xref: /titanic_50/usr/src/uts/common/fs/zfs/zfs_log.c (revision fd9ee8b58485b20072eeef1310a88ff348d5e7fa)
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) 2005, 2010, Oracle and/or its affiliates. All rights reserved.
23  */
24 
25 #include <sys/types.h>
26 #include <sys/param.h>
27 #include <sys/systm.h>
28 #include <sys/sysmacros.h>
29 #include <sys/cmn_err.h>
30 #include <sys/kmem.h>
31 #include <sys/thread.h>
32 #include <sys/file.h>
33 #include <sys/vfs.h>
34 #include <sys/zfs_znode.h>
35 #include <sys/zfs_dir.h>
36 #include <sys/zil.h>
37 #include <sys/zil_impl.h>
38 #include <sys/byteorder.h>
39 #include <sys/policy.h>
40 #include <sys/stat.h>
41 #include <sys/mode.h>
42 #include <sys/acl.h>
43 #include <sys/dmu.h>
44 #include <sys/spa.h>
45 #include <sys/zfs_fuid.h>
46 #include <sys/ddi.h>
47 #include <sys/dsl_dataset.h>
48 
49 /*
50  * These zfs_log_* functions must be called within a dmu tx, in one
51  * of 2 contexts depending on zilog->z_replay:
52  *
53  * Non replay mode
54  * ---------------
55  * We need to record the transaction so that if it is committed to
56  * the Intent Log then it can be replayed.  An intent log transaction
57  * structure (itx_t) is allocated and all the information necessary to
58  * possibly replay the transaction is saved in it. The itx is then assigned
59  * a sequence number and inserted in the in-memory list anchored in the zilog.
60  *
61  * Replay mode
62  * -----------
63  * We need to mark the intent log record as replayed in the log header.
64  * This is done in the same transaction as the replay so that they
65  * commit atomically.
66  */
67 
68 int
69 zfs_log_create_txtype(zil_create_t type, vsecattr_t *vsecp, vattr_t *vap)
70 {
71 	int isxvattr = (vap->va_mask & AT_XVATTR);
72 	switch (type) {
73 	case Z_FILE:
74 		if (vsecp == NULL && !isxvattr)
75 			return (TX_CREATE);
76 		if (vsecp && isxvattr)
77 			return (TX_CREATE_ACL_ATTR);
78 		if (vsecp)
79 			return (TX_CREATE_ACL);
80 		else
81 			return (TX_CREATE_ATTR);
82 		/*NOTREACHED*/
83 	case Z_DIR:
84 		if (vsecp == NULL && !isxvattr)
85 			return (TX_MKDIR);
86 		if (vsecp && isxvattr)
87 			return (TX_MKDIR_ACL_ATTR);
88 		if (vsecp)
89 			return (TX_MKDIR_ACL);
90 		else
91 			return (TX_MKDIR_ATTR);
92 	case Z_XATTRDIR:
93 		return (TX_MKXATTR);
94 	}
95 	ASSERT(0);
96 	return (TX_MAX_TYPE);
97 }
98 
99 /*
100  * build up the log data necessary for logging xvattr_t
101  * First lr_attr_t is initialized.  following the lr_attr_t
102  * is the mapsize and attribute bitmap copied from the xvattr_t.
103  * Following the bitmap and bitmapsize two 64 bit words are reserved
104  * for the create time which may be set.  Following the create time
105  * records a single 64 bit integer which has the bits to set on
106  * replay for the xvattr.
107  */
108 static void
109 zfs_log_xvattr(lr_attr_t *lrattr, xvattr_t *xvap)
110 {
111 	uint32_t	*bitmap;
112 	uint64_t	*attrs;
113 	uint64_t	*crtime;
114 	xoptattr_t	*xoap;
115 	void		*scanstamp;
116 	int		i;
117 
118 	xoap = xva_getxoptattr(xvap);
119 	ASSERT(xoap);
120 
121 	lrattr->lr_attr_masksize = xvap->xva_mapsize;
122 	bitmap = &lrattr->lr_attr_bitmap;
123 	for (i = 0; i != xvap->xva_mapsize; i++, bitmap++) {
124 		*bitmap = xvap->xva_reqattrmap[i];
125 	}
126 
127 	/* Now pack the attributes up in a single uint64_t */
128 	attrs = (uint64_t *)bitmap;
129 	crtime = attrs + 1;
130 	scanstamp = (caddr_t)(crtime + 2);
131 	*attrs = 0;
132 	if (XVA_ISSET_REQ(xvap, XAT_READONLY))
133 		*attrs |= (xoap->xoa_readonly == 0) ? 0 :
134 		    XAT0_READONLY;
135 	if (XVA_ISSET_REQ(xvap, XAT_HIDDEN))
136 		*attrs |= (xoap->xoa_hidden == 0) ? 0 :
137 		    XAT0_HIDDEN;
138 	if (XVA_ISSET_REQ(xvap, XAT_SYSTEM))
139 		*attrs |= (xoap->xoa_system == 0) ? 0 :
140 		    XAT0_SYSTEM;
141 	if (XVA_ISSET_REQ(xvap, XAT_ARCHIVE))
142 		*attrs |= (xoap->xoa_archive == 0) ? 0 :
143 		    XAT0_ARCHIVE;
144 	if (XVA_ISSET_REQ(xvap, XAT_IMMUTABLE))
145 		*attrs |= (xoap->xoa_immutable == 0) ? 0 :
146 		    XAT0_IMMUTABLE;
147 	if (XVA_ISSET_REQ(xvap, XAT_NOUNLINK))
148 		*attrs |= (xoap->xoa_nounlink == 0) ? 0 :
149 		    XAT0_NOUNLINK;
150 	if (XVA_ISSET_REQ(xvap, XAT_APPENDONLY))
151 		*attrs |= (xoap->xoa_appendonly == 0) ? 0 :
152 		    XAT0_APPENDONLY;
153 	if (XVA_ISSET_REQ(xvap, XAT_OPAQUE))
154 		*attrs |= (xoap->xoa_opaque == 0) ? 0 :
155 		    XAT0_APPENDONLY;
156 	if (XVA_ISSET_REQ(xvap, XAT_NODUMP))
157 		*attrs |= (xoap->xoa_nodump == 0) ? 0 :
158 		    XAT0_NODUMP;
159 	if (XVA_ISSET_REQ(xvap, XAT_AV_QUARANTINED))
160 		*attrs |= (xoap->xoa_av_quarantined == 0) ? 0 :
161 		    XAT0_AV_QUARANTINED;
162 	if (XVA_ISSET_REQ(xvap, XAT_AV_MODIFIED))
163 		*attrs |= (xoap->xoa_av_modified == 0) ? 0 :
164 		    XAT0_AV_MODIFIED;
165 	if (XVA_ISSET_REQ(xvap, XAT_CREATETIME))
166 		ZFS_TIME_ENCODE(&xoap->xoa_createtime, crtime);
167 	if (XVA_ISSET_REQ(xvap, XAT_AV_SCANSTAMP))
168 		bcopy(xoap->xoa_av_scanstamp, scanstamp, AV_SCANSTAMP_SZ);
169 	if (XVA_ISSET_REQ(xvap, XAT_REPARSE))
170 		*attrs |= (xoap->xoa_reparse == 0) ? 0 :
171 		    XAT0_REPARSE;
172 	if (XVA_ISSET_REQ(xvap, XAT_OFFLINE))
173 		*attrs |= (xoap->xoa_offline == 0) ? 0 :
174 		    XAT0_OFFLINE;
175 	if (XVA_ISSET_REQ(xvap, XAT_SPARSE))
176 		*attrs |= (xoap->xoa_sparse == 0) ? 0 :
177 		    XAT0_SPARSE;
178 }
179 
180 static void *
181 zfs_log_fuid_ids(zfs_fuid_info_t *fuidp, void *start)
182 {
183 	zfs_fuid_t *zfuid;
184 	uint64_t *fuidloc = start;
185 
186 	/* First copy in the ACE FUIDs */
187 	for (zfuid = list_head(&fuidp->z_fuids); zfuid;
188 	    zfuid = list_next(&fuidp->z_fuids, zfuid)) {
189 		*fuidloc++ = zfuid->z_logfuid;
190 	}
191 	return (fuidloc);
192 }
193 
194 
195 static void *
196 zfs_log_fuid_domains(zfs_fuid_info_t *fuidp, void *start)
197 {
198 	zfs_fuid_domain_t *zdomain;
199 
200 	/* now copy in the domain info, if any */
201 	if (fuidp->z_domain_str_sz != 0) {
202 		for (zdomain = list_head(&fuidp->z_domains); zdomain;
203 		    zdomain = list_next(&fuidp->z_domains, zdomain)) {
204 			bcopy((void *)zdomain->z_domain, start,
205 			    strlen(zdomain->z_domain) + 1);
206 			start = (caddr_t)start +
207 			    strlen(zdomain->z_domain) + 1;
208 		}
209 	}
210 	return (start);
211 }
212 
213 /*
214  * zfs_log_create() is used to handle TX_CREATE, TX_CREATE_ATTR, TX_MKDIR,
215  * TX_MKDIR_ATTR and TX_MKXATTR
216  * transactions.
217  *
218  * TX_CREATE and TX_MKDIR are standard creates, but they may have FUID
219  * domain information appended prior to the name.  In this case the
220  * uid/gid in the log record will be a log centric FUID.
221  *
222  * TX_CREATE_ACL_ATTR and TX_MKDIR_ACL_ATTR handle special creates that
223  * may contain attributes, ACL and optional fuid information.
224  *
225  * TX_CREATE_ACL and TX_MKDIR_ACL handle special creates that specify
226  * and ACL and normal users/groups in the ACEs.
227  *
228  * There may be an optional xvattr attribute information similar
229  * to zfs_log_setattr.
230  *
231  * Also, after the file name "domain" strings may be appended.
232  */
233 void
234 zfs_log_create(zilog_t *zilog, dmu_tx_t *tx, uint64_t txtype,
235     znode_t *dzp, znode_t *zp, char *name, vsecattr_t *vsecp,
236     zfs_fuid_info_t *fuidp, vattr_t *vap)
237 {
238 	itx_t *itx;
239 	lr_create_t *lr;
240 	lr_acl_create_t *lracl;
241 	size_t aclsize;
242 	size_t xvatsize = 0;
243 	size_t txsize;
244 	xvattr_t *xvap = (xvattr_t *)vap;
245 	void *end;
246 	size_t lrsize;
247 	size_t namesize = strlen(name) + 1;
248 	size_t fuidsz = 0;
249 
250 	if (zil_replaying(zilog, tx))
251 		return;
252 
253 	/*
254 	 * If we have FUIDs present then add in space for
255 	 * domains and ACE fuid's if any.
256 	 */
257 	if (fuidp) {
258 		fuidsz += fuidp->z_domain_str_sz;
259 		fuidsz += fuidp->z_fuid_cnt * sizeof (uint64_t);
260 	}
261 
262 	if (vap->va_mask & AT_XVATTR)
263 		xvatsize = ZIL_XVAT_SIZE(xvap->xva_mapsize);
264 
265 	if ((int)txtype == TX_CREATE_ATTR || (int)txtype == TX_MKDIR_ATTR ||
266 	    (int)txtype == TX_CREATE || (int)txtype == TX_MKDIR ||
267 	    (int)txtype == TX_MKXATTR) {
268 		txsize = sizeof (*lr) + namesize + fuidsz + xvatsize;
269 		lrsize = sizeof (*lr);
270 	} else {
271 		aclsize = (vsecp) ? vsecp->vsa_aclentsz : 0;
272 		txsize =
273 		    sizeof (lr_acl_create_t) + namesize + fuidsz +
274 		    ZIL_ACE_LENGTH(aclsize) + xvatsize;
275 		lrsize = sizeof (lr_acl_create_t);
276 	}
277 
278 	itx = zil_itx_create(txtype, txsize);
279 
280 	lr = (lr_create_t *)&itx->itx_lr;
281 	lr->lr_doid = dzp->z_id;
282 	lr->lr_foid = zp->z_id;
283 	lr->lr_mode = zp->z_mode;
284 	if (!IS_EPHEMERAL(zp->z_uid)) {
285 		lr->lr_uid = (uint64_t)zp->z_uid;
286 	} else {
287 		lr->lr_uid = fuidp->z_fuid_owner;
288 	}
289 	if (!IS_EPHEMERAL(zp->z_gid)) {
290 		lr->lr_gid = (uint64_t)zp->z_gid;
291 	} else {
292 		lr->lr_gid = fuidp->z_fuid_group;
293 	}
294 	(void) sa_lookup(zp->z_sa_hdl, SA_ZPL_GEN(zp->z_zfsvfs), &lr->lr_gen,
295 	    sizeof (uint64_t));
296 	(void) sa_lookup(zp->z_sa_hdl, SA_ZPL_CRTIME(zp->z_zfsvfs),
297 	    lr->lr_crtime, sizeof (uint64_t) * 2);
298 
299 	if (sa_lookup(zp->z_sa_hdl, SA_ZPL_RDEV(zp->z_zfsvfs), &lr->lr_rdev,
300 	    sizeof (lr->lr_rdev)) != 0)
301 		lr->lr_rdev = 0;
302 
303 	/*
304 	 * Fill in xvattr info if any
305 	 */
306 	if (vap->va_mask & AT_XVATTR) {
307 		zfs_log_xvattr((lr_attr_t *)((caddr_t)lr + lrsize), xvap);
308 		end = (caddr_t)lr + lrsize + xvatsize;
309 	} else {
310 		end = (caddr_t)lr + lrsize;
311 	}
312 
313 	/* Now fill in any ACL info */
314 
315 	if (vsecp) {
316 		lracl = (lr_acl_create_t *)&itx->itx_lr;
317 		lracl->lr_aclcnt = vsecp->vsa_aclcnt;
318 		lracl->lr_acl_bytes = aclsize;
319 		lracl->lr_domcnt = fuidp ? fuidp->z_domain_cnt : 0;
320 		lracl->lr_fuidcnt  = fuidp ? fuidp->z_fuid_cnt : 0;
321 		if (vsecp->vsa_aclflags & VSA_ACE_ACLFLAGS)
322 			lracl->lr_acl_flags = (uint64_t)vsecp->vsa_aclflags;
323 		else
324 			lracl->lr_acl_flags = 0;
325 
326 		bcopy(vsecp->vsa_aclentp, end, aclsize);
327 		end = (caddr_t)end + ZIL_ACE_LENGTH(aclsize);
328 	}
329 
330 	/* drop in FUID info */
331 	if (fuidp) {
332 		end = zfs_log_fuid_ids(fuidp, end);
333 		end = zfs_log_fuid_domains(fuidp, end);
334 	}
335 	/*
336 	 * Now place file name in log record
337 	 */
338 	bcopy(name, end, namesize);
339 
340 	zil_itx_assign(zilog, itx, tx);
341 }
342 
343 /*
344  * zfs_log_remove() handles both TX_REMOVE and TX_RMDIR transactions.
345  */
346 void
347 zfs_log_remove(zilog_t *zilog, dmu_tx_t *tx, uint64_t txtype,
348 	znode_t *dzp, char *name, uint64_t foid)
349 {
350 	itx_t *itx;
351 	lr_remove_t *lr;
352 	size_t namesize = strlen(name) + 1;
353 
354 	if (zil_replaying(zilog, tx))
355 		return;
356 
357 	itx = zil_itx_create(txtype, sizeof (*lr) + namesize);
358 	lr = (lr_remove_t *)&itx->itx_lr;
359 	lr->lr_doid = dzp->z_id;
360 	bcopy(name, (char *)(lr + 1), namesize);
361 
362 	itx->itx_oid = foid;
363 
364 	zil_itx_assign(zilog, itx, tx);
365 }
366 
367 /*
368  * zfs_log_link() handles TX_LINK transactions.
369  */
370 void
371 zfs_log_link(zilog_t *zilog, dmu_tx_t *tx, uint64_t txtype,
372 	znode_t *dzp, znode_t *zp, char *name)
373 {
374 	itx_t *itx;
375 	lr_link_t *lr;
376 	size_t namesize = strlen(name) + 1;
377 
378 	if (zil_replaying(zilog, tx))
379 		return;
380 
381 	itx = zil_itx_create(txtype, sizeof (*lr) + namesize);
382 	lr = (lr_link_t *)&itx->itx_lr;
383 	lr->lr_doid = dzp->z_id;
384 	lr->lr_link_obj = zp->z_id;
385 	bcopy(name, (char *)(lr + 1), namesize);
386 
387 	zil_itx_assign(zilog, itx, tx);
388 }
389 
390 /*
391  * zfs_log_symlink() handles TX_SYMLINK transactions.
392  */
393 void
394 zfs_log_symlink(zilog_t *zilog, dmu_tx_t *tx, uint64_t txtype,
395     znode_t *dzp, znode_t *zp, char *name, char *link)
396 {
397 	itx_t *itx;
398 	lr_create_t *lr;
399 	size_t namesize = strlen(name) + 1;
400 	size_t linksize = strlen(link) + 1;
401 
402 	if (zil_replaying(zilog, tx))
403 		return;
404 
405 	itx = zil_itx_create(txtype, sizeof (*lr) + namesize + linksize);
406 	lr = (lr_create_t *)&itx->itx_lr;
407 	lr->lr_doid = dzp->z_id;
408 	lr->lr_foid = zp->z_id;
409 	lr->lr_uid = zp->z_uid;
410 	lr->lr_gid = zp->z_gid;
411 	lr->lr_mode = zp->z_mode;
412 	(void) sa_lookup(zp->z_sa_hdl, SA_ZPL_GEN(zp->z_zfsvfs), &lr->lr_gen,
413 	    sizeof (uint64_t));
414 	(void) sa_lookup(zp->z_sa_hdl, SA_ZPL_CRTIME(zp->z_zfsvfs),
415 	    lr->lr_crtime, sizeof (uint64_t) * 2);
416 	bcopy(name, (char *)(lr + 1), namesize);
417 	bcopy(link, (char *)(lr + 1) + namesize, linksize);
418 
419 	zil_itx_assign(zilog, itx, tx);
420 }
421 
422 /*
423  * zfs_log_rename() handles TX_RENAME transactions.
424  */
425 void
426 zfs_log_rename(zilog_t *zilog, dmu_tx_t *tx, uint64_t txtype,
427 	znode_t *sdzp, char *sname, znode_t *tdzp, char *dname, znode_t *szp)
428 {
429 	itx_t *itx;
430 	lr_rename_t *lr;
431 	size_t snamesize = strlen(sname) + 1;
432 	size_t dnamesize = strlen(dname) + 1;
433 
434 	if (zil_replaying(zilog, tx))
435 		return;
436 
437 	itx = zil_itx_create(txtype, sizeof (*lr) + snamesize + dnamesize);
438 	lr = (lr_rename_t *)&itx->itx_lr;
439 	lr->lr_sdoid = sdzp->z_id;
440 	lr->lr_tdoid = tdzp->z_id;
441 	bcopy(sname, (char *)(lr + 1), snamesize);
442 	bcopy(dname, (char *)(lr + 1) + snamesize, dnamesize);
443 	itx->itx_oid = szp->z_id;
444 
445 	zil_itx_assign(zilog, itx, tx);
446 }
447 
448 /*
449  * zfs_log_write() handles TX_WRITE transactions.
450  */
451 ssize_t zfs_immediate_write_sz = 32768;
452 
453 void
454 zfs_log_write(zilog_t *zilog, dmu_tx_t *tx, int txtype,
455 	znode_t *zp, offset_t off, ssize_t resid, int ioflag)
456 {
457 	itx_wr_state_t write_state;
458 	boolean_t slogging;
459 	uintptr_t fsync_cnt;
460 	ssize_t immediate_write_sz;
461 
462 	if (zil_replaying(zilog, tx) || zp->z_unlinked)
463 		return;
464 
465 	immediate_write_sz = (zilog->zl_logbias == ZFS_LOGBIAS_THROUGHPUT)
466 	    ? 0 : zfs_immediate_write_sz;
467 
468 	slogging = spa_has_slogs(zilog->zl_spa) &&
469 	    (zilog->zl_logbias == ZFS_LOGBIAS_LATENCY);
470 	if (resid > immediate_write_sz && !slogging && resid <= zp->z_blksz)
471 		write_state = WR_INDIRECT;
472 	else if (ioflag & (FSYNC | FDSYNC))
473 		write_state = WR_COPIED;
474 	else
475 		write_state = WR_NEED_COPY;
476 
477 	if ((fsync_cnt = (uintptr_t)tsd_get(zfs_fsyncer_key)) != 0) {
478 		(void) tsd_set(zfs_fsyncer_key, (void *)(fsync_cnt - 1));
479 	}
480 
481 	while (resid) {
482 		itx_t *itx;
483 		lr_write_t *lr;
484 		ssize_t len;
485 
486 		/*
487 		 * If the write would overflow the largest block then split it.
488 		 */
489 		if (write_state != WR_INDIRECT && resid > ZIL_MAX_LOG_DATA)
490 			len = SPA_MAXBLOCKSIZE >> 1;
491 		else
492 			len = resid;
493 
494 		itx = zil_itx_create(txtype, sizeof (*lr) +
495 		    (write_state == WR_COPIED ? len : 0));
496 		lr = (lr_write_t *)&itx->itx_lr;
497 		if (write_state == WR_COPIED && dmu_read(zp->z_zfsvfs->z_os,
498 		    zp->z_id, off, len, lr + 1, DMU_READ_NO_PREFETCH) != 0) {
499 			zil_itx_destroy(itx);
500 			itx = zil_itx_create(txtype, sizeof (*lr));
501 			lr = (lr_write_t *)&itx->itx_lr;
502 			write_state = WR_NEED_COPY;
503 		}
504 
505 		itx->itx_wr_state = write_state;
506 		if (write_state == WR_NEED_COPY)
507 			itx->itx_sod += len;
508 		lr->lr_foid = zp->z_id;
509 		lr->lr_offset = off;
510 		lr->lr_length = len;
511 		lr->lr_blkoff = 0;
512 		BP_ZERO(&lr->lr_blkptr);
513 
514 		itx->itx_private = zp->z_zfsvfs;
515 
516 		if (!(ioflag & (FSYNC | FDSYNC)) && (zp->z_sync_cnt == 0) &&
517 		    (fsync_cnt == 0))
518 			itx->itx_sync = B_FALSE;
519 
520 		zil_itx_assign(zilog, itx, tx);
521 
522 		off += len;
523 		resid -= len;
524 	}
525 }
526 
527 /*
528  * zfs_log_truncate() handles TX_TRUNCATE transactions.
529  */
530 void
531 zfs_log_truncate(zilog_t *zilog, dmu_tx_t *tx, int txtype,
532 	znode_t *zp, uint64_t off, uint64_t len)
533 {
534 	itx_t *itx;
535 	lr_truncate_t *lr;
536 
537 	if (zil_replaying(zilog, tx) || zp->z_unlinked)
538 		return;
539 
540 	itx = zil_itx_create(txtype, sizeof (*lr));
541 	lr = (lr_truncate_t *)&itx->itx_lr;
542 	lr->lr_foid = zp->z_id;
543 	lr->lr_offset = off;
544 	lr->lr_length = len;
545 
546 	itx->itx_sync = (zp->z_sync_cnt != 0);
547 	zil_itx_assign(zilog, itx, tx);
548 }
549 
550 /*
551  * zfs_log_setattr() handles TX_SETATTR transactions.
552  */
553 void
554 zfs_log_setattr(zilog_t *zilog, dmu_tx_t *tx, int txtype,
555 	znode_t *zp, vattr_t *vap, uint_t mask_applied, zfs_fuid_info_t *fuidp)
556 {
557 	itx_t		*itx;
558 	lr_setattr_t	*lr;
559 	xvattr_t	*xvap = (xvattr_t *)vap;
560 	size_t		recsize = sizeof (lr_setattr_t);
561 	void		*start;
562 
563 	if (zil_replaying(zilog, tx) || zp->z_unlinked)
564 		return;
565 
566 	/*
567 	 * If XVATTR set, then log record size needs to allow
568 	 * for lr_attr_t + xvattr mask, mapsize and create time
569 	 * plus actual attribute values
570 	 */
571 	if (vap->va_mask & AT_XVATTR)
572 		recsize = sizeof (*lr) + ZIL_XVAT_SIZE(xvap->xva_mapsize);
573 
574 	if (fuidp)
575 		recsize += fuidp->z_domain_str_sz;
576 
577 	itx = zil_itx_create(txtype, recsize);
578 	lr = (lr_setattr_t *)&itx->itx_lr;
579 	lr->lr_foid = zp->z_id;
580 	lr->lr_mask = (uint64_t)mask_applied;
581 	lr->lr_mode = (uint64_t)vap->va_mode;
582 	if ((mask_applied & AT_UID) && IS_EPHEMERAL(vap->va_uid))
583 		lr->lr_uid = fuidp->z_fuid_owner;
584 	else
585 		lr->lr_uid = (uint64_t)vap->va_uid;
586 
587 	if ((mask_applied & AT_GID) && IS_EPHEMERAL(vap->va_gid))
588 		lr->lr_gid = fuidp->z_fuid_group;
589 	else
590 		lr->lr_gid = (uint64_t)vap->va_gid;
591 
592 	lr->lr_size = (uint64_t)vap->va_size;
593 	ZFS_TIME_ENCODE(&vap->va_atime, lr->lr_atime);
594 	ZFS_TIME_ENCODE(&vap->va_mtime, lr->lr_mtime);
595 	start = (lr_setattr_t *)(lr + 1);
596 	if (vap->va_mask & AT_XVATTR) {
597 		zfs_log_xvattr((lr_attr_t *)start, xvap);
598 		start = (caddr_t)start + ZIL_XVAT_SIZE(xvap->xva_mapsize);
599 	}
600 
601 	/*
602 	 * Now stick on domain information if any on end
603 	 */
604 
605 	if (fuidp)
606 		(void) zfs_log_fuid_domains(fuidp, start);
607 
608 	itx->itx_sync = (zp->z_sync_cnt != 0);
609 	zil_itx_assign(zilog, itx, tx);
610 }
611 
612 /*
613  * zfs_log_acl() handles TX_ACL transactions.
614  */
615 void
616 zfs_log_acl(zilog_t *zilog, dmu_tx_t *tx, znode_t *zp,
617     vsecattr_t *vsecp, zfs_fuid_info_t *fuidp)
618 {
619 	itx_t *itx;
620 	lr_acl_v0_t *lrv0;
621 	lr_acl_t *lr;
622 	int txtype;
623 	int lrsize;
624 	size_t txsize;
625 	size_t aclbytes = vsecp->vsa_aclentsz;
626 
627 	if (zil_replaying(zilog, tx) || zp->z_unlinked)
628 		return;
629 
630 	txtype = (zp->z_zfsvfs->z_version < ZPL_VERSION_FUID) ?
631 	    TX_ACL_V0 : TX_ACL;
632 
633 	if (txtype == TX_ACL)
634 		lrsize = sizeof (*lr);
635 	else
636 		lrsize = sizeof (*lrv0);
637 
638 	txsize = lrsize +
639 	    ((txtype == TX_ACL) ? ZIL_ACE_LENGTH(aclbytes) : aclbytes) +
640 	    (fuidp ? fuidp->z_domain_str_sz : 0) +
641 	    sizeof (uint64_t) * (fuidp ? fuidp->z_fuid_cnt : 0);
642 
643 	itx = zil_itx_create(txtype, txsize);
644 
645 	lr = (lr_acl_t *)&itx->itx_lr;
646 	lr->lr_foid = zp->z_id;
647 	if (txtype == TX_ACL) {
648 		lr->lr_acl_bytes = aclbytes;
649 		lr->lr_domcnt = fuidp ? fuidp->z_domain_cnt : 0;
650 		lr->lr_fuidcnt = fuidp ? fuidp->z_fuid_cnt : 0;
651 		if (vsecp->vsa_mask & VSA_ACE_ACLFLAGS)
652 			lr->lr_acl_flags = (uint64_t)vsecp->vsa_aclflags;
653 		else
654 			lr->lr_acl_flags = 0;
655 	}
656 	lr->lr_aclcnt = (uint64_t)vsecp->vsa_aclcnt;
657 
658 	if (txtype == TX_ACL_V0) {
659 		lrv0 = (lr_acl_v0_t *)lr;
660 		bcopy(vsecp->vsa_aclentp, (ace_t *)(lrv0 + 1), aclbytes);
661 	} else {
662 		void *start = (ace_t *)(lr + 1);
663 
664 		bcopy(vsecp->vsa_aclentp, start, aclbytes);
665 
666 		start = (caddr_t)start + ZIL_ACE_LENGTH(aclbytes);
667 
668 		if (fuidp) {
669 			start = zfs_log_fuid_ids(fuidp, start);
670 			(void) zfs_log_fuid_domains(fuidp, start);
671 		}
672 	}
673 
674 	itx->itx_sync = (zp->z_sync_cnt != 0);
675 	zil_itx_assign(zilog, itx, tx);
676 }
677