1/*
2 * CDDL HEADER START
3 *
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.
7 *
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.
12 *
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]
18 *
19 * CDDL HEADER END
20 */
21/*
22 * Copyright (c) 2010, Oracle and/or its affiliates. All rights reserved.
23 */
24
25#include <sys/zfs_context.h>
26#include <sys/vnode.h>
27#include <sys/sa.h>
28#include <sys/zfs_acl.h>
29#include <sys/zfs_sa.h>
30
31/*
32 * ZPL attribute registration table.
33 * Order of attributes doesn't matter
34 * a unique value will be assigned for each
35 * attribute that is file system specific
36 *
37 * This is just the set of ZPL attributes that this
38 * version of ZFS deals with natively.  The file system
39 * could have other attributes stored in files, but they will be
40 * ignored.  The SA framework will preserve them, just that
41 * this version of ZFS won't change or delete them.
42 */
43
44sa_attr_reg_t zfs_attr_table[ZPL_END+1] = {
45	{"ZPL_ATIME", sizeof (uint64_t) * 2, SA_UINT64_ARRAY, 0},
46	{"ZPL_MTIME", sizeof (uint64_t) * 2, SA_UINT64_ARRAY, 1},
47	{"ZPL_CTIME", sizeof (uint64_t) * 2, SA_UINT64_ARRAY, 2},
48	{"ZPL_CRTIME", sizeof (uint64_t) * 2, SA_UINT64_ARRAY, 3},
49	{"ZPL_GEN", sizeof (uint64_t), SA_UINT64_ARRAY, 4},
50	{"ZPL_MODE", sizeof (uint64_t), SA_UINT64_ARRAY, 5},
51	{"ZPL_SIZE", sizeof (uint64_t), SA_UINT64_ARRAY, 6},
52	{"ZPL_PARENT", sizeof (uint64_t), SA_UINT64_ARRAY, 7},
53	{"ZPL_LINKS", sizeof (uint64_t), SA_UINT64_ARRAY, 8},
54	{"ZPL_XATTR", sizeof (uint64_t), SA_UINT64_ARRAY, 9},
55	{"ZPL_RDEV", sizeof (uint64_t), SA_UINT64_ARRAY, 10},
56	{"ZPL_FLAGS", sizeof (uint64_t), SA_UINT64_ARRAY, 11},
57	{"ZPL_UID", sizeof (uint64_t), SA_UINT64_ARRAY, 12},
58	{"ZPL_GID", sizeof (uint64_t), SA_UINT64_ARRAY, 13},
59	{"ZPL_PAD", sizeof (uint64_t) * 4, SA_UINT64_ARRAY, 14},
60	{"ZPL_ZNODE_ACL", 88, SA_UINT8_ARRAY, 15},
61	{"ZPL_DACL_COUNT", sizeof (uint64_t), SA_UINT64_ARRAY, 0},
62	{"ZPL_SYMLINK", 0, SA_UINT8_ARRAY, 0},
63	{"ZPL_SCANSTAMP", 32, SA_UINT8_ARRAY, 0},
64	{"ZPL_DACL_ACES", 0, SA_ACL, 0},
65	{NULL, 0, 0, 0}
66};
67
68#ifdef _KERNEL
69
70int
71zfs_sa_readlink(znode_t *zp, uio_t *uio)
72{
73	dmu_buf_t *db = sa_get_db(zp->z_sa_hdl);
74	size_t bufsz;
75	int error;
76
77	bufsz = zp->z_size;
78	if (bufsz + ZFS_OLD_ZNODE_PHYS_SIZE <= db->db_size) {
79		error = uiomove((caddr_t)db->db_data +
80		    ZFS_OLD_ZNODE_PHYS_SIZE,
81		    MIN((size_t)bufsz, uio->uio_resid), UIO_READ, uio);
82	} else {
83		dmu_buf_t *dbp;
84		if ((error = dmu_buf_hold(zp->z_zfsvfs->z_os, zp->z_id,
85		    0, FTAG, &dbp, DMU_READ_NO_PREFETCH)) == 0) {
86			error = uiomove(dbp->db_data,
87			    MIN((size_t)bufsz, uio->uio_resid), UIO_READ, uio);
88			dmu_buf_rele(dbp, FTAG);
89		}
90	}
91	return (error);
92}
93
94void
95zfs_sa_symlink(znode_t *zp, char *link, int len, dmu_tx_t *tx)
96{
97	dmu_buf_t *db = sa_get_db(zp->z_sa_hdl);
98
99	if (ZFS_OLD_ZNODE_PHYS_SIZE + len <= dmu_bonus_max()) {
100		VERIFY(dmu_set_bonus(db,
101		    len + ZFS_OLD_ZNODE_PHYS_SIZE, tx) == 0);
102		if (len) {
103			bcopy(link, (caddr_t)db->db_data +
104			    ZFS_OLD_ZNODE_PHYS_SIZE, len);
105		}
106	} else {
107		dmu_buf_t *dbp;
108
109		zfs_grow_blocksize(zp, len, tx);
110		VERIFY(0 == dmu_buf_hold(zp->z_zfsvfs->z_os,
111		    zp->z_id, 0, FTAG, &dbp, DMU_READ_NO_PREFETCH));
112
113		dmu_buf_will_dirty(dbp, tx);
114
115		ASSERT3U(len, <=, dbp->db_size);
116		bcopy(link, dbp->db_data, len);
117		dmu_buf_rele(dbp, FTAG);
118	}
119}
120
121void
122zfs_sa_get_scanstamp(znode_t *zp, xvattr_t *xvap)
123{
124	zfsvfs_t *zfsvfs = zp->z_zfsvfs;
125	xoptattr_t *xoap;
126
127	ASSERT_VOP_LOCKED(ZTOV(zp), __func__);
128	VERIFY((xoap = xva_getxoptattr(xvap)) != NULL);
129	if (zp->z_is_sa) {
130		if (sa_lookup(zp->z_sa_hdl, SA_ZPL_SCANSTAMP(zfsvfs),
131		    &xoap->xoa_av_scanstamp,
132		    sizeof (xoap->xoa_av_scanstamp)) != 0)
133			return;
134	} else {
135		dmu_object_info_t doi;
136		dmu_buf_t *db = sa_get_db(zp->z_sa_hdl);
137		int len;
138
139		if (!(zp->z_pflags & ZFS_BONUS_SCANSTAMP))
140			return;
141
142		sa_object_info(zp->z_sa_hdl, &doi);
143		len = sizeof (xoap->xoa_av_scanstamp) +
144		    ZFS_OLD_ZNODE_PHYS_SIZE;
145
146		if (len <= doi.doi_bonus_size) {
147			(void) memcpy(xoap->xoa_av_scanstamp,
148			    (caddr_t)db->db_data + ZFS_OLD_ZNODE_PHYS_SIZE,
149			    sizeof (xoap->xoa_av_scanstamp));
150		}
151	}
152	XVA_SET_RTN(xvap, XAT_AV_SCANSTAMP);
153}
154
155void
156zfs_sa_set_scanstamp(znode_t *zp, xvattr_t *xvap, dmu_tx_t *tx)
157{
158	zfsvfs_t *zfsvfs = zp->z_zfsvfs;
159	xoptattr_t *xoap;
160
161	ASSERT_VOP_ELOCKED(ZTOV(zp), __func__);
162	VERIFY((xoap = xva_getxoptattr(xvap)) != NULL);
163	if (zp->z_is_sa)
164		VERIFY(0 == sa_update(zp->z_sa_hdl, SA_ZPL_SCANSTAMP(zfsvfs),
165		    &xoap->xoa_av_scanstamp,
166		    sizeof (xoap->xoa_av_scanstamp), tx));
167	else {
168		dmu_object_info_t doi;
169		dmu_buf_t *db = sa_get_db(zp->z_sa_hdl);
170		int len;
171
172		sa_object_info(zp->z_sa_hdl, &doi);
173		len = sizeof (xoap->xoa_av_scanstamp) +
174		    ZFS_OLD_ZNODE_PHYS_SIZE;
175		if (len > doi.doi_bonus_size)
176			VERIFY(dmu_set_bonus(db, len, tx) == 0);
177		(void) memcpy((caddr_t)db->db_data + ZFS_OLD_ZNODE_PHYS_SIZE,
178		    xoap->xoa_av_scanstamp, sizeof (xoap->xoa_av_scanstamp));
179
180		zp->z_pflags |= ZFS_BONUS_SCANSTAMP;
181		VERIFY(0 == sa_update(zp->z_sa_hdl, SA_ZPL_FLAGS(zfsvfs),
182		    &zp->z_pflags, sizeof (uint64_t), tx));
183	}
184}
185
186/*
187 * I'm not convinced we should do any of this upgrade.
188 * since the SA code can read both old/new znode formats
189 * with probably little to no performance difference.
190 *
191 * All new files will be created with the new format.
192 */
193
194void
195zfs_sa_upgrade(sa_handle_t *hdl, dmu_tx_t *tx)
196{
197	dmu_buf_t *db = sa_get_db(hdl);
198	znode_t *zp = sa_get_userdata(hdl);
199	zfsvfs_t *zfsvfs = zp->z_zfsvfs;
200	sa_bulk_attr_t bulk[20];
201	int count = 0;
202	sa_bulk_attr_t sa_attrs[20] = { 0 };
203	zfs_acl_locator_cb_t locate = { 0 };
204	uint64_t uid, gid, mode, rdev, xattr, parent;
205	uint64_t crtime[2], mtime[2], ctime[2];
206	zfs_acl_phys_t znode_acl;
207	char scanstamp[AV_SCANSTAMP_SZ];
208
209	/*
210	 * No upgrade if ACL isn't cached
211	 * since we won't know which locks are held
212	 * and ready the ACL would require special "locked"
213	 * interfaces that would be messy
214	 */
215	if (zp->z_acl_cached == NULL || ZTOV(zp)->v_type == VLNK)
216		return;
217
218	/*
219	 * If the vnode lock is held and we aren't the owner
220	 * then just return since we don't want to deadlock
221	 * trying to update the status of z_is_sa.  This
222	 * file can then be upgraded at a later time.
223	 *
224	 * Otherwise, we know we are doing the
225	 * sa_update() that caused us to enter this function.
226	 */
227	if (vn_lock(ZTOV(zp), LK_EXCLUSIVE | LK_NOWAIT) != 0)
228			return;
229
230	/* First do a bulk query of the attributes that aren't cached */
231	SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_MTIME(zfsvfs), NULL, &mtime, 16);
232	SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_CTIME(zfsvfs), NULL, &ctime, 16);
233	SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_CRTIME(zfsvfs), NULL, &crtime, 16);
234	SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_MODE(zfsvfs), NULL, &mode, 8);
235	SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_PARENT(zfsvfs), NULL, &parent, 8);
236	SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_XATTR(zfsvfs), NULL, &xattr, 8);
237	SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_RDEV(zfsvfs), NULL, &rdev, 8);
238	SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_UID(zfsvfs), NULL, &uid, 8);
239	SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_GID(zfsvfs), NULL, &gid, 8);
240	SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_ZNODE_ACL(zfsvfs), NULL,
241	    &znode_acl, 88);
242
243	if (sa_bulk_lookup_locked(hdl, bulk, count) != 0)
244		goto done;
245
246
247	/*
248	 * While the order here doesn't matter its best to try and organize
249	 * it is such a way to pick up an already existing layout number
250	 */
251	count = 0;
252	SA_ADD_BULK_ATTR(sa_attrs, count, SA_ZPL_MODE(zfsvfs), NULL, &mode, 8);
253	SA_ADD_BULK_ATTR(sa_attrs, count, SA_ZPL_SIZE(zfsvfs), NULL,
254	    &zp->z_size, 8);
255	SA_ADD_BULK_ATTR(sa_attrs, count, SA_ZPL_GEN(zfsvfs),
256	    NULL, &zp->z_gen, 8);
257	SA_ADD_BULK_ATTR(sa_attrs, count, SA_ZPL_UID(zfsvfs), NULL, &uid, 8);
258	SA_ADD_BULK_ATTR(sa_attrs, count, SA_ZPL_GID(zfsvfs), NULL, &gid, 8);
259	SA_ADD_BULK_ATTR(sa_attrs, count, SA_ZPL_PARENT(zfsvfs),
260	    NULL, &parent, 8);
261	SA_ADD_BULK_ATTR(sa_attrs, count, SA_ZPL_FLAGS(zfsvfs), NULL,
262	    &zp->z_pflags, 8);
263	SA_ADD_BULK_ATTR(sa_attrs, count, SA_ZPL_ATIME(zfsvfs), NULL,
264	    zp->z_atime, 16);
265	SA_ADD_BULK_ATTR(sa_attrs, count, SA_ZPL_MTIME(zfsvfs), NULL,
266	    &mtime, 16);
267	SA_ADD_BULK_ATTR(sa_attrs, count, SA_ZPL_CTIME(zfsvfs), NULL,
268	    &ctime, 16);
269	SA_ADD_BULK_ATTR(sa_attrs, count, SA_ZPL_CRTIME(zfsvfs), NULL,
270	    &crtime, 16);
271	SA_ADD_BULK_ATTR(sa_attrs, count, SA_ZPL_LINKS(zfsvfs), NULL,
272	    &zp->z_links, 8);
273	if (zp->z_vnode->v_type == VBLK || zp->z_vnode->v_type == VCHR)
274		SA_ADD_BULK_ATTR(sa_attrs, count, SA_ZPL_RDEV(zfsvfs), NULL,
275		    &rdev, 8);
276	SA_ADD_BULK_ATTR(sa_attrs, count, SA_ZPL_DACL_COUNT(zfsvfs), NULL,
277	    &zp->z_acl_cached->z_acl_count, 8);
278
279	if (zp->z_acl_cached->z_version < ZFS_ACL_VERSION_FUID)
280		zfs_acl_xform(zp, zp->z_acl_cached, CRED());
281
282	locate.cb_aclp = zp->z_acl_cached;
283	SA_ADD_BULK_ATTR(sa_attrs, count, SA_ZPL_DACL_ACES(zfsvfs),
284	    zfs_acl_data_locator, &locate, zp->z_acl_cached->z_acl_bytes);
285
286	if (xattr)
287		SA_ADD_BULK_ATTR(sa_attrs, count, SA_ZPL_XATTR(zfsvfs),
288		    NULL, &xattr, 8);
289
290	/* if scanstamp then add scanstamp */
291
292	if (zp->z_pflags & ZFS_BONUS_SCANSTAMP) {
293		bcopy((caddr_t)db->db_data + ZFS_OLD_ZNODE_PHYS_SIZE,
294		    scanstamp, AV_SCANSTAMP_SZ);
295		SA_ADD_BULK_ATTR(sa_attrs, count, SA_ZPL_SCANSTAMP(zfsvfs),
296		    NULL, scanstamp, AV_SCANSTAMP_SZ);
297		zp->z_pflags &= ~ZFS_BONUS_SCANSTAMP;
298	}
299
300	VERIFY(dmu_set_bonustype(db, DMU_OT_SA, tx) == 0);
301	VERIFY(sa_replace_all_by_template_locked(hdl, sa_attrs,
302	    count, tx) == 0);
303	if (znode_acl.z_acl_extern_obj)
304		VERIFY(0 == dmu_object_free(zfsvfs->z_os,
305		    znode_acl.z_acl_extern_obj, tx));
306
307	zp->z_is_sa = B_TRUE;
308done:
309	VOP_UNLOCK(ZTOV(zp), 0);
310}
311
312void
313zfs_sa_upgrade_txholds(dmu_tx_t *tx, znode_t *zp)
314{
315	if (!zp->z_zfsvfs->z_use_sa || zp->z_is_sa)
316		return;
317
318
319	dmu_tx_hold_sa(tx, zp->z_sa_hdl, B_TRUE);
320
321	if (zfs_external_acl(zp)) {
322		dmu_tx_hold_free(tx, zfs_external_acl(zp), 0,
323		    DMU_OBJECT_END);
324	}
325}
326
327#endif
328