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 on truncate too */
55 ret = should_remove_suid(dentry);
57 newattrs.ia_valid |= ret | ATTR_FORCE;
59 mutex_lock(&dentry->d_inode->i_mutex);
60 /* Note any delegations or leases have already been broken: */
61 ret = notify_change(dentry, &newattrs, NULL);
62 mutex_unlock(&dentry->d_inode->i_mutex);
66 long vfs_truncate(struct path *path, loff_t length)
71 inode = path->dentry->d_inode;
73 /* For directories it's -EISDIR, for other non-regulars - -EINVAL */
74 if (S_ISDIR(inode->i_mode))
76 if (!S_ISREG(inode->i_mode))
79 error = mnt_want_write(path->mnt);
83 error = inode_permission(inode, MAY_WRITE);
85 goto mnt_drop_write_and_out;
89 goto mnt_drop_write_and_out;
91 error = get_write_access(inode);
93 goto mnt_drop_write_and_out;
96 * Make sure that there are no leases. get_write_access() protects
97 * against the truncate racing with a lease-granting setlease().
99 error = break_lease(inode, O_WRONLY);
101 goto put_write_and_out;
103 error = locks_verify_truncate(inode, NULL, length);
105 error = security_path_truncate(path);
107 error = do_truncate(path->dentry, length, 0, NULL);
110 put_write_access(inode);
111 mnt_drop_write_and_out:
112 mnt_drop_write(path->mnt);
116 EXPORT_SYMBOL_GPL(vfs_truncate);
118 static long do_sys_truncate(const char __user *pathname, loff_t length)
120 unsigned int lookup_flags = LOOKUP_FOLLOW;
124 if (length < 0) /* sorry, but loff_t says... */
128 error = user_path_at(AT_FDCWD, pathname, lookup_flags, &path);
130 error = vfs_truncate(&path, length);
133 if (retry_estale(error, lookup_flags)) {
134 lookup_flags |= LOOKUP_REVAL;
140 SYSCALL_DEFINE2(truncate, const char __user *, path, long, length)
142 return do_sys_truncate(path, length);
146 COMPAT_SYSCALL_DEFINE2(truncate, const char __user *, path, compat_off_t, length)
148 return do_sys_truncate(path, length);
152 static long do_sys_ftruncate(unsigned int fd, loff_t length, int small)
155 struct dentry *dentry;
167 /* explicitly opened as large or we are on 64-bit box */
168 if (f.file->f_flags & O_LARGEFILE)
171 dentry = f.file->f_path.dentry;
172 inode = dentry->d_inode;
174 if (!S_ISREG(inode->i_mode) || !(f.file->f_mode & FMODE_WRITE))
178 /* Cannot ftruncate over 2^31 bytes without large file support */
179 if (small && length > MAX_NON_LFS)
183 if (IS_APPEND(inode))
186 sb_start_write(inode->i_sb);
187 error = locks_verify_truncate(inode, f.file, length);
189 error = security_path_truncate(&f.file->f_path);
191 error = do_truncate(dentry, length, ATTR_MTIME|ATTR_CTIME, f.file);
192 sb_end_write(inode->i_sb);
199 SYSCALL_DEFINE2(ftruncate, unsigned int, fd, unsigned long, length)
201 return do_sys_ftruncate(fd, length, 1);
205 COMPAT_SYSCALL_DEFINE2(ftruncate, unsigned int, fd, compat_ulong_t, length)
207 return do_sys_ftruncate(fd, length, 1);
211 /* LFS versions of truncate are only needed on 32 bit machines */
212 #if BITS_PER_LONG == 32
213 SYSCALL_DEFINE2(truncate64, const char __user *, path, loff_t, length)
215 return do_sys_truncate(path, length);
218 SYSCALL_DEFINE2(ftruncate64, unsigned int, fd, loff_t, length)
220 return do_sys_ftruncate(fd, length, 0);
222 #endif /* BITS_PER_LONG == 32 */
225 int vfs_fallocate(struct file *file, int mode, loff_t offset, loff_t len)
227 struct inode *inode = file_inode(file);
230 if (offset < 0 || len <= 0)
233 /* Return error if mode is not supported */
234 if (mode & ~FALLOC_FL_SUPPORTED_MASK)
237 /* Punch hole and zero range are mutually exclusive */
238 if ((mode & (FALLOC_FL_PUNCH_HOLE | FALLOC_FL_ZERO_RANGE)) ==
239 (FALLOC_FL_PUNCH_HOLE | FALLOC_FL_ZERO_RANGE))
242 /* Punch hole must have keep size set */
243 if ((mode & FALLOC_FL_PUNCH_HOLE) &&
244 !(mode & FALLOC_FL_KEEP_SIZE))
247 /* Collapse range should only be used exclusively. */
248 if ((mode & FALLOC_FL_COLLAPSE_RANGE) &&
249 (mode & ~FALLOC_FL_COLLAPSE_RANGE))
252 /* Insert range should only be used exclusively. */
253 if ((mode & FALLOC_FL_INSERT_RANGE) &&
254 (mode & ~FALLOC_FL_INSERT_RANGE))
257 if (!(file->f_mode & FMODE_WRITE))
261 * We can only allow pure fallocate on append only files
263 if ((mode & ~FALLOC_FL_KEEP_SIZE) && IS_APPEND(inode))
266 if (IS_IMMUTABLE(inode))
270 * We cannot allow any fallocate operation on an active swapfile
272 if (IS_SWAPFILE(inode))
276 * Revalidate the write permissions, in case security policy has
277 * changed since the files were opened.
279 ret = security_file_permission(file, MAY_WRITE);
283 if (S_ISFIFO(inode->i_mode))
287 * Let individual file system decide if it supports preallocation
288 * for directories or not.
290 if (!S_ISREG(inode->i_mode) && !S_ISDIR(inode->i_mode))
293 /* Check for wrap through zero too */
294 if (((offset + len) > inode->i_sb->s_maxbytes) || ((offset + len) < 0))
297 if (!file->f_op->fallocate)
300 sb_start_write(inode->i_sb);
301 ret = file->f_op->fallocate(file, mode, offset, len);
304 * Create inotify and fanotify events.
306 * To keep the logic simple always create events if fallocate succeeds.
307 * This implies that events are even created if the file size remains
308 * unchanged, e.g. when using flag FALLOC_FL_KEEP_SIZE.
311 fsnotify_modify(file);
313 sb_end_write(inode->i_sb);
316 EXPORT_SYMBOL_GPL(vfs_fallocate);
318 SYSCALL_DEFINE4(fallocate, int, fd, int, mode, loff_t, offset, loff_t, len)
320 struct fd f = fdget(fd);
324 error = vfs_fallocate(f.file, mode, offset, len);
331 * access() needs to use the real uid/gid, not the effective uid/gid.
332 * We do this by temporarily clearing all FS-related capabilities and
333 * switching the fsuid/fsgid around to the real ones.
335 SYSCALL_DEFINE3(faccessat, int, dfd, const char __user *, filename, int, mode)
337 const struct cred *old_cred;
338 struct cred *override_cred;
342 unsigned int lookup_flags = LOOKUP_FOLLOW;
344 if (mode & ~S_IRWXO) /* where's F_OK, X_OK, W_OK, R_OK? */
347 override_cred = prepare_creds();
351 override_cred->fsuid = override_cred->uid;
352 override_cred->fsgid = override_cred->gid;
354 if (!issecure(SECURE_NO_SETUID_FIXUP)) {
355 /* Clear the capabilities if we switch to a non-root user */
356 kuid_t root_uid = make_kuid(override_cred->user_ns, 0);
357 if (!uid_eq(override_cred->uid, root_uid))
358 cap_clear(override_cred->cap_effective);
360 override_cred->cap_effective =
361 override_cred->cap_permitted;
364 old_cred = override_creds(override_cred);
366 res = user_path_at(dfd, filename, lookup_flags, &path);
370 inode = d_backing_inode(path.dentry);
372 if ((mode & MAY_EXEC) && S_ISREG(inode->i_mode)) {
374 * MAY_EXEC on regular files is denied if the fs is mounted
375 * with the "noexec" flag.
378 if (path.mnt->mnt_flags & MNT_NOEXEC)
379 goto out_path_release;
382 res = inode_permission(inode, mode | MAY_ACCESS);
383 /* SuS v2 requires we report a read only fs too */
384 if (res || !(mode & S_IWOTH) || special_file(inode->i_mode))
385 goto out_path_release;
387 * This is a rare case where using __mnt_is_readonly()
388 * is OK without a mnt_want/drop_write() pair. Since
389 * no actual write to the fs is performed here, we do
390 * not need to telegraph to that to anyone.
392 * By doing this, we accept that this access is
393 * inherently racy and know that the fs may change
394 * state before we even see this result.
396 if (__mnt_is_readonly(path.mnt))
401 if (retry_estale(res, lookup_flags)) {
402 lookup_flags |= LOOKUP_REVAL;
406 revert_creds(old_cred);
407 put_cred(override_cred);
411 SYSCALL_DEFINE2(access, const char __user *, filename, int, mode)
413 return sys_faccessat(AT_FDCWD, filename, mode);
416 SYSCALL_DEFINE1(chdir, const char __user *, filename)
420 unsigned int lookup_flags = LOOKUP_FOLLOW | LOOKUP_DIRECTORY;
422 error = user_path_at(AT_FDCWD, filename, lookup_flags, &path);
426 error = inode_permission(path.dentry->d_inode, MAY_EXEC | MAY_CHDIR);
430 set_fs_pwd(current->fs, &path);
434 if (retry_estale(error, lookup_flags)) {
435 lookup_flags |= LOOKUP_REVAL;
442 SYSCALL_DEFINE1(fchdir, unsigned int, fd)
444 struct fd f = fdget_raw(fd);
452 inode = file_inode(f.file);
455 if (!S_ISDIR(inode->i_mode))
458 error = inode_permission(inode, MAY_EXEC | MAY_CHDIR);
460 set_fs_pwd(current->fs, &f.file->f_path);
467 SYSCALL_DEFINE1(chroot, const char __user *, filename)
471 unsigned int lookup_flags = LOOKUP_FOLLOW | LOOKUP_DIRECTORY;
473 error = user_path_at(AT_FDCWD, filename, lookup_flags, &path);
477 error = inode_permission(path.dentry->d_inode, MAY_EXEC | MAY_CHDIR);
482 if (!ns_capable(current_user_ns(), CAP_SYS_CHROOT))
484 error = security_path_chroot(&path);
488 set_fs_root(current->fs, &path);
492 if (retry_estale(error, lookup_flags)) {
493 lookup_flags |= LOOKUP_REVAL;
500 static int chmod_common(struct path *path, umode_t mode)
502 struct inode *inode = path->dentry->d_inode;
503 struct inode *delegated_inode = NULL;
504 struct iattr newattrs;
507 error = mnt_want_write(path->mnt);
511 mutex_lock(&inode->i_mutex);
512 error = security_path_chmod(path, mode);
515 newattrs.ia_mode = (mode & S_IALLUGO) | (inode->i_mode & ~S_IALLUGO);
516 newattrs.ia_valid = ATTR_MODE | ATTR_CTIME;
517 error = notify_change(path->dentry, &newattrs, &delegated_inode);
519 mutex_unlock(&inode->i_mutex);
520 if (delegated_inode) {
521 error = break_deleg_wait(&delegated_inode);
525 mnt_drop_write(path->mnt);
529 SYSCALL_DEFINE2(fchmod, unsigned int, fd, umode_t, mode)
531 struct fd f = fdget(fd);
536 err = chmod_common(&f.file->f_path, mode);
542 SYSCALL_DEFINE3(fchmodat, int, dfd, const char __user *, filename, umode_t, mode)
546 unsigned int lookup_flags = LOOKUP_FOLLOW;
548 error = user_path_at(dfd, filename, lookup_flags, &path);
550 error = chmod_common(&path, mode);
552 if (retry_estale(error, lookup_flags)) {
553 lookup_flags |= LOOKUP_REVAL;
560 SYSCALL_DEFINE2(chmod, const char __user *, filename, umode_t, mode)
562 return sys_fchmodat(AT_FDCWD, filename, mode);
565 static int chown_common(struct path *path, uid_t user, gid_t group)
567 struct inode *inode = path->dentry->d_inode;
568 struct inode *delegated_inode = NULL;
570 struct iattr newattrs;
574 uid = make_kuid(current_user_ns(), user);
575 gid = make_kgid(current_user_ns(), group);
578 newattrs.ia_valid = ATTR_CTIME;
579 if (user != (uid_t) -1) {
582 newattrs.ia_valid |= ATTR_UID;
583 newattrs.ia_uid = uid;
585 if (group != (gid_t) -1) {
588 newattrs.ia_valid |= ATTR_GID;
589 newattrs.ia_gid = gid;
591 if (!S_ISDIR(inode->i_mode))
593 ATTR_KILL_SUID | ATTR_KILL_SGID | ATTR_KILL_PRIV;
594 mutex_lock(&inode->i_mutex);
595 error = security_path_chown(path, uid, gid);
597 error = notify_change(path->dentry, &newattrs, &delegated_inode);
598 mutex_unlock(&inode->i_mutex);
599 if (delegated_inode) {
600 error = break_deleg_wait(&delegated_inode);
607 SYSCALL_DEFINE5(fchownat, int, dfd, const char __user *, filename, uid_t, user,
608 gid_t, group, int, flag)
614 if ((flag & ~(AT_SYMLINK_NOFOLLOW | AT_EMPTY_PATH)) != 0)
617 lookup_flags = (flag & AT_SYMLINK_NOFOLLOW) ? 0 : LOOKUP_FOLLOW;
618 if (flag & AT_EMPTY_PATH)
619 lookup_flags |= LOOKUP_EMPTY;
621 error = user_path_at(dfd, filename, lookup_flags, &path);
624 error = mnt_want_write(path.mnt);
627 error = chown_common(&path, user, group);
628 mnt_drop_write(path.mnt);
631 if (retry_estale(error, lookup_flags)) {
632 lookup_flags |= LOOKUP_REVAL;
639 SYSCALL_DEFINE3(chown, const char __user *, filename, uid_t, user, gid_t, group)
641 return sys_fchownat(AT_FDCWD, filename, user, group, 0);
644 SYSCALL_DEFINE3(lchown, const char __user *, filename, uid_t, user, gid_t, group)
646 return sys_fchownat(AT_FDCWD, filename, user, group,
647 AT_SYMLINK_NOFOLLOW);
650 SYSCALL_DEFINE3(fchown, unsigned int, fd, uid_t, user, gid_t, group)
652 struct fd f = fdget(fd);
658 error = mnt_want_write_file(f.file);
662 error = chown_common(&f.file->f_path, user, group);
663 mnt_drop_write_file(f.file);
670 int open_check_o_direct(struct file *f)
672 /* NB: we're sure to have correct a_ops only after f_op->open */
673 if (f->f_flags & O_DIRECT) {
674 if (!f->f_mapping->a_ops || !f->f_mapping->a_ops->direct_IO)
680 static int do_dentry_open(struct file *f,
681 int (*open)(struct inode *, struct file *),
682 const struct cred *cred)
684 static const struct file_operations empty_fops = {};
688 f->f_mode = OPEN_FMODE(f->f_flags) | FMODE_LSEEK |
689 FMODE_PREAD | FMODE_PWRITE;
691 path_get(&f->f_path);
692 inode = f->f_inode = f->f_path.dentry->d_inode;
693 f->f_mapping = inode->i_mapping;
695 if (unlikely(f->f_flags & O_PATH)) {
696 f->f_mode = FMODE_PATH;
697 f->f_op = &empty_fops;
701 if (f->f_mode & FMODE_WRITE && !special_file(inode->i_mode)) {
702 error = get_write_access(inode);
705 error = __mnt_want_write(f->f_path.mnt);
706 if (unlikely(error)) {
707 put_write_access(inode);
710 f->f_mode |= FMODE_WRITER;
713 /* POSIX.1-2008/SUSv4 Section XSI 2.9.7 */
714 if (S_ISREG(inode->i_mode))
715 f->f_mode |= FMODE_ATOMIC_POS;
717 f->f_op = fops_get(inode->i_fop);
718 if (unlikely(WARN_ON(!f->f_op))) {
723 error = security_file_open(f, cred);
727 error = break_lease(inode, f->f_flags);
732 open = f->f_op->open;
734 error = open(inode, f);
738 if ((f->f_mode & (FMODE_READ | FMODE_WRITE)) == FMODE_READ)
739 i_readcount_inc(inode);
740 if ((f->f_mode & FMODE_READ) &&
741 likely(f->f_op->read || f->f_op->read_iter))
742 f->f_mode |= FMODE_CAN_READ;
743 if ((f->f_mode & FMODE_WRITE) &&
744 likely(f->f_op->write || f->f_op->write_iter))
745 f->f_mode |= FMODE_CAN_WRITE;
747 f->f_flags &= ~(O_CREAT | O_EXCL | O_NOCTTY | O_TRUNC);
749 file_ra_state_init(&f->f_ra, f->f_mapping->host->i_mapping);
755 if (f->f_mode & FMODE_WRITER) {
756 put_write_access(inode);
757 __mnt_drop_write(f->f_path.mnt);
760 path_put(&f->f_path);
761 f->f_path.mnt = NULL;
762 f->f_path.dentry = NULL;
768 * finish_open - finish opening a file
769 * @file: file pointer
770 * @dentry: pointer to dentry
771 * @open: open callback
772 * @opened: state of open
774 * This can be used to finish opening a file passed to i_op->atomic_open().
776 * If the open callback is set to NULL, then the standard f_op->open()
777 * filesystem callback is substituted.
779 * NB: the dentry reference is _not_ consumed. If, for example, the dentry is
780 * the return value of d_splice_alias(), then the caller needs to perform dput()
781 * on it after finish_open().
783 * On successful return @file is a fully instantiated open file. After this, if
784 * an error occurs in ->atomic_open(), it needs to clean up with fput().
786 * Returns zero on success or -errno if the open failed.
788 int finish_open(struct file *file, struct dentry *dentry,
789 int (*open)(struct inode *, struct file *),
793 BUG_ON(*opened & FILE_OPENED); /* once it's opened, it's opened */
795 file->f_path.dentry = dentry;
796 error = do_dentry_open(file, open, current_cred());
798 *opened |= FILE_OPENED;
802 EXPORT_SYMBOL(finish_open);
805 * finish_no_open - finish ->atomic_open() without opening the file
807 * @file: file pointer
808 * @dentry: dentry or NULL (as returned from ->lookup())
810 * This can be used to set the result of a successful lookup in ->atomic_open().
812 * NB: unlike finish_open() this function does consume the dentry reference and
813 * the caller need not dput() it.
815 * Returns "1" which must be the return value of ->atomic_open() after having
816 * called this function.
818 int finish_no_open(struct file *file, struct dentry *dentry)
820 file->f_path.dentry = dentry;
823 EXPORT_SYMBOL(finish_no_open);
825 struct file *dentry_open(const struct path *path, int flags,
826 const struct cred *cred)
831 validate_creds(cred);
833 /* We must always pass in a valid mount pointer. */
836 f = get_empty_filp();
839 error = vfs_open(path, f, cred);
841 /* from now on we need fput() to dispose of f */
842 error = open_check_o_direct(f);
854 EXPORT_SYMBOL(dentry_open);
857 * vfs_open - open the file at the given path
858 * @path: path to open
859 * @filp: newly allocated file with f_flag initialized
860 * @cred: credentials to use
862 int vfs_open(const struct path *path, struct file *filp,
863 const struct cred *cred)
865 struct inode *inode = path->dentry->d_inode;
867 if (inode->i_op->dentry_open)
868 return inode->i_op->dentry_open(path->dentry, filp, cred);
870 filp->f_path = *path;
871 return do_dentry_open(filp, NULL, cred);
874 EXPORT_SYMBOL(vfs_open);
876 static inline int build_open_flags(int flags, umode_t mode, struct open_flags *op)
878 int lookup_flags = 0;
881 if (flags & (O_CREAT | __O_TMPFILE))
882 op->mode = (mode & S_IALLUGO) | S_IFREG;
886 /* Must never be set by userspace */
887 flags &= ~FMODE_NONOTIFY & ~O_CLOEXEC;
890 * O_SYNC is implemented as __O_SYNC|O_DSYNC. As many places only
891 * check for O_DSYNC if the need any syncing at all we enforce it's
892 * always set instead of having to deal with possibly weird behaviour
893 * for malicious applications setting only __O_SYNC.
895 if (flags & __O_SYNC)
898 if (flags & __O_TMPFILE) {
899 if ((flags & O_TMPFILE_MASK) != O_TMPFILE)
901 acc_mode = MAY_OPEN | ACC_MODE(flags);
902 if (!(acc_mode & MAY_WRITE))
904 } else if (flags & O_PATH) {
906 * If we have O_PATH in the open flag. Then we
907 * cannot have anything other than the below set of flags
909 flags &= O_DIRECTORY | O_NOFOLLOW | O_PATH;
912 acc_mode = MAY_OPEN | ACC_MODE(flags);
915 op->open_flag = flags;
917 /* O_TRUNC implies we need access checks for write permissions */
919 acc_mode |= MAY_WRITE;
921 /* Allow the LSM permission hook to distinguish append
922 access from general write access. */
923 if (flags & O_APPEND)
924 acc_mode |= MAY_APPEND;
926 op->acc_mode = acc_mode;
928 op->intent = flags & O_PATH ? 0 : LOOKUP_OPEN;
930 if (flags & O_CREAT) {
931 op->intent |= LOOKUP_CREATE;
933 op->intent |= LOOKUP_EXCL;
936 if (flags & O_DIRECTORY)
937 lookup_flags |= LOOKUP_DIRECTORY;
938 if (!(flags & O_NOFOLLOW))
939 lookup_flags |= LOOKUP_FOLLOW;
940 op->lookup_flags = lookup_flags;
945 * file_open_name - open file and return file pointer
947 * @name: struct filename containing path to open
948 * @flags: open flags as per the open(2) second argument
949 * @mode: mode for the new file if O_CREAT is set, else ignored
951 * This is the helper to open a file from kernelspace if you really
952 * have to. But in generally you should not do this, so please move
953 * along, nothing to see here..
955 struct file *file_open_name(struct filename *name, int flags, umode_t mode)
957 struct open_flags op;
958 int err = build_open_flags(flags, mode, &op);
959 return err ? ERR_PTR(err) : do_filp_open(AT_FDCWD, name, &op);
963 * filp_open - open file and return file pointer
965 * @filename: path to open
966 * @flags: open flags as per the open(2) second argument
967 * @mode: mode for the new file if O_CREAT is set, else ignored
969 * This is the helper to open a file from kernelspace if you really
970 * have to. But in generally you should not do this, so please move
971 * along, nothing to see here..
973 struct file *filp_open(const char *filename, int flags, umode_t mode)
975 struct filename *name = getname_kernel(filename);
976 struct file *file = ERR_CAST(name);
979 file = file_open_name(name, flags, mode);
984 EXPORT_SYMBOL(filp_open);
986 struct file *file_open_root(struct dentry *dentry, struct vfsmount *mnt,
987 const char *filename, int flags)
989 struct open_flags op;
990 int err = build_open_flags(flags, 0, &op);
994 return ERR_PTR(-EINVAL);
995 return do_file_open_root(dentry, mnt, filename, &op);
997 EXPORT_SYMBOL(file_open_root);
999 long do_sys_open(int dfd, const char __user *filename, int flags, umode_t mode)
1001 struct open_flags op;
1002 int fd = build_open_flags(flags, mode, &op);
1003 struct filename *tmp;
1008 tmp = getname(filename);
1010 return PTR_ERR(tmp);
1012 fd = get_unused_fd_flags(flags);
1014 struct file *f = do_filp_open(dfd, tmp, &op);
1027 SYSCALL_DEFINE3(open, const char __user *, filename, int, flags, umode_t, mode)
1029 if (force_o_largefile())
1030 flags |= O_LARGEFILE;
1032 return do_sys_open(AT_FDCWD, filename, flags, mode);
1035 SYSCALL_DEFINE4(openat, int, dfd, const char __user *, filename, int, flags,
1038 if (force_o_largefile())
1039 flags |= O_LARGEFILE;
1041 return do_sys_open(dfd, filename, flags, mode);
1047 * For backward compatibility? Maybe this should be moved
1048 * into arch/i386 instead?
1050 SYSCALL_DEFINE2(creat, const char __user *, pathname, umode_t, mode)
1052 return sys_open(pathname, O_CREAT | O_WRONLY | O_TRUNC, mode);
1058 * "id" is the POSIX thread ID. We use the
1059 * files pointer for this..
1061 int filp_close(struct file *filp, fl_owner_t id)
1065 if (!file_count(filp)) {
1066 printk(KERN_ERR "VFS: Close: file count is 0\n");
1070 if (filp->f_op->flush)
1071 retval = filp->f_op->flush(filp, id);
1073 if (likely(!(filp->f_mode & FMODE_PATH))) {
1074 dnotify_flush(filp, id);
1075 locks_remove_posix(filp, id);
1081 EXPORT_SYMBOL(filp_close);
1084 * Careful here! We test whether the file pointer is NULL before
1085 * releasing the fd. This ensures that one clone task can't release
1086 * an fd while another clone is opening it.
1088 SYSCALL_DEFINE1(close, unsigned int, fd)
1090 int retval = __close_fd(current->files, fd);
1092 /* can't restart close syscall because file table entry was cleared */
1093 if (unlikely(retval == -ERESTARTSYS ||
1094 retval == -ERESTARTNOINTR ||
1095 retval == -ERESTARTNOHAND ||
1096 retval == -ERESTART_RESTARTBLOCK))
1101 EXPORT_SYMBOL(sys_close);
1104 * This routine simulates a hangup on the tty, to arrange that users
1105 * are given clean terminals at login time.
1107 SYSCALL_DEFINE0(vhangup)
1109 if (capable(CAP_SYS_TTY_CONFIG)) {
1117 * Called when an inode is about to be open.
1118 * We use this to disallow opening large files on 32bit systems if
1119 * the caller didn't specify O_LARGEFILE. On 64bit systems we force
1120 * on this flag in sys_open.
1122 int generic_file_open(struct inode * inode, struct file * filp)
1124 if (!(filp->f_flags & O_LARGEFILE) && i_size_read(inode) > MAX_NON_LFS)
1129 EXPORT_SYMBOL(generic_file_open);
1132 * This is used by subsystems that don't want seekable
1133 * file descriptors. The function is not supposed to ever fail, the only
1134 * reason it returns an 'int' and not 'void' is so that it can be plugged
1135 * directly into file_operations structure.
1137 int nonseekable_open(struct inode *inode, struct file *filp)
1139 filp->f_mode &= ~(FMODE_LSEEK | FMODE_PREAD | FMODE_PWRITE);
1143 EXPORT_SYMBOL(nonseekable_open);