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>
34 int vfs_statfs(struct dentry *dentry, struct kstatfs *buf)
40 if (dentry->d_sb->s_op->statfs) {
41 memset(buf, 0, sizeof(*buf));
42 retval = security_sb_statfs(dentry);
45 retval = dentry->d_sb->s_op->statfs(dentry, buf);
46 if (retval == 0 && buf->f_frsize == 0)
47 buf->f_frsize = buf->f_bsize;
53 EXPORT_SYMBOL(vfs_statfs);
55 static int vfs_statfs_native(struct dentry *dentry, struct statfs *buf)
60 retval = vfs_statfs(dentry, &st);
64 if (sizeof(*buf) == sizeof(st))
65 memcpy(buf, &st, sizeof(st));
67 if (sizeof buf->f_blocks == 4) {
68 if ((st.f_blocks | st.f_bfree | st.f_bavail |
69 st.f_bsize | st.f_frsize) &
70 0xffffffff00000000ULL)
73 * f_files and f_ffree may be -1; it's okay to stuff
76 if (st.f_files != -1 &&
77 (st.f_files & 0xffffffff00000000ULL))
79 if (st.f_ffree != -1 &&
80 (st.f_ffree & 0xffffffff00000000ULL))
84 buf->f_type = st.f_type;
85 buf->f_bsize = st.f_bsize;
86 buf->f_blocks = st.f_blocks;
87 buf->f_bfree = st.f_bfree;
88 buf->f_bavail = st.f_bavail;
89 buf->f_files = st.f_files;
90 buf->f_ffree = st.f_ffree;
91 buf->f_fsid = st.f_fsid;
92 buf->f_namelen = st.f_namelen;
93 buf->f_frsize = st.f_frsize;
94 memset(buf->f_spare, 0, sizeof(buf->f_spare));
99 static int vfs_statfs64(struct dentry *dentry, struct statfs64 *buf)
104 retval = vfs_statfs(dentry, &st);
108 if (sizeof(*buf) == sizeof(st))
109 memcpy(buf, &st, sizeof(st));
111 buf->f_type = st.f_type;
112 buf->f_bsize = st.f_bsize;
113 buf->f_blocks = st.f_blocks;
114 buf->f_bfree = st.f_bfree;
115 buf->f_bavail = st.f_bavail;
116 buf->f_files = st.f_files;
117 buf->f_ffree = st.f_ffree;
118 buf->f_fsid = st.f_fsid;
119 buf->f_namelen = st.f_namelen;
120 buf->f_frsize = st.f_frsize;
121 memset(buf->f_spare, 0, sizeof(buf->f_spare));
126 SYSCALL_DEFINE2(statfs, const char __user *, pathname, struct statfs __user *, buf)
131 error = user_path(pathname, &path);
134 error = vfs_statfs_native(path.dentry, &tmp);
135 if (!error && copy_to_user(buf, &tmp, sizeof(tmp)))
142 SYSCALL_DEFINE3(statfs64, const char __user *, pathname, size_t, sz, struct statfs64 __user *, buf)
147 if (sz != sizeof(*buf))
149 error = user_path(pathname, &path);
152 error = vfs_statfs64(path.dentry, &tmp);
153 if (!error && copy_to_user(buf, &tmp, sizeof(tmp)))
160 SYSCALL_DEFINE2(fstatfs, unsigned int, fd, struct statfs __user *, buf)
170 error = vfs_statfs_native(file->f_path.dentry, &tmp);
171 if (!error && copy_to_user(buf, &tmp, sizeof(tmp)))
178 SYSCALL_DEFINE3(fstatfs64, unsigned int, fd, size_t, sz, struct statfs64 __user *, buf)
184 if (sz != sizeof(*buf))
191 error = vfs_statfs64(file->f_path.dentry, &tmp);
192 if (!error && copy_to_user(buf, &tmp, sizeof(tmp)))
199 int do_truncate(struct dentry *dentry, loff_t length, unsigned int time_attrs,
203 struct iattr newattrs;
205 /* Not pretty: "inode->i_size" shouldn't really be signed. But it is. */
209 newattrs.ia_size = length;
210 newattrs.ia_valid = ATTR_SIZE | time_attrs;
212 newattrs.ia_file = filp;
213 newattrs.ia_valid |= ATTR_FILE;
216 /* Remove suid/sgid on truncate too */
217 ret = should_remove_suid(dentry);
219 newattrs.ia_valid |= ret | ATTR_FORCE;
221 mutex_lock(&dentry->d_inode->i_mutex);
222 ret = notify_change(dentry, &newattrs);
223 mutex_unlock(&dentry->d_inode->i_mutex);
227 static long do_sys_truncate(const char __user *pathname, loff_t length)
234 if (length < 0) /* sorry, but loff_t says... */
237 error = user_path(pathname, &path);
240 inode = path.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 = mnt_want_write(path.mnt);
255 error = inode_permission(inode, MAY_WRITE);
257 goto mnt_drop_write_and_out;
260 if (IS_APPEND(inode))
261 goto mnt_drop_write_and_out;
263 error = get_write_access(inode);
265 goto mnt_drop_write_and_out;
268 * Make sure that there are no leases. get_write_access() protects
269 * against the truncate racing with a lease-granting setlease().
271 error = break_lease(inode, FMODE_WRITE);
273 goto put_write_and_out;
275 error = locks_verify_truncate(inode, NULL, length);
277 error = security_path_truncate(&path, length, 0);
280 error = do_truncate(path.dentry, length, 0, NULL);
284 put_write_access(inode);
285 mnt_drop_write_and_out:
286 mnt_drop_write(path.mnt);
293 SYSCALL_DEFINE2(truncate, const char __user *, path, unsigned long, length)
295 /* on 32-bit boxen it will cut the range 2^31--2^32-1 off */
296 return do_sys_truncate(path, (long)length);
299 static long do_sys_ftruncate(unsigned int fd, loff_t length, int small)
301 struct inode * inode;
302 struct dentry *dentry;
314 /* explicitly opened as large or we are on 64-bit box */
315 if (file->f_flags & O_LARGEFILE)
318 dentry = file->f_path.dentry;
319 inode = dentry->d_inode;
321 if (!S_ISREG(inode->i_mode) || !(file->f_mode & FMODE_WRITE))
325 /* Cannot ftruncate over 2^31 bytes without large file support */
326 if (small && length > MAX_NON_LFS)
330 if (IS_APPEND(inode))
333 error = locks_verify_truncate(inode, file, length);
335 error = security_path_truncate(&file->f_path, length,
336 ATTR_MTIME|ATTR_CTIME);
338 error = do_truncate(dentry, length, ATTR_MTIME|ATTR_CTIME, file);
345 SYSCALL_DEFINE2(ftruncate, unsigned int, fd, unsigned long, length)
347 long ret = do_sys_ftruncate(fd, length, 1);
348 /* avoid REGPARM breakage on x86: */
349 asmlinkage_protect(2, ret, fd, length);
353 /* LFS versions of truncate are only needed on 32 bit machines */
354 #if BITS_PER_LONG == 32
355 SYSCALL_DEFINE(truncate64)(const char __user * path, loff_t length)
357 return do_sys_truncate(path, length);
359 #ifdef CONFIG_HAVE_SYSCALL_WRAPPERS
360 asmlinkage long SyS_truncate64(long path, loff_t length)
362 return SYSC_truncate64((const char __user *) path, length);
364 SYSCALL_ALIAS(sys_truncate64, SyS_truncate64);
367 SYSCALL_DEFINE(ftruncate64)(unsigned int fd, loff_t length)
369 long ret = do_sys_ftruncate(fd, length, 0);
370 /* avoid REGPARM breakage on x86: */
371 asmlinkage_protect(2, ret, fd, length);
374 #ifdef CONFIG_HAVE_SYSCALL_WRAPPERS
375 asmlinkage long SyS_ftruncate64(long fd, loff_t length)
377 return SYSC_ftruncate64((unsigned int) fd, length);
379 SYSCALL_ALIAS(sys_ftruncate64, SyS_ftruncate64);
381 #endif /* BITS_PER_LONG == 32 */
384 int do_fallocate(struct file *file, int mode, loff_t offset, loff_t len)
386 struct inode *inode = file->f_path.dentry->d_inode;
389 if (offset < 0 || len <= 0)
392 /* Return error if mode is not supported */
393 if (mode && !(mode & FALLOC_FL_KEEP_SIZE))
396 if (!(file->f_mode & FMODE_WRITE))
399 * Revalidate the write permissions, in case security policy has
400 * changed since the files were opened.
402 ret = security_file_permission(file, MAY_WRITE);
406 if (S_ISFIFO(inode->i_mode))
410 * Let individual file system decide if it supports preallocation
411 * for directories or not.
413 if (!S_ISREG(inode->i_mode) && !S_ISDIR(inode->i_mode))
416 /* Check for wrap through zero too */
417 if (((offset + len) > inode->i_sb->s_maxbytes) || ((offset + len) < 0))
420 if (!inode->i_op->fallocate)
423 return inode->i_op->fallocate(inode, mode, offset, len);
426 SYSCALL_DEFINE(fallocate)(int fd, int mode, loff_t offset, loff_t len)
433 error = do_fallocate(file, mode, offset, len);
440 #ifdef CONFIG_HAVE_SYSCALL_WRAPPERS
441 asmlinkage long SyS_fallocate(long fd, long mode, loff_t offset, loff_t len)
443 return SYSC_fallocate((int)fd, (int)mode, offset, len);
445 SYSCALL_ALIAS(sys_fallocate, SyS_fallocate);
449 * access() needs to use the real uid/gid, not the effective uid/gid.
450 * We do this by temporarily clearing all FS-related capabilities and
451 * switching the fsuid/fsgid around to the real ones.
453 SYSCALL_DEFINE3(faccessat, int, dfd, const char __user *, filename, int, mode)
455 const struct cred *old_cred;
456 struct cred *override_cred;
461 if (mode & ~S_IRWXO) /* where's F_OK, X_OK, W_OK, R_OK? */
464 override_cred = prepare_creds();
468 override_cred->fsuid = override_cred->uid;
469 override_cred->fsgid = override_cred->gid;
471 if (!issecure(SECURE_NO_SETUID_FIXUP)) {
472 /* Clear the capabilities if we switch to a non-root user */
473 if (override_cred->uid)
474 cap_clear(override_cred->cap_effective);
476 override_cred->cap_effective =
477 override_cred->cap_permitted;
480 old_cred = override_creds(override_cred);
482 res = user_path_at(dfd, filename, LOOKUP_FOLLOW, &path);
486 inode = path.dentry->d_inode;
488 if ((mode & MAY_EXEC) && S_ISREG(inode->i_mode)) {
490 * MAY_EXEC on regular files is denied if the fs is mounted
491 * with the "noexec" flag.
494 if (path.mnt->mnt_flags & MNT_NOEXEC)
495 goto out_path_release;
498 res = inode_permission(inode, mode | MAY_ACCESS);
499 /* SuS v2 requires we report a read only fs too */
500 if (res || !(mode & S_IWOTH) || special_file(inode->i_mode))
501 goto out_path_release;
503 * This is a rare case where using __mnt_is_readonly()
504 * is OK without a mnt_want/drop_write() pair. Since
505 * no actual write to the fs is performed here, we do
506 * not need to telegraph to that to anyone.
508 * By doing this, we accept that this access is
509 * inherently racy and know that the fs may change
510 * state before we even see this result.
512 if (__mnt_is_readonly(path.mnt))
518 revert_creds(old_cred);
519 put_cred(override_cred);
523 SYSCALL_DEFINE2(access, const char __user *, filename, int, mode)
525 return sys_faccessat(AT_FDCWD, filename, mode);
528 SYSCALL_DEFINE1(chdir, const char __user *, filename)
533 error = user_path_dir(filename, &path);
537 error = inode_permission(path.dentry->d_inode, MAY_EXEC | MAY_ACCESS);
541 set_fs_pwd(current->fs, &path);
549 SYSCALL_DEFINE1(fchdir, unsigned int, fd)
560 inode = file->f_path.dentry->d_inode;
563 if (!S_ISDIR(inode->i_mode))
566 error = inode_permission(inode, MAY_EXEC | MAY_ACCESS);
568 set_fs_pwd(current->fs, &file->f_path);
575 SYSCALL_DEFINE1(chroot, const char __user *, filename)
580 error = user_path_dir(filename, &path);
584 error = inode_permission(path.dentry->d_inode, MAY_EXEC | MAY_ACCESS);
589 if (!capable(CAP_SYS_CHROOT))
592 set_fs_root(current->fs, &path);
600 SYSCALL_DEFINE2(fchmod, unsigned int, fd, mode_t, mode)
602 struct inode * inode;
603 struct dentry * dentry;
606 struct iattr newattrs;
612 dentry = file->f_path.dentry;
613 inode = dentry->d_inode;
615 audit_inode(NULL, dentry);
617 err = mnt_want_write_file(file);
620 mutex_lock(&inode->i_mutex);
621 if (mode == (mode_t) -1)
622 mode = inode->i_mode;
623 newattrs.ia_mode = (mode & S_IALLUGO) | (inode->i_mode & ~S_IALLUGO);
624 newattrs.ia_valid = ATTR_MODE | ATTR_CTIME;
625 err = notify_change(dentry, &newattrs);
626 mutex_unlock(&inode->i_mutex);
627 mnt_drop_write(file->f_path.mnt);
634 SYSCALL_DEFINE3(fchmodat, int, dfd, const char __user *, filename, mode_t, mode)
639 struct iattr newattrs;
641 error = user_path_at(dfd, filename, LOOKUP_FOLLOW, &path);
644 inode = path.dentry->d_inode;
646 error = mnt_want_write(path.mnt);
649 mutex_lock(&inode->i_mutex);
650 if (mode == (mode_t) -1)
651 mode = inode->i_mode;
652 newattrs.ia_mode = (mode & S_IALLUGO) | (inode->i_mode & ~S_IALLUGO);
653 newattrs.ia_valid = ATTR_MODE | ATTR_CTIME;
654 error = notify_change(path.dentry, &newattrs);
655 mutex_unlock(&inode->i_mutex);
656 mnt_drop_write(path.mnt);
663 SYSCALL_DEFINE2(chmod, const char __user *, filename, mode_t, mode)
665 return sys_fchmodat(AT_FDCWD, filename, mode);
668 static int chown_common(struct dentry * dentry, uid_t user, gid_t group)
670 struct inode *inode = dentry->d_inode;
672 struct iattr newattrs;
674 newattrs.ia_valid = ATTR_CTIME;
675 if (user != (uid_t) -1) {
676 newattrs.ia_valid |= ATTR_UID;
677 newattrs.ia_uid = user;
679 if (group != (gid_t) -1) {
680 newattrs.ia_valid |= ATTR_GID;
681 newattrs.ia_gid = group;
683 if (!S_ISDIR(inode->i_mode))
685 ATTR_KILL_SUID | ATTR_KILL_SGID | ATTR_KILL_PRIV;
686 mutex_lock(&inode->i_mutex);
687 error = notify_change(dentry, &newattrs);
688 mutex_unlock(&inode->i_mutex);
693 SYSCALL_DEFINE3(chown, const char __user *, filename, uid_t, user, gid_t, group)
698 error = user_path(filename, &path);
701 error = mnt_want_write(path.mnt);
704 error = chown_common(path.dentry, user, group);
705 mnt_drop_write(path.mnt);
712 SYSCALL_DEFINE5(fchownat, int, dfd, const char __user *, filename, uid_t, user,
713 gid_t, group, int, flag)
719 if ((flag & ~AT_SYMLINK_NOFOLLOW) != 0)
722 follow = (flag & AT_SYMLINK_NOFOLLOW) ? 0 : LOOKUP_FOLLOW;
723 error = user_path_at(dfd, filename, follow, &path);
726 error = mnt_want_write(path.mnt);
729 error = chown_common(path.dentry, user, group);
730 mnt_drop_write(path.mnt);
737 SYSCALL_DEFINE3(lchown, const char __user *, filename, uid_t, user, gid_t, group)
742 error = user_lpath(filename, &path);
745 error = mnt_want_write(path.mnt);
748 error = chown_common(path.dentry, user, group);
749 mnt_drop_write(path.mnt);
756 SYSCALL_DEFINE3(fchown, unsigned int, fd, uid_t, user, gid_t, group)
760 struct dentry * dentry;
766 error = mnt_want_write_file(file);
769 dentry = file->f_path.dentry;
770 audit_inode(NULL, dentry);
771 error = chown_common(dentry, user, group);
772 mnt_drop_write(file->f_path.mnt);
780 * You have to be very careful that these write
781 * counts get cleaned up in error cases and
782 * upon __fput(). This should probably never
783 * be called outside of __dentry_open().
785 static inline int __get_file_write_access(struct inode *inode,
786 struct vfsmount *mnt)
789 error = get_write_access(inode);
793 * Do not take mount writer counts on
794 * special files since no writes to
795 * the mount itself will occur.
797 if (!special_file(inode->i_mode)) {
799 * Balanced in __fput()
801 error = mnt_want_write(mnt);
803 put_write_access(inode);
808 static struct file *__dentry_open(struct dentry *dentry, struct vfsmount *mnt,
809 int flags, struct file *f,
810 int (*open)(struct inode *, struct file *),
811 const struct cred *cred)
817 f->f_mode = (__force fmode_t)((flags+1) & O_ACCMODE) | FMODE_LSEEK |
818 FMODE_PREAD | FMODE_PWRITE;
819 inode = dentry->d_inode;
820 if (f->f_mode & FMODE_WRITE) {
821 error = __get_file_write_access(inode, mnt);
824 if (!special_file(inode->i_mode))
828 f->f_mapping = inode->i_mapping;
829 f->f_path.dentry = dentry;
832 f->f_op = fops_get(inode->i_fop);
833 file_move(f, &inode->i_sb->s_files);
835 error = security_dentry_open(f, cred);
839 if (!open && f->f_op)
840 open = f->f_op->open;
842 error = open(inode, f);
847 f->f_flags &= ~(O_CREAT | O_EXCL | O_NOCTTY | O_TRUNC);
849 file_ra_state_init(&f->f_ra, f->f_mapping->host->i_mapping);
851 /* NB: we're sure to have correct a_ops only after f_op->open */
852 if (f->f_flags & O_DIRECT) {
853 if (!f->f_mapping->a_ops ||
854 ((!f->f_mapping->a_ops->direct_IO) &&
855 (!f->f_mapping->a_ops->get_xip_mem))) {
857 f = ERR_PTR(-EINVAL);
865 if (f->f_mode & FMODE_WRITE) {
866 put_write_access(inode);
867 if (!special_file(inode->i_mode)) {
869 * We don't consider this a real
870 * mnt_want/drop_write() pair
871 * because it all happenend right
872 * here, so just reset the state.
879 f->f_path.dentry = NULL;
880 f->f_path.mnt = NULL;
885 return ERR_PTR(error);
889 * lookup_instantiate_filp - instantiates the open intent filp
890 * @nd: pointer to nameidata
891 * @dentry: pointer to dentry
892 * @open: open callback
894 * Helper for filesystems that want to use lookup open intents and pass back
895 * a fully instantiated struct file to the caller.
896 * This function is meant to be called from within a filesystem's
898 * Beware of calling it for non-regular files! Those ->open methods might block
899 * (e.g. in fifo_open), leaving you with parent locked (and in case of fifo,
900 * leading to a deadlock, as nobody can open that fifo anymore, because
901 * another process to open fifo will block on locked parent when doing lookup).
902 * Note that in case of error, nd->intent.open.file is destroyed, but the
903 * path information remains valid.
904 * If the open callback is set to NULL, then the standard f_op->open()
905 * filesystem callback is substituted.
907 struct file *lookup_instantiate_filp(struct nameidata *nd, struct dentry *dentry,
908 int (*open)(struct inode *, struct file *))
910 const struct cred *cred = current_cred();
912 if (IS_ERR(nd->intent.open.file))
916 nd->intent.open.file = __dentry_open(dget(dentry), mntget(nd->path.mnt),
917 nd->intent.open.flags - 1,
918 nd->intent.open.file,
921 return nd->intent.open.file;
923 release_open_intent(nd);
924 nd->intent.open.file = (struct file *)dentry;
927 EXPORT_SYMBOL_GPL(lookup_instantiate_filp);
930 * nameidata_to_filp - convert a nameidata to an open filp.
931 * @nd: pointer to nameidata
934 * Note that this function destroys the original nameidata
936 struct file *nameidata_to_filp(struct nameidata *nd, int flags)
938 const struct cred *cred = current_cred();
941 /* Pick up the filp from the open intent */
942 filp = nd->intent.open.file;
943 /* Has the filesystem initialised the file for us? */
944 if (filp->f_path.dentry == NULL)
945 filp = __dentry_open(nd->path.dentry, nd->path.mnt, flags, filp,
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,
957 const struct cred *cred)
962 validate_creds(cred);
965 * We must always pass in a valid mount pointer. Historically
966 * callers got away with not passing it, but we must enforce this at
967 * the earliest possible point now to avoid strange problems deep in the
971 printk(KERN_WARNING "%s called with NULL vfsmount\n", __func__);
973 return ERR_PTR(-EINVAL);
977 f = get_empty_filp();
981 return ERR_PTR(error);
984 return __dentry_open(dentry, mnt, flags, f, NULL, cred);
986 EXPORT_SYMBOL(dentry_open);
988 static void __put_unused_fd(struct files_struct *files, unsigned int fd)
990 struct fdtable *fdt = files_fdtable(files);
991 __FD_CLR(fd, fdt->open_fds);
992 if (fd < files->next_fd)
996 void put_unused_fd(unsigned int fd)
998 struct files_struct *files = current->files;
999 spin_lock(&files->file_lock);
1000 __put_unused_fd(files, fd);
1001 spin_unlock(&files->file_lock);
1004 EXPORT_SYMBOL(put_unused_fd);
1007 * Install a file pointer in the fd array.
1009 * The VFS is full of places where we drop the files lock between
1010 * setting the open_fds bitmap and installing the file in the file
1011 * array. At any such point, we are vulnerable to a dup2() race
1012 * installing a file in the array before us. We need to detect this and
1013 * fput() the struct file we are about to overwrite in this case.
1015 * It should never happen - if we allow dup2() do it, _really_ bad things
1019 void fd_install(unsigned int fd, struct file *file)
1021 struct files_struct *files = current->files;
1022 struct fdtable *fdt;
1023 spin_lock(&files->file_lock);
1024 fdt = files_fdtable(files);
1025 BUG_ON(fdt->fd[fd] != NULL);
1026 rcu_assign_pointer(fdt->fd[fd], file);
1027 spin_unlock(&files->file_lock);
1030 EXPORT_SYMBOL(fd_install);
1032 long do_sys_open(int dfd, const char __user *filename, int flags, int mode)
1034 char *tmp = getname(filename);
1035 int fd = PTR_ERR(tmp);
1038 fd = get_unused_fd_flags(flags);
1040 struct file *f = do_filp_open(dfd, tmp, flags, mode, 0);
1045 fsnotify_open(f->f_path.dentry);
1054 SYSCALL_DEFINE3(open, const char __user *, filename, int, flags, int, mode)
1058 if (force_o_largefile())
1059 flags |= O_LARGEFILE;
1061 ret = do_sys_open(AT_FDCWD, filename, flags, mode);
1062 /* avoid REGPARM breakage on x86: */
1063 asmlinkage_protect(3, ret, filename, flags, mode);
1067 SYSCALL_DEFINE4(openat, int, dfd, const char __user *, filename, int, flags,
1072 if (force_o_largefile())
1073 flags |= O_LARGEFILE;
1075 ret = do_sys_open(dfd, filename, flags, mode);
1076 /* avoid REGPARM breakage on x86: */
1077 asmlinkage_protect(4, ret, dfd, filename, flags, mode);
1084 * For backward compatibility? Maybe this should be moved
1085 * into arch/i386 instead?
1087 SYSCALL_DEFINE2(creat, const char __user *, pathname, int, mode)
1089 return sys_open(pathname, O_CREAT | O_WRONLY | O_TRUNC, mode);
1095 * "id" is the POSIX thread ID. We use the
1096 * files pointer for this..
1098 int filp_close(struct file *filp, fl_owner_t id)
1102 if (!file_count(filp)) {
1103 printk(KERN_ERR "VFS: Close: file count is 0\n");
1107 if (filp->f_op && filp->f_op->flush)
1108 retval = filp->f_op->flush(filp, id);
1110 dnotify_flush(filp, id);
1111 locks_remove_posix(filp, id);
1116 EXPORT_SYMBOL(filp_close);
1119 * Careful here! We test whether the file pointer is NULL before
1120 * releasing the fd. This ensures that one clone task can't release
1121 * an fd while another clone is opening it.
1123 SYSCALL_DEFINE1(close, unsigned int, fd)
1126 struct files_struct *files = current->files;
1127 struct fdtable *fdt;
1130 spin_lock(&files->file_lock);
1131 fdt = files_fdtable(files);
1132 if (fd >= fdt->max_fds)
1137 rcu_assign_pointer(fdt->fd[fd], NULL);
1138 FD_CLR(fd, fdt->close_on_exec);
1139 __put_unused_fd(files, fd);
1140 spin_unlock(&files->file_lock);
1141 retval = filp_close(filp, files);
1143 /* can't restart close syscall because file table entry was cleared */
1144 if (unlikely(retval == -ERESTARTSYS ||
1145 retval == -ERESTARTNOINTR ||
1146 retval == -ERESTARTNOHAND ||
1147 retval == -ERESTART_RESTARTBLOCK))
1153 spin_unlock(&files->file_lock);
1156 EXPORT_SYMBOL(sys_close);
1159 * This routine simulates a hangup on the tty, to arrange that users
1160 * are given clean terminals at login time.
1162 SYSCALL_DEFINE0(vhangup)
1164 if (capable(CAP_SYS_TTY_CONFIG)) {
1172 * Called when an inode is about to be open.
1173 * We use this to disallow opening large files on 32bit systems if
1174 * the caller didn't specify O_LARGEFILE. On 64bit systems we force
1175 * on this flag in sys_open.
1177 int generic_file_open(struct inode * inode, struct file * filp)
1179 if (!(filp->f_flags & O_LARGEFILE) && i_size_read(inode) > MAX_NON_LFS)
1184 EXPORT_SYMBOL(generic_file_open);
1187 * This is used by subsystems that don't want seekable
1190 int nonseekable_open(struct inode *inode, struct file *filp)
1192 filp->f_mode &= ~(FMODE_LSEEK | FMODE_PREAD | FMODE_PWRITE);
1196 EXPORT_SYMBOL(nonseekable_open);