fs/open.c

   1 /*
   2  *  linux/fs/open.c
   3  *
   4  *  Copyright (C) 1991, 1992  Linus Torvalds
   5  */
   6
   7 #include <linux/string.h>
   8 #include <linux/mm.h>
   9 #include <linux/file.h>
  10 #include <linux/smp_lock.h>
  11 #include <linux/quotaops.h>
  12 #include <linux/fsnotify.h>
  13 #include <linux/module.h>
  14 #include <linux/slab.h>
  15 #include <linux/tty.h>
  16 #include <linux/namei.h>
  17 #include <linux/backing-dev.h>
  18 #include <linux/capability.h>
  19 #include <linux/security.h>
  20 #include <linux/mount.h>
  21 #include <linux/vfs.h>
  22 #include <linux/fcntl.h>
  23 #include <asm/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
  31 int vfs_statfs(struct dentry *dentry, struct kstatfs *buf)
  32 {
  33         int retval = -ENODEV;
  34
  35         if (dentry) {
  36                 retval = -ENOSYS;
  37                 if (dentry->d_sb->s_op->statfs) {
  38                         memset(buf, 0, sizeof(*buf));
  39                         retval = security_sb_statfs(dentry);
  40                         if (retval)
  41                                 return retval;
  42                         retval = dentry->d_sb->s_op->statfs(dentry, buf);
  43                         if (retval == 0 && buf->f_frsize == 0)
  44                                 buf->f_frsize = buf->f_bsize;
  45                 }
  46         }
  47         return retval;
  48 }
  49
  50 EXPORT_SYMBOL(vfs_statfs);
  51
  52 static int vfs_statfs_native(struct dentry *dentry, struct statfs *buf)
  53 {
  54         struct kstatfs st;
  55         int retval;
  56
  57         retval = vfs_statfs(dentry, &st);
  58         if (retval)
  59                 return retval;
  60
  61         if (sizeof(*buf) == sizeof(st))
  62                 memcpy(buf, &st, sizeof(st));
  63         else {
  64                 if (sizeof buf->f_blocks == 4) {
  65                         if ((st.f_blocks | st.f_bfree | st.f_bavail) &
  66                             0xffffffff00000000ULL)
  67                                 return -EOVERFLOW;
  68                         /*
  69                          * f_files and f_ffree may be -1; it's okay to stuff
  70                          * that into 32 bits
  71                          */
  72                         if (st.f_files != -1 &&
  73                             (st.f_files & 0xffffffff00000000ULL))
  74                                 return -EOVERFLOW;
  75                         if (st.f_ffree != -1 &&
  76                             (st.f_ffree & 0xffffffff00000000ULL))
  77                                 return -EOVERFLOW;
  78                 }
  79
  80                 buf->f_type = st.f_type;
  81                 buf->f_bsize = st.f_bsize;
  82                 buf->f_blocks = st.f_blocks;
  83                 buf->f_bfree = st.f_bfree;
  84                 buf->f_bavail = st.f_bavail;
  85                 buf->f_files = st.f_files;
  86                 buf->f_ffree = st.f_ffree;
  87                 buf->f_fsid = st.f_fsid;
  88                 buf->f_namelen = st.f_namelen;
  89                 buf->f_frsize = st.f_frsize;
  90                 memset(buf->f_spare, 0, sizeof(buf->f_spare));
  91         }
  92         return 0;
  93 }
  94
  95 static int vfs_statfs64(struct dentry *dentry, struct statfs64 *buf)
  96 {
  97         struct kstatfs st;
  98         int retval;
  99
 100         retval = vfs_statfs(dentry, &st);
 101         if (retval)
 102                 return retval;
 103
 104         if (sizeof(*buf) == sizeof(st))
 105                 memcpy(buf, &st, sizeof(st));
 106         else {
 107                 buf->f_type = st.f_type;
 108                 buf->f_bsize = st.f_bsize;
 109                 buf->f_blocks = st.f_blocks;
 110                 buf->f_bfree = st.f_bfree;
 111                 buf->f_bavail = st.f_bavail;
 112                 buf->f_files = st.f_files;
 113                 buf->f_ffree = st.f_ffree;
 114                 buf->f_fsid = st.f_fsid;
 115                 buf->f_namelen = st.f_namelen;
 116                 buf->f_frsize = st.f_frsize;
 117                 memset(buf->f_spare, 0, sizeof(buf->f_spare));
 118         }
 119         return 0;
 120 }
 121
 122 asmlinkage long sys_statfs(const char __user * path, struct statfs __user * buf)
 123 {
 124         struct nameidata nd;
 125         int error;
 126
 127         error = user_path_walk(path, &nd);
 128         if (!error) {
 129                 struct statfs tmp;
 130                 error = vfs_statfs_native(nd.dentry, &tmp);
 131                 if (!error && copy_to_user(buf, &tmp, sizeof(tmp)))
 132                         error = -EFAULT;
 133                 path_release(&nd);
 134         }
 135         return error;
 136 }
 137
 138
 139 asmlinkage long sys_statfs64(const char __user *path, size_t sz, struct statfs64 __user *buf)
 140 {
 141         struct nameidata nd;
 142         long error;
 143
 144         if (sz != sizeof(*buf))
 145                 return -EINVAL;
 146         error = user_path_walk(path, &nd);
 147         if (!error) {
 148                 struct statfs64 tmp;
 149                 error = vfs_statfs64(nd.dentry, &tmp);
 150                 if (!error && copy_to_user(buf, &tmp, sizeof(tmp)))
 151                         error = -EFAULT;
 152                 path_release(&nd);
 153         }
 154         return error;
 155 }
 156
 157
 158 asmlinkage long sys_fstatfs(unsigned int fd, struct statfs __user * buf)
 159 {
 160         struct file * file;
 161         struct statfs tmp;
 162         int error;
 163
 164         error = -EBADF;
 165         file = fget(fd);
 166         if (!file)
 167                 goto out;
 168         error = vfs_statfs_native(file->f_dentry, &tmp);
 169         if (!error && copy_to_user(buf, &tmp, sizeof(tmp)))
 170                 error = -EFAULT;
 171         fput(file);
 172 out:
 173         return error;
 174 }
 175
 176 asmlinkage long sys_fstatfs64(unsigned int fd, size_t sz, struct statfs64 __user *buf)
 177 {
 178         struct file * file;
 179         struct statfs64 tmp;
 180         int error;
 181
 182         if (sz != sizeof(*buf))
 183                 return -EINVAL;
 184
 185         error = -EBADF;
 186         file = fget(fd);
 187         if (!file)
 188                 goto out;
 189         error = vfs_statfs64(file->f_dentry, &tmp);
 190         if (!error && copy_to_user(buf, &tmp, sizeof(tmp)))
 191                 error = -EFAULT;
 192         fput(file);
 193 out:
 194         return error;
 195 }
 196
 197 int do_truncate(struct dentry *dentry, loff_t length, unsigned int time_attrs,
 198         struct file *filp)
 199 {
 200         int err;
 201         struct iattr newattrs;
 202
 203         /* Not pretty: "inode->i_size" shouldn't really be signed. But it is. */
 204         if (length < 0)
 205                 return -EINVAL;
 206
 207         newattrs.ia_size = length;
 208         newattrs.ia_valid = ATTR_SIZE | time_attrs;
 209         if (filp) {
 210                 newattrs.ia_file = filp;
 211                 newattrs.ia_valid |= ATTR_FILE;
 212         }
 213
 214         mutex_lock(&dentry->d_inode->i_mutex);
 215         err = notify_change(dentry, &newattrs);
 216         mutex_unlock(&dentry->d_inode->i_mutex);
 217         return err;
 218 }
 219
 220 static long do_sys_truncate(const char __user * path, loff_t length)
 221 {
 222         struct nameidata nd;
 223         struct inode * inode;
 224         int error;
 225
 226         error = -EINVAL;
 227         if (length < 0) /* sorry, but loff_t says... */
 228                 goto out;
 229
 230         error = user_path_walk(path, &nd);
 231         if (error)
 232                 goto out;
 233         inode = nd.dentry->d_inode;
 234
 235         /* For directories it's -EISDIR, for other non-regulars - -EINVAL */
 236         error = -EISDIR;
 237         if (S_ISDIR(inode->i_mode))
 238                 goto dput_and_out;
 239
 240         error = -EINVAL;
 241         if (!S_ISREG(inode->i_mode))
 242                 goto dput_and_out;
 243
 244         error = vfs_permission(&nd, MAY_WRITE);
 245         if (error)
 246                 goto dput_and_out;
 247
 248         error = -EROFS;
 249         if (IS_RDONLY(inode))
 250                 goto dput_and_out;
 251
 252         error = -EPERM;
 253         if (IS_IMMUTABLE(inode) || IS_APPEND(inode))
 254                 goto dput_and_out;
 255
 256         /*
 257          * Make sure that there are no leases.
 258          */
 259         error = break_lease(inode, FMODE_WRITE);
 260         if (error)
 261                 goto dput_and_out;
 262
 263         error = get_write_access(inode);
 264         if (error)
 265                 goto dput_and_out;
 266
 267         error = locks_verify_truncate(inode, NULL, length);
 268         if (!error) {
 269                 DQUOT_INIT(inode);
 270                 error = do_truncate(nd.dentry, length, 0, NULL);
 271         }
 272         put_write_access(inode);
 273
 274 dput_and_out:
 275         path_release(&nd);
 276 out:
 277         return error;
 278 }
 279
 280 asmlinkage long sys_truncate(const char __user * path, unsigned long length)
 281 {
 282         /* on 32-bit boxen it will cut the range 2^31--2^32-1 off */
 283         return do_sys_truncate(path, (long)length);
 284 }
 285
 286 static long do_sys_ftruncate(unsigned int fd, loff_t length, int small)
 287 {
 288         struct inode * inode;
 289         struct dentry *dentry;
 290         struct file * file;
 291         int error;
 292
 293         error = -EINVAL;
 294         if (length < 0)
 295                 goto out;
 296         error = -EBADF;
 297         file = fget(fd);
 298         if (!file)
 299                 goto out;
 300
 301         /* explicitly opened as large or we are on 64-bit box */
 302         if (file->f_flags & O_LARGEFILE)
 303                 small = 0;
 304
 305         dentry = file->f_dentry;
 306         inode = dentry->d_inode;
 307         error = -EINVAL;
 308         if (!S_ISREG(inode->i_mode) || !(file->f_mode & FMODE_WRITE))
 309                 goto out_putf;
 310
 311         error = -EINVAL;
 312         /* Cannot ftruncate over 2^31 bytes without large file support */
 313         if (small && length > MAX_NON_LFS)
 314                 goto out_putf;
 315
 316         error = -EPERM;
 317         if (IS_APPEND(inode))
 318                 goto out_putf;
 319
 320         error = locks_verify_truncate(inode, file, length);
 321         if (!error)
 322                 error = do_truncate(dentry, length, ATTR_MTIME|ATTR_CTIME, file);
 323 out_putf:
 324         fput(file);
 325 out:
 326         return error;
 327 }
 328
 329 asmlinkage long sys_ftruncate(unsigned int fd, unsigned long length)
 330 {
 331         long ret = do_sys_ftruncate(fd, length, 1);
 332         /* avoid REGPARM breakage on x86: */
 333         prevent_tail_call(ret);
 334         return ret;
 335 }
 336
 337 /* LFS versions of truncate are only needed on 32 bit machines */
 338 #if BITS_PER_LONG == 32
 339 asmlinkage long sys_truncate64(const char __user * path, loff_t length)
 340 {
 341         return do_sys_truncate(path, length);
 342 }
 343
 344 asmlinkage long sys_ftruncate64(unsigned int fd, loff_t length)
 345 {
 346         long ret = do_sys_ftruncate(fd, length, 0);
 347         /* avoid REGPARM breakage on x86: */
 348         prevent_tail_call(ret);
 349         return ret;
 350 }
 351 #endif
 352
 353 /*
 354  * access() needs to use the real uid/gid, not the effective uid/gid.
 355  * We do this by temporarily clearing all FS-related capabilities and
 356  * switching the fsuid/fsgid around to the real ones.
 357  */
 358 asmlinkage long sys_faccessat(int dfd, const char __user *filename, int mode)
 359 {
 360         struct nameidata nd;
 361         int old_fsuid, old_fsgid;
 362         kernel_cap_t old_cap;
 363         int res;
 364
 365         if (mode & ~S_IRWXO)    /* where's F_OK, X_OK, W_OK, R_OK? */
 366                 return -EINVAL;
 367
 368         old_fsuid = current->fsuid;
 369         old_fsgid = current->fsgid;
 370         old_cap = current->cap_effective;
 371
 372         current->fsuid = current->uid;
 373         current->fsgid = current->gid;
 374
 375         /*
 376          * Clear the capabilities if we switch to a non-root user
 377          *
 378          * FIXME: There is a race here against sys_capset.  The
 379          * capabilities can change yet we will restore the old
 380          * value below.  We should hold task_capabilities_lock,
 381          * but we cannot because user_path_walk can sleep.
 382          */
 383         if (current->uid)
 384                 cap_clear(current->cap_effective);
 385         else
 386                 current->cap_effective = current->cap_permitted;
 387
 388         res = __user_walk_fd(dfd, filename, LOOKUP_FOLLOW|LOOKUP_ACCESS, &nd);
 389         if (!res) {
 390                 res = vfs_permission(&nd, mode);
 391                 /* SuS v2 requires we report a read only fs too */
 392                 if(!res && (mode & S_IWOTH) && IS_RDONLY(nd.dentry->d_inode)
 393                    && !special_file(nd.dentry->d_inode->i_mode))
 394                         res = -EROFS;
 395                 path_release(&nd);
 396         }
 397
 398         current->fsuid = old_fsuid;
 399         current->fsgid = old_fsgid;
 400         current->cap_effective = old_cap;
 401
 402         return res;
 403 }
 404
 405 asmlinkage long sys_access(const char __user *filename, int mode)
 406 {
 407         return sys_faccessat(AT_FDCWD, filename, mode);
 408 }
 409
 410 asmlinkage long sys_chdir(const char __user * filename)
 411 {
 412         struct nameidata nd;
 413         int error;
 414
 415         error = __user_walk(filename,
 416                             LOOKUP_FOLLOW|LOOKUP_DIRECTORY|LOOKUP_CHDIR, &nd);
 417         if (error)
 418                 goto out;
 419
 420         error = vfs_permission(&nd, MAY_EXEC);
 421         if (error)
 422                 goto dput_and_out;
 423
 424         set_fs_pwd(current->fs, nd.mnt, nd.dentry);
 425
 426 dput_and_out:
 427         path_release(&nd);
 428 out:
 429         return error;
 430 }
 431
 432 asmlinkage long sys_fchdir(unsigned int fd)
 433 {
 434         struct file *file;
 435         struct dentry *dentry;
 436         struct inode *inode;
 437         struct vfsmount *mnt;
 438         int error;
 439
 440         error = -EBADF;
 441         file = fget(fd);
 442         if (!file)
 443                 goto out;
 444
 445         dentry = file->f_dentry;
 446         mnt = file->f_vfsmnt;
 447         inode = dentry->d_inode;
 448
 449         error = -ENOTDIR;
 450         if (!S_ISDIR(inode->i_mode))
 451                 goto out_putf;
 452
 453         error = file_permission(file, MAY_EXEC);
 454         if (!error)
 455                 set_fs_pwd(current->fs, mnt, dentry);
 456 out_putf:
 457         fput(file);
 458 out:
 459         return error;
 460 }
 461
 462 asmlinkage long sys_chroot(const char __user * filename)
 463 {
 464         struct nameidata nd;
 465         int error;
 466
 467         error = __user_walk(filename, LOOKUP_FOLLOW | LOOKUP_DIRECTORY | LOOKUP_NOALT, &nd);
 468         if (error)
 469                 goto out;
 470
 471         error = vfs_permission(&nd, MAY_EXEC);
 472         if (error)
 473                 goto dput_and_out;
 474
 475         error = -EPERM;
 476         if (!capable(CAP_SYS_CHROOT))
 477                 goto dput_and_out;
 478
 479         set_fs_root(current->fs, nd.mnt, nd.dentry);
 480         set_fs_altroot();
 481         error = 0;
 482 dput_and_out:
 483         path_release(&nd);
 484 out:
 485         return error;
 486 }
 487
 488 asmlinkage long sys_fchmod(unsigned int fd, mode_t mode)
 489 {
 490         struct inode * inode;
 491         struct dentry * dentry;
 492         struct file * file;
 493         int err = -EBADF;
 494         struct iattr newattrs;
 495
 496         file = fget(fd);
 497         if (!file)
 498                 goto out;
 499
 500         dentry = file->f_dentry;
 501         inode = dentry->d_inode;
 502
 503         audit_inode(NULL, inode);
 504
 505         err = -EROFS;
 506         if (IS_RDONLY(inode))
 507                 goto out_putf;
 508         err = -EPERM;
 509         if (IS_IMMUTABLE(inode) || IS_APPEND(inode))
 510                 goto out_putf;
 511         mutex_lock(&inode->i_mutex);
 512         if (mode == (mode_t) -1)
 513                 mode = inode->i_mode;
 514         newattrs.ia_mode = (mode & S_IALLUGO) | (inode->i_mode & ~S_IALLUGO);
 515         newattrs.ia_valid = ATTR_MODE | ATTR_CTIME;
 516         err = notify_change(dentry, &newattrs);
 517         mutex_unlock(&inode->i_mutex);
 518
 519 out_putf:
 520         fput(file);
 521 out:
 522         return err;
 523 }
 524
 525 asmlinkage long sys_fchmodat(int dfd, const char __user *filename,
 526                              mode_t mode)
 527 {
 528         struct nameidata nd;
 529         struct inode * inode;
 530         int error;
 531         struct iattr newattrs;
 532
 533         error = __user_walk_fd(dfd, filename, LOOKUP_FOLLOW, &nd);
 534         if (error)
 535                 goto out;
 536         inode = nd.dentry->d_inode;
 537
 538         error = -EROFS;
 539         if (IS_RDONLY(inode))
 540                 goto dput_and_out;
 541
 542         error = -EPERM;
 543         if (IS_IMMUTABLE(inode) || IS_APPEND(inode))
 544                 goto dput_and_out;
 545
 546         mutex_lock(&inode->i_mutex);
 547         if (mode == (mode_t) -1)
 548                 mode = inode->i_mode;
 549         newattrs.ia_mode = (mode & S_IALLUGO) | (inode->i_mode & ~S_IALLUGO);
 550         newattrs.ia_valid = ATTR_MODE | ATTR_CTIME;
 551         error = notify_change(nd.dentry, &newattrs);
 552         mutex_unlock(&inode->i_mutex);
 553
 554 dput_and_out:
 555         path_release(&nd);
 556 out:
 557         return error;
 558 }
 559
 560 asmlinkage long sys_chmod(const char __user *filename, mode_t mode)
 561 {
 562         return sys_fchmodat(AT_FDCWD, filename, mode);
 563 }
 564
 565 static int chown_common(struct dentry * dentry, uid_t user, gid_t group)
 566 {
 567         struct inode * inode;
 568         int error;
 569         struct iattr newattrs;
 570
 571         error = -ENOENT;
 572         if (!(inode = dentry->d_inode)) {
 573                 printk(KERN_ERR "chown_common: NULL inode\n");
 574                 goto out;
 575         }
 576         error = -EROFS;
 577         if (IS_RDONLY(inode))
 578                 goto out;
 579         error = -EPERM;
 580         if (IS_IMMUTABLE(inode) || IS_APPEND(inode))
 581                 goto out;
 582         newattrs.ia_valid =  ATTR_CTIME;
 583         if (user != (uid_t) -1) {
 584                 newattrs.ia_valid |= ATTR_UID;
 585                 newattrs.ia_uid = user;
 586         }
 587         if (group != (gid_t) -1) {
 588                 newattrs.ia_valid |= ATTR_GID;
 589                 newattrs.ia_gid = group;
 590         }
 591         if (!S_ISDIR(inode->i_mode))
 592                 newattrs.ia_valid |= ATTR_KILL_SUID|ATTR_KILL_SGID;
 593         mutex_lock(&inode->i_mutex);
 594         error = notify_change(dentry, &newattrs);
 595         mutex_unlock(&inode->i_mutex);
 596 out:
 597         return error;
 598 }
 599
 600 asmlinkage long sys_chown(const char __user * filename, uid_t user, gid_t group)
 601 {
 602         struct nameidata nd;
 603         int error;
 604
 605         error = user_path_walk(filename, &nd);
 606         if (!error) {
 607                 error = chown_common(nd.dentry, user, group);
 608                 path_release(&nd);
 609         }
 610         return error;
 611 }
 612
 613 asmlinkage long sys_fchownat(int dfd, const char __user *filename, uid_t user,
 614                              gid_t group, int flag)
 615 {
 616         struct nameidata nd;
 617         int error = -EINVAL;
 618         int follow;
 619
 620         if ((flag & ~AT_SYMLINK_NOFOLLOW) != 0)
 621                 goto out;
 622
 623         follow = (flag & AT_SYMLINK_NOFOLLOW) ? 0 : LOOKUP_FOLLOW;
 624         error = __user_walk_fd(dfd, filename, follow, &nd);
 625         if (!error) {
 626                 error = chown_common(nd.dentry, user, group);
 627                 path_release(&nd);
 628         }
 629 out:
 630         return error;
 631 }
 632
 633 asmlinkage long sys_lchown(const char __user * filename, uid_t user, gid_t group)
 634 {
 635         struct nameidata nd;
 636         int error;
 637
 638         error = user_path_walk_link(filename, &nd);
 639         if (!error) {
 640                 error = chown_common(nd.dentry, user, group);
 641                 path_release(&nd);
 642         }
 643         return error;
 644 }
 645
 646
 647 asmlinkage long sys_fchown(unsigned int fd, uid_t user, gid_t group)
 648 {
 649         struct file * file;
 650         int error = -EBADF;
 651
 652         file = fget(fd);
 653         if (file) {
 654                 struct dentry * dentry;
 655                 dentry = file->f_dentry;
 656                 audit_inode(NULL, dentry->d_inode);
 657                 error = chown_common(dentry, user, group);
 658                 fput(file);
 659         }
 660         return error;
 661 }
 662
 663 static struct file *__dentry_open(struct dentry *dentry, struct vfsmount *mnt,
 664                                         int flags, struct file *f,
 665                                         int (*open)(struct inode *, struct file *))
 666 {
 667         struct inode *inode;
 668         int error;
 669
 670         f->f_flags = flags;
 671         f->f_mode = ((flags+1) & O_ACCMODE) | FMODE_LSEEK |
 672                                 FMODE_PREAD | FMODE_PWRITE;
 673         inode = dentry->d_inode;
 674         if (f->f_mode & FMODE_WRITE) {
 675                 error = get_write_access(inode);
 676                 if (error)
 677                         goto cleanup_file;
 678         }
 679
 680         f->f_mapping = inode->i_mapping;
 681         f->f_dentry = dentry;
 682         f->f_vfsmnt = mnt;
 683         f->f_pos = 0;
 684         f->f_op = fops_get(inode->i_fop);
 685         file_move(f, &inode->i_sb->s_files);
 686
 687         if (!open && f->f_op)
 688                 open = f->f_op->open;
 689         if (open) {
 690                 error = open(inode, f);
 691                 if (error)
 692                         goto cleanup_all;
 693         }
 694
 695         f->f_flags &= ~(O_CREAT | O_EXCL | O_NOCTTY | O_TRUNC);
 696
 697         file_ra_state_init(&f->f_ra, f->f_mapping->host->i_mapping);
 698
 699         /* NB: we're sure to have correct a_ops only after f_op->open */
 700         if (f->f_flags & O_DIRECT) {
 701                 if (!f->f_mapping->a_ops ||
 702                     ((!f->f_mapping->a_ops->direct_IO) &&
 703                     (!f->f_mapping->a_ops->get_xip_page))) {
 704                         fput(f);
 705                         f = ERR_PTR(-EINVAL);
 706                 }
 707         }
 708
 709         return f;
 710
 711 cleanup_all:
 712         fops_put(f->f_op);
 713         if (f->f_mode & FMODE_WRITE)
 714                 put_write_access(inode);
 715         file_kill(f);
 716         f->f_dentry = NULL;
 717         f->f_vfsmnt = NULL;
 718 cleanup_file:
 719         put_filp(f);
 720         dput(dentry);
 721         mntput(mnt);
 722         return ERR_PTR(error);
 723 }
 724
 725 /*
 726  * Note that while the flag value (low two bits) for sys_open means:
 727  *      00 - read-only
 728  *      01 - write-only
 729  *      10 - read-write
 730  *      11 - special
 731  * it is changed into
 732  *      00 - no permissions needed
 733  *      01 - read-permission
 734  *      10 - write-permission
 735  *      11 - read-write
 736  * for the internal routines (ie open_namei()/follow_link() etc). 00 is
 737  * used by symlinks.
 738  */
 739 static struct file *do_filp_open(int dfd, const char *filename, int flags,
 740                                  int mode)
 741 {
 742         int namei_flags, error;
 743         struct nameidata nd;
 744
 745         namei_flags = flags;
 746         if ((namei_flags+1) & O_ACCMODE)
 747                 namei_flags++;
 748
 749         error = open_namei(dfd, filename, namei_flags, mode, &nd);
 750         if (!error)
 751                 return nameidata_to_filp(&nd, flags);
 752
 753         return ERR_PTR(error);
 754 }
 755
 756 struct file *filp_open(const char *filename, int flags, int mode)
 757 {
 758         return do_filp_open(AT_FDCWD, filename, flags, mode);
 759 }
 760 EXPORT_SYMBOL(filp_open);
 761
 762 /**
 763  * lookup_instantiate_filp - instantiates the open intent filp
 764  * @nd: pointer to nameidata
 765  * @dentry: pointer to dentry
 766  * @open: open callback
 767  *
 768  * Helper for filesystems that want to use lookup open intents and pass back
 769  * a fully instantiated struct file to the caller.
 770  * This function is meant to be called from within a filesystem's
 771  * lookup method.
 772  * Beware of calling it for non-regular files! Those ->open methods might block
 773  * (e.g. in fifo_open), leaving you with parent locked (and in case of fifo,
 774  * leading to a deadlock, as nobody can open that fifo anymore, because
 775  * another process to open fifo will block on locked parent when doing lookup).
 776  * Note that in case of error, nd->intent.open.file is destroyed, but the
 777  * path information remains valid.
 778  * If the open callback is set to NULL, then the standard f_op->open()
 779  * filesystem callback is substituted.
 780  */
 781 struct file *lookup_instantiate_filp(struct nameidata *nd, struct dentry *dentry,
 782                 int (*open)(struct inode *, struct file *))
 783 {
 784         if (IS_ERR(nd->intent.open.file))
 785                 goto out;
 786         if (IS_ERR(dentry))
 787                 goto out_err;
 788         nd->intent.open.file = __dentry_open(dget(dentry), mntget(nd->mnt),
 789                                              nd->intent.open.flags - 1,
 790                                              nd->intent.open.file,
 791                                              open);
 792 out:
 793         return nd->intent.open.file;
 794 out_err:
 795         release_open_intent(nd);
 796         nd->intent.open.file = (struct file *)dentry;
 797         goto out;
 798 }
 799 EXPORT_SYMBOL_GPL(lookup_instantiate_filp);
 800
 801 /**
 802  * nameidata_to_filp - convert a nameidata to an open filp.
 803  * @nd: pointer to nameidata
 804  * @flags: open flags
 805  *
 806  * Note that this function destroys the original nameidata
 807  */
 808 struct file *nameidata_to_filp(struct nameidata *nd, int flags)
 809 {
 810         struct file *filp;
 811
 812         /* Pick up the filp from the open intent */
 813         filp = nd->intent.open.file;
 814         /* Has the filesystem initialised the file for us? */
 815         if (filp->f_dentry == NULL)
 816                 filp = __dentry_open(nd->dentry, nd->mnt, flags, filp, NULL);
 817         else
 818                 path_release(nd);
 819         return filp;
 820 }
 821
 822 /*
 823  * dentry_open() will have done dput(dentry) and mntput(mnt) if it returns an
 824  * error.
 825  */
 826 struct file *dentry_open(struct dentry *dentry, struct vfsmount *mnt, int flags)
 827 {
 828         int error;
 829         struct file *f;
 830
 831         error = -ENFILE;
 832         f = get_empty_filp();
 833         if (f == NULL) {
 834                 dput(dentry);
 835                 mntput(mnt);
 836                 return ERR_PTR(error);
 837         }
 838
 839         return __dentry_open(dentry, mnt, flags, f, NULL);
 840 }
 841 EXPORT_SYMBOL(dentry_open);
 842
 843 /*
 844  * Find an empty file descriptor entry, and mark it busy.
 845  */
 846 int get_unused_fd(void)
 847 {
 848         struct files_struct * files = current->files;
 849         int fd, error;
 850         struct fdtable *fdt;
 851
 852         error = -EMFILE;
 853         spin_lock(&files->file_lock);
 854
 855 repeat:
 856         fdt = files_fdtable(files);
 857         fd = find_next_zero_bit(fdt->open_fds->fds_bits,
 858                                 fdt->max_fdset,
 859                                 files->next_fd);
 860
 861         /*
 862          * N.B. For clone tasks sharing a files structure, this test
 863          * will limit the total number of files that can be opened.
 864          */
 865         if (fd >= current->signal->rlim[RLIMIT_NOFILE].rlim_cur)
 866                 goto out;
 867
 868         /* Do we need to expand the fd array or fd set?  */
 869         error = expand_files(files, fd);
 870         if (error < 0)
 871                 goto out;
 872
 873         if (error) {
 874                 /*
 875                  * If we needed to expand the fs array we
 876                  * might have blocked - try again.
 877                  */
 878                 error = -EMFILE;
 879                 goto repeat;
 880         }
 881
 882         FD_SET(fd, fdt->open_fds);
 883         FD_CLR(fd, fdt->close_on_exec);
 884         files->next_fd = fd + 1;
 885 #if 1
 886         /* Sanity check */
 887         if (fdt->fd[fd] != NULL) {
 888                 printk(KERN_WARNING "get_unused_fd: slot %d not NULL!\n", fd);
 889                 fdt->fd[fd] = NULL;
 890         }
 891 #endif
 892         error = fd;
 893
 894 out:
 895         spin_unlock(&files->file_lock);
 896         return error;
 897 }
 898
 899 EXPORT_SYMBOL(get_unused_fd);
 900
 901 static void __put_unused_fd(struct files_struct *files, unsigned int fd)
 902 {
 903         struct fdtable *fdt = files_fdtable(files);
 904         __FD_CLR(fd, fdt->open_fds);
 905         if (fd < files->next_fd)
 906                 files->next_fd = fd;
 907 }
 908
 909 void fastcall put_unused_fd(unsigned int fd)
 910 {
 911         struct files_struct *files = current->files;
 912         spin_lock(&files->file_lock);
 913         __put_unused_fd(files, fd);
 914         spin_unlock(&files->file_lock);
 915 }
 916
 917 EXPORT_SYMBOL(put_unused_fd);
 918
 919 /*
 920  * Install a file pointer in the fd array.
 921  *
 922  * The VFS is full of places where we drop the files lock between
 923  * setting the open_fds bitmap and installing the file in the file
 924  * array.  At any such point, we are vulnerable to a dup2() race
 925  * installing a file in the array before us.  We need to detect this and
 926  * fput() the struct file we are about to overwrite in this case.
 927  *
 928  * It should never happen - if we allow dup2() do it, _really_ bad things
 929  * will follow.
 930  */
 931
 932 void fastcall fd_install(unsigned int fd, struct file * file)
 933 {
 934         struct files_struct *files = current->files;
 935         struct fdtable *fdt;
 936         spin_lock(&files->file_lock);
 937         fdt = files_fdtable(files);
 938         BUG_ON(fdt->fd[fd] != NULL);
 939         rcu_assign_pointer(fdt->fd[fd], file);
 940         spin_unlock(&files->file_lock);
 941 }
 942
 943 EXPORT_SYMBOL(fd_install);
 944
 945 long do_sys_open(int dfd, const char __user *filename, int flags, int mode)
 946 {
 947         char *tmp = getname(filename);
 948         int fd = PTR_ERR(tmp);
 949
 950         if (!IS_ERR(tmp)) {
 951                 fd = get_unused_fd();
 952                 if (fd >= 0) {
 953                         struct file *f = do_filp_open(dfd, tmp, flags, mode);
 954                         if (IS_ERR(f)) {
 955                                 put_unused_fd(fd);
 956                                 fd = PTR_ERR(f);
 957                         } else {
 958                                 fsnotify_open(f->f_dentry);
 959                                 fd_install(fd, f);
 960                         }
 961                 }
 962                 putname(tmp);
 963         }
 964         return fd;
 965 }
 966
 967 asmlinkage long sys_open(const char __user *filename, int flags, int mode)
 968 {
 969         long ret;
 970
 971         if (force_o_largefile())
 972                 flags |= O_LARGEFILE;
 973
 974         ret = do_sys_open(AT_FDCWD, filename, flags, mode);
 975         /* avoid REGPARM breakage on x86: */
 976         prevent_tail_call(ret);
 977         return ret;
 978 }
 979 EXPORT_SYMBOL_GPL(sys_open);
 980
 981 asmlinkage long sys_openat(int dfd, const char __user *filename, int flags,
 982                            int mode)
 983 {
 984         long ret;
 985
 986         if (force_o_largefile())
 987                 flags |= O_LARGEFILE;
 988
 989         ret = do_sys_open(dfd, filename, flags, mode);
 990         /* avoid REGPARM breakage on x86: */
 991         prevent_tail_call(ret);
 992         return ret;
 993 }
 994
 995 #ifndef __alpha__
 996
 997 /*
 998  * For backward compatibility?  Maybe this should be moved
 999  * into arch/i386 instead?
1000  */
1001 asmlinkage long sys_creat(const char __user * pathname, int mode)
1002 {
1003         return sys_open(pathname, O_CREAT | O_WRONLY | O_TRUNC, mode);
1004 }
1005
1006 #endif
1007
1008 /*
1009  * "id" is the POSIX thread ID. We use the
1010  * files pointer for this..
1011  */
1012 int filp_close(struct file *filp, fl_owner_t id)
1013 {
1014         int retval = 0;
1015
1016         if (!file_count(filp)) {
1017                 printk(KERN_ERR "VFS: Close: file count is 0\n");
1018                 return 0;
1019         }
1020
1021         if (filp->f_op && filp->f_op->flush)
1022                 retval = filp->f_op->flush(filp, id);
1023
1024         dnotify_flush(filp, id);
1025         locks_remove_posix(filp, id);
1026         fput(filp);
1027         return retval;
1028 }
1029
1030 EXPORT_SYMBOL(filp_close);
1031
1032 /*
1033  * Careful here! We test whether the file pointer is NULL before
1034  * releasing the fd. This ensures that one clone task can't release
1035  * an fd while another clone is opening it.
1036  */
1037 asmlinkage long sys_close(unsigned int fd)
1038 {
1039         struct file * filp;
1040         struct files_struct *files = current->files;
1041         struct fdtable *fdt;
1042         int retval;
1043
1044         spin_lock(&files->file_lock);
1045         fdt = files_fdtable(files);
1046         if (fd >= fdt->max_fds)
1047                 goto out_unlock;
1048         filp = fdt->fd[fd];
1049         if (!filp)
1050                 goto out_unlock;
1051         rcu_assign_pointer(fdt->fd[fd], NULL);
1052         FD_CLR(fd, fdt->close_on_exec);
1053         __put_unused_fd(files, fd);
1054         spin_unlock(&files->file_lock);
1055         retval = filp_close(filp, files);
1056
1057         /* can't restart close syscall because file table entry was cleared */
1058         if (unlikely(retval == -ERESTARTSYS ||
1059                      retval == -ERESTARTNOINTR ||
1060                      retval == -ERESTARTNOHAND ||
1061                      retval == -ERESTART_RESTARTBLOCK))
1062                 retval = -EINTR;
1063
1064         return retval;
1065
1066 out_unlock:
1067         spin_unlock(&files->file_lock);
1068         return -EBADF;
1069 }
1070
1071 EXPORT_SYMBOL(sys_close);
1072
1073 /*
1074  * This routine simulates a hangup on the tty, to arrange that users
1075  * are given clean terminals at login time.
1076  */
1077 asmlinkage long sys_vhangup(void)
1078 {
1079         if (capable(CAP_SYS_TTY_CONFIG)) {
1080                 tty_vhangup(current->signal->tty);
1081                 return 0;
1082         }
1083         return -EPERM;
1084 }
1085
1086 /*
1087  * Called when an inode is about to be open.
1088  * We use this to disallow opening large files on 32bit systems if
1089  * the caller didn't specify O_LARGEFILE.  On 64bit systems we force
1090  * on this flag in sys_open.
1091  */
1092 int generic_file_open(struct inode * inode, struct file * filp)
1093 {
1094         if (!(filp->f_flags & O_LARGEFILE) && i_size_read(inode) > MAX_NON_LFS)
1095                 return -EFBIG;
1096         return 0;
1097 }
1098
1099 EXPORT_SYMBOL(generic_file_open);
1100
1101 /*
1102  * This is used by subsystems that don't want seekable
1103  * file descriptors
1104  */
1105 int nonseekable_open(struct inode *inode, struct file *filp)
1106 {
1107         filp->f_mode &= ~(FMODE_LSEEK | FMODE_PREAD | FMODE_PWRITE);
1108         return 0;
1109 }
1110
1111 EXPORT_SYMBOL(nonseekable_open);