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();
250 rw_enter(&zp->z_xattr_lock, RW_READER);
252 if (zfsvfs->z_use_sa && zp->z_is_sa) {
253 error = zpl_xattr_list_sa(&xf);
258 error = zpl_xattr_list_dir(&xf, cr);
265 rw_exit(&zp->z_xattr_lock);
267 spl_fstrans_unmark(cookie);
274 zpl_xattr_get_dir(struct inode *ip, const char *name, void *value,
275 size_t size, cred_t *cr)
277 struct inode *dxip = NULL;
278 struct inode *xip = NULL;
282 /* Lookup the xattr directory */
283 error = -zfs_lookup(ip, NULL, &dxip, LOOKUP_XATTR, cr, NULL, NULL);
287 /* Lookup a specific xattr name in the directory */
288 error = -zfs_lookup(dxip, (char *)name, &xip, 0, cr, NULL, NULL);
293 error = i_size_read(xip);
297 if (size < i_size_read(xip)) {
302 error = zpl_read_common(xip, value, size, &pos, UIO_SYSSPACE, 0, cr);
314 zpl_xattr_get_sa(struct inode *ip, const char *name, void *value, size_t size)
316 znode_t *zp = ITOZ(ip);
321 ASSERT(RW_LOCK_HELD(&zp->z_xattr_lock));
323 mutex_enter(&zp->z_lock);
324 if (zp->z_xattr_cached == NULL)
325 error = -zfs_sa_get_xattr(zp);
326 mutex_exit(&zp->z_lock);
331 ASSERT(zp->z_xattr_cached);
332 error = -nvlist_lookup_byte_array(zp->z_xattr_cached, name,
333 &nv_value, &nv_size);
337 if (size == 0 || value == NULL)
343 memcpy(value, nv_value, nv_size);
349 __zpl_xattr_get(struct inode *ip, const char *name, void *value, size_t size,
352 znode_t *zp = ITOZ(ip);
353 zfsvfs_t *zfsvfs = ZTOZSB(zp);
356 ASSERT(RW_LOCK_HELD(&zp->z_xattr_lock));
358 if (zfsvfs->z_use_sa && zp->z_is_sa) {
359 error = zpl_xattr_get_sa(ip, name, value, size);
360 if (error != -ENOENT)
364 error = zpl_xattr_get_dir(ip, name, value, size, cr);
366 if (error == -ENOENT)
372 #define XATTR_NOENT 0x0
373 #define XATTR_IN_SA 0x1
374 #define XATTR_IN_DIR 0x2
375 /* check where the xattr resides */
377 __zpl_xattr_where(struct inode *ip, const char *name, int *where, cred_t *cr)
379 znode_t *zp = ITOZ(ip);
380 zfsvfs_t *zfsvfs = ZTOZSB(zp);
384 ASSERT(RW_LOCK_HELD(&zp->z_xattr_lock));
386 *where = XATTR_NOENT;
387 if (zfsvfs->z_use_sa && zp->z_is_sa) {
388 error = zpl_xattr_get_sa(ip, name, NULL, 0);
390 *where |= XATTR_IN_SA;
391 else if (error != -ENOENT)
395 error = zpl_xattr_get_dir(ip, name, NULL, 0, cr);
397 *where |= XATTR_IN_DIR;
398 else if (error != -ENOENT)
401 if (*where == (XATTR_IN_SA|XATTR_IN_DIR))
402 cmn_err(CE_WARN, "ZFS: inode %p has xattr \"%s\""
403 " in both SA and dir", ip, name);
404 if (*where == XATTR_NOENT)
412 zpl_xattr_get(struct inode *ip, const char *name, void *value, size_t size)
414 znode_t *zp = ITOZ(ip);
415 zfsvfs_t *zfsvfs = ZTOZSB(zp);
417 fstrans_cookie_t cookie;
421 cookie = spl_fstrans_mark();
424 rw_enter(&zp->z_xattr_lock, RW_READER);
425 error = __zpl_xattr_get(ip, name, value, size, cr);
426 rw_exit(&zp->z_xattr_lock);
428 spl_fstrans_unmark(cookie);
435 zpl_xattr_set_dir(struct inode *ip, const char *name, const void *value,
436 size_t size, int flags, cred_t *cr)
438 struct inode *dxip = NULL;
439 struct inode *xip = NULL;
442 int lookup_flags, error;
443 const int xattr_mode = S_IFREG | 0644;
447 * Lookup the xattr directory. When we're adding an entry pass
448 * CREATE_XATTR_DIR to ensure the xattr directory is created.
449 * When removing an entry this flag is not passed to avoid
450 * unnecessarily creating a new xattr directory.
452 lookup_flags = LOOKUP_XATTR;
454 lookup_flags |= CREATE_XATTR_DIR;
456 error = -zfs_lookup(ip, NULL, &dxip, lookup_flags, cr, NULL, NULL);
460 /* Lookup a specific xattr name in the directory */
461 error = -zfs_lookup(dxip, (char *)name, &xip, 0, cr, NULL, NULL);
462 if (error && (error != -ENOENT))
467 /* Remove a specific name xattr when value is set to NULL. */
470 error = -zfs_remove(dxip, (char *)name, cr, 0);
475 /* Lookup failed create a new xattr. */
477 vap = kmem_zalloc(sizeof (vattr_t), KM_SLEEP);
478 vap->va_mode = xattr_mode;
479 vap->va_mask = ATTR_MODE;
480 vap->va_uid = crgetfsuid(cr);
481 vap->va_gid = crgetfsgid(cr);
483 error = -zfs_create(dxip, (char *)name, vap, 0, 0644, &xip,
491 error = -zfs_freesp(ITOZ(xip), 0, 0, xattr_mode, TRUE);
495 wrote = zpl_write_common(xip, value, size, &pos, UIO_SYSSPACE, 0, cr);
502 ip->i_ctime = current_time(ip);
503 zfs_mark_inode_dirty(ip);
507 kmem_free(vap, sizeof (vattr_t));
515 if (error == -ENOENT)
518 ASSERT3S(error, <=, 0);
524 zpl_xattr_set_sa(struct inode *ip, const char *name, const void *value,
525 size_t size, int flags, cred_t *cr)
527 znode_t *zp = ITOZ(ip);
532 mutex_enter(&zp->z_lock);
533 if (zp->z_xattr_cached == NULL)
534 error = -zfs_sa_get_xattr(zp);
535 mutex_exit(&zp->z_lock);
540 ASSERT(zp->z_xattr_cached);
541 nvl = zp->z_xattr_cached;
544 error = -nvlist_remove(nvl, name, DATA_TYPE_BYTE_ARRAY);
545 if (error == -ENOENT)
546 error = zpl_xattr_set_dir(ip, name, NULL, 0, flags, cr);
548 /* Limited to 32k to keep nvpair memory allocations small */
549 if (size > DXATTR_MAX_ENTRY_SIZE)
552 /* Prevent the DXATTR SA from consuming the entire SA region */
553 error = -nvlist_size(nvl, &sa_size, NV_ENCODE_XDR);
557 if (sa_size > DXATTR_MAX_SA_SIZE)
560 error = -nvlist_add_byte_array(nvl, name,
561 (uchar_t *)value, size);
565 * Update the SA for additions, modifications, and removals. On
566 * error drop the inconsistent cached version of the nvlist, it
567 * will be reconstructed from the ARC when next accessed.
570 error = -zfs_sa_set_xattr(zp);
574 zp->z_xattr_cached = NULL;
577 ASSERT3S(error, <=, 0);
583 zpl_xattr_set(struct inode *ip, const char *name, const void *value,
584 size_t size, int flags)
586 znode_t *zp = ITOZ(ip);
587 zfsvfs_t *zfsvfs = ZTOZSB(zp);
589 fstrans_cookie_t cookie;
594 cookie = spl_fstrans_mark();
597 rw_enter(&ITOZ(ip)->z_xattr_lock, RW_WRITER);
600 * Before setting the xattr check to see if it already exists.
601 * This is done to ensure the following optional flags are honored.
603 * XATTR_CREATE: fail if xattr already exists
604 * XATTR_REPLACE: fail if xattr does not exist
606 * We also want to know if it resides in sa or dir, so we can make
607 * sure we don't end up with duplicate in both places.
609 error = __zpl_xattr_where(ip, name, &where, cr);
611 if (error != -ENODATA)
613 if (flags & XATTR_REPLACE)
616 /* The xattr to be removed already doesn't exist */
622 if (flags & XATTR_CREATE)
626 /* Preferentially store the xattr as a SA for better performance */
627 if (zfsvfs->z_use_sa && zp->z_is_sa &&
628 (zfsvfs->z_xattr_sa || (value == NULL && where & XATTR_IN_SA))) {
629 error = zpl_xattr_set_sa(ip, name, value, size, flags, cr);
632 * Successfully put into SA, we need to clear the one
635 if (where & XATTR_IN_DIR)
636 zpl_xattr_set_dir(ip, name, NULL, 0, 0, cr);
641 error = zpl_xattr_set_dir(ip, name, value, size, flags, cr);
643 * Successfully put into dir, we need to clear the one in SA.
645 if (error == 0 && (where & XATTR_IN_SA))
646 zpl_xattr_set_sa(ip, name, NULL, 0, 0, cr);
648 rw_exit(&ITOZ(ip)->z_xattr_lock);
650 spl_fstrans_unmark(cookie);
652 ASSERT3S(error, <=, 0);
658 * Extended user attributes
660 * "Extended user attributes may be assigned to files and directories for
661 * storing arbitrary additional information such as the mime type,
662 * character set or encoding of a file. The access permissions for user
663 * attributes are defined by the file permission bits: read permission
664 * is required to retrieve the attribute value, and writer permission is
665 * required to change it.
667 * The file permission bits of regular files and directories are
668 * interpreted differently from the file permission bits of special
669 * files and symbolic links. For regular files and directories the file
670 * permission bits define access to the file's contents, while for
671 * device special files they define access to the device described by
672 * the special file. The file permissions of symbolic links are not
673 * used in access checks. These differences would allow users to
674 * consume filesystem resources in a way not controllable by disk quotas
675 * for group or world writable special files and directories.
677 * For this reason, extended user attributes are allowed only for
678 * regular files and directories, and access to extended user attributes
679 * is restricted to the owner and to users with appropriate capabilities
680 * for directories with the sticky bit set (see the chmod(1) manual page
681 * for an explanation of the sticky bit)." - xattr(7)
683 * ZFS allows extended user attributes to be disabled administratively
684 * by setting the 'xattr=off' property on the dataset.
687 __zpl_xattr_user_list(struct inode *ip, char *list, size_t list_size,
688 const char *name, size_t name_len)
690 return (ITOZSB(ip)->z_flags & ZSB_XATTR);
692 ZPL_XATTR_LIST_WRAPPER(zpl_xattr_user_list);
695 __zpl_xattr_user_get(struct inode *ip, const char *name,
696 void *value, size_t size)
700 /* xattr_resolve_name will do this for us if this is defined */
701 #ifndef HAVE_XATTR_HANDLER_NAME
702 if (strcmp(name, "") == 0)
705 if (!(ITOZSB(ip)->z_flags & ZSB_XATTR))
706 return (-EOPNOTSUPP);
708 xattr_name = kmem_asprintf("%s%s", XATTR_USER_PREFIX, name);
709 error = zpl_xattr_get(ip, xattr_name, value, size);
714 ZPL_XATTR_GET_WRAPPER(zpl_xattr_user_get);
717 __zpl_xattr_user_set(struct inode *ip, const char *name,
718 const void *value, size_t size, int flags)
722 /* xattr_resolve_name will do this for us if this is defined */
723 #ifndef HAVE_XATTR_HANDLER_NAME
724 if (strcmp(name, "") == 0)
727 if (!(ITOZSB(ip)->z_flags & ZSB_XATTR))
728 return (-EOPNOTSUPP);
730 xattr_name = kmem_asprintf("%s%s", XATTR_USER_PREFIX, name);
731 error = zpl_xattr_set(ip, xattr_name, value, size, flags);
736 ZPL_XATTR_SET_WRAPPER(zpl_xattr_user_set);
738 xattr_handler_t zpl_xattr_user_handler =
740 .prefix = XATTR_USER_PREFIX,
741 .list = zpl_xattr_user_list,
742 .get = zpl_xattr_user_get,
743 .set = zpl_xattr_user_set,
747 * Trusted extended attributes
749 * "Trusted extended attributes are visible and accessible only to
750 * processes that have the CAP_SYS_ADMIN capability. Attributes in this
751 * class are used to implement mechanisms in user space (i.e., outside
752 * the kernel) which keep information in extended attributes to which
753 * ordinary processes should not have access." - xattr(7)
756 __zpl_xattr_trusted_list(struct inode *ip, char *list, size_t list_size,
757 const char *name, size_t name_len)
759 return (capable(CAP_SYS_ADMIN));
761 ZPL_XATTR_LIST_WRAPPER(zpl_xattr_trusted_list);
764 __zpl_xattr_trusted_get(struct inode *ip, const char *name,
765 void *value, size_t size)
770 if (!capable(CAP_SYS_ADMIN))
772 /* xattr_resolve_name will do this for us if this is defined */
773 #ifndef HAVE_XATTR_HANDLER_NAME
774 if (strcmp(name, "") == 0)
777 xattr_name = kmem_asprintf("%s%s", XATTR_TRUSTED_PREFIX, name);
778 error = zpl_xattr_get(ip, xattr_name, value, size);
783 ZPL_XATTR_GET_WRAPPER(zpl_xattr_trusted_get);
786 __zpl_xattr_trusted_set(struct inode *ip, const char *name,
787 const void *value, size_t size, int flags)
792 if (!capable(CAP_SYS_ADMIN))
794 /* xattr_resolve_name will do this for us if this is defined */
795 #ifndef HAVE_XATTR_HANDLER_NAME
796 if (strcmp(name, "") == 0)
799 xattr_name = kmem_asprintf("%s%s", XATTR_TRUSTED_PREFIX, name);
800 error = zpl_xattr_set(ip, xattr_name, value, size, flags);
805 ZPL_XATTR_SET_WRAPPER(zpl_xattr_trusted_set);
807 xattr_handler_t zpl_xattr_trusted_handler =
809 .prefix = XATTR_TRUSTED_PREFIX,
810 .list = zpl_xattr_trusted_list,
811 .get = zpl_xattr_trusted_get,
812 .set = zpl_xattr_trusted_set,
816 * Extended security attributes
818 * "The security attribute namespace is used by kernel security modules,
819 * such as Security Enhanced Linux, and also to implement file
820 * capabilities (see capabilities(7)). Read and write access
821 * permissions to security attributes depend on the policy implemented
822 * for each security attribute by the security module. When no security
823 * module is loaded, all processes have read access to extended security
824 * attributes, and write access is limited to processes that have the
825 * CAP_SYS_ADMIN capability." - xattr(7)
828 __zpl_xattr_security_list(struct inode *ip, char *list, size_t list_size,
829 const char *name, size_t name_len)
833 ZPL_XATTR_LIST_WRAPPER(zpl_xattr_security_list);
836 __zpl_xattr_security_get(struct inode *ip, const char *name,
837 void *value, size_t size)
841 /* xattr_resolve_name will do this for us if this is defined */
842 #ifndef HAVE_XATTR_HANDLER_NAME
843 if (strcmp(name, "") == 0)
846 xattr_name = kmem_asprintf("%s%s", XATTR_SECURITY_PREFIX, name);
847 error = zpl_xattr_get(ip, xattr_name, value, size);
852 ZPL_XATTR_GET_WRAPPER(zpl_xattr_security_get);
855 __zpl_xattr_security_set(struct inode *ip, const char *name,
856 const void *value, size_t size, int flags)
860 /* xattr_resolve_name will do this for us if this is defined */
861 #ifndef HAVE_XATTR_HANDLER_NAME
862 if (strcmp(name, "") == 0)
865 xattr_name = kmem_asprintf("%s%s", XATTR_SECURITY_PREFIX, name);
866 error = zpl_xattr_set(ip, xattr_name, value, size, flags);
871 ZPL_XATTR_SET_WRAPPER(zpl_xattr_security_set);
873 #ifdef HAVE_CALLBACK_SECURITY_INODE_INIT_SECURITY
875 __zpl_xattr_security_init(struct inode *ip, const struct xattr *xattrs,
878 const struct xattr *xattr;
881 for (xattr = xattrs; xattr->name != NULL; xattr++) {
882 error = __zpl_xattr_security_set(ip,
883 xattr->name, xattr->value, xattr->value_len, 0);
893 zpl_xattr_security_init(struct inode *ip, struct inode *dip,
894 const struct qstr *qstr)
896 return security_inode_init_security(ip, dip, qstr,
897 &__zpl_xattr_security_init, NULL);
902 zpl_xattr_security_init(struct inode *ip, struct inode *dip,
903 const struct qstr *qstr)
910 error = zpl_security_inode_init_security(ip, dip, qstr,
911 &name, &value, &len);
913 if (error == -EOPNOTSUPP)
919 error = __zpl_xattr_security_set(ip, name, value, len, 0);
926 #endif /* HAVE_CALLBACK_SECURITY_INODE_INIT_SECURITY */
929 * Security xattr namespace handlers.
931 xattr_handler_t zpl_xattr_security_handler = {
932 .prefix = XATTR_SECURITY_PREFIX,
933 .list = zpl_xattr_security_list,
934 .get = zpl_xattr_security_get,
935 .set = zpl_xattr_security_set,
939 * Extended system attributes
941 * "Extended system attributes are used by the kernel to store system
942 * objects such as Access Control Lists. Read and write access permissions
943 * to system attributes depend on the policy implemented for each system
944 * attribute implemented by filesystems in the kernel." - xattr(7)
946 #ifdef CONFIG_FS_POSIX_ACL
948 zpl_set_acl(struct inode *ip, struct posix_acl *acl, int type)
950 char *name, *value = NULL;
954 if (S_ISLNK(ip->i_mode))
955 return (-EOPNOTSUPP);
958 case ACL_TYPE_ACCESS:
959 name = XATTR_NAME_POSIX_ACL_ACCESS;
961 zpl_equivmode_t mode = ip->i_mode;
962 error = posix_acl_equiv_mode(acl, &mode);
967 * The mode bits will have been set by
968 * ->zfs_setattr()->zfs_acl_chmod_setattr()
969 * using the ZFS ACL conversion. If they
970 * differ from the Posix ACL conversion dirty
971 * the inode to write the Posix mode bits.
973 if (ip->i_mode != mode) {
975 ip->i_ctime = current_time(ip);
976 zfs_mark_inode_dirty(ip);
985 case ACL_TYPE_DEFAULT:
986 name = XATTR_NAME_POSIX_ACL_DEFAULT;
987 if (!S_ISDIR(ip->i_mode))
988 return (acl ? -EACCES : 0);
996 size = posix_acl_xattr_size(acl->a_count);
997 value = kmem_alloc(size, KM_SLEEP);
999 error = zpl_acl_to_xattr(acl, value, size);
1001 kmem_free(value, size);
1006 error = zpl_xattr_set(ip, name, value, size, 0);
1008 kmem_free(value, size);
1012 zpl_set_cached_acl(ip, type, acl);
1014 zpl_forget_cached_acl(ip, type);
1021 zpl_get_acl(struct inode *ip, int type)
1023 struct posix_acl *acl;
1029 * As of Linux 3.14, the kernel get_acl will check this for us.
1030 * Also as of Linux 4.7, comparing against ACL_NOT_CACHED is wrong
1031 * as the kernel get_acl will set it to temporary sentinel value.
1033 #ifndef HAVE_KERNEL_GET_ACL_HANDLE_CACHE
1034 acl = get_cached_acl(ip, type);
1035 if (acl != ACL_NOT_CACHED)
1040 case ACL_TYPE_ACCESS:
1041 name = XATTR_NAME_POSIX_ACL_ACCESS;
1043 case ACL_TYPE_DEFAULT:
1044 name = XATTR_NAME_POSIX_ACL_DEFAULT;
1047 return (ERR_PTR(-EINVAL));
1050 size = zpl_xattr_get(ip, name, NULL, 0);
1052 value = kmem_alloc(size, KM_SLEEP);
1053 size = zpl_xattr_get(ip, name, value, size);
1057 acl = zpl_acl_from_xattr(value, size);
1058 } else if (size == -ENODATA || size == -ENOSYS) {
1061 acl = ERR_PTR(-EIO);
1065 kmem_free(value, size);
1067 /* As of Linux 4.7, the kernel get_acl will set this for us */
1068 #ifndef HAVE_KERNEL_GET_ACL_HANDLE_CACHE
1070 zpl_set_cached_acl(ip, type, acl);
1076 #if !defined(HAVE_GET_ACL)
1078 __zpl_check_acl(struct inode *ip, int mask)
1080 struct posix_acl *acl;
1083 acl = zpl_get_acl(ip, ACL_TYPE_ACCESS);
1085 return (PTR_ERR(acl));
1088 error = posix_acl_permission(ip, acl, mask);
1089 zpl_posix_acl_release(acl);
1096 #if defined(HAVE_CHECK_ACL_WITH_FLAGS)
1098 zpl_check_acl(struct inode *ip, int mask, unsigned int flags)
1100 return (__zpl_check_acl(ip, mask));
1102 #elif defined(HAVE_CHECK_ACL)
1104 zpl_check_acl(struct inode *ip, int mask)
1106 return (__zpl_check_acl(ip, mask));
1108 #elif defined(HAVE_PERMISSION_WITH_NAMEIDATA)
1110 zpl_permission(struct inode *ip, int mask, struct nameidata *nd)
1112 return (generic_permission(ip, mask, __zpl_check_acl));
1114 #elif defined(HAVE_PERMISSION)
1116 zpl_permission(struct inode *ip, int mask)
1118 return (generic_permission(ip, mask, __zpl_check_acl));
1120 #endif /* HAVE_CHECK_ACL | HAVE_PERMISSION */
1121 #endif /* !HAVE_GET_ACL */
1124 zpl_init_acl(struct inode *ip, struct inode *dir)
1126 struct posix_acl *acl = NULL;
1129 if (ITOZSB(ip)->z_acl_type != ZFS_ACLTYPE_POSIXACL)
1132 if (!S_ISLNK(ip->i_mode)) {
1133 if (ITOZSB(ip)->z_acl_type == ZFS_ACLTYPE_POSIXACL) {
1134 acl = zpl_get_acl(dir, ACL_TYPE_DEFAULT);
1136 return (PTR_ERR(acl));
1140 ip->i_mode &= ~current_umask();
1141 ip->i_ctime = current_time(ip);
1142 zfs_mark_inode_dirty(ip);
1147 if ((ITOZSB(ip)->z_acl_type == ZFS_ACLTYPE_POSIXACL) && acl) {
1150 if (S_ISDIR(ip->i_mode)) {
1151 error = zpl_set_acl(ip, acl, ACL_TYPE_DEFAULT);
1157 error = __posix_acl_create(&acl, GFP_KERNEL, &mode);
1160 zfs_mark_inode_dirty(ip);
1162 error = zpl_set_acl(ip, acl, ACL_TYPE_ACCESS);
1166 zpl_posix_acl_release(acl);
1172 zpl_chmod_acl(struct inode *ip)
1174 struct posix_acl *acl;
1177 if (ITOZSB(ip)->z_acl_type != ZFS_ACLTYPE_POSIXACL)
1180 if (S_ISLNK(ip->i_mode))
1181 return (-EOPNOTSUPP);
1183 acl = zpl_get_acl(ip, ACL_TYPE_ACCESS);
1184 if (IS_ERR(acl) || !acl)
1185 return (PTR_ERR(acl));
1187 error = __posix_acl_chmod(&acl, GFP_KERNEL, ip->i_mode);
1189 error = zpl_set_acl(ip, acl, ACL_TYPE_ACCESS);
1191 zpl_posix_acl_release(acl);
1197 __zpl_xattr_acl_list_access(struct inode *ip, char *list, size_t list_size,
1198 const char *name, size_t name_len)
1200 char *xattr_name = XATTR_NAME_POSIX_ACL_ACCESS;
1201 size_t xattr_size = sizeof (XATTR_NAME_POSIX_ACL_ACCESS);
1203 if (ITOZSB(ip)->z_acl_type != ZFS_ACLTYPE_POSIXACL)
1206 if (list && xattr_size <= list_size)
1207 memcpy(list, xattr_name, xattr_size);
1209 return (xattr_size);
1211 ZPL_XATTR_LIST_WRAPPER(zpl_xattr_acl_list_access);
1214 __zpl_xattr_acl_list_default(struct inode *ip, char *list, size_t list_size,
1215 const char *name, size_t name_len)
1217 char *xattr_name = XATTR_NAME_POSIX_ACL_DEFAULT;
1218 size_t xattr_size = sizeof (XATTR_NAME_POSIX_ACL_DEFAULT);
1220 if (ITOZSB(ip)->z_acl_type != ZFS_ACLTYPE_POSIXACL)
1223 if (list && xattr_size <= list_size)
1224 memcpy(list, xattr_name, xattr_size);
1226 return (xattr_size);
1228 ZPL_XATTR_LIST_WRAPPER(zpl_xattr_acl_list_default);
1231 __zpl_xattr_acl_get_access(struct inode *ip, const char *name,
1232 void *buffer, size_t size)
1234 struct posix_acl *acl;
1235 int type = ACL_TYPE_ACCESS;
1237 /* xattr_resolve_name will do this for us if this is defined */
1238 #ifndef HAVE_XATTR_HANDLER_NAME
1239 if (strcmp(name, "") != 0)
1242 if (ITOZSB(ip)->z_acl_type != ZFS_ACLTYPE_POSIXACL)
1243 return (-EOPNOTSUPP);
1245 acl = zpl_get_acl(ip, type);
1247 return (PTR_ERR(acl));
1251 error = zpl_acl_to_xattr(acl, buffer, size);
1252 zpl_posix_acl_release(acl);
1256 ZPL_XATTR_GET_WRAPPER(zpl_xattr_acl_get_access);
1259 __zpl_xattr_acl_get_default(struct inode *ip, const char *name,
1260 void *buffer, size_t size)
1262 struct posix_acl *acl;
1263 int type = ACL_TYPE_DEFAULT;
1265 /* xattr_resolve_name will do this for us if this is defined */
1266 #ifndef HAVE_XATTR_HANDLER_NAME
1267 if (strcmp(name, "") != 0)
1270 if (ITOZSB(ip)->z_acl_type != ZFS_ACLTYPE_POSIXACL)
1271 return (-EOPNOTSUPP);
1273 acl = zpl_get_acl(ip, type);
1275 return (PTR_ERR(acl));
1279 error = zpl_acl_to_xattr(acl, buffer, size);
1280 zpl_posix_acl_release(acl);
1284 ZPL_XATTR_GET_WRAPPER(zpl_xattr_acl_get_default);
1287 __zpl_xattr_acl_set_access(struct inode *ip, const char *name,
1288 const void *value, size_t size, int flags)
1290 struct posix_acl *acl;
1291 int type = ACL_TYPE_ACCESS;
1293 /* xattr_resolve_name will do this for us if this is defined */
1294 #ifndef HAVE_XATTR_HANDLER_NAME
1295 if (strcmp(name, "") != 0)
1298 if (ITOZSB(ip)->z_acl_type != ZFS_ACLTYPE_POSIXACL)
1299 return (-EOPNOTSUPP);
1301 if (!zpl_inode_owner_or_capable(ip))
1305 acl = zpl_acl_from_xattr(value, size);
1307 return (PTR_ERR(acl));
1309 error = zpl_posix_acl_valid(ip, acl);
1311 zpl_posix_acl_release(acl);
1319 error = zpl_set_acl(ip, acl, type);
1320 zpl_posix_acl_release(acl);
1324 ZPL_XATTR_SET_WRAPPER(zpl_xattr_acl_set_access);
1327 __zpl_xattr_acl_set_default(struct inode *ip, const char *name,
1328 const void *value, size_t size, int flags)
1330 struct posix_acl *acl;
1331 int type = ACL_TYPE_DEFAULT;
1333 /* xattr_resolve_name will do this for us if this is defined */
1334 #ifndef HAVE_XATTR_HANDLER_NAME
1335 if (strcmp(name, "") != 0)
1338 if (ITOZSB(ip)->z_acl_type != ZFS_ACLTYPE_POSIXACL)
1339 return (-EOPNOTSUPP);
1341 if (!zpl_inode_owner_or_capable(ip))
1345 acl = zpl_acl_from_xattr(value, size);
1347 return (PTR_ERR(acl));
1349 error = zpl_posix_acl_valid(ip, acl);
1351 zpl_posix_acl_release(acl);
1359 error = zpl_set_acl(ip, acl, type);
1360 zpl_posix_acl_release(acl);
1364 ZPL_XATTR_SET_WRAPPER(zpl_xattr_acl_set_default);
1367 * ACL access xattr namespace handlers.
1369 * Use .name instead of .prefix when available. xattr_resolve_name will match
1370 * whole name and reject anything that has .name only as prefix.
1372 xattr_handler_t zpl_xattr_acl_access_handler =
1374 #ifdef HAVE_XATTR_HANDLER_NAME
1375 .name = XATTR_NAME_POSIX_ACL_ACCESS,
1377 .prefix = XATTR_NAME_POSIX_ACL_ACCESS,
1379 .list = zpl_xattr_acl_list_access,
1380 .get = zpl_xattr_acl_get_access,
1381 .set = zpl_xattr_acl_set_access,
1382 #if defined(HAVE_XATTR_LIST_SIMPLE) || \
1383 defined(HAVE_XATTR_LIST_DENTRY) || \
1384 defined(HAVE_XATTR_LIST_HANDLER)
1385 .flags = ACL_TYPE_ACCESS,
1390 * ACL default xattr namespace handlers.
1392 * Use .name instead of .prefix when available. xattr_resolve_name will match
1393 * whole name and reject anything that has .name only as prefix.
1395 xattr_handler_t zpl_xattr_acl_default_handler =
1397 #ifdef HAVE_XATTR_HANDLER_NAME
1398 .name = XATTR_NAME_POSIX_ACL_DEFAULT,
1400 .prefix = XATTR_NAME_POSIX_ACL_DEFAULT,
1402 .list = zpl_xattr_acl_list_default,
1403 .get = zpl_xattr_acl_get_default,
1404 .set = zpl_xattr_acl_set_default,
1405 #if defined(HAVE_XATTR_LIST_SIMPLE) || \
1406 defined(HAVE_XATTR_LIST_DENTRY) || \
1407 defined(HAVE_XATTR_LIST_HANDLER)
1408 .flags = ACL_TYPE_DEFAULT,
1412 #endif /* CONFIG_FS_POSIX_ACL */
1414 xattr_handler_t *zpl_xattr_handlers[] = {
1415 &zpl_xattr_security_handler,
1416 &zpl_xattr_trusted_handler,
1417 &zpl_xattr_user_handler,
1418 #ifdef CONFIG_FS_POSIX_ACL
1419 &zpl_xattr_acl_access_handler,
1420 &zpl_xattr_acl_default_handler,
1421 #endif /* CONFIG_FS_POSIX_ACL */
1425 static const struct xattr_handler *
1426 zpl_xattr_handler(const char *name)
1428 if (strncmp(name, XATTR_USER_PREFIX,
1429 XATTR_USER_PREFIX_LEN) == 0)
1430 return (&zpl_xattr_user_handler);
1432 if (strncmp(name, XATTR_TRUSTED_PREFIX,
1433 XATTR_TRUSTED_PREFIX_LEN) == 0)
1434 return (&zpl_xattr_trusted_handler);
1436 if (strncmp(name, XATTR_SECURITY_PREFIX,
1437 XATTR_SECURITY_PREFIX_LEN) == 0)
1438 return (&zpl_xattr_security_handler);
1440 #ifdef CONFIG_FS_POSIX_ACL
1441 if (strncmp(name, XATTR_NAME_POSIX_ACL_ACCESS,
1442 sizeof (XATTR_NAME_POSIX_ACL_ACCESS)) == 0)
1443 return (&zpl_xattr_acl_access_handler);
1445 if (strncmp(name, XATTR_NAME_POSIX_ACL_DEFAULT,
1446 sizeof (XATTR_NAME_POSIX_ACL_DEFAULT)) == 0)
1447 return (&zpl_xattr_acl_default_handler);
1448 #endif /* CONFIG_FS_POSIX_ACL */
1453 #if !defined(HAVE_POSIX_ACL_RELEASE) || defined(HAVE_POSIX_ACL_RELEASE_GPL_ONLY)
1454 struct acl_rel_struct {
1455 struct acl_rel_struct *next;
1456 struct posix_acl *acl;
1460 #define ACL_REL_GRACE (60*HZ)
1461 #define ACL_REL_WINDOW (1*HZ)
1462 #define ACL_REL_SCHED (ACL_REL_GRACE+ACL_REL_WINDOW)
1465 * Lockless multi-producer single-consumer fifo list.
1466 * Nodes are added to tail and removed from head. Tail pointer is our
1467 * synchronization point. It always points to the next pointer of the last
1468 * node, or head if list is empty.
1470 static struct acl_rel_struct *acl_rel_head = NULL;
1471 static struct acl_rel_struct **acl_rel_tail = &acl_rel_head;
1474 zpl_posix_acl_free(void *arg)
1476 struct acl_rel_struct *freelist = NULL;
1477 struct acl_rel_struct *a;
1479 boolean_t refire = B_FALSE;
1481 ASSERT3P(acl_rel_head, !=, NULL);
1482 while (acl_rel_head) {
1484 if (ddi_get_lbolt() - a->time >= ACL_REL_GRACE) {
1486 * If a is the last node we need to reset tail, but we
1487 * need to use cmpxchg to make sure it is still the
1490 if (acl_rel_tail == &a->next) {
1491 acl_rel_head = NULL;
1492 if (cmpxchg(&acl_rel_tail, &a->next,
1493 &acl_rel_head) == &a->next) {
1494 ASSERT3P(a->next, ==, NULL);
1501 * a is not last node, make sure next pointer is set
1502 * by the adder and advance the head.
1504 while (READ_ONCE(a->next) == NULL)
1506 acl_rel_head = a->next;
1511 * a is still in grace period. We are responsible to
1512 * reschedule the free task, since adder will only do
1513 * so if list is empty.
1515 new_time = a->time + ACL_REL_SCHED;
1522 taskq_dispatch_delay(system_delay_taskq, zpl_posix_acl_free,
1523 NULL, TQ_SLEEP, new_time);
1529 kmem_free(a, sizeof (struct acl_rel_struct));
1534 zpl_posix_acl_release_impl(struct posix_acl *acl)
1536 struct acl_rel_struct *a, **prev;
1538 a = kmem_alloc(sizeof (struct acl_rel_struct), KM_SLEEP);
1541 a->time = ddi_get_lbolt();
1542 /* atomically points tail to us and get the previous tail */
1543 prev = xchg(&acl_rel_tail, &a->next);
1544 ASSERT3P(*prev, ==, NULL);
1546 /* if it was empty before, schedule the free task */
1547 if (prev == &acl_rel_head)
1548 taskq_dispatch_delay(system_delay_taskq, zpl_posix_acl_free,
1549 NULL, TQ_SLEEP, ddi_get_lbolt() + ACL_REL_SCHED);