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_truncate2(struct vfsmount *mnt, struct dentry *dentry, loff_t length,
38 unsigned int time_attrs, struct file *filp)
41 struct iattr newattrs;
43 /* Not pretty: "inode->i_size" shouldn't really be signed. But it is. */
47 newattrs.ia_size = length;
48 newattrs.ia_valid = ATTR_SIZE | time_attrs;
50 newattrs.ia_file = filp;
51 newattrs.ia_valid |= ATTR_FILE;
54 /* Remove suid, sgid, and file capabilities on truncate too */
55 ret = dentry_needs_remove_privs(dentry);
59 newattrs.ia_valid |= ret | ATTR_FORCE;
61 mutex_lock(&dentry->d_inode->i_mutex);
62 /* Note any delegations or leases have already been broken: */
63 ret = notify_change2(mnt, dentry, &newattrs, NULL);
64 mutex_unlock(&dentry->d_inode->i_mutex);
67 int do_truncate(struct dentry *dentry, loff_t length, unsigned int time_attrs,
70 return do_truncate2(NULL, dentry, length, time_attrs, filp);
73 long vfs_truncate(struct path *path, loff_t length)
79 inode = path->dentry->d_inode;
82 /* For directories it's -EISDIR, for other non-regulars - -EINVAL */
83 if (S_ISDIR(inode->i_mode))
85 if (!S_ISREG(inode->i_mode))
88 error = mnt_want_write(path->mnt);
92 error = inode_permission2(mnt, inode, MAY_WRITE);
94 goto mnt_drop_write_and_out;
98 goto mnt_drop_write_and_out;
100 error = get_write_access(inode);
102 goto mnt_drop_write_and_out;
105 * Make sure that there are no leases. get_write_access() protects
106 * against the truncate racing with a lease-granting setlease().
108 error = break_lease(inode, O_WRONLY);
110 goto put_write_and_out;
112 error = locks_verify_truncate(inode, NULL, length);
114 error = security_path_truncate(path);
116 error = do_truncate2(mnt, path->dentry, length, 0, NULL);
119 put_write_access(inode);
120 mnt_drop_write_and_out:
121 mnt_drop_write(path->mnt);
125 EXPORT_SYMBOL_GPL(vfs_truncate);
127 static long do_sys_truncate(const char __user *pathname, loff_t length)
129 unsigned int lookup_flags = LOOKUP_FOLLOW;
133 if (length < 0) /* sorry, but loff_t says... */
137 error = user_path_at(AT_FDCWD, pathname, lookup_flags, &path);
139 error = vfs_truncate(&path, length);
142 if (retry_estale(error, lookup_flags)) {
143 lookup_flags |= LOOKUP_REVAL;
149 SYSCALL_DEFINE2(truncate, const char __user *, path, long, length)
151 return do_sys_truncate(path, length);
155 COMPAT_SYSCALL_DEFINE2(truncate, const char __user *, path, compat_off_t, length)
157 return do_sys_truncate(path, length);
161 static long do_sys_ftruncate(unsigned int fd, loff_t length, int small)
164 struct dentry *dentry;
165 struct vfsmount *mnt;
177 /* explicitly opened as large or we are on 64-bit box */
178 if (f.file->f_flags & O_LARGEFILE)
181 dentry = f.file->f_path.dentry;
182 mnt = f.file->f_path.mnt;
183 inode = dentry->d_inode;
185 if (!S_ISREG(inode->i_mode) || !(f.file->f_mode & FMODE_WRITE))
189 /* Cannot ftruncate over 2^31 bytes without large file support */
190 if (small && length > MAX_NON_LFS)
194 if (IS_APPEND(inode))
197 sb_start_write(inode->i_sb);
198 error = locks_verify_truncate(inode, f.file, length);
200 error = security_path_truncate(&f.file->f_path);
202 error = do_truncate2(mnt, dentry, length, ATTR_MTIME|ATTR_CTIME, f.file);
203 sb_end_write(inode->i_sb);
210 SYSCALL_DEFINE2(ftruncate, unsigned int, fd, unsigned long, length)
212 return do_sys_ftruncate(fd, length, 1);
216 COMPAT_SYSCALL_DEFINE2(ftruncate, unsigned int, fd, compat_ulong_t, length)
218 return do_sys_ftruncate(fd, length, 1);
222 /* LFS versions of truncate are only needed on 32 bit machines */
223 #if BITS_PER_LONG == 32
224 SYSCALL_DEFINE2(truncate64, const char __user *, path, loff_t, length)
226 return do_sys_truncate(path, length);
229 SYSCALL_DEFINE2(ftruncate64, unsigned int, fd, loff_t, length)
231 return do_sys_ftruncate(fd, length, 0);
233 #endif /* BITS_PER_LONG == 32 */
236 int vfs_fallocate(struct file *file, int mode, loff_t offset, loff_t len)
238 struct inode *inode = file_inode(file);
241 if (offset < 0 || len <= 0)
244 /* Return error if mode is not supported */
245 if (mode & ~FALLOC_FL_SUPPORTED_MASK)
248 /* Punch hole and zero range are mutually exclusive */
249 if ((mode & (FALLOC_FL_PUNCH_HOLE | FALLOC_FL_ZERO_RANGE)) ==
250 (FALLOC_FL_PUNCH_HOLE | FALLOC_FL_ZERO_RANGE))
253 /* Punch hole must have keep size set */
254 if ((mode & FALLOC_FL_PUNCH_HOLE) &&
255 !(mode & FALLOC_FL_KEEP_SIZE))
258 /* Collapse range should only be used exclusively. */
259 if ((mode & FALLOC_FL_COLLAPSE_RANGE) &&
260 (mode & ~FALLOC_FL_COLLAPSE_RANGE))
263 /* Insert range should only be used exclusively. */
264 if ((mode & FALLOC_FL_INSERT_RANGE) &&
265 (mode & ~FALLOC_FL_INSERT_RANGE))
268 if (!(file->f_mode & FMODE_WRITE))
272 * We can only allow pure fallocate on append only files
274 if ((mode & ~FALLOC_FL_KEEP_SIZE) && IS_APPEND(inode))
277 if (IS_IMMUTABLE(inode))
281 * We cannot allow any fallocate operation on an active swapfile
283 if (IS_SWAPFILE(inode))
287 * Revalidate the write permissions, in case security policy has
288 * changed since the files were opened.
290 ret = security_file_permission(file, MAY_WRITE);
294 if (S_ISFIFO(inode->i_mode))
298 * Let individual file system decide if it supports preallocation
299 * for directories or not.
301 if (!S_ISREG(inode->i_mode) && !S_ISDIR(inode->i_mode))
304 /* Check for wrap through zero too */
305 if (((offset + len) > inode->i_sb->s_maxbytes) || ((offset + len) < 0))
308 if (!file->f_op->fallocate)
311 sb_start_write(inode->i_sb);
312 ret = file->f_op->fallocate(file, mode, offset, len);
315 * Create inotify and fanotify events.
317 * To keep the logic simple always create events if fallocate succeeds.
318 * This implies that events are even created if the file size remains
319 * unchanged, e.g. when using flag FALLOC_FL_KEEP_SIZE.
322 fsnotify_modify(file);
324 sb_end_write(inode->i_sb);
327 EXPORT_SYMBOL_GPL(vfs_fallocate);
329 SYSCALL_DEFINE4(fallocate, int, fd, int, mode, loff_t, offset, loff_t, len)
331 struct fd f = fdget(fd);
335 error = vfs_fallocate(f.file, mode, offset, len);
342 * access() needs to use the real uid/gid, not the effective uid/gid.
343 * We do this by temporarily clearing all FS-related capabilities and
344 * switching the fsuid/fsgid around to the real ones.
346 SYSCALL_DEFINE3(faccessat, int, dfd, const char __user *, filename, int, mode)
348 const struct cred *old_cred;
349 struct cred *override_cred;
352 struct vfsmount *mnt;
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);
404 if ((mode & MAY_EXEC) && S_ISREG(inode->i_mode)) {
406 * MAY_EXEC on regular files is denied if the fs is mounted
407 * with the "noexec" flag.
410 if (path_noexec(&path))
411 goto out_path_release;
414 res = inode_permission2(mnt, inode, mode | MAY_ACCESS);
415 /* SuS v2 requires we report a read only fs too */
416 if (res || !(mode & S_IWOTH) || special_file(inode->i_mode))
417 goto out_path_release;
419 * This is a rare case where using __mnt_is_readonly()
420 * is OK without a mnt_want/drop_write() pair. Since
421 * no actual write to the fs is performed here, we do
422 * not need to telegraph to that to anyone.
424 * By doing this, we accept that this access is
425 * inherently racy and know that the fs may change
426 * state before we even see this result.
428 if (__mnt_is_readonly(path.mnt))
433 if (retry_estale(res, lookup_flags)) {
434 lookup_flags |= LOOKUP_REVAL;
438 revert_creds(old_cred);
439 put_cred(override_cred);
443 SYSCALL_DEFINE2(access, const char __user *, filename, int, mode)
445 return sys_faccessat(AT_FDCWD, filename, mode);
448 SYSCALL_DEFINE1(chdir, const char __user *, filename)
452 unsigned int lookup_flags = LOOKUP_FOLLOW | LOOKUP_DIRECTORY;
454 error = user_path_at(AT_FDCWD, filename, lookup_flags, &path);
458 error = inode_permission2(path.mnt, path.dentry->d_inode, MAY_EXEC | MAY_CHDIR);
462 set_fs_pwd(current->fs, &path);
466 if (retry_estale(error, lookup_flags)) {
467 lookup_flags |= LOOKUP_REVAL;
474 SYSCALL_DEFINE1(fchdir, unsigned int, fd)
476 struct fd f = fdget_raw(fd);
478 struct vfsmount *mnt;
485 inode = file_inode(f.file);
486 mnt = f.file->f_path.mnt;
489 if (!S_ISDIR(inode->i_mode))
492 error = inode_permission2(mnt, inode, MAY_EXEC | MAY_CHDIR);
494 set_fs_pwd(current->fs, &f.file->f_path);
501 SYSCALL_DEFINE1(chroot, const char __user *, filename)
505 unsigned int lookup_flags = LOOKUP_FOLLOW | LOOKUP_DIRECTORY;
507 error = user_path_at(AT_FDCWD, filename, lookup_flags, &path);
511 error = inode_permission2(path.mnt, path.dentry->d_inode, MAY_EXEC | MAY_CHDIR);
516 if (!ns_capable(current_user_ns(), CAP_SYS_CHROOT))
518 error = security_path_chroot(&path);
522 set_fs_root(current->fs, &path);
526 if (retry_estale(error, lookup_flags)) {
527 lookup_flags |= LOOKUP_REVAL;
534 static int chmod_common(struct path *path, umode_t mode)
536 struct inode *inode = path->dentry->d_inode;
537 struct inode *delegated_inode = NULL;
538 struct iattr newattrs;
541 error = mnt_want_write(path->mnt);
545 mutex_lock(&inode->i_mutex);
546 error = security_path_chmod(path, mode);
549 newattrs.ia_mode = (mode & S_IALLUGO) | (inode->i_mode & ~S_IALLUGO);
550 newattrs.ia_valid = ATTR_MODE | ATTR_CTIME;
551 error = notify_change2(path->mnt, path->dentry, &newattrs, &delegated_inode);
553 mutex_unlock(&inode->i_mutex);
554 if (delegated_inode) {
555 error = break_deleg_wait(&delegated_inode);
559 mnt_drop_write(path->mnt);
563 SYSCALL_DEFINE2(fchmod, unsigned int, fd, umode_t, mode)
565 struct fd f = fdget(fd);
570 err = chmod_common(&f.file->f_path, mode);
576 SYSCALL_DEFINE3(fchmodat, int, dfd, const char __user *, filename, umode_t, mode)
580 unsigned int lookup_flags = LOOKUP_FOLLOW;
582 error = user_path_at(dfd, filename, lookup_flags, &path);
584 error = chmod_common(&path, mode);
586 if (retry_estale(error, lookup_flags)) {
587 lookup_flags |= LOOKUP_REVAL;
594 SYSCALL_DEFINE2(chmod, const char __user *, filename, umode_t, mode)
596 return sys_fchmodat(AT_FDCWD, filename, mode);
599 static int chown_common(struct path *path, uid_t user, gid_t group)
601 struct inode *inode = path->dentry->d_inode;
602 struct inode *delegated_inode = NULL;
604 struct iattr newattrs;
608 uid = make_kuid(current_user_ns(), user);
609 gid = make_kgid(current_user_ns(), group);
612 newattrs.ia_valid = ATTR_CTIME;
613 if (user != (uid_t) -1) {
616 newattrs.ia_valid |= ATTR_UID;
617 newattrs.ia_uid = uid;
619 if (group != (gid_t) -1) {
622 newattrs.ia_valid |= ATTR_GID;
623 newattrs.ia_gid = gid;
625 if (!S_ISDIR(inode->i_mode))
627 ATTR_KILL_SUID | ATTR_KILL_SGID | ATTR_KILL_PRIV;
628 mutex_lock(&inode->i_mutex);
629 error = security_path_chown(path, uid, gid);
631 error = notify_change2(path->mnt, path->dentry, &newattrs, &delegated_inode);
632 mutex_unlock(&inode->i_mutex);
633 if (delegated_inode) {
634 error = break_deleg_wait(&delegated_inode);
641 SYSCALL_DEFINE5(fchownat, int, dfd, const char __user *, filename, uid_t, user,
642 gid_t, group, int, flag)
648 if ((flag & ~(AT_SYMLINK_NOFOLLOW | AT_EMPTY_PATH)) != 0)
651 lookup_flags = (flag & AT_SYMLINK_NOFOLLOW) ? 0 : LOOKUP_FOLLOW;
652 if (flag & AT_EMPTY_PATH)
653 lookup_flags |= LOOKUP_EMPTY;
655 error = user_path_at(dfd, filename, lookup_flags, &path);
658 error = mnt_want_write(path.mnt);
661 error = chown_common(&path, user, group);
662 mnt_drop_write(path.mnt);
665 if (retry_estale(error, lookup_flags)) {
666 lookup_flags |= LOOKUP_REVAL;
673 SYSCALL_DEFINE3(chown, const char __user *, filename, uid_t, user, gid_t, group)
675 return sys_fchownat(AT_FDCWD, filename, user, group, 0);
678 SYSCALL_DEFINE3(lchown, const char __user *, filename, uid_t, user, gid_t, group)
680 return sys_fchownat(AT_FDCWD, filename, user, group,
681 AT_SYMLINK_NOFOLLOW);
684 SYSCALL_DEFINE3(fchown, unsigned int, fd, uid_t, user, gid_t, group)
686 struct fd f = fdget(fd);
692 error = mnt_want_write_file(f.file);
696 error = chown_common(&f.file->f_path, user, group);
697 mnt_drop_write_file(f.file);
704 int open_check_o_direct(struct file *f)
706 /* NB: we're sure to have correct a_ops only after f_op->open */
707 if (f->f_flags & O_DIRECT) {
708 if (!f->f_mapping->a_ops || !f->f_mapping->a_ops->direct_IO)
714 static int do_dentry_open(struct file *f,
716 int (*open)(struct inode *, struct file *),
717 const struct cred *cred)
719 static const struct file_operations empty_fops = {};
722 f->f_mode = OPEN_FMODE(f->f_flags) | FMODE_LSEEK |
723 FMODE_PREAD | FMODE_PWRITE;
725 path_get(&f->f_path);
727 f->f_mapping = inode->i_mapping;
729 if (unlikely(f->f_flags & O_PATH)) {
730 f->f_mode = FMODE_PATH;
731 f->f_op = &empty_fops;
735 if (f->f_mode & FMODE_WRITE && !special_file(inode->i_mode)) {
736 error = get_write_access(inode);
739 error = __mnt_want_write(f->f_path.mnt);
740 if (unlikely(error)) {
741 put_write_access(inode);
744 f->f_mode |= FMODE_WRITER;
747 /* POSIX.1-2008/SUSv4 Section XSI 2.9.7 */
748 if (S_ISREG(inode->i_mode))
749 f->f_mode |= FMODE_ATOMIC_POS;
751 f->f_op = fops_get(inode->i_fop);
752 if (unlikely(WARN_ON(!f->f_op))) {
757 error = security_file_open(f, cred);
761 error = break_lease(inode, f->f_flags);
766 open = f->f_op->open;
768 error = open(inode, f);
772 if ((f->f_mode & (FMODE_READ | FMODE_WRITE)) == FMODE_READ)
773 i_readcount_inc(inode);
774 if ((f->f_mode & FMODE_READ) &&
775 likely(f->f_op->read || f->f_op->read_iter))
776 f->f_mode |= FMODE_CAN_READ;
777 if ((f->f_mode & FMODE_WRITE) &&
778 likely(f->f_op->write || f->f_op->write_iter))
779 f->f_mode |= FMODE_CAN_WRITE;
781 f->f_flags &= ~(O_CREAT | O_EXCL | O_NOCTTY | O_TRUNC);
783 file_ra_state_init(&f->f_ra, f->f_mapping->host->i_mapping);
789 if (f->f_mode & FMODE_WRITER) {
790 put_write_access(inode);
791 __mnt_drop_write(f->f_path.mnt);
794 path_put(&f->f_path);
795 f->f_path.mnt = NULL;
796 f->f_path.dentry = NULL;
802 * finish_open - finish opening a file
803 * @file: file pointer
804 * @dentry: pointer to dentry
805 * @open: open callback
806 * @opened: state of open
808 * This can be used to finish opening a file passed to i_op->atomic_open().
810 * If the open callback is set to NULL, then the standard f_op->open()
811 * filesystem callback is substituted.
813 * NB: the dentry reference is _not_ consumed. If, for example, the dentry is
814 * the return value of d_splice_alias(), then the caller needs to perform dput()
815 * on it after finish_open().
817 * On successful return @file is a fully instantiated open file. After this, if
818 * an error occurs in ->atomic_open(), it needs to clean up with fput().
820 * Returns zero on success or -errno if the open failed.
822 int finish_open(struct file *file, struct dentry *dentry,
823 int (*open)(struct inode *, struct file *),
827 BUG_ON(*opened & FILE_OPENED); /* once it's opened, it's opened */
829 file->f_path.dentry = dentry;
830 error = do_dentry_open(file, d_backing_inode(dentry), open,
833 *opened |= FILE_OPENED;
837 EXPORT_SYMBOL(finish_open);
840 * finish_no_open - finish ->atomic_open() without opening the file
842 * @file: file pointer
843 * @dentry: dentry or NULL (as returned from ->lookup())
845 * This can be used to set the result of a successful lookup in ->atomic_open().
847 * NB: unlike finish_open() this function does consume the dentry reference and
848 * the caller need not dput() it.
850 * Returns "1" which must be the return value of ->atomic_open() after having
851 * called this function.
853 int finish_no_open(struct file *file, struct dentry *dentry)
855 file->f_path.dentry = dentry;
858 EXPORT_SYMBOL(finish_no_open);
860 char *file_path(struct file *filp, char *buf, int buflen)
862 return d_path(&filp->f_path, buf, buflen);
864 EXPORT_SYMBOL(file_path);
867 * vfs_open - open the file at the given path
868 * @path: path to open
869 * @file: newly allocated file with f_flag initialized
870 * @cred: credentials to use
872 int vfs_open(const struct path *path, struct file *file,
873 const struct cred *cred)
875 struct inode *inode = vfs_select_inode(path->dentry, file->f_flags);
878 return PTR_ERR(inode);
880 file->f_path = *path;
881 return do_dentry_open(file, inode, NULL, cred);
884 struct file *dentry_open(const struct path *path, int flags,
885 const struct cred *cred)
890 validate_creds(cred);
892 /* We must always pass in a valid mount pointer. */
895 f = get_empty_filp();
898 error = vfs_open(path, f, cred);
900 /* from now on we need fput() to dispose of f */
901 error = open_check_o_direct(f);
913 EXPORT_SYMBOL(dentry_open);
915 static inline int build_open_flags(int flags, umode_t mode, struct open_flags *op)
917 int lookup_flags = 0;
921 * Clear out all open flags we don't know about so that we don't report
922 * them in fcntl(F_GETFD) or similar interfaces.
924 flags &= VALID_OPEN_FLAGS;
926 if (flags & (O_CREAT | __O_TMPFILE))
927 op->mode = (mode & S_IALLUGO) | S_IFREG;
931 /* Must never be set by userspace */
932 flags &= ~FMODE_NONOTIFY & ~O_CLOEXEC;
935 * O_SYNC is implemented as __O_SYNC|O_DSYNC. As many places only
936 * check for O_DSYNC if the need any syncing at all we enforce it's
937 * always set instead of having to deal with possibly weird behaviour
938 * for malicious applications setting only __O_SYNC.
940 if (flags & __O_SYNC)
943 if (flags & __O_TMPFILE) {
944 if ((flags & O_TMPFILE_MASK) != O_TMPFILE)
946 acc_mode = MAY_OPEN | ACC_MODE(flags);
947 if (!(acc_mode & MAY_WRITE))
949 } else if (flags & O_PATH) {
951 * If we have O_PATH in the open flag. Then we
952 * cannot have anything other than the below set of flags
954 flags &= O_DIRECTORY | O_NOFOLLOW | O_PATH;
957 acc_mode = MAY_OPEN | ACC_MODE(flags);
960 op->open_flag = flags;
962 /* O_TRUNC implies we need access checks for write permissions */
964 acc_mode |= MAY_WRITE;
966 /* Allow the LSM permission hook to distinguish append
967 access from general write access. */
968 if (flags & O_APPEND)
969 acc_mode |= MAY_APPEND;
971 op->acc_mode = acc_mode;
973 op->intent = flags & O_PATH ? 0 : LOOKUP_OPEN;
975 if (flags & O_CREAT) {
976 op->intent |= LOOKUP_CREATE;
978 op->intent |= LOOKUP_EXCL;
981 if (flags & O_DIRECTORY)
982 lookup_flags |= LOOKUP_DIRECTORY;
983 if (!(flags & O_NOFOLLOW))
984 lookup_flags |= LOOKUP_FOLLOW;
985 op->lookup_flags = lookup_flags;
990 * file_open_name - open file and return file pointer
992 * @name: struct filename containing path to open
993 * @flags: open flags as per the open(2) second argument
994 * @mode: mode for the new file if O_CREAT is set, else ignored
996 * This is the helper to open a file from kernelspace if you really
997 * have to. But in generally you should not do this, so please move
998 * along, nothing to see here..
1000 struct file *file_open_name(struct filename *name, int flags, umode_t mode)
1002 struct open_flags op;
1003 int err = build_open_flags(flags, mode, &op);
1004 return err ? ERR_PTR(err) : do_filp_open(AT_FDCWD, name, &op);
1008 * filp_open - open file and return file pointer
1010 * @filename: path to open
1011 * @flags: open flags as per the open(2) second argument
1012 * @mode: mode for the new file if O_CREAT is set, else ignored
1014 * This is the helper to open a file from kernelspace if you really
1015 * have to. But in generally you should not do this, so please move
1016 * along, nothing to see here..
1018 struct file *filp_open(const char *filename, int flags, umode_t mode)
1020 struct filename *name = getname_kernel(filename);
1021 struct file *file = ERR_CAST(name);
1023 if (!IS_ERR(name)) {
1024 file = file_open_name(name, flags, mode);
1029 EXPORT_SYMBOL(filp_open);
1031 struct file *file_open_root(struct dentry *dentry, struct vfsmount *mnt,
1032 const char *filename, int flags, umode_t mode)
1034 struct open_flags op;
1035 int err = build_open_flags(flags, mode, &op);
1037 return ERR_PTR(err);
1038 return do_file_open_root(dentry, mnt, filename, &op);
1040 EXPORT_SYMBOL(file_open_root);
1042 long do_sys_open(int dfd, const char __user *filename, int flags, umode_t mode)
1044 struct open_flags op;
1045 int fd = build_open_flags(flags, mode, &op);
1046 struct filename *tmp;
1051 tmp = getname(filename);
1053 return PTR_ERR(tmp);
1055 fd = get_unused_fd_flags(flags);
1057 struct file *f = do_filp_open(dfd, tmp, &op);
1070 SYSCALL_DEFINE3(open, const char __user *, filename, int, flags, umode_t, mode)
1072 if (force_o_largefile())
1073 flags |= O_LARGEFILE;
1075 return do_sys_open(AT_FDCWD, filename, flags, mode);
1078 SYSCALL_DEFINE4(openat, int, dfd, const char __user *, filename, int, flags,
1081 if (force_o_largefile())
1082 flags |= O_LARGEFILE;
1084 return do_sys_open(dfd, filename, flags, mode);
1090 * For backward compatibility? Maybe this should be moved
1091 * into arch/i386 instead?
1093 SYSCALL_DEFINE2(creat, const char __user *, pathname, umode_t, mode)
1095 return sys_open(pathname, O_CREAT | O_WRONLY | O_TRUNC, mode);
1101 * "id" is the POSIX thread ID. We use the
1102 * files pointer for this..
1104 int filp_close(struct file *filp, fl_owner_t id)
1108 if (!file_count(filp)) {
1109 printk(KERN_ERR "VFS: Close: file count is 0\n");
1113 if (filp->f_op->flush)
1114 retval = filp->f_op->flush(filp, id);
1116 if (likely(!(filp->f_mode & FMODE_PATH))) {
1117 dnotify_flush(filp, id);
1118 locks_remove_posix(filp, id);
1124 EXPORT_SYMBOL(filp_close);
1127 * Careful here! We test whether the file pointer is NULL before
1128 * releasing the fd. This ensures that one clone task can't release
1129 * an fd while another clone is opening it.
1131 SYSCALL_DEFINE1(close, unsigned int, fd)
1133 int retval = __close_fd(current->files, fd);
1135 /* can't restart close syscall because file table entry was cleared */
1136 if (unlikely(retval == -ERESTARTSYS ||
1137 retval == -ERESTARTNOINTR ||
1138 retval == -ERESTARTNOHAND ||
1139 retval == -ERESTART_RESTARTBLOCK))
1144 EXPORT_SYMBOL(sys_close);
1147 * This routine simulates a hangup on the tty, to arrange that users
1148 * are given clean terminals at login time.
1150 SYSCALL_DEFINE0(vhangup)
1152 if (capable(CAP_SYS_TTY_CONFIG)) {
1160 * Called when an inode is about to be open.
1161 * We use this to disallow opening large files on 32bit systems if
1162 * the caller didn't specify O_LARGEFILE. On 64bit systems we force
1163 * on this flag in sys_open.
1165 int generic_file_open(struct inode * inode, struct file * filp)
1167 if (!(filp->f_flags & O_LARGEFILE) && i_size_read(inode) > MAX_NON_LFS)
1172 EXPORT_SYMBOL(generic_file_open);
1175 * This is used by subsystems that don't want seekable
1176 * file descriptors. The function is not supposed to ever fail, the only
1177 * reason it returns an 'int' and not 'void' is so that it can be plugged
1178 * directly into file_operations structure.
1180 int nonseekable_open(struct inode *inode, struct file *filp)
1182 filp->f_mode &= ~(FMODE_LSEEK | FMODE_PREAD | FMODE_PWRITE);
1186 EXPORT_SYMBOL(nonseekable_open);
1189 * stream_open is used by subsystems that want stream-like file descriptors.
1190 * Such file descriptors are not seekable and don't have notion of position
1191 * (file.f_pos is always 0). Contrary to file descriptors of other regular
1192 * files, .read() and .write() can run simultaneously.
1194 * stream_open never fails and is marked to return int so that it could be
1195 * directly used as file_operations.open .
1197 int stream_open(struct inode *inode, struct file *filp)
1199 filp->f_mode &= ~(FMODE_LSEEK | FMODE_PREAD | FMODE_PWRITE | FMODE_ATOMIC_POS);
1200 filp->f_mode |= FMODE_STREAM;
1204 EXPORT_SYMBOL(stream_open);