xref: /illumos-gate/usr/src/uts/common/fs/zfs/zfs_log.c (revision 134a1f4e)
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 2010 Sun Microsystems, Inc.  All rights reserved.
23  * Use is subject to license terms.
24  */
25 
26 #include <sys/types.h>
27 #include <sys/param.h>
28 #include <sys/systm.h>
29 #include <sys/sysmacros.h>
30 #include <sys/cmn_err.h>
31 #include <sys/kmem.h>
32 #include <sys/thread.h>
33 #include <sys/file.h>
34 #include <sys/vfs.h>
35 #include <sys/zfs_znode.h>
36 #include <sys/zfs_dir.h>
37 #include <sys/zil.h>
38 #include <sys/zil_impl.h>
39 #include <sys/byteorder.h>
40 #include <sys/policy.h>
41 #include <sys/stat.h>
42 #include <sys/mode.h>
43 #include <sys/acl.h>
44 #include <sys/dmu.h>
45 #include <sys/spa.h>
46 #include <sys/zfs_fuid.h>
47 #include <sys/ddi.h>
48 #include <sys/dsl_dataset.h>
49 
50 /*
51  * These zfs_log_* functions must be called within a dmu tx, in one
52  * of 2 contexts depending on zilog->z_replay:
53  *
54  * Non replay mode
55  * ---------------
56  * We need to record the transaction so that if it is committed to
57  * the Intent Log then it can be replayed.  An intent log transaction
58  * structure (itx_t) is allocated and all the information necessary to
59  * possibly replay the transaction is saved in it. The itx is then assigned
60  * a sequence number and inserted in the in-memory list anchored in the zilog.
61  *
62  * Replay mode
63  * -----------
64  * We need to mark the intent log record as replayed in the log header.
65  * This is done in the same transaction as the replay so that they
66  * commit atomically.
67  */
68 
69 int
70 zfs_log_create_txtype(zil_create_t type, vsecattr_t *vsecp, vattr_t *vap)
71 {
72 	int isxvattr = (vap->va_mask & AT_XVATTR);
73 	switch (type) {
74 	case Z_FILE:
75 		if (vsecp == NULL && !isxvattr)
76 			return (TX_CREATE);
77 		if (vsecp && isxvattr)
78 			return (TX_CREATE_ACL_ATTR);
79 		if (vsecp)
80 			return (TX_CREATE_ACL);
81 		else
82 			return (TX_CREATE_ATTR);
83 		/*NOTREACHED*/
84 	case Z_DIR:
85 		if (vsecp == NULL && !isxvattr)
86 			return (TX_MKDIR);
87 		if (vsecp && isxvattr)
88 			return (TX_MKDIR_ACL_ATTR);
89 		if (vsecp)
90 			return (TX_MKDIR_ACL);
91 		else
92 			return (TX_MKDIR_ATTR);
93 	case Z_XATTRDIR:
94 		return (TX_MKXATTR);
95 	}
96 	ASSERT(0);
97 	return (TX_MAX_TYPE);
98 }
99 
100 /*
101  * build up the log data necessary for logging xvattr_t
102  * First lr_attr_t is initialized.  following the lr_attr_t
103  * is the mapsize and attribute bitmap copied from the xvattr_t.
104  * Following the bitmap and bitmapsize two 64 bit words are reserved
105  * for the create time which may be set.  Following the create time
106  * records a single 64 bit integer which has the bits to set on
107  * replay for the xvattr.
108  */
109 static void
110 zfs_log_xvattr(lr_attr_t *lrattr, xvattr_t *xvap)
111 {
112 	uint32_t	*bitmap;
113 	uint64_t	*attrs;
114 	uint64_t	*crtime;
115 	xoptattr_t	*xoap;
116 	void		*scanstamp;
117 	int		i;
118 
119 	xoap = xva_getxoptattr(xvap);
120 	ASSERT(xoap);
121 
122 	lrattr->lr_attr_masksize = xvap->xva_mapsize;
123 	bitmap = &lrattr->lr_attr_bitmap;
124 	for (i = 0; i != xvap->xva_mapsize; i++, bitmap++) {
125 		*bitmap = xvap->xva_reqattrmap[i];
126 	}
127 
128 	/* Now pack the attributes up in a single uint64_t */
129 	attrs = (uint64_t *)bitmap;
130 	crtime = attrs + 1;
131 	scanstamp = (caddr_t)(crtime + 2);
132 	*attrs = 0;
133 	if (XVA_ISSET_REQ(xvap, XAT_READONLY))
134 		*attrs |= (xoap->xoa_readonly == 0) ? 0 :
135 		    XAT0_READONLY;
136 	if (XVA_ISSET_REQ(xvap, XAT_HIDDEN))
137 		*attrs |= (xoap->xoa_hidden == 0) ? 0 :
138 		    XAT0_HIDDEN;
139 	if (XVA_ISSET_REQ(xvap, XAT_SYSTEM))
140 		*attrs |= (xoap->xoa_system == 0) ? 0 :
141 		    XAT0_SYSTEM;
142 	if (XVA_ISSET_REQ(xvap, XAT_ARCHIVE))
143 		*attrs |= (xoap->xoa_archive == 0) ? 0 :
144 		    XAT0_ARCHIVE;
145 	if (XVA_ISSET_REQ(xvap, XAT_IMMUTABLE))
146 		*attrs |= (xoap->xoa_immutable == 0) ? 0 :
147 		    XAT0_IMMUTABLE;
148 	if (XVA_ISSET_REQ(xvap, XAT_NOUNLINK))
149 		*attrs |= (xoap->xoa_nounlink == 0) ? 0 :
150 		    XAT0_NOUNLINK;
151 	if (XVA_ISSET_REQ(xvap, XAT_APPENDONLY))
152 		*attrs |= (xoap->xoa_appendonly == 0) ? 0 :
153 		    XAT0_APPENDONLY;
154 	if (XVA_ISSET_REQ(xvap, XAT_OPAQUE))
155 		*attrs |= (xoap->xoa_opaque == 0) ? 0 :
156 		    XAT0_APPENDONLY;
157 	if (XVA_ISSET_REQ(xvap, XAT_NODUMP))
158 		*attrs |= (xoap->xoa_nodump == 0) ? 0 :
159 		    XAT0_NODUMP;
160 	if (XVA_ISSET_REQ(xvap, XAT_AV_QUARANTINED))
161 		*attrs |= (xoap->xoa_av_quarantined == 0) ? 0 :
162 		    XAT0_AV_QUARANTINED;
163 	if (XVA_ISSET_REQ(xvap, XAT_AV_MODIFIED))
164 		*attrs |= (xoap->xoa_av_modified == 0) ? 0 :
165 		    XAT0_AV_MODIFIED;
166 	if (XVA_ISSET_REQ(xvap, XAT_CREATETIME))
167 		ZFS_TIME_ENCODE(&xoap->xoa_createtime, crtime);
168 	if (XVA_ISSET_REQ(xvap, XAT_AV_SCANSTAMP))
169 		bcopy(xoap->xoa_av_scanstamp, scanstamp, AV_SCANSTAMP_SZ);
170 	if (XVA_ISSET_REQ(xvap, XAT_REPARSE))
171 		*attrs |= (xoap->xoa_reparse == 0) ? 0 :
172 		    XAT0_REPARSE;
173 }
174 
175 static void *
176 zfs_log_fuid_ids(zfs_fuid_info_t *fuidp, void *start)
177 {
178 	zfs_fuid_t *zfuid;
179 	uint64_t *fuidloc = start;
180 
181 	/* First copy in the ACE FUIDs */
182 	for (zfuid = list_head(&fuidp->z_fuids); zfuid;
183 	    zfuid = list_next(&fuidp->z_fuids, zfuid)) {
184 		*fuidloc++ = zfuid->z_logfuid;
185 	}
186 	return (fuidloc);
187 }
188 
189 
190 static void *
191 zfs_log_fuid_domains(zfs_fuid_info_t *fuidp, void *start)
192 {
193 	zfs_fuid_domain_t *zdomain;
194 
195 	/* now copy in the domain info, if any */
196 	if (fuidp->z_domain_str_sz != 0) {
197 		for (zdomain = list_head(&fuidp->z_domains); zdomain;
198 		    zdomain = list_next(&fuidp->z_domains, zdomain)) {
199 			bcopy((void *)zdomain->z_domain, start,
200 			    strlen(zdomain->z_domain) + 1);
201 			start = (caddr_t)start +
202 			    strlen(zdomain->z_domain) + 1;
203 		}
204 	}
205 	return (start);
206 }
207 
208 /*
209  * zfs_log_create() is used to handle TX_CREATE, TX_CREATE_ATTR, TX_MKDIR,
210  * TX_MKDIR_ATTR and TX_MKXATTR
211  * transactions.
212  *
213  * TX_CREATE and TX_MKDIR are standard creates, but they may have FUID
214  * domain information appended prior to the name.  In this case the
215  * uid/gid in the log record will be a log centric FUID.
216  *
217  * TX_CREATE_ACL_ATTR and TX_MKDIR_ACL_ATTR handle special creates that
218  * may contain attributes, ACL and optional fuid information.
219  *
220  * TX_CREATE_ACL and TX_MKDIR_ACL handle special creates that specify
221  * and ACL and normal users/groups in the ACEs.
222  *
223  * There may be an optional xvattr attribute information similar
224  * to zfs_log_setattr.
225  *
226  * Also, after the file name "domain" strings may be appended.
227  */
228 void
229 zfs_log_create(zilog_t *zilog, dmu_tx_t *tx, uint64_t txtype,
230     znode_t *dzp, znode_t *zp, char *name, vsecattr_t *vsecp,
231     zfs_fuid_info_t *fuidp, vattr_t *vap)
232 {
233 	itx_t *itx;
234 	uint64_t seq;
235 	lr_create_t *lr;
236 	lr_acl_create_t *lracl;
237 	size_t aclsize;
238 	size_t xvatsize = 0;
239 	size_t txsize;
240 	xvattr_t *xvap = (xvattr_t *)vap;
241 	void *end;
242 	size_t lrsize;
243 	size_t namesize = strlen(name) + 1;
244 	size_t fuidsz = 0;
245 
246 	if (zil_replaying(zilog, tx))
247 		return;
248 
249 	/*
250 	 * If we have FUIDs present then add in space for
251 	 * domains and ACE fuid's if any.
252 	 */
253 	if (fuidp) {
254 		fuidsz += fuidp->z_domain_str_sz;
255 		fuidsz += fuidp->z_fuid_cnt * sizeof (uint64_t);
256 	}
257 
258 	if (vap->va_mask & AT_XVATTR)
259 		xvatsize = ZIL_XVAT_SIZE(xvap->xva_mapsize);
260 
261 	if ((int)txtype == TX_CREATE_ATTR || (int)txtype == TX_MKDIR_ATTR ||
262 	    (int)txtype == TX_CREATE || (int)txtype == TX_MKDIR ||
263 	    (int)txtype == TX_MKXATTR) {
264 		txsize = sizeof (*lr) + namesize + fuidsz + xvatsize;
265 		lrsize = sizeof (*lr);
266 	} else {
267 		aclsize = (vsecp) ? vsecp->vsa_aclentsz : 0;
268 		txsize =
269 		    sizeof (lr_acl_create_t) + namesize + fuidsz +
270 		    ZIL_ACE_LENGTH(aclsize) + xvatsize;
271 		lrsize = sizeof (lr_acl_create_t);
272 	}
273 
274 	itx = zil_itx_create(txtype, txsize);
275 
276 	lr = (lr_create_t *)&itx->itx_lr;
277 	lr->lr_doid = dzp->z_id;
278 	lr->lr_foid = zp->z_id;
279 	lr->lr_mode = zp->z_mode;
280 	if (!IS_EPHEMERAL(zp->z_uid)) {
281 		lr->lr_uid = (uint64_t)zp->z_uid;
282 	} else {
283 		lr->lr_uid = fuidp->z_fuid_owner;
284 	}
285 	if (!IS_EPHEMERAL(zp->z_gid)) {
286 		lr->lr_gid = (uint64_t)zp->z_gid;
287 	} else {
288 		lr->lr_gid = fuidp->z_fuid_group;
289 	}
290 	(void) sa_lookup(zp->z_sa_hdl, SA_ZPL_GEN(zp->z_zfsvfs), &lr->lr_gen,
291 	    sizeof (uint64_t));
292 	(void) sa_lookup(zp->z_sa_hdl, SA_ZPL_CRTIME(zp->z_zfsvfs),
293 	    lr->lr_crtime, sizeof (uint64_t) * 2);
294 
295 	if (sa_lookup(zp->z_sa_hdl, SA_ZPL_RDEV(zp->z_zfsvfs), &lr->lr_rdev,
296 	    sizeof (lr->lr_rdev)) != 0)
297 		lr->lr_rdev = 0;
298 
299 	/*
300 	 * Fill in xvattr info if any
301 	 */
302 	if (vap->va_mask & AT_XVATTR) {
303 		zfs_log_xvattr((lr_attr_t *)((caddr_t)lr + lrsize), xvap);
304 		end = (caddr_t)lr + lrsize + xvatsize;
305 	} else {
306 		end = (caddr_t)lr + lrsize;
307 	}
308 
309 	/* Now fill in any ACL info */
310 
311 	if (vsecp) {
312 		lracl = (lr_acl_create_t *)&itx->itx_lr;
313 		lracl->lr_aclcnt = vsecp->vsa_aclcnt;
314 		lracl->lr_acl_bytes = aclsize;
315 		lracl->lr_domcnt = fuidp ? fuidp->z_domain_cnt : 0;
316 		lracl->lr_fuidcnt  = fuidp ? fuidp->z_fuid_cnt : 0;
317 		if (vsecp->vsa_aclflags & VSA_ACE_ACLFLAGS)
318 			lracl->lr_acl_flags = (uint64_t)vsecp->vsa_aclflags;
319 		else
320 			lracl->lr_acl_flags = 0;
321 
322 		bcopy(vsecp->vsa_aclentp, end, aclsize);
323 		end = (caddr_t)end + ZIL_ACE_LENGTH(aclsize);
324 	}
325 
326 	/* drop in FUID info */
327 	if (fuidp) {
328 		end = zfs_log_fuid_ids(fuidp, end);
329 		end = zfs_log_fuid_domains(fuidp, end);
330 	}
331 	/*
332 	 * Now place file name in log record
333 	 */
334 	bcopy(name, end, namesize);
335 
336 	seq = zil_itx_assign(zilog, itx, tx);
337 	dzp->z_last_itx = seq;
338 	zp->z_last_itx = seq;
339 }
340 
341 /*
342  * zfs_log_remove() handles both TX_REMOVE and TX_RMDIR transactions.
343  */
344 void
345 zfs_log_remove(zilog_t *zilog, dmu_tx_t *tx, uint64_t txtype,
346 	znode_t *dzp, char *name)
347 {
348 	itx_t *itx;
349 	uint64_t seq;
350 	lr_remove_t *lr;
351 	size_t namesize = strlen(name) + 1;
352 
353 	if (zil_replaying(zilog, tx))
354 		return;
355 
356 	itx = zil_itx_create(txtype, sizeof (*lr) + namesize);
357 	lr = (lr_remove_t *)&itx->itx_lr;
358 	lr->lr_doid = dzp->z_id;
359 	bcopy(name, (char *)(lr + 1), namesize);
360 
361 	seq = zil_itx_assign(zilog, itx, tx);
362 	dzp->z_last_itx = seq;
363 }
364 
365 /*
366  * zfs_log_link() handles TX_LINK transactions.
367  */
368 void
369 zfs_log_link(zilog_t *zilog, dmu_tx_t *tx, uint64_t txtype,
370 	znode_t *dzp, znode_t *zp, char *name)
371 {
372 	itx_t *itx;
373 	uint64_t seq;
374 	lr_link_t *lr;
375 	size_t namesize = strlen(name) + 1;
376 
377 	if (zil_replaying(zilog, tx))
378 		return;
379 
380 	itx = zil_itx_create(txtype, sizeof (*lr) + namesize);
381 	lr = (lr_link_t *)&itx->itx_lr;
382 	lr->lr_doid = dzp->z_id;
383 	lr->lr_link_obj = zp->z_id;
384 	bcopy(name, (char *)(lr + 1), namesize);
385 
386 	seq = zil_itx_assign(zilog, itx, tx);
387 	dzp->z_last_itx = seq;
388 	zp->z_last_itx = seq;
389 }
390 
391 /*
392  * zfs_log_symlink() handles TX_SYMLINK transactions.
393  */
394 void
395 zfs_log_symlink(zilog_t *zilog, dmu_tx_t *tx, uint64_t txtype,
396     znode_t *dzp, znode_t *zp, char *name, char *link)
397 {
398 	itx_t *itx;
399 	uint64_t seq;
400 	lr_create_t *lr;
401 	size_t namesize = strlen(name) + 1;
402 	size_t linksize = strlen(link) + 1;
403 
404 	if (zil_replaying(zilog, tx))
405 		return;
406 
407 	itx = zil_itx_create(txtype, sizeof (*lr) + namesize + linksize);
408 	lr = (lr_create_t *)&itx->itx_lr;
409 	lr->lr_doid = dzp->z_id;
410 	lr->lr_foid = zp->z_id;
411 	lr->lr_uid = zp->z_uid;
412 	lr->lr_gid = zp->z_gid;
413 	lr->lr_mode = zp->z_mode;
414 	(void) sa_lookup(zp->z_sa_hdl, SA_ZPL_GEN(zp->z_zfsvfs), &lr->lr_gen,
415 	    sizeof (uint64_t));
416 	(void) sa_lookup(zp->z_sa_hdl, SA_ZPL_CRTIME(zp->z_zfsvfs),
417 	    lr->lr_crtime, sizeof (uint64_t) * 2);
418 	bcopy(name, (char *)(lr + 1), namesize);
419 	bcopy(link, (char *)(lr + 1) + namesize, linksize);
420 
421 	seq = zil_itx_assign(zilog, itx, tx);
422 	dzp->z_last_itx = seq;
423 	zp->z_last_itx = seq;
424 }
425 
426 /*
427  * zfs_log_rename() handles TX_RENAME transactions.
428  */
429 void
430 zfs_log_rename(zilog_t *zilog, dmu_tx_t *tx, uint64_t txtype,
431 	znode_t *sdzp, char *sname, znode_t *tdzp, char *dname, znode_t *szp)
432 {
433 	itx_t *itx;
434 	uint64_t seq;
435 	lr_rename_t *lr;
436 	size_t snamesize = strlen(sname) + 1;
437 	size_t dnamesize = strlen(dname) + 1;
438 
439 	if (zil_replaying(zilog, tx))
440 		return;
441 
442 	itx = zil_itx_create(txtype, sizeof (*lr) + snamesize + dnamesize);
443 	lr = (lr_rename_t *)&itx->itx_lr;
444 	lr->lr_sdoid = sdzp->z_id;
445 	lr->lr_tdoid = tdzp->z_id;
446 	bcopy(sname, (char *)(lr + 1), snamesize);
447 	bcopy(dname, (char *)(lr + 1) + snamesize, dnamesize);
448 
449 	seq = zil_itx_assign(zilog, itx, tx);
450 	sdzp->z_last_itx = seq;
451 	tdzp->z_last_itx = seq;
452 	szp->z_last_itx = seq;
453 }
454 
455 /*
456  * zfs_log_write() handles TX_WRITE transactions.
457  */
458 ssize_t zfs_immediate_write_sz = 32768;
459 
460 void
461 zfs_log_write(zilog_t *zilog, dmu_tx_t *tx, int txtype,
462 	znode_t *zp, offset_t off, ssize_t resid, int ioflag)
463 {
464 	itx_wr_state_t write_state;
465 	boolean_t slogging;
466 	uintptr_t fsync_cnt;
467 	ssize_t immediate_write_sz;
468 
469 	if (zil_replaying(zilog, tx) || zp->z_unlinked)
470 		return;
471 
472 	immediate_write_sz = (zilog->zl_logbias == ZFS_LOGBIAS_THROUGHPUT)
473 	    ? 0 : zfs_immediate_write_sz;
474 
475 	slogging = spa_has_slogs(zilog->zl_spa) &&
476 	    (zilog->zl_logbias == ZFS_LOGBIAS_LATENCY);
477 	if (resid > immediate_write_sz && !slogging && resid <= zp->z_blksz)
478 		write_state = WR_INDIRECT;
479 	else if (ioflag & (FSYNC | FDSYNC))
480 		write_state = WR_COPIED;
481 	else
482 		write_state = WR_NEED_COPY;
483 
484 	if ((fsync_cnt = (uintptr_t)tsd_get(zfs_fsyncer_key)) != 0) {
485 		(void) tsd_set(zfs_fsyncer_key, (void *)(fsync_cnt - 1));
486 	}
487 
488 	while (resid) {
489 		itx_t *itx;
490 		lr_write_t *lr;
491 		ssize_t len;
492 
493 		/*
494 		 * If the write would overflow the largest block then split it.
495 		 */
496 		if (write_state != WR_INDIRECT && resid > ZIL_MAX_LOG_DATA)
497 			len = SPA_MAXBLOCKSIZE >> 1;
498 		else
499 			len = resid;
500 
501 		itx = zil_itx_create(txtype, sizeof (*lr) +
502 		    (write_state == WR_COPIED ? len : 0));
503 		lr = (lr_write_t *)&itx->itx_lr;
504 		if (write_state == WR_COPIED && dmu_read(zp->z_zfsvfs->z_os,
505 		    zp->z_id, off, len, lr + 1, DMU_READ_NO_PREFETCH) != 0) {
506 			zil_itx_destroy(itx);
507 			itx = zil_itx_create(txtype, sizeof (*lr));
508 			lr = (lr_write_t *)&itx->itx_lr;
509 			write_state = WR_NEED_COPY;
510 		}
511 
512 		itx->itx_wr_state = write_state;
513 		if (write_state == WR_NEED_COPY)
514 			itx->itx_sod += len;
515 		lr->lr_foid = zp->z_id;
516 		lr->lr_offset = off;
517 		lr->lr_length = len;
518 		lr->lr_blkoff = 0;
519 		BP_ZERO(&lr->lr_blkptr);
520 
521 		itx->itx_private = zp->z_zfsvfs;
522 
523 		if ((zp->z_sync_cnt != 0) || (fsync_cnt != 0) ||
524 		    (ioflag & (FSYNC | FDSYNC)))
525 			itx->itx_sync = B_TRUE;
526 		else
527 			itx->itx_sync = B_FALSE;
528 
529 		zp->z_last_itx = zil_itx_assign(zilog, itx, tx);
530 
531 		off += len;
532 		resid -= len;
533 	}
534 }
535 
536 /*
537  * zfs_log_truncate() handles TX_TRUNCATE transactions.
538  */
539 void
540 zfs_log_truncate(zilog_t *zilog, dmu_tx_t *tx, int txtype,
541 	znode_t *zp, uint64_t off, uint64_t len)
542 {
543 	itx_t *itx;
544 	uint64_t seq;
545 	lr_truncate_t *lr;
546 
547 	if (zil_replaying(zilog, tx) || zp->z_unlinked)
548 		return;
549 
550 	itx = zil_itx_create(txtype, sizeof (*lr));
551 	lr = (lr_truncate_t *)&itx->itx_lr;
552 	lr->lr_foid = zp->z_id;
553 	lr->lr_offset = off;
554 	lr->lr_length = len;
555 
556 	itx->itx_sync = (zp->z_sync_cnt != 0);
557 	seq = zil_itx_assign(zilog, itx, tx);
558 	zp->z_last_itx = seq;
559 }
560 
561 /*
562  * zfs_log_setattr() handles TX_SETATTR transactions.
563  */
564 void
565 zfs_log_setattr(zilog_t *zilog, dmu_tx_t *tx, int txtype,
566 	znode_t *zp, vattr_t *vap, uint_t mask_applied, zfs_fuid_info_t *fuidp)
567 {
568 	itx_t		*itx;
569 	uint64_t	seq;
570 	lr_setattr_t	*lr;
571 	xvattr_t	*xvap = (xvattr_t *)vap;
572 	size_t		recsize = sizeof (lr_setattr_t);
573 	void		*start;
574 
575 	if (zil_replaying(zilog, tx) || zp->z_unlinked)
576 		return;
577 
578 	/*
579 	 * If XVATTR set, then log record size needs to allow
580 	 * for lr_attr_t + xvattr mask, mapsize and create time
581 	 * plus actual attribute values
582 	 */
583 	if (vap->va_mask & AT_XVATTR)
584 		recsize = sizeof (*lr) + ZIL_XVAT_SIZE(xvap->xva_mapsize);
585 
586 	if (fuidp)
587 		recsize += fuidp->z_domain_str_sz;
588 
589 	itx = zil_itx_create(txtype, recsize);
590 	lr = (lr_setattr_t *)&itx->itx_lr;
591 	lr->lr_foid = zp->z_id;
592 	lr->lr_mask = (uint64_t)mask_applied;
593 	lr->lr_mode = (uint64_t)vap->va_mode;
594 	if ((mask_applied & AT_UID) && IS_EPHEMERAL(vap->va_uid))
595 		lr->lr_uid = fuidp->z_fuid_owner;
596 	else
597 		lr->lr_uid = (uint64_t)vap->va_uid;
598 
599 	if ((mask_applied & AT_GID) && IS_EPHEMERAL(vap->va_gid))
600 		lr->lr_gid = fuidp->z_fuid_group;
601 	else
602 		lr->lr_gid = (uint64_t)vap->va_gid;
603 
604 	lr->lr_size = (uint64_t)vap->va_size;
605 	ZFS_TIME_ENCODE(&vap->va_atime, lr->lr_atime);
606 	ZFS_TIME_ENCODE(&vap->va_mtime, lr->lr_mtime);
607 	start = (lr_setattr_t *)(lr + 1);
608 	if (vap->va_mask & AT_XVATTR) {
609 		zfs_log_xvattr((lr_attr_t *)start, xvap);
610 		start = (caddr_t)start + ZIL_XVAT_SIZE(xvap->xva_mapsize);
611 	}
612 
613 	/*
614 	 * Now stick on domain information if any on end
615 	 */
616 
617 	if (fuidp)
618 		(void) zfs_log_fuid_domains(fuidp, start);
619 
620 	itx->itx_sync = (zp->z_sync_cnt != 0);
621 	seq = zil_itx_assign(zilog, itx, tx);
622 	zp->z_last_itx = seq;
623 }
624 
625 /*
626  * zfs_log_acl() handles TX_ACL transactions.
627  */
628 void
629 zfs_log_acl(zilog_t *zilog, dmu_tx_t *tx, znode_t *zp,
630     vsecattr_t *vsecp, zfs_fuid_info_t *fuidp)
631 {
632 	itx_t *itx;
633 	uint64_t seq;
634 	lr_acl_v0_t *lrv0;
635 	lr_acl_t *lr;
636 	int txtype;
637 	int lrsize;
638 	size_t txsize;
639 	size_t aclbytes = vsecp->vsa_aclentsz;
640 
641 	if (zil_replaying(zilog, tx) || zp->z_unlinked)
642 		return;
643 
644 	txtype = (zp->z_zfsvfs->z_version < ZPL_VERSION_FUID) ?
645 	    TX_ACL_V0 : TX_ACL;
646 
647 	if (txtype == TX_ACL)
648 		lrsize = sizeof (*lr);
649 	else
650 		lrsize = sizeof (*lrv0);
651 
652 	txsize = lrsize +
653 	    ((txtype == TX_ACL) ? ZIL_ACE_LENGTH(aclbytes) : aclbytes) +
654 	    (fuidp ? fuidp->z_domain_str_sz : 0) +
655 	    sizeof (uint64_t) * (fuidp ? fuidp->z_fuid_cnt : 0);
656 
657 	itx = zil_itx_create(txtype, txsize);
658 
659 	lr = (lr_acl_t *)&itx->itx_lr;
660 	lr->lr_foid = zp->z_id;
661 	if (txtype == TX_ACL) {
662 		lr->lr_acl_bytes = aclbytes;
663 		lr->lr_domcnt = fuidp ? fuidp->z_domain_cnt : 0;
664 		lr->lr_fuidcnt = fuidp ? fuidp->z_fuid_cnt : 0;
665 		if (vsecp->vsa_mask & VSA_ACE_ACLFLAGS)
666 			lr->lr_acl_flags = (uint64_t)vsecp->vsa_aclflags;
667 		else
668 			lr->lr_acl_flags = 0;
669 	}
670 	lr->lr_aclcnt = (uint64_t)vsecp->vsa_aclcnt;
671 
672 	if (txtype == TX_ACL_V0) {
673 		lrv0 = (lr_acl_v0_t *)lr;
674 		bcopy(vsecp->vsa_aclentp, (ace_t *)(lrv0 + 1), aclbytes);
675 	} else {
676 		void *start = (ace_t *)(lr + 1);
677 
678 		bcopy(vsecp->vsa_aclentp, start, aclbytes);
679 
680 		start = (caddr_t)start + ZIL_ACE_LENGTH(aclbytes);
681 
682 		if (fuidp) {
683 			start = zfs_log_fuid_ids(fuidp, start);
684 			(void) zfs_log_fuid_domains(fuidp, start);
685 		}
686 	}
687 
688 	itx->itx_sync = (zp->z_sync_cnt != 0);
689 	seq = zil_itx_assign(zilog, itx, tx);
690 	zp->z_last_itx = seq;
691 }
692