4 * The contents of this file are subject to the terms of the
5 * Common Development and Distribution License (the "License").
6 * You may not use this file except in compliance with the License.
8 * You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE
9 * or http://www.opensolaris.org/os/licensing.
10 * See the License for the specific language governing permissions
11 * and limitations under the License.
13 * When distributing Covered Code, include this CDDL HEADER in each
14 * file and include the License file at usr/src/OPENSOLARIS.LICENSE.
15 * If applicable, add the following below this CDDL HEADER, with the
16 * fields enclosed by brackets "[]" replaced with your own identifying
17 * information: Portions Copyright [yyyy] [name of copyright owner]
22 * Copyright (c) 2011, Lawrence Livermore National Security, LLC.
24 * Extended attributes (xattr) on Solaris are implemented as files
25 * which exist in a hidden xattr directory. These extended attributes
26 * can be accessed using the attropen() system call which opens
27 * the extended attribute. It can then be manipulated just like
28 * a standard file descriptor. This has a couple advantages such
29 * as practically no size limit on the file, and the extended
30 * attributes permissions may differ from those of the parent file.
31 * This interface is really quite clever, but it's also completely
32 * different than what is supported on Linux. It also comes with a
33 * steep performance penalty when accessing small xattrs because they
34 * are not stored with the parent file.
36 * Under Linux extended attributes are manipulated by the system
37 * calls getxattr(2), setxattr(2), and listxattr(2). They consider
38 * extended attributes to be name/value pairs where the name is a
39 * NULL terminated string. The name must also include one of the
40 * following namespace prefixes:
42 * user - No restrictions and is available to user applications.
43 * trusted - Restricted to kernel and root (CAP_SYS_ADMIN) use.
44 * system - Used for access control lists (system.nfs4_acl, etc).
45 * security - Used by SELinux to store a files security context.
47 * The value under Linux to limited to 65536 bytes of binary data.
48 * In practice, individual xattrs tend to be much smaller than this
49 * and are typically less than 100 bytes. A good example of this
50 * are the security.selinux xattrs which are less than 100 bytes and
51 * exist for every file when xattr labeling is enabled.
53 * The Linux xattr implementation has been written to take advantage of
54 * this typical usage. When the dataset property 'xattr=sa' is set,
55 * then xattrs will be preferentially stored as System Attributes (SA).
56 * This allows tiny xattrs (~100 bytes) to be stored with the dnode and
57 * up to 64k of xattrs to be stored in the spill block. If additional
58 * xattr space is required, which is unlikely under Linux, they will
59 * be stored using the traditional directory approach.
61 * This optimization results in roughly a 3x performance improvement
62 * when accessing xattrs because it avoids the need to perform a seek
63 * for every xattr value. When multiple xattrs are stored per-file
64 * the performance improvements are even greater because all of the
65 * xattrs stored in the spill block will be cached.
67 * However, by default SA based xattrs are disabled in the Linux port
68 * to maximize compatibility with other implementations. If you do
69 * enable SA based xattrs then they will not be visible on platforms
70 * which do not support this feature.
72 * NOTE: One additional consequence of the xattr directory implementation
73 * is that when an extended attribute is manipulated an inode is created.
74 * This inode will exist in the Linux inode cache but there will be no
75 * associated entry in the dentry cache which references it. This is
76 * safe but it may result in some confusion. Enabling SA based xattrs
77 * largely avoids the issue except in the overflow case.
80 #include <sys/zfs_vfsops.h>
81 #include <sys/zfs_vnops.h>
82 #include <sys/zfs_znode.h>
87 typedef struct xattr_filldir {
91 struct dentry *dentry;
94 static const struct xattr_handler *zpl_xattr_handler(const char *);
97 zpl_xattr_permission(xattr_filldir_t *xf, const char *name, int name_len)
99 static const struct xattr_handler *handler;
100 struct dentry *d = xf->dentry;
102 handler = zpl_xattr_handler(name);
107 #if defined(HAVE_XATTR_LIST_SIMPLE)
108 if (!handler->list(d))
110 #elif defined(HAVE_XATTR_LIST_DENTRY)
111 if (!handler->list(d, NULL, 0, name, name_len, 0))
113 #elif defined(HAVE_XATTR_LIST_HANDLER)
114 if (!handler->list(handler, d, NULL, 0, name, name_len))
116 #elif defined(HAVE_XATTR_LIST_INODE)
117 if (!handler->list(d->d_inode, NULL, 0, name, name_len))
126 * Determine is a given xattr name should be visible and if so copy it
127 * in to the provided buffer (xf->buf).
130 zpl_xattr_filldir(xattr_filldir_t *xf, const char *name, int name_len)
132 /* Check permissions using the per-namespace list xattr handler. */
133 if (!zpl_xattr_permission(xf, name, name_len))
136 /* When xf->buf is NULL only calculate the required size. */
138 if (xf->offset + name_len + 1 > xf->size)
141 memcpy(xf->buf + xf->offset, name, name_len);
142 xf->buf[xf->offset + name_len] = '\0';
145 xf->offset += (name_len + 1);
151 * Read as many directory entry names as will fit in to the provided buffer,
152 * or when no buffer is provided calculate the required buffer size.
155 zpl_xattr_readdir(struct inode *dxip, xattr_filldir_t *xf)
161 zap_cursor_init(&zc, ITOZSB(dxip)->z_os, ITOZ(dxip)->z_id);
163 while ((error = -zap_cursor_retrieve(&zc, &zap)) == 0) {
165 if (zap.za_integer_length != 8 || zap.za_num_integers != 1) {
170 error = zpl_xattr_filldir(xf, zap.za_name, strlen(zap.za_name));
174 zap_cursor_advance(&zc);
177 zap_cursor_fini(&zc);
179 if (error == -ENOENT)
186 zpl_xattr_list_dir(xattr_filldir_t *xf, cred_t *cr)
188 struct inode *ip = xf->dentry->d_inode;
189 struct inode *dxip = NULL;
192 /* Lookup the xattr directory */
193 error = -zfs_lookup(ip, NULL, &dxip, LOOKUP_XATTR, cr, NULL, NULL);
195 if (error == -ENOENT)
201 error = zpl_xattr_readdir(dxip, xf);
208 zpl_xattr_list_sa(xattr_filldir_t *xf)
210 znode_t *zp = ITOZ(xf->dentry->d_inode);
211 nvpair_t *nvp = NULL;
214 mutex_enter(&zp->z_lock);
215 if (zp->z_xattr_cached == NULL)
216 error = -zfs_sa_get_xattr(zp);
217 mutex_exit(&zp->z_lock);
222 ASSERT(zp->z_xattr_cached);
224 while ((nvp = nvlist_next_nvpair(zp->z_xattr_cached, nvp)) != NULL) {
225 ASSERT3U(nvpair_type(nvp), ==, DATA_TYPE_BYTE_ARRAY);
227 error = zpl_xattr_filldir(xf, nvpair_name(nvp),
228 strlen(nvpair_name(nvp)));
237 zpl_xattr_list(struct dentry *dentry, char *buffer, size_t buffer_size)
239 znode_t *zp = ITOZ(dentry->d_inode);
240 zfsvfs_t *zfsvfs = ZTOZSB(zp);
241 xattr_filldir_t xf = { buffer_size, 0, buffer, dentry };
243 fstrans_cookie_t cookie;
247 cookie = spl_fstrans_mark();
248 rrm_enter_read(&(zfsvfs)->z_teardown_lock, FTAG);
249 rw_enter(&zp->z_xattr_lock, RW_READER);
251 if (zfsvfs->z_use_sa && zp->z_is_sa) {
252 error = zpl_xattr_list_sa(&xf);
257 error = zpl_xattr_list_dir(&xf, cr);
264 rw_exit(&zp->z_xattr_lock);
265 rrm_exit(&(zfsvfs)->z_teardown_lock, FTAG);
266 spl_fstrans_unmark(cookie);
273 zpl_xattr_get_dir(struct inode *ip, const char *name, void *value,
274 size_t size, cred_t *cr)
276 struct inode *dxip = NULL;
277 struct inode *xip = NULL;
281 /* Lookup the xattr directory */
282 error = -zfs_lookup(ip, NULL, &dxip, LOOKUP_XATTR, cr, NULL, NULL);
286 /* Lookup a specific xattr name in the directory */
287 error = -zfs_lookup(dxip, (char *)name, &xip, 0, cr, NULL, NULL);
292 error = i_size_read(xip);
296 if (size < i_size_read(xip)) {
301 error = zpl_read_common(xip, value, size, &pos, UIO_SYSSPACE, 0, cr);
313 zpl_xattr_get_sa(struct inode *ip, const char *name, void *value, size_t size)
315 znode_t *zp = ITOZ(ip);
320 ASSERT(RW_LOCK_HELD(&zp->z_xattr_lock));
322 mutex_enter(&zp->z_lock);
323 if (zp->z_xattr_cached == NULL)
324 error = -zfs_sa_get_xattr(zp);
325 mutex_exit(&zp->z_lock);
330 ASSERT(zp->z_xattr_cached);
331 error = -nvlist_lookup_byte_array(zp->z_xattr_cached, name,
332 &nv_value, &nv_size);
342 memcpy(value, nv_value, nv_size);
348 __zpl_xattr_get(struct inode *ip, const char *name, void *value, size_t size,
351 znode_t *zp = ITOZ(ip);
352 zfsvfs_t *zfsvfs = ZTOZSB(zp);
355 ASSERT(RW_LOCK_HELD(&zp->z_xattr_lock));
357 if (zfsvfs->z_use_sa && zp->z_is_sa) {
358 error = zpl_xattr_get_sa(ip, name, value, size);
359 if (error != -ENOENT)
363 error = zpl_xattr_get_dir(ip, name, value, size, cr);
365 if (error == -ENOENT)
371 #define XATTR_NOENT 0x0
372 #define XATTR_IN_SA 0x1
373 #define XATTR_IN_DIR 0x2
374 /* check where the xattr resides */
376 __zpl_xattr_where(struct inode *ip, const char *name, int *where, cred_t *cr)
378 znode_t *zp = ITOZ(ip);
379 zfsvfs_t *zfsvfs = ZTOZSB(zp);
383 ASSERT(RW_LOCK_HELD(&zp->z_xattr_lock));
385 *where = XATTR_NOENT;
386 if (zfsvfs->z_use_sa && zp->z_is_sa) {
387 error = zpl_xattr_get_sa(ip, name, NULL, 0);
389 *where |= XATTR_IN_SA;
390 else if (error != -ENOENT)
394 error = zpl_xattr_get_dir(ip, name, NULL, 0, cr);
396 *where |= XATTR_IN_DIR;
397 else if (error != -ENOENT)
400 if (*where == (XATTR_IN_SA|XATTR_IN_DIR))
401 cmn_err(CE_WARN, "ZFS: inode %p has xattr \"%s\""
402 " in both SA and dir", ip, name);
403 if (*where == XATTR_NOENT)
411 zpl_xattr_get(struct inode *ip, const char *name, void *value, size_t size)
413 znode_t *zp = ITOZ(ip);
414 zfsvfs_t *zfsvfs = ZTOZSB(zp);
416 fstrans_cookie_t cookie;
420 cookie = spl_fstrans_mark();
421 rrm_enter_read(&(zfsvfs)->z_teardown_lock, FTAG);
422 rw_enter(&zp->z_xattr_lock, RW_READER);
423 error = __zpl_xattr_get(ip, name, value, size, cr);
424 rw_exit(&zp->z_xattr_lock);
425 rrm_exit(&(zfsvfs)->z_teardown_lock, FTAG);
426 spl_fstrans_unmark(cookie);
433 zpl_xattr_set_dir(struct inode *ip, const char *name, const void *value,
434 size_t size, int flags, cred_t *cr)
436 struct inode *dxip = NULL;
437 struct inode *xip = NULL;
440 int lookup_flags, error;
441 const int xattr_mode = S_IFREG | 0644;
445 * Lookup the xattr directory. When we're adding an entry pass
446 * CREATE_XATTR_DIR to ensure the xattr directory is created.
447 * When removing an entry this flag is not passed to avoid
448 * unnecessarily creating a new xattr directory.
450 lookup_flags = LOOKUP_XATTR;
452 lookup_flags |= CREATE_XATTR_DIR;
454 error = -zfs_lookup(ip, NULL, &dxip, lookup_flags, cr, NULL, NULL);
458 /* Lookup a specific xattr name in the directory */
459 error = -zfs_lookup(dxip, (char *)name, &xip, 0, cr, NULL, NULL);
460 if (error && (error != -ENOENT))
465 /* Remove a specific name xattr when value is set to NULL. */
468 error = -zfs_remove(dxip, (char *)name, cr, 0);
473 /* Lookup failed create a new xattr. */
475 vap = kmem_zalloc(sizeof (vattr_t), KM_SLEEP);
476 vap->va_mode = xattr_mode;
477 vap->va_mask = ATTR_MODE;
478 vap->va_uid = crgetfsuid(cr);
479 vap->va_gid = crgetfsgid(cr);
481 error = -zfs_create(dxip, (char *)name, vap, 0, 0644, &xip,
489 error = -zfs_freesp(ITOZ(xip), 0, 0, xattr_mode, TRUE);
493 wrote = zpl_write_common(xip, value, size, &pos, UIO_SYSSPACE, 0, cr);
499 kmem_free(vap, sizeof (vattr_t));
507 if (error == -ENOENT)
510 ASSERT3S(error, <=, 0);
516 zpl_xattr_set_sa(struct inode *ip, const char *name, const void *value,
517 size_t size, int flags, cred_t *cr)
519 znode_t *zp = ITOZ(ip);
524 mutex_enter(&zp->z_lock);
525 if (zp->z_xattr_cached == NULL)
526 error = -zfs_sa_get_xattr(zp);
527 mutex_exit(&zp->z_lock);
532 ASSERT(zp->z_xattr_cached);
533 nvl = zp->z_xattr_cached;
536 error = -nvlist_remove(nvl, name, DATA_TYPE_BYTE_ARRAY);
537 if (error == -ENOENT)
538 error = zpl_xattr_set_dir(ip, name, NULL, 0, flags, cr);
540 /* Limited to 32k to keep nvpair memory allocations small */
541 if (size > DXATTR_MAX_ENTRY_SIZE)
544 /* Prevent the DXATTR SA from consuming the entire SA region */
545 error = -nvlist_size(nvl, &sa_size, NV_ENCODE_XDR);
549 if (sa_size > DXATTR_MAX_SA_SIZE)
552 error = -nvlist_add_byte_array(nvl, name,
553 (uchar_t *)value, size);
557 * Update the SA for additions, modifications, and removals. On
558 * error drop the inconsistent cached version of the nvlist, it
559 * will be reconstructed from the ARC when next accessed.
562 error = -zfs_sa_set_xattr(zp);
566 zp->z_xattr_cached = NULL;
569 ASSERT3S(error, <=, 0);
575 zpl_xattr_set(struct inode *ip, const char *name, const void *value,
576 size_t size, int flags)
578 znode_t *zp = ITOZ(ip);
579 zfsvfs_t *zfsvfs = ZTOZSB(zp);
581 fstrans_cookie_t cookie;
586 cookie = spl_fstrans_mark();
587 rrm_enter_read(&(zfsvfs)->z_teardown_lock, FTAG);
588 rw_enter(&ITOZ(ip)->z_xattr_lock, RW_WRITER);
591 * Before setting the xattr check to see if it already exists.
592 * This is done to ensure the following optional flags are honored.
594 * XATTR_CREATE: fail if xattr already exists
595 * XATTR_REPLACE: fail if xattr does not exist
597 * We also want to know if it resides in sa or dir, so we can make
598 * sure we don't end up with duplicate in both places.
600 error = __zpl_xattr_where(ip, name, &where, cr);
602 if (error != -ENODATA)
604 if (flags & XATTR_REPLACE)
607 /* The xattr to be removed already doesn't exist */
613 if (flags & XATTR_CREATE)
617 /* Preferentially store the xattr as a SA for better performance */
618 if (zfsvfs->z_use_sa && zp->z_is_sa &&
619 (zfsvfs->z_xattr_sa || (value == NULL && where & XATTR_IN_SA))) {
620 error = zpl_xattr_set_sa(ip, name, value, size, flags, cr);
623 * Successfully put into SA, we need to clear the one
626 if (where & XATTR_IN_DIR)
627 zpl_xattr_set_dir(ip, name, NULL, 0, 0, cr);
632 error = zpl_xattr_set_dir(ip, name, value, size, flags, cr);
634 * Successfully put into dir, we need to clear the one in SA.
636 if (error == 0 && (where & XATTR_IN_SA))
637 zpl_xattr_set_sa(ip, name, NULL, 0, 0, cr);
639 rw_exit(&ITOZ(ip)->z_xattr_lock);
640 rrm_exit(&(zfsvfs)->z_teardown_lock, FTAG);
641 spl_fstrans_unmark(cookie);
643 ASSERT3S(error, <=, 0);
649 * Extended user attributes
651 * "Extended user attributes may be assigned to files and directories for
652 * storing arbitrary additional information such as the mime type,
653 * character set or encoding of a file. The access permissions for user
654 * attributes are defined by the file permission bits: read permission
655 * is required to retrieve the attribute value, and writer permission is
656 * required to change it.
658 * The file permission bits of regular files and directories are
659 * interpreted differently from the file permission bits of special
660 * files and symbolic links. For regular files and directories the file
661 * permission bits define access to the file's contents, while for
662 * device special files they define access to the device described by
663 * the special file. The file permissions of symbolic links are not
664 * used in access checks. These differences would allow users to
665 * consume filesystem resources in a way not controllable by disk quotas
666 * for group or world writable special files and directories.
668 * For this reason, extended user attributes are allowed only for
669 * regular files and directories, and access to extended user attributes
670 * is restricted to the owner and to users with appropriate capabilities
671 * for directories with the sticky bit set (see the chmod(1) manual page
672 * for an explanation of the sticky bit)." - xattr(7)
674 * ZFS allows extended user attributes to be disabled administratively
675 * by setting the 'xattr=off' property on the dataset.
678 __zpl_xattr_user_list(struct inode *ip, char *list, size_t list_size,
679 const char *name, size_t name_len)
681 return (ITOZSB(ip)->z_flags & ZSB_XATTR);
683 ZPL_XATTR_LIST_WRAPPER(zpl_xattr_user_list);
686 __zpl_xattr_user_get(struct inode *ip, const char *name,
687 void *value, size_t size)
691 /* xattr_resolve_name will do this for us if this is defined */
692 #ifndef HAVE_XATTR_HANDLER_NAME
693 if (strcmp(name, "") == 0)
696 if (!(ITOZSB(ip)->z_flags & ZSB_XATTR))
697 return (-EOPNOTSUPP);
699 xattr_name = kmem_asprintf("%s%s", XATTR_USER_PREFIX, name);
700 error = zpl_xattr_get(ip, xattr_name, value, size);
705 ZPL_XATTR_GET_WRAPPER(zpl_xattr_user_get);
708 __zpl_xattr_user_set(struct inode *ip, const char *name,
709 const void *value, size_t size, int flags)
713 /* xattr_resolve_name will do this for us if this is defined */
714 #ifndef HAVE_XATTR_HANDLER_NAME
715 if (strcmp(name, "") == 0)
718 if (!(ITOZSB(ip)->z_flags & ZSB_XATTR))
719 return (-EOPNOTSUPP);
721 xattr_name = kmem_asprintf("%s%s", XATTR_USER_PREFIX, name);
722 error = zpl_xattr_set(ip, xattr_name, value, size, flags);
727 ZPL_XATTR_SET_WRAPPER(zpl_xattr_user_set);
729 xattr_handler_t zpl_xattr_user_handler =
731 .prefix = XATTR_USER_PREFIX,
732 .list = zpl_xattr_user_list,
733 .get = zpl_xattr_user_get,
734 .set = zpl_xattr_user_set,
738 * Trusted extended attributes
740 * "Trusted extended attributes are visible and accessible only to
741 * processes that have the CAP_SYS_ADMIN capability. Attributes in this
742 * class are used to implement mechanisms in user space (i.e., outside
743 * the kernel) which keep information in extended attributes to which
744 * ordinary processes should not have access." - xattr(7)
747 __zpl_xattr_trusted_list(struct inode *ip, char *list, size_t list_size,
748 const char *name, size_t name_len)
750 return (capable(CAP_SYS_ADMIN));
752 ZPL_XATTR_LIST_WRAPPER(zpl_xattr_trusted_list);
755 __zpl_xattr_trusted_get(struct inode *ip, const char *name,
756 void *value, size_t size)
761 if (!capable(CAP_SYS_ADMIN))
763 /* xattr_resolve_name will do this for us if this is defined */
764 #ifndef HAVE_XATTR_HANDLER_NAME
765 if (strcmp(name, "") == 0)
768 xattr_name = kmem_asprintf("%s%s", XATTR_TRUSTED_PREFIX, name);
769 error = zpl_xattr_get(ip, xattr_name, value, size);
774 ZPL_XATTR_GET_WRAPPER(zpl_xattr_trusted_get);
777 __zpl_xattr_trusted_set(struct inode *ip, const char *name,
778 const void *value, size_t size, int flags)
783 if (!capable(CAP_SYS_ADMIN))
785 /* xattr_resolve_name will do this for us if this is defined */
786 #ifndef HAVE_XATTR_HANDLER_NAME
787 if (strcmp(name, "") == 0)
790 xattr_name = kmem_asprintf("%s%s", XATTR_TRUSTED_PREFIX, name);
791 error = zpl_xattr_set(ip, xattr_name, value, size, flags);
796 ZPL_XATTR_SET_WRAPPER(zpl_xattr_trusted_set);
798 xattr_handler_t zpl_xattr_trusted_handler =
800 .prefix = XATTR_TRUSTED_PREFIX,
801 .list = zpl_xattr_trusted_list,
802 .get = zpl_xattr_trusted_get,
803 .set = zpl_xattr_trusted_set,
807 * Extended security attributes
809 * "The security attribute namespace is used by kernel security modules,
810 * such as Security Enhanced Linux, and also to implement file
811 * capabilities (see capabilities(7)). Read and write access
812 * permissions to security attributes depend on the policy implemented
813 * for each security attribute by the security module. When no security
814 * module is loaded, all processes have read access to extended security
815 * attributes, and write access is limited to processes that have the
816 * CAP_SYS_ADMIN capability." - xattr(7)
819 __zpl_xattr_security_list(struct inode *ip, char *list, size_t list_size,
820 const char *name, size_t name_len)
824 ZPL_XATTR_LIST_WRAPPER(zpl_xattr_security_list);
827 __zpl_xattr_security_get(struct inode *ip, const char *name,
828 void *value, size_t size)
832 /* xattr_resolve_name will do this for us if this is defined */
833 #ifndef HAVE_XATTR_HANDLER_NAME
834 if (strcmp(name, "") == 0)
837 xattr_name = kmem_asprintf("%s%s", XATTR_SECURITY_PREFIX, name);
838 error = zpl_xattr_get(ip, xattr_name, value, size);
843 ZPL_XATTR_GET_WRAPPER(zpl_xattr_security_get);
846 __zpl_xattr_security_set(struct inode *ip, const char *name,
847 const void *value, size_t size, int flags)
851 /* xattr_resolve_name will do this for us if this is defined */
852 #ifndef HAVE_XATTR_HANDLER_NAME
853 if (strcmp(name, "") == 0)
856 xattr_name = kmem_asprintf("%s%s", XATTR_SECURITY_PREFIX, name);
857 error = zpl_xattr_set(ip, xattr_name, value, size, flags);
862 ZPL_XATTR_SET_WRAPPER(zpl_xattr_security_set);
864 #ifdef HAVE_CALLBACK_SECURITY_INODE_INIT_SECURITY
866 __zpl_xattr_security_init(struct inode *ip, const struct xattr *xattrs,
869 const struct xattr *xattr;
872 for (xattr = xattrs; xattr->name != NULL; xattr++) {
873 error = __zpl_xattr_security_set(ip,
874 xattr->name, xattr->value, xattr->value_len, 0);
884 zpl_xattr_security_init(struct inode *ip, struct inode *dip,
885 const struct qstr *qstr)
887 return security_inode_init_security(ip, dip, qstr,
888 &__zpl_xattr_security_init, NULL);
893 zpl_xattr_security_init(struct inode *ip, struct inode *dip,
894 const struct qstr *qstr)
901 error = zpl_security_inode_init_security(ip, dip, qstr,
902 &name, &value, &len);
904 if (error == -EOPNOTSUPP)
910 error = __zpl_xattr_security_set(ip, name, value, len, 0);
917 #endif /* HAVE_CALLBACK_SECURITY_INODE_INIT_SECURITY */
920 * Security xattr namespace handlers.
922 xattr_handler_t zpl_xattr_security_handler = {
923 .prefix = XATTR_SECURITY_PREFIX,
924 .list = zpl_xattr_security_list,
925 .get = zpl_xattr_security_get,
926 .set = zpl_xattr_security_set,
930 * Extended system attributes
932 * "Extended system attributes are used by the kernel to store system
933 * objects such as Access Control Lists. Read and write access permissions
934 * to system attributes depend on the policy implemented for each system
935 * attribute implemented by filesystems in the kernel." - xattr(7)
937 #ifdef CONFIG_FS_POSIX_ACL
939 zpl_set_acl(struct inode *ip, struct posix_acl *acl, int type)
941 char *name, *value = NULL;
945 if (S_ISLNK(ip->i_mode))
946 return (-EOPNOTSUPP);
949 case ACL_TYPE_ACCESS:
950 name = XATTR_NAME_POSIX_ACL_ACCESS;
952 zpl_equivmode_t mode = ip->i_mode;
953 error = posix_acl_equiv_mode(acl, &mode);
958 * The mode bits will have been set by
959 * ->zfs_setattr()->zfs_acl_chmod_setattr()
960 * using the ZFS ACL conversion. If they
961 * differ from the Posix ACL conversion dirty
962 * the inode to write the Posix mode bits.
964 if (ip->i_mode != mode) {
966 ip->i_ctime = current_time(ip);
967 zfs_mark_inode_dirty(ip);
976 case ACL_TYPE_DEFAULT:
977 name = XATTR_NAME_POSIX_ACL_DEFAULT;
978 if (!S_ISDIR(ip->i_mode))
979 return (acl ? -EACCES : 0);
987 size = posix_acl_xattr_size(acl->a_count);
988 value = kmem_alloc(size, KM_SLEEP);
990 error = zpl_acl_to_xattr(acl, value, size);
992 kmem_free(value, size);
997 error = zpl_xattr_set(ip, name, value, size, 0);
999 kmem_free(value, size);
1003 zpl_set_cached_acl(ip, type, acl);
1005 zpl_forget_cached_acl(ip, type);
1012 zpl_get_acl(struct inode *ip, int type)
1014 struct posix_acl *acl;
1020 * As of Linux 3.14, the kernel get_acl will check this for us.
1021 * Also as of Linux 4.7, comparing against ACL_NOT_CACHED is wrong
1022 * as the kernel get_acl will set it to temporary sentinel value.
1024 #ifndef HAVE_KERNEL_GET_ACL_HANDLE_CACHE
1025 acl = get_cached_acl(ip, type);
1026 if (acl != ACL_NOT_CACHED)
1031 case ACL_TYPE_ACCESS:
1032 name = XATTR_NAME_POSIX_ACL_ACCESS;
1034 case ACL_TYPE_DEFAULT:
1035 name = XATTR_NAME_POSIX_ACL_DEFAULT;
1038 return (ERR_PTR(-EINVAL));
1041 size = zpl_xattr_get(ip, name, NULL, 0);
1043 value = kmem_alloc(size, KM_SLEEP);
1044 size = zpl_xattr_get(ip, name, value, size);
1048 acl = zpl_acl_from_xattr(value, size);
1049 } else if (size == -ENODATA || size == -ENOSYS) {
1052 acl = ERR_PTR(-EIO);
1056 kmem_free(value, size);
1058 /* As of Linux 4.7, the kernel get_acl will set this for us */
1059 #ifndef HAVE_KERNEL_GET_ACL_HANDLE_CACHE
1061 zpl_set_cached_acl(ip, type, acl);
1067 #if !defined(HAVE_GET_ACL)
1069 __zpl_check_acl(struct inode *ip, int mask)
1071 struct posix_acl *acl;
1074 acl = zpl_get_acl(ip, ACL_TYPE_ACCESS);
1076 return (PTR_ERR(acl));
1079 error = posix_acl_permission(ip, acl, mask);
1080 zpl_posix_acl_release(acl);
1087 #if defined(HAVE_CHECK_ACL_WITH_FLAGS)
1089 zpl_check_acl(struct inode *ip, int mask, unsigned int flags)
1091 return (__zpl_check_acl(ip, mask));
1093 #elif defined(HAVE_CHECK_ACL)
1095 zpl_check_acl(struct inode *ip, int mask)
1097 return (__zpl_check_acl(ip, mask));
1099 #elif defined(HAVE_PERMISSION_WITH_NAMEIDATA)
1101 zpl_permission(struct inode *ip, int mask, struct nameidata *nd)
1103 return (generic_permission(ip, mask, __zpl_check_acl));
1105 #elif defined(HAVE_PERMISSION)
1107 zpl_permission(struct inode *ip, int mask)
1109 return (generic_permission(ip, mask, __zpl_check_acl));
1111 #endif /* HAVE_CHECK_ACL | HAVE_PERMISSION */
1112 #endif /* !HAVE_GET_ACL */
1115 zpl_init_acl(struct inode *ip, struct inode *dir)
1117 struct posix_acl *acl = NULL;
1120 if (ITOZSB(ip)->z_acl_type != ZFS_ACLTYPE_POSIXACL)
1123 if (!S_ISLNK(ip->i_mode)) {
1124 if (ITOZSB(ip)->z_acl_type == ZFS_ACLTYPE_POSIXACL) {
1125 acl = zpl_get_acl(dir, ACL_TYPE_DEFAULT);
1127 return (PTR_ERR(acl));
1131 ip->i_mode &= ~current_umask();
1132 ip->i_ctime = current_time(ip);
1133 zfs_mark_inode_dirty(ip);
1138 if ((ITOZSB(ip)->z_acl_type == ZFS_ACLTYPE_POSIXACL) && acl) {
1141 if (S_ISDIR(ip->i_mode)) {
1142 error = zpl_set_acl(ip, acl, ACL_TYPE_DEFAULT);
1148 error = __posix_acl_create(&acl, GFP_KERNEL, &mode);
1151 zfs_mark_inode_dirty(ip);
1153 error = zpl_set_acl(ip, acl, ACL_TYPE_ACCESS);
1157 zpl_posix_acl_release(acl);
1163 zpl_chmod_acl(struct inode *ip)
1165 struct posix_acl *acl;
1168 if (ITOZSB(ip)->z_acl_type != ZFS_ACLTYPE_POSIXACL)
1171 if (S_ISLNK(ip->i_mode))
1172 return (-EOPNOTSUPP);
1174 acl = zpl_get_acl(ip, ACL_TYPE_ACCESS);
1175 if (IS_ERR(acl) || !acl)
1176 return (PTR_ERR(acl));
1178 error = __posix_acl_chmod(&acl, GFP_KERNEL, ip->i_mode);
1180 error = zpl_set_acl(ip, acl, ACL_TYPE_ACCESS);
1182 zpl_posix_acl_release(acl);
1188 __zpl_xattr_acl_list_access(struct inode *ip, char *list, size_t list_size,
1189 const char *name, size_t name_len)
1191 char *xattr_name = XATTR_NAME_POSIX_ACL_ACCESS;
1192 size_t xattr_size = sizeof (XATTR_NAME_POSIX_ACL_ACCESS);
1194 if (ITOZSB(ip)->z_acl_type != ZFS_ACLTYPE_POSIXACL)
1197 if (list && xattr_size <= list_size)
1198 memcpy(list, xattr_name, xattr_size);
1200 return (xattr_size);
1202 ZPL_XATTR_LIST_WRAPPER(zpl_xattr_acl_list_access);
1205 __zpl_xattr_acl_list_default(struct inode *ip, char *list, size_t list_size,
1206 const char *name, size_t name_len)
1208 char *xattr_name = XATTR_NAME_POSIX_ACL_DEFAULT;
1209 size_t xattr_size = sizeof (XATTR_NAME_POSIX_ACL_DEFAULT);
1211 if (ITOZSB(ip)->z_acl_type != ZFS_ACLTYPE_POSIXACL)
1214 if (list && xattr_size <= list_size)
1215 memcpy(list, xattr_name, xattr_size);
1217 return (xattr_size);
1219 ZPL_XATTR_LIST_WRAPPER(zpl_xattr_acl_list_default);
1222 __zpl_xattr_acl_get_access(struct inode *ip, const char *name,
1223 void *buffer, size_t size)
1225 struct posix_acl *acl;
1226 int type = ACL_TYPE_ACCESS;
1228 /* xattr_resolve_name will do this for us if this is defined */
1229 #ifndef HAVE_XATTR_HANDLER_NAME
1230 if (strcmp(name, "") != 0)
1233 if (ITOZSB(ip)->z_acl_type != ZFS_ACLTYPE_POSIXACL)
1234 return (-EOPNOTSUPP);
1236 acl = zpl_get_acl(ip, type);
1238 return (PTR_ERR(acl));
1242 error = zpl_acl_to_xattr(acl, buffer, size);
1243 zpl_posix_acl_release(acl);
1247 ZPL_XATTR_GET_WRAPPER(zpl_xattr_acl_get_access);
1250 __zpl_xattr_acl_get_default(struct inode *ip, const char *name,
1251 void *buffer, size_t size)
1253 struct posix_acl *acl;
1254 int type = ACL_TYPE_DEFAULT;
1256 /* xattr_resolve_name will do this for us if this is defined */
1257 #ifndef HAVE_XATTR_HANDLER_NAME
1258 if (strcmp(name, "") != 0)
1261 if (ITOZSB(ip)->z_acl_type != ZFS_ACLTYPE_POSIXACL)
1262 return (-EOPNOTSUPP);
1264 acl = zpl_get_acl(ip, type);
1266 return (PTR_ERR(acl));
1270 error = zpl_acl_to_xattr(acl, buffer, size);
1271 zpl_posix_acl_release(acl);
1275 ZPL_XATTR_GET_WRAPPER(zpl_xattr_acl_get_default);
1278 __zpl_xattr_acl_set_access(struct inode *ip, const char *name,
1279 const void *value, size_t size, int flags)
1281 struct posix_acl *acl;
1282 int type = ACL_TYPE_ACCESS;
1284 /* xattr_resolve_name will do this for us if this is defined */
1285 #ifndef HAVE_XATTR_HANDLER_NAME
1286 if (strcmp(name, "") != 0)
1289 if (ITOZSB(ip)->z_acl_type != ZFS_ACLTYPE_POSIXACL)
1290 return (-EOPNOTSUPP);
1292 if (!zpl_inode_owner_or_capable(ip))
1296 acl = zpl_acl_from_xattr(value, size);
1298 return (PTR_ERR(acl));
1300 error = zpl_posix_acl_valid(ip, acl);
1302 zpl_posix_acl_release(acl);
1310 error = zpl_set_acl(ip, acl, type);
1311 zpl_posix_acl_release(acl);
1315 ZPL_XATTR_SET_WRAPPER(zpl_xattr_acl_set_access);
1318 __zpl_xattr_acl_set_default(struct inode *ip, const char *name,
1319 const void *value, size_t size, int flags)
1321 struct posix_acl *acl;
1322 int type = ACL_TYPE_DEFAULT;
1324 /* xattr_resolve_name will do this for us if this is defined */
1325 #ifndef HAVE_XATTR_HANDLER_NAME
1326 if (strcmp(name, "") != 0)
1329 if (ITOZSB(ip)->z_acl_type != ZFS_ACLTYPE_POSIXACL)
1330 return (-EOPNOTSUPP);
1332 if (!zpl_inode_owner_or_capable(ip))
1336 acl = zpl_acl_from_xattr(value, size);
1338 return (PTR_ERR(acl));
1340 error = zpl_posix_acl_valid(ip, acl);
1342 zpl_posix_acl_release(acl);
1350 error = zpl_set_acl(ip, acl, type);
1351 zpl_posix_acl_release(acl);
1355 ZPL_XATTR_SET_WRAPPER(zpl_xattr_acl_set_default);
1358 * ACL access xattr namespace handlers.
1360 * Use .name instead of .prefix when available. xattr_resolve_name will match
1361 * whole name and reject anything that has .name only as prefix.
1363 xattr_handler_t zpl_xattr_acl_access_handler =
1365 #ifdef HAVE_XATTR_HANDLER_NAME
1366 .name = XATTR_NAME_POSIX_ACL_ACCESS,
1368 .prefix = XATTR_NAME_POSIX_ACL_ACCESS,
1370 .list = zpl_xattr_acl_list_access,
1371 .get = zpl_xattr_acl_get_access,
1372 .set = zpl_xattr_acl_set_access,
1373 #if defined(HAVE_XATTR_LIST_SIMPLE) || \
1374 defined(HAVE_XATTR_LIST_DENTRY) || \
1375 defined(HAVE_XATTR_LIST_HANDLER)
1376 .flags = ACL_TYPE_ACCESS,
1381 * ACL default xattr namespace handlers.
1383 * Use .name instead of .prefix when available. xattr_resolve_name will match
1384 * whole name and reject anything that has .name only as prefix.
1386 xattr_handler_t zpl_xattr_acl_default_handler =
1388 #ifdef HAVE_XATTR_HANDLER_NAME
1389 .name = XATTR_NAME_POSIX_ACL_DEFAULT,
1391 .prefix = XATTR_NAME_POSIX_ACL_DEFAULT,
1393 .list = zpl_xattr_acl_list_default,
1394 .get = zpl_xattr_acl_get_default,
1395 .set = zpl_xattr_acl_set_default,
1396 #if defined(HAVE_XATTR_LIST_SIMPLE) || \
1397 defined(HAVE_XATTR_LIST_DENTRY) || \
1398 defined(HAVE_XATTR_LIST_HANDLER)
1399 .flags = ACL_TYPE_DEFAULT,
1403 #endif /* CONFIG_FS_POSIX_ACL */
1405 xattr_handler_t *zpl_xattr_handlers[] = {
1406 &zpl_xattr_security_handler,
1407 &zpl_xattr_trusted_handler,
1408 &zpl_xattr_user_handler,
1409 #ifdef CONFIG_FS_POSIX_ACL
1410 &zpl_xattr_acl_access_handler,
1411 &zpl_xattr_acl_default_handler,
1412 #endif /* CONFIG_FS_POSIX_ACL */
1416 static const struct xattr_handler *
1417 zpl_xattr_handler(const char *name)
1419 if (strncmp(name, XATTR_USER_PREFIX,
1420 XATTR_USER_PREFIX_LEN) == 0)
1421 return (&zpl_xattr_user_handler);
1423 if (strncmp(name, XATTR_TRUSTED_PREFIX,
1424 XATTR_TRUSTED_PREFIX_LEN) == 0)
1425 return (&zpl_xattr_trusted_handler);
1427 if (strncmp(name, XATTR_SECURITY_PREFIX,
1428 XATTR_SECURITY_PREFIX_LEN) == 0)
1429 return (&zpl_xattr_security_handler);
1431 #ifdef CONFIG_FS_POSIX_ACL
1432 if (strncmp(name, XATTR_NAME_POSIX_ACL_ACCESS,
1433 sizeof (XATTR_NAME_POSIX_ACL_ACCESS)) == 0)
1434 return (&zpl_xattr_acl_access_handler);
1436 if (strncmp(name, XATTR_NAME_POSIX_ACL_DEFAULT,
1437 sizeof (XATTR_NAME_POSIX_ACL_DEFAULT)) == 0)
1438 return (&zpl_xattr_acl_default_handler);
1439 #endif /* CONFIG_FS_POSIX_ACL */
1444 #if !defined(HAVE_POSIX_ACL_RELEASE) || defined(HAVE_POSIX_ACL_RELEASE_GPL_ONLY)
1445 struct acl_rel_struct {
1446 struct acl_rel_struct *next;
1447 struct posix_acl *acl;
1451 #define ACL_REL_GRACE (60*HZ)
1452 #define ACL_REL_WINDOW (1*HZ)
1453 #define ACL_REL_SCHED (ACL_REL_GRACE+ACL_REL_WINDOW)
1456 * Lockless multi-producer single-consumer fifo list.
1457 * Nodes are added to tail and removed from head. Tail pointer is our
1458 * synchronization point. It always points to the next pointer of the last
1459 * node, or head if list is empty.
1461 static struct acl_rel_struct *acl_rel_head = NULL;
1462 static struct acl_rel_struct **acl_rel_tail = &acl_rel_head;
1465 zpl_posix_acl_free(void *arg)
1467 struct acl_rel_struct *freelist = NULL;
1468 struct acl_rel_struct *a;
1470 boolean_t refire = B_FALSE;
1472 ASSERT3P(acl_rel_head, !=, NULL);
1473 while (acl_rel_head) {
1475 if (ddi_get_lbolt() - a->time >= ACL_REL_GRACE) {
1477 * If a is the last node we need to reset tail, but we
1478 * need to use cmpxchg to make sure it is still the
1481 if (acl_rel_tail == &a->next) {
1482 acl_rel_head = NULL;
1483 if (cmpxchg(&acl_rel_tail, &a->next,
1484 &acl_rel_head) == &a->next) {
1485 ASSERT3P(a->next, ==, NULL);
1492 * a is not last node, make sure next pointer is set
1493 * by the adder and advance the head.
1495 while (ACCESS_ONCE(a->next) == NULL)
1497 acl_rel_head = a->next;
1502 * a is still in grace period. We are responsible to
1503 * reschedule the free task, since adder will only do
1504 * so if list is empty.
1506 new_time = a->time + ACL_REL_SCHED;
1513 taskq_dispatch_delay(system_delay_taskq, zpl_posix_acl_free,
1514 NULL, TQ_SLEEP, new_time);
1520 kmem_free(a, sizeof (struct acl_rel_struct));
1525 zpl_posix_acl_release_impl(struct posix_acl *acl)
1527 struct acl_rel_struct *a, **prev;
1529 a = kmem_alloc(sizeof (struct acl_rel_struct), KM_SLEEP);
1532 a->time = ddi_get_lbolt();
1533 /* atomically points tail to us and get the previous tail */
1534 prev = xchg(&acl_rel_tail, &a->next);
1535 ASSERT3P(*prev, ==, NULL);
1537 /* if it was empty before, schedule the free task */
1538 if (prev == &acl_rel_head)
1539 taskq_dispatch_delay(system_delay_taskq, zpl_posix_acl_free,
1540 NULL, TQ_SLEEP, ddi_get_lbolt() + ACL_REL_SCHED);