xref: /titanic_52/usr/src/uts/common/fs/zfs/zfs_log.c (revision d78b796caab6ff2bb68085a7ec5d97a89a21cfdd)
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  * Copyright (c) 2015 by Delphix. All rights reserved.
24  * Copyright (c) 2014 Integros [integros.com]
25  */
26 
27 #include <sys/types.h>
28 #include <sys/param.h>
29 #include <sys/systm.h>
30 #include <sys/sysmacros.h>
31 #include <sys/cmn_err.h>
32 #include <sys/kmem.h>
33 #include <sys/thread.h>
34 #include <sys/file.h>
35 #include <sys/vfs.h>
36 #include <sys/zfs_znode.h>
37 #include <sys/zfs_dir.h>
38 #include <sys/zil.h>
39 #include <sys/zil_impl.h>
40 #include <sys/byteorder.h>
41 #include <sys/policy.h>
42 #include <sys/stat.h>
43 #include <sys/mode.h>
44 #include <sys/acl.h>
45 #include <sys/dmu.h>
46 #include <sys/spa.h>
47 #include <sys/zfs_fuid.h>
48 #include <sys/ddi.h>
49 #include <sys/dsl_dataset.h>
50 #include <sys/zfs_events.h>
51 
52 /*
53  * These zfs_log_* functions must be called within a dmu tx, in one
54  * of 2 contexts depending on zilog->z_replay:
55  *
56  * Non replay mode
57  * ---------------
58  * We need to record the transaction so that if it is committed to
59  * the Intent Log then it can be replayed.  An intent log transaction
60  * structure (itx_t) is allocated and all the information necessary to
61  * possibly replay the transaction is saved in it. The itx is then assigned
62  * a sequence number and inserted in the in-memory list anchored in the zilog.
63  *
64  * Replay mode
65  * -----------
66  * We need to mark the intent log record as replayed in the log header.
67  * This is done in the same transaction as the replay so that they
68  * commit atomically.
69  */
70 
71 int
72 zfs_log_create_txtype(zil_create_t type, vsecattr_t *vsecp, vattr_t *vap)
73 {
74 	int isxvattr = (vap->va_mask & AT_XVATTR);
75 	switch (type) {
76 	case Z_FILE:
77 		if (vsecp == NULL && !isxvattr)
78 			return (TX_CREATE);
79 		if (vsecp && isxvattr)
80 			return (TX_CREATE_ACL_ATTR);
81 		if (vsecp)
82 			return (TX_CREATE_ACL);
83 		else
84 			return (TX_CREATE_ATTR);
85 		/*NOTREACHED*/
86 	case Z_DIR:
87 		if (vsecp == NULL && !isxvattr)
88 			return (TX_MKDIR);
89 		if (vsecp && isxvattr)
90 			return (TX_MKDIR_ACL_ATTR);
91 		if (vsecp)
92 			return (TX_MKDIR_ACL);
93 		else
94 			return (TX_MKDIR_ATTR);
95 	case Z_XATTRDIR:
96 		return (TX_MKXATTR);
97 	}
98 	ASSERT(0);
99 	return (TX_MAX_TYPE);
100 }
101 
102 /*
103  * build up the log data necessary for logging xvattr_t
104  * First lr_attr_t is initialized.  following the lr_attr_t
105  * is the mapsize and attribute bitmap copied from the xvattr_t.
106  * Following the bitmap and bitmapsize two 64 bit words are reserved
107  * for the create time which may be set.  Following the create time
108  * records a single 64 bit integer which has the bits to set on
109  * replay for the xvattr.
110  */
111 static void
112 zfs_log_xvattr(lr_attr_t *lrattr, xvattr_t *xvap)
113 {
114 	uint32_t	*bitmap;
115 	uint64_t	*attrs;
116 	uint64_t	*crtime;
117 	xoptattr_t	*xoap;
118 	void		*scanstamp;
119 	int		i;
120 
121 	xoap = xva_getxoptattr(xvap);
122 	ASSERT(xoap);
123 
124 	lrattr->lr_attr_masksize = xvap->xva_mapsize;
125 	bitmap = &lrattr->lr_attr_bitmap;
126 	for (i = 0; i != xvap->xva_mapsize; i++, bitmap++) {
127 		*bitmap = xvap->xva_reqattrmap[i];
128 	}
129 
130 	/* Now pack the attributes up in a single uint64_t */
131 	attrs = (uint64_t *)bitmap;
132 	crtime = attrs + 1;
133 	scanstamp = (caddr_t)(crtime + 2);
134 	*attrs = 0;
135 	if (XVA_ISSET_REQ(xvap, XAT_READONLY))
136 		*attrs |= (xoap->xoa_readonly == 0) ? 0 :
137 		    XAT0_READONLY;
138 	if (XVA_ISSET_REQ(xvap, XAT_HIDDEN))
139 		*attrs |= (xoap->xoa_hidden == 0) ? 0 :
140 		    XAT0_HIDDEN;
141 	if (XVA_ISSET_REQ(xvap, XAT_SYSTEM))
142 		*attrs |= (xoap->xoa_system == 0) ? 0 :
143 		    XAT0_SYSTEM;
144 	if (XVA_ISSET_REQ(xvap, XAT_ARCHIVE))
145 		*attrs |= (xoap->xoa_archive == 0) ? 0 :
146 		    XAT0_ARCHIVE;
147 	if (XVA_ISSET_REQ(xvap, XAT_IMMUTABLE))
148 		*attrs |= (xoap->xoa_immutable == 0) ? 0 :
149 		    XAT0_IMMUTABLE;
150 	if (XVA_ISSET_REQ(xvap, XAT_NOUNLINK))
151 		*attrs |= (xoap->xoa_nounlink == 0) ? 0 :
152 		    XAT0_NOUNLINK;
153 	if (XVA_ISSET_REQ(xvap, XAT_APPENDONLY))
154 		*attrs |= (xoap->xoa_appendonly == 0) ? 0 :
155 		    XAT0_APPENDONLY;
156 	if (XVA_ISSET_REQ(xvap, XAT_OPAQUE))
157 		*attrs |= (xoap->xoa_opaque == 0) ? 0 :
158 		    XAT0_APPENDONLY;
159 	if (XVA_ISSET_REQ(xvap, XAT_NODUMP))
160 		*attrs |= (xoap->xoa_nodump == 0) ? 0 :
161 		    XAT0_NODUMP;
162 	if (XVA_ISSET_REQ(xvap, XAT_AV_QUARANTINED))
163 		*attrs |= (xoap->xoa_av_quarantined == 0) ? 0 :
164 		    XAT0_AV_QUARANTINED;
165 	if (XVA_ISSET_REQ(xvap, XAT_AV_MODIFIED))
166 		*attrs |= (xoap->xoa_av_modified == 0) ? 0 :
167 		    XAT0_AV_MODIFIED;
168 	if (XVA_ISSET_REQ(xvap, XAT_CREATETIME))
169 		ZFS_TIME_ENCODE(&xoap->xoa_createtime, crtime);
170 	if (XVA_ISSET_REQ(xvap, XAT_AV_SCANSTAMP))
171 		bcopy(xoap->xoa_av_scanstamp, scanstamp, AV_SCANSTAMP_SZ);
172 	if (XVA_ISSET_REQ(xvap, XAT_REPARSE))
173 		*attrs |= (xoap->xoa_reparse == 0) ? 0 :
174 		    XAT0_REPARSE;
175 	if (XVA_ISSET_REQ(xvap, XAT_OFFLINE))
176 		*attrs |= (xoap->xoa_offline == 0) ? 0 :
177 		    XAT0_OFFLINE;
178 	if (XVA_ISSET_REQ(xvap, XAT_SPARSE))
179 		*attrs |= (xoap->xoa_sparse == 0) ? 0 :
180 		    XAT0_SPARSE;
181 }
182 
183 static void *
184 zfs_log_fuid_ids(zfs_fuid_info_t *fuidp, void *start)
185 {
186 	zfs_fuid_t *zfuid;
187 	uint64_t *fuidloc = start;
188 
189 	/* First copy in the ACE FUIDs */
190 	for (zfuid = list_head(&fuidp->z_fuids); zfuid;
191 	    zfuid = list_next(&fuidp->z_fuids, zfuid)) {
192 		*fuidloc++ = zfuid->z_logfuid;
193 	}
194 	return (fuidloc);
195 }
196 
197 
198 static void *
199 zfs_log_fuid_domains(zfs_fuid_info_t *fuidp, void *start)
200 {
201 	zfs_fuid_domain_t *zdomain;
202 
203 	/* now copy in the domain info, if any */
204 	if (fuidp->z_domain_str_sz != 0) {
205 		for (zdomain = list_head(&fuidp->z_domains); zdomain;
206 		    zdomain = list_next(&fuidp->z_domains, zdomain)) {
207 			bcopy((void *)zdomain->z_domain, start,
208 			    strlen(zdomain->z_domain) + 1);
209 			start = (caddr_t)start +
210 			    strlen(zdomain->z_domain) + 1;
211 		}
212 	}
213 	return (start);
214 }
215 
216 /*
217  * Handles TX_CREATE, TX_CREATE_ATTR, TX_MKDIR, TX_MKDIR_ATTR and
218  * TK_MKXATTR transactions.
219  *
220  * TX_CREATE and TX_MKDIR are standard creates, but they may have FUID
221  * domain information appended prior to the name.  In this case the
222  * uid/gid in the log record will be a log centric FUID.
223  *
224  * TX_CREATE_ACL_ATTR and TX_MKDIR_ACL_ATTR handle special creates that
225  * may contain attributes, ACL and optional fuid information.
226  *
227  * TX_CREATE_ACL and TX_MKDIR_ACL handle special creates that specify
228  * and ACL and normal users/groups in the ACEs.
229  *
230  * There may be an optional xvattr attribute information similar
231  * to zfs_log_setattr.
232  *
233  * Also, after the file name "domain" strings may be appended.
234  */
235 void
236 zfs_log_create(zilog_t *zilog, dmu_tx_t *tx, uint64_t txtype,
237     znode_t *dzp, znode_t *zp, char *name, vsecattr_t *vsecp,
238     zfs_fuid_info_t *fuidp, vattr_t *vap)
239 {
240 	itx_t *itx;
241 	lr_create_t *lr;
242 	lr_acl_create_t *lracl;
243 	size_t aclsize = (vsecp != NULL) ? vsecp->vsa_aclentsz : 0;
244 	size_t xvatsize = 0;
245 	size_t txsize;
246 	xvattr_t *xvap = (xvattr_t *)vap;
247 	void *end;
248 	size_t lrsize;
249 	size_t namesize = strlen(name) + 1;
250 	size_t fuidsz = 0;
251 
252 	if (zil_replaying(zilog, tx))
253 		return;
254 
255 	/*
256 	 * If we have FUIDs present then add in space for
257 	 * domains and ACE fuid's if any.
258 	 */
259 	if (fuidp) {
260 		fuidsz += fuidp->z_domain_str_sz;
261 		fuidsz += fuidp->z_fuid_cnt * sizeof (uint64_t);
262 	}
263 
264 	if (vap->va_mask & AT_XVATTR)
265 		xvatsize = ZIL_XVAT_SIZE(xvap->xva_mapsize);
266 
267 	if ((int)txtype == TX_CREATE_ATTR || (int)txtype == TX_MKDIR_ATTR ||
268 	    (int)txtype == TX_CREATE || (int)txtype == TX_MKDIR ||
269 	    (int)txtype == TX_MKXATTR) {
270 		txsize = sizeof (*lr) + namesize + fuidsz + xvatsize;
271 		lrsize = sizeof (*lr);
272 	} else {
273 		txsize =
274 		    sizeof (lr_acl_create_t) + namesize + fuidsz +
275 		    ZIL_ACE_LENGTH(aclsize) + xvatsize;
276 		lrsize = sizeof (lr_acl_create_t);
277 	}
278 
279 	itx = zil_itx_create(txtype, txsize);
280 
281 	lr = (lr_create_t *)&itx->itx_lr;
282 	lr->lr_doid = dzp->z_id;
283 	lr->lr_foid = zp->z_id;
284 	lr->lr_mode = zp->z_mode;
285 	if (!IS_EPHEMERAL(zp->z_uid)) {
286 		lr->lr_uid = (uint64_t)zp->z_uid;
287 	} else {
288 		lr->lr_uid = fuidp->z_fuid_owner;
289 	}
290 	if (!IS_EPHEMERAL(zp->z_gid)) {
291 		lr->lr_gid = (uint64_t)zp->z_gid;
292 	} else {
293 		lr->lr_gid = fuidp->z_fuid_group;
294 	}
295 	(void) sa_lookup(zp->z_sa_hdl, SA_ZPL_GEN(zp->z_zfsvfs), &lr->lr_gen,
296 	    sizeof (uint64_t));
297 	(void) sa_lookup(zp->z_sa_hdl, SA_ZPL_CRTIME(zp->z_zfsvfs),
298 	    lr->lr_crtime, sizeof (uint64_t) * 2);
299 
300 	if (sa_lookup(zp->z_sa_hdl, SA_ZPL_RDEV(zp->z_zfsvfs), &lr->lr_rdev,
301 	    sizeof (lr->lr_rdev)) != 0)
302 		lr->lr_rdev = 0;
303 
304 	/*
305 	 * Fill in xvattr info if any
306 	 */
307 	if (vap->va_mask & AT_XVATTR) {
308 		zfs_log_xvattr((lr_attr_t *)((caddr_t)lr + lrsize), xvap);
309 		end = (caddr_t)lr + lrsize + xvatsize;
310 	} else {
311 		end = (caddr_t)lr + lrsize;
312 	}
313 
314 	/* Now fill in any ACL info */
315 
316 	if (vsecp) {
317 		lracl = (lr_acl_create_t *)&itx->itx_lr;
318 		lracl->lr_aclcnt = vsecp->vsa_aclcnt;
319 		lracl->lr_acl_bytes = aclsize;
320 		lracl->lr_domcnt = fuidp ? fuidp->z_domain_cnt : 0;
321 		lracl->lr_fuidcnt  = fuidp ? fuidp->z_fuid_cnt : 0;
322 		if (vsecp->vsa_aclflags & VSA_ACE_ACLFLAGS)
323 			lracl->lr_acl_flags = (uint64_t)vsecp->vsa_aclflags;
324 		else
325 			lracl->lr_acl_flags = 0;
326 
327 		bcopy(vsecp->vsa_aclentp, end, aclsize);
328 		end = (caddr_t)end + ZIL_ACE_LENGTH(aclsize);
329 	}
330 
331 	/* drop in FUID info */
332 	if (fuidp) {
333 		end = zfs_log_fuid_ids(fuidp, end);
334 		end = zfs_log_fuid_domains(fuidp, end);
335 	}
336 	/*
337 	 * Now place file name in log record
338 	 */
339 	bcopy(name, end, namesize);
340 
341 	zil_itx_assign(zilog, itx, tx);
342 
343 	rw_enter(&rz_zev_rwlock, RW_READER);
344 	if (rz_zev_callbacks && rz_zev_callbacks->rz_zev_znode_create)
345 		rz_zev_callbacks->rz_zev_znode_create(dzp, zp, name, txtype);
346 	rw_exit(&rz_zev_rwlock);
347 }
348 
349 /*
350  * Handles both TX_REMOVE and TX_RMDIR transactions.
351  */
352 void
353 zfs_log_remove(zilog_t *zilog, dmu_tx_t *tx, uint64_t txtype,
354     znode_t *dzp, char *name, uint64_t foid)
355 {
356 	itx_t *itx;
357 	lr_remove_t *lr;
358 	size_t namesize = strlen(name) + 1;
359 
360 	if (zil_replaying(zilog, tx))
361 		return;
362 
363 	itx = zil_itx_create(txtype, sizeof (*lr) + namesize);
364 	lr = (lr_remove_t *)&itx->itx_lr;
365 	lr->lr_doid = dzp->z_id;
366 	bcopy(name, (char *)(lr + 1), namesize);
367 
368 	itx->itx_oid = foid;
369 
370 	zil_itx_assign(zilog, itx, tx);
371 
372 	rw_enter(&rz_zev_rwlock, RW_READER);
373 	if (rz_zev_callbacks && rz_zev_callbacks->rz_zev_znode_remove)
374 		rz_zev_callbacks->rz_zev_znode_remove(dzp, name, txtype);
375 	rw_exit(&rz_zev_rwlock);
376 }
377 
378 /*
379  * Handles TX_LINK transactions.
380  */
381 void
382 zfs_log_link(zilog_t *zilog, dmu_tx_t *tx, uint64_t txtype,
383     znode_t *dzp, znode_t *zp, char *name)
384 {
385 	itx_t *itx;
386 	lr_link_t *lr;
387 	size_t namesize = strlen(name) + 1;
388 
389 	if (zil_replaying(zilog, tx))
390 		return;
391 
392 	itx = zil_itx_create(txtype, sizeof (*lr) + namesize);
393 	lr = (lr_link_t *)&itx->itx_lr;
394 	lr->lr_doid = dzp->z_id;
395 	lr->lr_link_obj = zp->z_id;
396 	bcopy(name, (char *)(lr + 1), namesize);
397 
398 	zil_itx_assign(zilog, itx, tx);
399 
400 	rw_enter(&rz_zev_rwlock, RW_READER);
401 	if (rz_zev_callbacks && rz_zev_callbacks->rz_zev_znode_link)
402 		rz_zev_callbacks->rz_zev_znode_link(dzp, zp, name);
403 	rw_exit(&rz_zev_rwlock);
404 }
405 
406 /*
407  * Handles TX_SYMLINK transactions.
408  */
409 void
410 zfs_log_symlink(zilog_t *zilog, dmu_tx_t *tx, uint64_t txtype,
411     znode_t *dzp, znode_t *zp, char *name, char *link)
412 {
413 	itx_t *itx;
414 	lr_create_t *lr;
415 	size_t namesize = strlen(name) + 1;
416 	size_t linksize = strlen(link) + 1;
417 
418 	if (zil_replaying(zilog, tx))
419 		return;
420 
421 	itx = zil_itx_create(txtype, sizeof (*lr) + namesize + linksize);
422 	lr = (lr_create_t *)&itx->itx_lr;
423 	lr->lr_doid = dzp->z_id;
424 	lr->lr_foid = zp->z_id;
425 	lr->lr_uid = zp->z_uid;
426 	lr->lr_gid = zp->z_gid;
427 	lr->lr_mode = zp->z_mode;
428 	(void) sa_lookup(zp->z_sa_hdl, SA_ZPL_GEN(zp->z_zfsvfs), &lr->lr_gen,
429 	    sizeof (uint64_t));
430 	(void) sa_lookup(zp->z_sa_hdl, SA_ZPL_CRTIME(zp->z_zfsvfs),
431 	    lr->lr_crtime, sizeof (uint64_t) * 2);
432 	bcopy(name, (char *)(lr + 1), namesize);
433 	bcopy(link, (char *)(lr + 1) + namesize, linksize);
434 
435 	zil_itx_assign(zilog, itx, tx);
436 
437 	rw_enter(&rz_zev_rwlock, RW_READER);
438 	if (rz_zev_callbacks && rz_zev_callbacks->rz_zev_znode_symlink)
439 		rz_zev_callbacks->rz_zev_znode_symlink(dzp, zp, name, link);
440 	rw_exit(&rz_zev_rwlock);
441 }
442 
443 /*
444  * Handles TX_RENAME transactions.
445  */
446 void
447 zfs_log_rename(zilog_t *zilog, dmu_tx_t *tx, uint64_t txtype,
448     znode_t *sdzp, char *sname, znode_t *tdzp, char *dname, znode_t *szp)
449 {
450 	itx_t *itx;
451 	lr_rename_t *lr;
452 	size_t snamesize = strlen(sname) + 1;
453 	size_t dnamesize = strlen(dname) + 1;
454 
455 	if (zil_replaying(zilog, tx))
456 		return;
457 
458 	itx = zil_itx_create(txtype, sizeof (*lr) + snamesize + dnamesize);
459 	lr = (lr_rename_t *)&itx->itx_lr;
460 	lr->lr_sdoid = sdzp->z_id;
461 	lr->lr_tdoid = tdzp->z_id;
462 	bcopy(sname, (char *)(lr + 1), snamesize);
463 	bcopy(dname, (char *)(lr + 1) + snamesize, dnamesize);
464 	itx->itx_oid = szp->z_id;
465 
466 	zil_itx_assign(zilog, itx, tx);
467 
468 	rw_enter(&rz_zev_rwlock, RW_READER);
469 	if (rz_zev_callbacks && rz_zev_callbacks->rz_zev_znode_rename)
470 		rz_zev_callbacks->rz_zev_znode_rename(sdzp,
471 			sname, tdzp, dname, szp);
472 	rw_exit(&rz_zev_rwlock);
473 }
474 
475 /*
476  * Handles TX_WRITE transactions.
477  */
478 ssize_t zfs_immediate_write_sz = 32768;
479 
480 void
481 zfs_log_write(zilog_t *zilog, dmu_tx_t *tx, int txtype,
482     znode_t *zp, offset_t off, ssize_t resid, int ioflag)
483 {
484 	itx_wr_state_t write_state;
485 	boolean_t slogging;
486 	uintptr_t fsync_cnt;
487 	ssize_t immediate_write_sz;
488 
489 	if (zil_replaying(zilog, tx) || zp->z_unlinked)
490 		return;
491 
492 	immediate_write_sz = (zilog->zl_logbias == ZFS_LOGBIAS_THROUGHPUT)
493 	    ? 0 : zfs_immediate_write_sz;
494 
495 	slogging = spa_has_slogs(zilog->zl_spa) &&
496 	    (zilog->zl_logbias == ZFS_LOGBIAS_LATENCY);
497 	if (resid > immediate_write_sz && !slogging && resid <= zp->z_blksz)
498 		write_state = WR_INDIRECT;
499 	else if (ioflag & (FSYNC | FDSYNC))
500 		write_state = WR_COPIED;
501 	else
502 		write_state = WR_NEED_COPY;
503 
504 	if ((fsync_cnt = (uintptr_t)tsd_get(zfs_fsyncer_key)) != 0) {
505 		(void) tsd_set(zfs_fsyncer_key, (void *)(fsync_cnt - 1));
506 	}
507 
508 	while (resid) {
509 		itx_t *itx;
510 		lr_write_t *lr;
511 		ssize_t len;
512 
513 		/*
514 		 * If the write would overflow the largest block then split it.
515 		 */
516 		if (write_state != WR_INDIRECT && resid > ZIL_MAX_LOG_DATA)
517 			len = SPA_OLD_MAXBLOCKSIZE >> 1;
518 		else
519 			len = resid;
520 
521 		itx = zil_itx_create(txtype, sizeof (*lr) +
522 		    (write_state == WR_COPIED ? len : 0));
523 		lr = (lr_write_t *)&itx->itx_lr;
524 		if (write_state == WR_COPIED && dmu_read(zp->z_zfsvfs->z_os,
525 		    zp->z_id, off, len, lr + 1, DMU_READ_NO_PREFETCH) != 0) {
526 			zil_itx_destroy(itx);
527 			itx = zil_itx_create(txtype, sizeof (*lr));
528 			lr = (lr_write_t *)&itx->itx_lr;
529 			write_state = WR_NEED_COPY;
530 		}
531 
532 		itx->itx_wr_state = write_state;
533 		if (write_state == WR_NEED_COPY)
534 			itx->itx_sod += len;
535 		lr->lr_foid = zp->z_id;
536 		lr->lr_offset = off;
537 		lr->lr_length = len;
538 		lr->lr_blkoff = 0;
539 		BP_ZERO(&lr->lr_blkptr);
540 
541 		itx->itx_private = zp->z_zfsvfs;
542 
543 		if (!(ioflag & (FSYNC | FDSYNC)) && (zp->z_sync_cnt == 0) &&
544 		    (fsync_cnt == 0))
545 			itx->itx_sync = B_FALSE;
546 
547 		zil_itx_assign(zilog, itx, tx);
548 
549 		if (ZTOV(zp)->v_type == VREG)
550 			atomic_add_32(&(zp->z_new_content), 1);
551 		rw_enter(&rz_zev_rwlock, RW_READER);
552 		if (rz_zev_callbacks && rz_zev_callbacks->rz_zev_znode_write)
553 			rz_zev_callbacks->rz_zev_znode_write(zp, off, len);
554 		rw_exit(&rz_zev_rwlock);
555 
556 		off += len;
557 		resid -= len;
558 	}
559 }
560 
561 /*
562  * Handles TX_TRUNCATE transactions.
563  */
564 void
565 zfs_log_truncate(zilog_t *zilog, dmu_tx_t *tx, int txtype,
566     znode_t *zp, uint64_t off, uint64_t len)
567 {
568 	itx_t *itx;
569 	lr_truncate_t *lr;
570 
571 	if (zil_replaying(zilog, tx) || zp->z_unlinked)
572 		return;
573 
574 	itx = zil_itx_create(txtype, sizeof (*lr));
575 	lr = (lr_truncate_t *)&itx->itx_lr;
576 	lr->lr_foid = zp->z_id;
577 	lr->lr_offset = off;
578 	lr->lr_length = len;
579 
580 	itx->itx_sync = (zp->z_sync_cnt != 0);
581 	zil_itx_assign(zilog, itx, tx);
582 
583 	if (ZTOV(zp)->v_type == VREG)
584 		atomic_add_32(&(zp->z_new_content), 1);
585 	rw_enter(&rz_zev_rwlock, RW_READER);
586 	if (rz_zev_callbacks && rz_zev_callbacks->rz_zev_znode_truncate)
587 		rz_zev_callbacks->rz_zev_znode_truncate(zp, off, len);
588 	rw_exit(&rz_zev_rwlock);
589 }
590 
591 /*
592  * Handles TX_SETATTR transactions.
593  */
594 void
595 zfs_log_setattr(zilog_t *zilog, dmu_tx_t *tx, int txtype,
596     znode_t *zp, vattr_t *vap, uint_t mask_applied, zfs_fuid_info_t *fuidp)
597 {
598 	itx_t		*itx;
599 	lr_setattr_t	*lr;
600 	xvattr_t	*xvap = (xvattr_t *)vap;
601 	size_t		recsize = sizeof (lr_setattr_t);
602 	void		*start;
603 
604 	if (zil_replaying(zilog, tx) || zp->z_unlinked)
605 		return;
606 
607 	/*
608 	 * If XVATTR set, then log record size needs to allow
609 	 * for lr_attr_t + xvattr mask, mapsize and create time
610 	 * plus actual attribute values
611 	 */
612 	if (vap->va_mask & AT_XVATTR)
613 		recsize = sizeof (*lr) + ZIL_XVAT_SIZE(xvap->xva_mapsize);
614 
615 	if (fuidp)
616 		recsize += fuidp->z_domain_str_sz;
617 
618 	itx = zil_itx_create(txtype, recsize);
619 	lr = (lr_setattr_t *)&itx->itx_lr;
620 	lr->lr_foid = zp->z_id;
621 	lr->lr_mask = (uint64_t)mask_applied;
622 	lr->lr_mode = (uint64_t)vap->va_mode;
623 	if ((mask_applied & AT_UID) && IS_EPHEMERAL(vap->va_uid))
624 		lr->lr_uid = fuidp->z_fuid_owner;
625 	else
626 		lr->lr_uid = (uint64_t)vap->va_uid;
627 
628 	if ((mask_applied & AT_GID) && IS_EPHEMERAL(vap->va_gid))
629 		lr->lr_gid = fuidp->z_fuid_group;
630 	else
631 		lr->lr_gid = (uint64_t)vap->va_gid;
632 
633 	lr->lr_size = (uint64_t)vap->va_size;
634 	ZFS_TIME_ENCODE(&vap->va_atime, lr->lr_atime);
635 	ZFS_TIME_ENCODE(&vap->va_mtime, lr->lr_mtime);
636 	start = (lr_setattr_t *)(lr + 1);
637 	if (vap->va_mask & AT_XVATTR) {
638 		zfs_log_xvattr((lr_attr_t *)start, xvap);
639 		start = (caddr_t)start + ZIL_XVAT_SIZE(xvap->xva_mapsize);
640 	}
641 
642 	/*
643 	 * Now stick on domain information if any on end
644 	 */
645 
646 	if (fuidp)
647 		(void) zfs_log_fuid_domains(fuidp, start);
648 
649 	itx->itx_sync = (zp->z_sync_cnt != 0);
650 	zil_itx_assign(zilog, itx, tx);
651 
652 	rw_enter(&rz_zev_rwlock, RW_READER);
653 	if (rz_zev_callbacks && rz_zev_callbacks->rz_zev_znode_setattr)
654 		rz_zev_callbacks->rz_zev_znode_setattr(zp);
655 	rw_exit(&rz_zev_rwlock);
656 }
657 
658 /*
659  * Handles TX_ACL transactions.
660  */
661 void
662 zfs_log_acl(zilog_t *zilog, dmu_tx_t *tx, znode_t *zp,
663     vsecattr_t *vsecp, zfs_fuid_info_t *fuidp)
664 {
665 	itx_t *itx;
666 	lr_acl_v0_t *lrv0;
667 	lr_acl_t *lr;
668 	int txtype;
669 	int lrsize;
670 	size_t txsize;
671 	size_t aclbytes = vsecp->vsa_aclentsz;
672 
673 	if (zil_replaying(zilog, tx) || zp->z_unlinked)
674 		return;
675 
676 	txtype = (zp->z_zfsvfs->z_version < ZPL_VERSION_FUID) ?
677 	    TX_ACL_V0 : TX_ACL;
678 
679 	if (txtype == TX_ACL)
680 		lrsize = sizeof (*lr);
681 	else
682 		lrsize = sizeof (*lrv0);
683 
684 	txsize = lrsize +
685 	    ((txtype == TX_ACL) ? ZIL_ACE_LENGTH(aclbytes) : aclbytes) +
686 	    (fuidp ? fuidp->z_domain_str_sz : 0) +
687 	    sizeof (uint64_t) * (fuidp ? fuidp->z_fuid_cnt : 0);
688 
689 	itx = zil_itx_create(txtype, txsize);
690 
691 	lr = (lr_acl_t *)&itx->itx_lr;
692 	lr->lr_foid = zp->z_id;
693 	if (txtype == TX_ACL) {
694 		lr->lr_acl_bytes = aclbytes;
695 		lr->lr_domcnt = fuidp ? fuidp->z_domain_cnt : 0;
696 		lr->lr_fuidcnt = fuidp ? fuidp->z_fuid_cnt : 0;
697 		if (vsecp->vsa_mask & VSA_ACE_ACLFLAGS)
698 			lr->lr_acl_flags = (uint64_t)vsecp->vsa_aclflags;
699 		else
700 			lr->lr_acl_flags = 0;
701 	}
702 	lr->lr_aclcnt = (uint64_t)vsecp->vsa_aclcnt;
703 
704 	if (txtype == TX_ACL_V0) {
705 		lrv0 = (lr_acl_v0_t *)lr;
706 		bcopy(vsecp->vsa_aclentp, (ace_t *)(lrv0 + 1), aclbytes);
707 	} else {
708 		void *start = (ace_t *)(lr + 1);
709 
710 		bcopy(vsecp->vsa_aclentp, start, aclbytes);
711 
712 		start = (caddr_t)start + ZIL_ACE_LENGTH(aclbytes);
713 
714 		if (fuidp) {
715 			start = zfs_log_fuid_ids(fuidp, start);
716 			(void) zfs_log_fuid_domains(fuidp, start);
717 		}
718 	}
719 
720 	itx->itx_sync = (zp->z_sync_cnt != 0);
721 	zil_itx_assign(zilog, itx, tx);
722 
723 	rw_enter(&rz_zev_rwlock, RW_READER);
724 	if (rz_zev_callbacks && rz_zev_callbacks->rz_zev_znode_acl)
725 		rz_zev_callbacks->rz_zev_znode_acl(zp);
726 	rw_exit(&rz_zev_rwlock);
727 }
728