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 <linux/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)
73 inode = path->dentry->d_inode;
75 /* For directories it's -EISDIR, for other non-regulars - -EINVAL */
76 if (S_ISDIR(inode->i_mode))
78 if (!S_ISREG(inode->i_mode))
81 error = mnt_want_write(path->mnt);
85 error = inode_permission(inode, MAY_WRITE);
87 goto mnt_drop_write_and_out;
91 goto mnt_drop_write_and_out;
93 error = get_write_access(inode);
95 goto mnt_drop_write_and_out;
98 * Make sure that there are no leases. get_write_access() protects
99 * against the truncate racing with a lease-granting setlease().
101 error = break_lease(inode, O_WRONLY);
103 goto put_write_and_out;
105 error = locks_verify_truncate(inode, NULL, length);
107 error = security_path_truncate(path);
109 error = do_truncate(path->dentry, length, 0, NULL);
112 put_write_access(inode);
113 mnt_drop_write_and_out:
114 mnt_drop_write(path->mnt);
118 EXPORT_SYMBOL_GPL(vfs_truncate);
120 long do_sys_truncate(const char __user *pathname, loff_t length)
122 unsigned int lookup_flags = LOOKUP_FOLLOW;
126 if (length < 0) /* sorry, but loff_t says... */
130 error = user_path_at(AT_FDCWD, pathname, lookup_flags, &path);
132 error = vfs_truncate(&path, length);
135 if (retry_estale(error, lookup_flags)) {
136 lookup_flags |= LOOKUP_REVAL;
142 SYSCALL_DEFINE2(truncate, const char __user *, path, long, length)
144 return do_sys_truncate(path, length);
148 COMPAT_SYSCALL_DEFINE2(truncate, const char __user *, path, compat_off_t, length)
150 return do_sys_truncate(path, length);
154 long do_sys_ftruncate(unsigned int fd, loff_t length, int small)
157 struct dentry *dentry;
169 /* explicitly opened as large or we are on 64-bit box */
170 if (f.file->f_flags & O_LARGEFILE)
173 dentry = f.file->f_path.dentry;
174 inode = dentry->d_inode;
176 if (!S_ISREG(inode->i_mode) || !(f.file->f_mode & FMODE_WRITE))
180 /* Cannot ftruncate over 2^31 bytes without large file support */
181 if (small && length > MAX_NON_LFS)
185 /* Check IS_APPEND on real upper inode */
186 if (IS_APPEND(file_inode(f.file)))
189 sb_start_write(inode->i_sb);
190 error = locks_verify_truncate(inode, f.file, length);
192 error = security_path_truncate(&f.file->f_path);
194 error = do_truncate(dentry, length, ATTR_MTIME|ATTR_CTIME, f.file);
195 sb_end_write(inode->i_sb);
202 SYSCALL_DEFINE2(ftruncate, unsigned int, fd, unsigned long, length)
204 return do_sys_ftruncate(fd, length, 1);
208 COMPAT_SYSCALL_DEFINE2(ftruncate, unsigned int, fd, compat_ulong_t, length)
210 return do_sys_ftruncate(fd, length, 1);
214 /* LFS versions of truncate are only needed on 32 bit machines */
215 #if BITS_PER_LONG == 32
216 SYSCALL_DEFINE2(truncate64, const char __user *, path, loff_t, length)
218 return do_sys_truncate(path, length);
221 SYSCALL_DEFINE2(ftruncate64, unsigned int, fd, loff_t, length)
223 return do_sys_ftruncate(fd, length, 0);
225 #endif /* BITS_PER_LONG == 32 */
228 int vfs_fallocate(struct file *file, int mode, loff_t offset, loff_t len)
230 struct inode *inode = file_inode(file);
233 if (offset < 0 || len <= 0)
236 /* Return error if mode is not supported */
237 if (mode & ~FALLOC_FL_SUPPORTED_MASK)
240 /* Punch hole and zero range are mutually exclusive */
241 if ((mode & (FALLOC_FL_PUNCH_HOLE | FALLOC_FL_ZERO_RANGE)) ==
242 (FALLOC_FL_PUNCH_HOLE | FALLOC_FL_ZERO_RANGE))
245 /* Punch hole must have keep size set */
246 if ((mode & FALLOC_FL_PUNCH_HOLE) &&
247 !(mode & FALLOC_FL_KEEP_SIZE))
250 /* Collapse range should only be used exclusively. */
251 if ((mode & FALLOC_FL_COLLAPSE_RANGE) &&
252 (mode & ~FALLOC_FL_COLLAPSE_RANGE))
255 /* Insert range should only be used exclusively. */
256 if ((mode & FALLOC_FL_INSERT_RANGE) &&
257 (mode & ~FALLOC_FL_INSERT_RANGE))
260 /* Unshare range should only be used with allocate mode. */
261 if ((mode & FALLOC_FL_UNSHARE_RANGE) &&
262 (mode & ~(FALLOC_FL_UNSHARE_RANGE | FALLOC_FL_KEEP_SIZE)))
265 if (!(file->f_mode & FMODE_WRITE))
269 * We can only allow pure fallocate on append only files
271 if ((mode & ~FALLOC_FL_KEEP_SIZE) && IS_APPEND(inode))
274 if (IS_IMMUTABLE(inode))
278 * We cannot allow any fallocate operation on an active swapfile
280 if (IS_SWAPFILE(inode))
284 * Revalidate the write permissions, in case security policy has
285 * changed since the files were opened.
287 ret = security_file_permission(file, MAY_WRITE);
291 if (S_ISFIFO(inode->i_mode))
294 if (S_ISDIR(inode->i_mode))
297 if (!S_ISREG(inode->i_mode) && !S_ISBLK(inode->i_mode))
300 /* Check for wrap through zero too */
301 if (((offset + len) > inode->i_sb->s_maxbytes) || ((offset + len) < 0))
304 if (!file->f_op->fallocate)
307 file_start_write(file);
308 ret = file->f_op->fallocate(file, mode, offset, len);
311 * Create inotify and fanotify events.
313 * To keep the logic simple always create events if fallocate succeeds.
314 * This implies that events are even created if the file size remains
315 * unchanged, e.g. when using flag FALLOC_FL_KEEP_SIZE.
318 fsnotify_modify(file);
320 file_end_write(file);
323 EXPORT_SYMBOL_GPL(vfs_fallocate);
325 int ksys_fallocate(int fd, int mode, loff_t offset, loff_t len)
327 struct fd f = fdget(fd);
331 error = vfs_fallocate(f.file, mode, offset, len);
337 SYSCALL_DEFINE4(fallocate, int, fd, int, mode, loff_t, offset, loff_t, len)
339 return ksys_fallocate(fd, mode, offset, len);
343 * access() needs to use the real uid/gid, not the effective uid/gid.
344 * We do this by temporarily clearing all FS-related capabilities and
345 * switching the fsuid/fsgid around to the real ones.
347 long do_faccessat(int dfd, const char __user *filename, int mode)
349 const struct cred *old_cred;
350 struct cred *override_cred;
354 unsigned int lookup_flags = LOOKUP_FOLLOW;
356 if (mode & ~S_IRWXO) /* where's F_OK, X_OK, W_OK, R_OK? */
359 override_cred = prepare_creds();
363 override_cred->fsuid = override_cred->uid;
364 override_cred->fsgid = override_cred->gid;
366 if (!issecure(SECURE_NO_SETUID_FIXUP)) {
367 /* Clear the capabilities if we switch to a non-root user */
368 kuid_t root_uid = make_kuid(override_cred->user_ns, 0);
369 if (!uid_eq(override_cred->uid, root_uid))
370 cap_clear(override_cred->cap_effective);
372 override_cred->cap_effective =
373 override_cred->cap_permitted;
377 * The new set of credentials can *only* be used in
378 * task-synchronous circumstances, and does not need
379 * RCU freeing, unless somebody then takes a separate
382 * NOTE! This is _only_ true because this credential
383 * is used purely for override_creds() that installs
384 * it as the subjective cred. Other threads will be
385 * accessing ->real_cred, not the subjective cred.
387 * If somebody _does_ make a copy of this (using the
388 * 'get_current_cred()' function), that will clear the
389 * non_rcu field, because now that other user may be
390 * expecting RCU freeing. But normal thread-synchronous
391 * cred accesses will keep things non-RCY.
393 override_cred->non_rcu = 1;
395 old_cred = override_creds(override_cred);
397 res = user_path_at(dfd, filename, lookup_flags, &path);
401 inode = d_backing_inode(path.dentry);
403 if ((mode & MAY_EXEC) && S_ISREG(inode->i_mode)) {
405 * MAY_EXEC on regular files is denied if the fs is mounted
406 * with the "noexec" flag.
409 if (path_noexec(&path))
410 goto out_path_release;
413 res = inode_permission(inode, mode | MAY_ACCESS);
414 /* SuS v2 requires we report a read only fs too */
415 if (res || !(mode & S_IWOTH) || special_file(inode->i_mode))
416 goto out_path_release;
418 * This is a rare case where using __mnt_is_readonly()
419 * is OK without a mnt_want/drop_write() pair. Since
420 * no actual write to the fs is performed here, we do
421 * not need to telegraph to that to anyone.
423 * By doing this, we accept that this access is
424 * inherently racy and know that the fs may change
425 * state before we even see this result.
427 if (__mnt_is_readonly(path.mnt))
432 if (retry_estale(res, lookup_flags)) {
433 lookup_flags |= LOOKUP_REVAL;
437 revert_creds(old_cred);
438 put_cred(override_cred);
442 SYSCALL_DEFINE3(faccessat, int, dfd, const char __user *, filename, int, mode)
444 return do_faccessat(dfd, filename, mode);
447 SYSCALL_DEFINE2(access, const char __user *, filename, int, mode)
449 return do_faccessat(AT_FDCWD, filename, mode);
452 int ksys_chdir(const char __user *filename)
456 unsigned int lookup_flags = LOOKUP_FOLLOW | LOOKUP_DIRECTORY;
458 error = user_path_at(AT_FDCWD, filename, lookup_flags, &path);
462 error = inode_permission(path.dentry->d_inode, MAY_EXEC | MAY_CHDIR);
466 set_fs_pwd(current->fs, &path);
470 if (retry_estale(error, lookup_flags)) {
471 lookup_flags |= LOOKUP_REVAL;
478 SYSCALL_DEFINE1(chdir, const char __user *, filename)
480 return ksys_chdir(filename);
483 SYSCALL_DEFINE1(fchdir, unsigned int, fd)
485 struct fd f = fdget_raw(fd);
493 if (!d_can_lookup(f.file->f_path.dentry))
496 error = inode_permission(file_inode(f.file), MAY_EXEC | MAY_CHDIR);
498 set_fs_pwd(current->fs, &f.file->f_path);
505 int ksys_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 SYSCALL_DEFINE1(chroot, const char __user *, filename)
540 return ksys_chroot(filename);
543 static int chmod_common(const struct path *path, umode_t mode)
545 struct inode *inode = path->dentry->d_inode;
546 struct inode *delegated_inode = NULL;
547 struct iattr newattrs;
550 error = mnt_want_write(path->mnt);
555 error = security_path_chmod(path, mode);
558 newattrs.ia_mode = (mode & S_IALLUGO) | (inode->i_mode & ~S_IALLUGO);
559 newattrs.ia_valid = ATTR_MODE | ATTR_CTIME;
560 error = notify_change(path->dentry, &newattrs, &delegated_inode);
563 if (delegated_inode) {
564 error = break_deleg_wait(&delegated_inode);
568 mnt_drop_write(path->mnt);
572 int ksys_fchmod(unsigned int fd, umode_t mode)
574 struct fd f = fdget(fd);
579 err = chmod_common(&f.file->f_path, mode);
585 SYSCALL_DEFINE2(fchmod, unsigned int, fd, umode_t, mode)
587 return ksys_fchmod(fd, mode);
590 int do_fchmodat(int dfd, const char __user *filename, umode_t mode)
594 unsigned int lookup_flags = LOOKUP_FOLLOW;
596 error = user_path_at(dfd, filename, lookup_flags, &path);
598 error = chmod_common(&path, mode);
600 if (retry_estale(error, lookup_flags)) {
601 lookup_flags |= LOOKUP_REVAL;
608 SYSCALL_DEFINE3(fchmodat, int, dfd, const char __user *, filename,
611 return do_fchmodat(dfd, filename, mode);
614 SYSCALL_DEFINE2(chmod, const char __user *, filename, umode_t, mode)
616 return do_fchmodat(AT_FDCWD, filename, mode);
619 static int chown_common(const struct path *path, uid_t user, gid_t group)
621 struct inode *inode = path->dentry->d_inode;
622 struct inode *delegated_inode = NULL;
624 struct iattr newattrs;
628 uid = make_kuid(current_user_ns(), user);
629 gid = make_kgid(current_user_ns(), group);
632 newattrs.ia_valid = ATTR_CTIME;
633 if (user != (uid_t) -1) {
636 newattrs.ia_valid |= ATTR_UID;
637 newattrs.ia_uid = uid;
639 if (group != (gid_t) -1) {
642 newattrs.ia_valid |= ATTR_GID;
643 newattrs.ia_gid = gid;
645 if (!S_ISDIR(inode->i_mode))
647 ATTR_KILL_SUID | ATTR_KILL_SGID | ATTR_KILL_PRIV;
649 error = security_path_chown(path, uid, gid);
651 error = notify_change(path->dentry, &newattrs, &delegated_inode);
653 if (delegated_inode) {
654 error = break_deleg_wait(&delegated_inode);
661 int do_fchownat(int dfd, const char __user *filename, uid_t user, gid_t group,
668 if ((flag & ~(AT_SYMLINK_NOFOLLOW | AT_EMPTY_PATH)) != 0)
671 lookup_flags = (flag & AT_SYMLINK_NOFOLLOW) ? 0 : LOOKUP_FOLLOW;
672 if (flag & AT_EMPTY_PATH)
673 lookup_flags |= LOOKUP_EMPTY;
675 error = user_path_at(dfd, filename, lookup_flags, &path);
678 error = mnt_want_write(path.mnt);
681 error = chown_common(&path, user, group);
682 mnt_drop_write(path.mnt);
685 if (retry_estale(error, lookup_flags)) {
686 lookup_flags |= LOOKUP_REVAL;
693 SYSCALL_DEFINE5(fchownat, int, dfd, const char __user *, filename, uid_t, user,
694 gid_t, group, int, flag)
696 return do_fchownat(dfd, filename, user, group, flag);
699 SYSCALL_DEFINE3(chown, const char __user *, filename, uid_t, user, gid_t, group)
701 return do_fchownat(AT_FDCWD, filename, user, group, 0);
704 SYSCALL_DEFINE3(lchown, const char __user *, filename, uid_t, user, gid_t, group)
706 return do_fchownat(AT_FDCWD, filename, user, group,
707 AT_SYMLINK_NOFOLLOW);
710 int ksys_fchown(unsigned int fd, uid_t user, gid_t group)
712 struct fd f = fdget(fd);
718 error = mnt_want_write_file(f.file);
722 error = chown_common(&f.file->f_path, user, group);
723 mnt_drop_write_file(f.file);
730 SYSCALL_DEFINE3(fchown, unsigned int, fd, uid_t, user, gid_t, group)
732 return ksys_fchown(fd, user, group);
735 static int do_dentry_open(struct file *f,
737 int (*open)(struct inode *, struct file *))
739 static const struct file_operations empty_fops = {};
742 path_get(&f->f_path);
744 f->f_mapping = inode->i_mapping;
746 /* Ensure that we skip any errors that predate opening of the file */
747 f->f_wb_err = filemap_sample_wb_err(f->f_mapping);
749 if (unlikely(f->f_flags & O_PATH)) {
750 f->f_mode = FMODE_PATH | FMODE_OPENED;
751 f->f_op = &empty_fops;
755 /* Any file opened for execve()/uselib() has to be a regular file. */
756 if (unlikely(f->f_flags & FMODE_EXEC && !S_ISREG(inode->i_mode))) {
761 if (f->f_mode & FMODE_WRITE && !special_file(inode->i_mode)) {
762 error = get_write_access(inode);
765 error = __mnt_want_write(f->f_path.mnt);
766 if (unlikely(error)) {
767 put_write_access(inode);
770 f->f_mode |= FMODE_WRITER;
773 /* POSIX.1-2008/SUSv4 Section XSI 2.9.7 */
774 if (S_ISREG(inode->i_mode) || S_ISDIR(inode->i_mode))
775 f->f_mode |= FMODE_ATOMIC_POS;
777 f->f_op = fops_get(inode->i_fop);
778 if (unlikely(WARN_ON(!f->f_op))) {
783 error = security_file_open(f);
787 error = break_lease(locks_inode(f), f->f_flags);
791 /* normally all 3 are set; ->open() can clear them if needed */
792 f->f_mode |= FMODE_LSEEK | FMODE_PREAD | FMODE_PWRITE;
794 open = f->f_op->open;
796 error = open(inode, f);
800 f->f_mode |= FMODE_OPENED;
801 if ((f->f_mode & (FMODE_READ | FMODE_WRITE)) == FMODE_READ)
802 i_readcount_inc(inode);
803 if ((f->f_mode & FMODE_READ) &&
804 likely(f->f_op->read || f->f_op->read_iter))
805 f->f_mode |= FMODE_CAN_READ;
806 if ((f->f_mode & FMODE_WRITE) &&
807 likely(f->f_op->write || f->f_op->write_iter))
808 f->f_mode |= FMODE_CAN_WRITE;
810 f->f_write_hint = WRITE_LIFE_NOT_SET;
811 f->f_flags &= ~(O_CREAT | O_EXCL | O_NOCTTY | O_TRUNC);
813 file_ra_state_init(&f->f_ra, f->f_mapping->host->i_mapping);
815 /* NB: we're sure to have correct a_ops only after f_op->open */
816 if (f->f_flags & O_DIRECT) {
817 if (!f->f_mapping->a_ops || !f->f_mapping->a_ops->direct_IO)
823 if (WARN_ON_ONCE(error > 0))
826 if (f->f_mode & FMODE_WRITER) {
827 put_write_access(inode);
828 __mnt_drop_write(f->f_path.mnt);
831 path_put(&f->f_path);
832 f->f_path.mnt = NULL;
833 f->f_path.dentry = NULL;
839 * finish_open - finish opening a file
840 * @file: file pointer
841 * @dentry: pointer to dentry
842 * @open: open callback
843 * @opened: state of open
845 * This can be used to finish opening a file passed to i_op->atomic_open().
847 * If the open callback is set to NULL, then the standard f_op->open()
848 * filesystem callback is substituted.
850 * NB: the dentry reference is _not_ consumed. If, for example, the dentry is
851 * the return value of d_splice_alias(), then the caller needs to perform dput()
852 * on it after finish_open().
854 * On successful return @file is a fully instantiated open file. After this, if
855 * an error occurs in ->atomic_open(), it needs to clean up with fput().
857 * Returns zero on success or -errno if the open failed.
859 int finish_open(struct file *file, struct dentry *dentry,
860 int (*open)(struct inode *, struct file *))
862 BUG_ON(file->f_mode & FMODE_OPENED); /* once it's opened, it's opened */
864 file->f_path.dentry = dentry;
865 return do_dentry_open(file, d_backing_inode(dentry), open);
867 EXPORT_SYMBOL(finish_open);
870 * finish_no_open - finish ->atomic_open() without opening the file
872 * @file: file pointer
873 * @dentry: dentry or NULL (as returned from ->lookup())
875 * This can be used to set the result of a successful lookup in ->atomic_open().
877 * NB: unlike finish_open() this function does consume the dentry reference and
878 * the caller need not dput() it.
880 * Returns "0" which must be the return value of ->atomic_open() after having
881 * called this function.
883 int finish_no_open(struct file *file, struct dentry *dentry)
885 file->f_path.dentry = dentry;
888 EXPORT_SYMBOL(finish_no_open);
890 char *file_path(struct file *filp, char *buf, int buflen)
892 return d_path(&filp->f_path, buf, buflen);
894 EXPORT_SYMBOL(file_path);
897 * vfs_open - open the file at the given path
898 * @path: path to open
899 * @file: newly allocated file with f_flag initialized
900 * @cred: credentials to use
902 int vfs_open(const struct path *path, struct file *file)
904 file->f_path = *path;
905 return do_dentry_open(file, d_backing_inode(path->dentry), NULL);
908 struct file *dentry_open(const struct path *path, int flags,
909 const struct cred *cred)
914 validate_creds(cred);
916 /* We must always pass in a valid mount pointer. */
919 f = alloc_empty_file(flags, cred);
921 error = vfs_open(path, f);
929 EXPORT_SYMBOL(dentry_open);
931 struct file *open_with_fake_path(const struct path *path, int flags,
932 struct inode *inode, const struct cred *cred)
934 struct file *f = alloc_empty_file_noaccount(flags, cred);
939 error = do_dentry_open(f, inode, NULL);
947 EXPORT_SYMBOL(open_with_fake_path);
949 static inline int build_open_flags(int flags, umode_t mode, struct open_flags *op)
951 int lookup_flags = 0;
952 int acc_mode = ACC_MODE(flags);
955 * Clear out all open flags we don't know about so that we don't report
956 * them in fcntl(F_GETFD) or similar interfaces.
958 flags &= VALID_OPEN_FLAGS;
960 if (flags & (O_CREAT | __O_TMPFILE))
961 op->mode = (mode & S_IALLUGO) | S_IFREG;
965 /* Must never be set by userspace */
966 flags &= ~FMODE_NONOTIFY & ~O_CLOEXEC;
969 * O_SYNC is implemented as __O_SYNC|O_DSYNC. As many places only
970 * check for O_DSYNC if the need any syncing at all we enforce it's
971 * always set instead of having to deal with possibly weird behaviour
972 * for malicious applications setting only __O_SYNC.
974 if (flags & __O_SYNC)
977 if (flags & __O_TMPFILE) {
978 if ((flags & O_TMPFILE_MASK) != O_TMPFILE)
980 if (!(acc_mode & MAY_WRITE))
982 } else if (flags & O_PATH) {
984 * If we have O_PATH in the open flag. Then we
985 * cannot have anything other than the below set of flags
987 flags &= O_DIRECTORY | O_NOFOLLOW | O_PATH;
991 op->open_flag = flags;
993 /* O_TRUNC implies we need access checks for write permissions */
995 acc_mode |= MAY_WRITE;
997 /* Allow the LSM permission hook to distinguish append
998 access from general write access. */
999 if (flags & O_APPEND)
1000 acc_mode |= MAY_APPEND;
1002 op->acc_mode = acc_mode;
1004 op->intent = flags & O_PATH ? 0 : LOOKUP_OPEN;
1006 if (flags & O_CREAT) {
1007 op->intent |= LOOKUP_CREATE;
1009 op->intent |= LOOKUP_EXCL;
1012 if (flags & O_DIRECTORY)
1013 lookup_flags |= LOOKUP_DIRECTORY;
1014 if (!(flags & O_NOFOLLOW))
1015 lookup_flags |= LOOKUP_FOLLOW;
1016 op->lookup_flags = lookup_flags;
1021 * file_open_name - open file and return file pointer
1023 * @name: struct filename containing path to open
1024 * @flags: open flags as per the open(2) second argument
1025 * @mode: mode for the new file if O_CREAT is set, else ignored
1027 * This is the helper to open a file from kernelspace if you really
1028 * have to. But in generally you should not do this, so please move
1029 * along, nothing to see here..
1031 struct file *file_open_name(struct filename *name, int flags, umode_t mode)
1033 struct open_flags op;
1034 int err = build_open_flags(flags, mode, &op);
1035 return err ? ERR_PTR(err) : do_filp_open(AT_FDCWD, name, &op);
1039 * filp_open - open file and return file pointer
1041 * @filename: path to open
1042 * @flags: open flags as per the open(2) second argument
1043 * @mode: mode for the new file if O_CREAT is set, else ignored
1045 * This is the helper to open a file from kernelspace if you really
1046 * have to. But in generally you should not do this, so please move
1047 * along, nothing to see here..
1049 struct file *filp_open(const char *filename, int flags, umode_t mode)
1051 struct filename *name = getname_kernel(filename);
1052 struct file *file = ERR_CAST(name);
1054 if (!IS_ERR(name)) {
1055 file = file_open_name(name, flags, mode);
1060 EXPORT_SYMBOL(filp_open);
1062 struct file *file_open_root(struct dentry *dentry, struct vfsmount *mnt,
1063 const char *filename, int flags, umode_t mode)
1065 struct open_flags op;
1066 int err = build_open_flags(flags, mode, &op);
1068 return ERR_PTR(err);
1069 return do_file_open_root(dentry, mnt, filename, &op);
1071 EXPORT_SYMBOL(file_open_root);
1073 long do_sys_open(int dfd, const char __user *filename, int flags, umode_t mode)
1075 struct open_flags op;
1076 int fd = build_open_flags(flags, mode, &op);
1077 struct filename *tmp;
1082 tmp = getname(filename);
1084 return PTR_ERR(tmp);
1086 fd = get_unused_fd_flags(flags);
1088 struct file *f = do_filp_open(dfd, tmp, &op);
1101 SYSCALL_DEFINE3(open, const char __user *, filename, int, flags, umode_t, mode)
1103 if (force_o_largefile())
1104 flags |= O_LARGEFILE;
1106 return do_sys_open(AT_FDCWD, filename, flags, mode);
1109 SYSCALL_DEFINE4(openat, int, dfd, const char __user *, filename, int, flags,
1112 if (force_o_largefile())
1113 flags |= O_LARGEFILE;
1115 return do_sys_open(dfd, filename, flags, mode);
1118 #ifdef CONFIG_COMPAT
1120 * Exactly like sys_open(), except that it doesn't set the
1123 COMPAT_SYSCALL_DEFINE3(open, const char __user *, filename, int, flags, umode_t, mode)
1125 return do_sys_open(AT_FDCWD, filename, flags, mode);
1129 * Exactly like sys_openat(), except that it doesn't set the
1132 COMPAT_SYSCALL_DEFINE4(openat, int, dfd, const char __user *, filename, int, flags, umode_t, mode)
1134 return do_sys_open(dfd, filename, flags, mode);
1141 * For backward compatibility? Maybe this should be moved
1142 * into arch/i386 instead?
1144 SYSCALL_DEFINE2(creat, const char __user *, pathname, umode_t, mode)
1146 return ksys_open(pathname, O_CREAT | O_WRONLY | O_TRUNC, mode);
1152 * "id" is the POSIX thread ID. We use the
1153 * files pointer for this..
1155 int filp_close(struct file *filp, fl_owner_t id)
1159 if (!file_count(filp)) {
1160 printk(KERN_ERR "VFS: Close: file count is 0\n");
1164 if (filp->f_op->flush)
1165 retval = filp->f_op->flush(filp, id);
1167 if (likely(!(filp->f_mode & FMODE_PATH))) {
1168 dnotify_flush(filp, id);
1169 locks_remove_posix(filp, id);
1175 EXPORT_SYMBOL(filp_close);
1178 * Careful here! We test whether the file pointer is NULL before
1179 * releasing the fd. This ensures that one clone task can't release
1180 * an fd while another clone is opening it.
1182 SYSCALL_DEFINE1(close, unsigned int, fd)
1184 int retval = __close_fd(current->files, fd);
1186 /* can't restart close syscall because file table entry was cleared */
1187 if (unlikely(retval == -ERESTARTSYS ||
1188 retval == -ERESTARTNOINTR ||
1189 retval == -ERESTARTNOHAND ||
1190 retval == -ERESTART_RESTARTBLOCK))
1197 * This routine simulates a hangup on the tty, to arrange that users
1198 * are given clean terminals at login time.
1200 SYSCALL_DEFINE0(vhangup)
1202 if (capable(CAP_SYS_TTY_CONFIG)) {
1210 * Called when an inode is about to be open.
1211 * We use this to disallow opening large files on 32bit systems if
1212 * the caller didn't specify O_LARGEFILE. On 64bit systems we force
1213 * on this flag in sys_open.
1215 int generic_file_open(struct inode * inode, struct file * filp)
1217 if (!(filp->f_flags & O_LARGEFILE) && i_size_read(inode) > MAX_NON_LFS)
1222 EXPORT_SYMBOL(generic_file_open);
1225 * This is used by subsystems that don't want seekable
1226 * file descriptors. The function is not supposed to ever fail, the only
1227 * reason it returns an 'int' and not 'void' is so that it can be plugged
1228 * directly into file_operations structure.
1230 int nonseekable_open(struct inode *inode, struct file *filp)
1232 filp->f_mode &= ~(FMODE_LSEEK | FMODE_PREAD | FMODE_PWRITE);
1236 EXPORT_SYMBOL(nonseekable_open);
1239 * stream_open is used by subsystems that want stream-like file descriptors.
1240 * Such file descriptors are not seekable and don't have notion of position
1241 * (file.f_pos is always 0). Contrary to file descriptors of other regular
1242 * files, .read() and .write() can run simultaneously.
1244 * stream_open never fails and is marked to return int so that it could be
1245 * directly used as file_operations.open .
1247 int stream_open(struct inode *inode, struct file *filp)
1249 filp->f_mode &= ~(FMODE_LSEEK | FMODE_PREAD | FMODE_PWRITE | FMODE_ATOMIC_POS);
1250 filp->f_mode |= FMODE_STREAM;
1254 EXPORT_SYMBOL(stream_open);