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;
376 old_cred = override_creds(override_cred);
378 res = user_path_at(dfd, filename, lookup_flags, &path);
382 inode = d_backing_inode(path.dentry);
384 if ((mode & MAY_EXEC) && S_ISREG(inode->i_mode)) {
386 * MAY_EXEC on regular files is denied if the fs is mounted
387 * with the "noexec" flag.
390 if (path_noexec(&path))
391 goto out_path_release;
394 res = inode_permission(inode, mode | MAY_ACCESS);
395 /* SuS v2 requires we report a read only fs too */
396 if (res || !(mode & S_IWOTH) || special_file(inode->i_mode))
397 goto out_path_release;
399 * This is a rare case where using __mnt_is_readonly()
400 * is OK without a mnt_want/drop_write() pair. Since
401 * no actual write to the fs is performed here, we do
402 * not need to telegraph to that to anyone.
404 * By doing this, we accept that this access is
405 * inherently racy and know that the fs may change
406 * state before we even see this result.
408 if (__mnt_is_readonly(path.mnt))
413 if (retry_estale(res, lookup_flags)) {
414 lookup_flags |= LOOKUP_REVAL;
418 revert_creds(old_cred);
419 put_cred(override_cred);
423 SYSCALL_DEFINE3(faccessat, int, dfd, const char __user *, filename, int, mode)
425 return do_faccessat(dfd, filename, mode);
428 SYSCALL_DEFINE2(access, const char __user *, filename, int, mode)
430 return do_faccessat(AT_FDCWD, filename, mode);
433 int ksys_chdir(const char __user *filename)
437 unsigned int lookup_flags = LOOKUP_FOLLOW | LOOKUP_DIRECTORY;
439 error = user_path_at(AT_FDCWD, filename, lookup_flags, &path);
443 error = inode_permission(path.dentry->d_inode, MAY_EXEC | MAY_CHDIR);
447 set_fs_pwd(current->fs, &path);
451 if (retry_estale(error, lookup_flags)) {
452 lookup_flags |= LOOKUP_REVAL;
459 SYSCALL_DEFINE1(chdir, const char __user *, filename)
461 return ksys_chdir(filename);
464 SYSCALL_DEFINE1(fchdir, unsigned int, fd)
466 struct fd f = fdget_raw(fd);
474 if (!d_can_lookup(f.file->f_path.dentry))
477 error = inode_permission(file_inode(f.file), MAY_EXEC | MAY_CHDIR);
479 set_fs_pwd(current->fs, &f.file->f_path);
486 int ksys_chroot(const char __user *filename)
490 unsigned int lookup_flags = LOOKUP_FOLLOW | LOOKUP_DIRECTORY;
492 error = user_path_at(AT_FDCWD, filename, lookup_flags, &path);
496 error = inode_permission(path.dentry->d_inode, MAY_EXEC | MAY_CHDIR);
501 if (!ns_capable(current_user_ns(), CAP_SYS_CHROOT))
503 error = security_path_chroot(&path);
507 set_fs_root(current->fs, &path);
511 if (retry_estale(error, lookup_flags)) {
512 lookup_flags |= LOOKUP_REVAL;
519 SYSCALL_DEFINE1(chroot, const char __user *, filename)
521 return ksys_chroot(filename);
524 static int chmod_common(const struct path *path, umode_t mode)
526 struct inode *inode = path->dentry->d_inode;
527 struct inode *delegated_inode = NULL;
528 struct iattr newattrs;
531 error = mnt_want_write(path->mnt);
536 error = security_path_chmod(path, mode);
539 newattrs.ia_mode = (mode & S_IALLUGO) | (inode->i_mode & ~S_IALLUGO);
540 newattrs.ia_valid = ATTR_MODE | ATTR_CTIME;
541 error = notify_change(path->dentry, &newattrs, &delegated_inode);
544 if (delegated_inode) {
545 error = break_deleg_wait(&delegated_inode);
549 mnt_drop_write(path->mnt);
553 int ksys_fchmod(unsigned int fd, umode_t mode)
555 struct fd f = fdget(fd);
560 err = chmod_common(&f.file->f_path, mode);
566 SYSCALL_DEFINE2(fchmod, unsigned int, fd, umode_t, mode)
568 return ksys_fchmod(fd, mode);
571 int do_fchmodat(int dfd, const char __user *filename, umode_t mode)
575 unsigned int lookup_flags = LOOKUP_FOLLOW;
577 error = user_path_at(dfd, filename, lookup_flags, &path);
579 error = chmod_common(&path, mode);
581 if (retry_estale(error, lookup_flags)) {
582 lookup_flags |= LOOKUP_REVAL;
589 SYSCALL_DEFINE3(fchmodat, int, dfd, const char __user *, filename,
592 return do_fchmodat(dfd, filename, mode);
595 SYSCALL_DEFINE2(chmod, const char __user *, filename, umode_t, mode)
597 return do_fchmodat(AT_FDCWD, filename, mode);
600 static int chown_common(const struct path *path, uid_t user, gid_t group)
602 struct inode *inode = path->dentry->d_inode;
603 struct inode *delegated_inode = NULL;
605 struct iattr newattrs;
609 uid = make_kuid(current_user_ns(), user);
610 gid = make_kgid(current_user_ns(), group);
613 newattrs.ia_valid = ATTR_CTIME;
614 if (user != (uid_t) -1) {
617 newattrs.ia_valid |= ATTR_UID;
618 newattrs.ia_uid = uid;
620 if (group != (gid_t) -1) {
623 newattrs.ia_valid |= ATTR_GID;
624 newattrs.ia_gid = gid;
626 if (!S_ISDIR(inode->i_mode))
628 ATTR_KILL_SUID | ATTR_KILL_SGID | ATTR_KILL_PRIV;
630 error = security_path_chown(path, uid, gid);
632 error = notify_change(path->dentry, &newattrs, &delegated_inode);
634 if (delegated_inode) {
635 error = break_deleg_wait(&delegated_inode);
642 int do_fchownat(int dfd, const char __user *filename, uid_t user, gid_t group,
649 if ((flag & ~(AT_SYMLINK_NOFOLLOW | AT_EMPTY_PATH)) != 0)
652 lookup_flags = (flag & AT_SYMLINK_NOFOLLOW) ? 0 : LOOKUP_FOLLOW;
653 if (flag & AT_EMPTY_PATH)
654 lookup_flags |= LOOKUP_EMPTY;
656 error = user_path_at(dfd, filename, lookup_flags, &path);
659 error = mnt_want_write(path.mnt);
662 error = chown_common(&path, user, group);
663 mnt_drop_write(path.mnt);
666 if (retry_estale(error, lookup_flags)) {
667 lookup_flags |= LOOKUP_REVAL;
674 SYSCALL_DEFINE5(fchownat, int, dfd, const char __user *, filename, uid_t, user,
675 gid_t, group, int, flag)
677 return do_fchownat(dfd, filename, user, group, flag);
680 SYSCALL_DEFINE3(chown, const char __user *, filename, uid_t, user, gid_t, group)
682 return do_fchownat(AT_FDCWD, filename, user, group, 0);
685 SYSCALL_DEFINE3(lchown, const char __user *, filename, uid_t, user, gid_t, group)
687 return do_fchownat(AT_FDCWD, filename, user, group,
688 AT_SYMLINK_NOFOLLOW);
691 int ksys_fchown(unsigned int fd, uid_t user, gid_t group)
693 struct fd f = fdget(fd);
699 error = mnt_want_write_file(f.file);
703 error = chown_common(&f.file->f_path, user, group);
704 mnt_drop_write_file(f.file);
711 SYSCALL_DEFINE3(fchown, unsigned int, fd, uid_t, user, gid_t, group)
713 return ksys_fchown(fd, user, group);
716 static int do_dentry_open(struct file *f,
718 int (*open)(struct inode *, struct file *))
720 static const struct file_operations empty_fops = {};
723 path_get(&f->f_path);
725 f->f_mapping = inode->i_mapping;
727 /* Ensure that we skip any errors that predate opening of the file */
728 f->f_wb_err = filemap_sample_wb_err(f->f_mapping);
730 if (unlikely(f->f_flags & O_PATH)) {
731 f->f_mode = FMODE_PATH | FMODE_OPENED;
732 f->f_op = &empty_fops;
736 /* Any file opened for execve()/uselib() has to be a regular file. */
737 if (unlikely(f->f_flags & FMODE_EXEC && !S_ISREG(inode->i_mode))) {
742 if (f->f_mode & FMODE_WRITE && !special_file(inode->i_mode)) {
743 error = get_write_access(inode);
746 error = __mnt_want_write(f->f_path.mnt);
747 if (unlikely(error)) {
748 put_write_access(inode);
751 f->f_mode |= FMODE_WRITER;
754 /* POSIX.1-2008/SUSv4 Section XSI 2.9.7 */
755 if (S_ISREG(inode->i_mode) || S_ISDIR(inode->i_mode))
756 f->f_mode |= FMODE_ATOMIC_POS;
758 f->f_op = fops_get(inode->i_fop);
759 if (unlikely(WARN_ON(!f->f_op))) {
764 error = security_file_open(f);
768 error = break_lease(locks_inode(f), f->f_flags);
772 /* normally all 3 are set; ->open() can clear them if needed */
773 f->f_mode |= FMODE_LSEEK | FMODE_PREAD | FMODE_PWRITE;
775 open = f->f_op->open;
777 error = open(inode, f);
781 f->f_mode |= FMODE_OPENED;
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_write_hint = WRITE_LIFE_NOT_SET;
792 f->f_flags &= ~(O_CREAT | O_EXCL | O_NOCTTY | O_TRUNC);
794 file_ra_state_init(&f->f_ra, f->f_mapping->host->i_mapping);
796 /* NB: we're sure to have correct a_ops only after f_op->open */
797 if (f->f_flags & O_DIRECT) {
798 if (!f->f_mapping->a_ops || !f->f_mapping->a_ops->direct_IO)
804 if (WARN_ON_ONCE(error > 0))
807 if (f->f_mode & FMODE_WRITER) {
808 put_write_access(inode);
809 __mnt_drop_write(f->f_path.mnt);
812 path_put(&f->f_path);
813 f->f_path.mnt = NULL;
814 f->f_path.dentry = NULL;
820 * finish_open - finish opening a file
821 * @file: file pointer
822 * @dentry: pointer to dentry
823 * @open: open callback
824 * @opened: state of open
826 * This can be used to finish opening a file passed to i_op->atomic_open().
828 * If the open callback is set to NULL, then the standard f_op->open()
829 * filesystem callback is substituted.
831 * NB: the dentry reference is _not_ consumed. If, for example, the dentry is
832 * the return value of d_splice_alias(), then the caller needs to perform dput()
833 * on it after finish_open().
835 * On successful return @file is a fully instantiated open file. After this, if
836 * an error occurs in ->atomic_open(), it needs to clean up with fput().
838 * Returns zero on success or -errno if the open failed.
840 int finish_open(struct file *file, struct dentry *dentry,
841 int (*open)(struct inode *, struct file *))
843 BUG_ON(file->f_mode & FMODE_OPENED); /* once it's opened, it's opened */
845 file->f_path.dentry = dentry;
846 return do_dentry_open(file, d_backing_inode(dentry), open);
848 EXPORT_SYMBOL(finish_open);
851 * finish_no_open - finish ->atomic_open() without opening the file
853 * @file: file pointer
854 * @dentry: dentry or NULL (as returned from ->lookup())
856 * This can be used to set the result of a successful lookup in ->atomic_open().
858 * NB: unlike finish_open() this function does consume the dentry reference and
859 * the caller need not dput() it.
861 * Returns "0" which must be the return value of ->atomic_open() after having
862 * called this function.
864 int finish_no_open(struct file *file, struct dentry *dentry)
866 file->f_path.dentry = dentry;
869 EXPORT_SYMBOL(finish_no_open);
871 char *file_path(struct file *filp, char *buf, int buflen)
873 return d_path(&filp->f_path, buf, buflen);
875 EXPORT_SYMBOL(file_path);
878 * vfs_open - open the file at the given path
879 * @path: path to open
880 * @file: newly allocated file with f_flag initialized
881 * @cred: credentials to use
883 int vfs_open(const struct path *path, struct file *file)
885 file->f_path = *path;
886 return do_dentry_open(file, d_backing_inode(path->dentry), NULL);
889 struct file *dentry_open(const struct path *path, int flags,
890 const struct cred *cred)
895 validate_creds(cred);
897 /* We must always pass in a valid mount pointer. */
900 f = alloc_empty_file(flags, cred);
902 error = vfs_open(path, f);
910 EXPORT_SYMBOL(dentry_open);
912 struct file *open_with_fake_path(const struct path *path, int flags,
913 struct inode *inode, const struct cred *cred)
915 struct file *f = alloc_empty_file_noaccount(flags, cred);
920 error = do_dentry_open(f, inode, NULL);
928 EXPORT_SYMBOL(open_with_fake_path);
930 static inline int build_open_flags(int flags, umode_t mode, struct open_flags *op)
932 int lookup_flags = 0;
933 int acc_mode = ACC_MODE(flags);
936 * Clear out all open flags we don't know about so that we don't report
937 * them in fcntl(F_GETFD) or similar interfaces.
939 flags &= VALID_OPEN_FLAGS;
941 if (flags & (O_CREAT | __O_TMPFILE))
942 op->mode = (mode & S_IALLUGO) | S_IFREG;
946 /* Must never be set by userspace */
947 flags &= ~FMODE_NONOTIFY & ~O_CLOEXEC;
950 * O_SYNC is implemented as __O_SYNC|O_DSYNC. As many places only
951 * check for O_DSYNC if the need any syncing at all we enforce it's
952 * always set instead of having to deal with possibly weird behaviour
953 * for malicious applications setting only __O_SYNC.
955 if (flags & __O_SYNC)
958 if (flags & __O_TMPFILE) {
959 if ((flags & O_TMPFILE_MASK) != O_TMPFILE)
961 if (!(acc_mode & MAY_WRITE))
963 } else if (flags & O_PATH) {
965 * If we have O_PATH in the open flag. Then we
966 * cannot have anything other than the below set of flags
968 flags &= O_DIRECTORY | O_NOFOLLOW | O_PATH;
972 op->open_flag = flags;
974 /* O_TRUNC implies we need access checks for write permissions */
976 acc_mode |= MAY_WRITE;
978 /* Allow the LSM permission hook to distinguish append
979 access from general write access. */
980 if (flags & O_APPEND)
981 acc_mode |= MAY_APPEND;
983 op->acc_mode = acc_mode;
985 op->intent = flags & O_PATH ? 0 : LOOKUP_OPEN;
987 if (flags & O_CREAT) {
988 op->intent |= LOOKUP_CREATE;
990 op->intent |= LOOKUP_EXCL;
993 if (flags & O_DIRECTORY)
994 lookup_flags |= LOOKUP_DIRECTORY;
995 if (!(flags & O_NOFOLLOW))
996 lookup_flags |= LOOKUP_FOLLOW;
997 op->lookup_flags = lookup_flags;
1002 * file_open_name - open file and return file pointer
1004 * @name: struct filename containing path to open
1005 * @flags: open flags as per the open(2) second argument
1006 * @mode: mode for the new file if O_CREAT is set, else ignored
1008 * This is the helper to open a file from kernelspace if you really
1009 * have to. But in generally you should not do this, so please move
1010 * along, nothing to see here..
1012 struct file *file_open_name(struct filename *name, int flags, umode_t mode)
1014 struct open_flags op;
1015 int err = build_open_flags(flags, mode, &op);
1016 return err ? ERR_PTR(err) : do_filp_open(AT_FDCWD, name, &op);
1020 * filp_open - open file and return file pointer
1022 * @filename: path to open
1023 * @flags: open flags as per the open(2) second argument
1024 * @mode: mode for the new file if O_CREAT is set, else ignored
1026 * This is the helper to open a file from kernelspace if you really
1027 * have to. But in generally you should not do this, so please move
1028 * along, nothing to see here..
1030 struct file *filp_open(const char *filename, int flags, umode_t mode)
1032 struct filename *name = getname_kernel(filename);
1033 struct file *file = ERR_CAST(name);
1035 if (!IS_ERR(name)) {
1036 file = file_open_name(name, flags, mode);
1041 EXPORT_SYMBOL(filp_open);
1043 struct file *file_open_root(struct dentry *dentry, struct vfsmount *mnt,
1044 const char *filename, int flags, umode_t mode)
1046 struct open_flags op;
1047 int err = build_open_flags(flags, mode, &op);
1049 return ERR_PTR(err);
1050 return do_file_open_root(dentry, mnt, filename, &op);
1052 EXPORT_SYMBOL(file_open_root);
1054 long do_sys_open(int dfd, const char __user *filename, int flags, umode_t mode)
1056 struct open_flags op;
1057 int fd = build_open_flags(flags, mode, &op);
1058 struct filename *tmp;
1063 tmp = getname(filename);
1065 return PTR_ERR(tmp);
1067 fd = get_unused_fd_flags(flags);
1069 struct file *f = do_filp_open(dfd, tmp, &op);
1082 SYSCALL_DEFINE3(open, const char __user *, filename, int, flags, umode_t, mode)
1084 if (force_o_largefile())
1085 flags |= O_LARGEFILE;
1087 return do_sys_open(AT_FDCWD, filename, flags, mode);
1090 SYSCALL_DEFINE4(openat, int, dfd, const char __user *, filename, int, flags,
1093 if (force_o_largefile())
1094 flags |= O_LARGEFILE;
1096 return do_sys_open(dfd, filename, flags, mode);
1099 #ifdef CONFIG_COMPAT
1101 * Exactly like sys_open(), except that it doesn't set the
1104 COMPAT_SYSCALL_DEFINE3(open, const char __user *, filename, int, flags, umode_t, mode)
1106 return do_sys_open(AT_FDCWD, filename, flags, mode);
1110 * Exactly like sys_openat(), except that it doesn't set the
1113 COMPAT_SYSCALL_DEFINE4(openat, int, dfd, const char __user *, filename, int, flags, umode_t, mode)
1115 return do_sys_open(dfd, filename, flags, mode);
1122 * For backward compatibility? Maybe this should be moved
1123 * into arch/i386 instead?
1125 SYSCALL_DEFINE2(creat, const char __user *, pathname, umode_t, mode)
1127 return ksys_open(pathname, O_CREAT | O_WRONLY | O_TRUNC, mode);
1133 * "id" is the POSIX thread ID. We use the
1134 * files pointer for this..
1136 int filp_close(struct file *filp, fl_owner_t id)
1140 if (!file_count(filp)) {
1141 printk(KERN_ERR "VFS: Close: file count is 0\n");
1145 if (filp->f_op->flush)
1146 retval = filp->f_op->flush(filp, id);
1148 if (likely(!(filp->f_mode & FMODE_PATH))) {
1149 dnotify_flush(filp, id);
1150 locks_remove_posix(filp, id);
1156 EXPORT_SYMBOL(filp_close);
1159 * Careful here! We test whether the file pointer is NULL before
1160 * releasing the fd. This ensures that one clone task can't release
1161 * an fd while another clone is opening it.
1163 SYSCALL_DEFINE1(close, unsigned int, fd)
1165 int retval = __close_fd(current->files, fd);
1167 /* can't restart close syscall because file table entry was cleared */
1168 if (unlikely(retval == -ERESTARTSYS ||
1169 retval == -ERESTARTNOINTR ||
1170 retval == -ERESTARTNOHAND ||
1171 retval == -ERESTART_RESTARTBLOCK))
1178 * This routine simulates a hangup on the tty, to arrange that users
1179 * are given clean terminals at login time.
1181 SYSCALL_DEFINE0(vhangup)
1183 if (capable(CAP_SYS_TTY_CONFIG)) {
1191 * Called when an inode is about to be open.
1192 * We use this to disallow opening large files on 32bit systems if
1193 * the caller didn't specify O_LARGEFILE. On 64bit systems we force
1194 * on this flag in sys_open.
1196 int generic_file_open(struct inode * inode, struct file * filp)
1198 if (!(filp->f_flags & O_LARGEFILE) && i_size_read(inode) > MAX_NON_LFS)
1203 EXPORT_SYMBOL(generic_file_open);
1206 * This is used by subsystems that don't want seekable
1207 * file descriptors. The function is not supposed to ever fail, the only
1208 * reason it returns an 'int' and not 'void' is so that it can be plugged
1209 * directly into file_operations structure.
1211 int nonseekable_open(struct inode *inode, struct file *filp)
1213 filp->f_mode &= ~(FMODE_LSEEK | FMODE_PREAD | FMODE_PWRITE);
1217 EXPORT_SYMBOL(nonseekable_open);
1220 * stream_open is used by subsystems that want stream-like file descriptors.
1221 * Such file descriptors are not seekable and don't have notion of position
1222 * (file.f_pos is always 0). Contrary to file descriptors of other regular
1223 * files, .read() and .write() can run simultaneously.
1225 * stream_open never fails and is marked to return int so that it could be
1226 * directly used as file_operations.open .
1228 int stream_open(struct inode *inode, struct file *filp)
1230 filp->f_mode &= ~(FMODE_LSEEK | FMODE_PREAD | FMODE_PWRITE | FMODE_ATOMIC_POS);
1231 filp->f_mode |= FMODE_STREAM;
1235 EXPORT_SYMBOL(stream_open);