xref: /titanic_44/usr/src/uts/common/fs/zfs/zfs_log.c (revision a776e90690d98f629ab57f6493ff5949af9d65bd)
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 2008 Sun Microsystems, Inc.  All rights reserved.
23  * Use is subject to license terms.
24  */
25 
26 #pragma ident	"%Z%%M%	%I%	%E% SMI"
27 
28 #include <sys/types.h>
29 #include <sys/param.h>
30 #include <sys/systm.h>
31 #include <sys/sysmacros.h>
32 #include <sys/cmn_err.h>
33 #include <sys/kmem.h>
34 #include <sys/thread.h>
35 #include <sys/file.h>
36 #include <sys/vfs.h>
37 #include <sys/zfs_znode.h>
38 #include <sys/zfs_dir.h>
39 #include <sys/zil.h>
40 #include <sys/zil_impl.h>
41 #include <sys/byteorder.h>
42 #include <sys/policy.h>
43 #include <sys/stat.h>
44 #include <sys/mode.h>
45 #include <sys/acl.h>
46 #include <sys/dmu.h>
47 #include <sys/spa.h>
48 #include <sys/zfs_fuid.h>
49 #include <sys/ddi.h>
50 
51 /*
52  * All the functions in this file are used to construct the log entries
53  * to record transactions. They allocate * an intent log transaction
54  * structure (itx_t) and save within it all the information necessary to
55  * possibly replay the transaction. The itx is then assigned a sequence
56  * number and inserted in the in-memory list anchored in the zilog.
57  */
58 
59 int
60 zfs_log_create_txtype(zil_create_t type, vsecattr_t *vsecp, vattr_t *vap)
61 {
62 	int isxvattr = (vap->va_mask & AT_XVATTR);
63 	switch (type) {
64 	case Z_FILE:
65 		if (vsecp == NULL && !isxvattr)
66 			return (TX_CREATE);
67 		if (vsecp && isxvattr)
68 			return (TX_CREATE_ACL_ATTR);
69 		if (vsecp)
70 			return (TX_CREATE_ACL);
71 		else
72 			return (TX_CREATE_ATTR);
73 		/*NOTREACHED*/
74 	case Z_DIR:
75 		if (vsecp == NULL && !isxvattr)
76 			return (TX_MKDIR);
77 		if (vsecp && isxvattr)
78 			return (TX_MKDIR_ACL_ATTR);
79 		if (vsecp)
80 			return (TX_MKDIR_ACL);
81 		else
82 			return (TX_MKDIR_ATTR);
83 	case Z_XATTRDIR:
84 		return (TX_MKXATTR);
85 	}
86 	ASSERT(0);
87 	return (TX_MAX_TYPE);
88 }
89 
90 /*
91  * build up the log data necessary for logging xvattr_t
92  * First lr_attr_t is initialized.  following the lr_attr_t
93  * is the mapsize and attribute bitmap copied from the xvattr_t.
94  * Following the bitmap and bitmapsize two 64 bit words are reserved
95  * for the create time which may be set.  Following the create time
96  * records a single 64 bit integer which has the bits to set on
97  * replay for the xvattr.
98  */
99 static void
100 zfs_log_xvattr(lr_attr_t *lrattr, xvattr_t *xvap)
101 {
102 	uint32_t	*bitmap;
103 	uint64_t	*attrs;
104 	uint64_t	*crtime;
105 	xoptattr_t	*xoap;
106 	void		*scanstamp;
107 	int		i;
108 
109 	xoap = xva_getxoptattr(xvap);
110 	ASSERT(xoap);
111 
112 	lrattr->lr_attr_masksize = xvap->xva_mapsize;
113 	bitmap = &lrattr->lr_attr_bitmap;
114 	for (i = 0; i != xvap->xva_mapsize; i++, bitmap++) {
115 		*bitmap = xvap->xva_reqattrmap[i];
116 	}
117 
118 	/* Now pack the attributes up in a single uint64_t */
119 	attrs = (uint64_t *)bitmap;
120 	crtime = attrs + 1;
121 	scanstamp = (caddr_t)(crtime + 2);
122 	*attrs = 0;
123 	if (XVA_ISSET_REQ(xvap, XAT_READONLY))
124 		*attrs |= (xoap->xoa_readonly == 0) ? 0 :
125 		    XAT0_READONLY;
126 	if (XVA_ISSET_REQ(xvap, XAT_HIDDEN))
127 		*attrs |= (xoap->xoa_hidden == 0) ? 0 :
128 		    XAT0_HIDDEN;
129 	if (XVA_ISSET_REQ(xvap, XAT_SYSTEM))
130 		*attrs |= (xoap->xoa_system == 0) ? 0 :
131 		    XAT0_SYSTEM;
132 	if (XVA_ISSET_REQ(xvap, XAT_ARCHIVE))
133 		*attrs |= (xoap->xoa_archive == 0) ? 0 :
134 		    XAT0_ARCHIVE;
135 	if (XVA_ISSET_REQ(xvap, XAT_IMMUTABLE))
136 		*attrs |= (xoap->xoa_immutable == 0) ? 0 :
137 		    XAT0_IMMUTABLE;
138 	if (XVA_ISSET_REQ(xvap, XAT_NOUNLINK))
139 		*attrs |= (xoap->xoa_nounlink == 0) ? 0 :
140 		    XAT0_NOUNLINK;
141 	if (XVA_ISSET_REQ(xvap, XAT_APPENDONLY))
142 		*attrs |= (xoap->xoa_appendonly == 0) ? 0 :
143 		    XAT0_APPENDONLY;
144 	if (XVA_ISSET_REQ(xvap, XAT_OPAQUE))
145 		*attrs |= (xoap->xoa_opaque == 0) ? 0 :
146 		    XAT0_APPENDONLY;
147 	if (XVA_ISSET_REQ(xvap, XAT_NODUMP))
148 		*attrs |= (xoap->xoa_nodump == 0) ? 0 :
149 		    XAT0_NODUMP;
150 	if (XVA_ISSET_REQ(xvap, XAT_AV_QUARANTINED))
151 		*attrs |= (xoap->xoa_av_quarantined == 0) ? 0 :
152 		    XAT0_AV_QUARANTINED;
153 	if (XVA_ISSET_REQ(xvap, XAT_AV_MODIFIED))
154 		*attrs |= (xoap->xoa_av_modified == 0) ? 0 :
155 		    XAT0_AV_MODIFIED;
156 	if (XVA_ISSET_REQ(xvap, XAT_CREATETIME))
157 		ZFS_TIME_ENCODE(&xoap->xoa_createtime, crtime);
158 	if (XVA_ISSET_REQ(xvap, XAT_AV_SCANSTAMP))
159 		bcopy(xoap->xoa_av_scanstamp, scanstamp, AV_SCANSTAMP_SZ);
160 }
161 
162 static void *
163 zfs_log_fuid_ids(zfs_fuid_info_t *fuidp, void *start)
164 {
165 	zfs_fuid_t *zfuid;
166 	uint64_t *fuidloc = start;
167 
168 	/* First copy in the ACE FUIDs */
169 	for (zfuid = list_head(&fuidp->z_fuids); zfuid;
170 	    zfuid = list_next(&fuidp->z_fuids, zfuid)) {
171 		*fuidloc++ = zfuid->z_logfuid;
172 	}
173 	return (fuidloc);
174 }
175 
176 
177 static void *
178 zfs_log_fuid_domains(zfs_fuid_info_t *fuidp, void *start)
179 {
180 	zfs_fuid_domain_t *zdomain;
181 
182 	/* now copy in the domain info, if any */
183 	if (fuidp->z_domain_str_sz != 0) {
184 		for (zdomain = list_head(&fuidp->z_domains); zdomain;
185 		    zdomain = list_next(&fuidp->z_domains, zdomain)) {
186 			bcopy((void *)zdomain->z_domain, start,
187 			    strlen(zdomain->z_domain) + 1);
188 			start = (caddr_t)start +
189 			    strlen(zdomain->z_domain) + 1;
190 		}
191 	}
192 	return (start);
193 }
194 
195 /*
196  * zfs_log_create() is used to handle TX_CREATE, TX_CREATE_ATTR, TX_MKDIR,
197  * TX_MKDIR_ATTR and TX_MKXATTR
198  * transactions.
199  *
200  * TX_CREATE and TX_MKDIR are standard creates, but they may have FUID
201  * domain information appended prior to the name.  In this case the
202  * uid/gid in the log record will be a log centric FUID.
203  *
204  * TX_CREATE_ACL_ATTR and TX_MKDIR_ACL_ATTR handle special creates that
205  * may contain attributes, ACL and optional fuid information.
206  *
207  * TX_CREATE_ACL and TX_MKDIR_ACL handle special creates that specify
208  * and ACL and normal users/groups in the ACEs.
209  *
210  * There may be an optional xvattr attribute information similar
211  * to zfs_log_setattr.
212  *
213  * Also, after the file name "domain" strings may be appended.
214  */
215 void
216 zfs_log_create(zilog_t *zilog, dmu_tx_t *tx, uint64_t txtype,
217     znode_t *dzp, znode_t *zp, char *name, vsecattr_t *vsecp,
218     zfs_fuid_info_t *fuidp, vattr_t *vap)
219 {
220 	itx_t *itx;
221 	uint64_t seq;
222 	lr_create_t *lr;
223 	lr_acl_create_t *lracl;
224 	size_t aclsize;
225 	size_t xvatsize = 0;
226 	size_t txsize;
227 	xvattr_t *xvap = (xvattr_t *)vap;
228 	void *end;
229 	size_t lrsize;
230 
231 	size_t namesize = strlen(name) + 1;
232 	size_t fuidsz = 0;
233 
234 	if (zilog == NULL)
235 		return;
236 
237 	/*
238 	 * If we have FUIDs present then add in space for
239 	 * domains and ACE fuid's if any.
240 	 */
241 	if (fuidp) {
242 		fuidsz += fuidp->z_domain_str_sz;
243 		fuidsz += fuidp->z_fuid_cnt * sizeof (uint64_t);
244 	}
245 
246 	if (vap->va_mask & AT_XVATTR)
247 		xvatsize = ZIL_XVAT_SIZE(xvap->xva_mapsize);
248 
249 	if ((int)txtype == TX_CREATE_ATTR || (int)txtype == TX_MKDIR_ATTR ||
250 	    (int)txtype == TX_CREATE || (int)txtype == TX_MKDIR ||
251 	    (int)txtype == TX_MKXATTR) {
252 		txsize = sizeof (*lr) + namesize + fuidsz + xvatsize;
253 		lrsize = sizeof (*lr);
254 	} else {
255 		aclsize = (vsecp) ? vsecp->vsa_aclentsz : 0;
256 		txsize =
257 		    sizeof (lr_acl_create_t) + namesize + fuidsz +
258 		    ZIL_ACE_LENGTH(aclsize) + xvatsize;
259 		lrsize = sizeof (lr_acl_create_t);
260 	}
261 
262 	itx = zil_itx_create(txtype, txsize);
263 
264 	lr = (lr_create_t *)&itx->itx_lr;
265 	lr->lr_doid = dzp->z_id;
266 	lr->lr_foid = zp->z_id;
267 	lr->lr_mode = zp->z_phys->zp_mode;
268 	if (!IS_EPHEMERAL(zp->z_phys->zp_uid)) {
269 		lr->lr_uid = (uint64_t)zp->z_phys->zp_uid;
270 	} else {
271 		lr->lr_uid = fuidp->z_fuid_owner;
272 	}
273 	if (!IS_EPHEMERAL(zp->z_phys->zp_gid)) {
274 		lr->lr_gid = (uint64_t)zp->z_phys->zp_gid;
275 	} else {
276 		lr->lr_gid = fuidp->z_fuid_group;
277 	}
278 	lr->lr_gen = zp->z_phys->zp_gen;
279 	lr->lr_crtime[0] = zp->z_phys->zp_crtime[0];
280 	lr->lr_crtime[1] = zp->z_phys->zp_crtime[1];
281 	lr->lr_rdev = zp->z_phys->zp_rdev;
282 
283 	/*
284 	 * Fill in xvattr info if any
285 	 */
286 	if (vap->va_mask & AT_XVATTR) {
287 		zfs_log_xvattr((lr_attr_t *)((caddr_t)lr + lrsize), xvap);
288 		end = (caddr_t)lr + lrsize + xvatsize;
289 	} else {
290 		end = (caddr_t)lr + lrsize;
291 	}
292 
293 	/* Now fill in any ACL info */
294 
295 	if (vsecp) {
296 		lracl = (lr_acl_create_t *)&itx->itx_lr;
297 		lracl->lr_aclcnt = vsecp->vsa_aclcnt;
298 		lracl->lr_acl_bytes = aclsize;
299 		lracl->lr_domcnt = fuidp ? fuidp->z_domain_cnt : 0;
300 		lracl->lr_fuidcnt  = fuidp ? fuidp->z_fuid_cnt : 0;
301 		if (vsecp->vsa_aclflags & VSA_ACE_ACLFLAGS)
302 			lracl->lr_acl_flags = (uint64_t)vsecp->vsa_aclflags;
303 		else
304 			lracl->lr_acl_flags = 0;
305 
306 		bcopy(vsecp->vsa_aclentp, end, aclsize);
307 		end = (caddr_t)end + ZIL_ACE_LENGTH(aclsize);
308 	}
309 
310 	/* drop in FUID info */
311 	if (fuidp) {
312 		end = zfs_log_fuid_ids(fuidp, end);
313 		end = zfs_log_fuid_domains(fuidp, end);
314 	}
315 	/*
316 	 * Now place file name in log record
317 	 */
318 	bcopy(name, end, namesize);
319 
320 	seq = zil_itx_assign(zilog, itx, tx);
321 	dzp->z_last_itx = seq;
322 	zp->z_last_itx = seq;
323 }
324 
325 /*
326  * zfs_log_remove() handles both TX_REMOVE and TX_RMDIR transactions.
327  */
328 void
329 zfs_log_remove(zilog_t *zilog, dmu_tx_t *tx, uint64_t txtype,
330 	znode_t *dzp, char *name)
331 {
332 	itx_t *itx;
333 	uint64_t seq;
334 	lr_remove_t *lr;
335 	size_t namesize = strlen(name) + 1;
336 
337 	if (zilog == NULL)
338 		return;
339 
340 	itx = zil_itx_create(txtype, sizeof (*lr) + namesize);
341 	lr = (lr_remove_t *)&itx->itx_lr;
342 	lr->lr_doid = dzp->z_id;
343 	bcopy(name, (char *)(lr + 1), namesize);
344 
345 	seq = zil_itx_assign(zilog, itx, tx);
346 	dzp->z_last_itx = seq;
347 }
348 
349 /*
350  * zfs_log_link() handles TX_LINK transactions.
351  */
352 void
353 zfs_log_link(zilog_t *zilog, dmu_tx_t *tx, uint64_t txtype,
354 	znode_t *dzp, znode_t *zp, char *name)
355 {
356 	itx_t *itx;
357 	uint64_t seq;
358 	lr_link_t *lr;
359 	size_t namesize = strlen(name) + 1;
360 
361 	if (zilog == NULL)
362 		return;
363 
364 	itx = zil_itx_create(txtype, sizeof (*lr) + namesize);
365 	lr = (lr_link_t *)&itx->itx_lr;
366 	lr->lr_doid = dzp->z_id;
367 	lr->lr_link_obj = zp->z_id;
368 	bcopy(name, (char *)(lr + 1), namesize);
369 
370 	seq = zil_itx_assign(zilog, itx, tx);
371 	dzp->z_last_itx = seq;
372 	zp->z_last_itx = seq;
373 }
374 
375 /*
376  * zfs_log_symlink() handles TX_SYMLINK transactions.
377  */
378 void
379 zfs_log_symlink(zilog_t *zilog, dmu_tx_t *tx, uint64_t txtype,
380     znode_t *dzp, znode_t *zp, char *name, char *link)
381 {
382 	itx_t *itx;
383 	uint64_t seq;
384 	lr_create_t *lr;
385 	size_t namesize = strlen(name) + 1;
386 	size_t linksize = strlen(link) + 1;
387 
388 	if (zilog == NULL)
389 		return;
390 
391 	itx = zil_itx_create(txtype, sizeof (*lr) + namesize + linksize);
392 	lr = (lr_create_t *)&itx->itx_lr;
393 	lr->lr_doid = dzp->z_id;
394 	lr->lr_foid = zp->z_id;
395 	lr->lr_mode = zp->z_phys->zp_mode;
396 	lr->lr_uid = zp->z_phys->zp_uid;
397 	lr->lr_gid = zp->z_phys->zp_gid;
398 	lr->lr_gen = zp->z_phys->zp_gen;
399 	lr->lr_crtime[0] = zp->z_phys->zp_crtime[0];
400 	lr->lr_crtime[1] = zp->z_phys->zp_crtime[1];
401 	bcopy(name, (char *)(lr + 1), namesize);
402 	bcopy(link, (char *)(lr + 1) + namesize, linksize);
403 
404 	seq = zil_itx_assign(zilog, itx, tx);
405 	dzp->z_last_itx = seq;
406 	zp->z_last_itx = seq;
407 }
408 
409 /*
410  * zfs_log_rename() handles TX_RENAME transactions.
411  */
412 void
413 zfs_log_rename(zilog_t *zilog, dmu_tx_t *tx, uint64_t txtype,
414 	znode_t *sdzp, char *sname, znode_t *tdzp, char *dname, znode_t *szp)
415 {
416 	itx_t *itx;
417 	uint64_t seq;
418 	lr_rename_t *lr;
419 	size_t snamesize = strlen(sname) + 1;
420 	size_t dnamesize = strlen(dname) + 1;
421 
422 	if (zilog == NULL)
423 		return;
424 
425 	itx = zil_itx_create(txtype, sizeof (*lr) + snamesize + dnamesize);
426 	lr = (lr_rename_t *)&itx->itx_lr;
427 	lr->lr_sdoid = sdzp->z_id;
428 	lr->lr_tdoid = tdzp->z_id;
429 	bcopy(sname, (char *)(lr + 1), snamesize);
430 	bcopy(dname, (char *)(lr + 1) + snamesize, dnamesize);
431 
432 	seq = zil_itx_assign(zilog, itx, tx);
433 	sdzp->z_last_itx = seq;
434 	tdzp->z_last_itx = seq;
435 	szp->z_last_itx = seq;
436 }
437 
438 /*
439  * zfs_log_write() handles TX_WRITE transactions.
440  */
441 ssize_t zfs_immediate_write_sz = 32768;
442 
443 #define	ZIL_MAX_LOG_DATA (SPA_MAXBLOCKSIZE - sizeof (zil_trailer_t) - \
444     sizeof (lr_write_t))
445 
446 void
447 zfs_log_write(zilog_t *zilog, dmu_tx_t *tx, int txtype,
448 	znode_t *zp, offset_t off, ssize_t resid, int ioflag)
449 {
450 	itx_wr_state_t write_state;
451 	boolean_t slogging;
452 	uintptr_t fsync_cnt;
453 
454 	if (zilog == NULL || zp->z_unlinked)
455 		return;
456 
457 	/*
458 	 * Writes are handled in three different ways:
459 	 *
460 	 * WR_INDIRECT:
461 	 *    If the write is greater than zfs_immediate_write_sz and there are
462 	 *    no separate logs in this pool then later *if* we need to log the
463 	 *    write then dmu_sync() is used to immediately write the block and
464 	 *    its block pointer is put in the log record.
465 	 * WR_COPIED:
466 	 *    If we know we'll immediately be committing the
467 	 *    transaction (FSYNC or FDSYNC), the we allocate a larger
468 	 *    log record here for the data and copy the data in.
469 	 * WR_NEED_COPY:
470 	 *    Otherwise we don't allocate a buffer, and *if* we need to
471 	 *    flush the write later then a buffer is allocated and
472 	 *    we retrieve the data using the dmu.
473 	 */
474 	slogging = spa_has_slogs(zilog->zl_spa);
475 	if (resid > zfs_immediate_write_sz && !slogging)
476 		write_state = WR_INDIRECT;
477 	else if (ioflag & (FSYNC | FDSYNC))
478 		write_state = WR_COPIED;
479 	else
480 		write_state = WR_NEED_COPY;
481 
482 	if ((fsync_cnt = (uintptr_t)tsd_get(zfs_fsyncer_key)) != 0) {
483 		(void) tsd_set(zfs_fsyncer_key, (void *)(fsync_cnt - 1));
484 	}
485 
486 	while (resid) {
487 		itx_t *itx;
488 		lr_write_t *lr;
489 		ssize_t len;
490 
491 		/*
492 		 * If there are slogs and the write would overflow the largest
493 		 * block, then because we don't want to use the main pool
494 		 * to dmu_sync, we have to split the write.
495 		 */
496 		if (slogging && 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) != 0) {
506 			kmem_free(itx, offsetof(itx_t, itx_lr) +
507 			    itx->itx_lr.lrc_reclen);
508 			itx = zil_itx_create(txtype, sizeof (*lr));
509 			lr = (lr_write_t *)&itx->itx_lr;
510 			write_state = WR_NEED_COPY;
511 		}
512 
513 		itx->itx_wr_state = write_state;
514 		if (write_state == WR_NEED_COPY)
515 			itx->itx_sod += len;
516 		lr->lr_foid = zp->z_id;
517 		lr->lr_offset = off;
518 		lr->lr_length = len;
519 		lr->lr_blkoff = 0;
520 		BP_ZERO(&lr->lr_blkptr);
521 
522 		itx->itx_private = zp->z_zfsvfs;
523 
524 		if ((zp->z_sync_cnt != 0) || (fsync_cnt != 0) ||
525 		    (ioflag & (FSYNC | FDSYNC)))
526 			itx->itx_sync = B_TRUE;
527 		else
528 			itx->itx_sync = B_FALSE;
529 
530 		zp->z_last_itx = zil_itx_assign(zilog, itx, tx);
531 
532 		off += len;
533 		resid -= len;
534 	}
535 }
536 
537 /*
538  * zfs_log_truncate() handles TX_TRUNCATE transactions.
539  */
540 void
541 zfs_log_truncate(zilog_t *zilog, dmu_tx_t *tx, int txtype,
542 	znode_t *zp, uint64_t off, uint64_t len)
543 {
544 	itx_t *itx;
545 	uint64_t seq;
546 	lr_truncate_t *lr;
547 
548 	if (zilog == NULL || zp->z_unlinked)
549 		return;
550 
551 	itx = zil_itx_create(txtype, sizeof (*lr));
552 	lr = (lr_truncate_t *)&itx->itx_lr;
553 	lr->lr_foid = zp->z_id;
554 	lr->lr_offset = off;
555 	lr->lr_length = len;
556 
557 	itx->itx_sync = (zp->z_sync_cnt != 0);
558 	seq = zil_itx_assign(zilog, itx, tx);
559 	zp->z_last_itx = seq;
560 }
561 
562 /*
563  * zfs_log_setattr() handles TX_SETATTR transactions.
564  */
565 void
566 zfs_log_setattr(zilog_t *zilog, dmu_tx_t *tx, int txtype,
567 	znode_t *zp, vattr_t *vap, uint_t mask_applied, zfs_fuid_info_t *fuidp)
568 {
569 	itx_t		*itx;
570 	uint64_t	seq;
571 	lr_setattr_t	*lr;
572 	xvattr_t	*xvap = (xvattr_t *)vap;
573 	size_t		recsize = sizeof (lr_setattr_t);
574 	void		*start;
575 
576 
577 	if (zilog == NULL || zp->z_unlinked)
578 		return;
579 
580 	/*
581 	 * If XVATTR set, then log record size needs to allow
582 	 * for lr_attr_t + xvattr mask, mapsize and create time
583 	 * plus actual attribute values
584 	 */
585 	if (vap->va_mask & AT_XVATTR)
586 		recsize = sizeof (*lr) + ZIL_XVAT_SIZE(xvap->xva_mapsize);
587 
588 	if (fuidp)
589 		recsize += fuidp->z_domain_str_sz;
590 
591 	itx = zil_itx_create(txtype, recsize);
592 	lr = (lr_setattr_t *)&itx->itx_lr;
593 	lr->lr_foid = zp->z_id;
594 	lr->lr_mask = (uint64_t)mask_applied;
595 	lr->lr_mode = (uint64_t)vap->va_mode;
596 	if ((mask_applied & AT_UID) && IS_EPHEMERAL(vap->va_uid))
597 		lr->lr_uid = fuidp->z_fuid_owner;
598 	else
599 		lr->lr_uid = (uint64_t)vap->va_uid;
600 
601 	if ((mask_applied & AT_GID) && IS_EPHEMERAL(vap->va_gid))
602 		lr->lr_gid = fuidp->z_fuid_group;
603 	else
604 		lr->lr_gid = (uint64_t)vap->va_gid;
605 
606 	lr->lr_size = (uint64_t)vap->va_size;
607 	ZFS_TIME_ENCODE(&vap->va_atime, lr->lr_atime);
608 	ZFS_TIME_ENCODE(&vap->va_mtime, lr->lr_mtime);
609 	start = (lr_setattr_t *)(lr + 1);
610 	if (vap->va_mask & AT_XVATTR) {
611 		zfs_log_xvattr((lr_attr_t *)start, xvap);
612 		start = (caddr_t)start + ZIL_XVAT_SIZE(xvap->xva_mapsize);
613 	}
614 
615 	/*
616 	 * Now stick on domain information if any on end
617 	 */
618 
619 	if (fuidp)
620 		(void) zfs_log_fuid_domains(fuidp, start);
621 
622 	itx->itx_sync = (zp->z_sync_cnt != 0);
623 	seq = zil_itx_assign(zilog, itx, tx);
624 	zp->z_last_itx = seq;
625 }
626 
627 /*
628  * zfs_log_acl() handles TX_ACL transactions.
629  */
630 void
631 zfs_log_acl(zilog_t *zilog, dmu_tx_t *tx, znode_t *zp,
632     vsecattr_t *vsecp, zfs_fuid_info_t *fuidp)
633 {
634 	itx_t *itx;
635 	uint64_t seq;
636 	lr_acl_v0_t *lrv0;
637 	lr_acl_t *lr;
638 	int txtype;
639 	int lrsize;
640 	size_t txsize;
641 	size_t aclbytes = vsecp->vsa_aclentsz;
642 
643 	txtype = (zp->z_zfsvfs->z_version == ZPL_VERSION_INITIAL) ?
644 	    TX_ACL_V0 : TX_ACL;
645 
646 	if (txtype == TX_ACL)
647 		lrsize = sizeof (*lr);
648 	else
649 		lrsize = sizeof (*lrv0);
650 
651 	if (zilog == NULL || zp->z_unlinked)
652 		return;
653 
654 	txsize = lrsize +
655 	    ((txtype == TX_ACL) ? ZIL_ACE_LENGTH(aclbytes) : aclbytes) +
656 	    (fuidp ? fuidp->z_domain_str_sz : 0) +
657 	    sizeof (uint64) * (fuidp ? fuidp->z_fuid_cnt : 0);
658 
659 	itx = zil_itx_create(txtype, txsize);
660 
661 	lr = (lr_acl_t *)&itx->itx_lr;
662 	lr->lr_foid = zp->z_id;
663 	if (txtype == TX_ACL) {
664 		lr->lr_acl_bytes = aclbytes;
665 		lr->lr_domcnt = fuidp ? fuidp->z_domain_cnt : 0;
666 		lr->lr_fuidcnt = fuidp ? fuidp->z_fuid_cnt : 0;
667 		if (vsecp->vsa_mask & VSA_ACE_ACLFLAGS)
668 			lr->lr_acl_flags = (uint64_t)vsecp->vsa_aclflags;
669 		else
670 			lr->lr_acl_flags = 0;
671 	}
672 	lr->lr_aclcnt = (uint64_t)vsecp->vsa_aclcnt;
673 
674 	if (txtype == TX_ACL_V0) {
675 		lrv0 = (lr_acl_v0_t *)lr;
676 		bcopy(vsecp->vsa_aclentp, (ace_t *)(lrv0 + 1), aclbytes);
677 	} else {
678 		void *start = (ace_t *)(lr + 1);
679 
680 		bcopy(vsecp->vsa_aclentp, start, aclbytes);
681 
682 		start = (caddr_t)start + ZIL_ACE_LENGTH(aclbytes);
683 
684 		if (fuidp) {
685 			start = zfs_log_fuid_ids(fuidp, start);
686 			(void) zfs_log_fuid_domains(fuidp, start);
687 		}
688 	}
689 
690 	itx->itx_sync = (zp->z_sync_cnt != 0);
691 	seq = zil_itx_assign(zilog, itx, tx);
692 	zp->z_last_itx = seq;
693 }
694