1 // SPDX-License-Identifier: GPL-2.0-only
25 	struct ovl_fs *ofs = OVL_FS(dentry->d_sb);  in ovl_setattr()
34 	if (attr->ia_valid & ATTR_SIZE) {  in ovl_setattr()
48 		if (attr->ia_valid & ATTR_SIZE) {  in ovl_setattr()
55 		if (attr->ia_valid & (ATTR_KILL_SUID|ATTR_KILL_SGID))  in ovl_setattr()
56 			attr->ia_valid &= ~ATTR_MODE;  in ovl_setattr()
61 		 * filesystem rely on attr->ia_file  in ovl_setattr()
63 		attr->ia_valid &= ~ATTR_FILE;  in ovl_setattr()
66 		 * If open(O_TRUNC) is done, VFS calls ->setattr with ATTR_OPEN  in ovl_setattr()
68 		 * filesystem during open -> do not pass ATTR_OPEN.  This  in ovl_setattr()
74 		attr->ia_valid &= ~ATTR_OPEN;  in ovl_setattr()
80 		inode_lock(upperdentry->d_inode);  in ovl_setattr()
81 		old_cred = ovl_override_creds(dentry->d_sb);  in ovl_setattr()
85 			ovl_copyattr(dentry->d_inode);  in ovl_setattr()
86 		inode_unlock(upperdentry->d_inode);  in ovl_setattr()
99 	struct ovl_fs *ofs = OVL_FS(dentry->d_sb);  in ovl_map_dev_ino()
102 	unsigned int xinoshift = 64 - xinobits;  in ovl_map_dev_ino()
108 		 * which is friendly to du -x.  in ovl_map_dev_ino()
110 		stat->dev = dentry->d_sb->s_dev;  in ovl_map_dev_ino()
118 		 * the non-persistent inode numbers range in case of overflow.  in ovl_map_dev_ino()
122 		if (likely(!(stat->ino >> xinoshift))) {  in ovl_map_dev_ino()
123 			stat->ino |= ((u64)fsid) << (xinoshift + 1);  in ovl_map_dev_ino()
124 			stat->dev = dentry->d_sb->s_dev;  in ovl_map_dev_ino()
128 					    dentry, stat->ino, xinobits);  in ovl_map_dev_ino()
133 	if (S_ISDIR(dentry->d_inode->i_mode)) {  in ovl_map_dev_ino()
136 		 * -xdev' will scan the entire overlay mount and won't cross the  in ovl_map_dev_ino()
143 		stat->dev = dentry->d_sb->s_dev;  in ovl_map_dev_ino()
144 		stat->ino = dentry->d_inode->i_ino;  in ovl_map_dev_ino()
147 		 * For non-samefs setup, if we cannot map all layers st_ino  in ovl_map_dev_ino()
152 		stat->dev = ofs->fs[fsid].pseudo_dev;  in ovl_map_dev_ino()
159 	struct dentry *dentry = path->dentry;  in ovl_getattr()
164 	bool is_dir = S_ISDIR(inode->i_mode);  in ovl_getattr()
172 	old_cred = ovl_override_creds(dentry->d_sb);  in ovl_getattr()
177 	/* Report the effective immutable/append-only STATX flags */  in ovl_getattr()
181 	 * For non-dir or same fs, we use st_ino of the copy up origin.  in ovl_getattr()
183 	 * With xino feature and non-samefs, we use st_ino of the copy up  in ovl_getattr()
186 	 * If lower filesystem supports NFS file handles, this also guaranties  in ovl_getattr()
189 	if (!is_dir || ovl_same_dev(OVL_FS(dentry->d_sb))) {  in ovl_getattr()
191 			fsid = ovl_layer_lower(dentry)->fsid;  in ovl_getattr()
204 			 * Lower hardlinks may be broken on copy up to different  in ovl_getattr()
205 			 * upper files, so we cannot use the lower origin st_ino  in ovl_getattr()
206 			 * for those different files, even for the same fs case.  in ovl_getattr()
209 			 * same dir on a lower layer. With the "verify_lower"  in ovl_getattr()
210 			 * feature, we do not use the lower origin st_ino, if  in ovl_getattr()
219 			    (!ovl_verify_lower(dentry->d_sb) &&  in ovl_getattr()
221 				fsid = ovl_layer_lower(dentry)->fsid;  in ovl_getattr()
222 				stat->ino = lowerstat.ino;  in ovl_getattr()
226 			 * If we are querying a metacopy dentry and lower  in ovl_getattr()
229 			 * vfs_getattr(). If lower itself is metacopy, then  in ovl_getattr()
234 				stat->blocks = lowerstat.blocks;  in ovl_getattr()
241 			 * If lower is not same as lowerdata or if there was  in ovl_getattr()
257 					round_up(stat->size, stat->blksize) >> 9;  in ovl_getattr()
259 			stat->blocks = lowerdatastat.blocks;  in ovl_getattr()
271 		stat->nlink = 1;  in ovl_getattr()
276 	 * and non-covered lower hardlinks. It does not include the upper  in ovl_getattr()
280 		stat->nlink = dentry->d_inode->i_nlink;  in ovl_getattr()
301 		return -ECHILD;  in ovl_permission()
312 	old_cred = ovl_override_creds(inode->i_sb);  in ovl_permission()
314 	    !special_file(realinode->i_mode) && mask & MAY_WRITE) {  in ovl_permission()
333 		return ERR_PTR(-ECHILD);  in ovl_get_link()
335 	old_cred = ovl_override_creds(dentry->d_sb);  in ovl_get_link()
344  * of the POSIX ACLs retrieved from the lower layer to this function to not
353 	for (unsigned int i = 0; i < acl->a_count; i++) {  in ovl_idmap_posix_acl()
357 		struct posix_acl_entry *e = &acl->a_entries[i];  in ovl_idmap_posix_acl()
358 		switch (e->e_tag) {  in ovl_idmap_posix_acl()
359 		case ACL_USER:  in ovl_idmap_posix_acl()
360 			vfsuid = make_vfsuid(idmap, fs_userns, e->e_uid);  in ovl_idmap_posix_acl()
361 			e->e_uid = vfsuid_into_kuid(vfsuid);  in ovl_idmap_posix_acl()
363 		case ACL_GROUP:  in ovl_idmap_posix_acl()
364 			vfsgid = make_vfsgid(idmap, fs_userns, e->e_gid);  in ovl_idmap_posix_acl()
365 			e->e_gid = vfsgid_into_kgid(vfsgid);  in ovl_idmap_posix_acl()
400 	struct inode *realinode = d_inode(path->dentry);  in ovl_get_acl_path()
402 	idmap = mnt_idmap(path->mnt);  in ovl_get_acl_path()
407 		real_acl = vfs_get_acl(idmap, path->dentry, acl_name);  in ovl_get_acl_path()
411 	if (!is_idmapped_mnt(path->mnt))  in ovl_get_acl_path()
416         * would alter the cached values filesystem wide for the lower  in ovl_get_acl_path()
423 		return ERR_PTR(-ENOMEM);  in ovl_get_acl_path()
435  * alter the cached values filesystem wide for the lower filesystem. Instead we
452 		return ERR_PTR(-ECHILD);  in do_ovl_get_acl()
464 			return ERR_PTR(-ECHILD);  in do_ovl_get_acl()
470 		old_cred = ovl_override_creds(inode->i_sb);  in do_ovl_get_acl()
485 	struct ovl_fs *ofs = OVL_FS(dentry->d_sb);  in ovl_set_or_remove_acl()
490 	 * If ACL is to be removed from a lower file, check if it exists in  in ovl_set_or_remove_acl()
498 		old_cred = ovl_override_creds(dentry->d_sb);  in ovl_set_or_remove_acl()
521 	old_cred = ovl_override_creds(dentry->d_sb);  in ovl_set_or_remove_acl()
544 		return -EOPNOTSUPP;  in ovl_set_acl()
545 	if (!realinode->i_op->set_acl)  in ovl_set_acl()
546 		return -EOPNOTSUPP;  in ovl_set_acl()
547 	if (type == ACL_TYPE_DEFAULT && !S_ISDIR(inode->i_mode))  in ovl_set_acl()
548 		return acl ? -EACCES : 0;  in ovl_set_acl()
550 		return -EPERM;  in ovl_set_acl()
556 	if (unlikely(inode->i_mode & S_ISGID) && type == ACL_TYPE_ACCESS &&  in ovl_set_acl()
557 	    !in_group_p(inode->i_gid) &&  in ovl_set_acl()
573 		struct ovl_fs *ofs = OVL_FS(inode->i_sb);  in ovl_update_time()
596 		return -EIO;  in ovl_fiemap()
598 	if (!realinode->i_op->fiemap)  in ovl_fiemap()
599 		return -EOPNOTSUPP;  in ovl_fiemap()
601 	old_cred = ovl_override_creds(inode->i_sb);  in ovl_fiemap()
602 	err = realinode->i_op->fiemap(realinode, fieinfo, start, len);  in ovl_fiemap()
630 		cmd = fa->fsx_valid ? FS_IOC_FSSETXATTR : FS_IOC_SETFLAGS;  in ovl_security_fileattr()
632 		cmd = fa->fsx_valid ? FS_IOC_FSGETXATTR : FS_IOC_GETFLAGS;  in ovl_security_fileattr()
648 	return vfs_fileattr_set(mnt_idmap(realpath->mnt), realpath->dentry, fa);  in ovl_real_fileattr_set()
668 		old_cred = ovl_override_creds(inode->i_sb);  in ovl_fileattr_set()
670 		 * Store immutable/append-only flags in xattr and clear them  in ovl_fileattr_set()
671 		 * in upper fileattr (in case they were set by older kernel)  in ovl_fileattr_set()
672 		 * so children of "ovl-immutable" directories lower aliases of  in ovl_fileattr_set()
673 		 * "ovl-immutable" hardlinks could be copied up.  in ovl_fileattr_set()
686 		flags = ovl_inode_real(inode)->i_flags & OVL_COPY_I_FLAGS_MASK;  in ovl_fileattr_set()
689 		flags |= inode->i_flags & OVL_PROT_I_FLAGS_MASK;  in ovl_fileattr_set()
705 	if (inode->i_flags & S_APPEND) {  in ovl_fileattr_prot_flags()
706 		fa->flags |= FS_APPEND_FL;  in ovl_fileattr_prot_flags()
707 		fa->fsx_xflags |= FS_XFLAG_APPEND;  in ovl_fileattr_prot_flags()
709 	if (inode->i_flags & S_IMMUTABLE) {  in ovl_fileattr_prot_flags()
710 		fa->flags |= FS_IMMUTABLE_FL;  in ovl_fileattr_prot_flags()
711 		fa->fsx_xflags |= FS_XFLAG_IMMUTABLE;  in ovl_fileattr_prot_flags()
723 	err = vfs_fileattr_get(realpath->dentry, fa);  in ovl_real_fileattr_get()
724 	if (err == -ENOIOCTLCMD)  in ovl_real_fileattr_get()
725 		err = -ENOTTY;  in ovl_real_fileattr_get()
738 	old_cred = ovl_override_creds(inode->i_sb);  in ovl_fileattr_get()
780 	/* For O_DIRECT dentry_open() checks f_mapping->a_ops->direct_IO */
786  * overlayfs instance as lower layer. We need to annotate the
797  * [...] &type->i_mutex_dir_key        (stack_depth=0)
802  * - inode->i_rwsem			(inode_lock[2])
803  * - upper_mnt->mnt_sb->s_writers	(ovl_want_write[0])
804  * - OVL_I(inode)->lock			(ovl_inode_lock[2])
805  * - OVL_I(lowerinode)->lock		(ovl_inode_lock[1])
808  * - inode->i_rwsem			(inode_lock[2])
809  * - OVL_I(inode)->lock			(ovl_inode_lock[2])
810  * - lowerinode->i_rwsem		(inode_lock[1])
811  * - OVL_I(lowerinode)->lock		(ovl_inode_lock[1])
813  * But lowerinode->i_rwsem SHOULD NOT be acquired while ovl_want_write() is
814  * held, because it is in reverse order of the non-nested case using the same
816  * - inode->i_rwsem			(inode_lock[1])
817  * - upper_mnt->mnt_sb->s_writers	(ovl_want_write[0])
818  * - OVL_I(inode)->lock			(ovl_inode_lock[1])
829 	int depth = inode->i_sb->s_stack_depth - 1;  in ovl_lockdep_annotate_inode_mutex_key()
834 	if (S_ISDIR(inode->i_mode))  in ovl_lockdep_annotate_inode_mutex_key()
835 		lockdep_set_class(&inode->i_rwsem, &ovl_i_mutex_dir_key[depth]);  in ovl_lockdep_annotate_inode_mutex_key()
837 		lockdep_set_class(&inode->i_rwsem, &ovl_i_mutex_key[depth]);  in ovl_lockdep_annotate_inode_mutex_key()
839 	lockdep_set_class(&OVL_I(inode)->lock, &ovl_i_lock_key[depth]);  in ovl_lockdep_annotate_inode_mutex_key()
845 	struct ovl_fs *ofs = OVL_FS(inode->i_sb);  in ovl_next_ino()
847 	inode->i_ino = atomic_long_inc_return(&ofs->last_ino);  in ovl_next_ino()
848 	if (unlikely(!inode->i_ino))  in ovl_next_ino()
849 		inode->i_ino = atomic_long_inc_return(&ofs->last_ino);  in ovl_next_ino()
854 	struct ovl_fs *ofs = OVL_FS(inode->i_sb);  in ovl_map_ino()
856 	unsigned int xinoshift = 64 - xinobits;  in ovl_map_ino()
865 	inode->i_ino = ino;  in ovl_map_ino()
869 		inode->i_ino |= (unsigned long)fsid << (xinoshift + 1);  in ovl_map_ino()
874 	 * For directory inodes on non-samefs with xino disabled or xino  in ovl_map_ino()
875 	 * overflow, we allocate a non-persistent inode number, to be used for  in ovl_map_ino()
882 	if (S_ISDIR(inode->i_mode)) {  in ovl_map_ino()
885 			inode->i_ino &= ~0UL >> xinobits;  in ovl_map_ino()
886 			inode->i_ino |= 1UL << xinoshift;  in ovl_map_ino()
897 	oi->__upperdentry = oip->upperdentry;  in ovl_inode_init()
898 	oi->oe = oip->oe;  in ovl_inode_init()
899 	oi->redirect = oip->redirect;  in ovl_inode_init()
900 	oi->lowerdata_redirect = oip->lowerdata_redirect;  in ovl_inode_init()
910 	inode->i_mode = mode;  in ovl_fill_inode()
911 	inode->i_flags |= S_NOCMTIME;  in ovl_fill_inode()
913 	inode->i_acl = inode->i_default_acl = ACL_DONT_CACHE;  in ovl_fill_inode()
919 	case S_IFREG:  in ovl_fill_inode()
920 		inode->i_op = &ovl_file_inode_operations;  in ovl_fill_inode()
921 		inode->i_fop = &ovl_file_operations;  in ovl_fill_inode()
922 		inode->i_mapping->a_ops = &ovl_aops;  in ovl_fill_inode()
925 	case S_IFDIR:  in ovl_fill_inode()
926 		inode->i_op = &ovl_dir_inode_operations;  in ovl_fill_inode()
927 		inode->i_fop = &ovl_dir_operations;  in ovl_fill_inode()
930 	case S_IFLNK:  in ovl_fill_inode()
931 		inode->i_op = &ovl_symlink_inode_operations;  in ovl_fill_inode()
935 		inode->i_op = &ovl_special_inode_operations;  in ovl_fill_inode()
943  * hardlinks and non-covered lower hardlinks. During the lifetime of a non-pure
946  * 1. Lower hardlink copy up
948  * 3. Lower hardlink whiteout or renamed over
950  * For the first, copy up case, the union nlink does not change, whether the
953  * lower inode nlink in the index inode xattr .overlay.nlink.
955  * For the second, upper hardlink case, the union nlink should be incremented
960  * For the last, lower cover up case, we simplify things by preceding the
968  * On-disk format for indexed nlink:
970  * nlink relative to the upper inode - "U[+-]NUM"
971  * nlink relative to the lower inode - "L[+-]NUM"
983 		       (int) (inode->i_nlink - realinode->i_nlink));  in ovl_set_nlink_common()
986 		return -EIO;  in ovl_set_nlink_common()
988 	return ovl_setxattr(OVL_FS(inode->i_sb), ovl_dentry_upper(dentry),  in ovl_set_nlink_common()
1011 	if (!lowerdentry || !upperdentry || d_inode(lowerdentry)->i_nlink == 1)  in ovl_get_nlink()
1015 				 &buf, sizeof(buf) - 1);  in ovl_get_nlink()
1021 	    (buf[1] != '+' && buf[1] != '-'))  in ovl_get_nlink()
1028 	nlink = d_inode(buf[0] == 'L' ? lowerdentry : upperdentry)->i_nlink;  in ovl_get_nlink()
1055 	return inode->i_private == data;  in ovl_inode_test()
1060 	inode->i_private = data;  in ovl_inode_set()
1069 	 * checks, so NULL lower/upper in dentry must match NULL lower/upper in  in ovl_verify_inode()
1071 	 * 'unknown' lower/upper.  in ovl_verify_inode()
1073 	if (S_ISDIR(inode->i_mode) && strict) {  in ovl_verify_inode()
1074 		/* Real lower dir moved to upper layer under us? */  in ovl_verify_inode()
1084 	 * Allow non-NULL lower inode in ovl_inode even if lowerdentry is NULL.  in ovl_verify_inode()
1086 	 * or hardlinked overlay dentry and lower dentry cannot be followed  in ovl_verify_inode()
1087 	 * by origin because lower fs does not support file handles.  in ovl_verify_inode()
1093 	 * Allow non-NULL __upperdentry in inode even if upperdentry is NULL.  in ovl_verify_inode()
1094 	 * This happens when finding a lower alias for a copied up hard link.  in ovl_verify_inode()
1114 		return ERR_PTR(-ESTALE);  in ovl_lookup_inode()
1148 		return ERR_PTR(-ENOTDIR);  in ovl_get_trap_inode()
1153 		return ERR_PTR(-ENOMEM);  in ovl_get_trap_inode()
1155 	if (!(trap->i_state & I_NEW)) {  in ovl_get_trap_inode()
1158 		return ERR_PTR(-ELOOP);  in ovl_get_trap_inode()
1161 	trap->i_mode = S_IFDIR;  in ovl_get_trap_inode()
1162 	trap->i_flags = S_DEAD;  in ovl_get_trap_inode()
1169  * Does overlay inode need to be hashed by lower inode?
1172 			     struct dentry *lower, bool index)  in ovl_hash_bylower()  argument
1177 	if (!lower)  in ovl_hash_bylower()
1188 	/* No, if lower hardlink is or will be broken on copy up */  in ovl_hash_bylower()
1190 	    !d_is_dir(lower) && d_inode(lower)->i_nlink > 1)  in ovl_hash_bylower()
1193 	/* No, if non-indexed upper with NFS export */  in ovl_hash_bylower()
1194 	if (ofs->config.nfs_export && upper)  in ovl_hash_bylower()
1197 	/* Otherwise, hash by lower inode for fsnotify */  in ovl_hash_bylower()
1214 	struct dentry *upperdentry = oip->upperdentry;  in ovl_get_inode()
1215 	struct ovl_path *lowerpath = ovl_lowerpath(oip->oe);  in ovl_get_inode()
1218 	struct dentry *lowerdentry = lowerpath ? lowerpath->dentry : NULL;  in ovl_get_inode()
1221 		.mnt = upperdentry ? ovl_upper_mnt(ofs) : lowerpath->layer->mnt,  in ovl_get_inode()
1224 					oip->index);  in ovl_get_inode()
1225 	int fsid = bylower ? lowerpath->layer->fsid : 0;  in ovl_get_inode()
1228 	int err = oip->newinode ? -EEXIST : -ENOMEM;  in ovl_get_inode()
1234 	 * Copy up origin (lower) may exist for non-indexed upper, but we must  in ovl_get_inode()
1235 	 * not use lower as hash key if this is a broken hardlink.  in ovl_get_inode()
1237 	is_dir = S_ISDIR(realinode->i_mode);  in ovl_get_inode()
1241 		unsigned int nlink = is_dir ? 1 : realinode->i_nlink;  in ovl_get_inode()
1243 		inode = ovl_iget5(sb, oip->newinode, key);  in ovl_get_inode()
1246 		if (!(inode->i_state & I_NEW)) {  in ovl_get_inode()
1254 				err = -ESTALE;  in ovl_get_inode()
1259 			ovl_free_entry(oip->oe);  in ovl_get_inode()
1260 			kfree(oip->redirect);  in ovl_get_inode()
1261 			kfree(oip->lowerdata_redirect);  in ovl_get_inode()
1265 		/* Recalculate nlink for non-dir due to indexing */  in ovl_get_inode()
1270 		ino = key->i_ino;  in ovl_get_inode()
1272 		/* Lower hardlink that will be broken on copy up */  in ovl_get_inode()
1275 			err = -ENOMEM;  in ovl_get_inode()
1278 		ino = realinode->i_ino;  in ovl_get_inode()
1279 		fsid = lowerpath->layer->fsid;  in ovl_get_inode()
1281 	ovl_fill_inode(inode, realinode->i_mode, realinode->i_rdev);  in ovl_get_inode()
1287 	if (oip->index)  in ovl_get_inode()
1293 	/* Check for non-merge dir that may have whiteouts */  in ovl_get_inode()
1295 		if (((upperdentry && lowerdentry) || ovl_numlower(oip->oe) > 1) ||  in ovl_get_inode()
1301 	/* Check for immutable/append-only inode flags in xattr */  in ovl_get_inode()
1305 	if (inode->i_state & I_NEW)  in ovl_get_inode()