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/thread.h> 33 #include <sys/file.h> 34 #include <sys/fcntl.h> 35 #include <sys/vfs.h> 36 #include <sys/fs/zfs.h> 37 #include <sys/zfs_znode.h> 38 #include <sys/zfs_dir.h> 39 #include <sys/zfs_acl.h> 40 #include <sys/zfs_fuid.h> 41 #include <sys/spa.h> 42 #include <sys/zil.h> 43 #include <sys/byteorder.h> 44 #include <sys/stat.h> 45 #include <sys/mode.h> 46 #include <sys/acl.h> 47 #include <sys/atomic.h> 48 #include <sys/cred.h> 49 50 /* 51 * Functions to replay ZFS intent log (ZIL) records 52 * The functions are called through a function vector (zfs_replay_vector) 53 * which is indexed by the transaction type. 54 */ 55 56 static void 57 zfs_init_vattr(vattr_t *vap, uint64_t mask, uint64_t mode, 58 uint64_t uid, uint64_t gid, uint64_t rdev, uint64_t nodeid) 59 { 60 bzero(vap, sizeof (*vap)); 61 vap->va_mask = (uint_t)mask; 62 vap->va_type = IFTOVT(mode); 63 vap->va_mode = mode & MODEMASK; 64 vap->va_uid = (uid_t)(IS_EPHEMERAL(uid)) ? -1 : uid; 65 vap->va_gid = (gid_t)(IS_EPHEMERAL(gid)) ? -1 : gid; 66 vap->va_rdev = zfs_cmpldev(rdev); 67 vap->va_nodeid = nodeid; 68 } 69 70 /* ARGSUSED */ 71 static int 72 zfs_replay_error(zfsvfs_t *zfsvfs, lr_t *lr, boolean_t byteswap) 73 { 74 return (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 } 133 134 static int 135 zfs_replay_domain_cnt(uint64_t uid, uint64_t gid) 136 { 137 uint64_t uid_idx; 138 uint64_t gid_idx; 139 int domcnt = 0; 140 141 uid_idx = FUID_INDEX(uid); 142 gid_idx = FUID_INDEX(gid); 143 if (uid_idx) 144 domcnt++; 145 if (gid_idx > 0 && gid_idx != uid_idx) 146 domcnt++; 147 148 return (domcnt); 149 } 150 151 static void * 152 zfs_replay_fuid_domain_common(zfs_fuid_info_t *fuid_infop, void *start, 153 int domcnt) 154 { 155 int i; 156 157 for (i = 0; i != domcnt; i++) { 158 fuid_infop->z_domain_table[i] = start; 159 start = (caddr_t)start + strlen(start) + 1; 160 } 161 162 return (start); 163 } 164 165 /* 166 * Set the uid/gid in the fuid_info structure. 167 */ 168 static void 169 zfs_replay_fuid_ugid(zfs_fuid_info_t *fuid_infop, uint64_t uid, uint64_t gid) 170 { 171 /* 172 * If owner or group are log specific FUIDs then slurp up 173 * domain information and build zfs_fuid_info_t 174 */ 175 if (IS_EPHEMERAL(uid)) 176 fuid_infop->z_fuid_owner = uid; 177 178 if (IS_EPHEMERAL(gid)) 179 fuid_infop->z_fuid_group = gid; 180 } 181 182 /* 183 * Load fuid domains into fuid_info_t 184 */ 185 static zfs_fuid_info_t * 186 zfs_replay_fuid_domain(void *buf, void **end, uint64_t uid, uint64_t gid) 187 { 188 int domcnt; 189 190 zfs_fuid_info_t *fuid_infop; 191 192 fuid_infop = zfs_fuid_info_alloc(); 193 194 domcnt = zfs_replay_domain_cnt(uid, gid); 195 196 if (domcnt == 0) 197 return (fuid_infop); 198 199 fuid_infop->z_domain_table = 200 kmem_zalloc(domcnt * sizeof (char **), KM_SLEEP); 201 202 zfs_replay_fuid_ugid(fuid_infop, uid, gid); 203 204 fuid_infop->z_domain_cnt = domcnt; 205 *end = zfs_replay_fuid_domain_common(fuid_infop, buf, domcnt); 206 return (fuid_infop); 207 } 208 209 /* 210 * load zfs_fuid_t's and fuid_domains into fuid_info_t 211 */ 212 static zfs_fuid_info_t * 213 zfs_replay_fuids(void *start, void **end, int idcnt, int domcnt, uint64_t uid, 214 uint64_t gid) 215 { 216 uint64_t *log_fuid = (uint64_t *)start; 217 zfs_fuid_info_t *fuid_infop; 218 int i; 219 220 fuid_infop = zfs_fuid_info_alloc(); 221 fuid_infop->z_domain_cnt = domcnt; 222 223 fuid_infop->z_domain_table = 224 kmem_zalloc(domcnt * sizeof (char **), KM_SLEEP); 225 226 for (i = 0; i != idcnt; i++) { 227 zfs_fuid_t *zfuid; 228 229 zfuid = kmem_alloc(sizeof (zfs_fuid_t), KM_SLEEP); 230 zfuid->z_logfuid = *log_fuid; 231 zfuid->z_id = -1; 232 zfuid->z_domidx = 0; 233 list_insert_tail(&fuid_infop->z_fuids, zfuid); 234 log_fuid++; 235 } 236 237 zfs_replay_fuid_ugid(fuid_infop, uid, gid); 238 239 *end = zfs_replay_fuid_domain_common(fuid_infop, log_fuid, domcnt); 240 return (fuid_infop); 241 } 242 243 static void 244 zfs_replay_swap_attrs(lr_attr_t *lrattr) 245 { 246 /* swap the lr_attr structure */ 247 byteswap_uint32_array(lrattr, sizeof (*lrattr)); 248 /* swap the bitmap */ 249 byteswap_uint32_array(lrattr + 1, (lrattr->lr_attr_masksize - 1) * 250 sizeof (uint32_t)); 251 /* swap the attributes, create time + 64 bit word for attributes */ 252 byteswap_uint64_array((caddr_t)(lrattr + 1) + (sizeof (uint32_t) * 253 (lrattr->lr_attr_masksize - 1)), 3 * sizeof (uint64_t)); 254 } 255 256 /* 257 * Replay file create with optional ACL, xvattr information as well 258 * as option FUID information. 259 */ 260 static int 261 zfs_replay_create_acl(zfsvfs_t *zfsvfs, 262 lr_acl_create_t *lracl, boolean_t byteswap) 263 { 264 char *name = NULL; /* location determined later */ 265 lr_create_t *lr = (lr_create_t *)lracl; 266 znode_t *dzp; 267 vnode_t *vp = NULL; 268 xvattr_t xva; 269 int vflg = 0; 270 vsecattr_t vsec = { 0 }; 271 lr_attr_t *lrattr; 272 void *aclstart; 273 void *fuidstart; 274 size_t xvatlen = 0; 275 uint64_t txtype; 276 int error; 277 278 if (byteswap) { 279 byteswap_uint64_array(lracl, sizeof (*lracl)); 280 txtype = (int)lr->lr_common.lrc_txtype; 281 if (txtype == TX_CREATE_ACL_ATTR || 282 txtype == TX_MKDIR_ACL_ATTR) { 283 lrattr = (lr_attr_t *)(caddr_t)(lracl + 1); 284 zfs_replay_swap_attrs(lrattr); 285 xvatlen = ZIL_XVAT_SIZE(lrattr->lr_attr_masksize); 286 } 287 288 aclstart = (caddr_t)(lracl + 1) + xvatlen; 289 zfs_ace_byteswap(aclstart, lracl->lr_acl_bytes, B_FALSE); 290 /* swap fuids */ 291 if (lracl->lr_fuidcnt) { 292 byteswap_uint64_array((caddr_t)aclstart + 293 ZIL_ACE_LENGTH(lracl->lr_acl_bytes), 294 lracl->lr_fuidcnt * sizeof (uint64_t)); 295 } 296 } 297 298 if ((error = zfs_zget(zfsvfs, lr->lr_doid, &dzp)) != 0) 299 return (error); 300 301 xva_init(&xva); 302 zfs_init_vattr(&xva.xva_vattr, AT_TYPE | AT_MODE | AT_UID | AT_GID, 303 lr->lr_mode, lr->lr_uid, lr->lr_gid, lr->lr_rdev, lr->lr_foid); 304 305 /* 306 * All forms of zfs create (create, mkdir, mkxattrdir, symlink) 307 * eventually end up in zfs_mknode(), which assigns the object's 308 * creation time and generation number. The generic VOP_CREATE() 309 * doesn't have either concept, so we smuggle the values inside 310 * the vattr's otherwise unused va_ctime and va_nblocks fields. 311 */ 312 ZFS_TIME_DECODE(&xva.xva_vattr.va_ctime, lr->lr_crtime); 313 xva.xva_vattr.va_nblocks = lr->lr_gen; 314 315 error = dmu_object_info(zfsvfs->z_os, lr->lr_foid, NULL); 316 if (error != ENOENT) 317 goto bail; 318 319 if (lr->lr_common.lrc_txtype & TX_CI) 320 vflg |= FIGNORECASE; 321 switch ((int)lr->lr_common.lrc_txtype) { 322 case TX_CREATE_ACL: 323 aclstart = (caddr_t)(lracl + 1); 324 fuidstart = (caddr_t)aclstart + 325 ZIL_ACE_LENGTH(lracl->lr_acl_bytes); 326 zfsvfs->z_fuid_replay = zfs_replay_fuids(fuidstart, 327 (void *)&name, lracl->lr_fuidcnt, lracl->lr_domcnt, 328 lr->lr_uid, lr->lr_gid); 329 /*FALLTHROUGH*/ 330 case TX_CREATE_ACL_ATTR: 331 if (name == NULL) { 332 lrattr = (lr_attr_t *)(caddr_t)(lracl + 1); 333 xvatlen = ZIL_XVAT_SIZE(lrattr->lr_attr_masksize); 334 xva.xva_vattr.va_mask |= AT_XVATTR; 335 zfs_replay_xvattr(lrattr, &xva); 336 } 337 vsec.vsa_mask = VSA_ACE | VSA_ACE_ACLFLAGS; 338 vsec.vsa_aclentp = (caddr_t)(lracl + 1) + xvatlen; 339 vsec.vsa_aclcnt = lracl->lr_aclcnt; 340 vsec.vsa_aclentsz = lracl->lr_acl_bytes; 341 vsec.vsa_aclflags = lracl->lr_acl_flags; 342 if (zfsvfs->z_fuid_replay == NULL) { 343 fuidstart = (caddr_t)(lracl + 1) + xvatlen + 344 ZIL_ACE_LENGTH(lracl->lr_acl_bytes); 345 zfsvfs->z_fuid_replay = 346 zfs_replay_fuids(fuidstart, 347 (void *)&name, lracl->lr_fuidcnt, lracl->lr_domcnt, 348 lr->lr_uid, lr->lr_gid); 349 } 350 351 error = VOP_CREATE(ZTOV(dzp), name, &xva.xva_vattr, 352 0, 0, &vp, kcred, vflg, NULL, &vsec); 353 break; 354 case TX_MKDIR_ACL: 355 aclstart = (caddr_t)(lracl + 1); 356 fuidstart = (caddr_t)aclstart + 357 ZIL_ACE_LENGTH(lracl->lr_acl_bytes); 358 zfsvfs->z_fuid_replay = zfs_replay_fuids(fuidstart, 359 (void *)&name, lracl->lr_fuidcnt, lracl->lr_domcnt, 360 lr->lr_uid, lr->lr_gid); 361 /*FALLTHROUGH*/ 362 case TX_MKDIR_ACL_ATTR: 363 if (name == NULL) { 364 lrattr = (lr_attr_t *)(caddr_t)(lracl + 1); 365 xvatlen = ZIL_XVAT_SIZE(lrattr->lr_attr_masksize); 366 zfs_replay_xvattr(lrattr, &xva); 367 } 368 vsec.vsa_mask = VSA_ACE | VSA_ACE_ACLFLAGS; 369 vsec.vsa_aclentp = (caddr_t)(lracl + 1) + xvatlen; 370 vsec.vsa_aclcnt = lracl->lr_aclcnt; 371 vsec.vsa_aclentsz = lracl->lr_acl_bytes; 372 vsec.vsa_aclflags = lracl->lr_acl_flags; 373 if (zfsvfs->z_fuid_replay == NULL) { 374 fuidstart = (caddr_t)(lracl + 1) + xvatlen + 375 ZIL_ACE_LENGTH(lracl->lr_acl_bytes); 376 zfsvfs->z_fuid_replay = 377 zfs_replay_fuids(fuidstart, 378 (void *)&name, lracl->lr_fuidcnt, lracl->lr_domcnt, 379 lr->lr_uid, lr->lr_gid); 380 } 381 error = VOP_MKDIR(ZTOV(dzp), name, &xva.xva_vattr, 382 &vp, kcred, NULL, vflg, &vsec); 383 break; 384 default: 385 error = ENOTSUP; 386 } 387 388 bail: 389 if (error == 0 && vp != NULL) 390 VN_RELE(vp); 391 392 VN_RELE(ZTOV(dzp)); 393 394 zfs_fuid_info_free(zfsvfs->z_fuid_replay); 395 zfsvfs->z_fuid_replay = NULL; 396 397 return (error); 398 } 399 400 static int 401 zfs_replay_create(zfsvfs_t *zfsvfs, lr_create_t *lr, boolean_t byteswap) 402 { 403 char *name = NULL; /* location determined later */ 404 char *link; /* symlink content follows name */ 405 znode_t *dzp; 406 vnode_t *vp = NULL; 407 xvattr_t xva; 408 int vflg = 0; 409 size_t lrsize = sizeof (lr_create_t); 410 lr_attr_t *lrattr; 411 void *start; 412 size_t xvatlen; 413 uint64_t txtype; 414 int error; 415 416 if (byteswap) { 417 byteswap_uint64_array(lr, sizeof (*lr)); 418 txtype = (int)lr->lr_common.lrc_txtype; 419 if (txtype == TX_CREATE_ATTR || txtype == TX_MKDIR_ATTR) 420 zfs_replay_swap_attrs((lr_attr_t *)(lr + 1)); 421 } 422 423 424 if ((error = zfs_zget(zfsvfs, lr->lr_doid, &dzp)) != 0) 425 return (error); 426 427 xva_init(&xva); 428 zfs_init_vattr(&xva.xva_vattr, AT_TYPE | AT_MODE | AT_UID | AT_GID, 429 lr->lr_mode, lr->lr_uid, lr->lr_gid, lr->lr_rdev, lr->lr_foid); 430 431 /* 432 * All forms of zfs create (create, mkdir, mkxattrdir, symlink) 433 * eventually end up in zfs_mknode(), which assigns the object's 434 * creation time and generation number. The generic VOP_CREATE() 435 * doesn't have either concept, so we smuggle the values inside 436 * the vattr's otherwise unused va_ctime and va_nblocks fields. 437 */ 438 ZFS_TIME_DECODE(&xva.xva_vattr.va_ctime, lr->lr_crtime); 439 xva.xva_vattr.va_nblocks = lr->lr_gen; 440 441 error = dmu_object_info(zfsvfs->z_os, lr->lr_foid, NULL); 442 if (error != ENOENT) 443 goto out; 444 445 if (lr->lr_common.lrc_txtype & TX_CI) 446 vflg |= FIGNORECASE; 447 448 /* 449 * Symlinks don't have fuid info, and CIFS never creates 450 * symlinks. 451 * 452 * The _ATTR versions will grab the fuid info in their subcases. 453 */ 454 if ((int)lr->lr_common.lrc_txtype != TX_SYMLINK && 455 (int)lr->lr_common.lrc_txtype != TX_MKDIR_ATTR && 456 (int)lr->lr_common.lrc_txtype != TX_CREATE_ATTR) { 457 start = (lr + 1); 458 zfsvfs->z_fuid_replay = 459 zfs_replay_fuid_domain(start, &start, 460 lr->lr_uid, lr->lr_gid); 461 } 462 463 switch ((int)lr->lr_common.lrc_txtype) { 464 case TX_CREATE_ATTR: 465 lrattr = (lr_attr_t *)(caddr_t)(lr + 1); 466 xvatlen = ZIL_XVAT_SIZE(lrattr->lr_attr_masksize); 467 zfs_replay_xvattr((lr_attr_t *)((caddr_t)lr + lrsize), &xva); 468 start = (caddr_t)(lr + 1) + xvatlen; 469 zfsvfs->z_fuid_replay = 470 zfs_replay_fuid_domain(start, &start, 471 lr->lr_uid, lr->lr_gid); 472 name = (char *)start; 473 474 /*FALLTHROUGH*/ 475 case TX_CREATE: 476 if (name == NULL) 477 name = (char *)start; 478 479 error = VOP_CREATE(ZTOV(dzp), name, &xva.xva_vattr, 480 0, 0, &vp, kcred, vflg, NULL, NULL); 481 break; 482 case TX_MKDIR_ATTR: 483 lrattr = (lr_attr_t *)(caddr_t)(lr + 1); 484 xvatlen = ZIL_XVAT_SIZE(lrattr->lr_attr_masksize); 485 zfs_replay_xvattr((lr_attr_t *)((caddr_t)lr + lrsize), &xva); 486 start = (caddr_t)(lr + 1) + xvatlen; 487 zfsvfs->z_fuid_replay = 488 zfs_replay_fuid_domain(start, &start, 489 lr->lr_uid, lr->lr_gid); 490 name = (char *)start; 491 492 /*FALLTHROUGH*/ 493 case TX_MKDIR: 494 if (name == NULL) 495 name = (char *)(lr + 1); 496 497 error = VOP_MKDIR(ZTOV(dzp), name, &xva.xva_vattr, 498 &vp, kcred, NULL, vflg, NULL); 499 break; 500 case TX_MKXATTR: 501 error = zfs_make_xattrdir(dzp, &xva.xva_vattr, &vp, kcred); 502 break; 503 case TX_SYMLINK: 504 name = (char *)(lr + 1); 505 link = name + strlen(name) + 1; 506 error = VOP_SYMLINK(ZTOV(dzp), name, &xva.xva_vattr, 507 link, kcred, NULL, vflg); 508 break; 509 default: 510 error = ENOTSUP; 511 } 512 513 out: 514 if (error == 0 && vp != NULL) 515 VN_RELE(vp); 516 517 VN_RELE(ZTOV(dzp)); 518 519 if (zfsvfs->z_fuid_replay) 520 zfs_fuid_info_free(zfsvfs->z_fuid_replay); 521 zfsvfs->z_fuid_replay = NULL; 522 return (error); 523 } 524 525 static int 526 zfs_replay_remove(zfsvfs_t *zfsvfs, lr_remove_t *lr, boolean_t byteswap) 527 { 528 char *name = (char *)(lr + 1); /* name follows lr_remove_t */ 529 znode_t *dzp; 530 int error; 531 int vflg = 0; 532 533 if (byteswap) 534 byteswap_uint64_array(lr, sizeof (*lr)); 535 536 if ((error = zfs_zget(zfsvfs, lr->lr_doid, &dzp)) != 0) 537 return (error); 538 539 if (lr->lr_common.lrc_txtype & TX_CI) 540 vflg |= FIGNORECASE; 541 542 switch ((int)lr->lr_common.lrc_txtype) { 543 case TX_REMOVE: 544 error = VOP_REMOVE(ZTOV(dzp), name, kcred, NULL, vflg); 545 break; 546 case TX_RMDIR: 547 error = VOP_RMDIR(ZTOV(dzp), name, NULL, kcred, NULL, vflg); 548 break; 549 default: 550 error = ENOTSUP; 551 } 552 553 VN_RELE(ZTOV(dzp)); 554 555 return (error); 556 } 557 558 static int 559 zfs_replay_link(zfsvfs_t *zfsvfs, lr_link_t *lr, boolean_t byteswap) 560 { 561 char *name = (char *)(lr + 1); /* name follows lr_link_t */ 562 znode_t *dzp, *zp; 563 int error; 564 int vflg = 0; 565 566 if (byteswap) 567 byteswap_uint64_array(lr, sizeof (*lr)); 568 569 if ((error = zfs_zget(zfsvfs, lr->lr_doid, &dzp)) != 0) 570 return (error); 571 572 if ((error = zfs_zget(zfsvfs, lr->lr_link_obj, &zp)) != 0) { 573 VN_RELE(ZTOV(dzp)); 574 return (error); 575 } 576 577 if (lr->lr_common.lrc_txtype & TX_CI) 578 vflg |= FIGNORECASE; 579 580 error = VOP_LINK(ZTOV(dzp), ZTOV(zp), name, kcred, NULL, vflg); 581 582 VN_RELE(ZTOV(zp)); 583 VN_RELE(ZTOV(dzp)); 584 585 return (error); 586 } 587 588 static int 589 zfs_replay_rename(zfsvfs_t *zfsvfs, lr_rename_t *lr, boolean_t byteswap) 590 { 591 char *sname = (char *)(lr + 1); /* sname and tname follow lr_rename_t */ 592 char *tname = sname + strlen(sname) + 1; 593 znode_t *sdzp, *tdzp; 594 int error; 595 int vflg = 0; 596 597 if (byteswap) 598 byteswap_uint64_array(lr, sizeof (*lr)); 599 600 if ((error = zfs_zget(zfsvfs, lr->lr_sdoid, &sdzp)) != 0) 601 return (error); 602 603 if ((error = zfs_zget(zfsvfs, lr->lr_tdoid, &tdzp)) != 0) { 604 VN_RELE(ZTOV(sdzp)); 605 return (error); 606 } 607 608 if (lr->lr_common.lrc_txtype & TX_CI) 609 vflg |= FIGNORECASE; 610 611 error = VOP_RENAME(ZTOV(sdzp), sname, ZTOV(tdzp), tname, kcred, 612 NULL, vflg); 613 614 VN_RELE(ZTOV(tdzp)); 615 VN_RELE(ZTOV(sdzp)); 616 617 return (error); 618 } 619 620 static int 621 zfs_replay_write(zfsvfs_t *zfsvfs, lr_write_t *lr, boolean_t byteswap) 622 { 623 char *data = (char *)(lr + 1); /* data follows lr_write_t */ 624 znode_t *zp; 625 int error; 626 ssize_t resid; 627 uint64_t orig_eof, eod, offset, length; 628 629 if (byteswap) 630 byteswap_uint64_array(lr, sizeof (*lr)); 631 632 if ((error = zfs_zget(zfsvfs, lr->lr_foid, &zp)) != 0) { 633 /* 634 * As we can log writes out of order, it's possible the 635 * file has been removed. In this case just drop the write 636 * and return success. 637 */ 638 if (error == ENOENT) 639 error = 0; 640 return (error); 641 } 642 643 offset = lr->lr_offset; 644 length = lr->lr_length; 645 eod = offset + length; /* end of data for this write */ 646 647 orig_eof = zp->z_phys->zp_size; 648 649 /* If it's a dmu_sync() block, write the whole block */ 650 if (lr->lr_common.lrc_reclen == sizeof (lr_write_t)) { 651 uint64_t blocksize = BP_GET_LSIZE(&lr->lr_blkptr); 652 if (length < blocksize) { 653 offset -= offset % blocksize; 654 length = blocksize; 655 } 656 } 657 658 error = vn_rdwr(UIO_WRITE, ZTOV(zp), data, length, offset, 659 UIO_SYSSPACE, 0, RLIM64_INFINITY, kcred, &resid); 660 661 /* 662 * This may be a write from a dmu_sync() for a whole block, 663 * and may extend beyond the current end of the file. 664 * We can't just replay what was written for this TX_WRITE as 665 * a future TX_WRITE2 may extend the eof and the data for that 666 * write needs to be there. So we write the whole block and 667 * reduce the eof. 668 */ 669 if (orig_eof < zp->z_phys->zp_size) /* file length grew ? */ 670 zp->z_phys->zp_size = eod; 671 672 VN_RELE(ZTOV(zp)); 673 674 return (error); 675 } 676 677 /* 678 * TX_WRITE2 are only generated when dmu_sync() returns EALREADY 679 * meaning the pool block is already being synced. So now that we always write 680 * out full blocks, all we have to do is expand the eof if 681 * the file is grown. 682 */ 683 static int 684 zfs_replay_write2(zfsvfs_t *zfsvfs, lr_write_t *lr, boolean_t byteswap) 685 { 686 znode_t *zp; 687 int error; 688 uint64_t end; 689 690 if (byteswap) 691 byteswap_uint64_array(lr, sizeof (*lr)); 692 693 if ((error = zfs_zget(zfsvfs, lr->lr_foid, &zp)) != 0) 694 return (error); 695 696 end = lr->lr_offset + lr->lr_length; 697 if (end > zp->z_phys->zp_size) { 698 ASSERT3U(end - zp->z_phys->zp_size, <, zp->z_blksz); 699 zp->z_phys->zp_size = end; 700 } 701 702 VN_RELE(ZTOV(zp)); 703 704 return (error); 705 } 706 707 static int 708 zfs_replay_truncate(zfsvfs_t *zfsvfs, lr_truncate_t *lr, boolean_t byteswap) 709 { 710 znode_t *zp; 711 flock64_t fl; 712 int error; 713 714 if (byteswap) 715 byteswap_uint64_array(lr, sizeof (*lr)); 716 717 if ((error = zfs_zget(zfsvfs, lr->lr_foid, &zp)) != 0) 718 return (error); 719 720 bzero(&fl, sizeof (fl)); 721 fl.l_type = F_WRLCK; 722 fl.l_whence = 0; 723 fl.l_start = lr->lr_offset; 724 fl.l_len = lr->lr_length; 725 726 error = VOP_SPACE(ZTOV(zp), F_FREESP, &fl, FWRITE | FOFFMAX, 727 lr->lr_offset, kcred, NULL); 728 729 VN_RELE(ZTOV(zp)); 730 731 return (error); 732 } 733 734 static int 735 zfs_replay_setattr(zfsvfs_t *zfsvfs, lr_setattr_t *lr, boolean_t byteswap) 736 { 737 znode_t *zp; 738 xvattr_t xva; 739 vattr_t *vap = &xva.xva_vattr; 740 int error; 741 void *start; 742 743 xva_init(&xva); 744 if (byteswap) { 745 byteswap_uint64_array(lr, sizeof (*lr)); 746 747 if ((lr->lr_mask & AT_XVATTR) && 748 zfsvfs->z_version >= ZPL_VERSION_INITIAL) 749 zfs_replay_swap_attrs((lr_attr_t *)(lr + 1)); 750 } 751 752 if ((error = zfs_zget(zfsvfs, lr->lr_foid, &zp)) != 0) 753 return (error); 754 755 zfs_init_vattr(vap, lr->lr_mask, lr->lr_mode, 756 lr->lr_uid, lr->lr_gid, 0, lr->lr_foid); 757 758 vap->va_size = lr->lr_size; 759 ZFS_TIME_DECODE(&vap->va_atime, lr->lr_atime); 760 ZFS_TIME_DECODE(&vap->va_mtime, lr->lr_mtime); 761 762 /* 763 * Fill in xvattr_t portions if necessary. 764 */ 765 766 start = (lr_setattr_t *)(lr + 1); 767 if (vap->va_mask & AT_XVATTR) { 768 zfs_replay_xvattr((lr_attr_t *)start, &xva); 769 start = (caddr_t)start + 770 ZIL_XVAT_SIZE(((lr_attr_t *)start)->lr_attr_masksize); 771 } else 772 xva.xva_vattr.va_mask &= ~AT_XVATTR; 773 774 zfsvfs->z_fuid_replay = zfs_replay_fuid_domain(start, &start, 775 lr->lr_uid, lr->lr_gid); 776 777 error = VOP_SETATTR(ZTOV(zp), vap, 0, kcred, NULL); 778 779 zfs_fuid_info_free(zfsvfs->z_fuid_replay); 780 zfsvfs->z_fuid_replay = NULL; 781 VN_RELE(ZTOV(zp)); 782 783 return (error); 784 } 785 786 static int 787 zfs_replay_acl_v0(zfsvfs_t *zfsvfs, lr_acl_v0_t *lr, boolean_t byteswap) 788 { 789 ace_t *ace = (ace_t *)(lr + 1); /* ace array follows lr_acl_t */ 790 vsecattr_t vsa; 791 znode_t *zp; 792 int error; 793 794 if (byteswap) { 795 byteswap_uint64_array(lr, sizeof (*lr)); 796 zfs_oldace_byteswap(ace, lr->lr_aclcnt); 797 } 798 799 if ((error = zfs_zget(zfsvfs, lr->lr_foid, &zp)) != 0) 800 return (error); 801 802 bzero(&vsa, sizeof (vsa)); 803 vsa.vsa_mask = VSA_ACE | VSA_ACECNT; 804 vsa.vsa_aclcnt = lr->lr_aclcnt; 805 vsa.vsa_aclentsz = sizeof (ace_t) * vsa.vsa_aclcnt; 806 vsa.vsa_aclflags = 0; 807 vsa.vsa_aclentp = ace; 808 809 error = VOP_SETSECATTR(ZTOV(zp), &vsa, 0, kcred, NULL); 810 811 VN_RELE(ZTOV(zp)); 812 813 return (error); 814 } 815 816 /* 817 * Replaying ACLs is complicated by FUID support. 818 * The log record may contain some optional data 819 * to be used for replaying FUID's. These pieces 820 * are the actual FUIDs that were created initially. 821 * The FUID table index may no longer be valid and 822 * during zfs_create() a new index may be assigned. 823 * Because of this the log will contain the original 824 * doman+rid in order to create a new FUID. 825 * 826 * The individual ACEs may contain an ephemeral uid/gid which is no 827 * longer valid and will need to be replaced with an actual FUID. 828 * 829 */ 830 static int 831 zfs_replay_acl(zfsvfs_t *zfsvfs, lr_acl_t *lr, boolean_t byteswap) 832 { 833 ace_t *ace = (ace_t *)(lr + 1); 834 vsecattr_t vsa; 835 znode_t *zp; 836 int error; 837 838 if (byteswap) { 839 byteswap_uint64_array(lr, sizeof (*lr)); 840 zfs_ace_byteswap(ace, lr->lr_acl_bytes, B_FALSE); 841 if (lr->lr_fuidcnt) { 842 byteswap_uint64_array((caddr_t)ace + 843 ZIL_ACE_LENGTH(lr->lr_acl_bytes), 844 lr->lr_fuidcnt * sizeof (uint64_t)); 845 } 846 } 847 848 if ((error = zfs_zget(zfsvfs, lr->lr_foid, &zp)) != 0) 849 return (error); 850 851 bzero(&vsa, sizeof (vsa)); 852 vsa.vsa_mask = VSA_ACE | VSA_ACECNT | VSA_ACE_ACLFLAGS; 853 vsa.vsa_aclcnt = lr->lr_aclcnt; 854 vsa.vsa_aclentp = ace; 855 vsa.vsa_aclentsz = lr->lr_acl_bytes; 856 vsa.vsa_aclflags = lr->lr_acl_flags; 857 858 if (lr->lr_fuidcnt) { 859 void *fuidstart = (caddr_t)ace + 860 ZIL_ACE_LENGTH(lr->lr_acl_bytes); 861 862 zfsvfs->z_fuid_replay = 863 zfs_replay_fuids(fuidstart, &fuidstart, 864 lr->lr_fuidcnt, lr->lr_domcnt, 0, 0); 865 } 866 867 error = VOP_SETSECATTR(ZTOV(zp), &vsa, 0, kcred, NULL); 868 869 if (zfsvfs->z_fuid_replay) 870 zfs_fuid_info_free(zfsvfs->z_fuid_replay); 871 872 zfsvfs->z_fuid_replay = NULL; 873 VN_RELE(ZTOV(zp)); 874 875 return (error); 876 } 877 878 /* 879 * Callback vectors for replaying records 880 */ 881 zil_replay_func_t *zfs_replay_vector[TX_MAX_TYPE] = { 882 zfs_replay_error, /* 0 no such transaction type */ 883 zfs_replay_create, /* TX_CREATE */ 884 zfs_replay_create, /* TX_MKDIR */ 885 zfs_replay_create, /* TX_MKXATTR */ 886 zfs_replay_create, /* TX_SYMLINK */ 887 zfs_replay_remove, /* TX_REMOVE */ 888 zfs_replay_remove, /* TX_RMDIR */ 889 zfs_replay_link, /* TX_LINK */ 890 zfs_replay_rename, /* TX_RENAME */ 891 zfs_replay_write, /* TX_WRITE */ 892 zfs_replay_truncate, /* TX_TRUNCATE */ 893 zfs_replay_setattr, /* TX_SETATTR */ 894 zfs_replay_acl_v0, /* TX_ACL_V0 */ 895 zfs_replay_acl, /* TX_ACL */ 896 zfs_replay_create_acl, /* TX_CREATE_ACL */ 897 zfs_replay_create, /* TX_CREATE_ATTR */ 898 zfs_replay_create_acl, /* TX_CREATE_ACL_ATTR */ 899 zfs_replay_create_acl, /* TX_MKDIR_ACL */ 900 zfs_replay_create, /* TX_MKDIR_ATTR */ 901 zfs_replay_create_acl, /* TX_MKDIR_ACL_ATTR */ 902 zfs_replay_write2, /* TX_WRITE2 */ 903 }; 904