4 * Copyright (C) 1991, 1992 Linus Torvalds
7 #include <linux/string.h>
9 #include <linux/file.h>
10 #include <linux/fdtable.h>
11 #include <linux/fsnotify.h>
12 #include <linux/module.h>
13 #include <linux/tty.h>
14 #include <linux/namei.h>
15 #include <linux/backing-dev.h>
16 #include <linux/capability.h>
17 #include <linux/securebits.h>
18 #include <linux/security.h>
19 #include <linux/mount.h>
20 #include <linux/fcntl.h>
21 #include <linux/slab.h>
22 #include <asm/uaccess.h>
24 #include <linux/personality.h>
25 #include <linux/pagemap.h>
26 #include <linux/syscalls.h>
27 #include <linux/rcupdate.h>
28 #include <linux/audit.h>
29 #include <linux/falloc.h>
30 #include <linux/fs_struct.h>
31 #include <linux/ima.h>
32 #include <linux/dnotify.h>
33 #include <linux/compat.h>
37 int do_truncate(struct dentry *dentry, loff_t length, unsigned int time_attrs,
41 struct iattr newattrs;
43 /* Not pretty: "inode->i_size" shouldn't really be signed. But it is. */
47 newattrs.ia_size = length;
48 newattrs.ia_valid = ATTR_SIZE | time_attrs;
50 newattrs.ia_file = filp;
51 newattrs.ia_valid |= ATTR_FILE;
54 /* Remove suid, sgid, and file capabilities on truncate too */
55 ret = dentry_needs_remove_privs(dentry);
59 newattrs.ia_valid |= ret | ATTR_FORCE;
61 mutex_lock(&dentry->d_inode->i_mutex);
62 /* Note any delegations or leases have already been broken: */
63 ret = notify_change(dentry, &newattrs, NULL);
64 mutex_unlock(&dentry->d_inode->i_mutex);
68 long vfs_truncate(struct path *path, loff_t length)
74 inode = path->dentry->d_inode;
77 /* For directories it's -EISDIR, for other non-regulars - -EINVAL */
78 if (S_ISDIR(inode->i_mode))
80 if (!S_ISREG(inode->i_mode))
83 error = mnt_want_write(path->mnt);
87 error = inode_permission2(mnt, inode, MAY_WRITE);
89 goto mnt_drop_write_and_out;
93 goto mnt_drop_write_and_out;
95 error = get_write_access(inode);
97 goto mnt_drop_write_and_out;
100 * Make sure that there are no leases. get_write_access() protects
101 * against the truncate racing with a lease-granting setlease().
103 error = break_lease(inode, O_WRONLY);
105 goto put_write_and_out;
107 error = locks_verify_truncate(inode, NULL, length);
109 error = security_path_truncate(path);
111 error = do_truncate(path->dentry, length, 0, NULL);
114 put_write_access(inode);
115 mnt_drop_write_and_out:
116 mnt_drop_write(path->mnt);
120 EXPORT_SYMBOL_GPL(vfs_truncate);
122 static long do_sys_truncate(const char __user *pathname, loff_t length)
124 unsigned int lookup_flags = LOOKUP_FOLLOW;
128 if (length < 0) /* sorry, but loff_t says... */
132 error = user_path_at(AT_FDCWD, pathname, lookup_flags, &path);
134 error = vfs_truncate(&path, length);
137 if (retry_estale(error, lookup_flags)) {
138 lookup_flags |= LOOKUP_REVAL;
144 SYSCALL_DEFINE2(truncate, const char __user *, path, long, length)
146 return do_sys_truncate(path, length);
150 COMPAT_SYSCALL_DEFINE2(truncate, const char __user *, path, compat_off_t, length)
152 return do_sys_truncate(path, length);
156 static long do_sys_ftruncate(unsigned int fd, loff_t length, int small)
159 struct dentry *dentry;
171 /* explicitly opened as large or we are on 64-bit box */
172 if (f.file->f_flags & O_LARGEFILE)
175 dentry = f.file->f_path.dentry;
176 inode = dentry->d_inode;
178 if (!S_ISREG(inode->i_mode) || !(f.file->f_mode & FMODE_WRITE))
182 /* Cannot ftruncate over 2^31 bytes without large file support */
183 if (small && length > MAX_NON_LFS)
187 if (IS_APPEND(inode))
190 sb_start_write(inode->i_sb);
191 error = locks_verify_truncate(inode, f.file, length);
193 error = security_path_truncate(&f.file->f_path);
195 error = do_truncate(dentry, length, ATTR_MTIME|ATTR_CTIME, f.file);
196 sb_end_write(inode->i_sb);
203 SYSCALL_DEFINE2(ftruncate, unsigned int, fd, unsigned long, length)
205 return do_sys_ftruncate(fd, length, 1);
209 COMPAT_SYSCALL_DEFINE2(ftruncate, unsigned int, fd, compat_ulong_t, length)
211 return do_sys_ftruncate(fd, length, 1);
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)
219 return do_sys_truncate(path, length);
222 SYSCALL_DEFINE2(ftruncate64, unsigned int, fd, loff_t, length)
224 return do_sys_ftruncate(fd, length, 0);
226 #endif /* BITS_PER_LONG == 32 */
229 int vfs_fallocate(struct file *file, int mode, loff_t offset, loff_t len)
231 struct inode *inode = file_inode(file);
234 if (offset < 0 || len <= 0)
237 /* Return error if mode is not supported */
238 if (mode & ~FALLOC_FL_SUPPORTED_MASK)
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))
246 /* Punch hole must have keep size set */
247 if ((mode & FALLOC_FL_PUNCH_HOLE) &&
248 !(mode & FALLOC_FL_KEEP_SIZE))
251 /* Collapse range should only be used exclusively. */
252 if ((mode & FALLOC_FL_COLLAPSE_RANGE) &&
253 (mode & ~FALLOC_FL_COLLAPSE_RANGE))
256 /* Insert range should only be used exclusively. */
257 if ((mode & FALLOC_FL_INSERT_RANGE) &&
258 (mode & ~FALLOC_FL_INSERT_RANGE))
261 if (!(file->f_mode & FMODE_WRITE))
265 * We can only allow pure fallocate on append only files
267 if ((mode & ~FALLOC_FL_KEEP_SIZE) && IS_APPEND(inode))
270 if (IS_IMMUTABLE(inode))
274 * We cannot allow any fallocate operation on an active swapfile
276 if (IS_SWAPFILE(inode))
280 * Revalidate the write permissions, in case security policy has
281 * changed since the files were opened.
283 ret = security_file_permission(file, MAY_WRITE);
287 if (S_ISFIFO(inode->i_mode))
291 * Let individual file system decide if it supports preallocation
292 * for directories or not.
294 if (!S_ISREG(inode->i_mode) && !S_ISDIR(inode->i_mode))
297 /* Check for wrap through zero too */
298 if (((offset + len) > inode->i_sb->s_maxbytes) || ((offset + len) < 0))
301 if (!file->f_op->fallocate)
304 sb_start_write(inode->i_sb);
305 ret = file->f_op->fallocate(file, mode, offset, len);
308 * Create inotify and fanotify events.
310 * To keep the logic simple always create events if fallocate succeeds.
311 * This implies that events are even created if the file size remains
312 * unchanged, e.g. when using flag FALLOC_FL_KEEP_SIZE.
315 fsnotify_modify(file);
317 sb_end_write(inode->i_sb);
320 EXPORT_SYMBOL_GPL(vfs_fallocate);
322 SYSCALL_DEFINE4(fallocate, int, fd, int, mode, loff_t, offset, loff_t, len)
324 struct fd f = fdget(fd);
328 error = vfs_fallocate(f.file, mode, offset, len);
335 * access() needs to use the real uid/gid, not the effective uid/gid.
336 * We do this by temporarily clearing all FS-related capabilities and
337 * switching the fsuid/fsgid around to the real ones.
339 SYSCALL_DEFINE3(faccessat, int, dfd, const char __user *, filename, int, mode)
341 const struct cred *old_cred;
342 struct cred *override_cred;
345 struct vfsmount *mnt;
347 unsigned int lookup_flags = LOOKUP_FOLLOW;
349 if (mode & ~S_IRWXO) /* where's F_OK, X_OK, W_OK, R_OK? */
352 override_cred = prepare_creds();
356 override_cred->fsuid = override_cred->uid;
357 override_cred->fsgid = override_cred->gid;
359 if (!issecure(SECURE_NO_SETUID_FIXUP)) {
360 /* Clear the capabilities if we switch to a non-root user */
361 kuid_t root_uid = make_kuid(override_cred->user_ns, 0);
362 if (!uid_eq(override_cred->uid, root_uid))
363 cap_clear(override_cred->cap_effective);
365 override_cred->cap_effective =
366 override_cred->cap_permitted;
369 old_cred = override_creds(override_cred);
371 res = user_path_at(dfd, filename, lookup_flags, &path);
375 inode = d_backing_inode(path.dentry);
378 if ((mode & MAY_EXEC) && S_ISREG(inode->i_mode)) {
380 * MAY_EXEC on regular files is denied if the fs is mounted
381 * with the "noexec" flag.
384 if (path_noexec(&path))
385 goto out_path_release;
388 res = inode_permission2(mnt, inode, mode | MAY_ACCESS);
389 /* SuS v2 requires we report a read only fs too */
390 if (res || !(mode & S_IWOTH) || special_file(inode->i_mode))
391 goto out_path_release;
393 * This is a rare case where using __mnt_is_readonly()
394 * is OK without a mnt_want/drop_write() pair. Since
395 * no actual write to the fs is performed here, we do
396 * not need to telegraph to that to anyone.
398 * By doing this, we accept that this access is
399 * inherently racy and know that the fs may change
400 * state before we even see this result.
402 if (__mnt_is_readonly(path.mnt))
407 if (retry_estale(res, lookup_flags)) {
408 lookup_flags |= LOOKUP_REVAL;
412 revert_creds(old_cred);
413 put_cred(override_cred);
417 SYSCALL_DEFINE2(access, const char __user *, filename, int, mode)
419 return sys_faccessat(AT_FDCWD, filename, mode);
422 SYSCALL_DEFINE1(chdir, const char __user *, filename)
426 unsigned int lookup_flags = LOOKUP_FOLLOW | LOOKUP_DIRECTORY;
428 error = user_path_at(AT_FDCWD, filename, lookup_flags, &path);
432 error = inode_permission2(path.mnt, path.dentry->d_inode, MAY_EXEC | MAY_CHDIR);
436 set_fs_pwd(current->fs, &path);
440 if (retry_estale(error, lookup_flags)) {
441 lookup_flags |= LOOKUP_REVAL;
448 SYSCALL_DEFINE1(fchdir, unsigned int, fd)
450 struct fd f = fdget_raw(fd);
452 struct vfsmount *mnt;
459 inode = file_inode(f.file);
460 mnt = f.file->f_path.mnt;
463 if (!S_ISDIR(inode->i_mode))
466 error = inode_permission2(mnt, inode, MAY_EXEC | MAY_CHDIR);
468 set_fs_pwd(current->fs, &f.file->f_path);
475 SYSCALL_DEFINE1(chroot, const char __user *, filename)
479 unsigned int lookup_flags = LOOKUP_FOLLOW | LOOKUP_DIRECTORY;
481 error = user_path_at(AT_FDCWD, filename, lookup_flags, &path);
485 error = inode_permission2(path.mnt, path.dentry->d_inode, MAY_EXEC | MAY_CHDIR);
490 if (!ns_capable(current_user_ns(), CAP_SYS_CHROOT))
492 error = security_path_chroot(&path);
496 set_fs_root(current->fs, &path);
500 if (retry_estale(error, lookup_flags)) {
501 lookup_flags |= LOOKUP_REVAL;
508 static int chmod_common(struct path *path, umode_t mode)
510 struct inode *inode = path->dentry->d_inode;
511 struct inode *delegated_inode = NULL;
512 struct iattr newattrs;
515 error = mnt_want_write(path->mnt);
519 mutex_lock(&inode->i_mutex);
520 error = security_path_chmod(path, mode);
523 newattrs.ia_mode = (mode & S_IALLUGO) | (inode->i_mode & ~S_IALLUGO);
524 newattrs.ia_valid = ATTR_MODE | ATTR_CTIME;
525 error = notify_change(path->dentry, &newattrs, &delegated_inode);
527 mutex_unlock(&inode->i_mutex);
528 if (delegated_inode) {
529 error = break_deleg_wait(&delegated_inode);
533 mnt_drop_write(path->mnt);
537 SYSCALL_DEFINE2(fchmod, unsigned int, fd, umode_t, mode)
539 struct fd f = fdget(fd);
544 err = chmod_common(&f.file->f_path, mode);
550 SYSCALL_DEFINE3(fchmodat, int, dfd, const char __user *, filename, umode_t, mode)
554 unsigned int lookup_flags = LOOKUP_FOLLOW;
556 error = user_path_at(dfd, filename, lookup_flags, &path);
558 error = chmod_common(&path, mode);
560 if (retry_estale(error, lookup_flags)) {
561 lookup_flags |= LOOKUP_REVAL;
568 SYSCALL_DEFINE2(chmod, const char __user *, filename, umode_t, mode)
570 return sys_fchmodat(AT_FDCWD, filename, mode);
573 static int chown_common(struct path *path, uid_t user, gid_t group)
575 struct inode *inode = path->dentry->d_inode;
576 struct inode *delegated_inode = NULL;
578 struct iattr newattrs;
582 uid = make_kuid(current_user_ns(), user);
583 gid = make_kgid(current_user_ns(), group);
586 newattrs.ia_valid = ATTR_CTIME;
587 if (user != (uid_t) -1) {
590 newattrs.ia_valid |= ATTR_UID;
591 newattrs.ia_uid = uid;
593 if (group != (gid_t) -1) {
596 newattrs.ia_valid |= ATTR_GID;
597 newattrs.ia_gid = gid;
599 if (!S_ISDIR(inode->i_mode))
601 ATTR_KILL_SUID | ATTR_KILL_SGID | ATTR_KILL_PRIV;
602 mutex_lock(&inode->i_mutex);
603 error = security_path_chown(path, uid, gid);
605 error = notify_change(path->dentry, &newattrs, &delegated_inode);
606 mutex_unlock(&inode->i_mutex);
607 if (delegated_inode) {
608 error = break_deleg_wait(&delegated_inode);
615 SYSCALL_DEFINE5(fchownat, int, dfd, const char __user *, filename, uid_t, user,
616 gid_t, group, int, flag)
622 if ((flag & ~(AT_SYMLINK_NOFOLLOW | AT_EMPTY_PATH)) != 0)
625 lookup_flags = (flag & AT_SYMLINK_NOFOLLOW) ? 0 : LOOKUP_FOLLOW;
626 if (flag & AT_EMPTY_PATH)
627 lookup_flags |= LOOKUP_EMPTY;
629 error = user_path_at(dfd, filename, lookup_flags, &path);
632 error = mnt_want_write(path.mnt);
635 error = chown_common(&path, user, group);
636 mnt_drop_write(path.mnt);
639 if (retry_estale(error, lookup_flags)) {
640 lookup_flags |= LOOKUP_REVAL;
647 SYSCALL_DEFINE3(chown, const char __user *, filename, uid_t, user, gid_t, group)
649 return sys_fchownat(AT_FDCWD, filename, user, group, 0);
652 SYSCALL_DEFINE3(lchown, const char __user *, filename, uid_t, user, gid_t, group)
654 return sys_fchownat(AT_FDCWD, filename, user, group,
655 AT_SYMLINK_NOFOLLOW);
658 SYSCALL_DEFINE3(fchown, unsigned int, fd, uid_t, user, gid_t, group)
660 struct fd f = fdget(fd);
666 error = mnt_want_write_file(f.file);
670 error = chown_common(&f.file->f_path, user, group);
671 mnt_drop_write_file(f.file);
678 int open_check_o_direct(struct file *f)
680 /* NB: we're sure to have correct a_ops only after f_op->open */
681 if (f->f_flags & O_DIRECT) {
682 if (!f->f_mapping->a_ops || !f->f_mapping->a_ops->direct_IO)
688 static int do_dentry_open(struct file *f,
690 int (*open)(struct inode *, struct file *),
691 const struct cred *cred)
693 static const struct file_operations empty_fops = {};
696 f->f_mode = OPEN_FMODE(f->f_flags) | FMODE_LSEEK |
697 FMODE_PREAD | FMODE_PWRITE;
699 path_get(&f->f_path);
701 f->f_mapping = inode->i_mapping;
703 if (unlikely(f->f_flags & O_PATH)) {
704 f->f_mode = FMODE_PATH;
705 f->f_op = &empty_fops;
709 if (f->f_mode & FMODE_WRITE && !special_file(inode->i_mode)) {
710 error = get_write_access(inode);
713 error = __mnt_want_write(f->f_path.mnt);
714 if (unlikely(error)) {
715 put_write_access(inode);
718 f->f_mode |= FMODE_WRITER;
721 /* POSIX.1-2008/SUSv4 Section XSI 2.9.7 */
722 if (S_ISREG(inode->i_mode))
723 f->f_mode |= FMODE_ATOMIC_POS;
725 f->f_op = fops_get(inode->i_fop);
726 if (unlikely(WARN_ON(!f->f_op))) {
731 error = security_file_open(f, cred);
735 error = break_lease(inode, f->f_flags);
740 open = f->f_op->open;
742 error = open(inode, f);
746 if ((f->f_mode & (FMODE_READ | FMODE_WRITE)) == FMODE_READ)
747 i_readcount_inc(inode);
748 if ((f->f_mode & FMODE_READ) &&
749 likely(f->f_op->read || f->f_op->read_iter))
750 f->f_mode |= FMODE_CAN_READ;
751 if ((f->f_mode & FMODE_WRITE) &&
752 likely(f->f_op->write || f->f_op->write_iter))
753 f->f_mode |= FMODE_CAN_WRITE;
755 f->f_flags &= ~(O_CREAT | O_EXCL | O_NOCTTY | O_TRUNC);
757 file_ra_state_init(&f->f_ra, f->f_mapping->host->i_mapping);
763 if (f->f_mode & FMODE_WRITER) {
764 put_write_access(inode);
765 __mnt_drop_write(f->f_path.mnt);
768 path_put(&f->f_path);
769 f->f_path.mnt = NULL;
770 f->f_path.dentry = NULL;
776 * finish_open - finish opening a file
777 * @file: file pointer
778 * @dentry: pointer to dentry
779 * @open: open callback
780 * @opened: state of open
782 * This can be used to finish opening a file passed to i_op->atomic_open().
784 * If the open callback is set to NULL, then the standard f_op->open()
785 * filesystem callback is substituted.
787 * NB: the dentry reference is _not_ consumed. If, for example, the dentry is
788 * the return value of d_splice_alias(), then the caller needs to perform dput()
789 * on it after finish_open().
791 * On successful return @file is a fully instantiated open file. After this, if
792 * an error occurs in ->atomic_open(), it needs to clean up with fput().
794 * Returns zero on success or -errno if the open failed.
796 int finish_open(struct file *file, struct dentry *dentry,
797 int (*open)(struct inode *, struct file *),
801 BUG_ON(*opened & FILE_OPENED); /* once it's opened, it's opened */
803 file->f_path.dentry = dentry;
804 error = do_dentry_open(file, d_backing_inode(dentry), open,
807 *opened |= FILE_OPENED;
811 EXPORT_SYMBOL(finish_open);
814 * finish_no_open - finish ->atomic_open() without opening the file
816 * @file: file pointer
817 * @dentry: dentry or NULL (as returned from ->lookup())
819 * This can be used to set the result of a successful lookup in ->atomic_open().
821 * NB: unlike finish_open() this function does consume the dentry reference and
822 * the caller need not dput() it.
824 * Returns "1" which must be the return value of ->atomic_open() after having
825 * called this function.
827 int finish_no_open(struct file *file, struct dentry *dentry)
829 file->f_path.dentry = dentry;
832 EXPORT_SYMBOL(finish_no_open);
834 char *file_path(struct file *filp, char *buf, int buflen)
836 return d_path(&filp->f_path, buf, buflen);
838 EXPORT_SYMBOL(file_path);
841 * vfs_open - open the file at the given path
842 * @path: path to open
843 * @file: newly allocated file with f_flag initialized
844 * @cred: credentials to use
846 int vfs_open(const struct path *path, struct file *file,
847 const struct cred *cred)
849 struct inode *inode = vfs_select_inode(path->dentry, file->f_flags);
852 return PTR_ERR(inode);
854 file->f_path = *path;
855 return do_dentry_open(file, inode, NULL, cred);
858 struct file *dentry_open(const struct path *path, int flags,
859 const struct cred *cred)
864 validate_creds(cred);
866 /* We must always pass in a valid mount pointer. */
869 f = get_empty_filp();
872 error = vfs_open(path, f, cred);
874 /* from now on we need fput() to dispose of f */
875 error = open_check_o_direct(f);
887 EXPORT_SYMBOL(dentry_open);
889 static inline int build_open_flags(int flags, umode_t mode, struct open_flags *op)
891 int lookup_flags = 0;
894 if (flags & (O_CREAT | __O_TMPFILE))
895 op->mode = (mode & S_IALLUGO) | S_IFREG;
899 /* Must never be set by userspace */
900 flags &= ~FMODE_NONOTIFY & ~O_CLOEXEC;
903 * O_SYNC is implemented as __O_SYNC|O_DSYNC. As many places only
904 * check for O_DSYNC if the need any syncing at all we enforce it's
905 * always set instead of having to deal with possibly weird behaviour
906 * for malicious applications setting only __O_SYNC.
908 if (flags & __O_SYNC)
911 if (flags & __O_TMPFILE) {
912 if ((flags & O_TMPFILE_MASK) != O_TMPFILE)
914 acc_mode = MAY_OPEN | ACC_MODE(flags);
915 if (!(acc_mode & MAY_WRITE))
917 } else if (flags & O_PATH) {
919 * If we have O_PATH in the open flag. Then we
920 * cannot have anything other than the below set of flags
922 flags &= O_DIRECTORY | O_NOFOLLOW | O_PATH;
925 acc_mode = MAY_OPEN | ACC_MODE(flags);
928 op->open_flag = flags;
930 /* O_TRUNC implies we need access checks for write permissions */
932 acc_mode |= MAY_WRITE;
934 /* Allow the LSM permission hook to distinguish append
935 access from general write access. */
936 if (flags & O_APPEND)
937 acc_mode |= MAY_APPEND;
939 op->acc_mode = acc_mode;
941 op->intent = flags & O_PATH ? 0 : LOOKUP_OPEN;
943 if (flags & O_CREAT) {
944 op->intent |= LOOKUP_CREATE;
946 op->intent |= LOOKUP_EXCL;
949 if (flags & O_DIRECTORY)
950 lookup_flags |= LOOKUP_DIRECTORY;
951 if (!(flags & O_NOFOLLOW))
952 lookup_flags |= LOOKUP_FOLLOW;
953 op->lookup_flags = lookup_flags;
958 * file_open_name - open file and return file pointer
960 * @name: struct filename containing path to open
961 * @flags: open flags as per the open(2) second argument
962 * @mode: mode for the new file if O_CREAT is set, else ignored
964 * This is the helper to open a file from kernelspace if you really
965 * have to. But in generally you should not do this, so please move
966 * along, nothing to see here..
968 struct file *file_open_name(struct filename *name, int flags, umode_t mode)
970 struct open_flags op;
971 int err = build_open_flags(flags, mode, &op);
972 return err ? ERR_PTR(err) : do_filp_open(AT_FDCWD, name, &op);
976 * filp_open - open file and return file pointer
978 * @filename: path to open
979 * @flags: open flags as per the open(2) second argument
980 * @mode: mode for the new file if O_CREAT is set, else ignored
982 * This is the helper to open a file from kernelspace if you really
983 * have to. But in generally you should not do this, so please move
984 * along, nothing to see here..
986 struct file *filp_open(const char *filename, int flags, umode_t mode)
988 struct filename *name = getname_kernel(filename);
989 struct file *file = ERR_CAST(name);
992 file = file_open_name(name, flags, mode);
997 EXPORT_SYMBOL(filp_open);
999 struct file *file_open_root(struct dentry *dentry, struct vfsmount *mnt,
1000 const char *filename, int flags, umode_t mode)
1002 struct open_flags op;
1003 int err = build_open_flags(flags, mode, &op);
1005 return ERR_PTR(err);
1006 return do_file_open_root(dentry, mnt, filename, &op);
1008 EXPORT_SYMBOL(file_open_root);
1010 long do_sys_open(int dfd, const char __user *filename, int flags, umode_t mode)
1012 struct open_flags op;
1013 int fd = build_open_flags(flags, mode, &op);
1014 struct filename *tmp;
1019 tmp = getname(filename);
1021 return PTR_ERR(tmp);
1023 fd = get_unused_fd_flags(flags);
1025 struct file *f = do_filp_open(dfd, tmp, &op);
1038 SYSCALL_DEFINE3(open, const char __user *, filename, int, flags, umode_t, mode)
1040 if (force_o_largefile())
1041 flags |= O_LARGEFILE;
1043 return do_sys_open(AT_FDCWD, filename, flags, mode);
1046 SYSCALL_DEFINE4(openat, int, dfd, const char __user *, filename, int, flags,
1049 if (force_o_largefile())
1050 flags |= O_LARGEFILE;
1052 return do_sys_open(dfd, filename, flags, mode);
1058 * For backward compatibility? Maybe this should be moved
1059 * into arch/i386 instead?
1061 SYSCALL_DEFINE2(creat, const char __user *, pathname, umode_t, mode)
1063 return sys_open(pathname, O_CREAT | O_WRONLY | O_TRUNC, mode);
1069 * "id" is the POSIX thread ID. We use the
1070 * files pointer for this..
1072 int filp_close(struct file *filp, fl_owner_t id)
1076 if (!file_count(filp)) {
1077 printk(KERN_ERR "VFS: Close: file count is 0\n");
1081 if (filp->f_op->flush)
1082 retval = filp->f_op->flush(filp, id);
1084 if (likely(!(filp->f_mode & FMODE_PATH))) {
1085 dnotify_flush(filp, id);
1086 locks_remove_posix(filp, id);
1092 EXPORT_SYMBOL(filp_close);
1095 * Careful here! We test whether the file pointer is NULL before
1096 * releasing the fd. This ensures that one clone task can't release
1097 * an fd while another clone is opening it.
1099 SYSCALL_DEFINE1(close, unsigned int, fd)
1101 int retval = __close_fd(current->files, fd);
1103 /* can't restart close syscall because file table entry was cleared */
1104 if (unlikely(retval == -ERESTARTSYS ||
1105 retval == -ERESTARTNOINTR ||
1106 retval == -ERESTARTNOHAND ||
1107 retval == -ERESTART_RESTARTBLOCK))
1112 EXPORT_SYMBOL(sys_close);
1115 * This routine simulates a hangup on the tty, to arrange that users
1116 * are given clean terminals at login time.
1118 SYSCALL_DEFINE0(vhangup)
1120 if (capable(CAP_SYS_TTY_CONFIG)) {
1128 * Called when an inode is about to be open.
1129 * We use this to disallow opening large files on 32bit systems if
1130 * the caller didn't specify O_LARGEFILE. On 64bit systems we force
1131 * on this flag in sys_open.
1133 int generic_file_open(struct inode * inode, struct file * filp)
1135 if (!(filp->f_flags & O_LARGEFILE) && i_size_read(inode) > MAX_NON_LFS)
1140 EXPORT_SYMBOL(generic_file_open);
1143 * This is used by subsystems that don't want seekable
1144 * file descriptors. The function is not supposed to ever fail, the only
1145 * reason it returns an 'int' and not 'void' is so that it can be plugged
1146 * directly into file_operations structure.
1148 int nonseekable_open(struct inode *inode, struct file *filp)
1150 filp->f_mode &= ~(FMODE_LSEEK | FMODE_PREAD | FMODE_PWRITE);
1154 EXPORT_SYMBOL(nonseekable_open);