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/audit.h>
31 #include <linux/vs_base.h>
32 #include <linux/vs_limit.h>
33 #include <linux/vs_dlimit.h>
34 #include <linux/vserver/xid.h>
36 #include <asm/unistd.h>
38 int vfs_statfs(struct dentry *dentry, struct kstatfs *buf)
43 struct super_block *sb = dentry->d_sb;
46 if (sb->s_op->statfs) {
47 memset(buf, 0, sizeof(*buf));
48 retval = security_sb_statfs(dentry);
51 retval = sb->s_op->statfs(dentry, buf);
52 if (retval == 0 && buf->f_frsize == 0)
53 buf->f_frsize = buf->f_bsize;
55 if (!vx_check(0, VX_ADMIN|VX_WATCH))
56 vx_vsi_statfs(sb, buf);
61 EXPORT_SYMBOL(vfs_statfs);
63 static int vfs_statfs_native(struct dentry *dentry, struct statfs *buf)
68 retval = vfs_statfs(dentry, &st);
72 if (sizeof(*buf) == sizeof(st))
73 memcpy(buf, &st, sizeof(st));
75 if (sizeof buf->f_blocks == 4) {
76 if ((st.f_blocks | st.f_bfree | st.f_bavail) &
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 asmlinkage long sys_statfs(const char __user * path, struct statfs __user * buf)
138 error = user_path_walk(path, &nd);
141 error = vfs_statfs_native(nd.dentry, &tmp);
142 if (!error && copy_to_user(buf, &tmp, sizeof(tmp)))
150 asmlinkage long sys_statfs64(const char __user *path, size_t sz, struct statfs64 __user *buf)
155 if (sz != sizeof(*buf))
157 error = user_path_walk(path, &nd);
160 error = vfs_statfs64(nd.dentry, &tmp);
161 if (!error && copy_to_user(buf, &tmp, sizeof(tmp)))
169 asmlinkage long sys_fstatfs(unsigned int fd, struct statfs __user * buf)
179 error = vfs_statfs_native(file->f_dentry, &tmp);
180 if (!error && copy_to_user(buf, &tmp, sizeof(tmp)))
187 asmlinkage long sys_fstatfs64(unsigned int fd, size_t sz, struct statfs64 __user *buf)
193 if (sz != sizeof(*buf))
200 error = vfs_statfs64(file->f_dentry, &tmp);
201 if (!error && copy_to_user(buf, &tmp, sizeof(tmp)))
208 int do_truncate(struct dentry *dentry, loff_t length, unsigned int time_attrs,
212 struct iattr newattrs;
214 /* Not pretty: "inode->i_size" shouldn't really be signed. But it is. */
218 newattrs.ia_size = length;
219 newattrs.ia_valid = ATTR_SIZE | time_attrs;
221 newattrs.ia_file = filp;
222 newattrs.ia_valid |= ATTR_FILE;
225 mutex_lock(&dentry->d_inode->i_mutex);
226 err = notify_change(dentry, &newattrs);
227 mutex_unlock(&dentry->d_inode->i_mutex);
231 static long do_sys_truncate(const char __user * path, loff_t length)
234 struct inode * inode;
238 if (length < 0) /* sorry, but loff_t says... */
241 error = user_path_walk(path, &nd);
244 inode = nd.dentry->d_inode;
246 /* For directories it's -EISDIR, for other non-regulars - -EINVAL */
248 if (S_ISDIR(inode->i_mode))
252 if (!S_ISREG(inode->i_mode))
255 error = vfs_permission(&nd, MAY_WRITE);
260 if (IS_RDONLY(inode) || MNT_IS_RDONLY(nd.mnt))
264 if (IS_IMMUTABLE(inode) || IS_APPEND(inode))
268 * Make sure that there are no leases.
270 error = break_lease(inode, FMODE_WRITE);
274 error = get_write_access(inode);
278 error = locks_verify_truncate(inode, NULL, length);
281 error = do_truncate(nd.dentry, length, 0, NULL);
283 put_write_access(inode);
291 asmlinkage long sys_truncate(const char __user * path, unsigned long length)
293 /* on 32-bit boxen it will cut the range 2^31--2^32-1 off */
294 return do_sys_truncate(path, (long)length);
297 static long do_sys_ftruncate(unsigned int fd, loff_t length, int small)
299 struct inode * inode;
300 struct dentry *dentry;
312 /* explicitly opened as large or we are on 64-bit box */
313 if (file->f_flags & O_LARGEFILE)
316 dentry = file->f_dentry;
317 inode = dentry->d_inode;
319 if (!S_ISREG(inode->i_mode) || !(file->f_mode & FMODE_WRITE))
323 /* Cannot ftruncate over 2^31 bytes without large file support */
324 if (small && length > MAX_NON_LFS)
328 if (IS_APPEND(inode))
331 error = locks_verify_truncate(inode, file, length);
333 error = do_truncate(dentry, length, ATTR_MTIME|ATTR_CTIME, file);
340 asmlinkage long sys_ftruncate(unsigned int fd, unsigned long length)
342 long ret = do_sys_ftruncate(fd, length, 1);
343 /* avoid REGPARM breakage on x86: */
344 prevent_tail_call(ret);
348 /* LFS versions of truncate are only needed on 32 bit machines */
349 #if BITS_PER_LONG == 32
350 asmlinkage long sys_truncate64(const char __user * path, loff_t length)
352 return do_sys_truncate(path, length);
355 asmlinkage long sys_ftruncate64(unsigned int fd, loff_t length)
357 long ret = do_sys_ftruncate(fd, length, 0);
358 /* avoid REGPARM breakage on x86: */
359 prevent_tail_call(ret);
364 #ifdef __ARCH_WANT_SYS_UTIME
367 * sys_utime() can be implemented in user-level using sys_utimes().
368 * Is this for backwards compatibility? If so, why not move it
369 * into the appropriate arch directory (for those architectures that
373 /* If times==NULL, set access and modification to current time,
374 * must be owner or have write permission.
375 * Else, update from *times, must be owner or super user.
377 asmlinkage long sys_utime(char __user * filename, struct utimbuf __user * times)
381 struct inode * inode;
382 struct iattr newattrs;
384 error = user_path_walk(filename, &nd);
387 inode = nd.dentry->d_inode;
390 if (IS_RDONLY(inode) || MNT_IS_RDONLY(nd.mnt))
393 /* Don't worry, the checks are done in inode_change_ok() */
394 newattrs.ia_valid = ATTR_CTIME | ATTR_MTIME | ATTR_ATIME;
397 if (IS_APPEND(inode) || IS_IMMUTABLE(inode))
400 error = get_user(newattrs.ia_atime.tv_sec, ×->actime);
401 newattrs.ia_atime.tv_nsec = 0;
403 error = get_user(newattrs.ia_mtime.tv_sec, ×->modtime);
404 newattrs.ia_mtime.tv_nsec = 0;
408 newattrs.ia_valid |= ATTR_ATIME_SET | ATTR_MTIME_SET;
411 if (IS_IMMUTABLE(inode))
414 if (current->fsuid != inode->i_uid &&
415 (error = vfs_permission(&nd, MAY_WRITE)) != 0)
418 mutex_lock(&inode->i_mutex);
419 error = notify_change(nd.dentry, &newattrs);
420 mutex_unlock(&inode->i_mutex);
429 /* If times==NULL, set access and modification to current time,
430 * must be owner or have write permission.
431 * Else, update from *times, must be owner or super user.
433 long do_utimes(int dfd, char __user *filename, struct timeval *times)
437 struct inode * inode;
438 struct iattr newattrs;
440 error = __user_walk_fd(dfd, filename, LOOKUP_FOLLOW, &nd);
444 inode = nd.dentry->d_inode;
447 if (IS_RDONLY(inode) || MNT_IS_RDONLY(nd.mnt))
450 /* Don't worry, the checks are done in inode_change_ok() */
451 newattrs.ia_valid = ATTR_CTIME | ATTR_MTIME | ATTR_ATIME;
454 if (IS_APPEND(inode) || IS_IMMUTABLE(inode))
457 newattrs.ia_atime.tv_sec = times[0].tv_sec;
458 newattrs.ia_atime.tv_nsec = times[0].tv_usec * 1000;
459 newattrs.ia_mtime.tv_sec = times[1].tv_sec;
460 newattrs.ia_mtime.tv_nsec = times[1].tv_usec * 1000;
461 newattrs.ia_valid |= ATTR_ATIME_SET | ATTR_MTIME_SET;
464 if (IS_IMMUTABLE(inode))
467 if (current->fsuid != inode->i_uid &&
468 (error = vfs_permission(&nd, MAY_WRITE)) != 0)
471 mutex_lock(&inode->i_mutex);
472 error = notify_change(nd.dentry, &newattrs);
473 mutex_unlock(&inode->i_mutex);
480 asmlinkage long sys_futimesat(int dfd, char __user *filename, struct timeval __user *utimes)
482 struct timeval times[2];
484 if (utimes && copy_from_user(×, utimes, sizeof(times)))
486 return do_utimes(dfd, filename, utimes ? times : NULL);
489 asmlinkage long sys_utimes(char __user *filename, struct timeval __user *utimes)
491 return sys_futimesat(AT_FDCWD, filename, utimes);
496 * access() needs to use the real uid/gid, not the effective uid/gid.
497 * We do this by temporarily clearing all FS-related capabilities and
498 * switching the fsuid/fsgid around to the real ones.
500 asmlinkage long sys_faccessat(int dfd, const char __user *filename, int mode)
503 int old_fsuid, old_fsgid;
504 kernel_cap_t old_cap;
507 if (mode & ~S_IRWXO) /* where's F_OK, X_OK, W_OK, R_OK? */
510 old_fsuid = current->fsuid;
511 old_fsgid = current->fsgid;
512 old_cap = current->cap_effective;
514 current->fsuid = current->uid;
515 current->fsgid = current->gid;
518 * Clear the capabilities if we switch to a non-root user
520 * FIXME: There is a race here against sys_capset. The
521 * capabilities can change yet we will restore the old
522 * value below. We should hold task_capabilities_lock,
523 * but we cannot because user_path_walk can sleep.
526 cap_clear(current->cap_effective);
528 current->cap_effective = current->cap_permitted;
530 res = __user_walk_fd(dfd, filename, LOOKUP_FOLLOW|LOOKUP_ACCESS, &nd);
532 res = vfs_permission(&nd, mode);
533 /* SuS v2 requires we report a read only fs too */
534 if(!res && (mode & S_IWOTH)
535 && (IS_RDONLY(nd.dentry->d_inode) || MNT_IS_RDONLY(nd.mnt))
536 && !special_file(nd.dentry->d_inode->i_mode))
541 current->fsuid = old_fsuid;
542 current->fsgid = old_fsgid;
543 current->cap_effective = old_cap;
548 asmlinkage long sys_access(const char __user *filename, int mode)
550 return sys_faccessat(AT_FDCWD, filename, mode);
553 asmlinkage long sys_chdir(const char __user * filename)
558 error = __user_walk(filename, LOOKUP_FOLLOW|LOOKUP_DIRECTORY, &nd);
562 error = vfs_permission(&nd, MAY_EXEC);
566 set_fs_pwd(current->fs, nd.mnt, nd.dentry);
574 EXPORT_SYMBOL_GPL(sys_chdir);
576 asmlinkage long sys_fchdir(unsigned int fd)
579 struct dentry *dentry;
581 struct vfsmount *mnt;
589 dentry = file->f_dentry;
590 mnt = file->f_vfsmnt;
591 inode = dentry->d_inode;
594 if (!S_ISDIR(inode->i_mode))
597 error = file_permission(file, MAY_EXEC);
599 set_fs_pwd(current->fs, mnt, dentry);
606 asmlinkage long sys_chroot(const char __user * filename)
611 error = __user_walk(filename, LOOKUP_FOLLOW | LOOKUP_DIRECTORY | LOOKUP_NOALT, &nd);
615 error = vfs_permission(&nd, MAY_EXEC);
620 if (!capable(CAP_SYS_CHROOT))
623 set_fs_root(current->fs, nd.mnt, nd.dentry);
632 EXPORT_SYMBOL_GPL(sys_chroot);
634 asmlinkage long sys_fchmod(unsigned int fd, mode_t mode)
636 struct inode * inode;
637 struct dentry * dentry;
640 struct iattr newattrs;
646 dentry = file->f_dentry;
647 inode = dentry->d_inode;
649 audit_inode(NULL, inode);
652 if (IS_RDONLY(inode) || MNT_IS_RDONLY(file->f_vfsmnt))
655 if (IS_IMMUTABLE(inode) || IS_APPEND(inode))
657 mutex_lock(&inode->i_mutex);
658 if (mode == (mode_t) -1)
659 mode = inode->i_mode;
660 newattrs.ia_mode = (mode & S_IALLUGO) | (inode->i_mode & ~S_IALLUGO);
661 newattrs.ia_valid = ATTR_MODE | ATTR_CTIME;
662 err = notify_change(dentry, &newattrs);
663 mutex_unlock(&inode->i_mutex);
671 asmlinkage long sys_fchmodat(int dfd, const char __user *filename,
675 struct inode * inode;
677 struct iattr newattrs;
679 error = __user_walk_fd(dfd, filename, LOOKUP_FOLLOW, &nd);
682 inode = nd.dentry->d_inode;
685 if (IS_RDONLY(inode) || MNT_IS_RDONLY(nd.mnt))
689 if (IS_IMMUTABLE(inode) || IS_APPEND(inode))
692 mutex_lock(&inode->i_mutex);
693 if (mode == (mode_t) -1)
694 mode = inode->i_mode;
695 newattrs.ia_mode = (mode & S_IALLUGO) | (inode->i_mode & ~S_IALLUGO);
696 newattrs.ia_valid = ATTR_MODE | ATTR_CTIME;
697 error = notify_change(nd.dentry, &newattrs);
698 mutex_unlock(&inode->i_mutex);
706 asmlinkage long sys_chmod(const char __user *filename, mode_t mode)
708 return sys_fchmodat(AT_FDCWD, filename, mode);
711 static int chown_common(struct dentry *dentry, struct vfsmount *mnt,
712 uid_t user, gid_t group)
714 struct inode * inode;
716 struct iattr newattrs;
719 if (!(inode = dentry->d_inode)) {
720 printk(KERN_ERR "chown_common: NULL inode\n");
724 if (IS_RDONLY(inode) || MNT_IS_RDONLY(mnt))
727 if (IS_IMMUTABLE(inode) || IS_APPEND(inode))
729 newattrs.ia_valid = ATTR_CTIME;
730 if (user != (uid_t) -1) {
731 newattrs.ia_valid |= ATTR_UID;
732 newattrs.ia_uid = vx_map_uid(user);
734 if (group != (gid_t) -1) {
735 newattrs.ia_valid |= ATTR_GID;
736 newattrs.ia_gid = vx_map_gid(group);
738 if (!S_ISDIR(inode->i_mode))
739 newattrs.ia_valid |= ATTR_KILL_SUID|ATTR_KILL_SGID;
740 mutex_lock(&inode->i_mutex);
741 error = notify_change(dentry, &newattrs);
742 mutex_unlock(&inode->i_mutex);
747 asmlinkage long sys_chown(const char __user * filename, uid_t user, gid_t group)
752 error = user_path_walk(filename, &nd);
754 error = chown_common(nd.dentry, nd.mnt, user, group);
760 asmlinkage long sys_fchownat(int dfd, const char __user *filename, uid_t user,
761 gid_t group, int flag)
767 if ((flag & ~AT_SYMLINK_NOFOLLOW) != 0)
770 follow = (flag & AT_SYMLINK_NOFOLLOW) ? 0 : LOOKUP_FOLLOW;
771 error = __user_walk_fd(dfd, filename, follow, &nd);
773 error = chown_common(nd.dentry, nd.mnt, user, group);
780 asmlinkage long sys_lchown(const char __user * filename, uid_t user, gid_t group)
785 error = user_path_walk_link(filename, &nd);
787 error = chown_common(nd.dentry, nd.mnt, user, group);
794 asmlinkage long sys_fchown(unsigned int fd, uid_t user, gid_t group)
801 struct dentry * dentry;
802 dentry = file->f_dentry;
803 audit_inode(NULL, dentry->d_inode);
804 error = chown_common(dentry, file->f_vfsmnt, user, group);
810 static struct file *__dentry_open(struct dentry *dentry, struct vfsmount *mnt,
811 int flags, struct file *f,
812 int (*open)(struct inode *, struct file *))
818 f->f_mode = ((flags+1) & O_ACCMODE) | FMODE_LSEEK |
819 FMODE_PREAD | FMODE_PWRITE;
820 inode = dentry->d_inode;
821 if (f->f_mode & FMODE_WRITE) {
822 error = get_write_access(inode);
827 f->f_mapping = inode->i_mapping;
828 f->f_dentry = dentry;
831 f->f_op = fops_get(inode->i_fop);
832 file_move(f, &inode->i_sb->s_files);
834 if (!open && f->f_op)
835 open = f->f_op->open;
837 error = open(inode, f);
842 f->f_flags &= ~(O_CREAT | O_EXCL | O_NOCTTY | O_TRUNC);
844 file_ra_state_init(&f->f_ra, f->f_mapping->host->i_mapping);
846 /* NB: we're sure to have correct a_ops only after f_op->open */
847 if (f->f_flags & O_DIRECT) {
848 if (!f->f_mapping->a_ops ||
849 ((!f->f_mapping->a_ops->direct_IO) &&
850 (!f->f_mapping->a_ops->get_xip_page))) {
852 f = ERR_PTR(-EINVAL);
860 if (f->f_mode & FMODE_WRITE)
861 put_write_access(inode);
869 return ERR_PTR(error);
873 * Note that while the flag value (low two bits) for sys_open means:
879 * 00 - no permissions needed
880 * 01 - read-permission
881 * 10 - write-permission
883 * for the internal routines (ie open_namei()/follow_link() etc). 00 is
886 static struct file *do_filp_open(int dfd, const char *filename, int flags,
889 int namei_flags, error;
893 if ((namei_flags+1) & O_ACCMODE)
896 error = open_namei(dfd, filename, namei_flags, mode, &nd);
898 return nameidata_to_filp(&nd, flags);
900 return ERR_PTR(error);
903 struct file *filp_open(const char *filename, int flags, int mode)
905 return do_filp_open(AT_FDCWD, filename, flags, mode);
907 EXPORT_SYMBOL(filp_open);
910 * lookup_instantiate_filp - instantiates the open intent filp
911 * @nd: pointer to nameidata
912 * @dentry: pointer to dentry
913 * @open: open callback
915 * Helper for filesystems that want to use lookup open intents and pass back
916 * a fully instantiated struct file to the caller.
917 * This function is meant to be called from within a filesystem's
919 * Beware of calling it for non-regular files! Those ->open methods might block
920 * (e.g. in fifo_open), leaving you with parent locked (and in case of fifo,
921 * leading to a deadlock, as nobody can open that fifo anymore, because
922 * another process to open fifo will block on locked parent when doing lookup).
923 * Note that in case of error, nd->intent.open.file is destroyed, but the
924 * path information remains valid.
925 * If the open callback is set to NULL, then the standard f_op->open()
926 * filesystem callback is substituted.
928 struct file *lookup_instantiate_filp(struct nameidata *nd, struct dentry *dentry,
929 int (*open)(struct inode *, struct file *))
931 if (IS_ERR(nd->intent.open.file))
935 nd->intent.open.file = __dentry_open(dget(dentry), mntget(nd->mnt),
936 nd->intent.open.flags - 1,
937 nd->intent.open.file,
940 return nd->intent.open.file;
942 release_open_intent(nd);
943 nd->intent.open.file = (struct file *)dentry;
946 EXPORT_SYMBOL_GPL(lookup_instantiate_filp);
949 * nameidata_to_filp - convert a nameidata to an open filp.
950 * @nd: pointer to nameidata
953 * Note that this function destroys the original nameidata
955 struct file *nameidata_to_filp(struct nameidata *nd, int flags)
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_dentry == NULL)
963 filp = __dentry_open(nd->dentry, nd->mnt, flags, filp, NULL);
970 * dentry_open() will have done dput(dentry) and mntput(mnt) if it returns an
973 struct file *dentry_open(struct dentry *dentry, struct vfsmount *mnt, int flags)
979 f = get_empty_filp();
983 return ERR_PTR(error);
986 return __dentry_open(dentry, mnt, flags, f, NULL);
988 EXPORT_SYMBOL(dentry_open);
991 * Find an empty file descriptor entry, and mark it busy.
993 int get_unused_fd(void)
995 struct files_struct * files = current->files;
1000 spin_lock(&files->file_lock);
1003 fdt = files_fdtable(files);
1004 fd = find_next_zero_bit(fdt->open_fds->fds_bits,
1009 * N.B. For clone tasks sharing a files structure, this test
1010 * will limit the total number of files that can be opened.
1012 if (fd >= current->signal->rlim[RLIMIT_NOFILE].rlim_cur)
1015 /* Do we need to expand the fd array or fd set? */
1016 error = expand_files(files, fd);
1022 * If we needed to expand the fs array we
1023 * might have blocked - try again.
1029 FD_SET(fd, fdt->open_fds);
1030 FD_CLR(fd, fdt->close_on_exec);
1031 files->next_fd = fd + 1;
1035 if (fdt->fd[fd] != NULL) {
1036 printk(KERN_WARNING "get_unused_fd: slot %d not NULL!\n", fd);
1043 spin_unlock(&files->file_lock);
1047 EXPORT_SYMBOL(get_unused_fd);
1049 static void __put_unused_fd(struct files_struct *files, unsigned int fd)
1051 struct fdtable *fdt = files_fdtable(files);
1052 __FD_CLR(fd, fdt->open_fds);
1053 if (fd < files->next_fd)
1054 files->next_fd = fd;
1058 void fastcall put_unused_fd(unsigned int fd)
1060 struct files_struct *files = current->files;
1061 spin_lock(&files->file_lock);
1062 __put_unused_fd(files, fd);
1063 spin_unlock(&files->file_lock);
1066 EXPORT_SYMBOL(put_unused_fd);
1069 * Install a file pointer in the fd array.
1071 * The VFS is full of places where we drop the files lock between
1072 * setting the open_fds bitmap and installing the file in the file
1073 * array. At any such point, we are vulnerable to a dup2() race
1074 * installing a file in the array before us. We need to detect this and
1075 * fput() the struct file we are about to overwrite in this case.
1077 * It should never happen - if we allow dup2() do it, _really_ bad things
1081 void fastcall fd_install(unsigned int fd, struct file * file)
1083 struct files_struct *files = current->files;
1084 struct fdtable *fdt;
1085 spin_lock(&files->file_lock);
1086 fdt = files_fdtable(files);
1087 BUG_ON(fdt->fd[fd] != NULL);
1088 rcu_assign_pointer(fdt->fd[fd], file);
1089 spin_unlock(&files->file_lock);
1092 EXPORT_SYMBOL(fd_install);
1094 long do_sys_open(int dfd, const char __user *filename, int flags, int mode)
1096 char *tmp = getname(filename);
1097 int fd = PTR_ERR(tmp);
1100 fd = get_unused_fd();
1102 struct file *f = do_filp_open(dfd, tmp, flags, mode);
1107 fsnotify_open(f->f_dentry);
1116 asmlinkage long sys_open(const char __user *filename, int flags, int mode)
1120 if (force_o_largefile())
1121 flags |= O_LARGEFILE;
1123 ret = do_sys_open(AT_FDCWD, filename, flags, mode);
1124 /* avoid REGPARM breakage on x86: */
1125 prevent_tail_call(ret);
1129 asmlinkage long sys_openat(int dfd, const char __user *filename, int flags,
1134 if (force_o_largefile())
1135 flags |= O_LARGEFILE;
1137 ret = do_sys_open(dfd, filename, flags, mode);
1138 /* avoid REGPARM breakage on x86: */
1139 prevent_tail_call(ret);
1146 * For backward compatibility? Maybe this should be moved
1147 * into arch/i386 instead?
1149 asmlinkage long sys_creat(const char __user * pathname, int mode)
1151 return sys_open(pathname, O_CREAT | O_WRONLY | O_TRUNC, mode);
1157 * "id" is the POSIX thread ID. We use the
1158 * files pointer for this..
1160 int filp_close(struct file *filp, fl_owner_t id)
1164 if (!file_count(filp)) {
1165 printk(KERN_ERR "VFS: Close: file count is 0\n");
1169 if (filp->f_op && filp->f_op->flush)
1170 retval = filp->f_op->flush(filp, id);
1172 dnotify_flush(filp, id);
1173 locks_remove_posix(filp, id);
1178 EXPORT_SYMBOL(filp_close);
1181 * Careful here! We test whether the file pointer is NULL before
1182 * releasing the fd. This ensures that one clone task can't release
1183 * an fd while another clone is opening it.
1185 asmlinkage long sys_close(unsigned int fd)
1188 struct files_struct *files = current->files;
1189 struct fdtable *fdt;
1191 spin_lock(&files->file_lock);
1192 fdt = files_fdtable(files);
1193 if (fd >= fdt->max_fds)
1198 rcu_assign_pointer(fdt->fd[fd], NULL);
1199 FD_CLR(fd, fdt->close_on_exec);
1200 __put_unused_fd(files, fd);
1201 spin_unlock(&files->file_lock);
1202 return filp_close(filp, files);
1205 spin_unlock(&files->file_lock);
1209 EXPORT_SYMBOL(sys_close);
1212 * This routine simulates a hangup on the tty, to arrange that users
1213 * are given clean terminals at login time.
1215 asmlinkage long sys_vhangup(void)
1217 if (capable(CAP_SYS_TTY_CONFIG)) {
1218 /* XXX: this needs locking */
1219 tty_vhangup(current->signal->tty);
1226 * Called when an inode is about to be open.
1227 * We use this to disallow opening large files on 32bit systems if
1228 * the caller didn't specify O_LARGEFILE. On 64bit systems we force
1229 * on this flag in sys_open.
1231 int generic_file_open(struct inode * inode, struct file * filp)
1233 if (!(filp->f_flags & O_LARGEFILE) && i_size_read(inode) > MAX_NON_LFS)
1238 EXPORT_SYMBOL(generic_file_open);
1241 * This is used by subsystems that don't want seekable
1244 int nonseekable_open(struct inode *inode, struct file *filp)
1246 filp->f_mode &= ~(FMODE_LSEEK | FMODE_PREAD | FMODE_PWRITE);
1250 EXPORT_SYMBOL(nonseekable_open);