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 2009 Sun Microsystems, Inc. All rights reserved.
23 * Use is subject to license terms.
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 (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 }
136
137 static int
zfs_replay_domain_cnt(uint64_t uid,uint64_t gid)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 *
zfs_replay_fuid_domain_common(zfs_fuid_info_t * fuid_infop,void * start,int domcnt)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
zfs_replay_fuid_ugid(zfs_fuid_info_t * fuid_infop,uint64_t uid,uint64_t gid)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 *
zfs_replay_fuid_domain(void * buf,void ** end,uint64_t uid,uint64_t gid)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 *
zfs_replay_fuids(void * start,void ** end,int idcnt,int domcnt,uint64_t uid,uint64_t gid)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
zfs_replay_swap_attrs(lr_attr_t * lrattr)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
zfs_replay_create_acl(zfsvfs_t * zfsvfs,lr_acl_create_t * lracl,boolean_t byteswap)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 vn_lock(ZTOV(dzp), LK_EXCLUSIVE | LK_RETRY);
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 #ifdef TODO
356 error = VOP_CREATE(ZTOV(dzp), name, &xva.xva_vattr,
357 0, 0, &vp, kcred, vflg, NULL, &vsec);
358 #else
359 panic("%s:%u: unsupported condition", __func__, __LINE__);
360 #endif
361 break;
362 case TX_MKDIR_ACL:
363 aclstart = (caddr_t)(lracl + 1);
364 fuidstart = (caddr_t)aclstart +
365 ZIL_ACE_LENGTH(lracl->lr_acl_bytes);
366 zfsvfs->z_fuid_replay = zfs_replay_fuids(fuidstart,
367 (void *)&name, lracl->lr_fuidcnt, lracl->lr_domcnt,
368 lr->lr_uid, lr->lr_gid);
369 /*FALLTHROUGH*/
370 case TX_MKDIR_ACL_ATTR:
371 if (name == NULL) {
372 lrattr = (lr_attr_t *)(caddr_t)(lracl + 1);
373 xvatlen = ZIL_XVAT_SIZE(lrattr->lr_attr_masksize);
374 zfs_replay_xvattr(lrattr, &xva);
375 }
376 vsec.vsa_mask = VSA_ACE | VSA_ACE_ACLFLAGS;
377 vsec.vsa_aclentp = (caddr_t)(lracl + 1) + xvatlen;
378 vsec.vsa_aclcnt = lracl->lr_aclcnt;
379 vsec.vsa_aclentsz = lracl->lr_acl_bytes;
380 vsec.vsa_aclflags = lracl->lr_acl_flags;
381 if (zfsvfs->z_fuid_replay == NULL) {
382 fuidstart = (caddr_t)(lracl + 1) + xvatlen +
383 ZIL_ACE_LENGTH(lracl->lr_acl_bytes);
384 zfsvfs->z_fuid_replay =
385 zfs_replay_fuids(fuidstart,
386 (void *)&name, lracl->lr_fuidcnt, lracl->lr_domcnt,
387 lr->lr_uid, lr->lr_gid);
388 }
389 #ifdef TODO
390 error = VOP_MKDIR(ZTOV(dzp), name, &xva.xva_vattr,
391 &vp, kcred, NULL, vflg, &vsec);
392 #else
393 panic("%s:%u: unsupported condition", __func__, __LINE__);
394 #endif
395 break;
396 default:
397 error = ENOTSUP;
398 }
399 VOP_UNLOCK(ZTOV(dzp));
400
401 bail:
402 if (error == 0 && vp != NULL)
403 VN_RELE(vp);
404
405 VN_RELE(ZTOV(dzp));
406
407 if (zfsvfs->z_fuid_replay)
408 zfs_fuid_info_free(zfsvfs->z_fuid_replay);
409 zfsvfs->z_fuid_replay = NULL;
410
411 return (error);
412 }
413
414 static int
zfs_replay_create(zfsvfs_t * zfsvfs,lr_create_t * lr,boolean_t byteswap)415 zfs_replay_create(zfsvfs_t *zfsvfs, lr_create_t *lr, boolean_t byteswap)
416 {
417 char *name = NULL; /* location determined later */
418 char *link; /* symlink content follows name */
419 znode_t *dzp;
420 vnode_t *vp = NULL;
421 xvattr_t xva;
422 int vflg = 0;
423 size_t lrsize = sizeof (lr_create_t);
424 lr_attr_t *lrattr;
425 void *start;
426 size_t xvatlen;
427 uint64_t txtype;
428 struct componentname cn;
429 int error;
430
431 txtype = (lr->lr_common.lrc_txtype & ~TX_CI);
432 if (byteswap) {
433 byteswap_uint64_array(lr, sizeof (*lr));
434 if (txtype == TX_CREATE_ATTR || txtype == TX_MKDIR_ATTR)
435 zfs_replay_swap_attrs((lr_attr_t *)(lr + 1));
436 }
437
438
439 if ((error = zfs_zget(zfsvfs, lr->lr_doid, &dzp)) != 0)
440 return (error);
441
442 xva_init(&xva);
443 zfs_init_vattr(&xva.xva_vattr, AT_TYPE | AT_MODE | AT_UID | AT_GID,
444 lr->lr_mode, lr->lr_uid, lr->lr_gid, lr->lr_rdev, lr->lr_foid);
445
446 /*
447 * All forms of zfs create (create, mkdir, mkxattrdir, symlink)
448 * eventually end up in zfs_mknode(), which assigns the object's
449 * creation time and generation number. The generic VOP_CREATE()
450 * doesn't have either concept, so we smuggle the values inside
451 * the vattr's otherwise unused va_ctime and va_nblocks fields.
452 */
453 ZFS_TIME_DECODE(&xva.xva_vattr.va_ctime, lr->lr_crtime);
454 xva.xva_vattr.va_nblocks = lr->lr_gen;
455
456 error = dmu_object_info(zfsvfs->z_os, lr->lr_foid, NULL);
457 if (error != ENOENT)
458 goto out;
459
460 if (lr->lr_common.lrc_txtype & TX_CI)
461 vflg |= FIGNORECASE;
462
463 /*
464 * Symlinks don't have fuid info, and CIFS never creates
465 * symlinks.
466 *
467 * The _ATTR versions will grab the fuid info in their subcases.
468 */
469 if ((int)lr->lr_common.lrc_txtype != TX_SYMLINK &&
470 (int)lr->lr_common.lrc_txtype != TX_MKDIR_ATTR &&
471 (int)lr->lr_common.lrc_txtype != TX_CREATE_ATTR) {
472 start = (lr + 1);
473 zfsvfs->z_fuid_replay =
474 zfs_replay_fuid_domain(start, &start,
475 lr->lr_uid, lr->lr_gid);
476 }
477
478 cn.cn_cred = kcred;
479 cn.cn_flags = 0;
480
481 vn_lock(ZTOV(dzp), LK_EXCLUSIVE | LK_RETRY);
482 switch (txtype) {
483 case TX_CREATE_ATTR:
484 lrattr = (lr_attr_t *)(caddr_t)(lr + 1);
485 xvatlen = ZIL_XVAT_SIZE(lrattr->lr_attr_masksize);
486 zfs_replay_xvattr((lr_attr_t *)((caddr_t)lr + lrsize), &xva);
487 start = (caddr_t)(lr + 1) + xvatlen;
488 zfsvfs->z_fuid_replay =
489 zfs_replay_fuid_domain(start, &start,
490 lr->lr_uid, lr->lr_gid);
491 name = (char *)start;
492
493 /*FALLTHROUGH*/
494 case TX_CREATE:
495 if (name == NULL)
496 name = (char *)start;
497
498 cn.cn_nameptr = name;
499 error = VOP_CREATE(ZTOV(dzp), &vp, &cn, &xva.xva_vattr /*,vflg*/);
500 break;
501 case TX_MKDIR_ATTR:
502 lrattr = (lr_attr_t *)(caddr_t)(lr + 1);
503 xvatlen = ZIL_XVAT_SIZE(lrattr->lr_attr_masksize);
504 zfs_replay_xvattr((lr_attr_t *)((caddr_t)lr + lrsize), &xva);
505 start = (caddr_t)(lr + 1) + xvatlen;
506 zfsvfs->z_fuid_replay =
507 zfs_replay_fuid_domain(start, &start,
508 lr->lr_uid, lr->lr_gid);
509 name = (char *)start;
510
511 /*FALLTHROUGH*/
512 case TX_MKDIR:
513 if (name == NULL)
514 name = (char *)(lr + 1);
515
516 cn.cn_nameptr = name;
517 error = VOP_MKDIR(ZTOV(dzp), &vp, &cn, &xva.xva_vattr /*,vflg*/);
518 break;
519 case TX_MKXATTR:
520 error = zfs_make_xattrdir(dzp, &xva.xva_vattr, &vp, kcred);
521 break;
522 case TX_SYMLINK:
523 name = (char *)(lr + 1);
524 link = name + strlen(name) + 1;
525 cn.cn_nameptr = name;
526 error = VOP_SYMLINK(ZTOV(dzp), &vp, &cn, &xva.xva_vattr, link /*,vflg*/);
527 break;
528 default:
529 error = ENOTSUP;
530 }
531 VOP_UNLOCK(ZTOV(dzp));
532
533 out:
534 if (error == 0 && vp != NULL)
535 VN_RELE(vp);
536
537 VN_RELE(ZTOV(dzp));
538
539 if (zfsvfs->z_fuid_replay)
540 zfs_fuid_info_free(zfsvfs->z_fuid_replay);
541 zfsvfs->z_fuid_replay = NULL;
542 return (error);
543 }
544
545 static int
zfs_replay_remove(zfsvfs_t * zfsvfs,lr_remove_t * lr,boolean_t byteswap)546 zfs_replay_remove(zfsvfs_t *zfsvfs, lr_remove_t *lr, boolean_t byteswap)
547 {
548 char *name = (char *)(lr + 1); /* name follows lr_remove_t */
549 znode_t *dzp;
550 struct componentname cn;
551 vnode_t *vp;
552 int error;
553 int vflg = 0;
554
555 if (byteswap)
556 byteswap_uint64_array(lr, sizeof (*lr));
557
558 if ((error = zfs_zget(zfsvfs, lr->lr_doid, &dzp)) != 0)
559 return (error);
560
561 if (lr->lr_common.lrc_txtype & TX_CI)
562 vflg |= FIGNORECASE;
563 cn.cn_nameptr = name;
564 cn.cn_namelen = strlen(name);
565 cn.cn_nameiop = DELETE;
566 cn.cn_flags = ISLASTCN;
567 //cn.cn_lkflags = LK_EXCLUSIVE | LK_RETRY;
568 cn.cn_cred = kcred;
569 vn_lock(ZTOV(dzp), LK_EXCLUSIVE | LK_RETRY);
570 error = VOP_LOOKUP(ZTOV(dzp), &vp, &cn);
571 if (error != 0) {
572 VOP_UNLOCK(ZTOV(dzp));
573 goto fail;
574 }
575 error = vn_lock(vp, LK_EXCLUSIVE);
576 if (error != 0) {
577 VOP_UNLOCK(ZTOV(dzp));
578 vrele(vp);
579 goto fail;
580 }
581
582 switch ((int)lr->lr_common.lrc_txtype) {
583 case TX_REMOVE:
584 error = VOP_REMOVE(ZTOV(dzp), vp, &cn /*,vflg*/);
585 break;
586 case TX_RMDIR:
587 error = VOP_RMDIR(ZTOV(dzp), vp, &cn /*,vflg*/);
588 break;
589 default:
590 error = ENOTSUP;
591 }
592 vput(vp);
593 VOP_UNLOCK(ZTOV(dzp));
594 fail:
595 VN_RELE(ZTOV(dzp));
596
597 return (error);
598 }
599
600 static int
zfs_replay_link(zfsvfs_t * zfsvfs,lr_link_t * lr,boolean_t byteswap)601 zfs_replay_link(zfsvfs_t *zfsvfs, lr_link_t *lr, boolean_t byteswap)
602 {
603 char *name = (char *)(lr + 1); /* name follows lr_link_t */
604 znode_t *dzp, *zp;
605 struct componentname cn;
606 int error;
607 int vflg = 0;
608
609 if (byteswap)
610 byteswap_uint64_array(lr, sizeof (*lr));
611
612 if ((error = zfs_zget(zfsvfs, lr->lr_doid, &dzp)) != 0)
613 return (error);
614
615 if ((error = zfs_zget(zfsvfs, lr->lr_link_obj, &zp)) != 0) {
616 VN_RELE(ZTOV(dzp));
617 return (error);
618 }
619
620 if (lr->lr_common.lrc_txtype & TX_CI)
621 vflg |= FIGNORECASE;
622 cn.cn_nameptr = name;
623 cn.cn_cred = kcred;
624 cn.cn_flags = 0;
625
626 vn_lock(ZTOV(dzp), LK_EXCLUSIVE | LK_RETRY);
627 vn_lock(ZTOV(zp), LK_EXCLUSIVE | LK_RETRY);
628 error = VOP_LINK(ZTOV(dzp), ZTOV(zp), &cn /*,vflg*/);
629 VOP_UNLOCK(ZTOV(zp));
630 VOP_UNLOCK(ZTOV(dzp));
631
632 VN_RELE(ZTOV(zp));
633 VN_RELE(ZTOV(dzp));
634
635 return (error);
636 }
637
638 static int
zfs_replay_rename(zfsvfs_t * zfsvfs,lr_rename_t * lr,boolean_t byteswap)639 zfs_replay_rename(zfsvfs_t *zfsvfs, lr_rename_t *lr, boolean_t byteswap)
640 {
641 char *sname = (char *)(lr + 1); /* sname and tname follow lr_rename_t */
642 char *tname = sname + strlen(sname) + 1;
643 znode_t *sdzp, *tdzp;
644 struct componentname scn, tcn;
645 vnode_t *svp, *tvp;
646 kthread_t *td = curthread;
647 int error;
648 int vflg = 0;
649
650 if (byteswap)
651 byteswap_uint64_array(lr, sizeof (*lr));
652
653 if ((error = zfs_zget(zfsvfs, lr->lr_sdoid, &sdzp)) != 0)
654 return (error);
655
656 if ((error = zfs_zget(zfsvfs, lr->lr_tdoid, &tdzp)) != 0) {
657 VN_RELE(ZTOV(sdzp));
658 return (error);
659 }
660
661 if (lr->lr_common.lrc_txtype & TX_CI)
662 vflg |= FIGNORECASE;
663 svp = tvp = NULL;
664
665 scn.cn_nameptr = sname;
666 scn.cn_namelen = strlen(sname);
667 scn.cn_nameiop = DELETE;
668 scn.cn_flags = ISLASTCN;
669 // scn.cn_lkflags = LK_EXCLUSIVE | LK_RETRY;
670 scn.cn_cred = kcred;
671 vn_lock(ZTOV(sdzp), LK_EXCLUSIVE | LK_RETRY);
672 error = VOP_LOOKUP(ZTOV(sdzp), &svp, &scn);
673 VOP_UNLOCK(ZTOV(sdzp));
674 if (error != 0)
675 goto fail;
676
677 tcn.cn_nameptr = tname;
678 tcn.cn_namelen = strlen(tname);
679 tcn.cn_nameiop = RENAME;
680 tcn.cn_flags = ISLASTCN;
681 // tcn.cn_lkflags = LK_EXCLUSIVE | LK_RETRY;
682 tcn.cn_cred = kcred;
683 vn_lock(ZTOV(tdzp), LK_EXCLUSIVE | LK_RETRY);
684 error = VOP_LOOKUP(ZTOV(tdzp), &tvp, &tcn);
685 if (error == EJUSTRETURN)
686 tvp = NULL;
687 else if (error != 0) {
688 VOP_UNLOCK(ZTOV(tdzp));
689 goto fail;
690 } else {
691 error = vn_lock(tvp, LK_EXCLUSIVE);
692 if (error != 0) {
693 VOP_UNLOCK(ZTOV(tdzp));
694 vrele(tvp);
695 goto fail;
696 }
697 }
698
699 error = VOP_RENAME(ZTOV(sdzp), svp, &scn, ZTOV(tdzp), tvp, &tcn /*,vflg*/);
700 return (error);
701 fail:
702 if (svp != NULL)
703 vrele(svp);
704 if (tvp != NULL)
705 vrele(tvp);
706 VN_RELE(ZTOV(tdzp));
707 VN_RELE(ZTOV(sdzp));
708
709 return (error);
710 }
711
712 static int
zfs_replay_write(zfsvfs_t * zfsvfs,lr_write_t * lr,boolean_t byteswap)713 zfs_replay_write(zfsvfs_t *zfsvfs, lr_write_t *lr, boolean_t byteswap)
714 {
715 char *data = (char *)(lr + 1); /* data follows lr_write_t */
716 znode_t *zp;
717 int error;
718 ssize_t resid;
719 uint64_t orig_eof, eod, offset, length;
720
721 if (byteswap)
722 byteswap_uint64_array(lr, sizeof (*lr));
723
724 if ((error = zfs_zget(zfsvfs, lr->lr_foid, &zp)) != 0) {
725 /*
726 * As we can log writes out of order, it's possible the
727 * file has been removed. In this case just drop the write
728 * and return success.
729 */
730 if (error == ENOENT)
731 error = 0;
732 return (error);
733 }
734
735 offset = lr->lr_offset;
736 length = lr->lr_length;
737 eod = offset + length; /* end of data for this write */
738
739 orig_eof = zp->z_phys->zp_size;
740
741 /* If it's a dmu_sync() block, write the whole block */
742 if (lr->lr_common.lrc_reclen == sizeof (lr_write_t)) {
743 uint64_t blocksize = BP_GET_LSIZE(&lr->lr_blkptr);
744 if (length < blocksize) {
745 offset -= offset % blocksize;
746 length = blocksize;
747 }
748 }
749
750 error = vn_rdwr(UIO_WRITE, ZTOV(zp), data, length, offset,
751 UIO_SYSSPACE, 0, RLIM64_INFINITY, kcred, &resid);
752
753 /*
754 * This may be a write from a dmu_sync() for a whole block,
755 * and may extend beyond the current end of the file.
756 * We can't just replay what was written for this TX_WRITE as
757 * a future TX_WRITE2 may extend the eof and the data for that
758 * write needs to be there. So we write the whole block and
759 * reduce the eof.
760 */
761 if (orig_eof < zp->z_phys->zp_size) /* file length grew ? */
762 zp->z_phys->zp_size = eod;
763
764 VN_RELE(ZTOV(zp));
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 end = lr->lr_offset + lr->lr_length;
789 if (end > zp->z_phys->zp_size) {
790 ASSERT3U(end - zp->z_phys->zp_size, <, zp->z_blksz);
791 zp->z_phys->zp_size = end;
792 }
793
794 VN_RELE(ZTOV(zp));
795
796 return (error);
797 }
798
799 static int
zfs_replay_truncate(zfsvfs_t * zfsvfs,lr_truncate_t * lr,boolean_t byteswap)800 zfs_replay_truncate(zfsvfs_t *zfsvfs, lr_truncate_t *lr, boolean_t byteswap)
801 {
802
803 #ifdef __NetBSD__
804 ZFS_LOG(0, "Unexpected code path, report to pjd@FreeBSD.org");
805 return (EOPNOTSUPP);
806 #else
807 if ((error = zfs_zget(zfsvfs, lr->lr_foid, &zp)) != 0)
808 return (error);
809
810 bzero(&fl, sizeof (fl));
811 fl.l_type = F_WRLCK;
812 fl.l_whence = 0;
813 fl.l_start = lr->lr_offset;
814 fl.l_len = lr->lr_length;
815
816 error = VOP_SPACE(ZTOV(zp), F_FREESP, &fl, FWRITE | FOFFMAX,
817 lr->lr_offset, kcred, NULL);
818
819 VN_RELE(ZTOV(zp));
820
821 return (error);
822 #endif
823 }
824
825 static int
zfs_replay_setattr(zfsvfs_t * zfsvfs,lr_setattr_t * lr,boolean_t byteswap)826 zfs_replay_setattr(zfsvfs_t *zfsvfs, lr_setattr_t *lr, boolean_t byteswap)
827 {
828 znode_t *zp;
829 xvattr_t xva;
830 vattr_t *vap = &xva.xva_vattr;
831 vnode_t *vp;
832 int error;
833 void *start;
834
835 xva_init(&xva);
836 if (byteswap) {
837 byteswap_uint64_array(lr, sizeof (*lr));
838
839 if ((lr->lr_mask & AT_XVATTR) &&
840 zfsvfs->z_version >= ZPL_VERSION_INITIAL)
841 zfs_replay_swap_attrs((lr_attr_t *)(lr + 1));
842 }
843
844 if ((error = zfs_zget(zfsvfs, lr->lr_foid, &zp)) != 0)
845 return (error);
846
847 zfs_init_vattr(vap, lr->lr_mask, lr->lr_mode,
848 lr->lr_uid, lr->lr_gid, 0, lr->lr_foid);
849
850 vap->va_size = lr->lr_size;
851 ZFS_TIME_DECODE(&vap->va_atime, lr->lr_atime);
852 ZFS_TIME_DECODE(&vap->va_mtime, lr->lr_mtime);
853
854 /*
855 * Fill in xvattr_t portions if necessary.
856 */
857
858 start = (lr_setattr_t *)(lr + 1);
859 if (vap->va_mask & AT_XVATTR) {
860 zfs_replay_xvattr((lr_attr_t *)start, &xva);
861 start = (caddr_t)start +
862 ZIL_XVAT_SIZE(((lr_attr_t *)start)->lr_attr_masksize);
863 } else
864 xva.xva_vattr.va_mask &= ~AT_XVATTR;
865
866 zfsvfs->z_fuid_replay = zfs_replay_fuid_domain(start, &start,
867 lr->lr_uid, lr->lr_gid);
868
869 vp = ZTOV(zp);
870 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
871 error = VOP_SETATTR(vp, vap, kcred);
872 VOP_UNLOCK(vp);
873
874 zfs_fuid_info_free(zfsvfs->z_fuid_replay);
875 zfsvfs->z_fuid_replay = NULL;
876 VN_RELE(vp);
877
878 return (error);
879 }
880
881 static int
zfs_replay_acl_v0(zfsvfs_t * zfsvfs,lr_acl_v0_t * lr,boolean_t byteswap)882 zfs_replay_acl_v0(zfsvfs_t *zfsvfs, lr_acl_v0_t *lr, boolean_t byteswap)
883 {
884 ace_t *ace = (ace_t *)(lr + 1); /* ace array follows lr_acl_t */
885 vsecattr_t vsa;
886 znode_t *zp;
887 int error;
888
889 if (byteswap) {
890 byteswap_uint64_array(lr, sizeof (*lr));
891 zfs_oldace_byteswap(ace, lr->lr_aclcnt);
892 }
893
894 if ((error = zfs_zget(zfsvfs, lr->lr_foid, &zp)) != 0)
895 return (error);
896
897 bzero(&vsa, sizeof (vsa));
898 vsa.vsa_mask = VSA_ACE | VSA_ACECNT;
899 vsa.vsa_aclcnt = lr->lr_aclcnt;
900 vsa.vsa_aclentsz = sizeof (ace_t) * vsa.vsa_aclcnt;
901 vsa.vsa_aclflags = 0;
902 vsa.vsa_aclentp = ace;
903
904 #ifdef TODO
905 error = VOP_SETSECATTR(ZTOV(zp), &vsa, 0, kcred, NULL);
906 #else
907 panic("%s:%u: unsupported condition", __func__, __LINE__);
908 #endif
909
910 VN_RELE(ZTOV(zp));
911
912 return (error);
913 }
914
915 /*
916 * Replaying ACLs is complicated by FUID support.
917 * The log record may contain some optional data
918 * to be used for replaying FUID's. These pieces
919 * are the actual FUIDs that were created initially.
920 * The FUID table index may no longer be valid and
921 * during zfs_create() a new index may be assigned.
922 * Because of this the log will contain the original
923 * doman+rid in order to create a new FUID.
924 *
925 * The individual ACEs may contain an ephemeral uid/gid which is no
926 * longer valid and will need to be replaced with an actual FUID.
927 *
928 */
929 static int
zfs_replay_acl(zfsvfs_t * zfsvfs,lr_acl_t * lr,boolean_t byteswap)930 zfs_replay_acl(zfsvfs_t *zfsvfs, lr_acl_t *lr, boolean_t byteswap)
931 {
932 ace_t *ace = (ace_t *)(lr + 1);
933 vsecattr_t vsa;
934 znode_t *zp;
935 int error;
936
937 if (byteswap) {
938 byteswap_uint64_array(lr, sizeof (*lr));
939 zfs_ace_byteswap(ace, lr->lr_acl_bytes, B_FALSE);
940 if (lr->lr_fuidcnt) {
941 byteswap_uint64_array((caddr_t)ace +
942 ZIL_ACE_LENGTH(lr->lr_acl_bytes),
943 lr->lr_fuidcnt * sizeof (uint64_t));
944 }
945 }
946
947 if ((error = zfs_zget(zfsvfs, lr->lr_foid, &zp)) != 0)
948 return (error);
949
950 #ifdef TODO
951 bzero(&vsa, sizeof (vsa));
952 vsa.vsa_mask = VSA_ACE | VSA_ACECNT | VSA_ACE_ACLFLAGS;
953 vsa.vsa_aclcnt = lr->lr_aclcnt;
954 vsa.vsa_aclentp = ace;
955 vsa.vsa_aclentsz = lr->lr_acl_bytes;
956 vsa.vsa_aclflags = lr->lr_acl_flags;
957
958 if (lr->lr_fuidcnt) {
959 void *fuidstart = (caddr_t)ace +
960 ZIL_ACE_LENGTH(lr->lr_acl_bytes);
961
962 zfsvfs->z_fuid_replay =
963 zfs_replay_fuids(fuidstart, &fuidstart,
964 lr->lr_fuidcnt, lr->lr_domcnt, 0, 0);
965 }
966
967 error = VOP_SETSECATTR(ZTOV(zp), &vsa, 0, kcred, NULL);
968
969 if (zfsvfs->z_fuid_replay)
970 zfs_fuid_info_free(zfsvfs->z_fuid_replay);
971 #else
972 error = EOPNOTSUPP;
973 #endif
974
975 zfsvfs->z_fuid_replay = NULL;
976 VN_RELE(ZTOV(zp));
977
978 return (error);
979 }
980
981 /*
982 * Callback vectors for replaying records
983 */
984 zil_replay_func_t *zfs_replay_vector[TX_MAX_TYPE] = {
985 zfs_replay_error, /* 0 no such transaction type */
986 zfs_replay_create, /* TX_CREATE */
987 zfs_replay_create, /* TX_MKDIR */
988 zfs_replay_create, /* TX_MKXATTR */
989 zfs_replay_create, /* TX_SYMLINK */
990 zfs_replay_remove, /* TX_REMOVE */
991 zfs_replay_remove, /* TX_RMDIR */
992 zfs_replay_link, /* TX_LINK */
993 zfs_replay_rename, /* TX_RENAME */
994 zfs_replay_write, /* TX_WRITE */
995 zfs_replay_truncate, /* TX_TRUNCATE */
996 zfs_replay_setattr, /* TX_SETATTR */
997 zfs_replay_acl_v0, /* TX_ACL_V0 */
998 zfs_replay_acl, /* TX_ACL */
999 zfs_replay_create_acl, /* TX_CREATE_ACL */
1000 zfs_replay_create, /* TX_CREATE_ATTR */
1001 zfs_replay_create_acl, /* TX_CREATE_ACL_ATTR */
1002 zfs_replay_create_acl, /* TX_MKDIR_ACL */
1003 zfs_replay_create, /* TX_MKDIR_ATTR */
1004 zfs_replay_create_acl, /* TX_MKDIR_ACL_ATTR */
1005 zfs_replay_write2, /* TX_WRITE2 */
1006 };
1007