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/quotaops.h>
12 #include <linux/fsnotify.h>
13 #include <linux/module.h>
14 #include <linux/slab.h>
15 #include <linux/tty.h>
16 #include <linux/namei.h>
17 #include <linux/backing-dev.h>
18 #include <linux/capability.h>
19 #include <linux/securebits.h>
20 #include <linux/security.h>
21 #include <linux/mount.h>
22 #include <linux/vfs.h>
23 #include <linux/fcntl.h>
24 #include <asm/uaccess.h>
26 #include <linux/personality.h>
27 #include <linux/pagemap.h>
28 #include <linux/syscalls.h>
29 #include <linux/rcupdate.h>
30 #include <linux/audit.h>
31 #include <linux/falloc.h>
32 #include <linux/fs_struct.h>
33 #include <linux/ima.h>
38 DEFINE_TRACE(fs_open);
39 DEFINE_TRACE(fs_close);
41 int vfs_statfs(struct dentry *dentry, struct kstatfs *buf)
47 if (dentry->d_sb->s_op->statfs) {
48 memset(buf, 0, sizeof(*buf));
49 retval = security_sb_statfs(dentry);
52 retval = dentry->d_sb->s_op->statfs(dentry, buf);
53 if (retval == 0 && buf->f_frsize == 0)
54 buf->f_frsize = buf->f_bsize;
60 EXPORT_SYMBOL(vfs_statfs);
62 static int vfs_statfs_native(struct dentry *dentry, struct statfs *buf)
67 retval = vfs_statfs(dentry, &st);
71 if (sizeof(*buf) == sizeof(st))
72 memcpy(buf, &st, sizeof(st));
74 if (sizeof buf->f_blocks == 4) {
75 if ((st.f_blocks | st.f_bfree | st.f_bavail |
76 st.f_bsize | st.f_frsize) &
77 0xffffffff00000000ULL)
80 * f_files and f_ffree may be -1; it's okay to stuff
83 if (st.f_files != -1 &&
84 (st.f_files & 0xffffffff00000000ULL))
86 if (st.f_ffree != -1 &&
87 (st.f_ffree & 0xffffffff00000000ULL))
91 buf->f_type = st.f_type;
92 buf->f_bsize = st.f_bsize;
93 buf->f_blocks = st.f_blocks;
94 buf->f_bfree = st.f_bfree;
95 buf->f_bavail = st.f_bavail;
96 buf->f_files = st.f_files;
97 buf->f_ffree = st.f_ffree;
98 buf->f_fsid = st.f_fsid;
99 buf->f_namelen = st.f_namelen;
100 buf->f_frsize = st.f_frsize;
101 memset(buf->f_spare, 0, sizeof(buf->f_spare));
106 static int vfs_statfs64(struct dentry *dentry, struct statfs64 *buf)
111 retval = vfs_statfs(dentry, &st);
115 if (sizeof(*buf) == sizeof(st))
116 memcpy(buf, &st, sizeof(st));
118 buf->f_type = st.f_type;
119 buf->f_bsize = st.f_bsize;
120 buf->f_blocks = st.f_blocks;
121 buf->f_bfree = st.f_bfree;
122 buf->f_bavail = st.f_bavail;
123 buf->f_files = st.f_files;
124 buf->f_ffree = st.f_ffree;
125 buf->f_fsid = st.f_fsid;
126 buf->f_namelen = st.f_namelen;
127 buf->f_frsize = st.f_frsize;
128 memset(buf->f_spare, 0, sizeof(buf->f_spare));
133 SYSCALL_DEFINE2(statfs, const char __user *, pathname, struct statfs __user *, buf)
138 error = user_path(pathname, &path);
141 error = vfs_statfs_native(path.dentry, &tmp);
142 if (!error && copy_to_user(buf, &tmp, sizeof(tmp)))
149 SYSCALL_DEFINE3(statfs64, const char __user *, pathname, size_t, sz, struct statfs64 __user *, buf)
154 if (sz != sizeof(*buf))
156 error = user_path(pathname, &path);
159 error = vfs_statfs64(path.dentry, &tmp);
160 if (!error && copy_to_user(buf, &tmp, sizeof(tmp)))
167 SYSCALL_DEFINE2(fstatfs, unsigned int, fd, struct statfs __user *, buf)
177 error = vfs_statfs_native(file->f_path.dentry, &tmp);
178 if (!error && copy_to_user(buf, &tmp, sizeof(tmp)))
185 SYSCALL_DEFINE3(fstatfs64, unsigned int, fd, size_t, sz, struct statfs64 __user *, buf)
191 if (sz != sizeof(*buf))
198 error = vfs_statfs64(file->f_path.dentry, &tmp);
199 if (!error && copy_to_user(buf, &tmp, sizeof(tmp)))
206 int do_truncate(struct dentry *dentry, loff_t length, unsigned int time_attrs,
210 struct iattr newattrs;
212 /* Not pretty: "inode->i_size" shouldn't really be signed. But it is. */
216 newattrs.ia_size = length;
217 newattrs.ia_valid = ATTR_SIZE | time_attrs;
219 newattrs.ia_file = filp;
220 newattrs.ia_valid |= ATTR_FILE;
223 /* Remove suid/sgid on truncate too */
224 ret = should_remove_suid(dentry);
226 newattrs.ia_valid |= ret | ATTR_FORCE;
228 mutex_lock(&dentry->d_inode->i_mutex);
229 ret = notify_change(dentry, &newattrs);
230 mutex_unlock(&dentry->d_inode->i_mutex);
234 static long do_sys_truncate(const char __user *pathname, loff_t length)
241 if (length < 0) /* sorry, but loff_t says... */
244 error = user_path(pathname, &path);
247 inode = path.dentry->d_inode;
249 /* For directories it's -EISDIR, for other non-regulars - -EINVAL */
251 if (S_ISDIR(inode->i_mode))
255 if (!S_ISREG(inode->i_mode))
258 error = mnt_want_write(path.mnt);
262 error = inode_permission(inode, MAY_WRITE);
264 goto mnt_drop_write_and_out;
267 if (IS_APPEND(inode))
268 goto mnt_drop_write_and_out;
270 error = get_write_access(inode);
272 goto mnt_drop_write_and_out;
275 * Make sure that there are no leases. get_write_access() protects
276 * against the truncate racing with a lease-granting setlease().
278 error = break_lease(inode, FMODE_WRITE);
280 goto put_write_and_out;
282 error = locks_verify_truncate(inode, NULL, length);
284 error = security_path_truncate(&path, length, 0);
287 error = do_truncate(path.dentry, length, 0, NULL);
291 put_write_access(inode);
292 mnt_drop_write_and_out:
293 mnt_drop_write(path.mnt);
300 SYSCALL_DEFINE2(truncate, const char __user *, path, long, length)
302 return do_sys_truncate(path, length);
305 static long do_sys_ftruncate(unsigned int fd, loff_t length, int small)
307 struct inode * inode;
308 struct dentry *dentry;
320 /* explicitly opened as large or we are on 64-bit box */
321 if (file->f_flags & O_LARGEFILE)
324 dentry = file->f_path.dentry;
325 inode = dentry->d_inode;
327 if (!S_ISREG(inode->i_mode) || !(file->f_mode & FMODE_WRITE))
331 /* Cannot ftruncate over 2^31 bytes without large file support */
332 if (small && length > MAX_NON_LFS)
336 if (IS_APPEND(inode))
339 error = locks_verify_truncate(inode, file, length);
341 error = security_path_truncate(&file->f_path, length,
342 ATTR_MTIME|ATTR_CTIME);
344 error = do_truncate(dentry, length, ATTR_MTIME|ATTR_CTIME, file);
351 SYSCALL_DEFINE2(ftruncate, unsigned int, fd, unsigned long, length)
353 long ret = do_sys_ftruncate(fd, length, 1);
354 /* avoid REGPARM breakage on x86: */
355 asmlinkage_protect(2, ret, fd, length);
359 /* LFS versions of truncate are only needed on 32 bit machines */
360 #if BITS_PER_LONG == 32
361 SYSCALL_DEFINE(truncate64)(const char __user * path, loff_t length)
363 return do_sys_truncate(path, length);
365 #ifdef CONFIG_HAVE_SYSCALL_WRAPPERS
366 asmlinkage long SyS_truncate64(long path, loff_t length)
368 return SYSC_truncate64((const char __user *) path, length);
370 SYSCALL_ALIAS(sys_truncate64, SyS_truncate64);
373 SYSCALL_DEFINE(ftruncate64)(unsigned int fd, loff_t length)
375 long ret = do_sys_ftruncate(fd, length, 0);
376 /* avoid REGPARM breakage on x86: */
377 asmlinkage_protect(2, ret, fd, length);
380 #ifdef CONFIG_HAVE_SYSCALL_WRAPPERS
381 asmlinkage long SyS_ftruncate64(long fd, loff_t length)
383 return SYSC_ftruncate64((unsigned int) fd, length);
385 SYSCALL_ALIAS(sys_ftruncate64, SyS_ftruncate64);
387 #endif /* BITS_PER_LONG == 32 */
390 int do_fallocate(struct file *file, int mode, loff_t offset, loff_t len)
392 struct inode *inode = file->f_path.dentry->d_inode;
395 if (offset < 0 || len <= 0)
398 /* Return error if mode is not supported */
399 if (mode && !(mode & FALLOC_FL_KEEP_SIZE))
402 if (!(file->f_mode & FMODE_WRITE))
405 * Revalidate the write permissions, in case security policy has
406 * changed since the files were opened.
408 ret = security_file_permission(file, MAY_WRITE);
412 if (S_ISFIFO(inode->i_mode))
416 * Let individual file system decide if it supports preallocation
417 * for directories or not.
419 if (!S_ISREG(inode->i_mode) && !S_ISDIR(inode->i_mode))
422 /* Check for wrap through zero too */
423 if (((offset + len) > inode->i_sb->s_maxbytes) || ((offset + len) < 0))
426 if (!inode->i_op->fallocate)
429 return inode->i_op->fallocate(inode, mode, offset, len);
432 SYSCALL_DEFINE(fallocate)(int fd, int mode, loff_t offset, loff_t len)
439 error = do_fallocate(file, mode, offset, len);
446 #ifdef CONFIG_HAVE_SYSCALL_WRAPPERS
447 asmlinkage long SyS_fallocate(long fd, long mode, loff_t offset, loff_t len)
449 return SYSC_fallocate((int)fd, (int)mode, offset, len);
451 SYSCALL_ALIAS(sys_fallocate, SyS_fallocate);
455 * access() needs to use the real uid/gid, not the effective uid/gid.
456 * We do this by temporarily clearing all FS-related capabilities and
457 * switching the fsuid/fsgid around to the real ones.
459 SYSCALL_DEFINE3(faccessat, int, dfd, const char __user *, filename, int, mode)
461 const struct cred *old_cred;
462 struct cred *override_cred;
467 if (mode & ~S_IRWXO) /* where's F_OK, X_OK, W_OK, R_OK? */
470 override_cred = prepare_creds();
474 override_cred->fsuid = override_cred->uid;
475 override_cred->fsgid = override_cred->gid;
477 if (!issecure(SECURE_NO_SETUID_FIXUP)) {
478 /* Clear the capabilities if we switch to a non-root user */
479 if (override_cred->uid)
480 cap_clear(override_cred->cap_effective);
482 override_cred->cap_effective =
483 override_cred->cap_permitted;
486 old_cred = override_creds(override_cred);
488 res = user_path_at(dfd, filename, LOOKUP_FOLLOW, &path);
492 inode = path.dentry->d_inode;
494 if ((mode & MAY_EXEC) && S_ISREG(inode->i_mode)) {
496 * MAY_EXEC on regular files is denied if the fs is mounted
497 * with the "noexec" flag.
500 if (path.mnt->mnt_flags & MNT_NOEXEC)
501 goto out_path_release;
504 res = inode_permission(inode, mode | MAY_ACCESS);
505 /* SuS v2 requires we report a read only fs too */
506 if (res || !(mode & S_IWOTH) || special_file(inode->i_mode))
507 goto out_path_release;
509 * This is a rare case where using __mnt_is_readonly()
510 * is OK without a mnt_want/drop_write() pair. Since
511 * no actual write to the fs is performed here, we do
512 * not need to telegraph to that to anyone.
514 * By doing this, we accept that this access is
515 * inherently racy and know that the fs may change
516 * state before we even see this result.
518 if (__mnt_is_readonly(path.mnt))
524 revert_creds(old_cred);
525 put_cred(override_cred);
529 SYSCALL_DEFINE2(access, const char __user *, filename, int, mode)
531 return sys_faccessat(AT_FDCWD, filename, mode);
534 SYSCALL_DEFINE1(chdir, const char __user *, filename)
539 error = user_path_dir(filename, &path);
543 error = inode_permission(path.dentry->d_inode, MAY_EXEC | MAY_ACCESS);
547 set_fs_pwd(current->fs, &path);
555 SYSCALL_DEFINE1(fchdir, unsigned int, fd)
566 inode = file->f_path.dentry->d_inode;
569 if (!S_ISDIR(inode->i_mode))
572 error = inode_permission(inode, MAY_EXEC | MAY_ACCESS);
574 set_fs_pwd(current->fs, &file->f_path);
581 SYSCALL_DEFINE1(chroot, const char __user *, filename)
586 error = user_path_dir(filename, &path);
590 error = inode_permission(path.dentry->d_inode, MAY_EXEC | MAY_ACCESS);
595 if (!capable(CAP_SYS_CHROOT))
597 error = security_path_chroot(&path);
601 set_fs_root(current->fs, &path);
609 SYSCALL_DEFINE2(fchmod, unsigned int, fd, mode_t, mode)
611 struct inode * inode;
612 struct dentry * dentry;
615 struct iattr newattrs;
621 dentry = file->f_path.dentry;
622 inode = dentry->d_inode;
624 audit_inode(NULL, dentry);
626 err = mnt_want_write_file(file);
629 mutex_lock(&inode->i_mutex);
630 err = security_path_chmod(dentry, file->f_vfsmnt, mode);
633 if (mode == (mode_t) -1)
634 mode = inode->i_mode;
635 newattrs.ia_mode = (mode & S_IALLUGO) | (inode->i_mode & ~S_IALLUGO);
636 newattrs.ia_valid = ATTR_MODE | ATTR_CTIME;
637 err = notify_change(dentry, &newattrs);
639 mutex_unlock(&inode->i_mutex);
640 mnt_drop_write(file->f_path.mnt);
647 SYSCALL_DEFINE3(fchmodat, int, dfd, const char __user *, filename, mode_t, mode)
652 struct iattr newattrs;
654 error = user_path_at(dfd, filename, LOOKUP_FOLLOW, &path);
657 inode = path.dentry->d_inode;
659 error = mnt_want_write(path.mnt);
662 mutex_lock(&inode->i_mutex);
663 error = security_path_chmod(path.dentry, path.mnt, mode);
666 if (mode == (mode_t) -1)
667 mode = inode->i_mode;
668 newattrs.ia_mode = (mode & S_IALLUGO) | (inode->i_mode & ~S_IALLUGO);
669 newattrs.ia_valid = ATTR_MODE | ATTR_CTIME;
670 error = notify_change(path.dentry, &newattrs);
672 mutex_unlock(&inode->i_mutex);
673 mnt_drop_write(path.mnt);
680 SYSCALL_DEFINE2(chmod, const char __user *, filename, mode_t, mode)
682 return sys_fchmodat(AT_FDCWD, filename, mode);
685 static int chown_common(struct path *path, uid_t user, gid_t group)
687 struct inode *inode = path->dentry->d_inode;
689 struct iattr newattrs;
691 newattrs.ia_valid = ATTR_CTIME;
692 if (user != (uid_t) -1) {
693 newattrs.ia_valid |= ATTR_UID;
694 newattrs.ia_uid = user;
696 if (group != (gid_t) -1) {
697 newattrs.ia_valid |= ATTR_GID;
698 newattrs.ia_gid = group;
700 if (!S_ISDIR(inode->i_mode))
702 ATTR_KILL_SUID | ATTR_KILL_SGID | ATTR_KILL_PRIV;
703 mutex_lock(&inode->i_mutex);
704 error = security_path_chown(path, user, group);
706 error = notify_change(path->dentry, &newattrs);
707 mutex_unlock(&inode->i_mutex);
712 SYSCALL_DEFINE3(chown, const char __user *, filename, uid_t, user, gid_t, group)
717 error = user_path(filename, &path);
720 error = mnt_want_write(path.mnt);
723 error = chown_common(&path, user, group);
724 mnt_drop_write(path.mnt);
731 SYSCALL_DEFINE5(fchownat, int, dfd, const char __user *, filename, uid_t, user,
732 gid_t, group, int, flag)
738 if ((flag & ~AT_SYMLINK_NOFOLLOW) != 0)
741 follow = (flag & AT_SYMLINK_NOFOLLOW) ? 0 : LOOKUP_FOLLOW;
742 error = user_path_at(dfd, filename, follow, &path);
745 error = mnt_want_write(path.mnt);
748 error = chown_common(&path, user, group);
749 mnt_drop_write(path.mnt);
756 SYSCALL_DEFINE3(lchown, const char __user *, filename, uid_t, user, gid_t, group)
761 error = user_lpath(filename, &path);
764 error = mnt_want_write(path.mnt);
767 error = chown_common(&path, user, group);
768 mnt_drop_write(path.mnt);
775 SYSCALL_DEFINE3(fchown, unsigned int, fd, uid_t, user, gid_t, group)
779 struct dentry * dentry;
785 error = mnt_want_write_file(file);
788 dentry = file->f_path.dentry;
789 audit_inode(NULL, dentry);
790 error = chown_common(&file->f_path, user, group);
791 mnt_drop_write(file->f_path.mnt);
799 * You have to be very careful that these write
800 * counts get cleaned up in error cases and
801 * upon __fput(). This should probably never
802 * be called outside of __dentry_open().
804 static inline int __get_file_write_access(struct inode *inode,
805 struct vfsmount *mnt)
808 error = get_write_access(inode);
812 * Do not take mount writer counts on
813 * special files since no writes to
814 * the mount itself will occur.
816 if (!special_file(inode->i_mode)) {
818 * Balanced in __fput()
820 error = mnt_want_write(mnt);
822 put_write_access(inode);
827 static struct file *__dentry_open(struct dentry *dentry, struct vfsmount *mnt,
829 int (*open)(struct inode *, struct file *),
830 const struct cred *cred)
835 f->f_mode = OPEN_FMODE(f->f_flags) | FMODE_LSEEK |
836 FMODE_PREAD | FMODE_PWRITE;
837 inode = dentry->d_inode;
838 if (f->f_mode & FMODE_WRITE) {
839 error = __get_file_write_access(inode, mnt);
842 if (!special_file(inode->i_mode))
846 f->f_mapping = inode->i_mapping;
847 f->f_path.dentry = dentry;
850 f->f_op = fops_get(inode->i_fop);
851 file_move(f, &inode->i_sb->s_files);
853 error = security_dentry_open(f, cred);
857 if (!open && f->f_op)
858 open = f->f_op->open;
860 error = open(inode, f);
866 f->f_flags &= ~(O_CREAT | O_EXCL | O_NOCTTY | O_TRUNC);
868 file_ra_state_init(&f->f_ra, f->f_mapping->host->i_mapping);
870 /* NB: we're sure to have correct a_ops only after f_op->open */
871 if (f->f_flags & O_DIRECT) {
872 if (!f->f_mapping->a_ops ||
873 ((!f->f_mapping->a_ops->direct_IO) &&
874 (!f->f_mapping->a_ops->get_xip_mem))) {
876 f = ERR_PTR(-EINVAL);
884 if (f->f_mode & FMODE_WRITE) {
885 put_write_access(inode);
886 if (!special_file(inode->i_mode)) {
888 * We don't consider this a real
889 * mnt_want/drop_write() pair
890 * because it all happenend right
891 * here, so just reset the state.
898 f->f_path.dentry = NULL;
899 f->f_path.mnt = NULL;
904 return ERR_PTR(error);
908 * lookup_instantiate_filp - instantiates the open intent filp
909 * @nd: pointer to nameidata
910 * @dentry: pointer to dentry
911 * @open: open callback
913 * Helper for filesystems that want to use lookup open intents and pass back
914 * a fully instantiated struct file to the caller.
915 * This function is meant to be called from within a filesystem's
917 * Beware of calling it for non-regular files! Those ->open methods might block
918 * (e.g. in fifo_open), leaving you with parent locked (and in case of fifo,
919 * leading to a deadlock, as nobody can open that fifo anymore, because
920 * another process to open fifo will block on locked parent when doing lookup).
921 * Note that in case of error, nd->intent.open.file is destroyed, but the
922 * path information remains valid.
923 * If the open callback is set to NULL, then the standard f_op->open()
924 * filesystem callback is substituted.
926 struct file *lookup_instantiate_filp(struct nameidata *nd, struct dentry *dentry,
927 int (*open)(struct inode *, struct file *))
929 const struct cred *cred = current_cred();
931 if (IS_ERR(nd->intent.open.file))
935 nd->intent.open.file = __dentry_open(dget(dentry), mntget(nd->path.mnt),
936 nd->intent.open.file,
939 return nd->intent.open.file;
941 release_open_intent(nd);
942 nd->intent.open.file = (struct file *)dentry;
945 EXPORT_SYMBOL_GPL(lookup_instantiate_filp);
948 * nameidata_to_filp - convert a nameidata to an open filp.
949 * @nd: pointer to nameidata
952 * Note that this function destroys the original nameidata
954 struct file *nameidata_to_filp(struct nameidata *nd)
956 const struct cred *cred = current_cred();
959 /* Pick up the filp from the open intent */
960 filp = nd->intent.open.file;
961 /* Has the filesystem initialised the file for us? */
962 if (filp->f_path.dentry == NULL)
963 filp = __dentry_open(nd->path.dentry, nd->path.mnt, filp,
971 * dentry_open() will have done dput(dentry) and mntput(mnt) if it returns an
974 struct file *dentry_open(struct dentry *dentry, struct vfsmount *mnt, int flags,
975 const struct cred *cred)
980 validate_creds(cred);
983 * We must always pass in a valid mount pointer. Historically
984 * callers got away with not passing it, but we must enforce this at
985 * the earliest possible point now to avoid strange problems deep in the
989 printk(KERN_WARNING "%s called with NULL vfsmount\n", __func__);
991 return ERR_PTR(-EINVAL);
995 f = get_empty_filp();
999 return ERR_PTR(error);
1003 return __dentry_open(dentry, mnt, f, NULL, cred);
1005 EXPORT_SYMBOL(dentry_open);
1007 static void __put_unused_fd(struct files_struct *files, unsigned int fd)
1009 struct fdtable *fdt = files_fdtable(files);
1010 __FD_CLR(fd, fdt->open_fds);
1011 if (fd < files->next_fd)
1012 files->next_fd = fd;
1015 void put_unused_fd(unsigned int fd)
1017 struct files_struct *files = current->files;
1018 spin_lock(&files->file_lock);
1019 __put_unused_fd(files, fd);
1020 spin_unlock(&files->file_lock);
1023 EXPORT_SYMBOL(put_unused_fd);
1026 * Install a file pointer in the fd array.
1028 * The VFS is full of places where we drop the files lock between
1029 * setting the open_fds bitmap and installing the file in the file
1030 * array. At any such point, we are vulnerable to a dup2() race
1031 * installing a file in the array before us. We need to detect this and
1032 * fput() the struct file we are about to overwrite in this case.
1034 * It should never happen - if we allow dup2() do it, _really_ bad things
1038 void fd_install(unsigned int fd, struct file *file)
1040 struct files_struct *files = current->files;
1041 struct fdtable *fdt;
1042 spin_lock(&files->file_lock);
1043 fdt = files_fdtable(files);
1044 BUG_ON(fdt->fd[fd] != NULL);
1045 rcu_assign_pointer(fdt->fd[fd], file);
1046 spin_unlock(&files->file_lock);
1049 EXPORT_SYMBOL(fd_install);
1051 long do_sys_open(int dfd, const char __user *filename, int flags, int mode)
1053 char *tmp = getname(filename);
1054 int fd = PTR_ERR(tmp);
1057 fd = get_unused_fd_flags(flags);
1059 struct file *f = do_filp_open(dfd, tmp, flags, mode, 0);
1064 fsnotify_open(f->f_path.dentry);
1067 trace_fs_open(fd, tmp);
1074 SYSCALL_DEFINE3(open, const char __user *, filename, int, flags, int, mode)
1078 if (force_o_largefile())
1079 flags |= O_LARGEFILE;
1081 ret = do_sys_open(AT_FDCWD, filename, flags, mode);
1082 /* avoid REGPARM breakage on x86: */
1083 asmlinkage_protect(3, ret, filename, flags, mode);
1087 SYSCALL_DEFINE4(openat, int, dfd, const char __user *, filename, int, flags,
1092 if (force_o_largefile())
1093 flags |= O_LARGEFILE;
1095 ret = do_sys_open(dfd, filename, flags, mode);
1096 /* avoid REGPARM breakage on x86: */
1097 asmlinkage_protect(4, ret, dfd, filename, flags, mode);
1104 * For backward compatibility? Maybe this should be moved
1105 * into arch/i386 instead?
1107 SYSCALL_DEFINE2(creat, const char __user *, pathname, int, mode)
1109 return sys_open(pathname, O_CREAT | O_WRONLY | O_TRUNC, mode);
1115 * "id" is the POSIX thread ID. We use the
1116 * files pointer for this..
1118 int filp_close(struct file *filp, fl_owner_t id)
1122 if (unlikely(!filp->f_count)) {
1123 printk(KERN_ERR "VFS: Close: file count is 0\n");
1127 if (filp->f_op && filp->f_op->flush)
1128 retval = filp->f_op->flush(filp, id);
1130 dnotify_flush(filp, id);
1131 locks_remove_posix(filp, id);
1136 EXPORT_SYMBOL(filp_close);
1139 * Careful here! We test whether the file pointer is NULL before
1140 * releasing the fd. This ensures that one clone task can't release
1141 * an fd while another clone is opening it.
1143 SYSCALL_DEFINE1(close, unsigned int, fd)
1146 struct files_struct *files = current->files;
1147 struct fdtable *fdt;
1150 spin_lock(&files->file_lock);
1151 fdt = files_fdtable(files);
1152 if (fd >= fdt->max_fds)
1158 rcu_assign_pointer(fdt->fd[fd], NULL);
1159 FD_CLR(fd, fdt->close_on_exec);
1160 __put_unused_fd(files, fd);
1161 spin_unlock(&files->file_lock);
1162 retval = filp_close(filp, files);
1164 /* can't restart close syscall because file table entry was cleared */
1165 if (unlikely(retval == -ERESTARTSYS ||
1166 retval == -ERESTARTNOINTR ||
1167 retval == -ERESTARTNOHAND ||
1168 retval == -ERESTART_RESTARTBLOCK))
1174 spin_unlock(&files->file_lock);
1177 EXPORT_SYMBOL(sys_close);
1180 * This routine simulates a hangup on the tty, to arrange that users
1181 * are given clean terminals at login time.
1183 SYSCALL_DEFINE0(vhangup)
1185 if (capable(CAP_SYS_TTY_CONFIG)) {
1193 * Called when an inode is about to be open.
1194 * We use this to disallow opening large files on 32bit systems if
1195 * the caller didn't specify O_LARGEFILE. On 64bit systems we force
1196 * on this flag in sys_open.
1198 int generic_file_open(struct inode * inode, struct file * filp)
1200 if (!(filp->f_flags & O_LARGEFILE) && i_size_read(inode) > MAX_NON_LFS)
1205 EXPORT_SYMBOL(generic_file_open);
1208 * This is used by subsystems that don't want seekable
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);