xref: /titanic_44/usr/src/uts/common/fs/zfs/zfs_log.c (revision 8aa47a6b5a3a9ec7a183509986554d4f8482d6ab)
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 #include <sys/zfs_events.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
zfs_log_create_txtype(zil_create_t type,vsecattr_t * vsecp,vattr_t * vap)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
zfs_log_xvattr(lr_attr_t * lrattr,xvattr_t * xvap)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 	if (XVA_ISSET_REQ(xvap, XAT_OFFLINE))
174 		*attrs |= (xoap->xoa_offline == 0) ? 0 :
175 		    XAT0_OFFLINE;
176 	if (XVA_ISSET_REQ(xvap, XAT_SPARSE))
177 		*attrs |= (xoap->xoa_sparse == 0) ? 0 :
178 		    XAT0_SPARSE;
179 }
180 
181 static void *
zfs_log_fuid_ids(zfs_fuid_info_t * fuidp,void * start)182 zfs_log_fuid_ids(zfs_fuid_info_t *fuidp, void *start)
183 {
184 	zfs_fuid_t *zfuid;
185 	uint64_t *fuidloc = start;
186 
187 	/* First copy in the ACE FUIDs */
188 	for (zfuid = list_head(&fuidp->z_fuids); zfuid;
189 	    zfuid = list_next(&fuidp->z_fuids, zfuid)) {
190 		*fuidloc++ = zfuid->z_logfuid;
191 	}
192 	return (fuidloc);
193 }
194 
195 
196 static void *
zfs_log_fuid_domains(zfs_fuid_info_t * fuidp,void * start)197 zfs_log_fuid_domains(zfs_fuid_info_t *fuidp, void *start)
198 {
199 	zfs_fuid_domain_t *zdomain;
200 
201 	/* now copy in the domain info, if any */
202 	if (fuidp->z_domain_str_sz != 0) {
203 		for (zdomain = list_head(&fuidp->z_domains); zdomain;
204 		    zdomain = list_next(&fuidp->z_domains, zdomain)) {
205 			bcopy((void *)zdomain->z_domain, start,
206 			    strlen(zdomain->z_domain) + 1);
207 			start = (caddr_t)start +
208 			    strlen(zdomain->z_domain) + 1;
209 		}
210 	}
211 	return (start);
212 }
213 
214 /*
215  * Handles TX_CREATE, TX_CREATE_ATTR, TX_MKDIR, TX_MKDIR_ATTR and
216  * TK_MKXATTR 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
zfs_log_create(zilog_t * zilog,dmu_tx_t * tx,uint64_t txtype,znode_t * dzp,znode_t * zp,char * name,vsecattr_t * vsecp,zfs_fuid_info_t * fuidp,vattr_t * vap)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 = (vsecp != NULL) ? vsecp->vsa_aclentsz : 0;
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 		txsize =
272 		    sizeof (lr_acl_create_t) + namesize + fuidsz +
273 		    ZIL_ACE_LENGTH(aclsize) + xvatsize;
274 		lrsize = sizeof (lr_acl_create_t);
275 	}
276 
277 	itx = zil_itx_create(txtype, txsize);
278 
279 	lr = (lr_create_t *)&itx->itx_lr;
280 	lr->lr_doid = dzp->z_id;
281 	lr->lr_foid = zp->z_id;
282 	lr->lr_mode = zp->z_mode;
283 	if (!IS_EPHEMERAL(zp->z_uid)) {
284 		lr->lr_uid = (uint64_t)zp->z_uid;
285 	} else {
286 		lr->lr_uid = fuidp->z_fuid_owner;
287 	}
288 	if (!IS_EPHEMERAL(zp->z_gid)) {
289 		lr->lr_gid = (uint64_t)zp->z_gid;
290 	} else {
291 		lr->lr_gid = fuidp->z_fuid_group;
292 	}
293 	(void) sa_lookup(zp->z_sa_hdl, SA_ZPL_GEN(zp->z_zfsvfs), &lr->lr_gen,
294 	    sizeof (uint64_t));
295 	(void) sa_lookup(zp->z_sa_hdl, SA_ZPL_CRTIME(zp->z_zfsvfs),
296 	    lr->lr_crtime, sizeof (uint64_t) * 2);
297 
298 	if (sa_lookup(zp->z_sa_hdl, SA_ZPL_RDEV(zp->z_zfsvfs), &lr->lr_rdev,
299 	    sizeof (lr->lr_rdev)) != 0)
300 		lr->lr_rdev = 0;
301 
302 	/*
303 	 * Fill in xvattr info if any
304 	 */
305 	if (vap->va_mask & AT_XVATTR) {
306 		zfs_log_xvattr((lr_attr_t *)((caddr_t)lr + lrsize), xvap);
307 		end = (caddr_t)lr + lrsize + xvatsize;
308 	} else {
309 		end = (caddr_t)lr + lrsize;
310 	}
311 
312 	/* Now fill in any ACL info */
313 
314 	if (vsecp) {
315 		lracl = (lr_acl_create_t *)&itx->itx_lr;
316 		lracl->lr_aclcnt = vsecp->vsa_aclcnt;
317 		lracl->lr_acl_bytes = aclsize;
318 		lracl->lr_domcnt = fuidp ? fuidp->z_domain_cnt : 0;
319 		lracl->lr_fuidcnt  = fuidp ? fuidp->z_fuid_cnt : 0;
320 		if (vsecp->vsa_aclflags & VSA_ACE_ACLFLAGS)
321 			lracl->lr_acl_flags = (uint64_t)vsecp->vsa_aclflags;
322 		else
323 			lracl->lr_acl_flags = 0;
324 
325 		bcopy(vsecp->vsa_aclentp, end, aclsize);
326 		end = (caddr_t)end + ZIL_ACE_LENGTH(aclsize);
327 	}
328 
329 	/* drop in FUID info */
330 	if (fuidp) {
331 		end = zfs_log_fuid_ids(fuidp, end);
332 		end = zfs_log_fuid_domains(fuidp, end);
333 	}
334 	/*
335 	 * Now place file name in log record
336 	 */
337 	bcopy(name, end, namesize);
338 
339 	zil_itx_assign(zilog, itx, tx);
340 
341 	rw_enter(&rz_zev_rwlock, RW_READER);
342 	if (rz_zev_callbacks && rz_zev_callbacks->rz_zev_znode_create)
343 		rz_zev_callbacks->rz_zev_znode_create(dzp, zp, tx, name,
344 		                                      txtype);
345 	rw_exit(&rz_zev_rwlock);
346 }
347 
348 /*
349  * Handles both TX_REMOVE and TX_RMDIR transactions.
350  */
351 void
zfs_log_remove(zilog_t * zilog,dmu_tx_t * tx,uint64_t txtype,znode_t * dzp,char * name,uint64_t foid)352 zfs_log_remove(zilog_t *zilog, dmu_tx_t *tx, uint64_t txtype,
353 	znode_t *dzp, char *name, uint64_t foid)
354 {
355 	itx_t *itx;
356 	lr_remove_t *lr;
357 	size_t namesize = strlen(name) + 1;
358 
359 	if (zil_replaying(zilog, tx))
360 		return;
361 
362 	itx = zil_itx_create(txtype, sizeof (*lr) + namesize);
363 	lr = (lr_remove_t *)&itx->itx_lr;
364 	lr->lr_doid = dzp->z_id;
365 	bcopy(name, (char *)(lr + 1), namesize);
366 
367 	itx->itx_oid = foid;
368 
369 	zil_itx_assign(zilog, itx, tx);
370 }
371 
372 /*
373  * Handles TX_LINK transactions.
374  */
375 void
zfs_log_link(zilog_t * zilog,dmu_tx_t * tx,uint64_t txtype,znode_t * dzp,znode_t * zp,char * name)376 zfs_log_link(zilog_t *zilog, dmu_tx_t *tx, uint64_t txtype,
377 	znode_t *dzp, znode_t *zp, char *name)
378 {
379 	itx_t *itx;
380 	lr_link_t *lr;
381 	size_t namesize = strlen(name) + 1;
382 
383 	if (zil_replaying(zilog, tx))
384 		return;
385 
386 	itx = zil_itx_create(txtype, sizeof (*lr) + namesize);
387 	lr = (lr_link_t *)&itx->itx_lr;
388 	lr->lr_doid = dzp->z_id;
389 	lr->lr_link_obj = zp->z_id;
390 	bcopy(name, (char *)(lr + 1), namesize);
391 
392 	zil_itx_assign(zilog, itx, tx);
393 
394 	rw_enter(&rz_zev_rwlock, RW_READER);
395 	if (rz_zev_callbacks && rz_zev_callbacks->rz_zev_znode_link)
396 		rz_zev_callbacks->rz_zev_znode_link(dzp, zp, tx, name);
397 	rw_exit(&rz_zev_rwlock);
398 }
399 
400 /*
401  * Handles TX_SYMLINK transactions.
402  */
403 void
zfs_log_symlink(zilog_t * zilog,dmu_tx_t * tx,uint64_t txtype,znode_t * dzp,znode_t * zp,char * name,char * link)404 zfs_log_symlink(zilog_t *zilog, dmu_tx_t *tx, uint64_t txtype,
405     znode_t *dzp, znode_t *zp, char *name, char *link)
406 {
407 	itx_t *itx;
408 	lr_create_t *lr;
409 	size_t namesize = strlen(name) + 1;
410 	size_t linksize = strlen(link) + 1;
411 
412 	if (zil_replaying(zilog, tx))
413 		return;
414 
415 	itx = zil_itx_create(txtype, sizeof (*lr) + namesize + linksize);
416 	lr = (lr_create_t *)&itx->itx_lr;
417 	lr->lr_doid = dzp->z_id;
418 	lr->lr_foid = zp->z_id;
419 	lr->lr_uid = zp->z_uid;
420 	lr->lr_gid = zp->z_gid;
421 	lr->lr_mode = zp->z_mode;
422 	(void) sa_lookup(zp->z_sa_hdl, SA_ZPL_GEN(zp->z_zfsvfs), &lr->lr_gen,
423 	    sizeof (uint64_t));
424 	(void) sa_lookup(zp->z_sa_hdl, SA_ZPL_CRTIME(zp->z_zfsvfs),
425 	    lr->lr_crtime, sizeof (uint64_t) * 2);
426 	bcopy(name, (char *)(lr + 1), namesize);
427 	bcopy(link, (char *)(lr + 1) + namesize, linksize);
428 
429 	zil_itx_assign(zilog, itx, tx);
430 
431 	rw_enter(&rz_zev_rwlock, RW_READER);
432 	if (rz_zev_callbacks && rz_zev_callbacks->rz_zev_znode_symlink)
433 		rz_zev_callbacks->rz_zev_znode_symlink(dzp, zp, tx, name, link);
434 	rw_exit(&rz_zev_rwlock);
435 }
436 
437 /*
438  * Handles TX_RENAME transactions.
439  */
440 void
zfs_log_rename(zilog_t * zilog,dmu_tx_t * tx,uint64_t txtype,znode_t * sdzp,char * sname,znode_t * tdzp,char * dname,znode_t * szp,znode_t * tzp)441 zfs_log_rename(zilog_t *zilog, dmu_tx_t *tx, uint64_t txtype,
442 	znode_t *sdzp, char *sname, znode_t *tdzp, char *dname,
443 	znode_t *szp, znode_t *tzp)
444 {
445 	itx_t *itx;
446 	lr_rename_t *lr;
447 	size_t snamesize = strlen(sname) + 1;
448 	size_t dnamesize = strlen(dname) + 1;
449 
450 	if (zil_replaying(zilog, tx))
451 		return;
452 
453 	itx = zil_itx_create(txtype, sizeof (*lr) + snamesize + dnamesize);
454 	lr = (lr_rename_t *)&itx->itx_lr;
455 	lr->lr_sdoid = sdzp->z_id;
456 	lr->lr_tdoid = tdzp->z_id;
457 	bcopy(sname, (char *)(lr + 1), snamesize);
458 	bcopy(dname, (char *)(lr + 1) + snamesize, dnamesize);
459 	itx->itx_oid = szp->z_id;
460 
461 	zil_itx_assign(zilog, itx, tx);
462 
463 	rw_enter(&rz_zev_rwlock, RW_READER);
464 	if (rz_zev_callbacks && rz_zev_callbacks->rz_zev_znode_rename)
465 		rz_zev_callbacks->rz_zev_znode_rename(sdzp,
466 			sname, tdzp, dname, szp, tzp, tx);
467 	rw_exit(&rz_zev_rwlock);
468 }
469 
470 /*
471  * Handles TX_WRITE transactions.
472  */
473 ssize_t zfs_immediate_write_sz = 32768;
474 
475 void
zfs_log_write(zilog_t * zilog,dmu_tx_t * tx,int txtype,znode_t * zp,offset_t off,ssize_t resid,int ioflag)476 zfs_log_write(zilog_t *zilog, dmu_tx_t *tx, int txtype,
477 	znode_t *zp, offset_t off, ssize_t resid, int ioflag)
478 {
479 	itx_wr_state_t write_state;
480 	boolean_t slogging;
481 	uintptr_t fsync_cnt;
482 	ssize_t immediate_write_sz;
483 
484 	if (zil_replaying(zilog, tx) || zp->z_unlinked)
485 		return;
486 
487 	immediate_write_sz = (zilog->zl_logbias == ZFS_LOGBIAS_THROUGHPUT)
488 	    ? 0 : zfs_immediate_write_sz;
489 
490 	slogging = spa_has_slogs(zilog->zl_spa) &&
491 	    (zilog->zl_logbias == ZFS_LOGBIAS_LATENCY);
492 	if (resid > immediate_write_sz && !slogging && resid <= zp->z_blksz)
493 		write_state = WR_INDIRECT;
494 	else if (ioflag & (FSYNC | FDSYNC))
495 		write_state = WR_COPIED;
496 	else
497 		write_state = WR_NEED_COPY;
498 
499 	if ((fsync_cnt = (uintptr_t)tsd_get(zfs_fsyncer_key)) != 0) {
500 		(void) tsd_set(zfs_fsyncer_key, (void *)(fsync_cnt - 1));
501 	}
502 
503 	while (resid) {
504 		itx_t *itx;
505 		lr_write_t *lr;
506 		ssize_t len;
507 
508 		/*
509 		 * If the write would overflow the largest block then split it.
510 		 */
511 		if (write_state != WR_INDIRECT && resid > ZIL_MAX_LOG_DATA)
512 			len = SPA_OLD_MAXBLOCKSIZE >> 1;
513 		else
514 			len = resid;
515 
516 		itx = zil_itx_create(txtype, sizeof (*lr) +
517 		    (write_state == WR_COPIED ? len : 0));
518 		lr = (lr_write_t *)&itx->itx_lr;
519 		if (write_state == WR_COPIED && dmu_read(zp->z_zfsvfs->z_os,
520 		    zp->z_id, off, len, lr + 1, DMU_READ_NO_PREFETCH) != 0) {
521 			zil_itx_destroy(itx);
522 			itx = zil_itx_create(txtype, sizeof (*lr));
523 			lr = (lr_write_t *)&itx->itx_lr;
524 			write_state = WR_NEED_COPY;
525 		}
526 
527 		itx->itx_wr_state = write_state;
528 		if (write_state == WR_NEED_COPY)
529 			itx->itx_sod += len;
530 		lr->lr_foid = zp->z_id;
531 		lr->lr_offset = off;
532 		lr->lr_length = len;
533 		lr->lr_blkoff = 0;
534 		BP_ZERO(&lr->lr_blkptr);
535 
536 		itx->itx_private = zp->z_zfsvfs;
537 
538 		if (!(ioflag & (FSYNC | FDSYNC)) && (zp->z_sync_cnt == 0) &&
539 		    (fsync_cnt == 0))
540 			itx->itx_sync = B_FALSE;
541 
542 		zil_itx_assign(zilog, itx, tx);
543 
544 		if (ZTOV(zp)->v_type == VREG)
545 			atomic_add_32(&(zp->z_new_content), 1);
546 		rw_enter(&rz_zev_rwlock, RW_READER);
547 		if (rz_zev_callbacks && rz_zev_callbacks->rz_zev_znode_write)
548 			rz_zev_callbacks->rz_zev_znode_write(zp, tx, off, len);
549 		rw_exit(&rz_zev_rwlock);
550 
551 		off += len;
552 		resid -= len;
553 	}
554 }
555 
556 /*
557  * Handles TX_TRUNCATE transactions.
558  */
559 void
zfs_log_truncate(zilog_t * zilog,dmu_tx_t * tx,int txtype,znode_t * zp,uint64_t off,uint64_t len)560 zfs_log_truncate(zilog_t *zilog, dmu_tx_t *tx, int txtype,
561 	znode_t *zp, uint64_t off, uint64_t len)
562 {
563 	itx_t *itx;
564 	lr_truncate_t *lr;
565 
566 	if (zil_replaying(zilog, tx) || zp->z_unlinked)
567 		return;
568 
569 	itx = zil_itx_create(txtype, sizeof (*lr));
570 	lr = (lr_truncate_t *)&itx->itx_lr;
571 	lr->lr_foid = zp->z_id;
572 	lr->lr_offset = off;
573 	lr->lr_length = len;
574 
575 	itx->itx_sync = (zp->z_sync_cnt != 0);
576 	zil_itx_assign(zilog, itx, tx);
577 
578 	if (ZTOV(zp)->v_type == VREG)
579 		atomic_add_32(&(zp->z_new_content), 1);
580 	rw_enter(&rz_zev_rwlock, RW_READER);
581 	if (rz_zev_callbacks && rz_zev_callbacks->rz_zev_znode_truncate)
582 		rz_zev_callbacks->rz_zev_znode_truncate(zp, tx, off, len);
583 	rw_exit(&rz_zev_rwlock);
584 }
585 
586 /*
587  * Handles TX_SETATTR transactions.
588  */
589 void
zfs_log_setattr(zilog_t * zilog,dmu_tx_t * tx,int txtype,znode_t * zp,vattr_t * vap,uint_t mask_applied,zfs_fuid_info_t * fuidp)590 zfs_log_setattr(zilog_t *zilog, dmu_tx_t *tx, int txtype,
591 	znode_t *zp, vattr_t *vap, uint_t mask_applied, zfs_fuid_info_t *fuidp)
592 {
593 	itx_t		*itx;
594 	lr_setattr_t	*lr;
595 	xvattr_t	*xvap = (xvattr_t *)vap;
596 	size_t		recsize = sizeof (lr_setattr_t);
597 	void		*start;
598 
599 	if (zil_replaying(zilog, tx) || zp->z_unlinked)
600 		return;
601 
602 	/*
603 	 * If XVATTR set, then log record size needs to allow
604 	 * for lr_attr_t + xvattr mask, mapsize and create time
605 	 * plus actual attribute values
606 	 */
607 	if (vap->va_mask & AT_XVATTR)
608 		recsize = sizeof (*lr) + ZIL_XVAT_SIZE(xvap->xva_mapsize);
609 
610 	if (fuidp)
611 		recsize += fuidp->z_domain_str_sz;
612 
613 	itx = zil_itx_create(txtype, recsize);
614 	lr = (lr_setattr_t *)&itx->itx_lr;
615 	lr->lr_foid = zp->z_id;
616 	lr->lr_mask = (uint64_t)mask_applied;
617 	lr->lr_mode = (uint64_t)vap->va_mode;
618 	if ((mask_applied & AT_UID) && IS_EPHEMERAL(vap->va_uid))
619 		lr->lr_uid = fuidp->z_fuid_owner;
620 	else
621 		lr->lr_uid = (uint64_t)vap->va_uid;
622 
623 	if ((mask_applied & AT_GID) && IS_EPHEMERAL(vap->va_gid))
624 		lr->lr_gid = fuidp->z_fuid_group;
625 	else
626 		lr->lr_gid = (uint64_t)vap->va_gid;
627 
628 	lr->lr_size = (uint64_t)vap->va_size;
629 	ZFS_TIME_ENCODE(&vap->va_atime, lr->lr_atime);
630 	ZFS_TIME_ENCODE(&vap->va_mtime, lr->lr_mtime);
631 	start = (lr_setattr_t *)(lr + 1);
632 	if (vap->va_mask & AT_XVATTR) {
633 		zfs_log_xvattr((lr_attr_t *)start, xvap);
634 		start = (caddr_t)start + ZIL_XVAT_SIZE(xvap->xva_mapsize);
635 	}
636 
637 	/*
638 	 * Now stick on domain information if any on end
639 	 */
640 
641 	if (fuidp)
642 		(void) zfs_log_fuid_domains(fuidp, start);
643 
644 	itx->itx_sync = (zp->z_sync_cnt != 0);
645 	zil_itx_assign(zilog, itx, tx);
646 }
647 
648 /*
649  * Handles TX_ACL transactions.
650  */
651 void
zfs_log_acl(zilog_t * zilog,dmu_tx_t * tx,znode_t * zp,vsecattr_t * vsecp,zfs_fuid_info_t * fuidp)652 zfs_log_acl(zilog_t *zilog, dmu_tx_t *tx, znode_t *zp,
653     vsecattr_t *vsecp, zfs_fuid_info_t *fuidp)
654 {
655 	itx_t *itx;
656 	lr_acl_v0_t *lrv0;
657 	lr_acl_t *lr;
658 	int txtype;
659 	int lrsize;
660 	size_t txsize;
661 	size_t aclbytes = vsecp->vsa_aclentsz;
662 
663 	if (zil_replaying(zilog, tx) || zp->z_unlinked)
664 		return;
665 
666 	txtype = (zp->z_zfsvfs->z_version < ZPL_VERSION_FUID) ?
667 	    TX_ACL_V0 : TX_ACL;
668 
669 	if (txtype == TX_ACL)
670 		lrsize = sizeof (*lr);
671 	else
672 		lrsize = sizeof (*lrv0);
673 
674 	txsize = lrsize +
675 	    ((txtype == TX_ACL) ? ZIL_ACE_LENGTH(aclbytes) : aclbytes) +
676 	    (fuidp ? fuidp->z_domain_str_sz : 0) +
677 	    sizeof (uint64_t) * (fuidp ? fuidp->z_fuid_cnt : 0);
678 
679 	itx = zil_itx_create(txtype, txsize);
680 
681 	lr = (lr_acl_t *)&itx->itx_lr;
682 	lr->lr_foid = zp->z_id;
683 	if (txtype == TX_ACL) {
684 		lr->lr_acl_bytes = aclbytes;
685 		lr->lr_domcnt = fuidp ? fuidp->z_domain_cnt : 0;
686 		lr->lr_fuidcnt = fuidp ? fuidp->z_fuid_cnt : 0;
687 		if (vsecp->vsa_mask & VSA_ACE_ACLFLAGS)
688 			lr->lr_acl_flags = (uint64_t)vsecp->vsa_aclflags;
689 		else
690 			lr->lr_acl_flags = 0;
691 	}
692 	lr->lr_aclcnt = (uint64_t)vsecp->vsa_aclcnt;
693 
694 	if (txtype == TX_ACL_V0) {
695 		lrv0 = (lr_acl_v0_t *)lr;
696 		bcopy(vsecp->vsa_aclentp, (ace_t *)(lrv0 + 1), aclbytes);
697 	} else {
698 		void *start = (ace_t *)(lr + 1);
699 
700 		bcopy(vsecp->vsa_aclentp, start, aclbytes);
701 
702 		start = (caddr_t)start + ZIL_ACE_LENGTH(aclbytes);
703 
704 		if (fuidp) {
705 			start = zfs_log_fuid_ids(fuidp, start);
706 			(void) zfs_log_fuid_domains(fuidp, start);
707 		}
708 	}
709 
710 	itx->itx_sync = (zp->z_sync_cnt != 0);
711 	zil_itx_assign(zilog, itx, tx);
712 
713 	rw_enter(&rz_zev_rwlock, RW_READER);
714 	if (rz_zev_callbacks && rz_zev_callbacks->rz_zev_znode_acl)
715 		rz_zev_callbacks->rz_zev_znode_acl(zp, tx);
716 	rw_exit(&rz_zev_rwlock);
717 }
718