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.
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.
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]
22 * Copyright (c) 2005, 2010, Oracle and/or its affiliates. All rights reserved.
23 * Copyright (c) 2013, 2015 by Delphix. All rights reserved.
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>
33 #include <sys/fcntl.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>
42 #include <sys/byteorder.h>
45 #include <sys/atomic.h>
47 #include <sys/namei.h>
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.
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)
60 vap->va_mask = (uint_t)mask;
62 vap->va_type = IFTOVT(mode);
64 vap->va_mode = mode & MODEMASK;
66 vap->va_uid = (uid_t)(IS_EPHEMERAL(uid)) ? -1 : uid;
68 vap->va_gid = (gid_t)(IS_EPHEMERAL(gid)) ? -1 : gid;
69 vap->va_rdev = zfs_cmpldev(rdev);
70 vap->va_nodeid = nodeid;
75 zfs_replay_error(void *arg1, void *arg2, boolean_t byteswap)
77 return (SET_ERROR(ENOTSUP));
81 zfs_replay_xvattr(lr_attr_t *lrattr, xvattr_t *xvap)
83 xoptattr_t *xoap = NULL;
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 */
96 ASSERT(lrattr->lr_attr_masksize == xvap->xva_mapsize);
98 bitmap = &lrattr->lr_attr_bitmap;
99 for (i = 0; i != lrattr->lr_attr_masksize; i++, bitmap++)
100 xvap->xva_reqattrmap[i] = *bitmap;
102 attrs = (uint64_t *)(lrattr + lrattr->lr_attr_masksize - 1);
104 scanstamp = (caddr_t)(crtime + 2);
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);
142 zfs_replay_domain_cnt(uint64_t uid, uint64_t gid)
148 uid_idx = FUID_INDEX(uid);
149 gid_idx = FUID_INDEX(gid);
152 if (gid_idx > 0 && gid_idx != uid_idx)
159 zfs_replay_fuid_domain_common(zfs_fuid_info_t *fuid_infop, void *start,
164 for (i = 0; i != domcnt; i++) {
165 fuid_infop->z_domain_table[i] = start;
166 start = (caddr_t)start + strlen(start) + 1;
173 * Set the uid/gid in the fuid_info structure.
176 zfs_replay_fuid_ugid(zfs_fuid_info_t *fuid_infop, uint64_t uid, uint64_t gid)
179 * If owner or group are log specific FUIDs then slurp up
180 * domain information and build zfs_fuid_info_t
182 if (IS_EPHEMERAL(uid))
183 fuid_infop->z_fuid_owner = uid;
185 if (IS_EPHEMERAL(gid))
186 fuid_infop->z_fuid_group = gid;
190 * Load fuid domains into fuid_info_t
192 static zfs_fuid_info_t *
193 zfs_replay_fuid_domain(void *buf, void **end, uint64_t uid, uint64_t gid)
197 zfs_fuid_info_t *fuid_infop;
199 fuid_infop = zfs_fuid_info_alloc();
201 domcnt = zfs_replay_domain_cnt(uid, gid);
206 fuid_infop->z_domain_table =
207 kmem_zalloc(domcnt * sizeof (char **), KM_SLEEP);
209 zfs_replay_fuid_ugid(fuid_infop, uid, gid);
211 fuid_infop->z_domain_cnt = domcnt;
212 *end = zfs_replay_fuid_domain_common(fuid_infop, buf, domcnt);
217 * load zfs_fuid_t's and fuid_domains into fuid_info_t
219 static zfs_fuid_info_t *
220 zfs_replay_fuids(void *start, void **end, int idcnt, int domcnt, uint64_t uid,
223 uint64_t *log_fuid = (uint64_t *)start;
224 zfs_fuid_info_t *fuid_infop;
227 fuid_infop = zfs_fuid_info_alloc();
228 fuid_infop->z_domain_cnt = domcnt;
230 fuid_infop->z_domain_table =
231 kmem_zalloc(domcnt * sizeof (char **), KM_SLEEP);
233 for (i = 0; i != idcnt; i++) {
236 zfuid = kmem_alloc(sizeof (zfs_fuid_t), KM_SLEEP);
237 zfuid->z_logfuid = *log_fuid;
240 list_insert_tail(&fuid_infop->z_fuids, zfuid);
244 zfs_replay_fuid_ugid(fuid_infop, uid, gid);
246 *end = zfs_replay_fuid_domain_common(fuid_infop, log_fuid, domcnt);
251 zfs_replay_swap_attrs(lr_attr_t *lrattr)
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) *
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));
264 * Replay file create with optional ACL, xvattr information as well
265 * as option FUID information.
268 zfs_replay_create_acl(void *arg1, void *arg2, boolean_t byteswap)
270 zfsvfs_t *zfsvfs = arg1;
271 lr_acl_create_t *lracl = arg2;
272 char *name = NULL; /* location determined later */
273 lr_create_t *lr = (lr_create_t *)lracl;
278 vsecattr_t vsec = { 0 };
288 txtype = (lr->lr_common.lrc_txtype & ~TX_CI);
290 byteswap_uint64_array(lracl, sizeof (*lracl));
291 if (txtype == TX_CREATE_ACL_ATTR ||
292 txtype == TX_MKDIR_ACL_ATTR) {
293 lrattr = (lr_attr_t *)(caddr_t)(lracl + 1);
294 zfs_replay_swap_attrs(lrattr);
295 xvatlen = ZIL_XVAT_SIZE(lrattr->lr_attr_masksize);
298 aclstart = (caddr_t)(lracl + 1) + xvatlen;
299 zfs_ace_byteswap(aclstart, lracl->lr_acl_bytes, B_FALSE);
301 if (lracl->lr_fuidcnt) {
302 byteswap_uint64_array((caddr_t)aclstart +
303 ZIL_ACE_LENGTH(lracl->lr_acl_bytes),
304 lracl->lr_fuidcnt * sizeof (uint64_t));
308 if ((error = zfs_zget(zfsvfs, lr->lr_doid, &dzp)) != 0)
311 objid = LR_FOID_GET_OBJ(lr->lr_foid);
312 dnodesize = LR_FOID_GET_SLOTS(lr->lr_foid) << DNODE_SHIFT;
315 zfs_init_vattr(&xva.xva_vattr, AT_TYPE | AT_MODE | AT_UID | AT_GID,
316 lr->lr_mode, lr->lr_uid, lr->lr_gid, lr->lr_rdev, objid);
319 * All forms of zfs create (create, mkdir, mkxattrdir, symlink)
320 * eventually end up in zfs_mknode(), which assigns the object's
321 * creation time, generation number, and dnode size. The generic
322 * zfs_create() has no concept of these attributes, so we smuggle
323 * the values inside the vattr's otherwise unused va_ctime,
324 * va_nblocks, and va_fsid fields.
327 ZFS_TIME_DECODE(&xva.xva_vattr.va_ctime, lr->lr_crtime);
328 xva.xva_vattr.va_nblocks = lr->lr_gen;
329 xva.xva_vattr.va_fsid = dnodesize;
331 error = dmu_object_info(zfsvfs->z_os, lr->lr_foid, NULL);
335 if (lr->lr_common.lrc_txtype & TX_CI)
339 aclstart = (caddr_t)(lracl + 1);
340 fuidstart = (caddr_t)aclstart +
341 ZIL_ACE_LENGTH(lracl->lr_acl_bytes);
342 zfsvfs->z_fuid_replay = zfs_replay_fuids(fuidstart,
343 (void *)&name, lracl->lr_fuidcnt, lracl->lr_domcnt,
344 lr->lr_uid, lr->lr_gid);
346 case TX_CREATE_ACL_ATTR:
348 lrattr = (lr_attr_t *)(caddr_t)(lracl + 1);
349 xvatlen = ZIL_XVAT_SIZE(lrattr->lr_attr_masksize);
350 xva.xva_vattr.va_mask |= AT_XVATTR;
351 zfs_replay_xvattr(lrattr, &xva);
353 vsec.vsa_mask = VSA_ACE | VSA_ACE_ACLFLAGS;
354 vsec.vsa_aclentp = (caddr_t)(lracl + 1) + xvatlen;
355 vsec.vsa_aclcnt = lracl->lr_aclcnt;
356 vsec.vsa_aclentsz = lracl->lr_acl_bytes;
357 vsec.vsa_aclflags = lracl->lr_acl_flags;
358 if (zfsvfs->z_fuid_replay == NULL) {
359 fuidstart = (caddr_t)(lracl + 1) + xvatlen +
360 ZIL_ACE_LENGTH(lracl->lr_acl_bytes);
361 zfsvfs->z_fuid_replay =
362 zfs_replay_fuids(fuidstart,
363 (void *)&name, lracl->lr_fuidcnt, lracl->lr_domcnt,
364 lr->lr_uid, lr->lr_gid);
368 error = VOP_CREATE(ZTOV(dzp), name, &xva.xva_vattr,
369 0, 0, &vp, kcred, vflg, NULL, &vsec);
371 panic("%s:%u: unsupported condition", __func__, __LINE__);
375 aclstart = (caddr_t)(lracl + 1);
376 fuidstart = (caddr_t)aclstart +
377 ZIL_ACE_LENGTH(lracl->lr_acl_bytes);
378 zfsvfs->z_fuid_replay = zfs_replay_fuids(fuidstart,
379 (void *)&name, lracl->lr_fuidcnt, lracl->lr_domcnt,
380 lr->lr_uid, lr->lr_gid);
382 case TX_MKDIR_ACL_ATTR:
384 lrattr = (lr_attr_t *)(caddr_t)(lracl + 1);
385 xvatlen = ZIL_XVAT_SIZE(lrattr->lr_attr_masksize);
386 zfs_replay_xvattr(lrattr, &xva);
388 vsec.vsa_mask = VSA_ACE | VSA_ACE_ACLFLAGS;
389 vsec.vsa_aclentp = (caddr_t)(lracl + 1) + xvatlen;
390 vsec.vsa_aclcnt = lracl->lr_aclcnt;
391 vsec.vsa_aclentsz = lracl->lr_acl_bytes;
392 vsec.vsa_aclflags = lracl->lr_acl_flags;
393 if (zfsvfs->z_fuid_replay == NULL) {
394 fuidstart = (caddr_t)(lracl + 1) + xvatlen +
395 ZIL_ACE_LENGTH(lracl->lr_acl_bytes);
396 zfsvfs->z_fuid_replay =
397 zfs_replay_fuids(fuidstart,
398 (void *)&name, lracl->lr_fuidcnt, lracl->lr_domcnt,
399 lr->lr_uid, lr->lr_gid);
402 error = VOP_MKDIR(ZTOV(dzp), name, &xva.xva_vattr,
403 &vp, kcred, NULL, vflg, &vsec);
405 panic("%s:%u: unsupported condition", __func__, __LINE__);
409 error = SET_ERROR(ENOTSUP);
413 if (error == 0 && vp != NULL)
418 if (zfsvfs->z_fuid_replay)
419 zfs_fuid_info_free(zfsvfs->z_fuid_replay);
420 zfsvfs->z_fuid_replay = NULL;
426 zfs_replay_create(void *arg1, void *arg2, boolean_t byteswap)
428 zfsvfs_t *zfsvfs = arg1;
429 lr_create_t *lr = arg2;
430 char *name = NULL; /* location determined later */
431 char *link; /* symlink content follows name */
436 size_t lrsize = sizeof (lr_create_t);
441 struct componentname cn;
444 txtype = (lr->lr_common.lrc_txtype & ~TX_CI);
446 byteswap_uint64_array(lr, sizeof (*lr));
447 if (txtype == TX_CREATE_ATTR || txtype == TX_MKDIR_ATTR)
448 zfs_replay_swap_attrs((lr_attr_t *)(lr + 1));
452 if ((error = zfs_zget(zfsvfs, lr->lr_doid, &dzp)) != 0)
455 uint64_t objid = LR_FOID_GET_OBJ(lr->lr_foid);
456 int dnodesize = LR_FOID_GET_SLOTS(lr->lr_foid) << DNODE_SHIFT;
459 zfs_init_vattr(&xva.xva_vattr, AT_TYPE | AT_MODE | AT_UID | AT_GID,
460 lr->lr_mode, lr->lr_uid, lr->lr_gid, lr->lr_rdev, objid);
463 * All forms of zfs create (create, mkdir, mkxattrdir, symlink)
464 * eventually end up in zfs_mknode(), which assigns the object's
465 * creation time, generation number, and dnode slot count. The
466 * generic zfs_create() has no concept of these attributes, so
467 * we smuggle the values inside * the vattr's otherwise unused
468 * va_ctime, va_nblocks, and va_nlink fields.
470 ZFS_TIME_DECODE(&xva.xva_vattr.va_ctime, lr->lr_crtime);
471 xva.xva_vattr.va_nblocks = lr->lr_gen;
472 xva.xva_vattr.va_fsid = dnodesize;
474 error = dmu_object_info(zfsvfs->z_os, objid, NULL);
478 if (lr->lr_common.lrc_txtype & TX_CI)
482 * Symlinks don't have fuid info, and CIFS never creates
485 * The _ATTR versions will grab the fuid info in their subcases.
487 if ((int)lr->lr_common.lrc_txtype != TX_SYMLINK &&
488 (int)lr->lr_common.lrc_txtype != TX_MKDIR_ATTR &&
489 (int)lr->lr_common.lrc_txtype != TX_CREATE_ATTR) {
491 zfsvfs->z_fuid_replay =
492 zfs_replay_fuid_domain(start, &start,
493 lr->lr_uid, lr->lr_gid);
497 cn.cn_thread = curthread;
498 cn.cn_flags = SAVENAME;
500 vn_lock(ZTOV(dzp), LK_EXCLUSIVE | LK_RETRY);
503 lrattr = (lr_attr_t *)(caddr_t)(lr + 1);
504 xvatlen = ZIL_XVAT_SIZE(lrattr->lr_attr_masksize);
505 zfs_replay_xvattr((lr_attr_t *)((caddr_t)lr + lrsize), &xva);
506 start = (caddr_t)(lr + 1) + xvatlen;
507 zfsvfs->z_fuid_replay =
508 zfs_replay_fuid_domain(start, &start,
509 lr->lr_uid, lr->lr_gid);
510 name = (char *)start;
515 name = (char *)start;
517 cn.cn_nameptr = name;
518 error = VOP_CREATE(ZTOV(dzp), &vp, &cn, &xva.xva_vattr /*,vflg*/);
521 lrattr = (lr_attr_t *)(caddr_t)(lr + 1);
522 xvatlen = ZIL_XVAT_SIZE(lrattr->lr_attr_masksize);
523 zfs_replay_xvattr((lr_attr_t *)((caddr_t)lr + lrsize), &xva);
524 start = (caddr_t)(lr + 1) + xvatlen;
525 zfsvfs->z_fuid_replay =
526 zfs_replay_fuid_domain(start, &start,
527 lr->lr_uid, lr->lr_gid);
528 name = (char *)start;
533 name = (char *)(lr + 1);
535 cn.cn_nameptr = name;
536 error = VOP_MKDIR(ZTOV(dzp), &vp, &cn, &xva.xva_vattr /*,vflg*/);
539 error = zfs_make_xattrdir(dzp, &xva.xva_vattr, &vp, kcred);
542 name = (char *)(lr + 1);
543 link = name + strlen(name) + 1;
544 cn.cn_nameptr = name;
545 error = VOP_SYMLINK(ZTOV(dzp), &vp, &cn, &xva.xva_vattr, link /*,vflg*/);
548 error = SET_ERROR(ENOTSUP);
550 VOP_UNLOCK(ZTOV(dzp), 0);
553 if (error == 0 && vp != NULL)
558 if (zfsvfs->z_fuid_replay)
559 zfs_fuid_info_free(zfsvfs->z_fuid_replay);
560 zfsvfs->z_fuid_replay = NULL;
565 zfs_replay_remove(void *arg1, void *arg2, boolean_t byteswap)
567 zfsvfs_t *zfsvfs = arg1;
568 lr_remove_t *lr = arg2;
569 char *name = (char *)(lr + 1); /* name follows lr_remove_t */
571 struct componentname cn;
577 byteswap_uint64_array(lr, sizeof (*lr));
579 if ((error = zfs_zget(zfsvfs, lr->lr_doid, &dzp)) != 0)
582 if (lr->lr_common.lrc_txtype & TX_CI)
584 cn.cn_nameptr = name;
585 cn.cn_namelen = strlen(name);
586 cn.cn_nameiop = DELETE;
587 cn.cn_flags = ISLASTCN | SAVENAME;
588 cn.cn_lkflags = LK_EXCLUSIVE | LK_RETRY;
590 cn.cn_thread = curthread;
591 vn_lock(ZTOV(dzp), LK_EXCLUSIVE | LK_RETRY);
592 error = VOP_LOOKUP(ZTOV(dzp), &vp, &cn);
594 VOP_UNLOCK(ZTOV(dzp), 0);
598 switch ((int)lr->lr_common.lrc_txtype) {
600 error = VOP_REMOVE(ZTOV(dzp), vp, &cn /*,vflg*/);
603 error = VOP_RMDIR(ZTOV(dzp), vp, &cn /*,vflg*/);
606 error = SET_ERROR(ENOTSUP);
609 VOP_UNLOCK(ZTOV(dzp), 0);
618 zfs_replay_link(void *arg1, void *arg2, boolean_t byteswap)
620 zfsvfs_t *zfsvfs = arg1;
621 lr_link_t *lr = arg2;
622 char *name = (char *)(lr + 1); /* name follows lr_link_t */
624 struct componentname cn;
629 byteswap_uint64_array(lr, sizeof (*lr));
631 if ((error = zfs_zget(zfsvfs, lr->lr_doid, &dzp)) != 0)
634 if ((error = zfs_zget(zfsvfs, lr->lr_link_obj, &zp)) != 0) {
639 if (lr->lr_common.lrc_txtype & TX_CI)
642 cn.cn_nameptr = name;
644 cn.cn_thread = curthread;
645 cn.cn_flags = SAVENAME;
647 vn_lock(ZTOV(dzp), LK_EXCLUSIVE | LK_RETRY);
648 vn_lock(ZTOV(zp), LK_EXCLUSIVE | LK_RETRY);
649 error = VOP_LINK(ZTOV(dzp), ZTOV(zp), &cn /*,vflg*/);
650 VOP_UNLOCK(ZTOV(zp), 0);
651 VOP_UNLOCK(ZTOV(dzp), 0);
660 zfs_replay_rename(void *arg1, void *arg2, boolean_t byteswap)
662 zfsvfs_t *zfsvfs = arg1;
663 lr_rename_t *lr = arg2;
664 char *sname = (char *)(lr + 1); /* sname and tname follow lr_rename_t */
665 char *tname = sname + strlen(sname) + 1;
666 znode_t *sdzp, *tdzp;
667 struct componentname scn, tcn;
669 kthread_t *td = curthread;
674 byteswap_uint64_array(lr, sizeof (*lr));
676 if ((error = zfs_zget(zfsvfs, lr->lr_sdoid, &sdzp)) != 0)
679 if ((error = zfs_zget(zfsvfs, lr->lr_tdoid, &tdzp)) != 0) {
684 if (lr->lr_common.lrc_txtype & TX_CI)
688 scn.cn_nameptr = sname;
689 scn.cn_namelen = strlen(sname);
690 scn.cn_nameiop = DELETE;
691 scn.cn_flags = ISLASTCN | SAVENAME;
692 scn.cn_lkflags = LK_EXCLUSIVE | LK_RETRY;
695 vn_lock(ZTOV(sdzp), LK_EXCLUSIVE | LK_RETRY);
696 error = VOP_LOOKUP(ZTOV(sdzp), &svp, &scn);
697 VOP_UNLOCK(ZTOV(sdzp), 0);
702 tcn.cn_nameptr = tname;
703 tcn.cn_namelen = strlen(tname);
704 tcn.cn_nameiop = RENAME;
705 tcn.cn_flags = ISLASTCN | SAVENAME;
706 tcn.cn_lkflags = LK_EXCLUSIVE | LK_RETRY;
709 vn_lock(ZTOV(tdzp), LK_EXCLUSIVE | LK_RETRY);
710 error = VOP_LOOKUP(ZTOV(tdzp), &tvp, &tcn);
711 if (error == EJUSTRETURN)
713 else if (error != 0) {
714 VOP_UNLOCK(ZTOV(tdzp), 0);
718 error = VOP_RENAME(ZTOV(sdzp), svp, &scn, ZTOV(tdzp), tvp, &tcn /*,vflg*/);
732 zfs_replay_write(void *arg1, void *arg2, boolean_t byteswap)
734 zfsvfs_t *zfsvfs = arg1;
735 lr_write_t *lr = arg2;
736 char *data = (char *)(lr + 1); /* data follows lr_write_t */
740 uint64_t eod, offset, length;
743 byteswap_uint64_array(lr, sizeof (*lr));
745 if ((error = zfs_zget(zfsvfs, lr->lr_foid, &zp)) != 0) {
747 * As we can log writes out of order, it's possible the
748 * file has been removed. In this case just drop the write
749 * and return success.
756 offset = lr->lr_offset;
757 length = lr->lr_length;
758 eod = offset + length; /* end of data for this write */
761 * This may be a write from a dmu_sync() for a whole block,
762 * and may extend beyond the current end of the file.
763 * We can't just replay what was written for this TX_WRITE as
764 * a future TX_WRITE2 may extend the eof and the data for that
765 * write needs to be there. So we write the whole block and
766 * reduce the eof. This needs to be done within the single dmu
767 * transaction created within vn_rdwr -> zfs_write. So a possible
768 * new end of file is passed through in zfsvfs->z_replay_eof
771 zfsvfs->z_replay_eof = 0; /* 0 means don't change end of file */
773 /* If it's a dmu_sync() block, write the whole block */
774 if (lr->lr_common.lrc_reclen == sizeof (lr_write_t)) {
775 uint64_t blocksize = BP_GET_LSIZE(&lr->lr_blkptr);
776 if (length < blocksize) {
777 offset -= offset % blocksize;
780 if (zp->z_size < eod)
781 zfsvfs->z_replay_eof = eod;
784 error = vn_rdwr(UIO_WRITE, ZTOV(zp), data, length, offset,
785 UIO_SYSSPACE, 0, RLIM64_INFINITY, kcred, &resid);
788 zfsvfs->z_replay_eof = 0; /* safety */
794 * TX_WRITE2 are only generated when dmu_sync() returns EALREADY
795 * meaning the pool block is already being synced. So now that we always write
796 * out full blocks, all we have to do is expand the eof if
800 zfs_replay_write2(void *arg1, void *arg2, boolean_t byteswap)
802 zfsvfs_t *zfsvfs = arg1;
803 lr_write_t *lr = arg2;
809 byteswap_uint64_array(lr, sizeof (*lr));
811 if ((error = zfs_zget(zfsvfs, lr->lr_foid, &zp)) != 0)
815 end = lr->lr_offset + lr->lr_length;
816 if (end > zp->z_size) {
817 dmu_tx_t *tx = dmu_tx_create(zfsvfs->z_os);
820 dmu_tx_hold_sa(tx, zp->z_sa_hdl, B_FALSE);
821 error = dmu_tx_assign(tx, TXG_WAIT);
824 if (error == ERESTART) {
832 (void) sa_update(zp->z_sa_hdl, SA_ZPL_SIZE(zfsvfs),
833 (void *)&zp->z_size, sizeof (uint64_t), tx);
835 /* Ensure the replayed seq is updated */
836 (void) zil_replaying(zfsvfs->z_log, tx);
847 zfs_replay_truncate(void *arg1, void *arg2, boolean_t byteswap)
850 zfsvfs_t *zfsvfs = arg1;
851 lr_truncate_t *lr = arg2;
857 byteswap_uint64_array(lr, sizeof (*lr));
859 if ((error = zfs_zget(zfsvfs, lr->lr_foid, &zp)) != 0)
862 bzero(&fl, sizeof (fl));
865 fl.l_start = lr->lr_offset;
866 fl.l_len = lr->lr_length;
868 error = VOP_SPACE(ZTOV(zp), F_FREESP, &fl, FWRITE | FOFFMAX,
869 lr->lr_offset, kcred, NULL);
875 ZFS_LOG(0, "Unexpected code path, report to pjd@FreeBSD.org");
881 zfs_replay_setattr(void *arg1, void *arg2, boolean_t byteswap)
883 zfsvfs_t *zfsvfs = arg1;
884 lr_setattr_t *lr = arg2;
887 vattr_t *vap = &xva.xva_vattr;
894 byteswap_uint64_array(lr, sizeof (*lr));
896 if ((lr->lr_mask & AT_XVATTR) &&
897 zfsvfs->z_version >= ZPL_VERSION_INITIAL)
898 zfs_replay_swap_attrs((lr_attr_t *)(lr + 1));
901 if ((error = zfs_zget(zfsvfs, lr->lr_foid, &zp)) != 0)
904 zfs_init_vattr(vap, lr->lr_mask, lr->lr_mode,
905 lr->lr_uid, lr->lr_gid, 0, lr->lr_foid);
907 vap->va_size = lr->lr_size;
908 ZFS_TIME_DECODE(&vap->va_atime, lr->lr_atime);
909 ZFS_TIME_DECODE(&vap->va_mtime, lr->lr_mtime);
912 * Fill in xvattr_t portions if necessary.
915 start = (lr_setattr_t *)(lr + 1);
916 if (vap->va_mask & AT_XVATTR) {
917 zfs_replay_xvattr((lr_attr_t *)start, &xva);
918 start = (caddr_t)start +
919 ZIL_XVAT_SIZE(((lr_attr_t *)start)->lr_attr_masksize);
921 xva.xva_vattr.va_mask &= ~AT_XVATTR;
923 zfsvfs->z_fuid_replay = zfs_replay_fuid_domain(start, &start,
924 lr->lr_uid, lr->lr_gid);
927 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
928 error = VOP_SETATTR(vp, vap, kcred);
931 zfs_fuid_info_free(zfsvfs->z_fuid_replay);
932 zfsvfs->z_fuid_replay = NULL;
938 extern int zfs_setsecattr(vnode_t *vp, vsecattr_t *vsecp, int flag, cred_t *cr,
939 caller_context_t *ct);
942 zfs_replay_acl_v0(void *arg1, void *arg2, boolean_t byteswap)
944 zfsvfs_t *zfsvfs = arg1;
945 lr_acl_v0_t *lr = arg2;
946 ace_t *ace = (ace_t *)(lr + 1); /* ace array follows lr_acl_t */
953 byteswap_uint64_array(lr, sizeof (*lr));
954 zfs_oldace_byteswap(ace, lr->lr_aclcnt);
957 if ((error = zfs_zget(zfsvfs, lr->lr_foid, &zp)) != 0)
960 bzero(&vsa, sizeof (vsa));
961 vsa.vsa_mask = VSA_ACE | VSA_ACECNT;
962 vsa.vsa_aclcnt = lr->lr_aclcnt;
963 vsa.vsa_aclentsz = sizeof (ace_t) * vsa.vsa_aclcnt;
964 vsa.vsa_aclflags = 0;
965 vsa.vsa_aclentp = ace;
968 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
969 error = zfs_setsecattr(vp, &vsa, 0, kcred, NULL);
978 * Replaying ACLs is complicated by FUID support.
979 * The log record may contain some optional data
980 * to be used for replaying FUID's. These pieces
981 * are the actual FUIDs that were created initially.
982 * The FUID table index may no longer be valid and
983 * during zfs_create() a new index may be assigned.
984 * Because of this the log will contain the original
985 * doman+rid in order to create a new FUID.
987 * The individual ACEs may contain an ephemeral uid/gid which is no
988 * longer valid and will need to be replaced with an actual FUID.
992 zfs_replay_acl(void *arg1, void *arg2, boolean_t byteswap)
994 zfsvfs_t *zfsvfs = arg1;
996 ace_t *ace = (ace_t *)(lr + 1);
1003 byteswap_uint64_array(lr, sizeof (*lr));
1004 zfs_ace_byteswap(ace, lr->lr_acl_bytes, B_FALSE);
1005 if (lr->lr_fuidcnt) {
1006 byteswap_uint64_array((caddr_t)ace +
1007 ZIL_ACE_LENGTH(lr->lr_acl_bytes),
1008 lr->lr_fuidcnt * sizeof (uint64_t));
1012 if ((error = zfs_zget(zfsvfs, lr->lr_foid, &zp)) != 0)
1015 bzero(&vsa, sizeof (vsa));
1016 vsa.vsa_mask = VSA_ACE | VSA_ACECNT | VSA_ACE_ACLFLAGS;
1017 vsa.vsa_aclcnt = lr->lr_aclcnt;
1018 vsa.vsa_aclentp = ace;
1019 vsa.vsa_aclentsz = lr->lr_acl_bytes;
1020 vsa.vsa_aclflags = lr->lr_acl_flags;
1022 if (lr->lr_fuidcnt) {
1023 void *fuidstart = (caddr_t)ace +
1024 ZIL_ACE_LENGTH(lr->lr_acl_bytes);
1026 zfsvfs->z_fuid_replay =
1027 zfs_replay_fuids(fuidstart, &fuidstart,
1028 lr->lr_fuidcnt, lr->lr_domcnt, 0, 0);
1032 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
1033 error = zfs_setsecattr(vp, &vsa, 0, kcred, NULL);
1036 if (zfsvfs->z_fuid_replay)
1037 zfs_fuid_info_free(zfsvfs->z_fuid_replay);
1039 zfsvfs->z_fuid_replay = NULL;
1046 * Callback vectors for replaying records
1048 zil_replay_func_t *zfs_replay_vector[TX_MAX_TYPE] = {
1049 zfs_replay_error, /* 0 no such transaction type */
1050 zfs_replay_create, /* TX_CREATE */
1051 zfs_replay_create, /* TX_MKDIR */
1052 zfs_replay_create, /* TX_MKXATTR */
1053 zfs_replay_create, /* TX_SYMLINK */
1054 zfs_replay_remove, /* TX_REMOVE */
1055 zfs_replay_remove, /* TX_RMDIR */
1056 zfs_replay_link, /* TX_LINK */
1057 zfs_replay_rename, /* TX_RENAME */
1058 zfs_replay_write, /* TX_WRITE */
1059 zfs_replay_truncate, /* TX_TRUNCATE */
1060 zfs_replay_setattr, /* TX_SETATTR */
1061 zfs_replay_acl_v0, /* TX_ACL_V0 */
1062 zfs_replay_acl, /* TX_ACL */
1063 zfs_replay_create_acl, /* TX_CREATE_ACL */
1064 zfs_replay_create, /* TX_CREATE_ATTR */
1065 zfs_replay_create_acl, /* TX_CREATE_ACL_ATTR */
1066 zfs_replay_create_acl, /* TX_MKDIR_ACL */
1067 zfs_replay_create, /* TX_MKDIR_ATTR */
1068 zfs_replay_create_acl, /* TX_MKDIR_ACL_ATTR */
1069 zfs_replay_write2, /* TX_WRITE2 */