4 * Copyright (C) 1991, 1992 Linus Torvalds
7 #include <linux/string.h>
9 #include <linux/utime.h>
10 #include <linux/file.h>
11 #include <linux/smp_lock.h>
12 #include <linux/quotaops.h>
13 #include <linux/fsnotify.h>
14 #include <linux/module.h>
15 #include <linux/slab.h>
16 #include <linux/tty.h>
17 #include <linux/namei.h>
18 #include <linux/backing-dev.h>
19 #include <linux/capability.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/vs_limit.h>
31 #include <linux/vs_dlimit.h>
32 #include <linux/vserver/xid.h>
34 #include <asm/unistd.h>
36 int vfs_statfs(struct super_block *sb, struct kstatfs *buf)
42 if (sb->s_op->statfs) {
43 memset(buf, 0, sizeof(*buf));
44 retval = security_sb_statfs(sb);
47 retval = sb->s_op->statfs(sb, buf);
48 if (retval == 0 && buf->f_frsize == 0)
49 buf->f_frsize = buf->f_bsize;
51 if (!vx_check(0, VX_ADMIN|VX_WATCH))
52 vx_vsi_statfs(sb, buf);
57 EXPORT_SYMBOL(vfs_statfs);
59 static int vfs_statfs_native(struct super_block *sb, struct statfs *buf)
64 retval = vfs_statfs(sb, &st);
68 if (sizeof(*buf) == sizeof(st))
69 memcpy(buf, &st, sizeof(st));
71 if (sizeof buf->f_blocks == 4) {
72 if ((st.f_blocks | st.f_bfree | st.f_bavail) &
73 0xffffffff00000000ULL)
76 * f_files and f_ffree may be -1; it's okay to stuff
79 if (st.f_files != -1 &&
80 (st.f_files & 0xffffffff00000000ULL))
82 if (st.f_ffree != -1 &&
83 (st.f_ffree & 0xffffffff00000000ULL))
87 buf->f_type = st.f_type;
88 buf->f_bsize = st.f_bsize;
89 buf->f_blocks = st.f_blocks;
90 buf->f_bfree = st.f_bfree;
91 buf->f_bavail = st.f_bavail;
92 buf->f_files = st.f_files;
93 buf->f_ffree = st.f_ffree;
94 buf->f_fsid = st.f_fsid;
95 buf->f_namelen = st.f_namelen;
96 buf->f_frsize = st.f_frsize;
97 memset(buf->f_spare, 0, sizeof(buf->f_spare));
102 static int vfs_statfs64(struct super_block *sb, struct statfs64 *buf)
107 retval = vfs_statfs(sb, &st);
111 if (sizeof(*buf) == sizeof(st))
112 memcpy(buf, &st, sizeof(st));
114 buf->f_type = st.f_type;
115 buf->f_bsize = st.f_bsize;
116 buf->f_blocks = st.f_blocks;
117 buf->f_bfree = st.f_bfree;
118 buf->f_bavail = st.f_bavail;
119 buf->f_files = st.f_files;
120 buf->f_ffree = st.f_ffree;
121 buf->f_fsid = st.f_fsid;
122 buf->f_namelen = st.f_namelen;
123 buf->f_frsize = st.f_frsize;
124 memset(buf->f_spare, 0, sizeof(buf->f_spare));
129 asmlinkage long sys_statfs(const char __user * path, struct statfs __user * buf)
134 error = user_path_walk(path, &nd);
137 error = vfs_statfs_native(nd.dentry->d_inode->i_sb, &tmp);
138 if (!error && copy_to_user(buf, &tmp, sizeof(tmp)))
146 asmlinkage long sys_statfs64(const char __user *path, size_t sz, struct statfs64 __user *buf)
151 if (sz != sizeof(*buf))
153 error = user_path_walk(path, &nd);
156 error = vfs_statfs64(nd.dentry->d_inode->i_sb, &tmp);
157 if (!error && copy_to_user(buf, &tmp, sizeof(tmp)))
165 asmlinkage long sys_fstatfs(unsigned int fd, struct statfs __user * buf)
175 error = vfs_statfs_native(file->f_dentry->d_inode->i_sb, &tmp);
176 if (!error && copy_to_user(buf, &tmp, sizeof(tmp)))
183 asmlinkage long sys_fstatfs64(unsigned int fd, size_t sz, struct statfs64 __user *buf)
189 if (sz != sizeof(*buf))
196 error = vfs_statfs64(file->f_dentry->d_inode->i_sb, &tmp);
197 if (!error && copy_to_user(buf, &tmp, sizeof(tmp)))
204 int do_truncate(struct dentry *dentry, loff_t length, unsigned int time_attrs,
208 struct iattr newattrs;
210 /* Not pretty: "inode->i_size" shouldn't really be signed. But it is. */
214 newattrs.ia_size = length;
215 newattrs.ia_valid = ATTR_SIZE | time_attrs;
217 newattrs.ia_file = filp;
218 newattrs.ia_valid |= ATTR_FILE;
221 mutex_lock(&dentry->d_inode->i_mutex);
222 err = notify_change(dentry, &newattrs);
223 mutex_unlock(&dentry->d_inode->i_mutex);
227 static long do_sys_truncate(const char __user * path, loff_t length)
230 struct inode * inode;
234 if (length < 0) /* sorry, but loff_t says... */
237 error = user_path_walk(path, &nd);
240 inode = nd.dentry->d_inode;
242 /* For directories it's -EISDIR, for other non-regulars - -EINVAL */
244 if (S_ISDIR(inode->i_mode))
248 if (!S_ISREG(inode->i_mode))
251 error = vfs_permission(&nd, MAY_WRITE);
256 if (IS_RDONLY(inode) || MNT_IS_RDONLY(nd.mnt))
260 if (IS_IMMUTABLE(inode) || IS_APPEND(inode))
264 * Make sure that there are no leases.
266 error = break_lease(inode, FMODE_WRITE);
270 error = get_write_access(inode);
274 error = locks_verify_truncate(inode, NULL, length);
277 error = do_truncate(nd.dentry, length, 0, NULL);
279 put_write_access(inode);
287 asmlinkage long sys_truncate(const char __user * path, unsigned long length)
289 /* on 32-bit boxen it will cut the range 2^31--2^32-1 off */
290 return do_sys_truncate(path, (long)length);
293 static long do_sys_ftruncate(unsigned int fd, loff_t length, int small)
295 struct inode * inode;
296 struct dentry *dentry;
308 /* explicitly opened as large or we are on 64-bit box */
309 if (file->f_flags & O_LARGEFILE)
312 dentry = file->f_dentry;
313 inode = dentry->d_inode;
315 if (!S_ISREG(inode->i_mode) || !(file->f_mode & FMODE_WRITE))
319 /* Cannot ftruncate over 2^31 bytes without large file support */
320 if (small && length > MAX_NON_LFS)
324 if (IS_APPEND(inode))
327 error = locks_verify_truncate(inode, file, length);
329 error = do_truncate(dentry, length, 0, file);
336 asmlinkage long sys_ftruncate(unsigned int fd, unsigned long length)
338 long ret = do_sys_ftruncate(fd, length, 1);
339 /* avoid REGPARM breakage on x86: */
340 prevent_tail_call(ret);
344 /* LFS versions of truncate are only needed on 32 bit machines */
345 #if BITS_PER_LONG == 32
346 asmlinkage long sys_truncate64(const char __user * path, loff_t length)
348 return do_sys_truncate(path, length);
351 asmlinkage long sys_ftruncate64(unsigned int fd, loff_t length)
353 long ret = do_sys_ftruncate(fd, length, 0);
354 /* avoid REGPARM breakage on x86: */
355 prevent_tail_call(ret);
360 #ifdef __ARCH_WANT_SYS_UTIME
363 * sys_utime() can be implemented in user-level using sys_utimes().
364 * Is this for backwards compatibility? If so, why not move it
365 * into the appropriate arch directory (for those architectures that
369 /* If times==NULL, set access and modification to current time,
370 * must be owner or have write permission.
371 * Else, update from *times, must be owner or super user.
373 asmlinkage long sys_utime(char __user * filename, struct utimbuf __user * times)
377 struct inode * inode;
378 struct iattr newattrs;
380 error = user_path_walk(filename, &nd);
383 inode = nd.dentry->d_inode;
386 if (IS_RDONLY(inode) || MNT_IS_RDONLY(nd.mnt))
389 /* Don't worry, the checks are done in inode_change_ok() */
390 newattrs.ia_valid = ATTR_CTIME | ATTR_MTIME | ATTR_ATIME;
393 if (IS_APPEND(inode) || IS_IMMUTABLE(inode))
396 error = get_user(newattrs.ia_atime.tv_sec, ×->actime);
397 newattrs.ia_atime.tv_nsec = 0;
399 error = get_user(newattrs.ia_mtime.tv_sec, ×->modtime);
400 newattrs.ia_mtime.tv_nsec = 0;
404 newattrs.ia_valid |= ATTR_ATIME_SET | ATTR_MTIME_SET;
407 if (IS_IMMUTABLE(inode))
410 if (current->fsuid != inode->i_uid &&
411 (error = vfs_permission(&nd, MAY_WRITE)) != 0)
414 mutex_lock(&inode->i_mutex);
415 error = notify_change(nd.dentry, &newattrs);
416 mutex_unlock(&inode->i_mutex);
425 /* If times==NULL, set access and modification to current time,
426 * must be owner or have write permission.
427 * Else, update from *times, must be owner or super user.
429 long do_utimes(int dfd, char __user *filename, struct timeval *times)
433 struct inode * inode;
434 struct iattr newattrs;
436 error = __user_walk_fd(dfd, filename, LOOKUP_FOLLOW, &nd);
440 inode = nd.dentry->d_inode;
443 if (IS_RDONLY(inode) || MNT_IS_RDONLY(nd.mnt))
446 /* Don't worry, the checks are done in inode_change_ok() */
447 newattrs.ia_valid = ATTR_CTIME | ATTR_MTIME | ATTR_ATIME;
450 if (IS_APPEND(inode) || IS_IMMUTABLE(inode))
453 newattrs.ia_atime.tv_sec = times[0].tv_sec;
454 newattrs.ia_atime.tv_nsec = times[0].tv_usec * 1000;
455 newattrs.ia_mtime.tv_sec = times[1].tv_sec;
456 newattrs.ia_mtime.tv_nsec = times[1].tv_usec * 1000;
457 newattrs.ia_valid |= ATTR_ATIME_SET | ATTR_MTIME_SET;
460 if (IS_IMMUTABLE(inode))
463 if (current->fsuid != inode->i_uid &&
464 (error = vfs_permission(&nd, MAY_WRITE)) != 0)
467 mutex_lock(&inode->i_mutex);
468 error = notify_change(nd.dentry, &newattrs);
469 mutex_unlock(&inode->i_mutex);
476 asmlinkage long sys_futimesat(int dfd, char __user *filename, struct timeval __user *utimes)
478 struct timeval times[2];
480 if (utimes && copy_from_user(×, utimes, sizeof(times)))
482 return do_utimes(dfd, filename, utimes ? times : NULL);
485 asmlinkage long sys_utimes(char __user *filename, struct timeval __user *utimes)
487 return sys_futimesat(AT_FDCWD, filename, utimes);
492 * access() needs to use the real uid/gid, not the effective uid/gid.
493 * We do this by temporarily clearing all FS-related capabilities and
494 * switching the fsuid/fsgid around to the real ones.
496 asmlinkage long sys_faccessat(int dfd, const char __user *filename, int mode)
499 int old_fsuid, old_fsgid;
500 kernel_cap_t old_cap;
503 if (mode & ~S_IRWXO) /* where's F_OK, X_OK, W_OK, R_OK? */
506 old_fsuid = current->fsuid;
507 old_fsgid = current->fsgid;
508 old_cap = current->cap_effective;
510 current->fsuid = current->uid;
511 current->fsgid = current->gid;
514 * Clear the capabilities if we switch to a non-root user
516 * FIXME: There is a race here against sys_capset. The
517 * capabilities can change yet we will restore the old
518 * value below. We should hold task_capabilities_lock,
519 * but we cannot because user_path_walk can sleep.
522 cap_clear(current->cap_effective);
524 current->cap_effective = current->cap_permitted;
526 res = __user_walk_fd(dfd, filename, LOOKUP_FOLLOW|LOOKUP_ACCESS, &nd);
528 res = vfs_permission(&nd, mode);
529 /* SuS v2 requires we report a read only fs too */
530 if(!res && (mode & S_IWOTH)
531 && (IS_RDONLY(nd.dentry->d_inode) || MNT_IS_RDONLY(nd.mnt))
532 && !special_file(nd.dentry->d_inode->i_mode))
537 current->fsuid = old_fsuid;
538 current->fsgid = old_fsgid;
539 current->cap_effective = old_cap;
544 asmlinkage long sys_access(const char __user *filename, int mode)
546 return sys_faccessat(AT_FDCWD, filename, mode);
549 asmlinkage long sys_chdir(const char __user * filename)
554 error = __user_walk(filename, LOOKUP_FOLLOW|LOOKUP_DIRECTORY, &nd);
558 error = vfs_permission(&nd, MAY_EXEC);
562 set_fs_pwd(current->fs, nd.mnt, nd.dentry);
570 asmlinkage long sys_fchdir(unsigned int fd)
573 struct dentry *dentry;
575 struct vfsmount *mnt;
583 dentry = file->f_dentry;
584 mnt = file->f_vfsmnt;
585 inode = dentry->d_inode;
588 if (!S_ISDIR(inode->i_mode))
591 error = file_permission(file, MAY_EXEC);
593 set_fs_pwd(current->fs, mnt, dentry);
600 asmlinkage long sys_chroot(const char __user * filename)
605 error = __user_walk(filename, LOOKUP_FOLLOW | LOOKUP_DIRECTORY | LOOKUP_NOALT, &nd);
609 error = vfs_permission(&nd, MAY_EXEC);
614 if (!capable(CAP_SYS_CHROOT))
617 set_fs_root(current->fs, nd.mnt, nd.dentry);
626 asmlinkage long sys_fchmod(unsigned int fd, mode_t mode)
628 struct inode * inode;
629 struct dentry * dentry;
632 struct iattr newattrs;
638 dentry = file->f_dentry;
639 inode = dentry->d_inode;
642 if (IS_RDONLY(inode) || MNT_IS_RDONLY(file->f_vfsmnt))
645 if (IS_IMMUTABLE(inode) || IS_APPEND(inode))
647 mutex_lock(&inode->i_mutex);
648 if (mode == (mode_t) -1)
649 mode = inode->i_mode;
650 newattrs.ia_mode = (mode & S_IALLUGO) | (inode->i_mode & ~S_IALLUGO);
651 newattrs.ia_valid = ATTR_MODE | ATTR_CTIME;
652 err = notify_change(dentry, &newattrs);
653 mutex_unlock(&inode->i_mutex);
661 asmlinkage long sys_fchmodat(int dfd, const char __user *filename,
665 struct inode * inode;
667 struct iattr newattrs;
669 error = __user_walk_fd(dfd, filename, LOOKUP_FOLLOW, &nd);
672 inode = nd.dentry->d_inode;
675 if (IS_RDONLY(inode) || MNT_IS_RDONLY(nd.mnt))
679 if (IS_IMMUTABLE(inode) || IS_APPEND(inode))
682 mutex_lock(&inode->i_mutex);
683 if (mode == (mode_t) -1)
684 mode = inode->i_mode;
685 newattrs.ia_mode = (mode & S_IALLUGO) | (inode->i_mode & ~S_IALLUGO);
686 newattrs.ia_valid = ATTR_MODE | ATTR_CTIME;
687 error = notify_change(nd.dentry, &newattrs);
688 mutex_unlock(&inode->i_mutex);
696 asmlinkage long sys_chmod(const char __user *filename, mode_t mode)
698 return sys_fchmodat(AT_FDCWD, filename, mode);
701 static int chown_common(struct dentry *dentry, struct vfsmount *mnt,
702 uid_t user, gid_t group)
704 struct inode * inode;
706 struct iattr newattrs;
709 if (!(inode = dentry->d_inode)) {
710 printk(KERN_ERR "chown_common: NULL inode\n");
714 if (IS_RDONLY(inode) || MNT_IS_RDONLY(mnt))
717 if (IS_IMMUTABLE(inode) || IS_APPEND(inode))
719 newattrs.ia_valid = ATTR_CTIME;
720 if (user != (uid_t) -1) {
721 newattrs.ia_valid |= ATTR_UID;
722 newattrs.ia_uid = vx_map_uid(user);
724 if (group != (gid_t) -1) {
725 newattrs.ia_valid |= ATTR_GID;
726 newattrs.ia_gid = vx_map_gid(group);
728 if (!S_ISDIR(inode->i_mode))
729 newattrs.ia_valid |= ATTR_KILL_SUID|ATTR_KILL_SGID;
730 mutex_lock(&inode->i_mutex);
731 error = notify_change(dentry, &newattrs);
732 mutex_unlock(&inode->i_mutex);
737 asmlinkage long sys_chown(const char __user * filename, uid_t user, gid_t group)
742 error = user_path_walk(filename, &nd);
744 error = chown_common(nd.dentry, nd.mnt, user, group);
750 asmlinkage long sys_fchownat(int dfd, const char __user *filename, uid_t user,
751 gid_t group, int flag)
757 if ((flag & ~AT_SYMLINK_NOFOLLOW) != 0)
760 follow = (flag & AT_SYMLINK_NOFOLLOW) ? 0 : LOOKUP_FOLLOW;
761 error = __user_walk_fd(dfd, filename, follow, &nd);
763 error = chown_common(nd.dentry, nd.mnt, user, group);
770 asmlinkage long sys_lchown(const char __user * filename, uid_t user, gid_t group)
775 error = user_path_walk_link(filename, &nd);
777 error = chown_common(nd.dentry, nd.mnt, user, group);
784 asmlinkage long sys_fchown(unsigned int fd, uid_t user, gid_t group)
791 error = chown_common(file->f_dentry, file->f_vfsmnt, user, group);
797 static struct file *__dentry_open(struct dentry *dentry, struct vfsmount *mnt,
798 int flags, struct file *f,
799 int (*open)(struct inode *, struct file *))
805 f->f_mode = ((flags+1) & O_ACCMODE) | FMODE_LSEEK |
806 FMODE_PREAD | FMODE_PWRITE;
807 inode = dentry->d_inode;
808 if (f->f_mode & FMODE_WRITE) {
809 error = get_write_access(inode);
814 f->f_mapping = inode->i_mapping;
815 f->f_dentry = dentry;
818 f->f_op = fops_get(inode->i_fop);
819 file_move(f, &inode->i_sb->s_files);
821 if (!open && f->f_op)
822 open = f->f_op->open;
824 error = open(inode, f);
829 f->f_flags &= ~(O_CREAT | O_EXCL | O_NOCTTY | O_TRUNC);
831 file_ra_state_init(&f->f_ra, f->f_mapping->host->i_mapping);
833 /* NB: we're sure to have correct a_ops only after f_op->open */
834 if (f->f_flags & O_DIRECT) {
835 if (!f->f_mapping->a_ops ||
836 ((!f->f_mapping->a_ops->direct_IO) &&
837 (!f->f_mapping->a_ops->get_xip_page))) {
839 f = ERR_PTR(-EINVAL);
847 if (f->f_mode & FMODE_WRITE)
848 put_write_access(inode);
856 return ERR_PTR(error);
860 * Note that while the flag value (low two bits) for sys_open means:
866 * 00 - no permissions needed
867 * 01 - read-permission
868 * 10 - write-permission
870 * for the internal routines (ie open_namei()/follow_link() etc). 00 is
873 static struct file *do_filp_open(int dfd, const char *filename, int flags,
876 int namei_flags, error;
880 if ((namei_flags+1) & O_ACCMODE)
883 error = open_namei(dfd, filename, namei_flags, mode, &nd);
885 return nameidata_to_filp(&nd, flags);
887 return ERR_PTR(error);
890 struct file *filp_open(const char *filename, int flags, int mode)
892 return do_filp_open(AT_FDCWD, filename, flags, mode);
894 EXPORT_SYMBOL(filp_open);
897 * lookup_instantiate_filp - instantiates the open intent filp
898 * @nd: pointer to nameidata
899 * @dentry: pointer to dentry
900 * @open: open callback
902 * Helper for filesystems that want to use lookup open intents and pass back
903 * a fully instantiated struct file to the caller.
904 * This function is meant to be called from within a filesystem's
906 * Note that in case of error, nd->intent.open.file is destroyed, but the
907 * path information remains valid.
908 * If the open callback is set to NULL, then the standard f_op->open()
909 * filesystem callback is substituted.
911 struct file *lookup_instantiate_filp(struct nameidata *nd, struct dentry *dentry,
912 int (*open)(struct inode *, struct file *))
914 if (IS_ERR(nd->intent.open.file))
918 nd->intent.open.file = __dentry_open(dget(dentry), mntget(nd->mnt),
919 nd->intent.open.flags - 1,
920 nd->intent.open.file,
923 return nd->intent.open.file;
925 release_open_intent(nd);
926 nd->intent.open.file = (struct file *)dentry;
929 EXPORT_SYMBOL_GPL(lookup_instantiate_filp);
932 * nameidata_to_filp - convert a nameidata to an open filp.
933 * @nd: pointer to nameidata
936 * Note that this function destroys the original nameidata
938 struct file *nameidata_to_filp(struct nameidata *nd, int flags)
942 /* Pick up the filp from the open intent */
943 filp = nd->intent.open.file;
944 /* Has the filesystem initialised the file for us? */
945 if (filp->f_dentry == NULL)
946 filp = __dentry_open(nd->dentry, nd->mnt, flags, filp, NULL);
953 * dentry_open() will have done dput(dentry) and mntput(mnt) if it returns an
956 struct file *dentry_open(struct dentry *dentry, struct vfsmount *mnt, int flags)
962 f = get_empty_filp();
966 return ERR_PTR(error);
969 return __dentry_open(dentry, mnt, flags, f, NULL);
971 EXPORT_SYMBOL(dentry_open);
974 * Find an empty file descriptor entry, and mark it busy.
976 int get_unused_fd(void)
978 struct files_struct * files = current->files;
983 spin_lock(&files->file_lock);
986 fdt = files_fdtable(files);
987 fd = find_next_zero_bit(fdt->open_fds->fds_bits,
992 * N.B. For clone tasks sharing a files structure, this test
993 * will limit the total number of files that can be opened.
995 if (fd >= current->signal->rlim[RLIMIT_NOFILE].rlim_cur)
998 /* Do we need to expand the fd array or fd set? */
999 error = expand_files(files, fd);
1005 * If we needed to expand the fs array we
1006 * might have blocked - try again.
1012 FD_SET(fd, fdt->open_fds);
1013 FD_CLR(fd, fdt->close_on_exec);
1014 fdt->next_fd = fd + 1;
1018 if (fdt->fd[fd] != NULL) {
1019 printk(KERN_WARNING "get_unused_fd: slot %d not NULL!\n", fd);
1026 spin_unlock(&files->file_lock);
1030 EXPORT_SYMBOL(get_unused_fd);
1032 static void __put_unused_fd(struct files_struct *files, unsigned int fd)
1034 struct fdtable *fdt = files_fdtable(files);
1035 __FD_CLR(fd, fdt->open_fds);
1036 if (fd < fdt->next_fd)
1041 void fastcall put_unused_fd(unsigned int fd)
1043 struct files_struct *files = current->files;
1044 spin_lock(&files->file_lock);
1045 __put_unused_fd(files, fd);
1046 spin_unlock(&files->file_lock);
1049 EXPORT_SYMBOL(put_unused_fd);
1052 * Install a file pointer in the fd array.
1054 * The VFS is full of places where we drop the files lock between
1055 * setting the open_fds bitmap and installing the file in the file
1056 * array. At any such point, we are vulnerable to a dup2() race
1057 * installing a file in the array before us. We need to detect this and
1058 * fput() the struct file we are about to overwrite in this case.
1060 * It should never happen - if we allow dup2() do it, _really_ bad things
1064 void fastcall fd_install(unsigned int fd, struct file * file)
1066 struct files_struct *files = current->files;
1067 struct fdtable *fdt;
1068 spin_lock(&files->file_lock);
1069 fdt = files_fdtable(files);
1070 BUG_ON(fdt->fd[fd] != NULL);
1071 rcu_assign_pointer(fdt->fd[fd], file);
1072 spin_unlock(&files->file_lock);
1075 EXPORT_SYMBOL(fd_install);
1077 long do_sys_open(int dfd, const char __user *filename, int flags, int mode)
1079 char *tmp = getname(filename);
1080 int fd = PTR_ERR(tmp);
1083 fd = get_unused_fd();
1085 struct file *f = do_filp_open(dfd, tmp, flags, mode);
1090 fsnotify_open(f->f_dentry);
1099 asmlinkage long sys_open(const char __user *filename, int flags, int mode)
1103 if (force_o_largefile())
1104 flags |= O_LARGEFILE;
1106 ret = do_sys_open(AT_FDCWD, filename, flags, mode);
1107 /* avoid REGPARM breakage on x86: */
1108 prevent_tail_call(ret);
1111 EXPORT_SYMBOL_GPL(sys_open);
1113 asmlinkage long sys_openat(int dfd, const char __user *filename, int flags,
1118 if (force_o_largefile())
1119 flags |= O_LARGEFILE;
1121 ret = do_sys_open(dfd, filename, flags, mode);
1122 /* avoid REGPARM breakage on x86: */
1123 prevent_tail_call(ret);
1126 EXPORT_SYMBOL_GPL(sys_openat);
1131 * For backward compatibility? Maybe this should be moved
1132 * into arch/i386 instead?
1134 asmlinkage long sys_creat(const char __user * pathname, int mode)
1136 return sys_open(pathname, O_CREAT | O_WRONLY | O_TRUNC, mode);
1142 * "id" is the POSIX thread ID. We use the
1143 * files pointer for this..
1145 int filp_close(struct file *filp, fl_owner_t id)
1149 if (!file_count(filp)) {
1150 printk(KERN_ERR "VFS: Close: file count is 0\n");
1154 if (filp->f_op && filp->f_op->flush)
1155 retval = filp->f_op->flush(filp);
1157 dnotify_flush(filp, id);
1158 locks_remove_posix(filp, id);
1163 EXPORT_SYMBOL(filp_close);
1166 * Careful here! We test whether the file pointer is NULL before
1167 * releasing the fd. This ensures that one clone task can't release
1168 * an fd while another clone is opening it.
1170 asmlinkage long sys_close(unsigned int fd)
1173 struct files_struct *files = current->files;
1174 struct fdtable *fdt;
1176 spin_lock(&files->file_lock);
1177 fdt = files_fdtable(files);
1178 if (fd >= fdt->max_fds)
1183 rcu_assign_pointer(fdt->fd[fd], NULL);
1184 FD_CLR(fd, fdt->close_on_exec);
1185 __put_unused_fd(files, fd);
1186 spin_unlock(&files->file_lock);
1187 return filp_close(filp, files);
1190 spin_unlock(&files->file_lock);
1194 EXPORT_SYMBOL(sys_close);
1197 * This routine simulates a hangup on the tty, to arrange that users
1198 * are given clean terminals at login time.
1200 asmlinkage long sys_vhangup(void)
1202 if (capable(CAP_SYS_TTY_CONFIG)) {
1203 tty_vhangup(current->signal->tty);
1210 * Called when an inode is about to be open.
1211 * We use this to disallow opening large files on 32bit systems if
1212 * the caller didn't specify O_LARGEFILE. On 64bit systems we force
1213 * on this flag in sys_open.
1215 int generic_file_open(struct inode * inode, struct file * filp)
1217 if (!(filp->f_flags & O_LARGEFILE) && i_size_read(inode) > MAX_NON_LFS)
1222 EXPORT_SYMBOL(generic_file_open);
1225 * This is used by subsystems that don't want seekable
1228 int nonseekable_open(struct inode *inode, struct file *filp)
1230 filp->f_mode &= ~(FMODE_LSEEK | FMODE_PREAD | FMODE_PWRITE);
1234 EXPORT_SYMBOL(nonseekable_open);