fs/open.c

   1 // SPDX-License-Identifier: GPL-2.0-only
   2 /*
   3  *  linux/fs/open.c
   4  *
   5  *  Copyright (C) 1991, 1992  Linus Torvalds
   6  */
   7
   8 #include <linux/string.h>
   9 #include <linux/mm.h>
  10 #include <linux/file.h>
  11 #include <linux/fdtable.h>
  12 #include <linux/fsnotify.h>
  13 #include <linux/module.h>
  14 #include <linux/tty.h>
  15 #include <linux/namei.h>
  16 #include <linux/backing-dev.h>
  17 #include <linux/capability.h>
  18 #include <linux/securebits.h>
  19 #include <linux/security.h>
  20 #include <linux/mount.h>
  21 #include <linux/fcntl.h>
  22 #include <linux/slab.h>
  23 #include <linux/uaccess.h>
  24 #include <linux/fs.h>
  25 #include <linux/personality.h>
  26 #include <linux/pagemap.h>
  27 #include <linux/syscalls.h>
  28 #include <linux/rcupdate.h>
  29 #include <linux/audit.h>
  30 #include <linux/falloc.h>
  31 #include <linux/fs_struct.h>
  32 #include <linux/ima.h>
  33 #include <linux/dnotify.h>
  34 #include <linux/compat.h>
  35 #include <linux/mnt_idmapping.h>
  36
  37 #include "internal.h"
  38
  39 int do_truncate(struct user_namespace *mnt_userns, struct dentry *dentry,
  40                 loff_t length, unsigned int time_attrs, struct file *filp)
  41 {
  42         int ret;
  43         struct iattr newattrs;
  44
  45         /* Not pretty: "inode->i_size" shouldn't really be signed. But it is. */
  46         if (length < 0)
  47                 return -EINVAL;
  48
  49         newattrs.ia_size = length;
  50         newattrs.ia_valid = ATTR_SIZE | time_attrs;
  51         if (filp) {
  52                 newattrs.ia_file = filp;
  53                 newattrs.ia_valid |= ATTR_FILE;
  54         }
  55
  56         /* Remove suid, sgid, and file capabilities on truncate too */
  57         ret = dentry_needs_remove_privs(dentry);
  58         if (ret < 0)
  59                 return ret;
  60         if (ret)
  61                 newattrs.ia_valid |= ret | ATTR_FORCE;
  62
  63         inode_lock(dentry->d_inode);
  64         /* Note any delegations or leases have already been broken: */
  65         ret = notify_change(mnt_userns, dentry, &newattrs, NULL);
  66         inode_unlock(dentry->d_inode);
  67         return ret;
  68 }
  69 EXPORT_SYMBOL_GPL(do_truncate);
  70
  71 long vfs_truncate(const struct path *path, loff_t length)
  72 {
  73         struct user_namespace *mnt_userns;
  74         struct inode *inode;
  75         long error;
  76
  77         inode = path->dentry->d_inode;
  78
  79         /* For directories it's -EISDIR, for other non-regulars - -EINVAL */
  80         if (S_ISDIR(inode->i_mode))
  81                 return -EISDIR;
  82         if (!S_ISREG(inode->i_mode))
  83                 return -EINVAL;
  84
  85         error = mnt_want_write(path->mnt);
  86         if (error)
  87                 goto out;
  88
  89         mnt_userns = mnt_user_ns(path->mnt);
  90         error = inode_permission(mnt_userns, inode, MAY_WRITE);
  91         if (error)
  92                 goto mnt_drop_write_and_out;
  93
  94         error = -EPERM;
  95         if (IS_APPEND(inode))
  96                 goto mnt_drop_write_and_out;
  97
  98         error = get_write_access(inode);
  99         if (error)
 100                 goto mnt_drop_write_and_out;
 101
 102         /*
 103          * Make sure that there are no leases.  get_write_access() protects
 104          * against the truncate racing with a lease-granting setlease().
 105          */
 106         error = break_lease(inode, O_WRONLY);
 107         if (error)
 108                 goto put_write_and_out;
 109
 110         error = security_path_truncate(path);
 111         if (!error)
 112                 error = do_truncate(mnt_userns, path->dentry, length, 0, NULL);
 113
 114 put_write_and_out:
 115         put_write_access(inode);
 116 mnt_drop_write_and_out:
 117         mnt_drop_write(path->mnt);
 118 out:
 119         return error;
 120 }
 121 EXPORT_SYMBOL_GPL(vfs_truncate);
 122
 123 long do_sys_truncate(const char __user *pathname, loff_t length)
 124 {
 125         unsigned int lookup_flags = LOOKUP_FOLLOW;
 126         struct path path;
 127         int error;
 128
 129         if (length < 0) /* sorry, but loff_t says... */
 130                 return -EINVAL;
 131
 132 retry:
 133         error = user_path_at(AT_FDCWD, pathname, lookup_flags, &path);
 134         if (!error) {
 135                 error = vfs_truncate(&path, length);
 136                 path_put(&path);
 137         }
 138         if (retry_estale(error, lookup_flags)) {
 139                 lookup_flags |= LOOKUP_REVAL;
 140                 goto retry;
 141         }
 142         return error;
 143 }
 144
 145 SYSCALL_DEFINE2(truncate, const char __user *, path, long, length)
 146 {
 147         return do_sys_truncate(path, length);
 148 }
 149
 150 #ifdef CONFIG_COMPAT
 151 COMPAT_SYSCALL_DEFINE2(truncate, const char __user *, path, compat_off_t, length)
 152 {
 153         return do_sys_truncate(path, length);
 154 }
 155 #endif
 156
 157 long do_sys_ftruncate(unsigned int fd, loff_t length, int small)
 158 {
 159         struct inode *inode;
 160         struct dentry *dentry;
 161         struct fd f;
 162         int error;
 163
 164         error = -EINVAL;
 165         if (length < 0)
 166                 goto out;
 167         error = -EBADF;
 168         f = fdget(fd);
 169         if (!f.file)
 170                 goto out;
 171
 172         /* explicitly opened as large or we are on 64-bit box */
 173         if (f.file->f_flags & O_LARGEFILE)
 174                 small = 0;
 175
 176         dentry = f.file->f_path.dentry;
 177         inode = dentry->d_inode;
 178         error = -EINVAL;
 179         if (!S_ISREG(inode->i_mode) || !(f.file->f_mode & FMODE_WRITE))
 180                 goto out_putf;
 181
 182         error = -EINVAL;
 183         /* Cannot ftruncate over 2^31 bytes without large file support */
 184         if (small && length > MAX_NON_LFS)
 185                 goto out_putf;
 186
 187         error = -EPERM;
 188         /* Check IS_APPEND on real upper inode */
 189         if (IS_APPEND(file_inode(f.file)))
 190                 goto out_putf;
 191         sb_start_write(inode->i_sb);
 192         error = security_path_truncate(&f.file->f_path);
 193         if (!error)
 194                 error = do_truncate(file_mnt_user_ns(f.file), dentry, length,
 195                                     ATTR_MTIME | ATTR_CTIME, f.file);
 196         sb_end_write(inode->i_sb);
 197 out_putf:
 198         fdput(f);
 199 out:
 200         return error;
 201 }
 202
 203 SYSCALL_DEFINE2(ftruncate, unsigned int, fd, unsigned long, length)
 204 {
 205         return do_sys_ftruncate(fd, length, 1);
 206 }
 207
 208 #ifdef CONFIG_COMPAT
 209 COMPAT_SYSCALL_DEFINE2(ftruncate, unsigned int, fd, compat_ulong_t, length)
 210 {
 211         return do_sys_ftruncate(fd, length, 1);
 212 }
 213 #endif
 214
 215 /* LFS versions of truncate are only needed on 32 bit machines */
 216 #if BITS_PER_LONG == 32
 217 SYSCALL_DEFINE2(truncate64, const char __user *, path, loff_t, length)
 218 {
 219         return do_sys_truncate(path, length);
 220 }
 221
 222 SYSCALL_DEFINE2(ftruncate64, unsigned int, fd, loff_t, length)
 223 {
 224         return do_sys_ftruncate(fd, length, 0);
 225 }
 226 #endif /* BITS_PER_LONG == 32 */
 227
 228
 229 int vfs_fallocate(struct file *file, int mode, loff_t offset, loff_t len)
 230 {
 231         struct inode *inode = file_inode(file);
 232         long ret;
 233
 234         if (offset < 0 || len <= 0)
 235                 return -EINVAL;
 236
 237         /* Return error if mode is not supported */
 238         if (mode & ~FALLOC_FL_SUPPORTED_MASK)
 239                 return -EOPNOTSUPP;
 240
 241         /* Punch hole and zero range are mutually exclusive */
 242         if ((mode & (FALLOC_FL_PUNCH_HOLE | FALLOC_FL_ZERO_RANGE)) ==
 243             (FALLOC_FL_PUNCH_HOLE | FALLOC_FL_ZERO_RANGE))
 244                 return -EOPNOTSUPP;
 245
 246         /* Punch hole must have keep size set */
 247         if ((mode & FALLOC_FL_PUNCH_HOLE) &&
 248             !(mode & FALLOC_FL_KEEP_SIZE))
 249                 return -EOPNOTSUPP;
 250
 251         /* Collapse range should only be used exclusively. */
 252         if ((mode & FALLOC_FL_COLLAPSE_RANGE) &&
 253             (mode & ~FALLOC_FL_COLLAPSE_RANGE))
 254                 return -EINVAL;
 255
 256         /* Insert range should only be used exclusively. */
 257         if ((mode & FALLOC_FL_INSERT_RANGE) &&
 258             (mode & ~FALLOC_FL_INSERT_RANGE))
 259                 return -EINVAL;
 260
 261         /* Unshare range should only be used with allocate mode. */
 262         if ((mode & FALLOC_FL_UNSHARE_RANGE) &&
 263             (mode & ~(FALLOC_FL_UNSHARE_RANGE | FALLOC_FL_KEEP_SIZE)))
 264                 return -EINVAL;
 265
 266         if (!(file->f_mode & FMODE_WRITE))
 267                 return -EBADF;
 268
 269         /*
 270          * We can only allow pure fallocate on append only files
 271          */
 272         if ((mode & ~FALLOC_FL_KEEP_SIZE) && IS_APPEND(inode))
 273                 return -EPERM;
 274
 275         if (IS_IMMUTABLE(inode))
 276                 return -EPERM;
 277
 278         /*
 279          * We cannot allow any fallocate operation on an active swapfile
 280          */
 281         if (IS_SWAPFILE(inode))
 282                 return -ETXTBSY;
 283
 284         /*
 285          * Revalidate the write permissions, in case security policy has
 286          * changed since the files were opened.
 287          */
 288         ret = security_file_permission(file, MAY_WRITE);
 289         if (ret)
 290                 return ret;
 291
 292         if (S_ISFIFO(inode->i_mode))
 293                 return -ESPIPE;
 294
 295         if (S_ISDIR(inode->i_mode))
 296                 return -EISDIR;
 297
 298         if (!S_ISREG(inode->i_mode) && !S_ISBLK(inode->i_mode))
 299                 return -ENODEV;
 300
 301         /* Check for wrap through zero too */
 302         if (((offset + len) > inode->i_sb->s_maxbytes) || ((offset + len) < 0))
 303                 return -EFBIG;
 304
 305         if (!file->f_op->fallocate)
 306                 return -EOPNOTSUPP;
 307
 308         file_start_write(file);
 309         ret = file->f_op->fallocate(file, mode, offset, len);
 310
 311         /*
 312          * Create inotify and fanotify events.
 313          *
 314          * To keep the logic simple always create events if fallocate succeeds.
 315          * This implies that events are even created if the file size remains
 316          * unchanged, e.g. when using flag FALLOC_FL_KEEP_SIZE.
 317          */
 318         if (ret == 0)
 319                 fsnotify_modify(file);
 320
 321         file_end_write(file);
 322         return ret;
 323 }
 324 EXPORT_SYMBOL_GPL(vfs_fallocate);
 325
 326 int ksys_fallocate(int fd, int mode, loff_t offset, loff_t len)
 327 {
 328         struct fd f = fdget(fd);
 329         int error = -EBADF;
 330
 331         if (f.file) {
 332                 error = vfs_fallocate(f.file, mode, offset, len);
 333                 fdput(f);
 334         }
 335         return error;
 336 }
 337
 338 SYSCALL_DEFINE4(fallocate, int, fd, int, mode, loff_t, offset, loff_t, len)
 339 {
 340         return ksys_fallocate(fd, mode, offset, len);
 341 }
 342
 343 /*
 344  * access() needs to use the real uid/gid, not the effective uid/gid.
 345  * We do this by temporarily clearing all FS-related capabilities and
 346  * switching the fsuid/fsgid around to the real ones.
 347  */
 348 static const struct cred *access_override_creds(void)
 349 {
 350         const struct cred *old_cred;
 351         struct cred *override_cred;
 352
 353         override_cred = prepare_creds();
 354         if (!override_cred)
 355                 return NULL;
 356
 357         override_cred->fsuid = override_cred->uid;
 358         override_cred->fsgid = override_cred->gid;
 359
 360         if (!issecure(SECURE_NO_SETUID_FIXUP)) {
 361                 /* Clear the capabilities if we switch to a non-root user */
 362                 kuid_t root_uid = make_kuid(override_cred->user_ns, 0);
 363                 if (!uid_eq(override_cred->uid, root_uid))
 364                         cap_clear(override_cred->cap_effective);
 365                 else
 366                         override_cred->cap_effective =
 367                                 override_cred->cap_permitted;
 368         }
 369
 370         /*
 371          * The new set of credentials can *only* be used in
 372          * task-synchronous circumstances, and does not need
 373          * RCU freeing, unless somebody then takes a separate
 374          * reference to it.
 375          *
 376          * NOTE! This is _only_ true because this credential
 377          * is used purely for override_creds() that installs
 378          * it as the subjective cred. Other threads will be
 379          * accessing ->real_cred, not the subjective cred.
 380          *
 381          * If somebody _does_ make a copy of this (using the
 382          * 'get_current_cred()' function), that will clear the
 383          * non_rcu field, because now that other user may be
 384          * expecting RCU freeing. But normal thread-synchronous
 385          * cred accesses will keep things non-RCY.
 386          */
 387         override_cred->non_rcu = 1;
 388
 389         old_cred = override_creds(override_cred);
 390
 391         /* override_cred() gets its own ref */
 392         put_cred(override_cred);
 393
 394         return old_cred;
 395 }
 396
 397 static long do_faccessat(int dfd, const char __user *filename, int mode, int flags)
 398 {
 399         struct path path;
 400         struct inode *inode;
 401         int res;
 402         unsigned int lookup_flags = LOOKUP_FOLLOW;
 403         const struct cred *old_cred = NULL;
 404
 405         if (mode & ~S_IRWXO)    /* where's F_OK, X_OK, W_OK, R_OK? */
 406                 return -EINVAL;
 407
 408         if (flags & ~(AT_EACCESS | AT_SYMLINK_NOFOLLOW | AT_EMPTY_PATH))
 409                 return -EINVAL;
 410
 411         if (flags & AT_SYMLINK_NOFOLLOW)
 412                 lookup_flags &= ~LOOKUP_FOLLOW;
 413         if (flags & AT_EMPTY_PATH)
 414                 lookup_flags |= LOOKUP_EMPTY;
 415
 416         if (!(flags & AT_EACCESS)) {
 417                 old_cred = access_override_creds();
 418                 if (!old_cred)
 419                         return -ENOMEM;
 420         }
 421
 422 retry:
 423         res = user_path_at(dfd, filename, lookup_flags, &path);
 424         if (res)
 425                 goto out;
 426
 427         inode = d_backing_inode(path.dentry);
 428
 429         if ((mode & MAY_EXEC) && S_ISREG(inode->i_mode)) {
 430                 /*
 431                  * MAY_EXEC on regular files is denied if the fs is mounted
 432                  * with the "noexec" flag.
 433                  */
 434                 res = -EACCES;
 435                 if (path_noexec(&path))
 436                         goto out_path_release;
 437         }
 438
 439         res = inode_permission(mnt_user_ns(path.mnt), inode, mode | MAY_ACCESS);
 440         /* SuS v2 requires we report a read only fs too */
 441         if (res || !(mode & S_IWOTH) || special_file(inode->i_mode))
 442                 goto out_path_release;
 443         /*
 444          * This is a rare case where using __mnt_is_readonly()
 445          * is OK without a mnt_want/drop_write() pair.  Since
 446          * no actual write to the fs is performed here, we do
 447          * not need to telegraph to that to anyone.
 448          *
 449          * By doing this, we accept that this access is
 450          * inherently racy and know that the fs may change
 451          * state before we even see this result.
 452          */
 453         if (__mnt_is_readonly(path.mnt))
 454                 res = -EROFS;
 455
 456 out_path_release:
 457         path_put(&path);
 458         if (retry_estale(res, lookup_flags)) {
 459                 lookup_flags |= LOOKUP_REVAL;
 460                 goto retry;
 461         }
 462 out:
 463         if (old_cred)
 464                 revert_creds(old_cred);
 465
 466         return res;
 467 }
 468
 469 SYSCALL_DEFINE3(faccessat, int, dfd, const char __user *, filename, int, mode)
 470 {
 471         return do_faccessat(dfd, filename, mode, 0);
 472 }
 473
 474 SYSCALL_DEFINE4(faccessat2, int, dfd, const char __user *, filename, int, mode,
 475                 int, flags)
 476 {
 477         return do_faccessat(dfd, filename, mode, flags);
 478 }
 479
 480 SYSCALL_DEFINE2(access, const char __user *, filename, int, mode)
 481 {
 482         return do_faccessat(AT_FDCWD, filename, mode, 0);
 483 }
 484
 485 SYSCALL_DEFINE1(chdir, const char __user *, filename)
 486 {
 487         struct path path;
 488         int error;
 489         unsigned int lookup_flags = LOOKUP_FOLLOW | LOOKUP_DIRECTORY;
 490 retry:
 491         error = user_path_at(AT_FDCWD, filename, lookup_flags, &path);
 492         if (error)
 493                 goto out;
 494
 495         error = path_permission(&path, MAY_EXEC | MAY_CHDIR);
 496         if (error)
 497                 goto dput_and_out;
 498
 499         set_fs_pwd(current->fs, &path);
 500
 501 dput_and_out:
 502         path_put(&path);
 503         if (retry_estale(error, lookup_flags)) {
 504                 lookup_flags |= LOOKUP_REVAL;
 505                 goto retry;
 506         }
 507 out:
 508         return error;
 509 }
 510
 511 SYSCALL_DEFINE1(fchdir, unsigned int, fd)
 512 {
 513         struct fd f = fdget_raw(fd);
 514         int error;
 515
 516         error = -EBADF;
 517         if (!f.file)
 518                 goto out;
 519
 520         error = -ENOTDIR;
 521         if (!d_can_lookup(f.file->f_path.dentry))
 522                 goto out_putf;
 523
 524         error = file_permission(f.file, MAY_EXEC | MAY_CHDIR);
 525         if (!error)
 526                 set_fs_pwd(current->fs, &f.file->f_path);
 527 out_putf:
 528         fdput(f);
 529 out:
 530         return error;
 531 }
 532
 533 SYSCALL_DEFINE1(chroot, const char __user *, filename)
 534 {
 535         struct path path;
 536         int error;
 537         unsigned int lookup_flags = LOOKUP_FOLLOW | LOOKUP_DIRECTORY;
 538 retry:
 539         error = user_path_at(AT_FDCWD, filename, lookup_flags, &path);
 540         if (error)
 541                 goto out;
 542
 543         error = path_permission(&path, MAY_EXEC | MAY_CHDIR);
 544         if (error)
 545                 goto dput_and_out;
 546
 547         error = -EPERM;
 548         if (!ns_capable(current_user_ns(), CAP_SYS_CHROOT))
 549                 goto dput_and_out;
 550         error = security_path_chroot(&path);
 551         if (error)
 552                 goto dput_and_out;
 553
 554         set_fs_root(current->fs, &path);
 555         error = 0;
 556 dput_and_out:
 557         path_put(&path);
 558         if (retry_estale(error, lookup_flags)) {
 559                 lookup_flags |= LOOKUP_REVAL;
 560                 goto retry;
 561         }
 562 out:
 563         return error;
 564 }
 565
 566 int chmod_common(const struct path *path, umode_t mode)
 567 {
 568         struct inode *inode = path->dentry->d_inode;
 569         struct inode *delegated_inode = NULL;
 570         struct iattr newattrs;
 571         int error;
 572
 573         error = mnt_want_write(path->mnt);
 574         if (error)
 575                 return error;
 576 retry_deleg:
 577         inode_lock(inode);
 578         error = security_path_chmod(path, mode);
 579         if (error)
 580                 goto out_unlock;
 581         newattrs.ia_mode = (mode & S_IALLUGO) | (inode->i_mode & ~S_IALLUGO);
 582         newattrs.ia_valid = ATTR_MODE | ATTR_CTIME;
 583         error = notify_change(mnt_user_ns(path->mnt), path->dentry,
 584                               &newattrs, &delegated_inode);
 585 out_unlock:
 586         inode_unlock(inode);
 587         if (delegated_inode) {
 588                 error = break_deleg_wait(&delegated_inode);
 589                 if (!error)
 590                         goto retry_deleg;
 591         }
 592         mnt_drop_write(path->mnt);
 593         return error;
 594 }
 595
 596 int vfs_fchmod(struct file *file, umode_t mode)
 597 {
 598         audit_file(file);
 599         return chmod_common(&file->f_path, mode);
 600 }
 601
 602 SYSCALL_DEFINE2(fchmod, unsigned int, fd, umode_t, mode)
 603 {
 604         struct fd f = fdget(fd);
 605         int err = -EBADF;
 606
 607         if (f.file) {
 608                 err = vfs_fchmod(f.file, mode);
 609                 fdput(f);
 610         }
 611         return err;
 612 }
 613
 614 static int do_fchmodat(int dfd, const char __user *filename, umode_t mode)
 615 {
 616         struct path path;
 617         int error;
 618         unsigned int lookup_flags = LOOKUP_FOLLOW;
 619 retry:
 620         error = user_path_at(dfd, filename, lookup_flags, &path);
 621         if (!error) {
 622                 error = chmod_common(&path, mode);
 623                 path_put(&path);
 624                 if (retry_estale(error, lookup_flags)) {
 625                         lookup_flags |= LOOKUP_REVAL;
 626                         goto retry;
 627                 }
 628         }
 629         return error;
 630 }
 631
 632 SYSCALL_DEFINE3(fchmodat, int, dfd, const char __user *, filename,
 633                 umode_t, mode)
 634 {
 635         return do_fchmodat(dfd, filename, mode);
 636 }
 637
 638 SYSCALL_DEFINE2(chmod, const char __user *, filename, umode_t, mode)
 639 {
 640         return do_fchmodat(AT_FDCWD, filename, mode);
 641 }
 642
 643 int chown_common(const struct path *path, uid_t user, gid_t group)
 644 {
 645         struct user_namespace *mnt_userns, *fs_userns;
 646         struct inode *inode = path->dentry->d_inode;
 647         struct inode *delegated_inode = NULL;
 648         int error;
 649         struct iattr newattrs;
 650         kuid_t uid;
 651         kgid_t gid;
 652
 653         uid = make_kuid(current_user_ns(), user);
 654         gid = make_kgid(current_user_ns(), group);
 655
 656         mnt_userns = mnt_user_ns(path->mnt);
 657         fs_userns = i_user_ns(inode);
 658         uid = mapped_kuid_user(mnt_userns, fs_userns, uid);
 659         gid = mapped_kgid_user(mnt_userns, fs_userns, gid);
 660
 661 retry_deleg:
 662         newattrs.ia_valid =  ATTR_CTIME;
 663         if (user != (uid_t) -1) {
 664                 if (!uid_valid(uid))
 665                         return -EINVAL;
 666                 newattrs.ia_valid |= ATTR_UID;
 667                 newattrs.ia_uid = uid;
 668         }
 669         if (group != (gid_t) -1) {
 670                 if (!gid_valid(gid))
 671                         return -EINVAL;
 672                 newattrs.ia_valid |= ATTR_GID;
 673                 newattrs.ia_gid = gid;
 674         }
 675         if (!S_ISDIR(inode->i_mode))
 676                 newattrs.ia_valid |=
 677                         ATTR_KILL_SUID | ATTR_KILL_SGID | ATTR_KILL_PRIV;
 678         inode_lock(inode);
 679         error = security_path_chown(path, uid, gid);
 680         if (!error)
 681                 error = notify_change(mnt_userns, path->dentry, &newattrs,
 682                                       &delegated_inode);
 683         inode_unlock(inode);
 684         if (delegated_inode) {
 685                 error = break_deleg_wait(&delegated_inode);
 686                 if (!error)
 687                         goto retry_deleg;
 688         }
 689         return error;
 690 }
 691
 692 int do_fchownat(int dfd, const char __user *filename, uid_t user, gid_t group,
 693                 int flag)
 694 {
 695         struct path path;
 696         int error = -EINVAL;
 697         int lookup_flags;
 698
 699         if ((flag & ~(AT_SYMLINK_NOFOLLOW | AT_EMPTY_PATH)) != 0)
 700                 goto out;
 701
 702         lookup_flags = (flag & AT_SYMLINK_NOFOLLOW) ? 0 : LOOKUP_FOLLOW;
 703         if (flag & AT_EMPTY_PATH)
 704                 lookup_flags |= LOOKUP_EMPTY;
 705 retry:
 706         error = user_path_at(dfd, filename, lookup_flags, &path);
 707         if (error)
 708                 goto out;
 709         error = mnt_want_write(path.mnt);
 710         if (error)
 711                 goto out_release;
 712         error = chown_common(&path, user, group);
 713         mnt_drop_write(path.mnt);
 714 out_release:
 715         path_put(&path);
 716         if (retry_estale(error, lookup_flags)) {
 717                 lookup_flags |= LOOKUP_REVAL;
 718                 goto retry;
 719         }
 720 out:
 721         return error;
 722 }
 723
 724 SYSCALL_DEFINE5(fchownat, int, dfd, const char __user *, filename, uid_t, user,
 725                 gid_t, group, int, flag)
 726 {
 727         return do_fchownat(dfd, filename, user, group, flag);
 728 }
 729
 730 SYSCALL_DEFINE3(chown, const char __user *, filename, uid_t, user, gid_t, group)
 731 {
 732         return do_fchownat(AT_FDCWD, filename, user, group, 0);
 733 }
 734
 735 SYSCALL_DEFINE3(lchown, const char __user *, filename, uid_t, user, gid_t, group)
 736 {
 737         return do_fchownat(AT_FDCWD, filename, user, group,
 738                            AT_SYMLINK_NOFOLLOW);
 739 }
 740
 741 int vfs_fchown(struct file *file, uid_t user, gid_t group)
 742 {
 743         int error;
 744
 745         error = mnt_want_write_file(file);
 746         if (error)
 747                 return error;
 748         audit_file(file);
 749         error = chown_common(&file->f_path, user, group);
 750         mnt_drop_write_file(file);
 751         return error;
 752 }
 753
 754 int ksys_fchown(unsigned int fd, uid_t user, gid_t group)
 755 {
 756         struct fd f = fdget(fd);
 757         int error = -EBADF;
 758
 759         if (f.file) {
 760                 error = vfs_fchown(f.file, user, group);
 761                 fdput(f);
 762         }
 763         return error;
 764 }
 765
 766 SYSCALL_DEFINE3(fchown, unsigned int, fd, uid_t, user, gid_t, group)
 767 {
 768         return ksys_fchown(fd, user, group);
 769 }
 770
 771 static int do_dentry_open(struct file *f,
 772                           struct inode *inode,
 773                           int (*open)(struct inode *, struct file *))
 774 {
 775         static const struct file_operations empty_fops = {};
 776         int error;
 777
 778         path_get(&f->f_path);
 779         f->f_inode = inode;
 780         f->f_mapping = inode->i_mapping;
 781         f->f_wb_err = filemap_sample_wb_err(f->f_mapping);
 782         f->f_sb_err = file_sample_sb_err(f);
 783
 784         if (unlikely(f->f_flags & O_PATH)) {
 785                 f->f_mode = FMODE_PATH | FMODE_OPENED;
 786                 f->f_op = &empty_fops;
 787                 return 0;
 788         }
 789
 790         if (f->f_mode & FMODE_WRITE && !special_file(inode->i_mode)) {
 791                 error = get_write_access(inode);
 792                 if (unlikely(error))
 793                         goto cleanup_file;
 794                 error = __mnt_want_write(f->f_path.mnt);
 795                 if (unlikely(error)) {
 796                         put_write_access(inode);
 797                         goto cleanup_file;
 798                 }
 799                 f->f_mode |= FMODE_WRITER;
 800         }
 801
 802         /* POSIX.1-2008/SUSv4 Section XSI 2.9.7 */
 803         if (S_ISREG(inode->i_mode) || S_ISDIR(inode->i_mode))
 804                 f->f_mode |= FMODE_ATOMIC_POS;
 805
 806         f->f_op = fops_get(inode->i_fop);
 807         if (WARN_ON(!f->f_op)) {
 808                 error = -ENODEV;
 809                 goto cleanup_all;
 810         }
 811
 812         error = security_file_open(f);
 813         if (error)
 814                 goto cleanup_all;
 815
 816         error = break_lease(locks_inode(f), f->f_flags);
 817         if (error)
 818                 goto cleanup_all;
 819
 820         /* normally all 3 are set; ->open() can clear them if needed */
 821         f->f_mode |= FMODE_LSEEK | FMODE_PREAD | FMODE_PWRITE;
 822         if (!open)
 823                 open = f->f_op->open;
 824         if (open) {
 825                 error = open(inode, f);
 826                 if (error)
 827                         goto cleanup_all;
 828         }
 829         f->f_mode |= FMODE_OPENED;
 830         if ((f->f_mode & (FMODE_READ | FMODE_WRITE)) == FMODE_READ)
 831                 i_readcount_inc(inode);
 832         if ((f->f_mode & FMODE_READ) &&
 833              likely(f->f_op->read || f->f_op->read_iter))
 834                 f->f_mode |= FMODE_CAN_READ;
 835         if ((f->f_mode & FMODE_WRITE) &&
 836              likely(f->f_op->write || f->f_op->write_iter))
 837                 f->f_mode |= FMODE_CAN_WRITE;
 838
 839         f->f_write_hint = WRITE_LIFE_NOT_SET;
 840         f->f_flags &= ~(O_CREAT | O_EXCL | O_NOCTTY | O_TRUNC);
 841
 842         file_ra_state_init(&f->f_ra, f->f_mapping->host->i_mapping);
 843
 844         /* NB: we're sure to have correct a_ops only after f_op->open */
 845         if (f->f_flags & O_DIRECT) {
 846                 if (!f->f_mapping->a_ops || !f->f_mapping->a_ops->direct_IO)
 847                         return -EINVAL;
 848         }
 849
 850         /*
 851          * XXX: Huge page cache doesn't support writing yet. Drop all page
 852          * cache for this file before processing writes.
 853          */
 854         if (f->f_mode & FMODE_WRITE) {
 855                 /*
 856                  * Paired with smp_mb() in collapse_file() to ensure nr_thps
 857                  * is up to date and the update to i_writecount by
 858                  * get_write_access() is visible. Ensures subsequent insertion
 859                  * of THPs into the page cache will fail.
 860                  */
 861                 smp_mb();
 862                 if (filemap_nr_thps(inode->i_mapping)) {
 863                         struct address_space *mapping = inode->i_mapping;
 864
 865                         filemap_invalidate_lock(inode->i_mapping);
 866                         /*
 867                          * unmap_mapping_range just need to be called once
 868                          * here, because the private pages is not need to be
 869                          * unmapped mapping (e.g. data segment of dynamic
 870                          * shared libraries here).
 871                          */
 872                         unmap_mapping_range(mapping, 0, 0, 0);
 873                         truncate_inode_pages(mapping, 0);
 874                         filemap_invalidate_unlock(inode->i_mapping);
 875                 }
 876         }
 877
 878         return 0;
 879
 880 cleanup_all:
 881         if (WARN_ON_ONCE(error > 0))
 882                 error = -EINVAL;
 883         fops_put(f->f_op);
 884         if (f->f_mode & FMODE_WRITER) {
 885                 put_write_access(inode);
 886                 __mnt_drop_write(f->f_path.mnt);
 887         }
 888 cleanup_file:
 889         path_put(&f->f_path);
 890         f->f_path.mnt = NULL;
 891         f->f_path.dentry = NULL;
 892         f->f_inode = NULL;
 893         return error;
 894 }
 895
 896 /**
 897  * finish_open - finish opening a file
 898  * @file: file pointer
 899  * @dentry: pointer to dentry
 900  * @open: open callback
 901  * @opened: state of open
 902  *
 903  * This can be used to finish opening a file passed to i_op->atomic_open().
 904  *
 905  * If the open callback is set to NULL, then the standard f_op->open()
 906  * filesystem callback is substituted.
 907  *
 908  * NB: the dentry reference is _not_ consumed.  If, for example, the dentry is
 909  * the return value of d_splice_alias(), then the caller needs to perform dput()
 910  * on it after finish_open().
 911  *
 912  * Returns zero on success or -errno if the open failed.
 913  */
 914 int finish_open(struct file *file, struct dentry *dentry,
 915                 int (*open)(struct inode *, struct file *))
 916 {
 917         BUG_ON(file->f_mode & FMODE_OPENED); /* once it's opened, it's opened */
 918
 919         file->f_path.dentry = dentry;
 920         return do_dentry_open(file, d_backing_inode(dentry), open);
 921 }
 922 EXPORT_SYMBOL(finish_open);
 923
 924 /**
 925  * finish_no_open - finish ->atomic_open() without opening the file
 926  *
 927  * @file: file pointer
 928  * @dentry: dentry or NULL (as returned from ->lookup())
 929  *
 930  * This can be used to set the result of a successful lookup in ->atomic_open().
 931  *
 932  * NB: unlike finish_open() this function does consume the dentry reference and
 933  * the caller need not dput() it.
 934  *
 935  * Returns "0" which must be the return value of ->atomic_open() after having
 936  * called this function.
 937  */
 938 int finish_no_open(struct file *file, struct dentry *dentry)
 939 {
 940         file->f_path.dentry = dentry;
 941         return 0;
 942 }
 943 EXPORT_SYMBOL(finish_no_open);
 944
 945 char *file_path(struct file *filp, char *buf, int buflen)
 946 {
 947         return d_path(&filp->f_path, buf, buflen);
 948 }
 949 EXPORT_SYMBOL(file_path);
 950
 951 /**
 952  * vfs_open - open the file at the given path
 953  * @path: path to open
 954  * @file: newly allocated file with f_flag initialized
 955  * @cred: credentials to use
 956  */
 957 int vfs_open(const struct path *path, struct file *file)
 958 {
 959         file->f_path = *path;
 960         return do_dentry_open(file, d_backing_inode(path->dentry), NULL);
 961 }
 962
 963 struct file *dentry_open(const struct path *path, int flags,
 964                          const struct cred *cred)
 965 {
 966         int error;
 967         struct file *f;
 968
 969         validate_creds(cred);
 970
 971         /* We must always pass in a valid mount pointer. */
 972         BUG_ON(!path->mnt);
 973
 974         f = alloc_empty_file(flags, cred);
 975         if (!IS_ERR(f)) {
 976                 error = vfs_open(path, f);
 977                 if (error) {
 978                         fput(f);
 979                         f = ERR_PTR(error);
 980                 }
 981         }
 982         return f;
 983 }
 984 EXPORT_SYMBOL(dentry_open);
 985
 986 struct file *open_with_fake_path(const struct path *path, int flags,
 987                                 struct inode *inode, const struct cred *cred)
 988 {
 989         struct file *f = alloc_empty_file_noaccount(flags, cred);
 990         if (!IS_ERR(f)) {
 991                 int error;
 992
 993                 f->f_path = *path;
 994                 error = do_dentry_open(f, inode, NULL);
 995                 if (error) {
 996                         fput(f);
 997                         f = ERR_PTR(error);
 998                 }
 999         }
1000         return f;
1001 }
1002 EXPORT_SYMBOL(open_with_fake_path);
1003
1004 #define WILL_CREATE(flags)      (flags & (O_CREAT | __O_TMPFILE))
1005 #define O_PATH_FLAGS            (O_DIRECTORY | O_NOFOLLOW | O_PATH | O_CLOEXEC)
1006
1007 inline struct open_how build_open_how(int flags, umode_t mode)
1008 {
1009         struct open_how how = {
1010                 .flags = flags & VALID_OPEN_FLAGS,
1011                 .mode = mode & S_IALLUGO,
1012         };
1013
1014         /* O_PATH beats everything else. */
1015         if (how.flags & O_PATH)
1016                 how.flags &= O_PATH_FLAGS;
1017         /* Modes should only be set for create-like flags. */
1018         if (!WILL_CREATE(how.flags))
1019                 how.mode = 0;
1020         return how;
1021 }
1022
1023 inline int build_open_flags(const struct open_how *how, struct open_flags *op)
1024 {
1025         u64 flags = how->flags;
1026         u64 strip = FMODE_NONOTIFY | O_CLOEXEC;
1027         int lookup_flags = 0;
1028         int acc_mode = ACC_MODE(flags);
1029
1030         BUILD_BUG_ON_MSG(upper_32_bits(VALID_OPEN_FLAGS),
1031                          "struct open_flags doesn't yet handle flags > 32 bits");
1032
1033         /*
1034          * Strip flags that either shouldn't be set by userspace like
1035          * FMODE_NONOTIFY or that aren't relevant in determining struct
1036          * open_flags like O_CLOEXEC.
1037          */
1038         flags &= ~strip;
1039
1040         /*
1041          * Older syscalls implicitly clear all of the invalid flags or argument
1042          * values before calling build_open_flags(), but openat2(2) checks all
1043          * of its arguments.
1044          */
1045         if (flags & ~VALID_OPEN_FLAGS)
1046                 return -EINVAL;
1047         if (how->resolve & ~VALID_RESOLVE_FLAGS)
1048                 return -EINVAL;
1049
1050         /* Scoping flags are mutually exclusive. */
1051         if ((how->resolve & RESOLVE_BENEATH) && (how->resolve & RESOLVE_IN_ROOT))
1052                 return -EINVAL;
1053
1054         /* Deal with the mode. */
1055         if (WILL_CREATE(flags)) {
1056                 if (how->mode & ~S_IALLUGO)
1057                         return -EINVAL;
1058                 op->mode = how->mode | S_IFREG;
1059         } else {
1060                 if (how->mode != 0)
1061                         return -EINVAL;
1062                 op->mode = 0;
1063         }
1064
1065         /*
1066          * In order to ensure programs get explicit errors when trying to use
1067          * O_TMPFILE on old kernels, O_TMPFILE is implemented such that it
1068          * looks like (O_DIRECTORY|O_RDWR & ~O_CREAT) to old kernels. But we
1069          * have to require userspace to explicitly set it.
1070          */
1071         if (flags & __O_TMPFILE) {
1072                 if ((flags & O_TMPFILE_MASK) != O_TMPFILE)
1073                         return -EINVAL;
1074                 if (!(acc_mode & MAY_WRITE))
1075                         return -EINVAL;
1076         }
1077         if (flags & O_PATH) {
1078                 /* O_PATH only permits certain other flags to be set. */
1079                 if (flags & ~O_PATH_FLAGS)
1080                         return -EINVAL;
1081                 acc_mode = 0;
1082         }
1083
1084         /*
1085          * O_SYNC is implemented as __O_SYNC|O_DSYNC.  As many places only
1086          * check for O_DSYNC if the need any syncing at all we enforce it's
1087          * always set instead of having to deal with possibly weird behaviour
1088          * for malicious applications setting only __O_SYNC.
1089          */
1090         if (flags & __O_SYNC)
1091                 flags |= O_DSYNC;
1092
1093         op->open_flag = flags;
1094
1095         /* O_TRUNC implies we need access checks for write permissions */
1096         if (flags & O_TRUNC)
1097                 acc_mode |= MAY_WRITE;
1098
1099         /* Allow the LSM permission hook to distinguish append
1100            access from general write access. */
1101         if (flags & O_APPEND)
1102                 acc_mode |= MAY_APPEND;
1103
1104         op->acc_mode = acc_mode;
1105
1106         op->intent = flags & O_PATH ? 0 : LOOKUP_OPEN;
1107
1108         if (flags & O_CREAT) {
1109                 op->intent |= LOOKUP_CREATE;
1110                 if (flags & O_EXCL) {
1111                         op->intent |= LOOKUP_EXCL;
1112                         flags |= O_NOFOLLOW;
1113                 }
1114         }
1115
1116         if (flags & O_DIRECTORY)
1117                 lookup_flags |= LOOKUP_DIRECTORY;
1118         if (!(flags & O_NOFOLLOW))
1119                 lookup_flags |= LOOKUP_FOLLOW;
1120
1121         if (how->resolve & RESOLVE_NO_XDEV)
1122                 lookup_flags |= LOOKUP_NO_XDEV;
1123         if (how->resolve & RESOLVE_NO_MAGICLINKS)
1124                 lookup_flags |= LOOKUP_NO_MAGICLINKS;
1125         if (how->resolve & RESOLVE_NO_SYMLINKS)
1126                 lookup_flags |= LOOKUP_NO_SYMLINKS;
1127         if (how->resolve & RESOLVE_BENEATH)
1128                 lookup_flags |= LOOKUP_BENEATH;
1129         if (how->resolve & RESOLVE_IN_ROOT)
1130                 lookup_flags |= LOOKUP_IN_ROOT;
1131         if (how->resolve & RESOLVE_CACHED) {
1132                 /* Don't bother even trying for create/truncate/tmpfile open */
1133                 if (flags & (O_TRUNC | O_CREAT | O_TMPFILE))
1134                         return -EAGAIN;
1135                 lookup_flags |= LOOKUP_CACHED;
1136         }
1137
1138         op->lookup_flags = lookup_flags;
1139         return 0;
1140 }
1141
1142 /**
1143  * file_open_name - open file and return file pointer
1144  *
1145  * @name:       struct filename containing path to open
1146  * @flags:      open flags as per the open(2) second argument
1147  * @mode:       mode for the new file if O_CREAT is set, else ignored
1148  *
1149  * This is the helper to open a file from kernelspace if you really
1150  * have to.  But in generally you should not do this, so please move
1151  * along, nothing to see here..
1152  */
1153 struct file *file_open_name(struct filename *name, int flags, umode_t mode)
1154 {
1155         struct open_flags op;
1156         struct open_how how = build_open_how(flags, mode);
1157         int err = build_open_flags(&how, &op);
1158         if (err)
1159                 return ERR_PTR(err);
1160         return do_filp_open(AT_FDCWD, name, &op);
1161 }
1162
1163 /**
1164  * filp_open - open file and return file pointer
1165  *
1166  * @filename:   path to open
1167  * @flags:      open flags as per the open(2) second argument
1168  * @mode:       mode for the new file if O_CREAT is set, else ignored
1169  *
1170  * This is the helper to open a file from kernelspace if you really
1171  * have to.  But in generally you should not do this, so please move
1172  * along, nothing to see here..
1173  */
1174 struct file *filp_open(const char *filename, int flags, umode_t mode)
1175 {
1176         struct filename *name = getname_kernel(filename);
1177         struct file *file = ERR_CAST(name);
1178
1179         if (!IS_ERR(name)) {
1180                 file = file_open_name(name, flags, mode);
1181                 putname(name);
1182         }
1183         return file;
1184 }
1185 EXPORT_SYMBOL(filp_open);
1186
1187 struct file *file_open_root(const struct path *root,
1188                             const char *filename, int flags, umode_t mode)
1189 {
1190         struct open_flags op;
1191         struct open_how how = build_open_how(flags, mode);
1192         int err = build_open_flags(&how, &op);
1193         if (err)
1194                 return ERR_PTR(err);
1195         return do_file_open_root(root, filename, &op);
1196 }
1197 EXPORT_SYMBOL(file_open_root);
1198
1199 static long do_sys_openat2(int dfd, const char __user *filename,
1200                            struct open_how *how)
1201 {
1202         struct open_flags op;
1203         int fd = build_open_flags(how, &op);
1204         struct filename *tmp;
1205
1206         if (fd)
1207                 return fd;
1208
1209         tmp = getname(filename);
1210         if (IS_ERR(tmp))
1211                 return PTR_ERR(tmp);
1212
1213         fd = get_unused_fd_flags(how->flags);
1214         if (fd >= 0) {
1215                 struct file *f = do_filp_open(dfd, tmp, &op);
1216                 if (IS_ERR(f)) {
1217                         put_unused_fd(fd);
1218                         fd = PTR_ERR(f);
1219                 } else {
1220                         fsnotify_open(f);
1221                         fd_install(fd, f);
1222                 }
1223         }
1224         putname(tmp);
1225         return fd;
1226 }
1227
1228 long do_sys_open(int dfd, const char __user *filename, int flags, umode_t mode)
1229 {
1230         struct open_how how = build_open_how(flags, mode);
1231         return do_sys_openat2(dfd, filename, &how);
1232 }
1233
1234
1235 SYSCALL_DEFINE3(open, const char __user *, filename, int, flags, umode_t, mode)
1236 {
1237         if (force_o_largefile())
1238                 flags |= O_LARGEFILE;
1239         return do_sys_open(AT_FDCWD, filename, flags, mode);
1240 }
1241
1242 SYSCALL_DEFINE4(openat, int, dfd, const char __user *, filename, int, flags,
1243                 umode_t, mode)
1244 {
1245         if (force_o_largefile())
1246                 flags |= O_LARGEFILE;
1247         return do_sys_open(dfd, filename, flags, mode);
1248 }
1249
1250 SYSCALL_DEFINE4(openat2, int, dfd, const char __user *, filename,
1251                 struct open_how __user *, how, size_t, usize)
1252 {
1253         int err;
1254         struct open_how tmp;
1255
1256         BUILD_BUG_ON(sizeof(struct open_how) < OPEN_HOW_SIZE_VER0);
1257         BUILD_BUG_ON(sizeof(struct open_how) != OPEN_HOW_SIZE_LATEST);
1258
1259         if (unlikely(usize < OPEN_HOW_SIZE_VER0))
1260                 return -EINVAL;
1261
1262         err = copy_struct_from_user(&tmp, sizeof(tmp), how, usize);
1263         if (err)
1264                 return err;
1265
1266         /* O_LARGEFILE is only allowed for non-O_PATH. */
1267         if (!(tmp.flags & O_PATH) && force_o_largefile())
1268                 tmp.flags |= O_LARGEFILE;
1269
1270         return do_sys_openat2(dfd, filename, &tmp);
1271 }
1272
1273 #ifdef CONFIG_COMPAT
1274 /*
1275  * Exactly like sys_open(), except that it doesn't set the
1276  * O_LARGEFILE flag.
1277  */
1278 COMPAT_SYSCALL_DEFINE3(open, const char __user *, filename, int, flags, umode_t, mode)
1279 {
1280         return do_sys_open(AT_FDCWD, filename, flags, mode);
1281 }
1282
1283 /*
1284  * Exactly like sys_openat(), except that it doesn't set the
1285  * O_LARGEFILE flag.
1286  */
1287 COMPAT_SYSCALL_DEFINE4(openat, int, dfd, const char __user *, filename, int, flags, umode_t, mode)
1288 {
1289         return do_sys_open(dfd, filename, flags, mode);
1290 }
1291 #endif
1292
1293 #ifndef __alpha__
1294
1295 /*
1296  * For backward compatibility?  Maybe this should be moved
1297  * into arch/i386 instead?
1298  */
1299 SYSCALL_DEFINE2(creat, const char __user *, pathname, umode_t, mode)
1300 {
1301         int flags = O_CREAT | O_WRONLY | O_TRUNC;
1302
1303         if (force_o_largefile())
1304                 flags |= O_LARGEFILE;
1305         return do_sys_open(AT_FDCWD, pathname, flags, mode);
1306 }
1307 #endif
1308
1309 /*
1310  * "id" is the POSIX thread ID. We use the
1311  * files pointer for this..
1312  */
1313 int filp_close(struct file *filp, fl_owner_t id)
1314 {
1315         int retval = 0;
1316
1317         if (!file_count(filp)) {
1318                 printk(KERN_ERR "VFS: Close: file count is 0\n");
1319                 return 0;
1320         }
1321
1322         if (filp->f_op->flush)
1323                 retval = filp->f_op->flush(filp, id);
1324
1325         if (likely(!(filp->f_mode & FMODE_PATH))) {
1326                 dnotify_flush(filp, id);
1327                 locks_remove_posix(filp, id);
1328         }
1329         fput(filp);
1330         return retval;
1331 }
1332
1333 EXPORT_SYMBOL(filp_close);
1334
1335 /*
1336  * Careful here! We test whether the file pointer is NULL before
1337  * releasing the fd. This ensures that one clone task can't release
1338  * an fd while another clone is opening it.
1339  */
1340 SYSCALL_DEFINE1(close, unsigned int, fd)
1341 {
1342         int retval = close_fd(fd);
1343
1344         /* can't restart close syscall because file table entry was cleared */
1345         if (unlikely(retval == -ERESTARTSYS ||
1346                      retval == -ERESTARTNOINTR ||
1347                      retval == -ERESTARTNOHAND ||
1348                      retval == -ERESTART_RESTARTBLOCK))
1349                 retval = -EINTR;
1350
1351         return retval;
1352 }
1353
1354 /**
1355  * close_range() - Close all file descriptors in a given range.
1356  *
1357  * @fd:     starting file descriptor to close
1358  * @max_fd: last file descriptor to close
1359  * @flags:  reserved for future extensions
1360  *
1361  * This closes a range of file descriptors. All file descriptors
1362  * from @fd up to and including @max_fd are closed.
1363  * Currently, errors to close a given file descriptor are ignored.
1364  */
1365 SYSCALL_DEFINE3(close_range, unsigned int, fd, unsigned int, max_fd,
1366                 unsigned int, flags)
1367 {
1368         return __close_range(fd, max_fd, flags);
1369 }
1370
1371 /*
1372  * This routine simulates a hangup on the tty, to arrange that users
1373  * are given clean terminals at login time.
1374  */
1375 SYSCALL_DEFINE0(vhangup)
1376 {
1377         if (capable(CAP_SYS_TTY_CONFIG)) {
1378                 tty_vhangup_self();
1379                 return 0;
1380         }
1381         return -EPERM;
1382 }
1383
1384 /*
1385  * Called when an inode is about to be open.
1386  * We use this to disallow opening large files on 32bit systems if
1387  * the caller didn't specify O_LARGEFILE.  On 64bit systems we force
1388  * on this flag in sys_open.
1389  */
1390 int generic_file_open(struct inode * inode, struct file * filp)
1391 {
1392         if (!(filp->f_flags & O_LARGEFILE) && i_size_read(inode) > MAX_NON_LFS)
1393                 return -EOVERFLOW;
1394         return 0;
1395 }
1396
1397 EXPORT_SYMBOL(generic_file_open);
1398
1399 /*
1400  * This is used by subsystems that don't want seekable
1401  * file descriptors. The function is not supposed to ever fail, the only
1402  * reason it returns an 'int' and not 'void' is so that it can be plugged
1403  * directly into file_operations structure.
1404  */
1405 int nonseekable_open(struct inode *inode, struct file *filp)
1406 {
1407         filp->f_mode &= ~(FMODE_LSEEK | FMODE_PREAD | FMODE_PWRITE);
1408         return 0;
1409 }
1410
1411 EXPORT_SYMBOL(nonseekable_open);
1412
1413 /*
1414  * stream_open is used by subsystems that want stream-like file descriptors.
1415  * Such file descriptors are not seekable and don't have notion of position
1416  * (file.f_pos is always 0 and ppos passed to .read()/.write() is always NULL).
1417  * Contrary to file descriptors of other regular files, .read() and .write()
1418  * can run simultaneously.
1419  *
1420  * stream_open never fails and is marked to return int so that it could be
1421  * directly used as file_operations.open .
1422  */
1423 int stream_open(struct inode *inode, struct file *filp)
1424 {
1425         filp->f_mode &= ~(FMODE_LSEEK | FMODE_PREAD | FMODE_PWRITE | FMODE_ATOMIC_POS);
1426         filp->f_mode |= FMODE_STREAM;
1427         return 0;
1428 }
1429
1430 EXPORT_SYMBOL(stream_open);