write(&mm->mmap_sem);
1386         if (!mm->dumpable) {
1387                 up_write(&mm->mmap_sem);
1388                 goto fail;
1389         }
1390         mm->dumpable = 0;
1391         init_completion(&mm->core_done);
1392         current->signal->group_exit = 1;
1393         current->signal->group_exit_code = exit_code;
1394         coredump_wait(mm);
1395
1396         if (current->rlim[RLIMIT_CORE].rlim_cur < binfmt->min_coredump)
1397                 goto fail_unlock;
1398
1399         /*
1400          * lock_kernel() because format_corename() is controlled by sysctl, which
1401          * uses lock_kernel()
1402          */
1403         lock_kernel();
1404         format_corename(corename, core_pattern, signr);
1405         unlock_kernel();
1406         file = filp_open(corename, O_CREAT | 2 | O_NOFOLLOW | O_LARGEFILE, 0600);
1407         if (IS_ERR(file))
1408                 goto fail_unlock;
1409         inode = file->f_dentry->d_inode;
1410         if (inode->i_nlink > 1)
1411                 goto close_fail;        /* multiple links - don't dump */
1412         if (d_unhashed(file->f_dentry))
1413                 goto close_fail;
1414
1415         if (!S_ISREG(inode->i_mode))
1416                 goto close_fail;
1417         if (!file->f_op)
1418                 goto close_fail;
1419         if (!file->f_op->write)
1420                 goto close_fail;
1421         if (do_truncate(file->f_dentry, 0) != 0)
1422                 goto close_fail;
1423
1424         retval = binfmt->core_dump(signr, regs, file);
1425
1426         current->signal->group_exit_code |= 0x80;
1427 close_fail:
1428         filp_close(file, NULL);
1429 fail_unlock:
1430         complete_all(&mm->core_done);
1431 fail:
1432         return retval;
1433 }