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/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/acl.h>
45 #include <sys/atomic.h>
46 #include <sys/cred.h>
47 #include <sys/namei.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
zfs_init_vattr(vattr_t * vap,uint64_t mask,uint64_t mode,uint64_t uid,uint64_t gid,uint64_t rdev,uint64_t nodeid)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 VATTR_NULL(vap);
60 vap->va_mask = (uint_t)mask;
61 if (mask & AT_TYPE)
62 vap->va_type = IFTOVT(mode);
63 if (mask & AT_MODE)
64 vap->va_mode = mode & MODEMASK;
65 if (mask & AT_UID)
66 vap->va_uid = (uid_t)(IS_EPHEMERAL(uid)) ? -1 : uid;
67 if (mask & AT_GID)
68 vap->va_gid = (gid_t)(IS_EPHEMERAL(gid)) ? -1 : gid;
69 vap->va_rdev = zfs_cmpldev(rdev);
70 vap->va_nodeid = nodeid;
71 }
72
73 /* ARGSUSED */
74 static int
zfs_replay_error(zfsvfs_t * zfsvfs,lr_t * lr,boolean_t byteswap)75 zfs_replay_error(zfsvfs_t *zfsvfs, lr_t *lr, boolean_t byteswap)
76 {
77 return (SET_ERROR(ENOTSUP));
78 }
79
80 static void
zfs_replay_xvattr(lr_attr_t * lrattr,xvattr_t * xvap)81 zfs_replay_xvattr(lr_attr_t *lrattr, xvattr_t *xvap)
82 {
83 xoptattr_t *xoap = NULL;
84 uint64_t *attrs;
85 uint64_t *crtime;
86 uint32_t *bitmap;
87 void *scanstamp;
88 int i;
89
90 xvap->xva_vattr.va_mask |= AT_XVATTR;
91 if ((xoap = xva_getxoptattr(xvap)) == NULL) {
92 xvap->xva_vattr.va_mask &= ~AT_XVATTR; /* shouldn't happen */
93 return;
94 }
95
96 ASSERT(lrattr->lr_attr_masksize == xvap->xva_mapsize);
97
98 bitmap = &lrattr->lr_attr_bitmap;
99 for (i = 0; i != lrattr->lr_attr_masksize; i++, bitmap++)
100 xvap->xva_reqattrmap[i] = *bitmap;
101
102 attrs = (uint64_t *)(lrattr + lrattr->lr_attr_masksize - 1);
103 crtime = attrs + 1;
104 scanstamp = (caddr_t)(crtime + 2);
105
106 if (XVA_ISSET_REQ(xvap, XAT_HIDDEN))
107 xoap->xoa_hidden = ((*attrs & XAT0_HIDDEN) != 0);
108 if (XVA_ISSET_REQ(xvap, XAT_SYSTEM))
109 xoap->xoa_system = ((*attrs & XAT0_SYSTEM) != 0);
110 if (XVA_ISSET_REQ(xvap, XAT_ARCHIVE))
111 xoap->xoa_archive = ((*attrs & XAT0_ARCHIVE) != 0);
112 if (XVA_ISSET_REQ(xvap, XAT_READONLY))
113 xoap->xoa_readonly = ((*attrs & XAT0_READONLY) != 0);
114 if (XVA_ISSET_REQ(xvap, XAT_IMMUTABLE))
115 xoap->xoa_immutable = ((*attrs & XAT0_IMMUTABLE) != 0);
116 if (XVA_ISSET_REQ(xvap, XAT_NOUNLINK))
117 xoap->xoa_nounlink = ((*attrs & XAT0_NOUNLINK) != 0);
118 if (XVA_ISSET_REQ(xvap, XAT_APPENDONLY))
119 xoap->xoa_appendonly = ((*attrs & XAT0_APPENDONLY) != 0);
120 if (XVA_ISSET_REQ(xvap, XAT_NODUMP))
121 xoap->xoa_nodump = ((*attrs & XAT0_NODUMP) != 0);
122 if (XVA_ISSET_REQ(xvap, XAT_OPAQUE))
123 xoap->xoa_opaque = ((*attrs & XAT0_OPAQUE) != 0);
124 if (XVA_ISSET_REQ(xvap, XAT_AV_MODIFIED))
125 xoap->xoa_av_modified = ((*attrs & XAT0_AV_MODIFIED) != 0);
126 if (XVA_ISSET_REQ(xvap, XAT_AV_QUARANTINED))
127 xoap->xoa_av_quarantined =
128 ((*attrs & XAT0_AV_QUARANTINED) != 0);
129 if (XVA_ISSET_REQ(xvap, XAT_CREATETIME))
130 ZFS_TIME_DECODE(&xoap->xoa_createtime, crtime);
131 if (XVA_ISSET_REQ(xvap, XAT_AV_SCANSTAMP))
132 bcopy(scanstamp, xoap->xoa_av_scanstamp, AV_SCANSTAMP_SZ);
133 if (XVA_ISSET_REQ(xvap, XAT_REPARSE))
134 xoap->xoa_reparse = ((*attrs & XAT0_REPARSE) != 0);
135 if (XVA_ISSET_REQ(xvap, XAT_OFFLINE))
136 xoap->xoa_offline = ((*attrs & XAT0_OFFLINE) != 0);
137 if (XVA_ISSET_REQ(xvap, XAT_SPARSE))
138 xoap->xoa_sparse = ((*attrs & XAT0_SPARSE) != 0);
139 }
140
141 static int
zfs_replay_domain_cnt(uint64_t uid,uint64_t gid)142 zfs_replay_domain_cnt(uint64_t uid, uint64_t gid)
143 {
144 uint64_t uid_idx;
145 uint64_t gid_idx;
146 int domcnt = 0;
147
148 uid_idx = FUID_INDEX(uid);
149 gid_idx = FUID_INDEX(gid);
150 if (uid_idx)
151 domcnt++;
152 if (gid_idx > 0 && gid_idx != uid_idx)
153 domcnt++;
154
155 return (domcnt);
156 }
157
158 static void *
zfs_replay_fuid_domain_common(zfs_fuid_info_t * fuid_infop,void * start,int domcnt)159 zfs_replay_fuid_domain_common(zfs_fuid_info_t *fuid_infop, void *start,
160 int domcnt)
161 {
162 int i;
163
164 for (i = 0; i != domcnt; i++) {
165 fuid_infop->z_domain_table[i] = start;
166 start = (caddr_t)start + strlen(start) + 1;
167 }
168
169 return (start);
170 }
171
172 /*
173 * Set the uid/gid in the fuid_info structure.
174 */
175 static void
zfs_replay_fuid_ugid(zfs_fuid_info_t * fuid_infop,uint64_t uid,uint64_t gid)176 zfs_replay_fuid_ugid(zfs_fuid_info_t *fuid_infop, uint64_t uid, uint64_t gid)
177 {
178 /*
179 * If owner or group are log specific FUIDs then slurp up
180 * domain information and build zfs_fuid_info_t
181 */
182 if (IS_EPHEMERAL(uid))
183 fuid_infop->z_fuid_owner = uid;
184
185 if (IS_EPHEMERAL(gid))
186 fuid_infop->z_fuid_group = gid;
187 }
188
189 /*
190 * Load fuid domains into fuid_info_t
191 */
192 static zfs_fuid_info_t *
zfs_replay_fuid_domain(void * buf,void ** end,uint64_t uid,uint64_t gid)193 zfs_replay_fuid_domain(void *buf, void **end, uint64_t uid, uint64_t gid)
194 {
195 int domcnt;
196
197 zfs_fuid_info_t *fuid_infop;
198
199 fuid_infop = zfs_fuid_info_alloc();
200
201 domcnt = zfs_replay_domain_cnt(uid, gid);
202
203 if (domcnt == 0)
204 return (fuid_infop);
205
206 fuid_infop->z_domain_table =
207 kmem_zalloc(domcnt * sizeof (char **), KM_SLEEP);
208
209 zfs_replay_fuid_ugid(fuid_infop, uid, gid);
210
211 fuid_infop->z_domain_cnt = domcnt;
212 *end = zfs_replay_fuid_domain_common(fuid_infop, buf, domcnt);
213 return (fuid_infop);
214 }
215
216 /*
217 * load zfs_fuid_t's and fuid_domains into fuid_info_t
218 */
219 static zfs_fuid_info_t *
zfs_replay_fuids(void * start,void ** end,int idcnt,int domcnt,uint64_t uid,uint64_t gid)220 zfs_replay_fuids(void *start, void **end, int idcnt, int domcnt, uint64_t uid,
221 uint64_t gid)
222 {
223 uint64_t *log_fuid = (uint64_t *)start;
224 zfs_fuid_info_t *fuid_infop;
225 int i;
226
227 fuid_infop = zfs_fuid_info_alloc();
228 fuid_infop->z_domain_cnt = domcnt;
229
230 fuid_infop->z_domain_table =
231 kmem_zalloc(domcnt * sizeof (char **), KM_SLEEP);
232
233 for (i = 0; i != idcnt; i++) {
234 zfs_fuid_t *zfuid;
235
236 zfuid = kmem_alloc(sizeof (zfs_fuid_t), KM_SLEEP);
237 zfuid->z_logfuid = *log_fuid;
238 zfuid->z_id = -1;
239 zfuid->z_domidx = 0;
240 list_insert_tail(&fuid_infop->z_fuids, zfuid);
241 log_fuid++;
242 }
243
244 zfs_replay_fuid_ugid(fuid_infop, uid, gid);
245
246 *end = zfs_replay_fuid_domain_common(fuid_infop, log_fuid, domcnt);
247 return (fuid_infop);
248 }
249
250 static void
zfs_replay_swap_attrs(lr_attr_t * lrattr)251 zfs_replay_swap_attrs(lr_attr_t *lrattr)
252 {
253 /* swap the lr_attr structure */
254 byteswap_uint32_array(lrattr, sizeof (*lrattr));
255 /* swap the bitmap */
256 byteswap_uint32_array(lrattr + 1, (lrattr->lr_attr_masksize - 1) *
257 sizeof (uint32_t));
258 /* swap the attributes, create time + 64 bit word for attributes */
259 byteswap_uint64_array((caddr_t)(lrattr + 1) + (sizeof (uint32_t) *
260 (lrattr->lr_attr_masksize - 1)), 3 * sizeof (uint64_t));
261 }
262
263 /*
264 * Replay file create with optional ACL, xvattr information as well
265 * as option FUID information.
266 */
267 static int
zfs_replay_create_acl(zfsvfs_t * zfsvfs,lr_acl_create_t * lracl,boolean_t byteswap)268 zfs_replay_create_acl(zfsvfs_t *zfsvfs,
269 lr_acl_create_t *lracl, boolean_t byteswap)
270 {
271 char *name = NULL; /* location determined later */
272 lr_create_t *lr = (lr_create_t *)lracl;
273 znode_t *dzp;
274 vnode_t *vp = NULL;
275 xvattr_t xva;
276 int vflg = 0;
277 vsecattr_t vsec = { 0 };
278 lr_attr_t *lrattr;
279 void *aclstart;
280 void *fuidstart;
281 size_t xvatlen = 0;
282 uint64_t txtype;
283 int error;
284
285 txtype = (lr->lr_common.lrc_txtype & ~TX_CI);
286 if (byteswap) {
287 byteswap_uint64_array(lracl, sizeof (*lracl));
288 if (txtype == TX_CREATE_ACL_ATTR ||
289 txtype == TX_MKDIR_ACL_ATTR) {
290 lrattr = (lr_attr_t *)(caddr_t)(lracl + 1);
291 zfs_replay_swap_attrs(lrattr);
292 xvatlen = ZIL_XVAT_SIZE(lrattr->lr_attr_masksize);
293 }
294
295 aclstart = (caddr_t)(lracl + 1) + xvatlen;
296 zfs_ace_byteswap(aclstart, lracl->lr_acl_bytes, B_FALSE);
297 /* swap fuids */
298 if (lracl->lr_fuidcnt) {
299 byteswap_uint64_array((caddr_t)aclstart +
300 ZIL_ACE_LENGTH(lracl->lr_acl_bytes),
301 lracl->lr_fuidcnt * sizeof (uint64_t));
302 }
303 }
304
305 if ((error = zfs_zget(zfsvfs, lr->lr_doid, &dzp)) != 0)
306 return (error);
307
308 xva_init(&xva);
309 zfs_init_vattr(&xva.xva_vattr, AT_TYPE | AT_MODE | AT_UID | AT_GID,
310 lr->lr_mode, lr->lr_uid, lr->lr_gid, lr->lr_rdev, lr->lr_foid);
311
312 /*
313 * All forms of zfs create (create, mkdir, mkxattrdir, symlink)
314 * eventually end up in zfs_mknode(), which assigns the object's
315 * creation time and generation number. The generic VOP_CREATE()
316 * doesn't have either concept, so we smuggle the values inside
317 * the vattr's otherwise unused va_ctime and va_nblocks fields.
318 */
319 ZFS_TIME_DECODE(&xva.xva_vattr.va_ctime, lr->lr_crtime);
320 xva.xva_vattr.va_nblocks = lr->lr_gen;
321
322 error = dmu_object_info(zfsvfs->z_os, lr->lr_foid, NULL);
323 if (error != ENOENT)
324 goto bail;
325
326 if (lr->lr_common.lrc_txtype & TX_CI)
327 vflg |= FIGNORECASE;
328 switch (txtype) {
329 case TX_CREATE_ACL:
330 aclstart = (caddr_t)(lracl + 1);
331 fuidstart = (caddr_t)aclstart +
332 ZIL_ACE_LENGTH(lracl->lr_acl_bytes);
333 zfsvfs->z_fuid_replay = zfs_replay_fuids(fuidstart,
334 (void *)&name, lracl->lr_fuidcnt, lracl->lr_domcnt,
335 lr->lr_uid, lr->lr_gid);
336 /*FALLTHROUGH*/
337 case TX_CREATE_ACL_ATTR:
338 if (name == NULL) {
339 lrattr = (lr_attr_t *)(caddr_t)(lracl + 1);
340 xvatlen = ZIL_XVAT_SIZE(lrattr->lr_attr_masksize);
341 xva.xva_vattr.va_mask |= AT_XVATTR;
342 zfs_replay_xvattr(lrattr, &xva);
343 }
344 vsec.vsa_mask = VSA_ACE | VSA_ACE_ACLFLAGS;
345 vsec.vsa_aclentp = (caddr_t)(lracl + 1) + xvatlen;
346 vsec.vsa_aclcnt = lracl->lr_aclcnt;
347 vsec.vsa_aclentsz = lracl->lr_acl_bytes;
348 vsec.vsa_aclflags = lracl->lr_acl_flags;
349 if (zfsvfs->z_fuid_replay == NULL) {
350 fuidstart = (caddr_t)(lracl + 1) + xvatlen +
351 ZIL_ACE_LENGTH(lracl->lr_acl_bytes);
352 zfsvfs->z_fuid_replay =
353 zfs_replay_fuids(fuidstart,
354 (void *)&name, lracl->lr_fuidcnt, lracl->lr_domcnt,
355 lr->lr_uid, lr->lr_gid);
356 }
357
358 #ifdef TODO
359 error = VOP_CREATE(ZTOV(dzp), name, &xva.xva_vattr,
360 0, 0, &vp, kcred, vflg, NULL, &vsec);
361 #else
362 panic("%s:%u: unsupported condition", __func__, __LINE__);
363 #endif
364 break;
365 case TX_MKDIR_ACL:
366 aclstart = (caddr_t)(lracl + 1);
367 fuidstart = (caddr_t)aclstart +
368 ZIL_ACE_LENGTH(lracl->lr_acl_bytes);
369 zfsvfs->z_fuid_replay = zfs_replay_fuids(fuidstart,
370 (void *)&name, lracl->lr_fuidcnt, lracl->lr_domcnt,
371 lr->lr_uid, lr->lr_gid);
372 /*FALLTHROUGH*/
373 case TX_MKDIR_ACL_ATTR:
374 if (name == NULL) {
375 lrattr = (lr_attr_t *)(caddr_t)(lracl + 1);
376 xvatlen = ZIL_XVAT_SIZE(lrattr->lr_attr_masksize);
377 zfs_replay_xvattr(lrattr, &xva);
378 }
379 vsec.vsa_mask = VSA_ACE | VSA_ACE_ACLFLAGS;
380 vsec.vsa_aclentp = (caddr_t)(lracl + 1) + xvatlen;
381 vsec.vsa_aclcnt = lracl->lr_aclcnt;
382 vsec.vsa_aclentsz = lracl->lr_acl_bytes;
383 vsec.vsa_aclflags = lracl->lr_acl_flags;
384 if (zfsvfs->z_fuid_replay == NULL) {
385 fuidstart = (caddr_t)(lracl + 1) + xvatlen +
386 ZIL_ACE_LENGTH(lracl->lr_acl_bytes);
387 zfsvfs->z_fuid_replay =
388 zfs_replay_fuids(fuidstart,
389 (void *)&name, lracl->lr_fuidcnt, lracl->lr_domcnt,
390 lr->lr_uid, lr->lr_gid);
391 }
392 #ifdef TODO
393 error = VOP_MKDIR(ZTOV(dzp), name, &xva.xva_vattr,
394 &vp, kcred, NULL, vflg, &vsec);
395 #else
396 panic("%s:%u: unsupported condition", __func__, __LINE__);
397 #endif
398 break;
399 default:
400 error = SET_ERROR(ENOTSUP);
401 }
402
403 bail:
404 if (error == 0 && vp != NULL)
405 VN_RELE(vp);
406
407 VN_RELE(ZTOV(dzp));
408
409 if (zfsvfs->z_fuid_replay)
410 zfs_fuid_info_free(zfsvfs->z_fuid_replay);
411 zfsvfs->z_fuid_replay = NULL;
412
413 return (error);
414 }
415
416 static int
zfs_replay_create(zfsvfs_t * zfsvfs,lr_create_t * lr,boolean_t byteswap)417 zfs_replay_create(zfsvfs_t *zfsvfs, lr_create_t *lr, boolean_t byteswap)
418 {
419 char *name = NULL; /* location determined later */
420 char *link; /* symlink content follows name */
421 znode_t *dzp;
422 vnode_t *vp = NULL;
423 xvattr_t xva;
424 int vflg = 0;
425 size_t lrsize = sizeof (lr_create_t);
426 lr_attr_t *lrattr;
427 void *start;
428 size_t xvatlen;
429 uint64_t txtype;
430 struct componentname cn;
431 int error;
432
433 txtype = (lr->lr_common.lrc_txtype & ~TX_CI);
434 if (byteswap) {
435 byteswap_uint64_array(lr, sizeof (*lr));
436 if (txtype == TX_CREATE_ATTR || txtype == TX_MKDIR_ATTR)
437 zfs_replay_swap_attrs((lr_attr_t *)(lr + 1));
438 }
439
440
441 if ((error = zfs_zget(zfsvfs, lr->lr_doid, &dzp)) != 0)
442 return (error);
443
444 xva_init(&xva);
445 zfs_init_vattr(&xva.xva_vattr, AT_TYPE | AT_MODE | AT_UID | AT_GID,
446 lr->lr_mode, lr->lr_uid, lr->lr_gid, lr->lr_rdev, lr->lr_foid);
447
448 /*
449 * All forms of zfs create (create, mkdir, mkxattrdir, symlink)
450 * eventually end up in zfs_mknode(), which assigns the object's
451 * creation time and generation number. The generic VOP_CREATE()
452 * doesn't have either concept, so we smuggle the values inside
453 * the vattr's otherwise unused va_ctime and va_nblocks fields.
454 */
455 ZFS_TIME_DECODE(&xva.xva_vattr.va_ctime, lr->lr_crtime);
456 xva.xva_vattr.va_nblocks = lr->lr_gen;
457
458 error = dmu_object_info(zfsvfs->z_os, lr->lr_foid, NULL);
459 if (error != ENOENT)
460 goto out;
461
462 if (lr->lr_common.lrc_txtype & TX_CI)
463 vflg |= FIGNORECASE;
464
465 /*
466 * Symlinks don't have fuid info, and CIFS never creates
467 * symlinks.
468 *
469 * The _ATTR versions will grab the fuid info in their subcases.
470 */
471 if ((int)lr->lr_common.lrc_txtype != TX_SYMLINK &&
472 (int)lr->lr_common.lrc_txtype != TX_MKDIR_ATTR &&
473 (int)lr->lr_common.lrc_txtype != TX_CREATE_ATTR) {
474 start = (lr + 1);
475 zfsvfs->z_fuid_replay =
476 zfs_replay_fuid_domain(start, &start,
477 lr->lr_uid, lr->lr_gid);
478 }
479
480 cn.cn_cred = kcred;
481 cn.cn_thread = curthread;
482 cn.cn_flags = SAVENAME;
483
484 vn_lock(ZTOV(dzp), LK_EXCLUSIVE | LK_RETRY);
485 switch (txtype) {
486 case TX_CREATE_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_CREATE:
498 if (name == NULL)
499 name = (char *)start;
500
501 cn.cn_nameptr = name;
502 error = VOP_CREATE(ZTOV(dzp), &vp, &cn, &xva.xva_vattr /*,vflg*/);
503 break;
504 case TX_MKDIR_ATTR:
505 lrattr = (lr_attr_t *)(caddr_t)(lr + 1);
506 xvatlen = ZIL_XVAT_SIZE(lrattr->lr_attr_masksize);
507 zfs_replay_xvattr((lr_attr_t *)((caddr_t)lr + lrsize), &xva);
508 start = (caddr_t)(lr + 1) + xvatlen;
509 zfsvfs->z_fuid_replay =
510 zfs_replay_fuid_domain(start, &start,
511 lr->lr_uid, lr->lr_gid);
512 name = (char *)start;
513
514 /*FALLTHROUGH*/
515 case TX_MKDIR:
516 if (name == NULL)
517 name = (char *)(lr + 1);
518
519 cn.cn_nameptr = name;
520 error = VOP_MKDIR(ZTOV(dzp), &vp, &cn, &xva.xva_vattr /*,vflg*/);
521 break;
522 case TX_MKXATTR:
523 error = zfs_make_xattrdir(dzp, &xva.xva_vattr, &vp, kcred);
524 break;
525 case TX_SYMLINK:
526 name = (char *)(lr + 1);
527 link = name + strlen(name) + 1;
528 cn.cn_nameptr = name;
529 error = VOP_SYMLINK(ZTOV(dzp), &vp, &cn, &xva.xva_vattr, link /*,vflg*/);
530 break;
531 default:
532 error = SET_ERROR(ENOTSUP);
533 }
534 VOP_UNLOCK(ZTOV(dzp), 0);
535
536 out:
537 if (error == 0 && vp != NULL)
538 VN_URELE(vp);
539
540 VN_RELE(ZTOV(dzp));
541
542 if (zfsvfs->z_fuid_replay)
543 zfs_fuid_info_free(zfsvfs->z_fuid_replay);
544 zfsvfs->z_fuid_replay = NULL;
545 return (error);
546 }
547
548 static int
zfs_replay_remove(zfsvfs_t * zfsvfs,lr_remove_t * lr,boolean_t byteswap)549 zfs_replay_remove(zfsvfs_t *zfsvfs, lr_remove_t *lr, boolean_t byteswap)
550 {
551 char *name = (char *)(lr + 1); /* name follows lr_remove_t */
552 znode_t *dzp;
553 struct componentname cn;
554 vnode_t *vp;
555 int error;
556 int vflg = 0;
557
558 if (byteswap)
559 byteswap_uint64_array(lr, sizeof (*lr));
560
561 if ((error = zfs_zget(zfsvfs, lr->lr_doid, &dzp)) != 0)
562 return (error);
563
564 if (lr->lr_common.lrc_txtype & TX_CI)
565 vflg |= FIGNORECASE;
566 cn.cn_nameptr = name;
567 cn.cn_namelen = strlen(name);
568 cn.cn_nameiop = DELETE;
569 cn.cn_flags = ISLASTCN | SAVENAME;
570 cn.cn_lkflags = LK_EXCLUSIVE | LK_RETRY;
571 cn.cn_cred = kcred;
572 cn.cn_thread = curthread;
573 vn_lock(ZTOV(dzp), LK_EXCLUSIVE | LK_RETRY);
574 error = VOP_LOOKUP(ZTOV(dzp), &vp, &cn);
575 if (error != 0) {
576 VOP_UNLOCK(ZTOV(dzp), 0);
577 goto fail;
578 }
579
580 switch ((int)lr->lr_common.lrc_txtype) {
581 case TX_REMOVE:
582 error = VOP_REMOVE(ZTOV(dzp), vp, &cn /*,vflg*/);
583 break;
584 case TX_RMDIR:
585 error = VOP_RMDIR(ZTOV(dzp), vp, &cn /*,vflg*/);
586 break;
587 default:
588 error = SET_ERROR(ENOTSUP);
589 }
590 vput(vp);
591 VOP_UNLOCK(ZTOV(dzp), 0);
592
593 fail:
594 VN_RELE(ZTOV(dzp));
595
596 return (error);
597 }
598
599 static int
zfs_replay_link(zfsvfs_t * zfsvfs,lr_link_t * lr,boolean_t byteswap)600 zfs_replay_link(zfsvfs_t *zfsvfs, lr_link_t *lr, boolean_t byteswap)
601 {
602 char *name = (char *)(lr + 1); /* name follows lr_link_t */
603 znode_t *dzp, *zp;
604 struct componentname cn;
605 int error;
606 int vflg = 0;
607
608 if (byteswap)
609 byteswap_uint64_array(lr, sizeof (*lr));
610
611 if ((error = zfs_zget(zfsvfs, lr->lr_doid, &dzp)) != 0)
612 return (error);
613
614 if ((error = zfs_zget(zfsvfs, lr->lr_link_obj, &zp)) != 0) {
615 VN_RELE(ZTOV(dzp));
616 return (error);
617 }
618
619 if (lr->lr_common.lrc_txtype & TX_CI)
620 vflg |= FIGNORECASE;
621
622 cn.cn_nameptr = name;
623 cn.cn_cred = kcred;
624 cn.cn_thread = curthread;
625 cn.cn_flags = SAVENAME;
626
627 vn_lock(ZTOV(dzp), LK_EXCLUSIVE | LK_RETRY);
628 vn_lock(ZTOV(zp), LK_EXCLUSIVE | LK_RETRY);
629 error = VOP_LINK(ZTOV(dzp), ZTOV(zp), &cn /*,vflg*/);
630 VOP_UNLOCK(ZTOV(zp), 0);
631 VOP_UNLOCK(ZTOV(dzp), 0);
632
633 VN_RELE(ZTOV(zp));
634 VN_RELE(ZTOV(dzp));
635
636 return (error);
637 }
638
639 static int
zfs_replay_rename(zfsvfs_t * zfsvfs,lr_rename_t * lr,boolean_t byteswap)640 zfs_replay_rename(zfsvfs_t *zfsvfs, lr_rename_t *lr, boolean_t byteswap)
641 {
642 char *sname = (char *)(lr + 1); /* sname and tname follow lr_rename_t */
643 char *tname = sname + strlen(sname) + 1;
644 znode_t *sdzp, *tdzp;
645 struct componentname scn, tcn;
646 vnode_t *svp, *tvp;
647 kthread_t *td = curthread;
648 int error;
649 int vflg = 0;
650
651 if (byteswap)
652 byteswap_uint64_array(lr, sizeof (*lr));
653
654 if ((error = zfs_zget(zfsvfs, lr->lr_sdoid, &sdzp)) != 0)
655 return (error);
656
657 if ((error = zfs_zget(zfsvfs, lr->lr_tdoid, &tdzp)) != 0) {
658 VN_RELE(ZTOV(sdzp));
659 return (error);
660 }
661
662 if (lr->lr_common.lrc_txtype & TX_CI)
663 vflg |= FIGNORECASE;
664 svp = tvp = NULL;
665
666 scn.cn_nameptr = sname;
667 scn.cn_namelen = strlen(sname);
668 scn.cn_nameiop = DELETE;
669 scn.cn_flags = ISLASTCN | SAVENAME;
670 scn.cn_lkflags = LK_EXCLUSIVE | LK_RETRY;
671 scn.cn_cred = kcred;
672 scn.cn_thread = td;
673 vn_lock(ZTOV(sdzp), LK_EXCLUSIVE | LK_RETRY);
674 error = VOP_LOOKUP(ZTOV(sdzp), &svp, &scn);
675 VOP_UNLOCK(ZTOV(sdzp), 0);
676 if (error != 0)
677 goto fail;
678 VOP_UNLOCK(svp, 0);
679
680 tcn.cn_nameptr = tname;
681 tcn.cn_namelen = strlen(tname);
682 tcn.cn_nameiop = RENAME;
683 tcn.cn_flags = ISLASTCN | SAVENAME;
684 tcn.cn_lkflags = LK_EXCLUSIVE | LK_RETRY;
685 tcn.cn_cred = kcred;
686 tcn.cn_thread = td;
687 vn_lock(ZTOV(tdzp), LK_EXCLUSIVE | LK_RETRY);
688 error = VOP_LOOKUP(ZTOV(tdzp), &tvp, &tcn);
689 if (error == EJUSTRETURN)
690 tvp = NULL;
691 else if (error != 0) {
692 VOP_UNLOCK(ZTOV(tdzp), 0);
693 goto fail;
694 }
695
696 error = VOP_RENAME(ZTOV(sdzp), svp, &scn, ZTOV(tdzp), tvp, &tcn /*,vflg*/);
697 return (error);
698 fail:
699 if (svp != NULL)
700 vrele(svp);
701 if (tvp != NULL)
702 vrele(tvp);
703 VN_RELE(ZTOV(tdzp));
704 VN_RELE(ZTOV(sdzp));
705
706 return (error);
707 }
708
709 static int
zfs_replay_write(zfsvfs_t * zfsvfs,lr_write_t * lr,boolean_t byteswap)710 zfs_replay_write(zfsvfs_t *zfsvfs, lr_write_t *lr, boolean_t byteswap)
711 {
712 char *data = (char *)(lr + 1); /* data follows lr_write_t */
713 znode_t *zp;
714 int error;
715 ssize_t resid;
716 uint64_t eod, offset, length;
717
718 if (byteswap)
719 byteswap_uint64_array(lr, sizeof (*lr));
720
721 if ((error = zfs_zget(zfsvfs, lr->lr_foid, &zp)) != 0) {
722 /*
723 * As we can log writes out of order, it's possible the
724 * file has been removed. In this case just drop the write
725 * and return success.
726 */
727 if (error == ENOENT)
728 error = 0;
729 return (error);
730 }
731
732 offset = lr->lr_offset;
733 length = lr->lr_length;
734 eod = offset + length; /* end of data for this write */
735
736 /*
737 * This may be a write from a dmu_sync() for a whole block,
738 * and may extend beyond the current end of the file.
739 * We can't just replay what was written for this TX_WRITE as
740 * a future TX_WRITE2 may extend the eof and the data for that
741 * write needs to be there. So we write the whole block and
742 * reduce the eof. This needs to be done within the single dmu
743 * transaction created within vn_rdwr -> zfs_write. So a possible
744 * new end of file is passed through in zfsvfs->z_replay_eof
745 */
746
747 zfsvfs->z_replay_eof = 0; /* 0 means don't change end of file */
748
749 /* If it's a dmu_sync() block, write the whole block */
750 if (lr->lr_common.lrc_reclen == sizeof (lr_write_t)) {
751 uint64_t blocksize = BP_GET_LSIZE(&lr->lr_blkptr);
752 if (length < blocksize) {
753 offset -= offset % blocksize;
754 length = blocksize;
755 }
756 if (zp->z_size < eod)
757 zfsvfs->z_replay_eof = eod;
758 }
759
760 error = vn_rdwr(UIO_WRITE, ZTOV(zp), data, length, offset,
761 UIO_SYSSPACE, 0, RLIM64_INFINITY, kcred, &resid);
762
763 VN_RELE(ZTOV(zp));
764 zfsvfs->z_replay_eof = 0; /* safety */
765
766 return (error);
767 }
768
769 /*
770 * TX_WRITE2 are only generated when dmu_sync() returns EALREADY
771 * meaning the pool block is already being synced. So now that we always write
772 * out full blocks, all we have to do is expand the eof if
773 * the file is grown.
774 */
775 static int
zfs_replay_write2(zfsvfs_t * zfsvfs,lr_write_t * lr,boolean_t byteswap)776 zfs_replay_write2(zfsvfs_t *zfsvfs, lr_write_t *lr, boolean_t byteswap)
777 {
778 znode_t *zp;
779 int error;
780 uint64_t end;
781
782 if (byteswap)
783 byteswap_uint64_array(lr, sizeof (*lr));
784
785 if ((error = zfs_zget(zfsvfs, lr->lr_foid, &zp)) != 0)
786 return (error);
787
788 top:
789 end = lr->lr_offset + lr->lr_length;
790 if (end > zp->z_size) {
791 dmu_tx_t *tx = dmu_tx_create(zfsvfs->z_os);
792
793 zp->z_size = end;
794 dmu_tx_hold_sa(tx, zp->z_sa_hdl, B_FALSE);
795 error = dmu_tx_assign(tx, TXG_WAIT);
796 if (error) {
797 VN_RELE(ZTOV(zp));
798 if (error == ERESTART) {
799 dmu_tx_wait(tx);
800 dmu_tx_abort(tx);
801 goto top;
802 }
803 dmu_tx_abort(tx);
804 return (error);
805 }
806 (void) sa_update(zp->z_sa_hdl, SA_ZPL_SIZE(zfsvfs),
807 (void *)&zp->z_size, sizeof (uint64_t), tx);
808
809 /* Ensure the replayed seq is updated */
810 (void) zil_replaying(zfsvfs->z_log, tx);
811
812 dmu_tx_commit(tx);
813 }
814
815 VN_RELE(ZTOV(zp));
816
817 return (error);
818 }
819
820 static int
zfs_replay_truncate(zfsvfs_t * zfsvfs,lr_truncate_t * lr,boolean_t byteswap)821 zfs_replay_truncate(zfsvfs_t *zfsvfs, lr_truncate_t *lr, boolean_t byteswap)
822 {
823 #ifdef illumos
824 znode_t *zp;
825 flock64_t fl;
826 int error;
827
828 if (byteswap)
829 byteswap_uint64_array(lr, sizeof (*lr));
830
831 if ((error = zfs_zget(zfsvfs, lr->lr_foid, &zp)) != 0)
832 return (error);
833
834 bzero(&fl, sizeof (fl));
835 fl.l_type = F_WRLCK;
836 fl.l_whence = 0;
837 fl.l_start = lr->lr_offset;
838 fl.l_len = lr->lr_length;
839
840 error = VOP_SPACE(ZTOV(zp), F_FREESP, &fl, FWRITE | FOFFMAX,
841 lr->lr_offset, kcred, NULL);
842
843 VN_RELE(ZTOV(zp));
844
845 return (error);
846 #else
847 ZFS_LOG(0, "Unexpected code path, report to pjd@FreeBSD.org");
848 return (EOPNOTSUPP);
849 #endif
850 }
851
852 static int
zfs_replay_setattr(zfsvfs_t * zfsvfs,lr_setattr_t * lr,boolean_t byteswap)853 zfs_replay_setattr(zfsvfs_t *zfsvfs, lr_setattr_t *lr, boolean_t byteswap)
854 {
855 znode_t *zp;
856 xvattr_t xva;
857 vattr_t *vap = &xva.xva_vattr;
858 vnode_t *vp;
859 int error;
860 void *start;
861
862 xva_init(&xva);
863 if (byteswap) {
864 byteswap_uint64_array(lr, sizeof (*lr));
865
866 if ((lr->lr_mask & AT_XVATTR) &&
867 zfsvfs->z_version >= ZPL_VERSION_INITIAL)
868 zfs_replay_swap_attrs((lr_attr_t *)(lr + 1));
869 }
870
871 if ((error = zfs_zget(zfsvfs, lr->lr_foid, &zp)) != 0)
872 return (error);
873
874 zfs_init_vattr(vap, lr->lr_mask, lr->lr_mode,
875 lr->lr_uid, lr->lr_gid, 0, lr->lr_foid);
876
877 vap->va_size = lr->lr_size;
878 ZFS_TIME_DECODE(&vap->va_atime, lr->lr_atime);
879 ZFS_TIME_DECODE(&vap->va_mtime, lr->lr_mtime);
880
881 /*
882 * Fill in xvattr_t portions if necessary.
883 */
884
885 start = (lr_setattr_t *)(lr + 1);
886 if (vap->va_mask & AT_XVATTR) {
887 zfs_replay_xvattr((lr_attr_t *)start, &xva);
888 start = (caddr_t)start +
889 ZIL_XVAT_SIZE(((lr_attr_t *)start)->lr_attr_masksize);
890 } else
891 xva.xva_vattr.va_mask &= ~AT_XVATTR;
892
893 zfsvfs->z_fuid_replay = zfs_replay_fuid_domain(start, &start,
894 lr->lr_uid, lr->lr_gid);
895
896 vp = ZTOV(zp);
897 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
898 error = VOP_SETATTR(vp, vap, kcred);
899 VOP_UNLOCK(vp, 0);
900
901 zfs_fuid_info_free(zfsvfs->z_fuid_replay);
902 zfsvfs->z_fuid_replay = NULL;
903 VN_RELE(vp);
904
905 return (error);
906 }
907
908 extern int zfs_setsecattr(vnode_t *vp, vsecattr_t *vsecp, int flag, cred_t *cr,
909 caller_context_t *ct);
910
911 static int
zfs_replay_acl_v0(zfsvfs_t * zfsvfs,lr_acl_v0_t * lr,boolean_t byteswap)912 zfs_replay_acl_v0(zfsvfs_t *zfsvfs, lr_acl_v0_t *lr, boolean_t byteswap)
913 {
914 ace_t *ace = (ace_t *)(lr + 1); /* ace array follows lr_acl_t */
915 vsecattr_t vsa;
916 vnode_t *vp;
917 znode_t *zp;
918 int error;
919
920 if (byteswap) {
921 byteswap_uint64_array(lr, sizeof (*lr));
922 zfs_oldace_byteswap(ace, lr->lr_aclcnt);
923 }
924
925 if ((error = zfs_zget(zfsvfs, lr->lr_foid, &zp)) != 0)
926 return (error);
927
928 bzero(&vsa, sizeof (vsa));
929 vsa.vsa_mask = VSA_ACE | VSA_ACECNT;
930 vsa.vsa_aclcnt = lr->lr_aclcnt;
931 vsa.vsa_aclentsz = sizeof (ace_t) * vsa.vsa_aclcnt;
932 vsa.vsa_aclflags = 0;
933 vsa.vsa_aclentp = ace;
934
935 vp = ZTOV(zp);
936 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
937 error = zfs_setsecattr(vp, &vsa, 0, kcred, NULL);
938 VOP_UNLOCK(vp, 0);
939
940 VN_RELE(vp);
941
942 return (error);
943 }
944
945 /*
946 * Replaying ACLs is complicated by FUID support.
947 * The log record may contain some optional data
948 * to be used for replaying FUID's. These pieces
949 * are the actual FUIDs that were created initially.
950 * The FUID table index may no longer be valid and
951 * during zfs_create() a new index may be assigned.
952 * Because of this the log will contain the original
953 * doman+rid in order to create a new FUID.
954 *
955 * The individual ACEs may contain an ephemeral uid/gid which is no
956 * longer valid and will need to be replaced with an actual FUID.
957 *
958 */
959 static int
zfs_replay_acl(zfsvfs_t * zfsvfs,lr_acl_t * lr,boolean_t byteswap)960 zfs_replay_acl(zfsvfs_t *zfsvfs, lr_acl_t *lr, boolean_t byteswap)
961 {
962 ace_t *ace = (ace_t *)(lr + 1);
963 vsecattr_t vsa;
964 znode_t *zp;
965 vnode_t *vp;
966 int error;
967
968 if (byteswap) {
969 byteswap_uint64_array(lr, sizeof (*lr));
970 zfs_ace_byteswap(ace, lr->lr_acl_bytes, B_FALSE);
971 if (lr->lr_fuidcnt) {
972 byteswap_uint64_array((caddr_t)ace +
973 ZIL_ACE_LENGTH(lr->lr_acl_bytes),
974 lr->lr_fuidcnt * sizeof (uint64_t));
975 }
976 }
977
978 if ((error = zfs_zget(zfsvfs, lr->lr_foid, &zp)) != 0)
979 return (error);
980
981 bzero(&vsa, sizeof (vsa));
982 vsa.vsa_mask = VSA_ACE | VSA_ACECNT | VSA_ACE_ACLFLAGS;
983 vsa.vsa_aclcnt = lr->lr_aclcnt;
984 vsa.vsa_aclentp = ace;
985 vsa.vsa_aclentsz = lr->lr_acl_bytes;
986 vsa.vsa_aclflags = lr->lr_acl_flags;
987
988 if (lr->lr_fuidcnt) {
989 void *fuidstart = (caddr_t)ace +
990 ZIL_ACE_LENGTH(lr->lr_acl_bytes);
991
992 zfsvfs->z_fuid_replay =
993 zfs_replay_fuids(fuidstart, &fuidstart,
994 lr->lr_fuidcnt, lr->lr_domcnt, 0, 0);
995 }
996
997 vp = ZTOV(zp);
998 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
999 error = zfs_setsecattr(vp, &vsa, 0, kcred, NULL);
1000 VOP_UNLOCK(vp, 0);
1001
1002 if (zfsvfs->z_fuid_replay)
1003 zfs_fuid_info_free(zfsvfs->z_fuid_replay);
1004
1005 zfsvfs->z_fuid_replay = NULL;
1006 VN_RELE(vp);
1007
1008 return (error);
1009 }
1010
1011 /*
1012 * Callback vectors for replaying records
1013 */
1014 zil_replay_func_t *zfs_replay_vector[TX_MAX_TYPE] = {
1015 zfs_replay_error, /* 0 no such transaction type */
1016 zfs_replay_create, /* TX_CREATE */
1017 zfs_replay_create, /* TX_MKDIR */
1018 zfs_replay_create, /* TX_MKXATTR */
1019 zfs_replay_create, /* TX_SYMLINK */
1020 zfs_replay_remove, /* TX_REMOVE */
1021 zfs_replay_remove, /* TX_RMDIR */
1022 zfs_replay_link, /* TX_LINK */
1023 zfs_replay_rename, /* TX_RENAME */
1024 zfs_replay_write, /* TX_WRITE */
1025 zfs_replay_truncate, /* TX_TRUNCATE */
1026 zfs_replay_setattr, /* TX_SETATTR */
1027 zfs_replay_acl_v0, /* TX_ACL_V0 */
1028 zfs_replay_acl, /* TX_ACL */
1029 zfs_replay_create_acl, /* TX_CREATE_ACL */
1030 zfs_replay_create, /* TX_CREATE_ATTR */
1031 zfs_replay_create_acl, /* TX_CREATE_ACL_ATTR */
1032 zfs_replay_create_acl, /* TX_MKDIR_ACL */
1033 zfs_replay_create, /* TX_MKDIR_ATTR */
1034 zfs_replay_create_acl, /* TX_MKDIR_ACL_ATTR */
1035 zfs_replay_write2, /* TX_WRITE2 */
1036 };
1037