xref: /illumos-gate/usr/src/uts/common/fs/zfs/zfs_replay.c (revision c7a0678d)
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, 2015 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(void *arg1, void *arg2, 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(void *arg1, void *arg2, boolean_t byteswap)
266 {
267 	zfsvfs_t *zfsvfs = arg1;
268 	lr_acl_create_t *lracl = arg2;
269 	char *name = NULL;		/* location determined later */
270 	lr_create_t *lr = (lr_create_t *)lracl;
271 	znode_t *dzp;
272 	vnode_t *vp = NULL;
273 	xvattr_t xva;
274 	int vflg = 0;
275 	vsecattr_t vsec = { 0 };
276 	lr_attr_t *lrattr;
277 	void *aclstart;
278 	void *fuidstart;
279 	size_t xvatlen = 0;
280 	uint64_t txtype;
281 	int error;
282 
283 	txtype = (lr->lr_common.lrc_txtype & ~TX_CI);
284 	if (byteswap) {
285 		byteswap_uint64_array(lracl, sizeof (*lracl));
286 		if (txtype == TX_CREATE_ACL_ATTR ||
287 		    txtype == TX_MKDIR_ACL_ATTR) {
288 			lrattr = (lr_attr_t *)(caddr_t)(lracl + 1);
289 			zfs_replay_swap_attrs(lrattr);
290 			xvatlen = ZIL_XVAT_SIZE(lrattr->lr_attr_masksize);
291 		}
292 
293 		aclstart = (caddr_t)(lracl + 1) + xvatlen;
294 		zfs_ace_byteswap(aclstart, lracl->lr_acl_bytes, B_FALSE);
295 		/* swap fuids */
296 		if (lracl->lr_fuidcnt) {
297 			byteswap_uint64_array((caddr_t)aclstart +
298 			    ZIL_ACE_LENGTH(lracl->lr_acl_bytes),
299 			    lracl->lr_fuidcnt * sizeof (uint64_t));
300 		}
301 	}
302 
303 	if ((error = zfs_zget(zfsvfs, lr->lr_doid, &dzp)) != 0)
304 		return (error);
305 
306 	xva_init(&xva);
307 	zfs_init_vattr(&xva.xva_vattr, AT_TYPE | AT_MODE | AT_UID | AT_GID,
308 	    lr->lr_mode, lr->lr_uid, lr->lr_gid, lr->lr_rdev, lr->lr_foid);
309 
310 	/*
311 	 * All forms of zfs create (create, mkdir, mkxattrdir, symlink)
312 	 * eventually end up in zfs_mknode(), which assigns the object's
313 	 * creation time and generation number.  The generic VOP_CREATE()
314 	 * doesn't have either concept, so we smuggle the values inside
315 	 * the vattr's otherwise unused va_ctime and va_nblocks fields.
316 	 */
317 	ZFS_TIME_DECODE(&xva.xva_vattr.va_ctime, lr->lr_crtime);
318 	xva.xva_vattr.va_nblocks = lr->lr_gen;
319 
320 	error = dmu_object_info(zfsvfs->z_os, lr->lr_foid, NULL);
321 	if (error != ENOENT)
322 		goto bail;
323 
324 	if (lr->lr_common.lrc_txtype & TX_CI)
325 		vflg |= FIGNORECASE;
326 	switch (txtype) {
327 	case TX_CREATE_ACL:
328 		aclstart = (caddr_t)(lracl + 1);
329 		fuidstart = (caddr_t)aclstart +
330 		    ZIL_ACE_LENGTH(lracl->lr_acl_bytes);
331 		zfsvfs->z_fuid_replay = zfs_replay_fuids(fuidstart,
332 		    (void *)&name, lracl->lr_fuidcnt, lracl->lr_domcnt,
333 		    lr->lr_uid, lr->lr_gid);
334 		/*FALLTHROUGH*/
335 	case TX_CREATE_ACL_ATTR:
336 		if (name == NULL) {
337 			lrattr = (lr_attr_t *)(caddr_t)(lracl + 1);
338 			xvatlen = ZIL_XVAT_SIZE(lrattr->lr_attr_masksize);
339 			xva.xva_vattr.va_mask |= AT_XVATTR;
340 			zfs_replay_xvattr(lrattr, &xva);
341 		}
342 		vsec.vsa_mask = VSA_ACE | VSA_ACE_ACLFLAGS;
343 		vsec.vsa_aclentp = (caddr_t)(lracl + 1) + xvatlen;
344 		vsec.vsa_aclcnt = lracl->lr_aclcnt;
345 		vsec.vsa_aclentsz = lracl->lr_acl_bytes;
346 		vsec.vsa_aclflags = lracl->lr_acl_flags;
347 		if (zfsvfs->z_fuid_replay == NULL) {
348 			fuidstart = (caddr_t)(lracl + 1) + xvatlen +
349 			    ZIL_ACE_LENGTH(lracl->lr_acl_bytes);
350 			zfsvfs->z_fuid_replay =
351 			    zfs_replay_fuids(fuidstart,
352 			    (void *)&name, lracl->lr_fuidcnt, lracl->lr_domcnt,
353 			    lr->lr_uid, lr->lr_gid);
354 		}
355 
356 		error = VOP_CREATE(ZTOV(dzp), name, &xva.xva_vattr,
357 		    0, 0, &vp, kcred, vflg, NULL, &vsec);
358 		break;
359 	case TX_MKDIR_ACL:
360 		aclstart = (caddr_t)(lracl + 1);
361 		fuidstart = (caddr_t)aclstart +
362 		    ZIL_ACE_LENGTH(lracl->lr_acl_bytes);
363 		zfsvfs->z_fuid_replay = zfs_replay_fuids(fuidstart,
364 		    (void *)&name, lracl->lr_fuidcnt, lracl->lr_domcnt,
365 		    lr->lr_uid, lr->lr_gid);
366 		/*FALLTHROUGH*/
367 	case TX_MKDIR_ACL_ATTR:
368 		if (name == NULL) {
369 			lrattr = (lr_attr_t *)(caddr_t)(lracl + 1);
370 			xvatlen = ZIL_XVAT_SIZE(lrattr->lr_attr_masksize);
371 			zfs_replay_xvattr(lrattr, &xva);
372 		}
373 		vsec.vsa_mask = VSA_ACE | VSA_ACE_ACLFLAGS;
374 		vsec.vsa_aclentp = (caddr_t)(lracl + 1) + xvatlen;
375 		vsec.vsa_aclcnt = lracl->lr_aclcnt;
376 		vsec.vsa_aclentsz = lracl->lr_acl_bytes;
377 		vsec.vsa_aclflags = lracl->lr_acl_flags;
378 		if (zfsvfs->z_fuid_replay == NULL) {
379 			fuidstart = (caddr_t)(lracl + 1) + xvatlen +
380 			    ZIL_ACE_LENGTH(lracl->lr_acl_bytes);
381 			zfsvfs->z_fuid_replay =
382 			    zfs_replay_fuids(fuidstart,
383 			    (void *)&name, lracl->lr_fuidcnt, lracl->lr_domcnt,
384 			    lr->lr_uid, lr->lr_gid);
385 		}
386 		error = VOP_MKDIR(ZTOV(dzp), name, &xva.xva_vattr,
387 		    &vp, kcred, NULL, vflg, &vsec);
388 		break;
389 	default:
390 		error = SET_ERROR(ENOTSUP);
391 	}
392 
393 bail:
394 	if (error == 0 && vp != NULL)
395 		VN_RELE(vp);
396 
397 	VN_RELE(ZTOV(dzp));
398 
399 	if (zfsvfs->z_fuid_replay)
400 		zfs_fuid_info_free(zfsvfs->z_fuid_replay);
401 	zfsvfs->z_fuid_replay = NULL;
402 
403 	return (error);
404 }
405 
406 static int
407 zfs_replay_create(void *arg1, void *arg2, boolean_t byteswap)
408 {
409 	zfsvfs_t *zfsvfs = arg1;
410 	lr_create_t *lr = arg2;
411 	char *name = NULL;		/* location determined later */
412 	char *link;			/* symlink content follows name */
413 	znode_t *dzp;
414 	vnode_t *vp = NULL;
415 	xvattr_t xva;
416 	int vflg = 0;
417 	size_t lrsize = sizeof (lr_create_t);
418 	lr_attr_t *lrattr;
419 	void *start;
420 	size_t xvatlen;
421 	uint64_t txtype;
422 	int error;
423 
424 	txtype = (lr->lr_common.lrc_txtype & ~TX_CI);
425 	if (byteswap) {
426 		byteswap_uint64_array(lr, sizeof (*lr));
427 		if (txtype == TX_CREATE_ATTR || txtype == TX_MKDIR_ATTR)
428 			zfs_replay_swap_attrs((lr_attr_t *)(lr + 1));
429 	}
430 
431 
432 	if ((error = zfs_zget(zfsvfs, lr->lr_doid, &dzp)) != 0)
433 		return (error);
434 
435 	xva_init(&xva);
436 	zfs_init_vattr(&xva.xva_vattr, AT_TYPE | AT_MODE | AT_UID | AT_GID,
437 	    lr->lr_mode, lr->lr_uid, lr->lr_gid, lr->lr_rdev, lr->lr_foid);
438 
439 	/*
440 	 * All forms of zfs create (create, mkdir, mkxattrdir, symlink)
441 	 * eventually end up in zfs_mknode(), which assigns the object's
442 	 * creation time and generation number.  The generic VOP_CREATE()
443 	 * doesn't have either concept, so we smuggle the values inside
444 	 * the vattr's otherwise unused va_ctime and va_nblocks fields.
445 	 */
446 	ZFS_TIME_DECODE(&xva.xva_vattr.va_ctime, lr->lr_crtime);
447 	xva.xva_vattr.va_nblocks = lr->lr_gen;
448 
449 	error = dmu_object_info(zfsvfs->z_os, lr->lr_foid, NULL);
450 	if (error != ENOENT)
451 		goto out;
452 
453 	if (lr->lr_common.lrc_txtype & TX_CI)
454 		vflg |= FIGNORECASE;
455 
456 	/*
457 	 * Symlinks don't have fuid info, and CIFS never creates
458 	 * symlinks.
459 	 *
460 	 * The _ATTR versions will grab the fuid info in their subcases.
461 	 */
462 	if ((int)lr->lr_common.lrc_txtype != TX_SYMLINK &&
463 	    (int)lr->lr_common.lrc_txtype != TX_MKDIR_ATTR &&
464 	    (int)lr->lr_common.lrc_txtype != TX_CREATE_ATTR) {
465 		start = (lr + 1);
466 		zfsvfs->z_fuid_replay =
467 		    zfs_replay_fuid_domain(start, &start,
468 		    lr->lr_uid, lr->lr_gid);
469 	}
470 
471 	switch (txtype) {
472 	case TX_CREATE_ATTR:
473 		lrattr = (lr_attr_t *)(caddr_t)(lr + 1);
474 		xvatlen = ZIL_XVAT_SIZE(lrattr->lr_attr_masksize);
475 		zfs_replay_xvattr((lr_attr_t *)((caddr_t)lr + lrsize), &xva);
476 		start = (caddr_t)(lr + 1) + xvatlen;
477 		zfsvfs->z_fuid_replay =
478 		    zfs_replay_fuid_domain(start, &start,
479 		    lr->lr_uid, lr->lr_gid);
480 		name = (char *)start;
481 
482 		/*FALLTHROUGH*/
483 	case TX_CREATE:
484 		if (name == NULL)
485 			name = (char *)start;
486 
487 		error = VOP_CREATE(ZTOV(dzp), name, &xva.xva_vattr,
488 		    0, 0, &vp, kcred, vflg, NULL, NULL);
489 		break;
490 	case TX_MKDIR_ATTR:
491 		lrattr = (lr_attr_t *)(caddr_t)(lr + 1);
492 		xvatlen = ZIL_XVAT_SIZE(lrattr->lr_attr_masksize);
493 		zfs_replay_xvattr((lr_attr_t *)((caddr_t)lr + lrsize), &xva);
494 		start = (caddr_t)(lr + 1) + xvatlen;
495 		zfsvfs->z_fuid_replay =
496 		    zfs_replay_fuid_domain(start, &start,
497 		    lr->lr_uid, lr->lr_gid);
498 		name = (char *)start;
499 
500 		/*FALLTHROUGH*/
501 	case TX_MKDIR:
502 		if (name == NULL)
503 			name = (char *)(lr + 1);
504 
505 		error = VOP_MKDIR(ZTOV(dzp), name, &xva.xva_vattr,
506 		    &vp, kcred, NULL, vflg, NULL);
507 		break;
508 	case TX_MKXATTR:
509 		error = zfs_make_xattrdir(dzp, &xva.xva_vattr, &vp, kcred);
510 		break;
511 	case TX_SYMLINK:
512 		name = (char *)(lr + 1);
513 		link = name + strlen(name) + 1;
514 		error = VOP_SYMLINK(ZTOV(dzp), name, &xva.xva_vattr,
515 		    link, kcred, NULL, vflg);
516 		break;
517 	default:
518 		error = SET_ERROR(ENOTSUP);
519 	}
520 
521 out:
522 	if (error == 0 && vp != NULL)
523 		VN_RELE(vp);
524 
525 	VN_RELE(ZTOV(dzp));
526 
527 	if (zfsvfs->z_fuid_replay)
528 		zfs_fuid_info_free(zfsvfs->z_fuid_replay);
529 	zfsvfs->z_fuid_replay = NULL;
530 	return (error);
531 }
532 
533 static int
534 zfs_replay_remove(void *arg1, void *arg2, boolean_t byteswap)
535 {
536 	zfsvfs_t *zfsvfs = arg1;
537 	lr_remove_t *lr = arg2;
538 	char *name = (char *)(lr + 1);	/* name follows lr_remove_t */
539 	znode_t *dzp;
540 	int error;
541 	int vflg = 0;
542 
543 	if (byteswap)
544 		byteswap_uint64_array(lr, sizeof (*lr));
545 
546 	if ((error = zfs_zget(zfsvfs, lr->lr_doid, &dzp)) != 0)
547 		return (error);
548 
549 	if (lr->lr_common.lrc_txtype & TX_CI)
550 		vflg |= FIGNORECASE;
551 
552 	switch ((int)lr->lr_common.lrc_txtype) {
553 	case TX_REMOVE:
554 		error = VOP_REMOVE(ZTOV(dzp), name, kcred, NULL, vflg);
555 		break;
556 	case TX_RMDIR:
557 		error = VOP_RMDIR(ZTOV(dzp), name, NULL, kcred, NULL, vflg);
558 		break;
559 	default:
560 		error = SET_ERROR(ENOTSUP);
561 	}
562 
563 	VN_RELE(ZTOV(dzp));
564 
565 	return (error);
566 }
567 
568 static int
569 zfs_replay_link(void *arg1, void *arg2, boolean_t byteswap)
570 {
571 	zfsvfs_t *zfsvfs = arg1;
572 	lr_link_t *lr = arg2;
573 	char *name = (char *)(lr + 1);	/* name follows lr_link_t */
574 	znode_t *dzp, *zp;
575 	int error;
576 	int vflg = 0;
577 
578 	if (byteswap)
579 		byteswap_uint64_array(lr, sizeof (*lr));
580 
581 	if ((error = zfs_zget(zfsvfs, lr->lr_doid, &dzp)) != 0)
582 		return (error);
583 
584 	if ((error = zfs_zget(zfsvfs, lr->lr_link_obj, &zp)) != 0) {
585 		VN_RELE(ZTOV(dzp));
586 		return (error);
587 	}
588 
589 	if (lr->lr_common.lrc_txtype & TX_CI)
590 		vflg |= FIGNORECASE;
591 
592 	error = VOP_LINK(ZTOV(dzp), ZTOV(zp), name, kcred, NULL, vflg);
593 
594 	VN_RELE(ZTOV(zp));
595 	VN_RELE(ZTOV(dzp));
596 
597 	return (error);
598 }
599 
600 static int
601 zfs_replay_rename(void *arg1, void *arg2, boolean_t byteswap)
602 {
603 	zfsvfs_t *zfsvfs = arg1;
604 	lr_rename_t *lr = arg2;
605 	char *sname = (char *)(lr + 1);	/* sname and tname follow lr_rename_t */
606 	char *tname = sname + strlen(sname) + 1;
607 	znode_t *sdzp, *tdzp;
608 	int error;
609 	int vflg = 0;
610 
611 	if (byteswap)
612 		byteswap_uint64_array(lr, sizeof (*lr));
613 
614 	if ((error = zfs_zget(zfsvfs, lr->lr_sdoid, &sdzp)) != 0)
615 		return (error);
616 
617 	if ((error = zfs_zget(zfsvfs, lr->lr_tdoid, &tdzp)) != 0) {
618 		VN_RELE(ZTOV(sdzp));
619 		return (error);
620 	}
621 
622 	if (lr->lr_common.lrc_txtype & TX_CI)
623 		vflg |= FIGNORECASE;
624 
625 	error = VOP_RENAME(ZTOV(sdzp), sname, ZTOV(tdzp), tname, kcred,
626 	    NULL, vflg);
627 
628 	VN_RELE(ZTOV(tdzp));
629 	VN_RELE(ZTOV(sdzp));
630 
631 	return (error);
632 }
633 
634 static int
635 zfs_replay_write(void *arg1, void *arg2, boolean_t byteswap)
636 {
637 	zfsvfs_t *zfsvfs = arg1;
638 	lr_write_t *lr = arg2;
639 	char *data = (char *)(lr + 1);	/* data follows lr_write_t */
640 	znode_t	*zp;
641 	int error;
642 	ssize_t resid;
643 	uint64_t eod, offset, length;
644 
645 	if (byteswap)
646 		byteswap_uint64_array(lr, sizeof (*lr));
647 
648 	if ((error = zfs_zget(zfsvfs, lr->lr_foid, &zp)) != 0) {
649 		/*
650 		 * As we can log writes out of order, it's possible the
651 		 * file has been removed. In this case just drop the write
652 		 * and return success.
653 		 */
654 		if (error == ENOENT)
655 			error = 0;
656 		return (error);
657 	}
658 
659 	offset = lr->lr_offset;
660 	length = lr->lr_length;
661 	eod = offset + length;	/* end of data for this write */
662 
663 	/*
664 	 * This may be a write from a dmu_sync() for a whole block,
665 	 * and may extend beyond the current end of the file.
666 	 * We can't just replay what was written for this TX_WRITE as
667 	 * a future TX_WRITE2 may extend the eof and the data for that
668 	 * write needs to be there. So we write the whole block and
669 	 * reduce the eof. This needs to be done within the single dmu
670 	 * transaction created within vn_rdwr -> zfs_write. So a possible
671 	 * new end of file is passed through in zfsvfs->z_replay_eof
672 	 */
673 
674 	zfsvfs->z_replay_eof = 0; /* 0 means don't change end of file */
675 
676 	/* If it's a dmu_sync() block, write the whole block */
677 	if (lr->lr_common.lrc_reclen == sizeof (lr_write_t)) {
678 		uint64_t blocksize = BP_GET_LSIZE(&lr->lr_blkptr);
679 		if (length < blocksize) {
680 			offset -= offset % blocksize;
681 			length = blocksize;
682 		}
683 		if (zp->z_size < eod)
684 			zfsvfs->z_replay_eof = eod;
685 	}
686 
687 	error = vn_rdwr(UIO_WRITE, ZTOV(zp), data, length, offset,
688 	    UIO_SYSSPACE, 0, RLIM64_INFINITY, kcred, &resid);
689 
690 	VN_RELE(ZTOV(zp));
691 	zfsvfs->z_replay_eof = 0;	/* safety */
692 
693 	return (error);
694 }
695 
696 /*
697  * TX_WRITE2 are only generated when dmu_sync() returns EALREADY
698  * meaning the pool block is already being synced. So now that we always write
699  * out full blocks, all we have to do is expand the eof if
700  * the file is grown.
701  */
702 static int
703 zfs_replay_write2(void *arg1, void *arg2, boolean_t byteswap)
704 {
705 	zfsvfs_t *zfsvfs = arg1;
706 	lr_write_t *lr = arg2;
707 	znode_t	*zp;
708 	int error;
709 	uint64_t end;
710 
711 	if (byteswap)
712 		byteswap_uint64_array(lr, sizeof (*lr));
713 
714 	if ((error = zfs_zget(zfsvfs, lr->lr_foid, &zp)) != 0)
715 		return (error);
716 
717 top:
718 	end = lr->lr_offset + lr->lr_length;
719 	if (end > zp->z_size) {
720 		dmu_tx_t *tx = dmu_tx_create(zfsvfs->z_os);
721 
722 		zp->z_size = end;
723 		dmu_tx_hold_sa(tx, zp->z_sa_hdl, B_FALSE);
724 		error = dmu_tx_assign(tx, TXG_WAIT);
725 		if (error) {
726 			VN_RELE(ZTOV(zp));
727 			if (error == ERESTART) {
728 				dmu_tx_wait(tx);
729 				dmu_tx_abort(tx);
730 				goto top;
731 			}
732 			dmu_tx_abort(tx);
733 			return (error);
734 		}
735 		(void) sa_update(zp->z_sa_hdl, SA_ZPL_SIZE(zfsvfs),
736 		    (void *)&zp->z_size, sizeof (uint64_t), tx);
737 
738 		/* Ensure the replayed seq is updated */
739 		(void) zil_replaying(zfsvfs->z_log, tx);
740 
741 		dmu_tx_commit(tx);
742 	}
743 
744 	VN_RELE(ZTOV(zp));
745 
746 	return (error);
747 }
748 
749 static int
750 zfs_replay_truncate(void *arg1, void *arg2, boolean_t byteswap)
751 {
752 	zfsvfs_t *zfsvfs = arg1;
753 	lr_truncate_t *lr = arg2;
754 	znode_t *zp;
755 	flock64_t fl;
756 	int error;
757 
758 	if (byteswap)
759 		byteswap_uint64_array(lr, sizeof (*lr));
760 
761 	if ((error = zfs_zget(zfsvfs, lr->lr_foid, &zp)) != 0)
762 		return (error);
763 
764 	bzero(&fl, sizeof (fl));
765 	fl.l_type = F_WRLCK;
766 	fl.l_whence = 0;
767 	fl.l_start = lr->lr_offset;
768 	fl.l_len = lr->lr_length;
769 
770 	error = VOP_SPACE(ZTOV(zp), F_FREESP, &fl, FWRITE | FOFFMAX,
771 	    lr->lr_offset, kcred, NULL);
772 
773 	VN_RELE(ZTOV(zp));
774 
775 	return (error);
776 }
777 
778 static int
779 zfs_replay_setattr(void *arg1, void *arg2, boolean_t byteswap)
780 {
781 	zfsvfs_t *zfsvfs = arg1;
782 	lr_setattr_t *lr = arg2;
783 	znode_t *zp;
784 	xvattr_t xva;
785 	vattr_t *vap = &xva.xva_vattr;
786 	int error;
787 	void *start;
788 
789 	xva_init(&xva);
790 	if (byteswap) {
791 		byteswap_uint64_array(lr, sizeof (*lr));
792 
793 		if ((lr->lr_mask & AT_XVATTR) &&
794 		    zfsvfs->z_version >= ZPL_VERSION_INITIAL)
795 			zfs_replay_swap_attrs((lr_attr_t *)(lr + 1));
796 	}
797 
798 	if ((error = zfs_zget(zfsvfs, lr->lr_foid, &zp)) != 0)
799 		return (error);
800 
801 	zfs_init_vattr(vap, lr->lr_mask, lr->lr_mode,
802 	    lr->lr_uid, lr->lr_gid, 0, lr->lr_foid);
803 
804 	vap->va_size = lr->lr_size;
805 	ZFS_TIME_DECODE(&vap->va_atime, lr->lr_atime);
806 	ZFS_TIME_DECODE(&vap->va_mtime, lr->lr_mtime);
807 
808 	/*
809 	 * Fill in xvattr_t portions if necessary.
810 	 */
811 
812 	start = (lr_setattr_t *)(lr + 1);
813 	if (vap->va_mask & AT_XVATTR) {
814 		zfs_replay_xvattr((lr_attr_t *)start, &xva);
815 		start = (caddr_t)start +
816 		    ZIL_XVAT_SIZE(((lr_attr_t *)start)->lr_attr_masksize);
817 	} else
818 		xva.xva_vattr.va_mask &= ~AT_XVATTR;
819 
820 	zfsvfs->z_fuid_replay = zfs_replay_fuid_domain(start, &start,
821 	    lr->lr_uid, lr->lr_gid);
822 
823 	error = VOP_SETATTR(ZTOV(zp), vap, 0, kcred, NULL);
824 
825 	zfs_fuid_info_free(zfsvfs->z_fuid_replay);
826 	zfsvfs->z_fuid_replay = NULL;
827 	VN_RELE(ZTOV(zp));
828 
829 	return (error);
830 }
831 
832 static int
833 zfs_replay_acl_v0(void *arg1, void *arg2, boolean_t byteswap)
834 {
835 	zfsvfs_t *zfsvfs = arg1;
836 	lr_acl_v0_t *lr = arg2;
837 	ace_t *ace = (ace_t *)(lr + 1);	/* ace array follows lr_acl_t */
838 	vsecattr_t vsa;
839 	znode_t *zp;
840 	int error;
841 
842 	if (byteswap) {
843 		byteswap_uint64_array(lr, sizeof (*lr));
844 		zfs_oldace_byteswap(ace, lr->lr_aclcnt);
845 	}
846 
847 	if ((error = zfs_zget(zfsvfs, lr->lr_foid, &zp)) != 0)
848 		return (error);
849 
850 	bzero(&vsa, sizeof (vsa));
851 	vsa.vsa_mask = VSA_ACE | VSA_ACECNT;
852 	vsa.vsa_aclcnt = lr->lr_aclcnt;
853 	vsa.vsa_aclentsz = sizeof (ace_t) * vsa.vsa_aclcnt;
854 	vsa.vsa_aclflags = 0;
855 	vsa.vsa_aclentp = ace;
856 
857 	error = VOP_SETSECATTR(ZTOV(zp), &vsa, 0, kcred, NULL);
858 
859 	VN_RELE(ZTOV(zp));
860 
861 	return (error);
862 }
863 
864 /*
865  * Replaying ACLs is complicated by FUID support.
866  * The log record may contain some optional data
867  * to be used for replaying FUID's.  These pieces
868  * are the actual FUIDs that were created initially.
869  * The FUID table index may no longer be valid and
870  * during zfs_create() a new index may be assigned.
871  * Because of this the log will contain the original
872  * doman+rid in order to create a new FUID.
873  *
874  * The individual ACEs may contain an ephemeral uid/gid which is no
875  * longer valid and will need to be replaced with an actual FUID.
876  *
877  */
878 static int
879 zfs_replay_acl(void *arg1, void *arg2, boolean_t byteswap)
880 {
881 	zfsvfs_t *zfsvfs = arg1;
882 	lr_acl_t *lr = arg2;
883 	ace_t *ace = (ace_t *)(lr + 1);
884 	vsecattr_t vsa;
885 	znode_t *zp;
886 	int error;
887 
888 	if (byteswap) {
889 		byteswap_uint64_array(lr, sizeof (*lr));
890 		zfs_ace_byteswap(ace, lr->lr_acl_bytes, B_FALSE);
891 		if (lr->lr_fuidcnt) {
892 			byteswap_uint64_array((caddr_t)ace +
893 			    ZIL_ACE_LENGTH(lr->lr_acl_bytes),
894 			    lr->lr_fuidcnt * sizeof (uint64_t));
895 		}
896 	}
897 
898 	if ((error = zfs_zget(zfsvfs, lr->lr_foid, &zp)) != 0)
899 		return (error);
900 
901 	bzero(&vsa, sizeof (vsa));
902 	vsa.vsa_mask = VSA_ACE | VSA_ACECNT | VSA_ACE_ACLFLAGS;
903 	vsa.vsa_aclcnt = lr->lr_aclcnt;
904 	vsa.vsa_aclentp = ace;
905 	vsa.vsa_aclentsz = lr->lr_acl_bytes;
906 	vsa.vsa_aclflags = lr->lr_acl_flags;
907 
908 	if (lr->lr_fuidcnt) {
909 		void *fuidstart = (caddr_t)ace +
910 		    ZIL_ACE_LENGTH(lr->lr_acl_bytes);
911 
912 		zfsvfs->z_fuid_replay =
913 		    zfs_replay_fuids(fuidstart, &fuidstart,
914 		    lr->lr_fuidcnt, lr->lr_domcnt, 0, 0);
915 	}
916 
917 	error = VOP_SETSECATTR(ZTOV(zp), &vsa, 0, kcred, NULL);
918 
919 	if (zfsvfs->z_fuid_replay)
920 		zfs_fuid_info_free(zfsvfs->z_fuid_replay);
921 
922 	zfsvfs->z_fuid_replay = NULL;
923 	VN_RELE(ZTOV(zp));
924 
925 	return (error);
926 }
927 
928 /*
929  * Callback vectors for replaying records
930  */
931 zil_replay_func_t *zfs_replay_vector[TX_MAX_TYPE] = {
932 	zfs_replay_error,	/* 0 no such transaction type */
933 	zfs_replay_create,	/* TX_CREATE */
934 	zfs_replay_create,	/* TX_MKDIR */
935 	zfs_replay_create,	/* TX_MKXATTR */
936 	zfs_replay_create,	/* TX_SYMLINK */
937 	zfs_replay_remove,	/* TX_REMOVE */
938 	zfs_replay_remove,	/* TX_RMDIR */
939 	zfs_replay_link,	/* TX_LINK */
940 	zfs_replay_rename,	/* TX_RENAME */
941 	zfs_replay_write,	/* TX_WRITE */
942 	zfs_replay_truncate,	/* TX_TRUNCATE */
943 	zfs_replay_setattr,	/* TX_SETATTR */
944 	zfs_replay_acl_v0,	/* TX_ACL_V0 */
945 	zfs_replay_acl,		/* TX_ACL */
946 	zfs_replay_create_acl,	/* TX_CREATE_ACL */
947 	zfs_replay_create,	/* TX_CREATE_ATTR */
948 	zfs_replay_create_acl,	/* TX_CREATE_ACL_ATTR */
949 	zfs_replay_create_acl,	/* TX_MKDIR_ACL */
950 	zfs_replay_create,	/* TX_MKDIR_ATTR */
951 	zfs_replay_create_acl,	/* TX_MKDIR_ACL_ATTR */
952 	zfs_replay_write2,	/* TX_WRITE2 */
953 };
954