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 https://opensource.org/licenses/CDDL-1.0.
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) 2015, 2018 by Delphix. All rights reserved.
24 * Copyright (c) 2022 by Pawel Jakub Dawidek
28 #include <sys/types.h>
29 #include <sys/param.h>
30 #include <sys/sysmacros.h>
31 #include <sys/cmn_err.h>
33 #include <sys/thread.h>
36 #include <sys/zfs_znode.h>
37 #include <sys/zfs_dir.h>
39 #include <sys/zil_impl.h>
40 #include <sys/byteorder.h>
41 #include <sys/policy.h>
47 #include <sys/zfs_fuid.h>
48 #include <sys/dsl_dataset.h>
51 * These zfs_log_* functions must be called within a dmu tx, in one
52 * of 2 contexts depending on zilog->z_replay:
56 * We need to record the transaction so that if it is committed to
57 * the Intent Log then it can be replayed. An intent log transaction
58 * structure (itx_t) is allocated and all the information necessary to
59 * possibly replay the transaction is saved in it. The itx is then assigned
60 * a sequence number and inserted in the in-memory list anchored in the zilog.
64 * We need to mark the intent log record as replayed in the log header.
65 * This is done in the same transaction as the replay so that they
70 zfs_log_create_txtype(zil_create_t type, vsecattr_t *vsecp, vattr_t *vap)
72 int isxvattr = (vap->va_mask & ATTR_XVATTR);
75 if (vsecp == NULL && !isxvattr)
77 if (vsecp && isxvattr)
78 return (TX_CREATE_ACL_ATTR);
80 return (TX_CREATE_ACL);
82 return (TX_CREATE_ATTR);
84 if (vsecp == NULL && !isxvattr)
86 if (vsecp && isxvattr)
87 return (TX_MKDIR_ACL_ATTR);
89 return (TX_MKDIR_ACL);
91 return (TX_MKDIR_ATTR);
100 * build up the log data necessary for logging xvattr_t
101 * First lr_attr_t is initialized. following the lr_attr_t
102 * is the mapsize and attribute bitmap copied from the xvattr_t.
103 * Following the bitmap and bitmapsize two 64 bit words are reserved
104 * for the create time which may be set. Following the create time
105 * records a single 64 bit integer which has the bits to set on
106 * replay for the xvattr.
109 zfs_log_xvattr(lr_attr_t *lrattr, xvattr_t *xvap)
113 xoap = xva_getxoptattr(xvap);
116 lrattr->lr_attr_masksize = xvap->xva_mapsize;
117 uint32_t *bitmap = &lrattr->lr_attr_bitmap;
118 for (int i = 0; i != xvap->xva_mapsize; i++, bitmap++)
119 *bitmap = xvap->xva_reqattrmap[i];
121 lr_attr_end_t *end = (lr_attr_end_t *)bitmap;
122 end->lr_attr_attrs = 0;
123 end->lr_attr_crtime[0] = 0;
124 end->lr_attr_crtime[1] = 0;
125 memset(end->lr_attr_scanstamp, 0, AV_SCANSTAMP_SZ);
127 if (XVA_ISSET_REQ(xvap, XAT_READONLY))
128 end->lr_attr_attrs |= (xoap->xoa_readonly == 0) ? 0 :
130 if (XVA_ISSET_REQ(xvap, XAT_HIDDEN))
131 end->lr_attr_attrs |= (xoap->xoa_hidden == 0) ? 0 :
133 if (XVA_ISSET_REQ(xvap, XAT_SYSTEM))
134 end->lr_attr_attrs |= (xoap->xoa_system == 0) ? 0 :
136 if (XVA_ISSET_REQ(xvap, XAT_ARCHIVE))
137 end->lr_attr_attrs |= (xoap->xoa_archive == 0) ? 0 :
139 if (XVA_ISSET_REQ(xvap, XAT_IMMUTABLE))
140 end->lr_attr_attrs |= (xoap->xoa_immutable == 0) ? 0 :
142 if (XVA_ISSET_REQ(xvap, XAT_NOUNLINK))
143 end->lr_attr_attrs |= (xoap->xoa_nounlink == 0) ? 0 :
145 if (XVA_ISSET_REQ(xvap, XAT_APPENDONLY))
146 end->lr_attr_attrs |= (xoap->xoa_appendonly == 0) ? 0 :
148 if (XVA_ISSET_REQ(xvap, XAT_OPAQUE))
149 end->lr_attr_attrs |= (xoap->xoa_opaque == 0) ? 0 :
151 if (XVA_ISSET_REQ(xvap, XAT_NODUMP))
152 end->lr_attr_attrs |= (xoap->xoa_nodump == 0) ? 0 :
154 if (XVA_ISSET_REQ(xvap, XAT_AV_QUARANTINED))
155 end->lr_attr_attrs |= (xoap->xoa_av_quarantined == 0) ? 0 :
157 if (XVA_ISSET_REQ(xvap, XAT_AV_MODIFIED))
158 end->lr_attr_attrs |= (xoap->xoa_av_modified == 0) ? 0 :
160 if (XVA_ISSET_REQ(xvap, XAT_CREATETIME))
161 ZFS_TIME_ENCODE(&xoap->xoa_createtime, end->lr_attr_crtime);
162 if (XVA_ISSET_REQ(xvap, XAT_AV_SCANSTAMP)) {
163 ASSERT(!XVA_ISSET_REQ(xvap, XAT_PROJID));
165 memcpy(end->lr_attr_scanstamp, xoap->xoa_av_scanstamp,
167 } else if (XVA_ISSET_REQ(xvap, XAT_PROJID)) {
169 * XAT_PROJID and XAT_AV_SCANSTAMP will never be valid
170 * at the same time, so we can share the same space.
172 memcpy(end->lr_attr_scanstamp, &xoap->xoa_projid,
175 if (XVA_ISSET_REQ(xvap, XAT_REPARSE))
176 end->lr_attr_attrs |= (xoap->xoa_reparse == 0) ? 0 :
178 if (XVA_ISSET_REQ(xvap, XAT_OFFLINE))
179 end->lr_attr_attrs |= (xoap->xoa_offline == 0) ? 0 :
181 if (XVA_ISSET_REQ(xvap, XAT_SPARSE))
182 end->lr_attr_attrs |= (xoap->xoa_sparse == 0) ? 0 :
184 if (XVA_ISSET_REQ(xvap, XAT_PROJINHERIT))
185 end->lr_attr_attrs |= (xoap->xoa_projinherit == 0) ? 0 :
190 zfs_log_fuid_ids(zfs_fuid_info_t *fuidp, void *start)
193 uint64_t *fuidloc = start;
195 /* First copy in the ACE FUIDs */
196 for (zfuid = list_head(&fuidp->z_fuids); zfuid;
197 zfuid = list_next(&fuidp->z_fuids, zfuid)) {
198 *fuidloc++ = zfuid->z_logfuid;
205 zfs_log_fuid_domains(zfs_fuid_info_t *fuidp, void *start)
207 zfs_fuid_domain_t *zdomain;
209 /* now copy in the domain info, if any */
210 if (fuidp->z_domain_str_sz != 0) {
211 for (zdomain = list_head(&fuidp->z_domains); zdomain;
212 zdomain = list_next(&fuidp->z_domains, zdomain)) {
213 memcpy(start, zdomain->z_domain,
214 strlen(zdomain->z_domain) + 1);
215 start = (caddr_t)start +
216 strlen(zdomain->z_domain) + 1;
223 * If zp is an xattr node, check whether the xattr owner is unlinked.
224 * We don't want to log anything if the owner is unlinked.
227 zfs_xattr_owner_unlinked(znode_t *zp)
235 * zrele drops the vnode lock which violates the VOP locking contract
236 * on FreeBSD. See comment at the top of zfs_replay.c for more detail.
239 * if zp is XATTR node, keep walking up via z_xattr_parent until we
242 while (tzp->z_pflags & ZFS_XATTR) {
243 ASSERT3U(zp->z_xattr_parent, !=, 0);
244 if (zfs_zget(ZTOZSB(tzp), tzp->z_xattr_parent, &dzp) != 0) {
252 unlinked = tzp->z_unlinked;
259 * if zp is XATTR node, keep walking up via z_xattr_parent until we
262 while (zp->z_pflags & ZFS_XATTR) {
263 ASSERT3U(zp->z_xattr_parent, !=, 0);
264 if (zfs_zget(ZTOZSB(zp), zp->z_xattr_parent, &dzp) != 0) {
271 unlinked = zp->z_unlinked;
279 * Handles TX_CREATE, TX_CREATE_ATTR, TX_MKDIR, TX_MKDIR_ATTR and
280 * TK_MKXATTR transactions.
282 * TX_CREATE and TX_MKDIR are standard creates, but they may have FUID
283 * domain information appended prior to the name. In this case the
284 * uid/gid in the log record will be a log centric FUID.
286 * TX_CREATE_ACL_ATTR and TX_MKDIR_ACL_ATTR handle special creates that
287 * may contain attributes, ACL and optional fuid information.
289 * TX_CREATE_ACL and TX_MKDIR_ACL handle special creates that specify
290 * and ACL and normal users/groups in the ACEs.
292 * There may be an optional xvattr attribute information similar
293 * to zfs_log_setattr.
295 * Also, after the file name "domain" strings may be appended.
298 zfs_log_create(zilog_t *zilog, dmu_tx_t *tx, uint64_t txtype,
299 znode_t *dzp, znode_t *zp, const char *name, vsecattr_t *vsecp,
300 zfs_fuid_info_t *fuidp, vattr_t *vap)
304 lr_acl_create_t *lracl;
308 xvattr_t *xvap = (xvattr_t *)vap;
311 size_t namesize = strlen(name) + 1;
314 if (zil_replaying(zilog, tx) || zfs_xattr_owner_unlinked(dzp))
318 * If we have FUIDs present then add in space for
319 * domains and ACE fuid's if any.
322 fuidsz += fuidp->z_domain_str_sz;
323 fuidsz += fuidp->z_fuid_cnt * sizeof (uint64_t);
326 if (vap->va_mask & ATTR_XVATTR)
327 xvatsize = ZIL_XVAT_SIZE(xvap->xva_mapsize);
329 if ((int)txtype == TX_CREATE_ATTR || (int)txtype == TX_MKDIR_ATTR ||
330 (int)txtype == TX_CREATE || (int)txtype == TX_MKDIR ||
331 (int)txtype == TX_MKXATTR) {
332 txsize = sizeof (*lr) + namesize + fuidsz + xvatsize;
333 lrsize = sizeof (*lr);
336 sizeof (lr_acl_create_t) + namesize + fuidsz +
337 ZIL_ACE_LENGTH(aclsize) + xvatsize;
338 lrsize = sizeof (lr_acl_create_t);
341 itx = zil_itx_create(txtype, txsize);
343 lr = (lr_create_t *)&itx->itx_lr;
344 lr->lr_doid = dzp->z_id;
345 lr->lr_foid = zp->z_id;
346 /* Store dnode slot count in 8 bits above object id. */
347 LR_FOID_SET_SLOTS(lr->lr_foid, zp->z_dnodesize >> DNODE_SHIFT);
348 lr->lr_mode = zp->z_mode;
349 if (!IS_EPHEMERAL(KUID_TO_SUID(ZTOUID(zp)))) {
350 lr->lr_uid = (uint64_t)KUID_TO_SUID(ZTOUID(zp));
352 lr->lr_uid = fuidp->z_fuid_owner;
354 if (!IS_EPHEMERAL(KGID_TO_SGID(ZTOGID(zp)))) {
355 lr->lr_gid = (uint64_t)KGID_TO_SGID(ZTOGID(zp));
357 lr->lr_gid = fuidp->z_fuid_group;
359 (void) sa_lookup(zp->z_sa_hdl, SA_ZPL_GEN(ZTOZSB(zp)), &lr->lr_gen,
361 (void) sa_lookup(zp->z_sa_hdl, SA_ZPL_CRTIME(ZTOZSB(zp)),
362 lr->lr_crtime, sizeof (uint64_t) * 2);
364 if (sa_lookup(zp->z_sa_hdl, SA_ZPL_RDEV(ZTOZSB(zp)), &lr->lr_rdev,
365 sizeof (lr->lr_rdev)) != 0)
369 * Fill in xvattr info if any
371 if (vap->va_mask & ATTR_XVATTR) {
372 zfs_log_xvattr((lr_attr_t *)((caddr_t)lr + lrsize), xvap);
373 end = (caddr_t)lr + lrsize + xvatsize;
375 end = (caddr_t)lr + lrsize;
378 /* Now fill in any ACL info */
381 lracl = (lr_acl_create_t *)&itx->itx_lr;
382 lracl->lr_aclcnt = vsecp->vsa_aclcnt;
383 lracl->lr_acl_bytes = aclsize;
384 lracl->lr_domcnt = fuidp ? fuidp->z_domain_cnt : 0;
385 lracl->lr_fuidcnt = fuidp ? fuidp->z_fuid_cnt : 0;
386 if (vsecp->vsa_aclflags & VSA_ACE_ACLFLAGS)
387 lracl->lr_acl_flags = (uint64_t)vsecp->vsa_aclflags;
389 lracl->lr_acl_flags = 0;
391 memcpy(end, vsecp->vsa_aclentp, aclsize);
392 end = (caddr_t)end + ZIL_ACE_LENGTH(aclsize);
395 /* drop in FUID info */
397 end = zfs_log_fuid_ids(fuidp, end);
398 end = zfs_log_fuid_domains(fuidp, end);
401 * Now place file name in log record
403 memcpy(end, name, namesize);
405 zil_itx_assign(zilog, itx, tx);
409 * Handles both TX_REMOVE and TX_RMDIR transactions.
412 zfs_log_remove(zilog_t *zilog, dmu_tx_t *tx, uint64_t txtype,
413 znode_t *dzp, const char *name, uint64_t foid, boolean_t unlinked)
417 size_t namesize = strlen(name) + 1;
419 if (zil_replaying(zilog, tx) || zfs_xattr_owner_unlinked(dzp))
422 itx = zil_itx_create(txtype, sizeof (*lr) + namesize);
423 lr = (lr_remove_t *)&itx->itx_lr;
424 lr->lr_doid = dzp->z_id;
425 memcpy(lr + 1, name, namesize);
430 * Object ids can be re-instantiated in the next txg so
431 * remove any async transactions to avoid future leaks.
432 * This can happen if a fsync occurs on the re-instantiated
433 * object for a WR_INDIRECT or WR_NEED_COPY write, which gets
434 * the new file data and flushes a write record for the old object.
437 ASSERT((txtype & ~TX_CI) == TX_REMOVE);
438 zil_remove_async(zilog, foid);
440 zil_itx_assign(zilog, itx, tx);
444 * Handles TX_LINK transactions.
447 zfs_log_link(zilog_t *zilog, dmu_tx_t *tx, uint64_t txtype,
448 znode_t *dzp, znode_t *zp, const char *name)
452 size_t namesize = strlen(name) + 1;
454 if (zil_replaying(zilog, tx))
457 itx = zil_itx_create(txtype, sizeof (*lr) + namesize);
458 lr = (lr_link_t *)&itx->itx_lr;
459 lr->lr_doid = dzp->z_id;
460 lr->lr_link_obj = zp->z_id;
461 memcpy(lr + 1, name, namesize);
463 zil_itx_assign(zilog, itx, tx);
467 * Handles TX_SYMLINK transactions.
470 zfs_log_symlink(zilog_t *zilog, dmu_tx_t *tx, uint64_t txtype,
471 znode_t *dzp, znode_t *zp, const char *name, const char *link)
475 size_t namesize = strlen(name) + 1;
476 size_t linksize = strlen(link) + 1;
478 if (zil_replaying(zilog, tx))
481 itx = zil_itx_create(txtype, sizeof (*lr) + namesize + linksize);
482 lr = (lr_create_t *)&itx->itx_lr;
483 lr->lr_doid = dzp->z_id;
484 lr->lr_foid = zp->z_id;
485 lr->lr_uid = KUID_TO_SUID(ZTOUID(zp));
486 lr->lr_gid = KGID_TO_SGID(ZTOGID(zp));
487 lr->lr_mode = zp->z_mode;
488 (void) sa_lookup(zp->z_sa_hdl, SA_ZPL_GEN(ZTOZSB(zp)), &lr->lr_gen,
490 (void) sa_lookup(zp->z_sa_hdl, SA_ZPL_CRTIME(ZTOZSB(zp)),
491 lr->lr_crtime, sizeof (uint64_t) * 2);
492 memcpy((char *)(lr + 1), name, namesize);
493 memcpy((char *)(lr + 1) + namesize, link, linksize);
495 zil_itx_assign(zilog, itx, tx);
499 do_zfs_log_rename(zilog_t *zilog, dmu_tx_t *tx, uint64_t txtype, znode_t *sdzp,
500 const char *sname, znode_t *tdzp, const char *dname, znode_t *szp)
504 size_t snamesize = strlen(sname) + 1;
505 size_t dnamesize = strlen(dname) + 1;
507 if (zil_replaying(zilog, tx))
510 itx = zil_itx_create(txtype, sizeof (*lr) + snamesize + dnamesize);
511 lr = (lr_rename_t *)&itx->itx_lr;
512 lr->lr_sdoid = sdzp->z_id;
513 lr->lr_tdoid = tdzp->z_id;
514 memcpy((char *)(lr + 1), sname, snamesize);
515 memcpy((char *)(lr + 1) + snamesize, dname, dnamesize);
516 itx->itx_oid = szp->z_id;
518 zil_itx_assign(zilog, itx, tx);
522 * Handles TX_RENAME transactions.
525 zfs_log_rename(zilog_t *zilog, dmu_tx_t *tx, uint64_t txtype, znode_t *sdzp,
526 const char *sname, znode_t *tdzp, const char *dname, znode_t *szp)
529 do_zfs_log_rename(zilog, tx, txtype, sdzp, sname, tdzp, dname, szp);
533 * Handles TX_RENAME_EXCHANGE transactions.
536 zfs_log_rename_exchange(zilog_t *zilog, dmu_tx_t *tx, uint64_t txtype,
537 znode_t *sdzp, const char *sname, znode_t *tdzp, const char *dname,
540 txtype |= TX_RENAME_EXCHANGE;
541 do_zfs_log_rename(zilog, tx, txtype, sdzp, sname, tdzp, dname, szp);
545 * Handles TX_RENAME_WHITEOUT transactions.
547 * Unfortunately we cannot reuse do_zfs_log_rename because we we need to call
548 * zfs_mknode() on replay which requires stashing bits as with TX_CREATE.
551 zfs_log_rename_whiteout(zilog_t *zilog, dmu_tx_t *tx, uint64_t txtype,
552 znode_t *sdzp, const char *sname, znode_t *tdzp, const char *dname,
553 znode_t *szp, znode_t *wzp)
556 lr_rename_whiteout_t *lr;
557 size_t snamesize = strlen(sname) + 1;
558 size_t dnamesize = strlen(dname) + 1;
560 if (zil_replaying(zilog, tx))
563 txtype |= TX_RENAME_WHITEOUT;
564 itx = zil_itx_create(txtype, sizeof (*lr) + snamesize + dnamesize);
565 lr = (lr_rename_whiteout_t *)&itx->itx_lr;
566 lr->lr_rename.lr_sdoid = sdzp->z_id;
567 lr->lr_rename.lr_tdoid = tdzp->z_id;
570 * RENAME_WHITEOUT will create an entry at the source znode, so we need
571 * to store the same data that the equivalent call to zfs_log_create()
574 lr->lr_wfoid = wzp->z_id;
575 LR_FOID_SET_SLOTS(lr->lr_wfoid, wzp->z_dnodesize >> DNODE_SHIFT);
576 (void) sa_lookup(wzp->z_sa_hdl, SA_ZPL_GEN(ZTOZSB(wzp)), &lr->lr_wgen,
578 (void) sa_lookup(wzp->z_sa_hdl, SA_ZPL_CRTIME(ZTOZSB(wzp)),
579 lr->lr_wcrtime, sizeof (uint64_t) * 2);
580 lr->lr_wmode = wzp->z_mode;
581 lr->lr_wuid = (uint64_t)KUID_TO_SUID(ZTOUID(wzp));
582 lr->lr_wgid = (uint64_t)KGID_TO_SGID(ZTOGID(wzp));
585 * This rdev will always be makdevice(0, 0) but because the ZIL log and
586 * replay code needs to be platform independent (and there is no
587 * platform independent makdev()) we need to copy the one created
588 * during the rename operation.
590 (void) sa_lookup(wzp->z_sa_hdl, SA_ZPL_RDEV(ZTOZSB(wzp)), &lr->lr_wrdev,
591 sizeof (lr->lr_wrdev));
593 memcpy((char *)(lr + 1), sname, snamesize);
594 memcpy((char *)(lr + 1) + snamesize, dname, dnamesize);
595 itx->itx_oid = szp->z_id;
597 zil_itx_assign(zilog, itx, tx);
601 * zfs_log_write() handles TX_WRITE transactions. The specified callback is
602 * called as soon as the write is on stable storage (be it via a DMU sync or a
605 static int64_t zfs_immediate_write_sz = 32768;
608 zfs_log_write(zilog_t *zilog, dmu_tx_t *tx, int txtype,
609 znode_t *zp, offset_t off, ssize_t resid, int ioflag,
610 zil_callback_t callback, void *callback_data)
612 dmu_buf_impl_t *db = (dmu_buf_impl_t *)sa_get_db(zp->z_sa_hdl);
613 uint32_t blocksize = zp->z_blksz;
614 itx_wr_state_t write_state;
617 ssize_t size = resid;
619 if (zil_replaying(zilog, tx) || zp->z_unlinked ||
620 zfs_xattr_owner_unlinked(zp)) {
621 if (callback != NULL)
622 callback(callback_data);
626 if (zilog->zl_logbias == ZFS_LOGBIAS_THROUGHPUT)
627 write_state = WR_INDIRECT;
628 else if (!spa_has_slogs(zilog->zl_spa) &&
629 resid >= zfs_immediate_write_sz)
630 write_state = WR_INDIRECT;
631 else if (ioflag & (O_SYNC | O_DSYNC))
632 write_state = WR_COPIED;
634 write_state = WR_NEED_COPY;
636 if ((fsync_cnt = (uintptr_t)tsd_get(zfs_fsyncer_key)) != 0) {
637 (void) tsd_set(zfs_fsyncer_key, (void *)(fsync_cnt - 1));
640 (void) sa_lookup(zp->z_sa_hdl, SA_ZPL_GEN(ZTOZSB(zp)), &gen,
646 itx_wr_state_t wr_state = write_state;
650 * A WR_COPIED record must fit entirely in one log block.
651 * Large writes can use WR_NEED_COPY, which the ZIL will
652 * split into multiple records across several log blocks
655 if (wr_state == WR_COPIED &&
656 resid > zil_max_copied_data(zilog))
657 wr_state = WR_NEED_COPY;
658 else if (wr_state == WR_INDIRECT)
659 len = MIN(blocksize - P2PHASE(off, blocksize), resid);
661 itx = zil_itx_create(txtype, sizeof (*lr) +
662 (wr_state == WR_COPIED ? len : 0));
663 lr = (lr_write_t *)&itx->itx_lr;
666 * For WR_COPIED records, copy the data into the lr_write_t.
668 if (wr_state == WR_COPIED) {
671 err = dmu_read_by_dnode(DB_DNODE(db), off, len, lr + 1,
672 DMU_READ_NO_PREFETCH);
674 zil_itx_destroy(itx);
675 itx = zil_itx_create(txtype, sizeof (*lr));
676 lr = (lr_write_t *)&itx->itx_lr;
677 wr_state = WR_NEED_COPY;
682 itx->itx_wr_state = wr_state;
683 lr->lr_foid = zp->z_id;
687 BP_ZERO(&lr->lr_blkptr);
689 itx->itx_private = ZTOZSB(zp);
692 if (!(ioflag & (O_SYNC | O_DSYNC)) && (zp->z_sync_cnt == 0) &&
694 itx->itx_sync = B_FALSE;
696 itx->itx_callback = callback;
697 itx->itx_callback_data = callback_data;
698 zil_itx_assign(zilog, itx, tx);
704 if (write_state == WR_COPIED || write_state == WR_NEED_COPY) {
705 dsl_pool_wrlog_count(zilog->zl_dmu_pool, size, tx->tx_txg);
710 * Handles TX_TRUNCATE transactions.
713 zfs_log_truncate(zilog_t *zilog, dmu_tx_t *tx, int txtype,
714 znode_t *zp, uint64_t off, uint64_t len)
719 if (zil_replaying(zilog, tx) || zp->z_unlinked ||
720 zfs_xattr_owner_unlinked(zp))
723 itx = zil_itx_create(txtype, sizeof (*lr));
724 lr = (lr_truncate_t *)&itx->itx_lr;
725 lr->lr_foid = zp->z_id;
729 itx->itx_sync = (zp->z_sync_cnt != 0);
730 zil_itx_assign(zilog, itx, tx);
734 * Handles TX_SETATTR transactions.
737 zfs_log_setattr(zilog_t *zilog, dmu_tx_t *tx, int txtype,
738 znode_t *zp, vattr_t *vap, uint_t mask_applied, zfs_fuid_info_t *fuidp)
742 xvattr_t *xvap = (xvattr_t *)vap;
743 size_t recsize = sizeof (lr_setattr_t);
746 if (zil_replaying(zilog, tx) || zp->z_unlinked)
750 * If XVATTR set, then log record size needs to allow
751 * for lr_attr_t + xvattr mask, mapsize and create time
752 * plus actual attribute values
754 if (vap->va_mask & ATTR_XVATTR)
755 recsize = sizeof (*lr) + ZIL_XVAT_SIZE(xvap->xva_mapsize);
758 recsize += fuidp->z_domain_str_sz;
760 itx = zil_itx_create(txtype, recsize);
761 lr = (lr_setattr_t *)&itx->itx_lr;
762 lr->lr_foid = zp->z_id;
763 lr->lr_mask = (uint64_t)mask_applied;
764 lr->lr_mode = (uint64_t)vap->va_mode;
765 if ((mask_applied & ATTR_UID) && IS_EPHEMERAL(vap->va_uid))
766 lr->lr_uid = fuidp->z_fuid_owner;
768 lr->lr_uid = (uint64_t)vap->va_uid;
770 if ((mask_applied & ATTR_GID) && IS_EPHEMERAL(vap->va_gid))
771 lr->lr_gid = fuidp->z_fuid_group;
773 lr->lr_gid = (uint64_t)vap->va_gid;
775 lr->lr_size = (uint64_t)vap->va_size;
776 ZFS_TIME_ENCODE(&vap->va_atime, lr->lr_atime);
777 ZFS_TIME_ENCODE(&vap->va_mtime, lr->lr_mtime);
778 start = (lr_setattr_t *)(lr + 1);
779 if (vap->va_mask & ATTR_XVATTR) {
780 zfs_log_xvattr((lr_attr_t *)start, xvap);
781 start = (caddr_t)start + ZIL_XVAT_SIZE(xvap->xva_mapsize);
785 * Now stick on domain information if any on end
789 (void) zfs_log_fuid_domains(fuidp, start);
791 itx->itx_sync = (zp->z_sync_cnt != 0);
792 zil_itx_assign(zilog, itx, tx);
796 * Handles TX_SETSAXATTR transactions.
799 zfs_log_setsaxattr(zilog_t *zilog, dmu_tx_t *tx, int txtype,
800 znode_t *zp, const char *name, const void *value, size_t size)
804 size_t recsize = sizeof (lr_setsaxattr_t);
808 if (zil_replaying(zilog, tx) || zp->z_unlinked)
811 namelen = strlen(name) + 1;
812 recsize += (namelen + size);
813 itx = zil_itx_create(txtype, recsize);
814 lr = (lr_setsaxattr_t *)&itx->itx_lr;
815 lr->lr_foid = zp->z_id;
816 xattrstart = (char *)(lr + 1);
817 memcpy(xattrstart, name, namelen);
819 memcpy((char *)xattrstart + namelen, value, size);
825 itx->itx_sync = (zp->z_sync_cnt != 0);
826 zil_itx_assign(zilog, itx, tx);
830 * Handles TX_ACL transactions.
833 zfs_log_acl(zilog_t *zilog, dmu_tx_t *tx, znode_t *zp,
834 vsecattr_t *vsecp, zfs_fuid_info_t *fuidp)
842 size_t aclbytes = vsecp->vsa_aclentsz;
844 if (zil_replaying(zilog, tx) || zp->z_unlinked)
847 txtype = (ZTOZSB(zp)->z_version < ZPL_VERSION_FUID) ?
850 if (txtype == TX_ACL)
851 lrsize = sizeof (*lr);
853 lrsize = sizeof (*lrv0);
856 ((txtype == TX_ACL) ? ZIL_ACE_LENGTH(aclbytes) : aclbytes) +
857 (fuidp ? fuidp->z_domain_str_sz : 0) +
858 sizeof (uint64_t) * (fuidp ? fuidp->z_fuid_cnt : 0);
860 itx = zil_itx_create(txtype, txsize);
862 lr = (lr_acl_t *)&itx->itx_lr;
863 lr->lr_foid = zp->z_id;
864 if (txtype == TX_ACL) {
865 lr->lr_acl_bytes = aclbytes;
866 lr->lr_domcnt = fuidp ? fuidp->z_domain_cnt : 0;
867 lr->lr_fuidcnt = fuidp ? fuidp->z_fuid_cnt : 0;
868 if (vsecp->vsa_mask & VSA_ACE_ACLFLAGS)
869 lr->lr_acl_flags = (uint64_t)vsecp->vsa_aclflags;
871 lr->lr_acl_flags = 0;
873 lr->lr_aclcnt = (uint64_t)vsecp->vsa_aclcnt;
875 if (txtype == TX_ACL_V0) {
876 lrv0 = (lr_acl_v0_t *)lr;
877 memcpy(lrv0 + 1, vsecp->vsa_aclentp, aclbytes);
879 void *start = (ace_t *)(lr + 1);
881 memcpy(start, vsecp->vsa_aclentp, aclbytes);
883 start = (caddr_t)start + ZIL_ACE_LENGTH(aclbytes);
886 start = zfs_log_fuid_ids(fuidp, start);
887 (void) zfs_log_fuid_domains(fuidp, start);
891 itx->itx_sync = (zp->z_sync_cnt != 0);
892 zil_itx_assign(zilog, itx, tx);
896 * Handles TX_CLONE_RANGE transactions.
899 zfs_log_clone_range(zilog_t *zilog, dmu_tx_t *tx, int txtype, znode_t *zp,
900 uint64_t off, uint64_t len, uint64_t blksz, const blkptr_t *bps,
904 lr_clone_range_t *lr;
905 uint64_t partlen, max_log_data;
908 if (zil_replaying(zilog, tx) || zp->z_unlinked)
911 max_log_data = zil_max_log_data(zilog, sizeof (lr_clone_range_t));
914 partnbps = MIN(nbps, max_log_data / sizeof (bps[0]));
916 for (i = 0; i < partnbps; i++) {
917 partlen += BP_GET_LSIZE(&bps[i]);
919 partlen = MIN(partlen, len);
921 itx = zil_itx_create(txtype,
922 sizeof (*lr) + sizeof (bps[0]) * partnbps);
923 lr = (lr_clone_range_t *)&itx->itx_lr;
924 lr->lr_foid = zp->z_id;
926 lr->lr_length = partlen;
927 lr->lr_blksz = blksz;
928 lr->lr_nbps = partnbps;
929 memcpy(lr->lr_bps, bps, sizeof (bps[0]) * partnbps);
931 itx->itx_sync = (zp->z_sync_cnt != 0);
933 zil_itx_assign(zilog, itx, tx);
936 ASSERT3U(nbps, >=, partnbps);
939 ASSERT3U(len, >=, partlen);
944 ZFS_MODULE_PARAM(zfs, zfs_, immediate_write_sz, S64, ZMOD_RW,
945 "Largest data block to write to zil");