1 // SPDX-License-Identifier: GPL-2.0-only
5 * Copyright (C) 1991, 1992 Linus Torvalds
8 #include <linux/string.h>
10 #include <linux/file.h>
11 #include <linux/fdtable.h>
12 #include <linux/fsnotify.h>
13 #include <linux/module.h>
14 #include <linux/tty.h>
15 #include <linux/namei.h>
16 #include <linux/backing-dev.h>
17 #include <linux/capability.h>
18 #include <linux/securebits.h>
19 #include <linux/security.h>
20 #include <linux/mount.h>
21 #include <linux/fcntl.h>
22 #include <linux/slab.h>
23 #include <linux/uaccess.h>
25 #include <linux/personality.h>
26 #include <linux/pagemap.h>
27 #include <linux/syscalls.h>
28 #include <linux/rcupdate.h>
29 #include <linux/audit.h>
30 #include <linux/falloc.h>
31 #include <linux/fs_struct.h>
32 #include <linux/ima.h>
33 #include <linux/dnotify.h>
34 #include <linux/compat.h>
38 int do_truncate(struct dentry *dentry, loff_t length, unsigned int time_attrs,
42 struct iattr newattrs;
44 /* Not pretty: "inode->i_size" shouldn't really be signed. But it is. */
48 newattrs.ia_size = length;
49 newattrs.ia_valid = ATTR_SIZE | time_attrs;
51 newattrs.ia_file = filp;
52 newattrs.ia_valid |= ATTR_FILE;
55 /* Remove suid, sgid, and file capabilities on truncate too */
56 ret = dentry_needs_remove_privs(dentry);
60 newattrs.ia_valid |= ret | ATTR_FORCE;
62 inode_lock(dentry->d_inode);
63 /* Note any delegations or leases have already been broken: */
64 ret = notify_change(dentry, &newattrs, NULL);
65 inode_unlock(dentry->d_inode);
69 long vfs_truncate(const struct path *path, loff_t length)
74 inode = path->dentry->d_inode;
76 /* For directories it's -EISDIR, for other non-regulars - -EINVAL */
77 if (S_ISDIR(inode->i_mode))
79 if (!S_ISREG(inode->i_mode))
82 error = mnt_want_write(path->mnt);
86 error = inode_permission(inode, MAY_WRITE);
88 goto mnt_drop_write_and_out;
92 goto mnt_drop_write_and_out;
94 error = get_write_access(inode);
96 goto mnt_drop_write_and_out;
99 * Make sure that there are no leases. get_write_access() protects
100 * against the truncate racing with a lease-granting setlease().
102 error = break_lease(inode, O_WRONLY);
104 goto put_write_and_out;
106 error = locks_verify_truncate(inode, NULL, length);
108 error = security_path_truncate(path);
110 error = do_truncate(path->dentry, length, 0, NULL);
113 put_write_access(inode);
114 mnt_drop_write_and_out:
115 mnt_drop_write(path->mnt);
119 EXPORT_SYMBOL_GPL(vfs_truncate);
121 long do_sys_truncate(const char __user *pathname, loff_t length)
123 unsigned int lookup_flags = LOOKUP_FOLLOW;
127 if (length < 0) /* sorry, but loff_t says... */
131 error = user_path_at(AT_FDCWD, pathname, lookup_flags, &path);
133 error = vfs_truncate(&path, length);
136 if (retry_estale(error, lookup_flags)) {
137 lookup_flags |= LOOKUP_REVAL;
143 SYSCALL_DEFINE2(truncate, const char __user *, path, long, length)
145 return do_sys_truncate(path, length);
149 COMPAT_SYSCALL_DEFINE2(truncate, const char __user *, path, compat_off_t, length)
151 return do_sys_truncate(path, length);
155 long do_sys_ftruncate(unsigned int fd, loff_t length, int small)
158 struct dentry *dentry;
170 /* explicitly opened as large or we are on 64-bit box */
171 if (f.file->f_flags & O_LARGEFILE)
174 dentry = f.file->f_path.dentry;
175 inode = dentry->d_inode;
177 if (!S_ISREG(inode->i_mode) || !(f.file->f_mode & FMODE_WRITE))
181 /* Cannot ftruncate over 2^31 bytes without large file support */
182 if (small && length > MAX_NON_LFS)
186 /* Check IS_APPEND on real upper inode */
187 if (IS_APPEND(file_inode(f.file)))
190 sb_start_write(inode->i_sb);
191 error = locks_verify_truncate(inode, f.file, length);
193 error = security_path_truncate(&f.file->f_path);
195 error = do_truncate(dentry, length, ATTR_MTIME|ATTR_CTIME, f.file);
196 sb_end_write(inode->i_sb);
203 SYSCALL_DEFINE2(ftruncate, unsigned int, fd, unsigned long, length)
205 return do_sys_ftruncate(fd, length, 1);
209 COMPAT_SYSCALL_DEFINE2(ftruncate, unsigned int, fd, compat_ulong_t, length)
211 return do_sys_ftruncate(fd, length, 1);
215 /* LFS versions of truncate are only needed on 32 bit machines */
216 #if BITS_PER_LONG == 32
217 SYSCALL_DEFINE2(truncate64, const char __user *, path, loff_t, length)
219 return do_sys_truncate(path, length);
222 SYSCALL_DEFINE2(ftruncate64, unsigned int, fd, loff_t, length)
224 return do_sys_ftruncate(fd, length, 0);
226 #endif /* BITS_PER_LONG == 32 */
229 int vfs_fallocate(struct file *file, int mode, loff_t offset, loff_t len)
231 struct inode *inode = file_inode(file);
234 if (offset < 0 || len <= 0)
237 /* Return error if mode is not supported */
238 if (mode & ~FALLOC_FL_SUPPORTED_MASK)
241 /* Punch hole and zero range are mutually exclusive */
242 if ((mode & (FALLOC_FL_PUNCH_HOLE | FALLOC_FL_ZERO_RANGE)) ==
243 (FALLOC_FL_PUNCH_HOLE | FALLOC_FL_ZERO_RANGE))
246 /* Punch hole must have keep size set */
247 if ((mode & FALLOC_FL_PUNCH_HOLE) &&
248 !(mode & FALLOC_FL_KEEP_SIZE))
251 /* Collapse range should only be used exclusively. */
252 if ((mode & FALLOC_FL_COLLAPSE_RANGE) &&
253 (mode & ~FALLOC_FL_COLLAPSE_RANGE))
256 /* Insert range should only be used exclusively. */
257 if ((mode & FALLOC_FL_INSERT_RANGE) &&
258 (mode & ~FALLOC_FL_INSERT_RANGE))
261 /* Unshare range should only be used with allocate mode. */
262 if ((mode & FALLOC_FL_UNSHARE_RANGE) &&
263 (mode & ~(FALLOC_FL_UNSHARE_RANGE | FALLOC_FL_KEEP_SIZE)))
266 if (!(file->f_mode & FMODE_WRITE))
270 * We can only allow pure fallocate on append only files
272 if ((mode & ~FALLOC_FL_KEEP_SIZE) && IS_APPEND(inode))
275 if (IS_IMMUTABLE(inode))
279 * We cannot allow any fallocate operation on an active swapfile
281 if (IS_SWAPFILE(inode))
285 * Revalidate the write permissions, in case security policy has
286 * changed since the files were opened.
288 ret = security_file_permission(file, MAY_WRITE);
292 if (S_ISFIFO(inode->i_mode))
295 if (S_ISDIR(inode->i_mode))
298 if (!S_ISREG(inode->i_mode) && !S_ISBLK(inode->i_mode))
301 /* Check for wrap through zero too */
302 if (((offset + len) > inode->i_sb->s_maxbytes) || ((offset + len) < 0))
305 if (!file->f_op->fallocate)
308 file_start_write(file);
309 ret = file->f_op->fallocate(file, mode, offset, len);
312 * Create inotify and fanotify events.
314 * To keep the logic simple always create events if fallocate succeeds.
315 * This implies that events are even created if the file size remains
316 * unchanged, e.g. when using flag FALLOC_FL_KEEP_SIZE.
319 fsnotify_modify(file);
321 file_end_write(file);
324 EXPORT_SYMBOL_GPL(vfs_fallocate);
326 int ksys_fallocate(int fd, int mode, loff_t offset, loff_t len)
328 struct fd f = fdget(fd);
332 error = vfs_fallocate(f.file, mode, offset, len);
338 SYSCALL_DEFINE4(fallocate, int, fd, int, mode, loff_t, offset, loff_t, len)
340 return ksys_fallocate(fd, mode, offset, len);
344 * access() needs to use the real uid/gid, not the effective uid/gid.
345 * We do this by temporarily clearing all FS-related capabilities and
346 * switching the fsuid/fsgid around to the real ones.
348 long do_faccessat(int dfd, const char __user *filename, int mode)
350 const struct cred *old_cred;
351 struct cred *override_cred;
355 unsigned int lookup_flags = LOOKUP_FOLLOW;
357 if (mode & ~S_IRWXO) /* where's F_OK, X_OK, W_OK, R_OK? */
360 override_cred = prepare_creds();
364 override_cred->fsuid = override_cred->uid;
365 override_cred->fsgid = override_cred->gid;
367 if (!issecure(SECURE_NO_SETUID_FIXUP)) {
368 /* Clear the capabilities if we switch to a non-root user */
369 kuid_t root_uid = make_kuid(override_cred->user_ns, 0);
370 if (!uid_eq(override_cred->uid, root_uid))
371 cap_clear(override_cred->cap_effective);
373 override_cred->cap_effective =
374 override_cred->cap_permitted;
378 * The new set of credentials can *only* be used in
379 * task-synchronous circumstances, and does not need
380 * RCU freeing, unless somebody then takes a separate
383 * NOTE! This is _only_ true because this credential
384 * is used purely for override_creds() that installs
385 * it as the subjective cred. Other threads will be
386 * accessing ->real_cred, not the subjective cred.
388 * If somebody _does_ make a copy of this (using the
389 * 'get_current_cred()' function), that will clear the
390 * non_rcu field, because now that other user may be
391 * expecting RCU freeing. But normal thread-synchronous
392 * cred accesses will keep things non-RCY.
394 override_cred->non_rcu = 1;
396 old_cred = override_creds(override_cred);
398 res = user_path_at(dfd, filename, lookup_flags, &path);
402 inode = d_backing_inode(path.dentry);
404 if ((mode & MAY_EXEC) && S_ISREG(inode->i_mode)) {
406 * MAY_EXEC on regular files is denied if the fs is mounted
407 * with the "noexec" flag.
410 if (path_noexec(&path))
411 goto out_path_release;
414 res = inode_permission(inode, mode | MAY_ACCESS);
415 /* SuS v2 requires we report a read only fs too */
416 if (res || !(mode & S_IWOTH) || special_file(inode->i_mode))
417 goto out_path_release;
419 * This is a rare case where using __mnt_is_readonly()
420 * is OK without a mnt_want/drop_write() pair. Since
421 * no actual write to the fs is performed here, we do
422 * not need to telegraph to that to anyone.
424 * By doing this, we accept that this access is
425 * inherently racy and know that the fs may change
426 * state before we even see this result.
428 if (__mnt_is_readonly(path.mnt))
433 if (retry_estale(res, lookup_flags)) {
434 lookup_flags |= LOOKUP_REVAL;
438 revert_creds(old_cred);
439 put_cred(override_cred);
443 SYSCALL_DEFINE3(faccessat, int, dfd, const char __user *, filename, int, mode)
445 return do_faccessat(dfd, filename, mode);
448 SYSCALL_DEFINE2(access, const char __user *, filename, int, mode)
450 return do_faccessat(AT_FDCWD, filename, mode);
453 int ksys_chdir(const char __user *filename)
457 unsigned int lookup_flags = LOOKUP_FOLLOW | LOOKUP_DIRECTORY;
459 error = user_path_at(AT_FDCWD, filename, lookup_flags, &path);
463 error = inode_permission(path.dentry->d_inode, MAY_EXEC | MAY_CHDIR);
467 set_fs_pwd(current->fs, &path);
471 if (retry_estale(error, lookup_flags)) {
472 lookup_flags |= LOOKUP_REVAL;
479 SYSCALL_DEFINE1(chdir, const char __user *, filename)
481 return ksys_chdir(filename);
484 SYSCALL_DEFINE1(fchdir, unsigned int, fd)
486 struct fd f = fdget_raw(fd);
494 if (!d_can_lookup(f.file->f_path.dentry))
497 error = inode_permission(file_inode(f.file), MAY_EXEC | MAY_CHDIR);
499 set_fs_pwd(current->fs, &f.file->f_path);
506 int ksys_chroot(const char __user *filename)
510 unsigned int lookup_flags = LOOKUP_FOLLOW | LOOKUP_DIRECTORY;
512 error = user_path_at(AT_FDCWD, filename, lookup_flags, &path);
516 error = inode_permission(path.dentry->d_inode, MAY_EXEC | MAY_CHDIR);
521 if (!ns_capable(current_user_ns(), CAP_SYS_CHROOT))
523 error = security_path_chroot(&path);
527 set_fs_root(current->fs, &path);
531 if (retry_estale(error, lookup_flags)) {
532 lookup_flags |= LOOKUP_REVAL;
539 SYSCALL_DEFINE1(chroot, const char __user *, filename)
541 return ksys_chroot(filename);
544 static int chmod_common(const struct path *path, umode_t mode)
546 struct inode *inode = path->dentry->d_inode;
547 struct inode *delegated_inode = NULL;
548 struct iattr newattrs;
551 error = mnt_want_write(path->mnt);
556 error = security_path_chmod(path, mode);
559 newattrs.ia_mode = (mode & S_IALLUGO) | (inode->i_mode & ~S_IALLUGO);
560 newattrs.ia_valid = ATTR_MODE | ATTR_CTIME;
561 error = notify_change(path->dentry, &newattrs, &delegated_inode);
564 if (delegated_inode) {
565 error = break_deleg_wait(&delegated_inode);
569 mnt_drop_write(path->mnt);
573 int ksys_fchmod(unsigned int fd, umode_t mode)
575 struct fd f = fdget(fd);
580 err = chmod_common(&f.file->f_path, mode);
586 SYSCALL_DEFINE2(fchmod, unsigned int, fd, umode_t, mode)
588 return ksys_fchmod(fd, mode);
591 int do_fchmodat(int dfd, const char __user *filename, umode_t mode)
595 unsigned int lookup_flags = LOOKUP_FOLLOW;
597 error = user_path_at(dfd, filename, lookup_flags, &path);
599 error = chmod_common(&path, mode);
601 if (retry_estale(error, lookup_flags)) {
602 lookup_flags |= LOOKUP_REVAL;
609 SYSCALL_DEFINE3(fchmodat, int, dfd, const char __user *, filename,
612 return do_fchmodat(dfd, filename, mode);
615 SYSCALL_DEFINE2(chmod, const char __user *, filename, umode_t, mode)
617 return do_fchmodat(AT_FDCWD, filename, mode);
620 static int chown_common(const struct path *path, uid_t user, gid_t group)
622 struct inode *inode = path->dentry->d_inode;
623 struct inode *delegated_inode = NULL;
625 struct iattr newattrs;
629 uid = make_kuid(current_user_ns(), user);
630 gid = make_kgid(current_user_ns(), group);
633 newattrs.ia_valid = ATTR_CTIME;
634 if (user != (uid_t) -1) {
637 newattrs.ia_valid |= ATTR_UID;
638 newattrs.ia_uid = uid;
640 if (group != (gid_t) -1) {
643 newattrs.ia_valid |= ATTR_GID;
644 newattrs.ia_gid = gid;
646 if (!S_ISDIR(inode->i_mode))
648 ATTR_KILL_SUID | ATTR_KILL_SGID | ATTR_KILL_PRIV;
650 error = security_path_chown(path, uid, gid);
652 error = notify_change(path->dentry, &newattrs, &delegated_inode);
654 if (delegated_inode) {
655 error = break_deleg_wait(&delegated_inode);
662 int do_fchownat(int dfd, const char __user *filename, uid_t user, gid_t group,
669 if ((flag & ~(AT_SYMLINK_NOFOLLOW | AT_EMPTY_PATH)) != 0)
672 lookup_flags = (flag & AT_SYMLINK_NOFOLLOW) ? 0 : LOOKUP_FOLLOW;
673 if (flag & AT_EMPTY_PATH)
674 lookup_flags |= LOOKUP_EMPTY;
676 error = user_path_at(dfd, filename, lookup_flags, &path);
679 error = mnt_want_write(path.mnt);
682 error = chown_common(&path, user, group);
683 mnt_drop_write(path.mnt);
686 if (retry_estale(error, lookup_flags)) {
687 lookup_flags |= LOOKUP_REVAL;
694 SYSCALL_DEFINE5(fchownat, int, dfd, const char __user *, filename, uid_t, user,
695 gid_t, group, int, flag)
697 return do_fchownat(dfd, filename, user, group, flag);
700 SYSCALL_DEFINE3(chown, const char __user *, filename, uid_t, user, gid_t, group)
702 return do_fchownat(AT_FDCWD, filename, user, group, 0);
705 SYSCALL_DEFINE3(lchown, const char __user *, filename, uid_t, user, gid_t, group)
707 return do_fchownat(AT_FDCWD, filename, user, group,
708 AT_SYMLINK_NOFOLLOW);
711 int ksys_fchown(unsigned int fd, uid_t user, gid_t group)
713 struct fd f = fdget(fd);
719 error = mnt_want_write_file(f.file);
723 error = chown_common(&f.file->f_path, user, group);
724 mnt_drop_write_file(f.file);
731 SYSCALL_DEFINE3(fchown, unsigned int, fd, uid_t, user, gid_t, group)
733 return ksys_fchown(fd, user, group);
736 static int do_dentry_open(struct file *f,
738 int (*open)(struct inode *, struct file *))
740 static const struct file_operations empty_fops = {};
743 path_get(&f->f_path);
745 f->f_mapping = inode->i_mapping;
747 /* Ensure that we skip any errors that predate opening of the file */
748 f->f_wb_err = filemap_sample_wb_err(f->f_mapping);
750 if (unlikely(f->f_flags & O_PATH)) {
751 f->f_mode = FMODE_PATH | FMODE_OPENED;
752 f->f_op = &empty_fops;
756 /* Any file opened for execve()/uselib() has to be a regular file. */
757 if (unlikely(f->f_flags & FMODE_EXEC && !S_ISREG(inode->i_mode))) {
762 if (f->f_mode & FMODE_WRITE && !special_file(inode->i_mode)) {
763 error = get_write_access(inode);
766 error = __mnt_want_write(f->f_path.mnt);
767 if (unlikely(error)) {
768 put_write_access(inode);
771 f->f_mode |= FMODE_WRITER;
774 /* POSIX.1-2008/SUSv4 Section XSI 2.9.7 */
775 if (S_ISREG(inode->i_mode) || S_ISDIR(inode->i_mode))
776 f->f_mode |= FMODE_ATOMIC_POS;
778 f->f_op = fops_get(inode->i_fop);
779 if (WARN_ON(!f->f_op)) {
784 error = security_file_open(f);
788 error = break_lease(locks_inode(f), f->f_flags);
792 /* normally all 3 are set; ->open() can clear them if needed */
793 f->f_mode |= FMODE_LSEEK | FMODE_PREAD | FMODE_PWRITE;
795 open = f->f_op->open;
797 error = open(inode, f);
801 f->f_mode |= FMODE_OPENED;
802 if ((f->f_mode & (FMODE_READ | FMODE_WRITE)) == FMODE_READ)
803 i_readcount_inc(inode);
804 if ((f->f_mode & FMODE_READ) &&
805 likely(f->f_op->read || f->f_op->read_iter))
806 f->f_mode |= FMODE_CAN_READ;
807 if ((f->f_mode & FMODE_WRITE) &&
808 likely(f->f_op->write || f->f_op->write_iter))
809 f->f_mode |= FMODE_CAN_WRITE;
811 f->f_write_hint = WRITE_LIFE_NOT_SET;
812 f->f_flags &= ~(O_CREAT | O_EXCL | O_NOCTTY | O_TRUNC);
814 file_ra_state_init(&f->f_ra, f->f_mapping->host->i_mapping);
816 /* NB: we're sure to have correct a_ops only after f_op->open */
817 if (f->f_flags & O_DIRECT) {
818 if (!f->f_mapping->a_ops || !f->f_mapping->a_ops->direct_IO)
823 * XXX: Huge page cache doesn't support writing yet. Drop all page
824 * cache for this file before processing writes.
826 if ((f->f_mode & FMODE_WRITE) && filemap_nr_thps(inode->i_mapping))
827 truncate_pagecache(inode, 0);
832 if (WARN_ON_ONCE(error > 0))
835 if (f->f_mode & FMODE_WRITER) {
836 put_write_access(inode);
837 __mnt_drop_write(f->f_path.mnt);
840 path_put(&f->f_path);
841 f->f_path.mnt = NULL;
842 f->f_path.dentry = NULL;
848 * finish_open - finish opening a file
849 * @file: file pointer
850 * @dentry: pointer to dentry
851 * @open: open callback
852 * @opened: state of open
854 * This can be used to finish opening a file passed to i_op->atomic_open().
856 * If the open callback is set to NULL, then the standard f_op->open()
857 * filesystem callback is substituted.
859 * NB: the dentry reference is _not_ consumed. If, for example, the dentry is
860 * the return value of d_splice_alias(), then the caller needs to perform dput()
861 * on it after finish_open().
863 * On successful return @file is a fully instantiated open file. After this, if
864 * an error occurs in ->atomic_open(), it needs to clean up with fput().
866 * Returns zero on success or -errno if the open failed.
868 int finish_open(struct file *file, struct dentry *dentry,
869 int (*open)(struct inode *, struct file *))
871 BUG_ON(file->f_mode & FMODE_OPENED); /* once it's opened, it's opened */
873 file->f_path.dentry = dentry;
874 return do_dentry_open(file, d_backing_inode(dentry), open);
876 EXPORT_SYMBOL(finish_open);
879 * finish_no_open - finish ->atomic_open() without opening the file
881 * @file: file pointer
882 * @dentry: dentry or NULL (as returned from ->lookup())
884 * This can be used to set the result of a successful lookup in ->atomic_open().
886 * NB: unlike finish_open() this function does consume the dentry reference and
887 * the caller need not dput() it.
889 * Returns "0" which must be the return value of ->atomic_open() after having
890 * called this function.
892 int finish_no_open(struct file *file, struct dentry *dentry)
894 file->f_path.dentry = dentry;
897 EXPORT_SYMBOL(finish_no_open);
899 char *file_path(struct file *filp, char *buf, int buflen)
901 return d_path(&filp->f_path, buf, buflen);
903 EXPORT_SYMBOL(file_path);
906 * vfs_open - open the file at the given path
907 * @path: path to open
908 * @file: newly allocated file with f_flag initialized
909 * @cred: credentials to use
911 int vfs_open(const struct path *path, struct file *file)
913 file->f_path = *path;
914 return do_dentry_open(file, d_backing_inode(path->dentry), NULL);
917 struct file *dentry_open(const struct path *path, int flags,
918 const struct cred *cred)
923 validate_creds(cred);
925 /* We must always pass in a valid mount pointer. */
928 f = alloc_empty_file(flags, cred);
930 error = vfs_open(path, f);
938 EXPORT_SYMBOL(dentry_open);
940 struct file *open_with_fake_path(const struct path *path, int flags,
941 struct inode *inode, const struct cred *cred)
943 struct file *f = alloc_empty_file_noaccount(flags, cred);
948 error = do_dentry_open(f, inode, NULL);
956 EXPORT_SYMBOL(open_with_fake_path);
958 static inline int build_open_flags(int flags, umode_t mode, struct open_flags *op)
960 int lookup_flags = 0;
961 int acc_mode = ACC_MODE(flags);
964 * Clear out all open flags we don't know about so that we don't report
965 * them in fcntl(F_GETFD) or similar interfaces.
967 flags &= VALID_OPEN_FLAGS;
969 if (flags & (O_CREAT | __O_TMPFILE))
970 op->mode = (mode & S_IALLUGO) | S_IFREG;
974 /* Must never be set by userspace */
975 flags &= ~FMODE_NONOTIFY & ~O_CLOEXEC;
978 * O_SYNC is implemented as __O_SYNC|O_DSYNC. As many places only
979 * check for O_DSYNC if the need any syncing at all we enforce it's
980 * always set instead of having to deal with possibly weird behaviour
981 * for malicious applications setting only __O_SYNC.
983 if (flags & __O_SYNC)
986 if (flags & __O_TMPFILE) {
987 if ((flags & O_TMPFILE_MASK) != O_TMPFILE)
989 if (!(acc_mode & MAY_WRITE))
991 } else if (flags & O_PATH) {
993 * If we have O_PATH in the open flag. Then we
994 * cannot have anything other than the below set of flags
996 flags &= O_DIRECTORY | O_NOFOLLOW | O_PATH;
1000 op->open_flag = flags;
1002 /* O_TRUNC implies we need access checks for write permissions */
1003 if (flags & O_TRUNC)
1004 acc_mode |= MAY_WRITE;
1006 /* Allow the LSM permission hook to distinguish append
1007 access from general write access. */
1008 if (flags & O_APPEND)
1009 acc_mode |= MAY_APPEND;
1011 op->acc_mode = acc_mode;
1013 op->intent = flags & O_PATH ? 0 : LOOKUP_OPEN;
1015 if (flags & O_CREAT) {
1016 op->intent |= LOOKUP_CREATE;
1018 op->intent |= LOOKUP_EXCL;
1021 if (flags & O_DIRECTORY)
1022 lookup_flags |= LOOKUP_DIRECTORY;
1023 if (!(flags & O_NOFOLLOW))
1024 lookup_flags |= LOOKUP_FOLLOW;
1025 op->lookup_flags = lookup_flags;
1030 * file_open_name - open file and return file pointer
1032 * @name: struct filename containing path to open
1033 * @flags: open flags as per the open(2) second argument
1034 * @mode: mode for the new file if O_CREAT is set, else ignored
1036 * This is the helper to open a file from kernelspace if you really
1037 * have to. But in generally you should not do this, so please move
1038 * along, nothing to see here..
1040 struct file *file_open_name(struct filename *name, int flags, umode_t mode)
1042 struct open_flags op;
1043 int err = build_open_flags(flags, mode, &op);
1044 return err ? ERR_PTR(err) : do_filp_open(AT_FDCWD, name, &op);
1048 * filp_open - open file and return file pointer
1050 * @filename: path to open
1051 * @flags: open flags as per the open(2) second argument
1052 * @mode: mode for the new file if O_CREAT is set, else ignored
1054 * This is the helper to open a file from kernelspace if you really
1055 * have to. But in generally you should not do this, so please move
1056 * along, nothing to see here..
1058 struct file *filp_open(const char *filename, int flags, umode_t mode)
1060 struct filename *name = getname_kernel(filename);
1061 struct file *file = ERR_CAST(name);
1063 if (!IS_ERR(name)) {
1064 file = file_open_name(name, flags, mode);
1069 EXPORT_SYMBOL(filp_open);
1071 struct file *file_open_root(struct dentry *dentry, struct vfsmount *mnt,
1072 const char *filename, int flags, umode_t mode)
1074 struct open_flags op;
1075 int err = build_open_flags(flags, mode, &op);
1077 return ERR_PTR(err);
1078 return do_file_open_root(dentry, mnt, filename, &op);
1080 EXPORT_SYMBOL(file_open_root);
1082 long do_sys_open(int dfd, const char __user *filename, int flags, umode_t mode)
1084 struct open_flags op;
1085 int fd = build_open_flags(flags, mode, &op);
1086 struct filename *tmp;
1091 tmp = getname(filename);
1093 return PTR_ERR(tmp);
1095 fd = get_unused_fd_flags(flags);
1097 struct file *f = do_filp_open(dfd, tmp, &op);
1110 SYSCALL_DEFINE3(open, const char __user *, filename, int, flags, umode_t, mode)
1112 if (force_o_largefile())
1113 flags |= O_LARGEFILE;
1115 return do_sys_open(AT_FDCWD, filename, flags, mode);
1118 SYSCALL_DEFINE4(openat, int, dfd, const char __user *, filename, int, flags,
1121 if (force_o_largefile())
1122 flags |= O_LARGEFILE;
1124 return do_sys_open(dfd, filename, flags, mode);
1127 #ifdef CONFIG_COMPAT
1129 * Exactly like sys_open(), except that it doesn't set the
1132 COMPAT_SYSCALL_DEFINE3(open, const char __user *, filename, int, flags, umode_t, mode)
1134 return do_sys_open(AT_FDCWD, filename, flags, mode);
1138 * Exactly like sys_openat(), except that it doesn't set the
1141 COMPAT_SYSCALL_DEFINE4(openat, int, dfd, const char __user *, filename, int, flags, umode_t, mode)
1143 return do_sys_open(dfd, filename, flags, mode);
1150 * For backward compatibility? Maybe this should be moved
1151 * into arch/i386 instead?
1153 SYSCALL_DEFINE2(creat, const char __user *, pathname, umode_t, mode)
1155 return ksys_open(pathname, O_CREAT | O_WRONLY | O_TRUNC, mode);
1161 * "id" is the POSIX thread ID. We use the
1162 * files pointer for this..
1164 int filp_close(struct file *filp, fl_owner_t id)
1168 if (!file_count(filp)) {
1169 printk(KERN_ERR "VFS: Close: file count is 0\n");
1173 if (filp->f_op->flush)
1174 retval = filp->f_op->flush(filp, id);
1176 if (likely(!(filp->f_mode & FMODE_PATH))) {
1177 dnotify_flush(filp, id);
1178 locks_remove_posix(filp, id);
1184 EXPORT_SYMBOL(filp_close);
1187 * Careful here! We test whether the file pointer is NULL before
1188 * releasing the fd. This ensures that one clone task can't release
1189 * an fd while another clone is opening it.
1191 SYSCALL_DEFINE1(close, unsigned int, fd)
1193 int retval = __close_fd(current->files, fd);
1195 /* can't restart close syscall because file table entry was cleared */
1196 if (unlikely(retval == -ERESTARTSYS ||
1197 retval == -ERESTARTNOINTR ||
1198 retval == -ERESTARTNOHAND ||
1199 retval == -ERESTART_RESTARTBLOCK))
1206 * This routine simulates a hangup on the tty, to arrange that users
1207 * are given clean terminals at login time.
1209 SYSCALL_DEFINE0(vhangup)
1211 if (capable(CAP_SYS_TTY_CONFIG)) {
1219 * Called when an inode is about to be open.
1220 * We use this to disallow opening large files on 32bit systems if
1221 * the caller didn't specify O_LARGEFILE. On 64bit systems we force
1222 * on this flag in sys_open.
1224 int generic_file_open(struct inode * inode, struct file * filp)
1226 if (!(filp->f_flags & O_LARGEFILE) && i_size_read(inode) > MAX_NON_LFS)
1231 EXPORT_SYMBOL(generic_file_open);
1234 * This is used by subsystems that don't want seekable
1235 * file descriptors. The function is not supposed to ever fail, the only
1236 * reason it returns an 'int' and not 'void' is so that it can be plugged
1237 * directly into file_operations structure.
1239 int nonseekable_open(struct inode *inode, struct file *filp)
1241 filp->f_mode &= ~(FMODE_LSEEK | FMODE_PREAD | FMODE_PWRITE);
1245 EXPORT_SYMBOL(nonseekable_open);
1248 * stream_open is used by subsystems that want stream-like file descriptors.
1249 * Such file descriptors are not seekable and don't have notion of position
1250 * (file.f_pos is always 0 and ppos passed to .read()/.write() is always NULL).
1251 * Contrary to file descriptors of other regular files, .read() and .write()
1252 * can run simultaneously.
1254 * stream_open never fails and is marked to return int so that it could be
1255 * directly used as file_operations.open .
1257 int stream_open(struct inode *inode, struct file *filp)
1259 filp->f_mode &= ~(FMODE_LSEEK | FMODE_PREAD | FMODE_PWRITE | FMODE_ATOMIC_POS);
1260 filp->f_mode |= FMODE_STREAM;
1264 EXPORT_SYMBOL(stream_open);