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
  36 #include "internal.h"
  37
  38 int do_truncate(struct dentry *dentry, loff_t length, unsigned int time_attrs,
  39         struct file *filp)
  40 {
  41         int ret;
  42         struct iattr newattrs;
  43
  44         /* Not pretty: "inode->i_size" shouldn't really be signed. But it is. */
  45         if (length < 0)
  46                 return -EINVAL;
  47
  48         newattrs.ia_size = length;
  49         newattrs.ia_valid = ATTR_SIZE | time_attrs;
  50         if (filp) {
  51                 newattrs.ia_file = filp;
  52                 newattrs.ia_valid |= ATTR_FILE;
  53         }
  54
  55         /* Remove suid, sgid, and file capabilities on truncate too */
  56         ret = dentry_needs_remove_privs(dentry);
  57         if (ret < 0)
  58                 return ret;
  59         if (ret)
  60                 newattrs.ia_valid |= ret | ATTR_FORCE;
  61
  62         inode_lock(dentry->d_inode);
  63         /* Note any delegations or leases have already been broken: */
  64         ret = notify_change(dentry, &newattrs, NULL);
  65         inode_unlock(dentry->d_inode);
  66         return ret;
  67 }
  68
  69 long vfs_truncate(const struct path *path, loff_t length)
  70 {
  71         struct inode *inode;
  72         long error;
  73
  74         inode = path->dentry->d_inode;
  75
  76         /* For directories it's -EISDIR, for other non-regulars - -EINVAL */
  77         if (S_ISDIR(inode->i_mode))
  78                 return -EISDIR;
  79         if (!S_ISREG(inode->i_mode))
  80                 return -EINVAL;
  81
  82         error = mnt_want_write(path->mnt);
  83         if (error)
  84                 goto out;
  85
  86         error = inode_permission(inode, MAY_WRITE);
  87         if (error)
  88                 goto mnt_drop_write_and_out;
  89
  90         error = -EPERM;
  91         if (IS_APPEND(inode))
  92                 goto mnt_drop_write_and_out;
  93
  94         error = get_write_access(inode);
  95         if (error)
  96                 goto mnt_drop_write_and_out;
  97
  98         /*
  99          * Make sure that there are no leases.  get_write_access() protects
 100          * against the truncate racing with a lease-granting setlease().
 101          */
 102         error = break_lease(inode, O_WRONLY);
 103         if (error)
 104                 goto put_write_and_out;
 105
 106         error = locks_verify_truncate(inode, NULL, length);
 107         if (!error)
 108                 error = security_path_truncate(path);
 109         if (!error)
 110                 error = do_truncate(path->dentry, length, 0, NULL);
 111
 112 put_write_and_out:
 113         put_write_access(inode);
 114 mnt_drop_write_and_out:
 115         mnt_drop_write(path->mnt);
 116 out:
 117         return error;
 118 }
 119 EXPORT_SYMBOL_GPL(vfs_truncate);
 120
 121 long do_sys_truncate(const char __user *pathname, loff_t length)
 122 {
 123         unsigned int lookup_flags = LOOKUP_FOLLOW;
 124         struct path path;
 125         int error;
 126
 127         if (length < 0) /* sorry, but loff_t says... */
 128                 return -EINVAL;
 129
 130 retry:
 131         error = user_path_at(AT_FDCWD, pathname, lookup_flags, &path);
 132         if (!error) {
 133                 error = vfs_truncate(&path, length);
 134                 path_put(&path);
 135         }
 136         if (retry_estale(error, lookup_flags)) {
 137                 lookup_flags |= LOOKUP_REVAL;
 138                 goto retry;
 139         }
 140         return error;
 141 }
 142
 143 SYSCALL_DEFINE2(truncate, const char __user *, path, long, length)
 144 {
 145         return do_sys_truncate(path, length);
 146 }
 147
 148 #ifdef CONFIG_COMPAT
 149 COMPAT_SYSCALL_DEFINE2(truncate, const char __user *, path, compat_off_t, length)
 150 {
 151         return do_sys_truncate(path, length);
 152 }
 153 #endif
 154
 155 long do_sys_ftruncate(unsigned int fd, loff_t length, int small)
 156 {
 157         struct inode *inode;
 158         struct dentry *dentry;
 159         struct fd f;
 160         int error;
 161
 162         error = -EINVAL;
 163         if (length < 0)
 164                 goto out;
 165         error = -EBADF;
 166         f = fdget(fd);
 167         if (!f.file)
 168                 goto out;
 169
 170         /* explicitly opened as large or we are on 64-bit box */
 171         if (f.file->f_flags & O_LARGEFILE)
 172                 small = 0;
 173
 174         dentry = f.file->f_path.dentry;
 175         inode = dentry->d_inode;
 176         error = -EINVAL;
 177         if (!S_ISREG(inode->i_mode) || !(f.file->f_mode & FMODE_WRITE))
 178                 goto out_putf;
 179
 180         error = -EINVAL;
 181         /* Cannot ftruncate over 2^31 bytes without large file support */
 182         if (small && length > MAX_NON_LFS)
 183                 goto out_putf;
 184
 185         error = -EPERM;
 186         /* Check IS_APPEND on real upper inode */
 187         if (IS_APPEND(file_inode(f.file)))
 188                 goto out_putf;
 189
 190         sb_start_write(inode->i_sb);
 191         error = locks_verify_truncate(inode, f.file, length);
 192         if (!error)
 193                 error = security_path_truncate(&f.file->f_path);
 194         if (!error)
 195                 error = do_truncate(dentry, length, 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(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 = inode_permission(path.dentry->d_inode, 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 = inode_permission(file_inode(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 = inode_permission(path.dentry->d_inode, 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(path->dentry, &newattrs, &delegated_inode);
 584 out_unlock:
 585         inode_unlock(inode);
 586         if (delegated_inode) {
 587                 error = break_deleg_wait(&delegated_inode);
 588                 if (!error)
 589                         goto retry_deleg;
 590         }
 591         mnt_drop_write(path->mnt);
 592         return error;
 593 }
 594
 595 int vfs_fchmod(struct file *file, umode_t mode)
 596 {
 597         audit_file(file);
 598         return chmod_common(&file->f_path, mode);
 599 }
 600
 601 SYSCALL_DEFINE2(fchmod, unsigned int, fd, umode_t, mode)
 602 {
 603         struct fd f = fdget(fd);
 604         int err = -EBADF;
 605
 606         if (f.file) {
 607                 err = vfs_fchmod(f.file, mode);
 608                 fdput(f);
 609         }
 610         return err;
 611 }
 612
 613 static int do_fchmodat(int dfd, const char __user *filename, umode_t mode)
 614 {
 615         struct path path;
 616         int error;
 617         unsigned int lookup_flags = LOOKUP_FOLLOW;
 618 retry:
 619         error = user_path_at(dfd, filename, lookup_flags, &path);
 620         if (!error) {
 621                 error = chmod_common(&path, mode);
 622                 path_put(&path);
 623                 if (retry_estale(error, lookup_flags)) {
 624                         lookup_flags |= LOOKUP_REVAL;
 625                         goto retry;
 626                 }
 627         }
 628         return error;
 629 }
 630
 631 SYSCALL_DEFINE3(fchmodat, int, dfd, const char __user *, filename,
 632                 umode_t, mode)
 633 {
 634         return do_fchmodat(dfd, filename, mode);
 635 }
 636
 637 SYSCALL_DEFINE2(chmod, const char __user *, filename, umode_t, mode)
 638 {
 639         return do_fchmodat(AT_FDCWD, filename, mode);
 640 }
 641
 642 int chown_common(const struct path *path, uid_t user, gid_t group)
 643 {
 644         struct inode *inode = path->dentry->d_inode;
 645         struct inode *delegated_inode = NULL;
 646         int error;
 647         struct iattr newattrs;
 648         kuid_t uid;
 649         kgid_t gid;
 650
 651         uid = make_kuid(current_user_ns(), user);
 652         gid = make_kgid(current_user_ns(), group);
 653
 654 retry_deleg:
 655         newattrs.ia_valid =  ATTR_CTIME;
 656         if (user != (uid_t) -1) {
 657                 if (!uid_valid(uid))
 658                         return -EINVAL;
 659                 newattrs.ia_valid |= ATTR_UID;
 660                 newattrs.ia_uid = uid;
 661         }
 662         if (group != (gid_t) -1) {
 663                 if (!gid_valid(gid))
 664                         return -EINVAL;
 665                 newattrs.ia_valid |= ATTR_GID;
 666                 newattrs.ia_gid = gid;
 667         }
 668         if (!S_ISDIR(inode->i_mode))
 669                 newattrs.ia_valid |=
 670                         ATTR_KILL_SUID | ATTR_KILL_SGID | ATTR_KILL_PRIV;
 671         inode_lock(inode);
 672         error = security_path_chown(path, uid, gid);
 673         if (!error)
 674                 error = notify_change(path->dentry, &newattrs, &delegated_inode);
 675         inode_unlock(inode);
 676         if (delegated_inode) {
 677                 error = break_deleg_wait(&delegated_inode);
 678                 if (!error)
 679                         goto retry_deleg;
 680         }
 681         return error;
 682 }
 683
 684 int do_fchownat(int dfd, const char __user *filename, uid_t user, gid_t group,
 685                 int flag)
 686 {
 687         struct path path;
 688         int error = -EINVAL;
 689         int lookup_flags;
 690
 691         if ((flag & ~(AT_SYMLINK_NOFOLLOW | AT_EMPTY_PATH)) != 0)
 692                 goto out;
 693
 694         lookup_flags = (flag & AT_SYMLINK_NOFOLLOW) ? 0 : LOOKUP_FOLLOW;
 695         if (flag & AT_EMPTY_PATH)
 696                 lookup_flags |= LOOKUP_EMPTY;
 697 retry:
 698         error = user_path_at(dfd, filename, lookup_flags, &path);
 699         if (error)
 700                 goto out;
 701         error = mnt_want_write(path.mnt);
 702         if (error)
 703                 goto out_release;
 704         error = chown_common(&path, user, group);
 705         mnt_drop_write(path.mnt);
 706 out_release:
 707         path_put(&path);
 708         if (retry_estale(error, lookup_flags)) {
 709                 lookup_flags |= LOOKUP_REVAL;
 710                 goto retry;
 711         }
 712 out:
 713         return error;
 714 }
 715
 716 SYSCALL_DEFINE5(fchownat, int, dfd, const char __user *, filename, uid_t, user,
 717                 gid_t, group, int, flag)
 718 {
 719         return do_fchownat(dfd, filename, user, group, flag);
 720 }
 721
 722 SYSCALL_DEFINE3(chown, const char __user *, filename, uid_t, user, gid_t, group)
 723 {
 724         return do_fchownat(AT_FDCWD, filename, user, group, 0);
 725 }
 726
 727 SYSCALL_DEFINE3(lchown, const char __user *, filename, uid_t, user, gid_t, group)
 728 {
 729         return do_fchownat(AT_FDCWD, filename, user, group,
 730                            AT_SYMLINK_NOFOLLOW);
 731 }
 732
 733 int vfs_fchown(struct file *file, uid_t user, gid_t group)
 734 {
 735         int error;
 736
 737         error = mnt_want_write_file(file);
 738         if (error)
 739                 return error;
 740         audit_file(file);
 741         error = chown_common(&file->f_path, user, group);
 742         mnt_drop_write_file(file);
 743         return error;
 744 }
 745
 746 int ksys_fchown(unsigned int fd, uid_t user, gid_t group)
 747 {
 748         struct fd f = fdget(fd);
 749         int error = -EBADF;
 750
 751         if (f.file) {
 752                 error = vfs_fchown(f.file, user, group);
 753                 fdput(f);
 754         }
 755         return error;
 756 }
 757
 758 SYSCALL_DEFINE3(fchown, unsigned int, fd, uid_t, user, gid_t, group)
 759 {
 760         return ksys_fchown(fd, user, group);
 761 }
 762
 763 static int do_dentry_open(struct file *f,
 764                           struct inode *inode,
 765                           int (*open)(struct inode *, struct file *))
 766 {
 767         static const struct file_operations empty_fops = {};
 768         int error;
 769
 770         path_get(&f->f_path);
 771         f->f_inode = inode;
 772         f->f_mapping = inode->i_mapping;
 773         f->f_wb_err = filemap_sample_wb_err(f->f_mapping);
 774         f->f_sb_err = file_sample_sb_err(f);
 775
 776         if (unlikely(f->f_flags & O_PATH)) {
 777                 f->f_mode = FMODE_PATH | FMODE_OPENED;
 778                 f->f_op = &empty_fops;
 779                 return 0;
 780         }
 781
 782         /* Any file opened for execve()/uselib() has to be a regular file. */
 783         if (unlikely(f->f_flags & FMODE_EXEC && !S_ISREG(inode->i_mode))) {
 784                 error = -EACCES;
 785                 goto cleanup_file;
 786         }
 787
 788         if (f->f_mode & FMODE_WRITE && !special_file(inode->i_mode)) {
 789                 error = get_write_access(inode);
 790                 if (unlikely(error))
 791                         goto cleanup_file;
 792                 error = __mnt_want_write(f->f_path.mnt);
 793                 if (unlikely(error)) {
 794                         put_write_access(inode);
 795                         goto cleanup_file;
 796                 }
 797                 f->f_mode |= FMODE_WRITER;
 798         }
 799
 800         /* POSIX.1-2008/SUSv4 Section XSI 2.9.7 */
 801         if (S_ISREG(inode->i_mode) || S_ISDIR(inode->i_mode))
 802                 f->f_mode |= FMODE_ATOMIC_POS;
 803
 804         f->f_op = fops_get(inode->i_fop);
 805         if (WARN_ON(!f->f_op)) {
 806                 error = -ENODEV;
 807                 goto cleanup_all;
 808         }
 809
 810         error = security_file_open(f);
 811         if (error)
 812                 goto cleanup_all;
 813
 814         error = break_lease(locks_inode(f), f->f_flags);
 815         if (error)
 816                 goto cleanup_all;
 817
 818         /* normally all 3 are set; ->open() can clear them if needed */
 819         f->f_mode |= FMODE_LSEEK | FMODE_PREAD | FMODE_PWRITE;
 820         if (!open)
 821                 open = f->f_op->open;
 822         if (open) {
 823                 error = open(inode, f);
 824                 if (error)
 825                         goto cleanup_all;
 826         }
 827         f->f_mode |= FMODE_OPENED;
 828         if ((f->f_mode & (FMODE_READ | FMODE_WRITE)) == FMODE_READ)
 829                 i_readcount_inc(inode);
 830         if ((f->f_mode & FMODE_READ) &&
 831              likely(f->f_op->read || f->f_op->read_iter))
 832                 f->f_mode |= FMODE_CAN_READ;
 833         if ((f->f_mode & FMODE_WRITE) &&
 834              likely(f->f_op->write || f->f_op->write_iter))
 835                 f->f_mode |= FMODE_CAN_WRITE;
 836
 837         f->f_write_hint = WRITE_LIFE_NOT_SET;
 838         f->f_flags &= ~(O_CREAT | O_EXCL | O_NOCTTY | O_TRUNC);
 839
 840         file_ra_state_init(&f->f_ra, f->f_mapping->host->i_mapping);
 841
 842         /* NB: we're sure to have correct a_ops only after f_op->open */
 843         if (f->f_flags & O_DIRECT) {
 844                 if (!f->f_mapping->a_ops || !f->f_mapping->a_ops->direct_IO)
 845                         return -EINVAL;
 846         }
 847
 848         /*
 849          * XXX: Huge page cache doesn't support writing yet. Drop all page
 850          * cache for this file before processing writes.
 851          */
 852         if ((f->f_mode & FMODE_WRITE) && filemap_nr_thps(inode->i_mapping))
 853                 truncate_pagecache(inode, 0);
 854
 855         return 0;
 856
 857 cleanup_all:
 858         if (WARN_ON_ONCE(error > 0))
 859                 error = -EINVAL;
 860         fops_put(f->f_op);
 861         if (f->f_mode & FMODE_WRITER) {
 862                 put_write_access(inode);
 863                 __mnt_drop_write(f->f_path.mnt);
 864         }
 865 cleanup_file:
 866         path_put(&f->f_path);
 867         f->f_path.mnt = NULL;
 868         f->f_path.dentry = NULL;
 869         f->f_inode = NULL;
 870         return error;
 871 }
 872
 873 /**
 874  * finish_open - finish opening a file
 875  * @file: file pointer
 876  * @dentry: pointer to dentry
 877  * @open: open callback
 878  * @opened: state of open
 879  *
 880  * This can be used to finish opening a file passed to i_op->atomic_open().
 881  *
 882  * If the open callback is set to NULL, then the standard f_op->open()
 883  * filesystem callback is substituted.
 884  *
 885  * NB: the dentry reference is _not_ consumed.  If, for example, the dentry is
 886  * the return value of d_splice_alias(), then the caller needs to perform dput()
 887  * on it after finish_open().
 888  *
 889  * Returns zero on success or -errno if the open failed.
 890  */
 891 int finish_open(struct file *file, struct dentry *dentry,
 892                 int (*open)(struct inode *, struct file *))
 893 {
 894         BUG_ON(file->f_mode & FMODE_OPENED); /* once it's opened, it's opened */
 895
 896         file->f_path.dentry = dentry;
 897         return do_dentry_open(file, d_backing_inode(dentry), open);
 898 }
 899 EXPORT_SYMBOL(finish_open);
 900
 901 /**
 902  * finish_no_open - finish ->atomic_open() without opening the file
 903  *
 904  * @file: file pointer
 905  * @dentry: dentry or NULL (as returned from ->lookup())
 906  *
 907  * This can be used to set the result of a successful lookup in ->atomic_open().
 908  *
 909  * NB: unlike finish_open() this function does consume the dentry reference and
 910  * the caller need not dput() it.
 911  *
 912  * Returns "0" which must be the return value of ->atomic_open() after having
 913  * called this function.
 914  */
 915 int finish_no_open(struct file *file, struct dentry *dentry)
 916 {
 917         file->f_path.dentry = dentry;
 918         return 0;
 919 }
 920 EXPORT_SYMBOL(finish_no_open);
 921
 922 char *file_path(struct file *filp, char *buf, int buflen)
 923 {
 924         return d_path(&filp->f_path, buf, buflen);
 925 }
 926 EXPORT_SYMBOL(file_path);
 927
 928 /**
 929  * vfs_open - open the file at the given path
 930  * @path: path to open
 931  * @file: newly allocated file with f_flag initialized
 932  * @cred: credentials to use
 933  */
 934 int vfs_open(const struct path *path, struct file *file)
 935 {
 936         file->f_path = *path;
 937         return do_dentry_open(file, d_backing_inode(path->dentry), NULL);
 938 }
 939
 940 struct file *dentry_open(const struct path *path, int flags,
 941                          const struct cred *cred)
 942 {
 943         int error;
 944         struct file *f;
 945
 946         validate_creds(cred);
 947
 948         /* We must always pass in a valid mount pointer. */
 949         BUG_ON(!path->mnt);
 950
 951         f = alloc_empty_file(flags, cred);
 952         if (!IS_ERR(f)) {
 953                 error = vfs_open(path, f);
 954                 if (error) {
 955                         fput(f);
 956                         f = ERR_PTR(error);
 957                 }
 958         }
 959         return f;
 960 }
 961 EXPORT_SYMBOL(dentry_open);
 962
 963 struct file *open_with_fake_path(const struct path *path, int flags,
 964                                 struct inode *inode, const struct cred *cred)
 965 {
 966         struct file *f = alloc_empty_file_noaccount(flags, cred);
 967         if (!IS_ERR(f)) {
 968                 int error;
 969
 970                 f->f_path = *path;
 971                 error = do_dentry_open(f, inode, NULL);
 972                 if (error) {
 973                         fput(f);
 974                         f = ERR_PTR(error);
 975                 }
 976         }
 977         return f;
 978 }
 979 EXPORT_SYMBOL(open_with_fake_path);
 980
 981 #define WILL_CREATE(flags)      (flags & (O_CREAT | __O_TMPFILE))
 982 #define O_PATH_FLAGS            (O_DIRECTORY | O_NOFOLLOW | O_PATH | O_CLOEXEC)
 983
 984 inline struct open_how build_open_how(int flags, umode_t mode)
 985 {
 986         struct open_how how = {
 987                 .flags = flags & VALID_OPEN_FLAGS,
 988                 .mode = mode & S_IALLUGO,
 989         };
 990
 991         /* O_PATH beats everything else. */
 992         if (how.flags & O_PATH)
 993                 how.flags &= O_PATH_FLAGS;
 994         /* Modes should only be set for create-like flags. */
 995         if (!WILL_CREATE(how.flags))
 996                 how.mode = 0;
 997         return how;
 998 }
 999
1000 inline int build_open_flags(const struct open_how *how, struct open_flags *op)
1001 {
1002         int flags = how->flags;
1003         int lookup_flags = 0;
1004         int acc_mode = ACC_MODE(flags);
1005
1006         /* Must never be set by userspace */
1007         flags &= ~(FMODE_NONOTIFY | O_CLOEXEC);
1008
1009         /*
1010          * Older syscalls implicitly clear all of the invalid flags or argument
1011          * values before calling build_open_flags(), but openat2(2) checks all
1012          * of its arguments.
1013          */
1014         if (flags & ~VALID_OPEN_FLAGS)
1015                 return -EINVAL;
1016         if (how->resolve & ~VALID_RESOLVE_FLAGS)
1017                 return -EINVAL;
1018
1019         /* Deal with the mode. */
1020         if (WILL_CREATE(flags)) {
1021                 if (how->mode & ~S_IALLUGO)
1022                         return -EINVAL;
1023                 op->mode = how->mode | S_IFREG;
1024         } else {
1025                 if (how->mode != 0)
1026                         return -EINVAL;
1027                 op->mode = 0;
1028         }
1029
1030         /*
1031          * In order to ensure programs get explicit errors when trying to use
1032          * O_TMPFILE on old kernels, O_TMPFILE is implemented such that it
1033          * looks like (O_DIRECTORY|O_RDWR & ~O_CREAT) to old kernels. But we
1034          * have to require userspace to explicitly set it.
1035          */
1036         if (flags & __O_TMPFILE) {
1037                 if ((flags & O_TMPFILE_MASK) != O_TMPFILE)
1038                         return -EINVAL;
1039                 if (!(acc_mode & MAY_WRITE))
1040                         return -EINVAL;
1041         }
1042         if (flags & O_PATH) {
1043                 /* O_PATH only permits certain other flags to be set. */
1044                 if (flags & ~O_PATH_FLAGS)
1045                         return -EINVAL;
1046                 acc_mode = 0;
1047         }
1048
1049         /*
1050          * O_SYNC is implemented as __O_SYNC|O_DSYNC.  As many places only
1051          * check for O_DSYNC if the need any syncing at all we enforce it's
1052          * always set instead of having to deal with possibly weird behaviour
1053          * for malicious applications setting only __O_SYNC.
1054          */
1055         if (flags & __O_SYNC)
1056                 flags |= O_DSYNC;
1057
1058         op->open_flag = flags;
1059
1060         /* O_TRUNC implies we need access checks for write permissions */
1061         if (flags & O_TRUNC)
1062                 acc_mode |= MAY_WRITE;
1063
1064         /* Allow the LSM permission hook to distinguish append
1065            access from general write access. */
1066         if (flags & O_APPEND)
1067                 acc_mode |= MAY_APPEND;
1068
1069         op->acc_mode = acc_mode;
1070
1071         op->intent = flags & O_PATH ? 0 : LOOKUP_OPEN;
1072
1073         if (flags & O_CREAT) {
1074                 op->intent |= LOOKUP_CREATE;
1075                 if (flags & O_EXCL) {
1076                         op->intent |= LOOKUP_EXCL;
1077                         flags |= O_NOFOLLOW;
1078                 }
1079         }
1080
1081         if (flags & O_DIRECTORY)
1082                 lookup_flags |= LOOKUP_DIRECTORY;
1083         if (!(flags & O_NOFOLLOW))
1084                 lookup_flags |= LOOKUP_FOLLOW;
1085
1086         if (how->resolve & RESOLVE_NO_XDEV)
1087                 lookup_flags |= LOOKUP_NO_XDEV;
1088         if (how->resolve & RESOLVE_NO_MAGICLINKS)
1089                 lookup_flags |= LOOKUP_NO_MAGICLINKS;
1090         if (how->resolve & RESOLVE_NO_SYMLINKS)
1091                 lookup_flags |= LOOKUP_NO_SYMLINKS;
1092         if (how->resolve & RESOLVE_BENEATH)
1093                 lookup_flags |= LOOKUP_BENEATH;
1094         if (how->resolve & RESOLVE_IN_ROOT)
1095                 lookup_flags |= LOOKUP_IN_ROOT;
1096
1097         op->lookup_flags = lookup_flags;
1098         return 0;
1099 }
1100
1101 /**
1102  * file_open_name - open file and return file pointer
1103  *
1104  * @name:       struct filename containing path to open
1105  * @flags:      open flags as per the open(2) second argument
1106  * @mode:       mode for the new file if O_CREAT is set, else ignored
1107  *
1108  * This is the helper to open a file from kernelspace if you really
1109  * have to.  But in generally you should not do this, so please move
1110  * along, nothing to see here..
1111  */
1112 struct file *file_open_name(struct filename *name, int flags, umode_t mode)
1113 {
1114         struct open_flags op;
1115         struct open_how how = build_open_how(flags, mode);
1116         int err = build_open_flags(&how, &op);
1117         if (err)
1118                 return ERR_PTR(err);
1119         return do_filp_open(AT_FDCWD, name, &op);
1120 }
1121
1122 /**
1123  * filp_open - open file and return file pointer
1124  *
1125  * @filename:   path to open
1126  * @flags:      open flags as per the open(2) second argument
1127  * @mode:       mode for the new file if O_CREAT is set, else ignored
1128  *
1129  * This is the helper to open a file from kernelspace if you really
1130  * have to.  But in generally you should not do this, so please move
1131  * along, nothing to see here..
1132  */
1133 struct file *filp_open(const char *filename, int flags, umode_t mode)
1134 {
1135         struct filename *name = getname_kernel(filename);
1136         struct file *file = ERR_CAST(name);
1137
1138         if (!IS_ERR(name)) {
1139                 file = file_open_name(name, flags, mode);
1140                 putname(name);
1141         }
1142         return file;
1143 }
1144 EXPORT_SYMBOL(filp_open);
1145
1146 struct file *file_open_root(struct dentry *dentry, struct vfsmount *mnt,
1147                             const char *filename, int flags, umode_t mode)
1148 {
1149         struct open_flags op;
1150         struct open_how how = build_open_how(flags, mode);
1151         int err = build_open_flags(&how, &op);
1152         if (err)
1153                 return ERR_PTR(err);
1154         return do_file_open_root(dentry, mnt, filename, &op);
1155 }
1156 EXPORT_SYMBOL(file_open_root);
1157
1158 static long do_sys_openat2(int dfd, const char __user *filename,
1159                            struct open_how *how)
1160 {
1161         struct open_flags op;
1162         int fd = build_open_flags(how, &op);
1163         struct filename *tmp;
1164
1165         if (fd)
1166                 return fd;
1167
1168         tmp = getname(filename);
1169         if (IS_ERR(tmp))
1170                 return PTR_ERR(tmp);
1171
1172         fd = get_unused_fd_flags(how->flags);
1173         if (fd >= 0) {
1174                 struct file *f = do_filp_open(dfd, tmp, &op);
1175                 if (IS_ERR(f)) {
1176                         put_unused_fd(fd);
1177                         fd = PTR_ERR(f);
1178                 } else {
1179                         fsnotify_open(f);
1180                         fd_install(fd, f);
1181                 }
1182         }
1183         putname(tmp);
1184         return fd;
1185 }
1186
1187 long do_sys_open(int dfd, const char __user *filename, int flags, umode_t mode)
1188 {
1189         struct open_how how = build_open_how(flags, mode);
1190         return do_sys_openat2(dfd, filename, &how);
1191 }
1192
1193
1194 SYSCALL_DEFINE3(open, const char __user *, filename, int, flags, umode_t, mode)
1195 {
1196         if (force_o_largefile())
1197                 flags |= O_LARGEFILE;
1198         return do_sys_open(AT_FDCWD, filename, flags, mode);
1199 }
1200
1201 SYSCALL_DEFINE4(openat, int, dfd, const char __user *, filename, int, flags,
1202                 umode_t, mode)
1203 {
1204         if (force_o_largefile())
1205                 flags |= O_LARGEFILE;
1206         return do_sys_open(dfd, filename, flags, mode);
1207 }
1208
1209 SYSCALL_DEFINE4(openat2, int, dfd, const char __user *, filename,
1210                 struct open_how __user *, how, size_t, usize)
1211 {
1212         int err;
1213         struct open_how tmp;
1214
1215         BUILD_BUG_ON(sizeof(struct open_how) < OPEN_HOW_SIZE_VER0);
1216         BUILD_BUG_ON(sizeof(struct open_how) != OPEN_HOW_SIZE_LATEST);
1217
1218         if (unlikely(usize < OPEN_HOW_SIZE_VER0))
1219                 return -EINVAL;
1220
1221         err = copy_struct_from_user(&tmp, sizeof(tmp), how, usize);
1222         if (err)
1223                 return err;
1224
1225         /* O_LARGEFILE is only allowed for non-O_PATH. */
1226         if (!(tmp.flags & O_PATH) && force_o_largefile())
1227                 tmp.flags |= O_LARGEFILE;
1228
1229         return do_sys_openat2(dfd, filename, &tmp);
1230 }
1231
1232 #ifdef CONFIG_COMPAT
1233 /*
1234  * Exactly like sys_open(), except that it doesn't set the
1235  * O_LARGEFILE flag.
1236  */
1237 COMPAT_SYSCALL_DEFINE3(open, const char __user *, filename, int, flags, umode_t, mode)
1238 {
1239         return do_sys_open(AT_FDCWD, filename, flags, mode);
1240 }
1241
1242 /*
1243  * Exactly like sys_openat(), except that it doesn't set the
1244  * O_LARGEFILE flag.
1245  */
1246 COMPAT_SYSCALL_DEFINE4(openat, int, dfd, const char __user *, filename, int, flags, umode_t, mode)
1247 {
1248         return do_sys_open(dfd, filename, flags, mode);
1249 }
1250 #endif
1251
1252 #ifndef __alpha__
1253
1254 /*
1255  * For backward compatibility?  Maybe this should be moved
1256  * into arch/i386 instead?
1257  */
1258 SYSCALL_DEFINE2(creat, const char __user *, pathname, umode_t, mode)
1259 {
1260         int flags = O_CREAT | O_WRONLY | O_TRUNC;
1261
1262         if (force_o_largefile())
1263                 flags |= O_LARGEFILE;
1264         return do_sys_open(AT_FDCWD, pathname, flags, mode);
1265 }
1266 #endif
1267
1268 /*
1269  * "id" is the POSIX thread ID. We use the
1270  * files pointer for this..
1271  */
1272 int filp_close(struct file *filp, fl_owner_t id)
1273 {
1274         int retval = 0;
1275
1276         if (!file_count(filp)) {
1277                 printk(KERN_ERR "VFS: Close: file count is 0\n");
1278                 return 0;
1279         }
1280
1281         if (filp->f_op->flush)
1282                 retval = filp->f_op->flush(filp, id);
1283
1284         if (likely(!(filp->f_mode & FMODE_PATH))) {
1285                 dnotify_flush(filp, id);
1286                 locks_remove_posix(filp, id);
1287         }
1288         fput(filp);
1289         return retval;
1290 }
1291
1292 EXPORT_SYMBOL(filp_close);
1293
1294 /*
1295  * Careful here! We test whether the file pointer is NULL before
1296  * releasing the fd. This ensures that one clone task can't release
1297  * an fd while another clone is opening it.
1298  */
1299 SYSCALL_DEFINE1(close, unsigned int, fd)
1300 {
1301         int retval = __close_fd(current->files, fd);
1302
1303         /* can't restart close syscall because file table entry was cleared */
1304         if (unlikely(retval == -ERESTARTSYS ||
1305                      retval == -ERESTARTNOINTR ||
1306                      retval == -ERESTARTNOHAND ||
1307                      retval == -ERESTART_RESTARTBLOCK))
1308                 retval = -EINTR;
1309
1310         return retval;
1311 }
1312
1313 /**
1314  * close_range() - Close all file descriptors in a given range.
1315  *
1316  * @fd:     starting file descriptor to close
1317  * @max_fd: last file descriptor to close
1318  * @flags:  reserved for future extensions
1319  *
1320  * This closes a range of file descriptors. All file descriptors
1321  * from @fd up to and including @max_fd are closed.
1322  * Currently, errors to close a given file descriptor are ignored.
1323  */
1324 SYSCALL_DEFINE3(close_range, unsigned int, fd, unsigned int, max_fd,
1325                 unsigned int, flags)
1326 {
1327         return __close_range(fd, max_fd, flags);
1328 }
1329
1330 /*
1331  * This routine simulates a hangup on the tty, to arrange that users
1332  * are given clean terminals at login time.
1333  */
1334 SYSCALL_DEFINE0(vhangup)
1335 {
1336         if (capable(CAP_SYS_TTY_CONFIG)) {
1337                 tty_vhangup_self();
1338                 return 0;
1339         }
1340         return -EPERM;
1341 }
1342
1343 /*
1344  * Called when an inode is about to be open.
1345  * We use this to disallow opening large files on 32bit systems if
1346  * the caller didn't specify O_LARGEFILE.  On 64bit systems we force
1347  * on this flag in sys_open.
1348  */
1349 int generic_file_open(struct inode * inode, struct file * filp)
1350 {
1351         if (!(filp->f_flags & O_LARGEFILE) && i_size_read(inode) > MAX_NON_LFS)
1352                 return -EOVERFLOW;
1353         return 0;
1354 }
1355
1356 EXPORT_SYMBOL(generic_file_open);
1357
1358 /*
1359  * This is used by subsystems that don't want seekable
1360  * file descriptors. The function is not supposed to ever fail, the only
1361  * reason it returns an 'int' and not 'void' is so that it can be plugged
1362  * directly into file_operations structure.
1363  */
1364 int nonseekable_open(struct inode *inode, struct file *filp)
1365 {
1366         filp->f_mode &= ~(FMODE_LSEEK | FMODE_PREAD | FMODE_PWRITE);
1367         return 0;
1368 }
1369
1370 EXPORT_SYMBOL(nonseekable_open);
1371
1372 /*
1373  * stream_open is used by subsystems that want stream-like file descriptors.
1374  * Such file descriptors are not seekable and don't have notion of position
1375  * (file.f_pos is always 0 and ppos passed to .read()/.write() is always NULL).
1376  * Contrary to file descriptors of other regular files, .read() and .write()
1377  * can run simultaneously.
1378  *
1379  * stream_open never fails and is marked to return int so that it could be
1380  * directly used as file_operations.open .
1381  */
1382 int stream_open(struct inode *inode, struct file *filp)
1383 {
1384         filp->f_mode &= ~(FMODE_LSEEK | FMODE_PREAD | FMODE_PWRITE | FMODE_ATOMIC_POS);
1385         filp->f_mode |= FMODE_STREAM;
1386         return 0;
1387 }
1388
1389 EXPORT_SYMBOL(stream_open);