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 inode_lock(dentry->d_inode);
62 /* Note any delegations or leases have already been broken: */
63 ret = notify_change(dentry, &newattrs, NULL);
64 inode_unlock(dentry->d_inode);
68 long vfs_truncate(const struct path *path, loff_t length)
71 struct dentry *upperdentry;
74 inode = path->dentry->d_inode;
76 /* For directories it's -EISDIR, for other non-regulars - -EINVAL */
77 if (S_ISDIR(inode->i_mode))
79 if (!S_ISREG(inode->i_mode))
82 error = mnt_want_write(path->mnt);
86 error = inode_permission(inode, MAY_WRITE);
88 goto mnt_drop_write_and_out;
92 goto mnt_drop_write_and_out;
95 * If this is an overlayfs then do as if opening the file so we get
96 * write access on the upper inode, not on the overlay inode. For
97 * non-overlay filesystems d_real() is an identity function.
99 upperdentry = d_real(path->dentry, NULL, O_WRONLY);
100 error = PTR_ERR(upperdentry);
101 if (IS_ERR(upperdentry))
102 goto mnt_drop_write_and_out;
104 error = get_write_access(upperdentry->d_inode);
106 goto mnt_drop_write_and_out;
109 * Make sure that there are no leases. get_write_access() protects
110 * against the truncate racing with a lease-granting setlease().
112 error = break_lease(inode, O_WRONLY);
114 goto put_write_and_out;
116 error = locks_verify_truncate(inode, NULL, length);
118 error = security_path_truncate(path);
120 error = do_truncate(path->dentry, length, 0, NULL);
123 put_write_access(upperdentry->d_inode);
124 mnt_drop_write_and_out:
125 mnt_drop_write(path->mnt);
129 EXPORT_SYMBOL_GPL(vfs_truncate);
131 static long do_sys_truncate(const char __user *pathname, loff_t length)
133 unsigned int lookup_flags = LOOKUP_FOLLOW;
137 if (length < 0) /* sorry, but loff_t says... */
141 error = user_path_at(AT_FDCWD, pathname, lookup_flags, &path);
143 error = vfs_truncate(&path, length);
146 if (retry_estale(error, lookup_flags)) {
147 lookup_flags |= LOOKUP_REVAL;
153 SYSCALL_DEFINE2(truncate, const char __user *, path, long, length)
155 return do_sys_truncate(path, length);
159 COMPAT_SYSCALL_DEFINE2(truncate, const char __user *, path, compat_off_t, length)
161 return do_sys_truncate(path, length);
165 static long do_sys_ftruncate(unsigned int fd, loff_t length, int small)
168 struct dentry *dentry;
180 /* explicitly opened as large or we are on 64-bit box */
181 if (f.file->f_flags & O_LARGEFILE)
184 dentry = f.file->f_path.dentry;
185 inode = dentry->d_inode;
187 if (!S_ISREG(inode->i_mode) || !(f.file->f_mode & FMODE_WRITE))
191 /* Cannot ftruncate over 2^31 bytes without large file support */
192 if (small && length > MAX_NON_LFS)
196 if (IS_APPEND(inode))
199 sb_start_write(inode->i_sb);
200 error = locks_verify_truncate(inode, f.file, length);
202 error = security_path_truncate(&f.file->f_path);
204 error = do_truncate(dentry, length, ATTR_MTIME|ATTR_CTIME, f.file);
205 sb_end_write(inode->i_sb);
212 SYSCALL_DEFINE2(ftruncate, unsigned int, fd, unsigned long, length)
214 return do_sys_ftruncate(fd, length, 1);
218 COMPAT_SYSCALL_DEFINE2(ftruncate, unsigned int, fd, compat_ulong_t, length)
220 return do_sys_ftruncate(fd, length, 1);
224 /* LFS versions of truncate are only needed on 32 bit machines */
225 #if BITS_PER_LONG == 32
226 SYSCALL_DEFINE2(truncate64, const char __user *, path, loff_t, length)
228 return do_sys_truncate(path, length);
231 SYSCALL_DEFINE2(ftruncate64, unsigned int, fd, loff_t, length)
233 return do_sys_ftruncate(fd, length, 0);
235 #endif /* BITS_PER_LONG == 32 */
238 int vfs_fallocate(struct file *file, int mode, loff_t offset, loff_t len)
240 struct inode *inode = file_inode(file);
243 if (offset < 0 || len <= 0)
246 /* Return error if mode is not supported */
247 if (mode & ~FALLOC_FL_SUPPORTED_MASK)
250 /* Punch hole and zero range are mutually exclusive */
251 if ((mode & (FALLOC_FL_PUNCH_HOLE | FALLOC_FL_ZERO_RANGE)) ==
252 (FALLOC_FL_PUNCH_HOLE | FALLOC_FL_ZERO_RANGE))
255 /* Punch hole must have keep size set */
256 if ((mode & FALLOC_FL_PUNCH_HOLE) &&
257 !(mode & FALLOC_FL_KEEP_SIZE))
260 /* Collapse range should only be used exclusively. */
261 if ((mode & FALLOC_FL_COLLAPSE_RANGE) &&
262 (mode & ~FALLOC_FL_COLLAPSE_RANGE))
265 /* Insert range should only be used exclusively. */
266 if ((mode & FALLOC_FL_INSERT_RANGE) &&
267 (mode & ~FALLOC_FL_INSERT_RANGE))
270 /* Unshare range should only be used with allocate mode. */
271 if ((mode & FALLOC_FL_UNSHARE_RANGE) &&
272 (mode & ~(FALLOC_FL_UNSHARE_RANGE | FALLOC_FL_KEEP_SIZE)))
275 if (!(file->f_mode & FMODE_WRITE))
279 * We can only allow pure fallocate on append only files
281 if ((mode & ~FALLOC_FL_KEEP_SIZE) && IS_APPEND(inode))
284 if (IS_IMMUTABLE(inode))
288 * We cannot allow any fallocate operation on an active swapfile
290 if (IS_SWAPFILE(inode))
294 * Revalidate the write permissions, in case security policy has
295 * changed since the files were opened.
297 ret = security_file_permission(file, MAY_WRITE);
301 if (S_ISFIFO(inode->i_mode))
305 * Let individual file system decide if it supports preallocation
306 * for directories or not.
308 if (!S_ISREG(inode->i_mode) && !S_ISDIR(inode->i_mode) &&
309 !S_ISBLK(inode->i_mode))
312 /* Check for wrap through zero too */
313 if (((offset + len) > inode->i_sb->s_maxbytes) || ((offset + len) < 0))
316 if (!file->f_op->fallocate)
319 sb_start_write(inode->i_sb);
320 ret = file->f_op->fallocate(file, mode, offset, len);
323 * Create inotify and fanotify events.
325 * To keep the logic simple always create events if fallocate succeeds.
326 * This implies that events are even created if the file size remains
327 * unchanged, e.g. when using flag FALLOC_FL_KEEP_SIZE.
330 fsnotify_modify(file);
332 sb_end_write(inode->i_sb);
335 EXPORT_SYMBOL_GPL(vfs_fallocate);
337 SYSCALL_DEFINE4(fallocate, int, fd, int, mode, loff_t, offset, loff_t, len)
339 struct fd f = fdget(fd);
343 error = vfs_fallocate(f.file, mode, offset, len);
350 * access() needs to use the real uid/gid, not the effective uid/gid.
351 * We do this by temporarily clearing all FS-related capabilities and
352 * switching the fsuid/fsgid around to the real ones.
354 SYSCALL_DEFINE3(faccessat, int, dfd, const char __user *, filename, int, mode)
356 const struct cred *old_cred;
357 struct cred *override_cred;
361 unsigned int lookup_flags = LOOKUP_FOLLOW;
363 if (mode & ~S_IRWXO) /* where's F_OK, X_OK, W_OK, R_OK? */
366 override_cred = prepare_creds();
370 override_cred->fsuid = override_cred->uid;
371 override_cred->fsgid = override_cred->gid;
373 if (!issecure(SECURE_NO_SETUID_FIXUP)) {
374 /* Clear the capabilities if we switch to a non-root user */
375 kuid_t root_uid = make_kuid(override_cred->user_ns, 0);
376 if (!uid_eq(override_cred->uid, root_uid))
377 cap_clear(override_cred->cap_effective);
379 override_cred->cap_effective =
380 override_cred->cap_permitted;
384 * The new set of credentials can *only* be used in
385 * task-synchronous circumstances, and does not need
386 * RCU freeing, unless somebody then takes a separate
389 * NOTE! This is _only_ true because this credential
390 * is used purely for override_creds() that installs
391 * it as the subjective cred. Other threads will be
392 * accessing ->real_cred, not the subjective cred.
394 * If somebody _does_ make a copy of this (using the
395 * 'get_current_cred()' function), that will clear the
396 * non_rcu field, because now that other user may be
397 * expecting RCU freeing. But normal thread-synchronous
398 * cred accesses will keep things non-RCY.
400 override_cred->non_rcu = 1;
402 old_cred = override_creds(override_cred);
404 res = user_path_at(dfd, filename, lookup_flags, &path);
408 inode = d_backing_inode(path.dentry);
410 if ((mode & MAY_EXEC) && S_ISREG(inode->i_mode)) {
412 * MAY_EXEC on regular files is denied if the fs is mounted
413 * with the "noexec" flag.
416 if (path_noexec(&path))
417 goto out_path_release;
420 res = inode_permission(inode, mode | MAY_ACCESS);
421 /* SuS v2 requires we report a read only fs too */
422 if (res || !(mode & S_IWOTH) || special_file(inode->i_mode))
423 goto out_path_release;
425 * This is a rare case where using __mnt_is_readonly()
426 * is OK without a mnt_want/drop_write() pair. Since
427 * no actual write to the fs is performed here, we do
428 * not need to telegraph to that to anyone.
430 * By doing this, we accept that this access is
431 * inherently racy and know that the fs may change
432 * state before we even see this result.
434 if (__mnt_is_readonly(path.mnt))
439 if (retry_estale(res, lookup_flags)) {
440 lookup_flags |= LOOKUP_REVAL;
444 revert_creds(old_cred);
445 put_cred(override_cred);
449 SYSCALL_DEFINE2(access, const char __user *, filename, int, mode)
451 return sys_faccessat(AT_FDCWD, filename, mode);
454 SYSCALL_DEFINE1(chdir, const char __user *, filename)
458 unsigned int lookup_flags = LOOKUP_FOLLOW | LOOKUP_DIRECTORY;
460 error = user_path_at(AT_FDCWD, filename, lookup_flags, &path);
464 error = inode_permission(path.dentry->d_inode, MAY_EXEC | MAY_CHDIR);
468 set_fs_pwd(current->fs, &path);
472 if (retry_estale(error, lookup_flags)) {
473 lookup_flags |= LOOKUP_REVAL;
480 SYSCALL_DEFINE1(fchdir, unsigned int, fd)
482 struct fd f = fdget_raw(fd);
490 inode = file_inode(f.file);
493 if (!S_ISDIR(inode->i_mode))
496 error = inode_permission(inode, MAY_EXEC | MAY_CHDIR);
498 set_fs_pwd(current->fs, &f.file->f_path);
505 SYSCALL_DEFINE1(chroot, const char __user *, filename)
509 unsigned int lookup_flags = LOOKUP_FOLLOW | LOOKUP_DIRECTORY;
511 error = user_path_at(AT_FDCWD, filename, lookup_flags, &path);
515 error = inode_permission(path.dentry->d_inode, MAY_EXEC | MAY_CHDIR);
520 if (!ns_capable(current_user_ns(), CAP_SYS_CHROOT))
522 error = security_path_chroot(&path);
526 set_fs_root(current->fs, &path);
530 if (retry_estale(error, lookup_flags)) {
531 lookup_flags |= LOOKUP_REVAL;
538 static int chmod_common(const struct path *path, umode_t mode)
540 struct inode *inode = path->dentry->d_inode;
541 struct inode *delegated_inode = NULL;
542 struct iattr newattrs;
545 error = mnt_want_write(path->mnt);
550 error = security_path_chmod(path, mode);
553 newattrs.ia_mode = (mode & S_IALLUGO) | (inode->i_mode & ~S_IALLUGO);
554 newattrs.ia_valid = ATTR_MODE | ATTR_CTIME;
555 error = notify_change(path->dentry, &newattrs, &delegated_inode);
558 if (delegated_inode) {
559 error = break_deleg_wait(&delegated_inode);
563 mnt_drop_write(path->mnt);
567 SYSCALL_DEFINE2(fchmod, unsigned int, fd, umode_t, mode)
569 struct fd f = fdget(fd);
574 err = chmod_common(&f.file->f_path, mode);
580 SYSCALL_DEFINE3(fchmodat, int, dfd, const char __user *, filename, umode_t, mode)
584 unsigned int lookup_flags = LOOKUP_FOLLOW;
586 error = user_path_at(dfd, filename, lookup_flags, &path);
588 error = chmod_common(&path, mode);
590 if (retry_estale(error, lookup_flags)) {
591 lookup_flags |= LOOKUP_REVAL;
598 SYSCALL_DEFINE2(chmod, const char __user *, filename, umode_t, mode)
600 return sys_fchmodat(AT_FDCWD, filename, mode);
603 static int chown_common(const struct path *path, uid_t user, gid_t group)
605 struct inode *inode = path->dentry->d_inode;
606 struct inode *delegated_inode = NULL;
608 struct iattr newattrs;
612 uid = make_kuid(current_user_ns(), user);
613 gid = make_kgid(current_user_ns(), group);
616 newattrs.ia_valid = ATTR_CTIME;
617 if (user != (uid_t) -1) {
620 newattrs.ia_valid |= ATTR_UID;
621 newattrs.ia_uid = uid;
623 if (group != (gid_t) -1) {
626 newattrs.ia_valid |= ATTR_GID;
627 newattrs.ia_gid = gid;
629 if (!S_ISDIR(inode->i_mode))
631 ATTR_KILL_SUID | ATTR_KILL_SGID | ATTR_KILL_PRIV;
633 error = security_path_chown(path, uid, gid);
635 error = notify_change(path->dentry, &newattrs, &delegated_inode);
637 if (delegated_inode) {
638 error = break_deleg_wait(&delegated_inode);
645 SYSCALL_DEFINE5(fchownat, int, dfd, const char __user *, filename, uid_t, user,
646 gid_t, group, int, flag)
652 if ((flag & ~(AT_SYMLINK_NOFOLLOW | AT_EMPTY_PATH)) != 0)
655 lookup_flags = (flag & AT_SYMLINK_NOFOLLOW) ? 0 : LOOKUP_FOLLOW;
656 if (flag & AT_EMPTY_PATH)
657 lookup_flags |= LOOKUP_EMPTY;
659 error = user_path_at(dfd, filename, lookup_flags, &path);
662 error = mnt_want_write(path.mnt);
665 error = chown_common(&path, user, group);
666 mnt_drop_write(path.mnt);
669 if (retry_estale(error, lookup_flags)) {
670 lookup_flags |= LOOKUP_REVAL;
677 SYSCALL_DEFINE3(chown, const char __user *, filename, uid_t, user, gid_t, group)
679 return sys_fchownat(AT_FDCWD, filename, user, group, 0);
682 SYSCALL_DEFINE3(lchown, const char __user *, filename, uid_t, user, gid_t, group)
684 return sys_fchownat(AT_FDCWD, filename, user, group,
685 AT_SYMLINK_NOFOLLOW);
688 SYSCALL_DEFINE3(fchown, unsigned int, fd, uid_t, user, gid_t, group)
690 struct fd f = fdget(fd);
696 error = mnt_want_write_file(f.file);
700 error = chown_common(&f.file->f_path, user, group);
701 mnt_drop_write_file(f.file);
708 int open_check_o_direct(struct file *f)
710 /* NB: we're sure to have correct a_ops only after f_op->open */
711 if (f->f_flags & O_DIRECT) {
712 if (!f->f_mapping->a_ops || !f->f_mapping->a_ops->direct_IO)
718 static int do_dentry_open(struct file *f,
720 int (*open)(struct inode *, struct file *),
721 const struct cred *cred)
723 static const struct file_operations empty_fops = {};
726 f->f_mode = OPEN_FMODE(f->f_flags) | FMODE_LSEEK |
727 FMODE_PREAD | FMODE_PWRITE;
729 path_get(&f->f_path);
731 f->f_mapping = inode->i_mapping;
733 if (unlikely(f->f_flags & O_PATH)) {
734 f->f_mode = FMODE_PATH;
735 f->f_op = &empty_fops;
739 /* Any file opened for execve()/uselib() has to be a regular file. */
740 if (unlikely(f->f_flags & FMODE_EXEC && !S_ISREG(inode->i_mode))) {
745 if (f->f_mode & FMODE_WRITE && !special_file(inode->i_mode)) {
746 error = get_write_access(inode);
749 error = __mnt_want_write(f->f_path.mnt);
750 if (unlikely(error)) {
751 put_write_access(inode);
754 f->f_mode |= FMODE_WRITER;
757 /* POSIX.1-2008/SUSv4 Section XSI 2.9.7 */
758 if (S_ISREG(inode->i_mode) || S_ISDIR(inode->i_mode))
759 f->f_mode |= FMODE_ATOMIC_POS;
761 f->f_op = fops_get(inode->i_fop);
762 if (unlikely(WARN_ON(!f->f_op))) {
767 error = security_file_open(f, cred);
771 error = break_lease(locks_inode(f), f->f_flags);
776 open = f->f_op->open;
778 error = open(inode, f);
782 if ((f->f_mode & (FMODE_READ | FMODE_WRITE)) == FMODE_READ)
783 i_readcount_inc(inode);
784 if ((f->f_mode & FMODE_READ) &&
785 likely(f->f_op->read || f->f_op->read_iter))
786 f->f_mode |= FMODE_CAN_READ;
787 if ((f->f_mode & FMODE_WRITE) &&
788 likely(f->f_op->write || f->f_op->write_iter))
789 f->f_mode |= FMODE_CAN_WRITE;
791 f->f_flags &= ~(O_CREAT | O_EXCL | O_NOCTTY | O_TRUNC);
793 file_ra_state_init(&f->f_ra, f->f_mapping->host->i_mapping);
799 if (f->f_mode & FMODE_WRITER) {
800 put_write_access(inode);
801 __mnt_drop_write(f->f_path.mnt);
804 path_put(&f->f_path);
805 f->f_path.mnt = NULL;
806 f->f_path.dentry = NULL;
812 * finish_open - finish opening a file
813 * @file: file pointer
814 * @dentry: pointer to dentry
815 * @open: open callback
816 * @opened: state of open
818 * This can be used to finish opening a file passed to i_op->atomic_open().
820 * If the open callback is set to NULL, then the standard f_op->open()
821 * filesystem callback is substituted.
823 * NB: the dentry reference is _not_ consumed. If, for example, the dentry is
824 * the return value of d_splice_alias(), then the caller needs to perform dput()
825 * on it after finish_open().
827 * Returns zero on success or -errno if the open failed.
829 int finish_open(struct file *file, struct dentry *dentry,
830 int (*open)(struct inode *, struct file *),
834 BUG_ON(*opened & FILE_OPENED); /* once it's opened, it's opened */
836 file->f_path.dentry = dentry;
837 error = do_dentry_open(file, d_backing_inode(dentry), open,
840 *opened |= FILE_OPENED;
844 EXPORT_SYMBOL(finish_open);
847 * finish_no_open - finish ->atomic_open() without opening the file
849 * @file: file pointer
850 * @dentry: dentry or NULL (as returned from ->lookup())
852 * This can be used to set the result of a successful lookup in ->atomic_open().
854 * NB: unlike finish_open() this function does consume the dentry reference and
855 * the caller need not dput() it.
857 * Returns "1" which must be the return value of ->atomic_open() after having
858 * called this function.
860 int finish_no_open(struct file *file, struct dentry *dentry)
862 file->f_path.dentry = dentry;
865 EXPORT_SYMBOL(finish_no_open);
867 char *file_path(struct file *filp, char *buf, int buflen)
869 return d_path(&filp->f_path, buf, buflen);
871 EXPORT_SYMBOL(file_path);
874 * vfs_open - open the file at the given path
875 * @path: path to open
876 * @file: newly allocated file with f_flag initialized
877 * @cred: credentials to use
879 int vfs_open(const struct path *path, struct file *file,
880 const struct cred *cred)
882 struct dentry *dentry = d_real(path->dentry, NULL, file->f_flags);
885 return PTR_ERR(dentry);
887 file->f_path = *path;
888 return do_dentry_open(file, d_backing_inode(dentry), NULL, cred);
891 struct file *dentry_open(const struct path *path, int flags,
892 const struct cred *cred)
897 validate_creds(cred);
899 /* We must always pass in a valid mount pointer. */
902 f = get_empty_filp();
905 error = vfs_open(path, f, cred);
907 /* from now on we need fput() to dispose of f */
908 error = open_check_o_direct(f);
920 EXPORT_SYMBOL(dentry_open);
922 static inline int build_open_flags(int flags, umode_t mode, struct open_flags *op)
924 int lookup_flags = 0;
925 int acc_mode = ACC_MODE(flags);
928 * Clear out all open flags we don't know about so that we don't report
929 * them in fcntl(F_GETFD) or similar interfaces.
931 flags &= VALID_OPEN_FLAGS;
933 if (flags & (O_CREAT | __O_TMPFILE))
934 op->mode = (mode & S_IALLUGO) | S_IFREG;
938 /* Must never be set by userspace */
939 flags &= ~FMODE_NONOTIFY & ~O_CLOEXEC;
942 * O_SYNC is implemented as __O_SYNC|O_DSYNC. As many places only
943 * check for O_DSYNC if the need any syncing at all we enforce it's
944 * always set instead of having to deal with possibly weird behaviour
945 * for malicious applications setting only __O_SYNC.
947 if (flags & __O_SYNC)
950 if (flags & __O_TMPFILE) {
951 if ((flags & O_TMPFILE_MASK) != O_TMPFILE)
953 if (!(acc_mode & MAY_WRITE))
955 } else if (flags & O_PATH) {
957 * If we have O_PATH in the open flag. Then we
958 * cannot have anything other than the below set of flags
960 flags &= O_DIRECTORY | O_NOFOLLOW | O_PATH;
964 op->open_flag = flags;
966 /* O_TRUNC implies we need access checks for write permissions */
968 acc_mode |= MAY_WRITE;
970 /* Allow the LSM permission hook to distinguish append
971 access from general write access. */
972 if (flags & O_APPEND)
973 acc_mode |= MAY_APPEND;
975 op->acc_mode = acc_mode;
977 op->intent = flags & O_PATH ? 0 : LOOKUP_OPEN;
979 if (flags & O_CREAT) {
980 op->intent |= LOOKUP_CREATE;
982 op->intent |= LOOKUP_EXCL;
985 if (flags & O_DIRECTORY)
986 lookup_flags |= LOOKUP_DIRECTORY;
987 if (!(flags & O_NOFOLLOW))
988 lookup_flags |= LOOKUP_FOLLOW;
989 op->lookup_flags = lookup_flags;
994 * file_open_name - open file and return file pointer
996 * @name: struct filename containing path to open
997 * @flags: open flags as per the open(2) second argument
998 * @mode: mode for the new file if O_CREAT is set, else ignored
1000 * This is the helper to open a file from kernelspace if you really
1001 * have to. But in generally you should not do this, so please move
1002 * along, nothing to see here..
1004 struct file *file_open_name(struct filename *name, int flags, umode_t mode)
1006 struct open_flags op;
1007 int err = build_open_flags(flags, mode, &op);
1008 return err ? ERR_PTR(err) : do_filp_open(AT_FDCWD, name, &op);
1012 * filp_open - open file and return file pointer
1014 * @filename: path to open
1015 * @flags: open flags as per the open(2) second argument
1016 * @mode: mode for the new file if O_CREAT is set, else ignored
1018 * This is the helper to open a file from kernelspace if you really
1019 * have to. But in generally you should not do this, so please move
1020 * along, nothing to see here..
1022 struct file *filp_open(const char *filename, int flags, umode_t mode)
1024 struct filename *name = getname_kernel(filename);
1025 struct file *file = ERR_CAST(name);
1027 if (!IS_ERR(name)) {
1028 file = file_open_name(name, flags, mode);
1033 EXPORT_SYMBOL(filp_open);
1035 struct file *file_open_root(struct dentry *dentry, struct vfsmount *mnt,
1036 const char *filename, int flags, umode_t mode)
1038 struct open_flags op;
1039 int err = build_open_flags(flags, mode, &op);
1041 return ERR_PTR(err);
1042 return do_file_open_root(dentry, mnt, filename, &op);
1044 EXPORT_SYMBOL(file_open_root);
1046 struct file *filp_clone_open(struct file *oldfile)
1051 file = get_empty_filp();
1055 file->f_flags = oldfile->f_flags;
1056 retval = vfs_open(&oldfile->f_path, file, oldfile->f_cred);
1059 return ERR_PTR(retval);
1064 EXPORT_SYMBOL(filp_clone_open);
1066 long do_sys_open(int dfd, const char __user *filename, int flags, umode_t mode)
1068 struct open_flags op;
1069 int fd = build_open_flags(flags, mode, &op);
1070 struct filename *tmp;
1075 tmp = getname(filename);
1077 return PTR_ERR(tmp);
1079 fd = get_unused_fd_flags(flags);
1081 struct file *f = do_filp_open(dfd, tmp, &op);
1094 SYSCALL_DEFINE3(open, const char __user *, filename, int, flags, umode_t, mode)
1096 if (force_o_largefile())
1097 flags |= O_LARGEFILE;
1099 return do_sys_open(AT_FDCWD, filename, flags, mode);
1102 SYSCALL_DEFINE4(openat, int, dfd, const char __user *, filename, int, flags,
1105 if (force_o_largefile())
1106 flags |= O_LARGEFILE;
1108 return do_sys_open(dfd, filename, flags, mode);
1114 * For backward compatibility? Maybe this should be moved
1115 * into arch/i386 instead?
1117 SYSCALL_DEFINE2(creat, const char __user *, pathname, umode_t, mode)
1119 return sys_open(pathname, O_CREAT | O_WRONLY | O_TRUNC, mode);
1125 * "id" is the POSIX thread ID. We use the
1126 * files pointer for this..
1128 int filp_close(struct file *filp, fl_owner_t id)
1132 if (!file_count(filp)) {
1133 printk(KERN_ERR "VFS: Close: file count is 0\n");
1137 if (filp->f_op->flush)
1138 retval = filp->f_op->flush(filp, id);
1140 if (likely(!(filp->f_mode & FMODE_PATH))) {
1141 dnotify_flush(filp, id);
1142 locks_remove_posix(filp, id);
1148 EXPORT_SYMBOL(filp_close);
1151 * Careful here! We test whether the file pointer is NULL before
1152 * releasing the fd. This ensures that one clone task can't release
1153 * an fd while another clone is opening it.
1155 SYSCALL_DEFINE1(close, unsigned int, fd)
1157 int retval = __close_fd(current->files, fd);
1159 /* can't restart close syscall because file table entry was cleared */
1160 if (unlikely(retval == -ERESTARTSYS ||
1161 retval == -ERESTARTNOINTR ||
1162 retval == -ERESTARTNOHAND ||
1163 retval == -ERESTART_RESTARTBLOCK))
1168 EXPORT_SYMBOL(sys_close);
1171 * This routine simulates a hangup on the tty, to arrange that users
1172 * are given clean terminals at login time.
1174 SYSCALL_DEFINE0(vhangup)
1176 if (capable(CAP_SYS_TTY_CONFIG)) {
1184 * Called when an inode is about to be open.
1185 * We use this to disallow opening large files on 32bit systems if
1186 * the caller didn't specify O_LARGEFILE. On 64bit systems we force
1187 * on this flag in sys_open.
1189 int generic_file_open(struct inode * inode, struct file * filp)
1191 if (!(filp->f_flags & O_LARGEFILE) && i_size_read(inode) > MAX_NON_LFS)
1196 EXPORT_SYMBOL(generic_file_open);
1199 * This is used by subsystems that don't want seekable
1200 * file descriptors. The function is not supposed to ever fail, the only
1201 * reason it returns an 'int' and not 'void' is so that it can be plugged
1202 * directly into file_operations structure.
1204 int nonseekable_open(struct inode *inode, struct file *filp)
1206 filp->f_mode &= ~(FMODE_LSEEK | FMODE_PREAD | FMODE_PWRITE);
1210 EXPORT_SYMBOL(nonseekable_open);
1213 * stream_open is used by subsystems that want stream-like file descriptors.
1214 * Such file descriptors are not seekable and don't have notion of position
1215 * (file.f_pos is always 0). Contrary to file descriptors of other regular
1216 * files, .read() and .write() can run simultaneously.
1218 * stream_open never fails and is marked to return int so that it could be
1219 * directly used as file_operations.open .
1221 int stream_open(struct inode *inode, struct file *filp)
1223 filp->f_mode &= ~(FMODE_LSEEK | FMODE_PREAD | FMODE_PWRITE | FMODE_ATOMIC_POS);
1224 filp->f_mode |= FMODE_STREAM;
1228 EXPORT_SYMBOL(stream_open);