xref: /titanic_41/usr/src/uts/common/fs/zfs/zfs_replay.c (revision 1e4c938b57d1656808e4112127ff1dce3eba5314)
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/fcntl.h>
34 #include <sys/vfs.h>
35 #include <sys/fs/zfs.h>
36 #include <sys/zfs_znode.h>
37 #include <sys/zfs_dir.h>
38 #include <sys/zfs_acl.h>
39 #include <sys/zfs_fuid.h>
40 #include <sys/spa.h>
41 #include <sys/zil.h>
42 #include <sys/byteorder.h>
43 #include <sys/stat.h>
44 #include <sys/mode.h>
45 #include <sys/acl.h>
46 #include <sys/atomic.h>
47 #include <sys/cred.h>
48 
49 /*
50  * Functions to replay ZFS intent log (ZIL) records
51  * The functions are called through a function vector (zfs_replay_vector)
52  * which is indexed by the transaction type.
53  */
54 
55 static void
56 zfs_init_vattr(vattr_t *vap, uint64_t mask, uint64_t mode,
57 	uint64_t uid, uint64_t gid, uint64_t rdev, uint64_t nodeid)
58 {
59 	bzero(vap, sizeof (*vap));
60 	vap->va_mask = (uint_t)mask;
61 	vap->va_type = IFTOVT(mode);
62 	vap->va_mode = mode & MODEMASK;
63 	vap->va_uid = (uid_t)(IS_EPHEMERAL(uid)) ? -1 : uid;
64 	vap->va_gid = (gid_t)(IS_EPHEMERAL(gid)) ? -1 : gid;
65 	vap->va_rdev = zfs_cmpldev(rdev);
66 	vap->va_nodeid = nodeid;
67 }
68 
69 /* ARGSUSED */
70 static int
71 zfs_replay_error(zfsvfs_t *zfsvfs, lr_t *lr, boolean_t byteswap)
72 {
73 	return (ENOTSUP);
74 }
75 
76 static void
77 zfs_replay_xvattr(lr_attr_t *lrattr, xvattr_t *xvap)
78 {
79 	xoptattr_t *xoap = NULL;
80 	uint64_t *attrs;
81 	uint64_t *crtime;
82 	uint32_t *bitmap;
83 	void *scanstamp;
84 	int i;
85 
86 	xvap->xva_vattr.va_mask |= AT_XVATTR;
87 	if ((xoap = xva_getxoptattr(xvap)) == NULL) {
88 		xvap->xva_vattr.va_mask &= ~AT_XVATTR; /* shouldn't happen */
89 		return;
90 	}
91 
92 	ASSERT(lrattr->lr_attr_masksize == xvap->xva_mapsize);
93 
94 	bitmap = &lrattr->lr_attr_bitmap;
95 	for (i = 0; i != lrattr->lr_attr_masksize; i++, bitmap++)
96 		xvap->xva_reqattrmap[i] = *bitmap;
97 
98 	attrs = (uint64_t *)(lrattr + lrattr->lr_attr_masksize - 1);
99 	crtime = attrs + 1;
100 	scanstamp = (caddr_t)(crtime + 2);
101 
102 	if (XVA_ISSET_REQ(xvap, XAT_HIDDEN))
103 		xoap->xoa_hidden = ((*attrs & XAT0_HIDDEN) != 0);
104 	if (XVA_ISSET_REQ(xvap, XAT_SYSTEM))
105 		xoap->xoa_system = ((*attrs & XAT0_SYSTEM) != 0);
106 	if (XVA_ISSET_REQ(xvap, XAT_ARCHIVE))
107 		xoap->xoa_archive = ((*attrs & XAT0_ARCHIVE) != 0);
108 	if (XVA_ISSET_REQ(xvap, XAT_READONLY))
109 		xoap->xoa_readonly = ((*attrs & XAT0_READONLY) != 0);
110 	if (XVA_ISSET_REQ(xvap, XAT_IMMUTABLE))
111 		xoap->xoa_immutable = ((*attrs & XAT0_IMMUTABLE) != 0);
112 	if (XVA_ISSET_REQ(xvap, XAT_NOUNLINK))
113 		xoap->xoa_nounlink = ((*attrs & XAT0_NOUNLINK) != 0);
114 	if (XVA_ISSET_REQ(xvap, XAT_APPENDONLY))
115 		xoap->xoa_appendonly = ((*attrs & XAT0_APPENDONLY) != 0);
116 	if (XVA_ISSET_REQ(xvap, XAT_NODUMP))
117 		xoap->xoa_nodump = ((*attrs & XAT0_NODUMP) != 0);
118 	if (XVA_ISSET_REQ(xvap, XAT_OPAQUE))
119 		xoap->xoa_opaque = ((*attrs & XAT0_OPAQUE) != 0);
120 	if (XVA_ISSET_REQ(xvap, XAT_AV_MODIFIED))
121 		xoap->xoa_av_modified = ((*attrs & XAT0_AV_MODIFIED) != 0);
122 	if (XVA_ISSET_REQ(xvap, XAT_AV_QUARANTINED))
123 		xoap->xoa_av_quarantined =
124 		    ((*attrs & XAT0_AV_QUARANTINED) != 0);
125 	if (XVA_ISSET_REQ(xvap, XAT_CREATETIME))
126 		ZFS_TIME_DECODE(&xoap->xoa_createtime, crtime);
127 	if (XVA_ISSET_REQ(xvap, XAT_AV_SCANSTAMP))
128 		bcopy(scanstamp, xoap->xoa_av_scanstamp, AV_SCANSTAMP_SZ);
129 	if (XVA_ISSET_REQ(xvap, XAT_REPARSE))
130 		xoap->xoa_reparse = ((*attrs & XAT0_REPARSE) != 0);
131 	if (XVA_ISSET_REQ(xvap, XAT_OFFLINE))
132 		xoap->xoa_offline = ((*attrs & XAT0_OFFLINE) != 0);
133 	if (XVA_ISSET_REQ(xvap, XAT_SPARSE))
134 		xoap->xoa_sparse = ((*attrs & XAT0_SPARSE) != 0);
135 }
136 
137 static int
138 zfs_replay_domain_cnt(uint64_t uid, uint64_t gid)
139 {
140 	uint64_t uid_idx;
141 	uint64_t gid_idx;
142 	int domcnt = 0;
143 
144 	uid_idx = FUID_INDEX(uid);
145 	gid_idx = FUID_INDEX(gid);
146 	if (uid_idx)
147 		domcnt++;
148 	if (gid_idx > 0 && gid_idx != uid_idx)
149 		domcnt++;
150 
151 	return (domcnt);
152 }
153 
154 static void *
155 zfs_replay_fuid_domain_common(zfs_fuid_info_t *fuid_infop, void *start,
156     int domcnt)
157 {
158 	int i;
159 
160 	for (i = 0; i != domcnt; i++) {
161 		fuid_infop->z_domain_table[i] = start;
162 		start = (caddr_t)start + strlen(start) + 1;
163 	}
164 
165 	return (start);
166 }
167 
168 /*
169  * Set the uid/gid in the fuid_info structure.
170  */
171 static void
172 zfs_replay_fuid_ugid(zfs_fuid_info_t *fuid_infop, uint64_t uid, uint64_t gid)
173 {
174 	/*
175 	 * If owner or group are log specific FUIDs then slurp up
176 	 * domain information and build zfs_fuid_info_t
177 	 */
178 	if (IS_EPHEMERAL(uid))
179 		fuid_infop->z_fuid_owner = uid;
180 
181 	if (IS_EPHEMERAL(gid))
182 		fuid_infop->z_fuid_group = gid;
183 }
184 
185 /*
186  * Load fuid domains into fuid_info_t
187  */
188 static zfs_fuid_info_t *
189 zfs_replay_fuid_domain(void *buf, void **end, uint64_t uid, uint64_t gid)
190 {
191 	int domcnt;
192 
193 	zfs_fuid_info_t *fuid_infop;
194 
195 	fuid_infop = zfs_fuid_info_alloc();
196 
197 	domcnt = zfs_replay_domain_cnt(uid, gid);
198 
199 	if (domcnt == 0)
200 		return (fuid_infop);
201 
202 	fuid_infop->z_domain_table =
203 	    kmem_zalloc(domcnt * sizeof (char **), KM_SLEEP);
204 
205 	zfs_replay_fuid_ugid(fuid_infop, uid, gid);
206 
207 	fuid_infop->z_domain_cnt = domcnt;
208 	*end = zfs_replay_fuid_domain_common(fuid_infop, buf, domcnt);
209 	return (fuid_infop);
210 }
211 
212 /*
213  * load zfs_fuid_t's and fuid_domains into fuid_info_t
214  */
215 static zfs_fuid_info_t *
216 zfs_replay_fuids(void *start, void **end, int idcnt, int domcnt, uint64_t uid,
217     uint64_t gid)
218 {
219 	uint64_t *log_fuid = (uint64_t *)start;
220 	zfs_fuid_info_t *fuid_infop;
221 	int i;
222 
223 	fuid_infop = zfs_fuid_info_alloc();
224 	fuid_infop->z_domain_cnt = domcnt;
225 
226 	fuid_infop->z_domain_table =
227 	    kmem_zalloc(domcnt * sizeof (char **), KM_SLEEP);
228 
229 	for (i = 0; i != idcnt; i++) {
230 		zfs_fuid_t *zfuid;
231 
232 		zfuid = kmem_alloc(sizeof (zfs_fuid_t), KM_SLEEP);
233 		zfuid->z_logfuid = *log_fuid;
234 		zfuid->z_id = -1;
235 		zfuid->z_domidx = 0;
236 		list_insert_tail(&fuid_infop->z_fuids, zfuid);
237 		log_fuid++;
238 	}
239 
240 	zfs_replay_fuid_ugid(fuid_infop, uid, gid);
241 
242 	*end = zfs_replay_fuid_domain_common(fuid_infop, log_fuid, domcnt);
243 	return (fuid_infop);
244 }
245 
246 static void
247 zfs_replay_swap_attrs(lr_attr_t *lrattr)
248 {
249 	/* swap the lr_attr structure */
250 	byteswap_uint32_array(lrattr, sizeof (*lrattr));
251 	/* swap the bitmap */
252 	byteswap_uint32_array(lrattr + 1, (lrattr->lr_attr_masksize - 1) *
253 	    sizeof (uint32_t));
254 	/* swap the attributes, create time + 64 bit word for attributes */
255 	byteswap_uint64_array((caddr_t)(lrattr + 1) + (sizeof (uint32_t) *
256 	    (lrattr->lr_attr_masksize - 1)), 3 * sizeof (uint64_t));
257 }
258 
259 /*
260  * Replay file create with optional ACL, xvattr information as well
261  * as option FUID information.
262  */
263 static int
264 zfs_replay_create_acl(zfsvfs_t *zfsvfs,
265     lr_acl_create_t *lracl, boolean_t byteswap)
266 {
267 	char *name = NULL;		/* location determined later */
268 	lr_create_t *lr = (lr_create_t *)lracl;
269 	znode_t *dzp;
270 	vnode_t *vp = NULL;
271 	xvattr_t xva;
272 	int vflg = 0;
273 	vsecattr_t vsec = { 0 };
274 	lr_attr_t *lrattr;
275 	void *aclstart;
276 	void *fuidstart;
277 	size_t xvatlen = 0;
278 	uint64_t txtype;
279 	int error;
280 
281 	txtype = (lr->lr_common.lrc_txtype & ~TX_CI);
282 	if (byteswap) {
283 		byteswap_uint64_array(lracl, sizeof (*lracl));
284 		if (txtype == TX_CREATE_ACL_ATTR ||
285 		    txtype == TX_MKDIR_ACL_ATTR) {
286 			lrattr = (lr_attr_t *)(caddr_t)(lracl + 1);
287 			zfs_replay_swap_attrs(lrattr);
288 			xvatlen = ZIL_XVAT_SIZE(lrattr->lr_attr_masksize);
289 		}
290 
291 		aclstart = (caddr_t)(lracl + 1) + xvatlen;
292 		zfs_ace_byteswap(aclstart, lracl->lr_acl_bytes, B_FALSE);
293 		/* swap fuids */
294 		if (lracl->lr_fuidcnt) {
295 			byteswap_uint64_array((caddr_t)aclstart +
296 			    ZIL_ACE_LENGTH(lracl->lr_acl_bytes),
297 			    lracl->lr_fuidcnt * sizeof (uint64_t));
298 		}
299 	}
300 
301 	if ((error = zfs_zget(zfsvfs, lr->lr_doid, &dzp)) != 0)
302 		return (error);
303 
304 	xva_init(&xva);
305 	zfs_init_vattr(&xva.xva_vattr, AT_TYPE | AT_MODE | AT_UID | AT_GID,
306 	    lr->lr_mode, lr->lr_uid, lr->lr_gid, lr->lr_rdev, lr->lr_foid);
307 
308 	/*
309 	 * All forms of zfs create (create, mkdir, mkxattrdir, symlink)
310 	 * eventually end up in zfs_mknode(), which assigns the object's
311 	 * creation time and generation number.  The generic VOP_CREATE()
312 	 * doesn't have either concept, so we smuggle the values inside
313 	 * the vattr's otherwise unused va_ctime and va_nblocks fields.
314 	 */
315 	ZFS_TIME_DECODE(&xva.xva_vattr.va_ctime, lr->lr_crtime);
316 	xva.xva_vattr.va_nblocks = lr->lr_gen;
317 
318 	error = dmu_object_info(zfsvfs->z_os, lr->lr_foid, NULL);
319 	if (error != ENOENT)
320 		goto bail;
321 
322 	if (lr->lr_common.lrc_txtype & TX_CI)
323 		vflg |= FIGNORECASE;
324 	switch (txtype) {
325 	case TX_CREATE_ACL:
326 		aclstart = (caddr_t)(lracl + 1);
327 		fuidstart = (caddr_t)aclstart +
328 		    ZIL_ACE_LENGTH(lracl->lr_acl_bytes);
329 		zfsvfs->z_fuid_replay = zfs_replay_fuids(fuidstart,
330 		    (void *)&name, lracl->lr_fuidcnt, lracl->lr_domcnt,
331 		    lr->lr_uid, lr->lr_gid);
332 		/*FALLTHROUGH*/
333 	case TX_CREATE_ACL_ATTR:
334 		if (name == NULL) {
335 			lrattr = (lr_attr_t *)(caddr_t)(lracl + 1);
336 			xvatlen = ZIL_XVAT_SIZE(lrattr->lr_attr_masksize);
337 			xva.xva_vattr.va_mask |= AT_XVATTR;
338 			zfs_replay_xvattr(lrattr, &xva);
339 		}
340 		vsec.vsa_mask = VSA_ACE | VSA_ACE_ACLFLAGS;
341 		vsec.vsa_aclentp = (caddr_t)(lracl + 1) + xvatlen;
342 		vsec.vsa_aclcnt = lracl->lr_aclcnt;
343 		vsec.vsa_aclentsz = lracl->lr_acl_bytes;
344 		vsec.vsa_aclflags = lracl->lr_acl_flags;
345 		if (zfsvfs->z_fuid_replay == NULL) {
346 			fuidstart = (caddr_t)(lracl + 1) + xvatlen +
347 			    ZIL_ACE_LENGTH(lracl->lr_acl_bytes);
348 			zfsvfs->z_fuid_replay =
349 			    zfs_replay_fuids(fuidstart,
350 			    (void *)&name, lracl->lr_fuidcnt, lracl->lr_domcnt,
351 			    lr->lr_uid, lr->lr_gid);
352 		}
353 
354 		error = VOP_CREATE(ZTOV(dzp), name, &xva.xva_vattr,
355 		    0, 0, &vp, kcred, vflg, NULL, &vsec);
356 		break;
357 	case TX_MKDIR_ACL:
358 		aclstart = (caddr_t)(lracl + 1);
359 		fuidstart = (caddr_t)aclstart +
360 		    ZIL_ACE_LENGTH(lracl->lr_acl_bytes);
361 		zfsvfs->z_fuid_replay = zfs_replay_fuids(fuidstart,
362 		    (void *)&name, lracl->lr_fuidcnt, lracl->lr_domcnt,
363 		    lr->lr_uid, lr->lr_gid);
364 		/*FALLTHROUGH*/
365 	case TX_MKDIR_ACL_ATTR:
366 		if (name == NULL) {
367 			lrattr = (lr_attr_t *)(caddr_t)(lracl + 1);
368 			xvatlen = ZIL_XVAT_SIZE(lrattr->lr_attr_masksize);
369 			zfs_replay_xvattr(lrattr, &xva);
370 		}
371 		vsec.vsa_mask = VSA_ACE | VSA_ACE_ACLFLAGS;
372 		vsec.vsa_aclentp = (caddr_t)(lracl + 1) + xvatlen;
373 		vsec.vsa_aclcnt = lracl->lr_aclcnt;
374 		vsec.vsa_aclentsz = lracl->lr_acl_bytes;
375 		vsec.vsa_aclflags = lracl->lr_acl_flags;
376 		if (zfsvfs->z_fuid_replay == NULL) {
377 			fuidstart = (caddr_t)(lracl + 1) + xvatlen +
378 			    ZIL_ACE_LENGTH(lracl->lr_acl_bytes);
379 			zfsvfs->z_fuid_replay =
380 			    zfs_replay_fuids(fuidstart,
381 			    (void *)&name, lracl->lr_fuidcnt, lracl->lr_domcnt,
382 			    lr->lr_uid, lr->lr_gid);
383 		}
384 		error = VOP_MKDIR(ZTOV(dzp), name, &xva.xva_vattr,
385 		    &vp, kcred, NULL, vflg, &vsec);
386 		break;
387 	default:
388 		error = ENOTSUP;
389 	}
390 
391 bail:
392 	if (error == 0 && vp != NULL)
393 		VN_RELE(vp);
394 
395 	VN_RELE(ZTOV(dzp));
396 
397 	if (zfsvfs->z_fuid_replay)
398 		zfs_fuid_info_free(zfsvfs->z_fuid_replay);
399 	zfsvfs->z_fuid_replay = NULL;
400 
401 	return (error);
402 }
403 
404 static int
405 zfs_replay_create(zfsvfs_t *zfsvfs, lr_create_t *lr, boolean_t byteswap)
406 {
407 	char *name = NULL;		/* location determined later */
408 	char *link;			/* symlink content follows name */
409 	znode_t *dzp;
410 	vnode_t *vp = NULL;
411 	xvattr_t xva;
412 	int vflg = 0;
413 	size_t lrsize = sizeof (lr_create_t);
414 	lr_attr_t *lrattr;
415 	void *start;
416 	size_t xvatlen;
417 	uint64_t txtype;
418 	int error;
419 
420 	txtype = (lr->lr_common.lrc_txtype & ~TX_CI);
421 	if (byteswap) {
422 		byteswap_uint64_array(lr, sizeof (*lr));
423 		if (txtype == TX_CREATE_ATTR || txtype == TX_MKDIR_ATTR)
424 			zfs_replay_swap_attrs((lr_attr_t *)(lr + 1));
425 	}
426 
427 
428 	if ((error = zfs_zget(zfsvfs, lr->lr_doid, &dzp)) != 0)
429 		return (error);
430 
431 	xva_init(&xva);
432 	zfs_init_vattr(&xva.xva_vattr, AT_TYPE | AT_MODE | AT_UID | AT_GID,
433 	    lr->lr_mode, lr->lr_uid, lr->lr_gid, lr->lr_rdev, lr->lr_foid);
434 
435 	/*
436 	 * All forms of zfs create (create, mkdir, mkxattrdir, symlink)
437 	 * eventually end up in zfs_mknode(), which assigns the object's
438 	 * creation time and generation number.  The generic VOP_CREATE()
439 	 * doesn't have either concept, so we smuggle the values inside
440 	 * the vattr's otherwise unused va_ctime and va_nblocks fields.
441 	 */
442 	ZFS_TIME_DECODE(&xva.xva_vattr.va_ctime, lr->lr_crtime);
443 	xva.xva_vattr.va_nblocks = lr->lr_gen;
444 
445 	error = dmu_object_info(zfsvfs->z_os, lr->lr_foid, NULL);
446 	if (error != ENOENT)
447 		goto out;
448 
449 	if (lr->lr_common.lrc_txtype & TX_CI)
450 		vflg |= FIGNORECASE;
451 
452 	/*
453 	 * Symlinks don't have fuid info, and CIFS never creates
454 	 * symlinks.
455 	 *
456 	 * The _ATTR versions will grab the fuid info in their subcases.
457 	 */
458 	if ((int)lr->lr_common.lrc_txtype != TX_SYMLINK &&
459 	    (int)lr->lr_common.lrc_txtype != TX_MKDIR_ATTR &&
460 	    (int)lr->lr_common.lrc_txtype != TX_CREATE_ATTR) {
461 		start = (lr + 1);
462 		zfsvfs->z_fuid_replay =
463 		    zfs_replay_fuid_domain(start, &start,
464 		    lr->lr_uid, lr->lr_gid);
465 	}
466 
467 	switch (txtype) {
468 	case TX_CREATE_ATTR:
469 		lrattr = (lr_attr_t *)(caddr_t)(lr + 1);
470 		xvatlen = ZIL_XVAT_SIZE(lrattr->lr_attr_masksize);
471 		zfs_replay_xvattr((lr_attr_t *)((caddr_t)lr + lrsize), &xva);
472 		start = (caddr_t)(lr + 1) + xvatlen;
473 		zfsvfs->z_fuid_replay =
474 		    zfs_replay_fuid_domain(start, &start,
475 		    lr->lr_uid, lr->lr_gid);
476 		name = (char *)start;
477 
478 		/*FALLTHROUGH*/
479 	case TX_CREATE:
480 		if (name == NULL)
481 			name = (char *)start;
482 
483 		error = VOP_CREATE(ZTOV(dzp), name, &xva.xva_vattr,
484 		    0, 0, &vp, kcred, vflg, NULL, NULL);
485 		break;
486 	case TX_MKDIR_ATTR:
487 		lrattr = (lr_attr_t *)(caddr_t)(lr + 1);
488 		xvatlen = ZIL_XVAT_SIZE(lrattr->lr_attr_masksize);
489 		zfs_replay_xvattr((lr_attr_t *)((caddr_t)lr + lrsize), &xva);
490 		start = (caddr_t)(lr + 1) + xvatlen;
491 		zfsvfs->z_fuid_replay =
492 		    zfs_replay_fuid_domain(start, &start,
493 		    lr->lr_uid, lr->lr_gid);
494 		name = (char *)start;
495 
496 		/*FALLTHROUGH*/
497 	case TX_MKDIR:
498 		if (name == NULL)
499 			name = (char *)(lr + 1);
500 
501 		error = VOP_MKDIR(ZTOV(dzp), name, &xva.xva_vattr,
502 		    &vp, kcred, NULL, vflg, NULL);
503 		break;
504 	case TX_MKXATTR:
505 		error = zfs_make_xattrdir(dzp, &xva.xva_vattr, &vp, kcred);
506 		break;
507 	case TX_SYMLINK:
508 		name = (char *)(lr + 1);
509 		link = name + strlen(name) + 1;
510 		error = VOP_SYMLINK(ZTOV(dzp), name, &xva.xva_vattr,
511 		    link, kcred, NULL, vflg);
512 		break;
513 	default:
514 		error = ENOTSUP;
515 	}
516 
517 out:
518 	if (error == 0 && vp != NULL)
519 		VN_RELE(vp);
520 
521 	VN_RELE(ZTOV(dzp));
522 
523 	if (zfsvfs->z_fuid_replay)
524 		zfs_fuid_info_free(zfsvfs->z_fuid_replay);
525 	zfsvfs->z_fuid_replay = NULL;
526 	return (error);
527 }
528 
529 static int
530 zfs_replay_remove(zfsvfs_t *zfsvfs, lr_remove_t *lr, boolean_t byteswap)
531 {
532 	char *name = (char *)(lr + 1);	/* name follows lr_remove_t */
533 	znode_t *dzp;
534 	int error;
535 	int vflg = 0;
536 
537 	if (byteswap)
538 		byteswap_uint64_array(lr, sizeof (*lr));
539 
540 	if ((error = zfs_zget(zfsvfs, lr->lr_doid, &dzp)) != 0)
541 		return (error);
542 
543 	if (lr->lr_common.lrc_txtype & TX_CI)
544 		vflg |= FIGNORECASE;
545 
546 	switch ((int)lr->lr_common.lrc_txtype) {
547 	case TX_REMOVE:
548 		error = VOP_REMOVE(ZTOV(dzp), name, kcred, NULL, vflg);
549 		break;
550 	case TX_RMDIR:
551 		error = VOP_RMDIR(ZTOV(dzp), name, NULL, kcred, NULL, vflg);
552 		break;
553 	default:
554 		error = ENOTSUP;
555 	}
556 
557 	VN_RELE(ZTOV(dzp));
558 
559 	return (error);
560 }
561 
562 static int
563 zfs_replay_link(zfsvfs_t *zfsvfs, lr_link_t *lr, boolean_t byteswap)
564 {
565 	char *name = (char *)(lr + 1);	/* name follows lr_link_t */
566 	znode_t *dzp, *zp;
567 	int error;
568 	int vflg = 0;
569 
570 	if (byteswap)
571 		byteswap_uint64_array(lr, sizeof (*lr));
572 
573 	if ((error = zfs_zget(zfsvfs, lr->lr_doid, &dzp)) != 0)
574 		return (error);
575 
576 	if ((error = zfs_zget(zfsvfs, lr->lr_link_obj, &zp)) != 0) {
577 		VN_RELE(ZTOV(dzp));
578 		return (error);
579 	}
580 
581 	if (lr->lr_common.lrc_txtype & TX_CI)
582 		vflg |= FIGNORECASE;
583 
584 	error = VOP_LINK(ZTOV(dzp), ZTOV(zp), name, kcred, NULL, vflg);
585 
586 	VN_RELE(ZTOV(zp));
587 	VN_RELE(ZTOV(dzp));
588 
589 	return (error);
590 }
591 
592 static int
593 zfs_replay_rename(zfsvfs_t *zfsvfs, lr_rename_t *lr, boolean_t byteswap)
594 {
595 	char *sname = (char *)(lr + 1);	/* sname and tname follow lr_rename_t */
596 	char *tname = sname + strlen(sname) + 1;
597 	znode_t *sdzp, *tdzp;
598 	int error;
599 	int vflg = 0;
600 
601 	if (byteswap)
602 		byteswap_uint64_array(lr, sizeof (*lr));
603 
604 	if ((error = zfs_zget(zfsvfs, lr->lr_sdoid, &sdzp)) != 0)
605 		return (error);
606 
607 	if ((error = zfs_zget(zfsvfs, lr->lr_tdoid, &tdzp)) != 0) {
608 		VN_RELE(ZTOV(sdzp));
609 		return (error);
610 	}
611 
612 	if (lr->lr_common.lrc_txtype & TX_CI)
613 		vflg |= FIGNORECASE;
614 
615 	error = VOP_RENAME(ZTOV(sdzp), sname, ZTOV(tdzp), tname, kcred,
616 	    NULL, vflg);
617 
618 	VN_RELE(ZTOV(tdzp));
619 	VN_RELE(ZTOV(sdzp));
620 
621 	return (error);
622 }
623 
624 static int
625 zfs_replay_write(zfsvfs_t *zfsvfs, lr_write_t *lr, boolean_t byteswap)
626 {
627 	char *data = (char *)(lr + 1);	/* data follows lr_write_t */
628 	znode_t	*zp;
629 	int error;
630 	ssize_t resid;
631 	uint64_t eod, offset, length;
632 
633 	if (byteswap)
634 		byteswap_uint64_array(lr, sizeof (*lr));
635 
636 	if ((error = zfs_zget(zfsvfs, lr->lr_foid, &zp)) != 0) {
637 		/*
638 		 * As we can log writes out of order, it's possible the
639 		 * file has been removed. In this case just drop the write
640 		 * and return success.
641 		 */
642 		if (error == ENOENT)
643 			error = 0;
644 		return (error);
645 	}
646 
647 	offset = lr->lr_offset;
648 	length = lr->lr_length;
649 	eod = offset + length;	/* end of data for this write */
650 
651 	/*
652 	 * This may be a write from a dmu_sync() for a whole block,
653 	 * and may extend beyond the current end of the file.
654 	 * We can't just replay what was written for this TX_WRITE as
655 	 * a future TX_WRITE2 may extend the eof and the data for that
656 	 * write needs to be there. So we write the whole block and
657 	 * reduce the eof. This needs to be done within the single dmu
658 	 * transaction created within vn_rdwr -> zfs_write. So a possible
659 	 * new end of file is passed through in zfsvfs->z_replay_eof
660 	 */
661 
662 	zfsvfs->z_replay_eof = 0; /* 0 means don't change end of file */
663 
664 	/* If it's a dmu_sync() block, write the whole block */
665 	if (lr->lr_common.lrc_reclen == sizeof (lr_write_t)) {
666 		uint64_t blocksize = BP_GET_LSIZE(&lr->lr_blkptr);
667 		if (length < blocksize) {
668 			offset -= offset % blocksize;
669 			length = blocksize;
670 		}
671 		if (zp->z_size < eod)
672 			zfsvfs->z_replay_eof = eod;
673 	}
674 
675 	error = vn_rdwr(UIO_WRITE, ZTOV(zp), data, length, offset,
676 	    UIO_SYSSPACE, 0, RLIM64_INFINITY, kcred, &resid);
677 
678 	VN_RELE(ZTOV(zp));
679 	zfsvfs->z_replay_eof = 0;	/* safety */
680 
681 	return (error);
682 }
683 
684 /*
685  * TX_WRITE2 are only generated when dmu_sync() returns EALREADY
686  * meaning the pool block is already being synced. So now that we always write
687  * out full blocks, all we have to do is expand the eof if
688  * the file is grown.
689  */
690 static int
691 zfs_replay_write2(zfsvfs_t *zfsvfs, lr_write_t *lr, boolean_t byteswap)
692 {
693 	znode_t	*zp;
694 	int error;
695 	uint64_t end;
696 
697 	if (byteswap)
698 		byteswap_uint64_array(lr, sizeof (*lr));
699 
700 	if ((error = zfs_zget(zfsvfs, lr->lr_foid, &zp)) != 0)
701 		return (error);
702 
703 top:
704 	end = lr->lr_offset + lr->lr_length;
705 	if (end > zp->z_size) {
706 		dmu_tx_t *tx = dmu_tx_create(zfsvfs->z_os);
707 
708 		zp->z_size = end;
709 		dmu_tx_hold_sa(tx, zp->z_sa_hdl, B_FALSE);
710 		error = dmu_tx_assign(tx, TXG_WAIT);
711 		if (error) {
712 			VN_RELE(ZTOV(zp));
713 			if (error == ERESTART) {
714 				dmu_tx_wait(tx);
715 				dmu_tx_abort(tx);
716 				goto top;
717 			}
718 			dmu_tx_abort(tx);
719 			return (error);
720 		}
721 		(void) sa_update(zp->z_sa_hdl, SA_ZPL_SIZE(zfsvfs),
722 		    (void *)&zp->z_size, sizeof (uint64_t), tx);
723 
724 		/* Ensure the replayed seq is updated */
725 		(void) zil_replaying(zfsvfs->z_log, tx);
726 
727 		dmu_tx_commit(tx);
728 	}
729 
730 	VN_RELE(ZTOV(zp));
731 
732 	return (error);
733 }
734 
735 static int
736 zfs_replay_truncate(zfsvfs_t *zfsvfs, lr_truncate_t *lr, boolean_t byteswap)
737 {
738 	znode_t *zp;
739 	flock64_t fl;
740 	int error;
741 
742 	if (byteswap)
743 		byteswap_uint64_array(lr, sizeof (*lr));
744 
745 	if ((error = zfs_zget(zfsvfs, lr->lr_foid, &zp)) != 0)
746 		return (error);
747 
748 	bzero(&fl, sizeof (fl));
749 	fl.l_type = F_WRLCK;
750 	fl.l_whence = 0;
751 	fl.l_start = lr->lr_offset;
752 	fl.l_len = lr->lr_length;
753 
754 	error = VOP_SPACE(ZTOV(zp), F_FREESP, &fl, FWRITE | FOFFMAX,
755 	    lr->lr_offset, kcred, NULL);
756 
757 	VN_RELE(ZTOV(zp));
758 
759 	return (error);
760 }
761 
762 static int
763 zfs_replay_setattr(zfsvfs_t *zfsvfs, lr_setattr_t *lr, boolean_t byteswap)
764 {
765 	znode_t *zp;
766 	xvattr_t xva;
767 	vattr_t *vap = &xva.xva_vattr;
768 	int error;
769 	void *start;
770 
771 	xva_init(&xva);
772 	if (byteswap) {
773 		byteswap_uint64_array(lr, sizeof (*lr));
774 
775 		if ((lr->lr_mask & AT_XVATTR) &&
776 		    zfsvfs->z_version >= ZPL_VERSION_INITIAL)
777 			zfs_replay_swap_attrs((lr_attr_t *)(lr + 1));
778 	}
779 
780 	if ((error = zfs_zget(zfsvfs, lr->lr_foid, &zp)) != 0)
781 		return (error);
782 
783 	zfs_init_vattr(vap, lr->lr_mask, lr->lr_mode,
784 	    lr->lr_uid, lr->lr_gid, 0, lr->lr_foid);
785 
786 	vap->va_size = lr->lr_size;
787 	ZFS_TIME_DECODE(&vap->va_atime, lr->lr_atime);
788 	ZFS_TIME_DECODE(&vap->va_mtime, lr->lr_mtime);
789 
790 	/*
791 	 * Fill in xvattr_t portions if necessary.
792 	 */
793 
794 	start = (lr_setattr_t *)(lr + 1);
795 	if (vap->va_mask & AT_XVATTR) {
796 		zfs_replay_xvattr((lr_attr_t *)start, &xva);
797 		start = (caddr_t)start +
798 		    ZIL_XVAT_SIZE(((lr_attr_t *)start)->lr_attr_masksize);
799 	} else
800 		xva.xva_vattr.va_mask &= ~AT_XVATTR;
801 
802 	zfsvfs->z_fuid_replay = zfs_replay_fuid_domain(start, &start,
803 	    lr->lr_uid, lr->lr_gid);
804 
805 	error = VOP_SETATTR(ZTOV(zp), vap, 0, kcred, NULL);
806 
807 	zfs_fuid_info_free(zfsvfs->z_fuid_replay);
808 	zfsvfs->z_fuid_replay = NULL;
809 	VN_RELE(ZTOV(zp));
810 
811 	return (error);
812 }
813 
814 static int
815 zfs_replay_acl_v0(zfsvfs_t *zfsvfs, lr_acl_v0_t *lr, boolean_t byteswap)
816 {
817 	ace_t *ace = (ace_t *)(lr + 1);	/* ace array follows lr_acl_t */
818 	vsecattr_t vsa;
819 	znode_t *zp;
820 	int error;
821 
822 	if (byteswap) {
823 		byteswap_uint64_array(lr, sizeof (*lr));
824 		zfs_oldace_byteswap(ace, lr->lr_aclcnt);
825 	}
826 
827 	if ((error = zfs_zget(zfsvfs, lr->lr_foid, &zp)) != 0)
828 		return (error);
829 
830 	bzero(&vsa, sizeof (vsa));
831 	vsa.vsa_mask = VSA_ACE | VSA_ACECNT;
832 	vsa.vsa_aclcnt = lr->lr_aclcnt;
833 	vsa.vsa_aclentsz = sizeof (ace_t) * vsa.vsa_aclcnt;
834 	vsa.vsa_aclflags = 0;
835 	vsa.vsa_aclentp = ace;
836 
837 	error = VOP_SETSECATTR(ZTOV(zp), &vsa, 0, kcred, NULL);
838 
839 	VN_RELE(ZTOV(zp));
840 
841 	return (error);
842 }
843 
844 /*
845  * Replaying ACLs is complicated by FUID support.
846  * The log record may contain some optional data
847  * to be used for replaying FUID's.  These pieces
848  * are the actual FUIDs that were created initially.
849  * The FUID table index may no longer be valid and
850  * during zfs_create() a new index may be assigned.
851  * Because of this the log will contain the original
852  * doman+rid in order to create a new FUID.
853  *
854  * The individual ACEs may contain an ephemeral uid/gid which is no
855  * longer valid and will need to be replaced with an actual FUID.
856  *
857  */
858 static int
859 zfs_replay_acl(zfsvfs_t *zfsvfs, lr_acl_t *lr, boolean_t byteswap)
860 {
861 	ace_t *ace = (ace_t *)(lr + 1);
862 	vsecattr_t vsa;
863 	znode_t *zp;
864 	int error;
865 
866 	if (byteswap) {
867 		byteswap_uint64_array(lr, sizeof (*lr));
868 		zfs_ace_byteswap(ace, lr->lr_acl_bytes, B_FALSE);
869 		if (lr->lr_fuidcnt) {
870 			byteswap_uint64_array((caddr_t)ace +
871 			    ZIL_ACE_LENGTH(lr->lr_acl_bytes),
872 			    lr->lr_fuidcnt * sizeof (uint64_t));
873 		}
874 	}
875 
876 	if ((error = zfs_zget(zfsvfs, lr->lr_foid, &zp)) != 0)
877 		return (error);
878 
879 	bzero(&vsa, sizeof (vsa));
880 	vsa.vsa_mask = VSA_ACE | VSA_ACECNT | VSA_ACE_ACLFLAGS;
881 	vsa.vsa_aclcnt = lr->lr_aclcnt;
882 	vsa.vsa_aclentp = ace;
883 	vsa.vsa_aclentsz = lr->lr_acl_bytes;
884 	vsa.vsa_aclflags = lr->lr_acl_flags;
885 
886 	if (lr->lr_fuidcnt) {
887 		void *fuidstart = (caddr_t)ace +
888 		    ZIL_ACE_LENGTH(lr->lr_acl_bytes);
889 
890 		zfsvfs->z_fuid_replay =
891 		    zfs_replay_fuids(fuidstart, &fuidstart,
892 		    lr->lr_fuidcnt, lr->lr_domcnt, 0, 0);
893 	}
894 
895 	error = VOP_SETSECATTR(ZTOV(zp), &vsa, 0, kcred, NULL);
896 
897 	if (zfsvfs->z_fuid_replay)
898 		zfs_fuid_info_free(zfsvfs->z_fuid_replay);
899 
900 	zfsvfs->z_fuid_replay = NULL;
901 	VN_RELE(ZTOV(zp));
902 
903 	return (error);
904 }
905 
906 /*
907  * Callback vectors for replaying records
908  */
909 zil_replay_func_t *zfs_replay_vector[TX_MAX_TYPE] = {
910 	zfs_replay_error,	/* 0 no such transaction type */
911 	zfs_replay_create,	/* TX_CREATE */
912 	zfs_replay_create,	/* TX_MKDIR */
913 	zfs_replay_create,	/* TX_MKXATTR */
914 	zfs_replay_create,	/* TX_SYMLINK */
915 	zfs_replay_remove,	/* TX_REMOVE */
916 	zfs_replay_remove,	/* TX_RMDIR */
917 	zfs_replay_link,	/* TX_LINK */
918 	zfs_replay_rename,	/* TX_RENAME */
919 	zfs_replay_write,	/* TX_WRITE */
920 	zfs_replay_truncate,	/* TX_TRUNCATE */
921 	zfs_replay_setattr,	/* TX_SETATTR */
922 	zfs_replay_acl_v0,	/* TX_ACL_V0 */
923 	zfs_replay_acl,		/* TX_ACL */
924 	zfs_replay_create_acl,	/* TX_CREATE_ACL */
925 	zfs_replay_create,	/* TX_CREATE_ATTR */
926 	zfs_replay_create_acl,	/* TX_CREATE_ACL_ATTR */
927 	zfs_replay_create_acl,	/* TX_MKDIR_ACL */
928 	zfs_replay_create,	/* TX_MKDIR_ATTR */
929 	zfs_replay_create_acl,	/* TX_MKDIR_ACL_ATTR */
930 	zfs_replay_write2,	/* TX_WRITE2 */
931 };
932