xref: /titanic_41/usr/src/uts/common/fs/zfs/zfs_replay.c (revision 982e63f0dc541ca7cc5d69cf6504d5218a48434d)
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) 2013 by Delphix. All rights reserved.
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/fcntl.h>
35 #include <sys/vfs.h>
36 #include <sys/fs/zfs.h>
37 #include <sys/zfs_znode.h>
38 #include <sys/zfs_dir.h>
39 #include <sys/zfs_acl.h>
40 #include <sys/zfs_fuid.h>
41 #include <sys/spa.h>
42 #include <sys/zil.h>
43 #include <sys/byteorder.h>
44 #include <sys/stat.h>
45 #include <sys/mode.h>
46 #include <sys/acl.h>
47 #include <sys/atomic.h>
48 #include <sys/cred.h>
49 
50 /*
51  * Functions to replay ZFS intent log (ZIL) records
52  * The functions are called through a function vector (zfs_replay_vector)
53  * which is indexed by the transaction type.
54  */
55 
56 static void
57 zfs_init_vattr(vattr_t *vap, uint64_t mask, uint64_t mode,
58 	uint64_t uid, uint64_t gid, uint64_t rdev, uint64_t nodeid)
59 {
60 	bzero(vap, sizeof (*vap));
61 	vap->va_mask = (uint_t)mask;
62 	vap->va_type = IFTOVT(mode);
63 	vap->va_mode = mode & MODEMASK;
64 	vap->va_uid = (uid_t)(IS_EPHEMERAL(uid)) ? -1 : uid;
65 	vap->va_gid = (gid_t)(IS_EPHEMERAL(gid)) ? -1 : gid;
66 	vap->va_rdev = zfs_cmpldev(rdev);
67 	vap->va_nodeid = nodeid;
68 }
69 
70 /* ARGSUSED */
71 static int
72 zfs_replay_error(zfsvfs_t *zfsvfs, lr_t *lr, boolean_t byteswap)
73 {
74 	return (SET_ERROR(ENOTSUP));
75 }
76 
77 static void
78 zfs_replay_xvattr(lr_attr_t *lrattr, xvattr_t *xvap)
79 {
80 	xoptattr_t *xoap = NULL;
81 	uint64_t *attrs;
82 	uint64_t *crtime;
83 	uint32_t *bitmap;
84 	void *scanstamp;
85 	int i;
86 
87 	xvap->xva_vattr.va_mask |= AT_XVATTR;
88 	if ((xoap = xva_getxoptattr(xvap)) == NULL) {
89 		xvap->xva_vattr.va_mask &= ~AT_XVATTR; /* shouldn't happen */
90 		return;
91 	}
92 
93 	ASSERT(lrattr->lr_attr_masksize == xvap->xva_mapsize);
94 
95 	bitmap = &lrattr->lr_attr_bitmap;
96 	for (i = 0; i != lrattr->lr_attr_masksize; i++, bitmap++)
97 		xvap->xva_reqattrmap[i] = *bitmap;
98 
99 	attrs = (uint64_t *)(lrattr + lrattr->lr_attr_masksize - 1);
100 	crtime = attrs + 1;
101 	scanstamp = (caddr_t)(crtime + 2);
102 
103 	if (XVA_ISSET_REQ(xvap, XAT_HIDDEN))
104 		xoap->xoa_hidden = ((*attrs & XAT0_HIDDEN) != 0);
105 	if (XVA_ISSET_REQ(xvap, XAT_SYSTEM))
106 		xoap->xoa_system = ((*attrs & XAT0_SYSTEM) != 0);
107 	if (XVA_ISSET_REQ(xvap, XAT_ARCHIVE))
108 		xoap->xoa_archive = ((*attrs & XAT0_ARCHIVE) != 0);
109 	if (XVA_ISSET_REQ(xvap, XAT_READONLY))
110 		xoap->xoa_readonly = ((*attrs & XAT0_READONLY) != 0);
111 	if (XVA_ISSET_REQ(xvap, XAT_IMMUTABLE))
112 		xoap->xoa_immutable = ((*attrs & XAT0_IMMUTABLE) != 0);
113 	if (XVA_ISSET_REQ(xvap, XAT_NOUNLINK))
114 		xoap->xoa_nounlink = ((*attrs & XAT0_NOUNLINK) != 0);
115 	if (XVA_ISSET_REQ(xvap, XAT_APPENDONLY))
116 		xoap->xoa_appendonly = ((*attrs & XAT0_APPENDONLY) != 0);
117 	if (XVA_ISSET_REQ(xvap, XAT_NODUMP))
118 		xoap->xoa_nodump = ((*attrs & XAT0_NODUMP) != 0);
119 	if (XVA_ISSET_REQ(xvap, XAT_OPAQUE))
120 		xoap->xoa_opaque = ((*attrs & XAT0_OPAQUE) != 0);
121 	if (XVA_ISSET_REQ(xvap, XAT_AV_MODIFIED))
122 		xoap->xoa_av_modified = ((*attrs & XAT0_AV_MODIFIED) != 0);
123 	if (XVA_ISSET_REQ(xvap, XAT_AV_QUARANTINED))
124 		xoap->xoa_av_quarantined =
125 		    ((*attrs & XAT0_AV_QUARANTINED) != 0);
126 	if (XVA_ISSET_REQ(xvap, XAT_CREATETIME))
127 		ZFS_TIME_DECODE(&xoap->xoa_createtime, crtime);
128 	if (XVA_ISSET_REQ(xvap, XAT_AV_SCANSTAMP))
129 		bcopy(scanstamp, xoap->xoa_av_scanstamp, AV_SCANSTAMP_SZ);
130 	if (XVA_ISSET_REQ(xvap, XAT_REPARSE))
131 		xoap->xoa_reparse = ((*attrs & XAT0_REPARSE) != 0);
132 	if (XVA_ISSET_REQ(xvap, XAT_OFFLINE))
133 		xoap->xoa_offline = ((*attrs & XAT0_OFFLINE) != 0);
134 	if (XVA_ISSET_REQ(xvap, XAT_SPARSE))
135 		xoap->xoa_sparse = ((*attrs & XAT0_SPARSE) != 0);
136 }
137 
138 static int
139 zfs_replay_domain_cnt(uint64_t uid, uint64_t gid)
140 {
141 	uint64_t uid_idx;
142 	uint64_t gid_idx;
143 	int domcnt = 0;
144 
145 	uid_idx = FUID_INDEX(uid);
146 	gid_idx = FUID_INDEX(gid);
147 	if (uid_idx)
148 		domcnt++;
149 	if (gid_idx > 0 && gid_idx != uid_idx)
150 		domcnt++;
151 
152 	return (domcnt);
153 }
154 
155 static void *
156 zfs_replay_fuid_domain_common(zfs_fuid_info_t *fuid_infop, void *start,
157     int domcnt)
158 {
159 	int i;
160 
161 	for (i = 0; i != domcnt; i++) {
162 		fuid_infop->z_domain_table[i] = start;
163 		start = (caddr_t)start + strlen(start) + 1;
164 	}
165 
166 	return (start);
167 }
168 
169 /*
170  * Set the uid/gid in the fuid_info structure.
171  */
172 static void
173 zfs_replay_fuid_ugid(zfs_fuid_info_t *fuid_infop, uint64_t uid, uint64_t gid)
174 {
175 	/*
176 	 * If owner or group are log specific FUIDs then slurp up
177 	 * domain information and build zfs_fuid_info_t
178 	 */
179 	if (IS_EPHEMERAL(uid))
180 		fuid_infop->z_fuid_owner = uid;
181 
182 	if (IS_EPHEMERAL(gid))
183 		fuid_infop->z_fuid_group = gid;
184 }
185 
186 /*
187  * Load fuid domains into fuid_info_t
188  */
189 static zfs_fuid_info_t *
190 zfs_replay_fuid_domain(void *buf, void **end, uint64_t uid, uint64_t gid)
191 {
192 	int domcnt;
193 
194 	zfs_fuid_info_t *fuid_infop;
195 
196 	fuid_infop = zfs_fuid_info_alloc();
197 
198 	domcnt = zfs_replay_domain_cnt(uid, gid);
199 
200 	if (domcnt == 0)
201 		return (fuid_infop);
202 
203 	fuid_infop->z_domain_table =
204 	    kmem_zalloc(domcnt * sizeof (char **), KM_SLEEP);
205 
206 	zfs_replay_fuid_ugid(fuid_infop, uid, gid);
207 
208 	fuid_infop->z_domain_cnt = domcnt;
209 	*end = zfs_replay_fuid_domain_common(fuid_infop, buf, domcnt);
210 	return (fuid_infop);
211 }
212 
213 /*
214  * load zfs_fuid_t's and fuid_domains into fuid_info_t
215  */
216 static zfs_fuid_info_t *
217 zfs_replay_fuids(void *start, void **end, int idcnt, int domcnt, uint64_t uid,
218     uint64_t gid)
219 {
220 	uint64_t *log_fuid = (uint64_t *)start;
221 	zfs_fuid_info_t *fuid_infop;
222 	int i;
223 
224 	fuid_infop = zfs_fuid_info_alloc();
225 	fuid_infop->z_domain_cnt = domcnt;
226 
227 	fuid_infop->z_domain_table =
228 	    kmem_zalloc(domcnt * sizeof (char **), KM_SLEEP);
229 
230 	for (i = 0; i != idcnt; i++) {
231 		zfs_fuid_t *zfuid;
232 
233 		zfuid = kmem_alloc(sizeof (zfs_fuid_t), KM_SLEEP);
234 		zfuid->z_logfuid = *log_fuid;
235 		zfuid->z_id = -1;
236 		zfuid->z_domidx = 0;
237 		list_insert_tail(&fuid_infop->z_fuids, zfuid);
238 		log_fuid++;
239 	}
240 
241 	zfs_replay_fuid_ugid(fuid_infop, uid, gid);
242 
243 	*end = zfs_replay_fuid_domain_common(fuid_infop, log_fuid, domcnt);
244 	return (fuid_infop);
245 }
246 
247 static void
248 zfs_replay_swap_attrs(lr_attr_t *lrattr)
249 {
250 	/* swap the lr_attr structure */
251 	byteswap_uint32_array(lrattr, sizeof (*lrattr));
252 	/* swap the bitmap */
253 	byteswap_uint32_array(lrattr + 1, (lrattr->lr_attr_masksize - 1) *
254 	    sizeof (uint32_t));
255 	/* swap the attributes, create time + 64 bit word for attributes */
256 	byteswap_uint64_array((caddr_t)(lrattr + 1) + (sizeof (uint32_t) *
257 	    (lrattr->lr_attr_masksize - 1)), 3 * sizeof (uint64_t));
258 }
259 
260 /*
261  * Replay file create with optional ACL, xvattr information as well
262  * as option FUID information.
263  */
264 static int
265 zfs_replay_create_acl(zfsvfs_t *zfsvfs,
266     lr_acl_create_t *lracl, boolean_t byteswap)
267 {
268 	char *name = NULL;		/* location determined later */
269 	lr_create_t *lr = (lr_create_t *)lracl;
270 	znode_t *dzp;
271 	vnode_t *vp = NULL;
272 	xvattr_t xva;
273 	int vflg = 0;
274 	vsecattr_t vsec = { 0 };
275 	lr_attr_t *lrattr;
276 	void *aclstart;
277 	void *fuidstart;
278 	size_t xvatlen = 0;
279 	uint64_t txtype;
280 	int error;
281 
282 	txtype = (lr->lr_common.lrc_txtype & ~TX_CI);
283 	if (byteswap) {
284 		byteswap_uint64_array(lracl, sizeof (*lracl));
285 		if (txtype == TX_CREATE_ACL_ATTR ||
286 		    txtype == TX_MKDIR_ACL_ATTR) {
287 			lrattr = (lr_attr_t *)(caddr_t)(lracl + 1);
288 			zfs_replay_swap_attrs(lrattr);
289 			xvatlen = ZIL_XVAT_SIZE(lrattr->lr_attr_masksize);
290 		}
291 
292 		aclstart = (caddr_t)(lracl + 1) + xvatlen;
293 		zfs_ace_byteswap(aclstart, lracl->lr_acl_bytes, B_FALSE);
294 		/* swap fuids */
295 		if (lracl->lr_fuidcnt) {
296 			byteswap_uint64_array((caddr_t)aclstart +
297 			    ZIL_ACE_LENGTH(lracl->lr_acl_bytes),
298 			    lracl->lr_fuidcnt * sizeof (uint64_t));
299 		}
300 	}
301 
302 	if ((error = zfs_zget(zfsvfs, lr->lr_doid, &dzp)) != 0)
303 		return (error);
304 
305 	xva_init(&xva);
306 	zfs_init_vattr(&xva.xva_vattr, AT_TYPE | AT_MODE | AT_UID | AT_GID,
307 	    lr->lr_mode, lr->lr_uid, lr->lr_gid, lr->lr_rdev, lr->lr_foid);
308 
309 	/*
310 	 * All forms of zfs create (create, mkdir, mkxattrdir, symlink)
311 	 * eventually end up in zfs_mknode(), which assigns the object's
312 	 * creation time and generation number.  The generic VOP_CREATE()
313 	 * doesn't have either concept, so we smuggle the values inside
314 	 * the vattr's otherwise unused va_ctime and va_nblocks fields.
315 	 */
316 	ZFS_TIME_DECODE(&xva.xva_vattr.va_ctime, lr->lr_crtime);
317 	xva.xva_vattr.va_nblocks = lr->lr_gen;
318 
319 	error = dmu_object_info(zfsvfs->z_os, lr->lr_foid, NULL);
320 	if (error != ENOENT)
321 		goto bail;
322 
323 	if (lr->lr_common.lrc_txtype & TX_CI)
324 		vflg |= FIGNORECASE;
325 	switch (txtype) {
326 	case TX_CREATE_ACL:
327 		aclstart = (caddr_t)(lracl + 1);
328 		fuidstart = (caddr_t)aclstart +
329 		    ZIL_ACE_LENGTH(lracl->lr_acl_bytes);
330 		zfsvfs->z_fuid_replay = zfs_replay_fuids(fuidstart,
331 		    (void *)&name, lracl->lr_fuidcnt, lracl->lr_domcnt,
332 		    lr->lr_uid, lr->lr_gid);
333 		/*FALLTHROUGH*/
334 	case TX_CREATE_ACL_ATTR:
335 		if (name == NULL) {
336 			lrattr = (lr_attr_t *)(caddr_t)(lracl + 1);
337 			xvatlen = ZIL_XVAT_SIZE(lrattr->lr_attr_masksize);
338 			xva.xva_vattr.va_mask |= AT_XVATTR;
339 			zfs_replay_xvattr(lrattr, &xva);
340 		}
341 		vsec.vsa_mask = VSA_ACE | VSA_ACE_ACLFLAGS;
342 		vsec.vsa_aclentp = (caddr_t)(lracl + 1) + xvatlen;
343 		vsec.vsa_aclcnt = lracl->lr_aclcnt;
344 		vsec.vsa_aclentsz = lracl->lr_acl_bytes;
345 		vsec.vsa_aclflags = lracl->lr_acl_flags;
346 		if (zfsvfs->z_fuid_replay == NULL) {
347 			fuidstart = (caddr_t)(lracl + 1) + xvatlen +
348 			    ZIL_ACE_LENGTH(lracl->lr_acl_bytes);
349 			zfsvfs->z_fuid_replay =
350 			    zfs_replay_fuids(fuidstart,
351 			    (void *)&name, lracl->lr_fuidcnt, lracl->lr_domcnt,
352 			    lr->lr_uid, lr->lr_gid);
353 		}
354 
355 		error = VOP_CREATE(ZTOV(dzp), name, &xva.xva_vattr,
356 		    0, 0, &vp, kcred, vflg, NULL, &vsec);
357 		break;
358 	case TX_MKDIR_ACL:
359 		aclstart = (caddr_t)(lracl + 1);
360 		fuidstart = (caddr_t)aclstart +
361 		    ZIL_ACE_LENGTH(lracl->lr_acl_bytes);
362 		zfsvfs->z_fuid_replay = zfs_replay_fuids(fuidstart,
363 		    (void *)&name, lracl->lr_fuidcnt, lracl->lr_domcnt,
364 		    lr->lr_uid, lr->lr_gid);
365 		/*FALLTHROUGH*/
366 	case TX_MKDIR_ACL_ATTR:
367 		if (name == NULL) {
368 			lrattr = (lr_attr_t *)(caddr_t)(lracl + 1);
369 			xvatlen = ZIL_XVAT_SIZE(lrattr->lr_attr_masksize);
370 			zfs_replay_xvattr(lrattr, &xva);
371 		}
372 		vsec.vsa_mask = VSA_ACE | VSA_ACE_ACLFLAGS;
373 		vsec.vsa_aclentp = (caddr_t)(lracl + 1) + xvatlen;
374 		vsec.vsa_aclcnt = lracl->lr_aclcnt;
375 		vsec.vsa_aclentsz = lracl->lr_acl_bytes;
376 		vsec.vsa_aclflags = lracl->lr_acl_flags;
377 		if (zfsvfs->z_fuid_replay == NULL) {
378 			fuidstart = (caddr_t)(lracl + 1) + xvatlen +
379 			    ZIL_ACE_LENGTH(lracl->lr_acl_bytes);
380 			zfsvfs->z_fuid_replay =
381 			    zfs_replay_fuids(fuidstart,
382 			    (void *)&name, lracl->lr_fuidcnt, lracl->lr_domcnt,
383 			    lr->lr_uid, lr->lr_gid);
384 		}
385 		error = VOP_MKDIR(ZTOV(dzp), name, &xva.xva_vattr,
386 		    &vp, kcred, NULL, vflg, &vsec);
387 		break;
388 	default:
389 		error = SET_ERROR(ENOTSUP);
390 	}
391 
392 bail:
393 	if (error == 0 && vp != NULL)
394 		VN_RELE(vp);
395 
396 	VN_RELE(ZTOV(dzp));
397 
398 	if (zfsvfs->z_fuid_replay)
399 		zfs_fuid_info_free(zfsvfs->z_fuid_replay);
400 	zfsvfs->z_fuid_replay = NULL;
401 
402 	return (error);
403 }
404 
405 static int
406 zfs_replay_create(zfsvfs_t *zfsvfs, lr_create_t *lr, boolean_t byteswap)
407 {
408 	char *name = NULL;		/* location determined later */
409 	char *link;			/* symlink content follows name */
410 	znode_t *dzp;
411 	vnode_t *vp = NULL;
412 	xvattr_t xva;
413 	int vflg = 0;
414 	size_t lrsize = sizeof (lr_create_t);
415 	lr_attr_t *lrattr;
416 	void *start;
417 	size_t xvatlen;
418 	uint64_t txtype;
419 	int error;
420 
421 	txtype = (lr->lr_common.lrc_txtype & ~TX_CI);
422 	if (byteswap) {
423 		byteswap_uint64_array(lr, sizeof (*lr));
424 		if (txtype == TX_CREATE_ATTR || txtype == TX_MKDIR_ATTR)
425 			zfs_replay_swap_attrs((lr_attr_t *)(lr + 1));
426 	}
427 
428 
429 	if ((error = zfs_zget(zfsvfs, lr->lr_doid, &dzp)) != 0)
430 		return (error);
431 
432 	xva_init(&xva);
433 	zfs_init_vattr(&xva.xva_vattr, AT_TYPE | AT_MODE | AT_UID | AT_GID,
434 	    lr->lr_mode, lr->lr_uid, lr->lr_gid, lr->lr_rdev, lr->lr_foid);
435 
436 	/*
437 	 * All forms of zfs create (create, mkdir, mkxattrdir, symlink)
438 	 * eventually end up in zfs_mknode(), which assigns the object's
439 	 * creation time and generation number.  The generic VOP_CREATE()
440 	 * doesn't have either concept, so we smuggle the values inside
441 	 * the vattr's otherwise unused va_ctime and va_nblocks fields.
442 	 */
443 	ZFS_TIME_DECODE(&xva.xva_vattr.va_ctime, lr->lr_crtime);
444 	xva.xva_vattr.va_nblocks = lr->lr_gen;
445 
446 	error = dmu_object_info(zfsvfs->z_os, lr->lr_foid, NULL);
447 	if (error != ENOENT)
448 		goto out;
449 
450 	if (lr->lr_common.lrc_txtype & TX_CI)
451 		vflg |= FIGNORECASE;
452 
453 	/*
454 	 * Symlinks don't have fuid info, and CIFS never creates
455 	 * symlinks.
456 	 *
457 	 * The _ATTR versions will grab the fuid info in their subcases.
458 	 */
459 	if ((int)lr->lr_common.lrc_txtype != TX_SYMLINK &&
460 	    (int)lr->lr_common.lrc_txtype != TX_MKDIR_ATTR &&
461 	    (int)lr->lr_common.lrc_txtype != TX_CREATE_ATTR) {
462 		start = (lr + 1);
463 		zfsvfs->z_fuid_replay =
464 		    zfs_replay_fuid_domain(start, &start,
465 		    lr->lr_uid, lr->lr_gid);
466 	}
467 
468 	switch (txtype) {
469 	case TX_CREATE_ATTR:
470 		lrattr = (lr_attr_t *)(caddr_t)(lr + 1);
471 		xvatlen = ZIL_XVAT_SIZE(lrattr->lr_attr_masksize);
472 		zfs_replay_xvattr((lr_attr_t *)((caddr_t)lr + lrsize), &xva);
473 		start = (caddr_t)(lr + 1) + xvatlen;
474 		zfsvfs->z_fuid_replay =
475 		    zfs_replay_fuid_domain(start, &start,
476 		    lr->lr_uid, lr->lr_gid);
477 		name = (char *)start;
478 
479 		/*FALLTHROUGH*/
480 	case TX_CREATE:
481 		if (name == NULL)
482 			name = (char *)start;
483 
484 		error = VOP_CREATE(ZTOV(dzp), name, &xva.xva_vattr,
485 		    0, 0, &vp, kcred, vflg, NULL, NULL);
486 		break;
487 	case TX_MKDIR_ATTR:
488 		lrattr = (lr_attr_t *)(caddr_t)(lr + 1);
489 		xvatlen = ZIL_XVAT_SIZE(lrattr->lr_attr_masksize);
490 		zfs_replay_xvattr((lr_attr_t *)((caddr_t)lr + lrsize), &xva);
491 		start = (caddr_t)(lr + 1) + xvatlen;
492 		zfsvfs->z_fuid_replay =
493 		    zfs_replay_fuid_domain(start, &start,
494 		    lr->lr_uid, lr->lr_gid);
495 		name = (char *)start;
496 
497 		/*FALLTHROUGH*/
498 	case TX_MKDIR:
499 		if (name == NULL)
500 			name = (char *)(lr + 1);
501 
502 		error = VOP_MKDIR(ZTOV(dzp), name, &xva.xva_vattr,
503 		    &vp, kcred, NULL, vflg, NULL);
504 		break;
505 	case TX_MKXATTR:
506 		error = zfs_make_xattrdir(dzp, &xva.xva_vattr, &vp, kcred);
507 		break;
508 	case TX_SYMLINK:
509 		name = (char *)(lr + 1);
510 		link = name + strlen(name) + 1;
511 		error = VOP_SYMLINK(ZTOV(dzp), name, &xva.xva_vattr,
512 		    link, kcred, NULL, vflg);
513 		break;
514 	default:
515 		error = SET_ERROR(ENOTSUP);
516 	}
517 
518 out:
519 	if (error == 0 && vp != NULL)
520 		VN_RELE(vp);
521 
522 	VN_RELE(ZTOV(dzp));
523 
524 	if (zfsvfs->z_fuid_replay)
525 		zfs_fuid_info_free(zfsvfs->z_fuid_replay);
526 	zfsvfs->z_fuid_replay = NULL;
527 	return (error);
528 }
529 
530 static int
531 zfs_replay_remove(zfsvfs_t *zfsvfs, lr_remove_t *lr, boolean_t byteswap)
532 {
533 	char *name = (char *)(lr + 1);	/* name follows lr_remove_t */
534 	znode_t *dzp;
535 	int error;
536 	int vflg = 0;
537 
538 	if (byteswap)
539 		byteswap_uint64_array(lr, sizeof (*lr));
540 
541 	if ((error = zfs_zget(zfsvfs, lr->lr_doid, &dzp)) != 0)
542 		return (error);
543 
544 	if (lr->lr_common.lrc_txtype & TX_CI)
545 		vflg |= FIGNORECASE;
546 
547 	switch ((int)lr->lr_common.lrc_txtype) {
548 	case TX_REMOVE:
549 		error = VOP_REMOVE(ZTOV(dzp), name, kcred, NULL, vflg);
550 		break;
551 	case TX_RMDIR:
552 		error = VOP_RMDIR(ZTOV(dzp), name, NULL, kcred, NULL, vflg);
553 		break;
554 	default:
555 		error = SET_ERROR(ENOTSUP);
556 	}
557 
558 	VN_RELE(ZTOV(dzp));
559 
560 	return (error);
561 }
562 
563 static int
564 zfs_replay_link(zfsvfs_t *zfsvfs, lr_link_t *lr, boolean_t byteswap)
565 {
566 	char *name = (char *)(lr + 1);	/* name follows lr_link_t */
567 	znode_t *dzp, *zp;
568 	int error;
569 	int vflg = 0;
570 
571 	if (byteswap)
572 		byteswap_uint64_array(lr, sizeof (*lr));
573 
574 	if ((error = zfs_zget(zfsvfs, lr->lr_doid, &dzp)) != 0)
575 		return (error);
576 
577 	if ((error = zfs_zget(zfsvfs, lr->lr_link_obj, &zp)) != 0) {
578 		VN_RELE(ZTOV(dzp));
579 		return (error);
580 	}
581 
582 	if (lr->lr_common.lrc_txtype & TX_CI)
583 		vflg |= FIGNORECASE;
584 
585 	error = VOP_LINK(ZTOV(dzp), ZTOV(zp), name, kcred, NULL, vflg);
586 
587 	VN_RELE(ZTOV(zp));
588 	VN_RELE(ZTOV(dzp));
589 
590 	return (error);
591 }
592 
593 static int
594 zfs_replay_rename(zfsvfs_t *zfsvfs, lr_rename_t *lr, boolean_t byteswap)
595 {
596 	char *sname = (char *)(lr + 1);	/* sname and tname follow lr_rename_t */
597 	char *tname = sname + strlen(sname) + 1;
598 	znode_t *sdzp, *tdzp;
599 	int error;
600 	int vflg = 0;
601 
602 	if (byteswap)
603 		byteswap_uint64_array(lr, sizeof (*lr));
604 
605 	if ((error = zfs_zget(zfsvfs, lr->lr_sdoid, &sdzp)) != 0)
606 		return (error);
607 
608 	if ((error = zfs_zget(zfsvfs, lr->lr_tdoid, &tdzp)) != 0) {
609 		VN_RELE(ZTOV(sdzp));
610 		return (error);
611 	}
612 
613 	if (lr->lr_common.lrc_txtype & TX_CI)
614 		vflg |= FIGNORECASE;
615 
616 	error = VOP_RENAME(ZTOV(sdzp), sname, ZTOV(tdzp), tname, kcred,
617 	    NULL, vflg);
618 
619 	VN_RELE(ZTOV(tdzp));
620 	VN_RELE(ZTOV(sdzp));
621 
622 	return (error);
623 }
624 
625 static int
626 zfs_replay_write(zfsvfs_t *zfsvfs, lr_write_t *lr, boolean_t byteswap)
627 {
628 	char *data = (char *)(lr + 1);	/* data follows lr_write_t */
629 	znode_t	*zp;
630 	int error;
631 	ssize_t resid;
632 	uint64_t eod, offset, length;
633 
634 	if (byteswap)
635 		byteswap_uint64_array(lr, sizeof (*lr));
636 
637 	if ((error = zfs_zget(zfsvfs, lr->lr_foid, &zp)) != 0) {
638 		/*
639 		 * As we can log writes out of order, it's possible the
640 		 * file has been removed. In this case just drop the write
641 		 * and return success.
642 		 */
643 		if (error == ENOENT)
644 			error = 0;
645 		return (error);
646 	}
647 
648 	offset = lr->lr_offset;
649 	length = lr->lr_length;
650 	eod = offset + length;	/* end of data for this write */
651 
652 	/*
653 	 * This may be a write from a dmu_sync() for a whole block,
654 	 * and may extend beyond the current end of the file.
655 	 * We can't just replay what was written for this TX_WRITE as
656 	 * a future TX_WRITE2 may extend the eof and the data for that
657 	 * write needs to be there. So we write the whole block and
658 	 * reduce the eof. This needs to be done within the single dmu
659 	 * transaction created within vn_rdwr -> zfs_write. So a possible
660 	 * new end of file is passed through in zfsvfs->z_replay_eof
661 	 */
662 
663 	zfsvfs->z_replay_eof = 0; /* 0 means don't change end of file */
664 
665 	/* If it's a dmu_sync() block, write the whole block */
666 	if (lr->lr_common.lrc_reclen == sizeof (lr_write_t)) {
667 		uint64_t blocksize = BP_GET_LSIZE(&lr->lr_blkptr);
668 		if (length < blocksize) {
669 			offset -= offset % blocksize;
670 			length = blocksize;
671 		}
672 		if (zp->z_size < eod)
673 			zfsvfs->z_replay_eof = eod;
674 	}
675 
676 	error = vn_rdwr(UIO_WRITE, ZTOV(zp), data, length, offset,
677 	    UIO_SYSSPACE, 0, RLIM64_INFINITY, kcred, &resid);
678 
679 	VN_RELE(ZTOV(zp));
680 	zfsvfs->z_replay_eof = 0;	/* safety */
681 
682 	return (error);
683 }
684 
685 /*
686  * TX_WRITE2 are only generated when dmu_sync() returns EALREADY
687  * meaning the pool block is already being synced. So now that we always write
688  * out full blocks, all we have to do is expand the eof if
689  * the file is grown.
690  */
691 static int
692 zfs_replay_write2(zfsvfs_t *zfsvfs, lr_write_t *lr, boolean_t byteswap)
693 {
694 	znode_t	*zp;
695 	int error;
696 	uint64_t end;
697 
698 	if (byteswap)
699 		byteswap_uint64_array(lr, sizeof (*lr));
700 
701 	if ((error = zfs_zget(zfsvfs, lr->lr_foid, &zp)) != 0)
702 		return (error);
703 
704 top:
705 	end = lr->lr_offset + lr->lr_length;
706 	if (end > zp->z_size) {
707 		dmu_tx_t *tx = dmu_tx_create(zfsvfs->z_os);
708 
709 		zp->z_size = end;
710 		dmu_tx_hold_sa(tx, zp->z_sa_hdl, B_FALSE);
711 		error = dmu_tx_assign(tx, TXG_WAIT);
712 		if (error) {
713 			VN_RELE(ZTOV(zp));
714 			if (error == ERESTART) {
715 				dmu_tx_wait(tx);
716 				dmu_tx_abort(tx);
717 				goto top;
718 			}
719 			dmu_tx_abort(tx);
720 			return (error);
721 		}
722 		(void) sa_update(zp->z_sa_hdl, SA_ZPL_SIZE(zfsvfs),
723 		    (void *)&zp->z_size, sizeof (uint64_t), tx);
724 
725 		/* Ensure the replayed seq is updated */
726 		(void) zil_replaying(zfsvfs->z_log, tx);
727 
728 		dmu_tx_commit(tx);
729 	}
730 
731 	VN_RELE(ZTOV(zp));
732 
733 	return (error);
734 }
735 
736 static int
737 zfs_replay_truncate(zfsvfs_t *zfsvfs, lr_truncate_t *lr, boolean_t byteswap)
738 {
739 	znode_t *zp;
740 	flock64_t fl;
741 	int error;
742 
743 	if (byteswap)
744 		byteswap_uint64_array(lr, sizeof (*lr));
745 
746 	if ((error = zfs_zget(zfsvfs, lr->lr_foid, &zp)) != 0)
747 		return (error);
748 
749 	bzero(&fl, sizeof (fl));
750 	fl.l_type = F_WRLCK;
751 	fl.l_whence = 0;
752 	fl.l_start = lr->lr_offset;
753 	fl.l_len = lr->lr_length;
754 
755 	error = VOP_SPACE(ZTOV(zp), F_FREESP, &fl, FWRITE | FOFFMAX,
756 	    lr->lr_offset, kcred, NULL);
757 
758 	VN_RELE(ZTOV(zp));
759 
760 	return (error);
761 }
762 
763 static int
764 zfs_replay_setattr(zfsvfs_t *zfsvfs, lr_setattr_t *lr, boolean_t byteswap)
765 {
766 	znode_t *zp;
767 	xvattr_t xva;
768 	vattr_t *vap = &xva.xva_vattr;
769 	int error;
770 	void *start;
771 
772 	xva_init(&xva);
773 	if (byteswap) {
774 		byteswap_uint64_array(lr, sizeof (*lr));
775 
776 		if ((lr->lr_mask & AT_XVATTR) &&
777 		    zfsvfs->z_version >= ZPL_VERSION_INITIAL)
778 			zfs_replay_swap_attrs((lr_attr_t *)(lr + 1));
779 	}
780 
781 	if ((error = zfs_zget(zfsvfs, lr->lr_foid, &zp)) != 0)
782 		return (error);
783 
784 	zfs_init_vattr(vap, lr->lr_mask, lr->lr_mode,
785 	    lr->lr_uid, lr->lr_gid, 0, lr->lr_foid);
786 
787 	vap->va_size = lr->lr_size;
788 	ZFS_TIME_DECODE(&vap->va_atime, lr->lr_atime);
789 	ZFS_TIME_DECODE(&vap->va_mtime, lr->lr_mtime);
790 
791 	/*
792 	 * Fill in xvattr_t portions if necessary.
793 	 */
794 
795 	start = (lr_setattr_t *)(lr + 1);
796 	if (vap->va_mask & AT_XVATTR) {
797 		zfs_replay_xvattr((lr_attr_t *)start, &xva);
798 		start = (caddr_t)start +
799 		    ZIL_XVAT_SIZE(((lr_attr_t *)start)->lr_attr_masksize);
800 	} else
801 		xva.xva_vattr.va_mask &= ~AT_XVATTR;
802 
803 	zfsvfs->z_fuid_replay = zfs_replay_fuid_domain(start, &start,
804 	    lr->lr_uid, lr->lr_gid);
805 
806 	error = VOP_SETATTR(ZTOV(zp), vap, 0, kcred, NULL);
807 
808 	zfs_fuid_info_free(zfsvfs->z_fuid_replay);
809 	zfsvfs->z_fuid_replay = NULL;
810 	VN_RELE(ZTOV(zp));
811 
812 	return (error);
813 }
814 
815 static int
816 zfs_replay_acl_v0(zfsvfs_t *zfsvfs, lr_acl_v0_t *lr, boolean_t byteswap)
817 {
818 	ace_t *ace = (ace_t *)(lr + 1);	/* ace array follows lr_acl_t */
819 	vsecattr_t vsa;
820 	znode_t *zp;
821 	int error;
822 
823 	if (byteswap) {
824 		byteswap_uint64_array(lr, sizeof (*lr));
825 		zfs_oldace_byteswap(ace, lr->lr_aclcnt);
826 	}
827 
828 	if ((error = zfs_zget(zfsvfs, lr->lr_foid, &zp)) != 0)
829 		return (error);
830 
831 	bzero(&vsa, sizeof (vsa));
832 	vsa.vsa_mask = VSA_ACE | VSA_ACECNT;
833 	vsa.vsa_aclcnt = lr->lr_aclcnt;
834 	vsa.vsa_aclentsz = sizeof (ace_t) * vsa.vsa_aclcnt;
835 	vsa.vsa_aclflags = 0;
836 	vsa.vsa_aclentp = ace;
837 
838 	error = VOP_SETSECATTR(ZTOV(zp), &vsa, 0, kcred, NULL);
839 
840 	VN_RELE(ZTOV(zp));
841 
842 	return (error);
843 }
844 
845 /*
846  * Replaying ACLs is complicated by FUID support.
847  * The log record may contain some optional data
848  * to be used for replaying FUID's.  These pieces
849  * are the actual FUIDs that were created initially.
850  * The FUID table index may no longer be valid and
851  * during zfs_create() a new index may be assigned.
852  * Because of this the log will contain the original
853  * doman+rid in order to create a new FUID.
854  *
855  * The individual ACEs may contain an ephemeral uid/gid which is no
856  * longer valid and will need to be replaced with an actual FUID.
857  *
858  */
859 static int
860 zfs_replay_acl(zfsvfs_t *zfsvfs, lr_acl_t *lr, boolean_t byteswap)
861 {
862 	ace_t *ace = (ace_t *)(lr + 1);
863 	vsecattr_t vsa;
864 	znode_t *zp;
865 	int error;
866 
867 	if (byteswap) {
868 		byteswap_uint64_array(lr, sizeof (*lr));
869 		zfs_ace_byteswap(ace, lr->lr_acl_bytes, B_FALSE);
870 		if (lr->lr_fuidcnt) {
871 			byteswap_uint64_array((caddr_t)ace +
872 			    ZIL_ACE_LENGTH(lr->lr_acl_bytes),
873 			    lr->lr_fuidcnt * sizeof (uint64_t));
874 		}
875 	}
876 
877 	if ((error = zfs_zget(zfsvfs, lr->lr_foid, &zp)) != 0)
878 		return (error);
879 
880 	bzero(&vsa, sizeof (vsa));
881 	vsa.vsa_mask = VSA_ACE | VSA_ACECNT | VSA_ACE_ACLFLAGS;
882 	vsa.vsa_aclcnt = lr->lr_aclcnt;
883 	vsa.vsa_aclentp = ace;
884 	vsa.vsa_aclentsz = lr->lr_acl_bytes;
885 	vsa.vsa_aclflags = lr->lr_acl_flags;
886 
887 	if (lr->lr_fuidcnt) {
888 		void *fuidstart = (caddr_t)ace +
889 		    ZIL_ACE_LENGTH(lr->lr_acl_bytes);
890 
891 		zfsvfs->z_fuid_replay =
892 		    zfs_replay_fuids(fuidstart, &fuidstart,
893 		    lr->lr_fuidcnt, lr->lr_domcnt, 0, 0);
894 	}
895 
896 	error = VOP_SETSECATTR(ZTOV(zp), &vsa, 0, kcred, NULL);
897 
898 	if (zfsvfs->z_fuid_replay)
899 		zfs_fuid_info_free(zfsvfs->z_fuid_replay);
900 
901 	zfsvfs->z_fuid_replay = NULL;
902 	VN_RELE(ZTOV(zp));
903 
904 	return (error);
905 }
906 
907 /*
908  * Callback vectors for replaying records
909  */
910 zil_replay_func_t *zfs_replay_vector[TX_MAX_TYPE] = {
911 	zfs_replay_error,	/* 0 no such transaction type */
912 	zfs_replay_create,	/* TX_CREATE */
913 	zfs_replay_create,	/* TX_MKDIR */
914 	zfs_replay_create,	/* TX_MKXATTR */
915 	zfs_replay_create,	/* TX_SYMLINK */
916 	zfs_replay_remove,	/* TX_REMOVE */
917 	zfs_replay_remove,	/* TX_RMDIR */
918 	zfs_replay_link,	/* TX_LINK */
919 	zfs_replay_rename,	/* TX_RENAME */
920 	zfs_replay_write,	/* TX_WRITE */
921 	zfs_replay_truncate,	/* TX_TRUNCATE */
922 	zfs_replay_setattr,	/* TX_SETATTR */
923 	zfs_replay_acl_v0,	/* TX_ACL_V0 */
924 	zfs_replay_acl,		/* TX_ACL */
925 	zfs_replay_create_acl,	/* TX_CREATE_ACL */
926 	zfs_replay_create,	/* TX_CREATE_ATTR */
927 	zfs_replay_create_acl,	/* TX_CREATE_ACL_ATTR */
928 	zfs_replay_create_acl,	/* TX_MKDIR_ACL */
929 	zfs_replay_create,	/* TX_MKDIR_ATTR */
930 	zfs_replay_create_acl,	/* TX_MKDIR_ACL_ATTR */
931 	zfs_replay_write2,	/* TX_WRITE2 */
932 };
933