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.
27 #include <sys/types.h>
28 #include <sys/param.h>
29 #include <sys/systm.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>
49 #include <sys/dsl_dataset.h>
52 * These zfs_log_* functions must be called within a dmu tx, in one
53 * of 2 contexts depending on zilog->z_replay:
57 * We need to record the transaction so that if it is committed to
58 * the Intent Log then it can be replayed. An intent log transaction
59 * structure (itx_t) is allocated and all the information necessary to
60 * possibly replay the transaction is saved in it. The itx is then assigned
61 * a sequence number and inserted in the in-memory list anchored in the zilog.
65 * We need to mark the intent log record as replayed in the log header.
66 * This is done in the same transaction as the replay so that they
71 zfs_log_create_txtype(zil_create_t type, vsecattr_t *vsecp, vattr_t *vap)
73 int isxvattr = (vap->va_mask & AT_XVATTR);
76 if (vsecp == NULL && !isxvattr)
78 if (vsecp && isxvattr)
79 return (TX_CREATE_ACL_ATTR);
81 return (TX_CREATE_ACL);
83 return (TX_CREATE_ATTR);
86 if (vsecp == NULL && !isxvattr)
88 if (vsecp && isxvattr)
89 return (TX_MKDIR_ACL_ATTR);
91 return (TX_MKDIR_ACL);
93 return (TX_MKDIR_ATTR);
102 * build up the log data necessary for logging xvattr_t
103 * First lr_attr_t is initialized. following the lr_attr_t
104 * is the mapsize and attribute bitmap copied from the xvattr_t.
105 * Following the bitmap and bitmapsize two 64 bit words are reserved
106 * for the create time which may be set. Following the create time
107 * records a single 64 bit integer which has the bits to set on
108 * replay for the xvattr.
111 zfs_log_xvattr(lr_attr_t *lrattr, xvattr_t *xvap)
120 xoap = xva_getxoptattr(xvap);
123 lrattr->lr_attr_masksize = xvap->xva_mapsize;
124 bitmap = &lrattr->lr_attr_bitmap;
125 for (i = 0; i != xvap->xva_mapsize; i++, bitmap++) {
126 *bitmap = xvap->xva_reqattrmap[i];
129 /* Now pack the attributes up in a single uint64_t */
130 attrs = (uint64_t *)bitmap;
132 scanstamp = (caddr_t)(crtime + 2);
134 if (XVA_ISSET_REQ(xvap, XAT_READONLY))
135 *attrs |= (xoap->xoa_readonly == 0) ? 0 :
137 if (XVA_ISSET_REQ(xvap, XAT_HIDDEN))
138 *attrs |= (xoap->xoa_hidden == 0) ? 0 :
140 if (XVA_ISSET_REQ(xvap, XAT_SYSTEM))
141 *attrs |= (xoap->xoa_system == 0) ? 0 :
143 if (XVA_ISSET_REQ(xvap, XAT_ARCHIVE))
144 *attrs |= (xoap->xoa_archive == 0) ? 0 :
146 if (XVA_ISSET_REQ(xvap, XAT_IMMUTABLE))
147 *attrs |= (xoap->xoa_immutable == 0) ? 0 :
149 if (XVA_ISSET_REQ(xvap, XAT_NOUNLINK))
150 *attrs |= (xoap->xoa_nounlink == 0) ? 0 :
152 if (XVA_ISSET_REQ(xvap, XAT_APPENDONLY))
153 *attrs |= (xoap->xoa_appendonly == 0) ? 0 :
155 if (XVA_ISSET_REQ(xvap, XAT_OPAQUE))
156 *attrs |= (xoap->xoa_opaque == 0) ? 0 :
158 if (XVA_ISSET_REQ(xvap, XAT_NODUMP))
159 *attrs |= (xoap->xoa_nodump == 0) ? 0 :
161 if (XVA_ISSET_REQ(xvap, XAT_AV_QUARANTINED))
162 *attrs |= (xoap->xoa_av_quarantined == 0) ? 0 :
164 if (XVA_ISSET_REQ(xvap, XAT_AV_MODIFIED))
165 *attrs |= (xoap->xoa_av_modified == 0) ? 0 :
167 if (XVA_ISSET_REQ(xvap, XAT_CREATETIME))
168 ZFS_TIME_ENCODE(&xoap->xoa_createtime, crtime);
169 if (XVA_ISSET_REQ(xvap, XAT_AV_SCANSTAMP))
170 bcopy(xoap->xoa_av_scanstamp, scanstamp, AV_SCANSTAMP_SZ);
171 if (XVA_ISSET_REQ(xvap, XAT_REPARSE))
172 *attrs |= (xoap->xoa_reparse == 0) ? 0 :
174 if (XVA_ISSET_REQ(xvap, XAT_OFFLINE))
175 *attrs |= (xoap->xoa_offline == 0) ? 0 :
177 if (XVA_ISSET_REQ(xvap, XAT_SPARSE))
178 *attrs |= (xoap->xoa_sparse == 0) ? 0 :
183 zfs_log_fuid_ids(zfs_fuid_info_t *fuidp, void *start)
186 uint64_t *fuidloc = start;
188 /* First copy in the ACE FUIDs */
189 for (zfuid = list_head(&fuidp->z_fuids); zfuid;
190 zfuid = list_next(&fuidp->z_fuids, zfuid)) {
191 *fuidloc++ = zfuid->z_logfuid;
198 zfs_log_fuid_domains(zfs_fuid_info_t *fuidp, void *start)
200 zfs_fuid_domain_t *zdomain;
202 /* now copy in the domain info, if any */
203 if (fuidp->z_domain_str_sz != 0) {
204 for (zdomain = list_head(&fuidp->z_domains); zdomain;
205 zdomain = list_next(&fuidp->z_domains, zdomain)) {
206 bcopy((void *)zdomain->z_domain, start,
207 strlen(zdomain->z_domain) + 1);
208 start = (caddr_t)start +
209 strlen(zdomain->z_domain) + 1;
216 * zfs_log_create() is used to handle TX_CREATE, TX_CREATE_ATTR, TX_MKDIR,
217 * TX_MKDIR_ATTR and TX_MKXATTR
220 * TX_CREATE and TX_MKDIR are standard creates, but they may have FUID
221 * domain information appended prior to the name. In this case the
222 * uid/gid in the log record will be a log centric FUID.
224 * TX_CREATE_ACL_ATTR and TX_MKDIR_ACL_ATTR handle special creates that
225 * may contain attributes, ACL and optional fuid information.
227 * TX_CREATE_ACL and TX_MKDIR_ACL handle special creates that specify
228 * and ACL and normal users/groups in the ACEs.
230 * There may be an optional xvattr attribute information similar
231 * to zfs_log_setattr.
233 * Also, after the file name "domain" strings may be appended.
236 zfs_log_create(zilog_t *zilog, dmu_tx_t *tx, uint64_t txtype,
237 znode_t *dzp, znode_t *zp, char *name, vsecattr_t *vsecp,
238 zfs_fuid_info_t *fuidp, vattr_t *vap)
242 lr_acl_create_t *lracl;
246 xvattr_t *xvap = (xvattr_t *)vap;
249 size_t namesize = strlen(name) + 1;
252 if (zil_replaying(zilog, tx))
256 * If we have FUIDs present then add in space for
257 * domains and ACE fuid's if any.
260 fuidsz += fuidp->z_domain_str_sz;
261 fuidsz += fuidp->z_fuid_cnt * sizeof (uint64_t);
264 if (vap->va_mask & AT_XVATTR)
265 xvatsize = ZIL_XVAT_SIZE(xvap->xva_mapsize);
267 if ((int)txtype == TX_CREATE_ATTR || (int)txtype == TX_MKDIR_ATTR ||
268 (int)txtype == TX_CREATE || (int)txtype == TX_MKDIR ||
269 (int)txtype == TX_MKXATTR) {
270 txsize = sizeof (*lr) + namesize + fuidsz + xvatsize;
271 lrsize = sizeof (*lr);
273 aclsize = (vsecp) ? vsecp->vsa_aclentsz : 0;
275 sizeof (lr_acl_create_t) + namesize + fuidsz +
276 ZIL_ACE_LENGTH(aclsize) + xvatsize;
277 lrsize = sizeof (lr_acl_create_t);
280 itx = zil_itx_create(txtype, txsize);
282 lr = (lr_create_t *)&itx->itx_lr;
283 lr->lr_doid = dzp->z_id;
284 lr->lr_foid = zp->z_id;
285 lr->lr_mode = zp->z_mode;
286 if (!IS_EPHEMERAL(zp->z_uid)) {
287 lr->lr_uid = (uint64_t)zp->z_uid;
289 lr->lr_uid = fuidp->z_fuid_owner;
291 if (!IS_EPHEMERAL(zp->z_gid)) {
292 lr->lr_gid = (uint64_t)zp->z_gid;
294 lr->lr_gid = fuidp->z_fuid_group;
296 (void) sa_lookup(zp->z_sa_hdl, SA_ZPL_GEN(zp->z_zfsvfs), &lr->lr_gen,
298 (void) sa_lookup(zp->z_sa_hdl, SA_ZPL_CRTIME(zp->z_zfsvfs),
299 lr->lr_crtime, sizeof (uint64_t) * 2);
301 if (sa_lookup(zp->z_sa_hdl, SA_ZPL_RDEV(zp->z_zfsvfs), &lr->lr_rdev,
302 sizeof (lr->lr_rdev)) != 0)
306 * Fill in xvattr info if any
308 if (vap->va_mask & AT_XVATTR) {
309 zfs_log_xvattr((lr_attr_t *)((caddr_t)lr + lrsize), xvap);
310 end = (caddr_t)lr + lrsize + xvatsize;
312 end = (caddr_t)lr + lrsize;
315 /* Now fill in any ACL info */
318 lracl = (lr_acl_create_t *)&itx->itx_lr;
319 lracl->lr_aclcnt = vsecp->vsa_aclcnt;
320 lracl->lr_acl_bytes = aclsize;
321 lracl->lr_domcnt = fuidp ? fuidp->z_domain_cnt : 0;
322 lracl->lr_fuidcnt = fuidp ? fuidp->z_fuid_cnt : 0;
323 if (vsecp->vsa_aclflags & VSA_ACE_ACLFLAGS)
324 lracl->lr_acl_flags = (uint64_t)vsecp->vsa_aclflags;
326 lracl->lr_acl_flags = 0;
328 bcopy(vsecp->vsa_aclentp, end, aclsize);
329 end = (caddr_t)end + ZIL_ACE_LENGTH(aclsize);
332 /* drop in FUID info */
334 end = zfs_log_fuid_ids(fuidp, end);
335 end = zfs_log_fuid_domains(fuidp, end);
338 * Now place file name in log record
340 bcopy(name, end, namesize);
342 zil_itx_assign(zilog, itx, tx);
346 * zfs_log_remove() handles both TX_REMOVE and TX_RMDIR transactions.
349 zfs_log_remove(zilog_t *zilog, dmu_tx_t *tx, uint64_t txtype,
350 znode_t *dzp, char *name, uint64_t foid)
354 size_t namesize = strlen(name) + 1;
356 if (zil_replaying(zilog, tx))
359 itx = zil_itx_create(txtype, sizeof (*lr) + namesize);
360 lr = (lr_remove_t *)&itx->itx_lr;
361 lr->lr_doid = dzp->z_id;
362 bcopy(name, (char *)(lr + 1), namesize);
366 zil_itx_assign(zilog, itx, tx);
370 * zfs_log_link() handles TX_LINK transactions.
373 zfs_log_link(zilog_t *zilog, dmu_tx_t *tx, uint64_t txtype,
374 znode_t *dzp, znode_t *zp, char *name)
378 size_t namesize = strlen(name) + 1;
380 if (zil_replaying(zilog, tx))
383 itx = zil_itx_create(txtype, sizeof (*lr) + namesize);
384 lr = (lr_link_t *)&itx->itx_lr;
385 lr->lr_doid = dzp->z_id;
386 lr->lr_link_obj = zp->z_id;
387 bcopy(name, (char *)(lr + 1), namesize);
389 zil_itx_assign(zilog, itx, tx);
393 * zfs_log_symlink() handles TX_SYMLINK transactions.
396 zfs_log_symlink(zilog_t *zilog, dmu_tx_t *tx, uint64_t txtype,
397 znode_t *dzp, znode_t *zp, char *name, char *link)
401 size_t namesize = strlen(name) + 1;
402 size_t linksize = strlen(link) + 1;
404 if (zil_replaying(zilog, tx))
407 itx = zil_itx_create(txtype, sizeof (*lr) + namesize + linksize);
408 lr = (lr_create_t *)&itx->itx_lr;
409 lr->lr_doid = dzp->z_id;
410 lr->lr_foid = zp->z_id;
411 lr->lr_uid = zp->z_uid;
412 lr->lr_gid = zp->z_gid;
413 lr->lr_mode = zp->z_mode;
414 (void) sa_lookup(zp->z_sa_hdl, SA_ZPL_GEN(zp->z_zfsvfs), &lr->lr_gen,
416 (void) sa_lookup(zp->z_sa_hdl, SA_ZPL_CRTIME(zp->z_zfsvfs),
417 lr->lr_crtime, sizeof (uint64_t) * 2);
418 bcopy(name, (char *)(lr + 1), namesize);
419 bcopy(link, (char *)(lr + 1) + namesize, linksize);
421 zil_itx_assign(zilog, itx, tx);
425 * zfs_log_rename() handles TX_RENAME transactions.
428 zfs_log_rename(zilog_t *zilog, dmu_tx_t *tx, uint64_t txtype,
429 znode_t *sdzp, char *sname, znode_t *tdzp, char *dname, znode_t *szp)
433 size_t snamesize = strlen(sname) + 1;
434 size_t dnamesize = strlen(dname) + 1;
436 if (zil_replaying(zilog, tx))
439 itx = zil_itx_create(txtype, sizeof (*lr) + snamesize + dnamesize);
440 lr = (lr_rename_t *)&itx->itx_lr;
441 lr->lr_sdoid = sdzp->z_id;
442 lr->lr_tdoid = tdzp->z_id;
443 bcopy(sname, (char *)(lr + 1), snamesize);
444 bcopy(dname, (char *)(lr + 1) + snamesize, dnamesize);
445 itx->itx_oid = szp->z_id;
447 zil_itx_assign(zilog, itx, tx);
451 * zfs_log_write() handles TX_WRITE transactions.
453 ssize_t zfs_immediate_write_sz = 32768;
456 zfs_log_write(zilog_t *zilog, dmu_tx_t *tx, int txtype,
457 znode_t *zp, offset_t off, ssize_t resid, int ioflag)
459 itx_wr_state_t write_state;
462 ssize_t immediate_write_sz;
464 if (zil_replaying(zilog, tx) || zp->z_unlinked)
467 immediate_write_sz = (zilog->zl_logbias == ZFS_LOGBIAS_THROUGHPUT)
468 ? 0 : zfs_immediate_write_sz;
470 slogging = spa_has_slogs(zilog->zl_spa) &&
471 (zilog->zl_logbias == ZFS_LOGBIAS_LATENCY);
472 if (resid > immediate_write_sz && !slogging && resid <= zp->z_blksz)
473 write_state = WR_INDIRECT;
474 else if (ioflag & (FSYNC | FDSYNC))
475 write_state = WR_COPIED;
477 write_state = WR_NEED_COPY;
479 if ((fsync_cnt = (uintptr_t)tsd_get(zfs_fsyncer_key)) != 0) {
480 (void) tsd_set(zfs_fsyncer_key, (void *)(fsync_cnt - 1));
489 * If the write would overflow the largest block then split it.
491 if (write_state != WR_INDIRECT && resid > ZIL_MAX_LOG_DATA)
492 len = SPA_MAXBLOCKSIZE >> 1;
496 itx = zil_itx_create(txtype, sizeof (*lr) +
497 (write_state == WR_COPIED ? len : 0));
498 lr = (lr_write_t *)&itx->itx_lr;
499 if (write_state == WR_COPIED && dmu_read(zp->z_zfsvfs->z_os,
500 zp->z_id, off, len, lr + 1, DMU_READ_NO_PREFETCH) != 0) {
501 zil_itx_destroy(itx);
502 itx = zil_itx_create(txtype, sizeof (*lr));
503 lr = (lr_write_t *)&itx->itx_lr;
504 write_state = WR_NEED_COPY;
507 itx->itx_wr_state = write_state;
508 if (write_state == WR_NEED_COPY)
510 lr->lr_foid = zp->z_id;
514 BP_ZERO(&lr->lr_blkptr);
516 itx->itx_private = zp->z_zfsvfs;
518 if (!(ioflag & (FSYNC | FDSYNC)) && (zp->z_sync_cnt == 0) &&
520 itx->itx_sync = B_FALSE;
522 zil_itx_assign(zilog, itx, tx);
530 * zfs_log_truncate() handles TX_TRUNCATE transactions.
533 zfs_log_truncate(zilog_t *zilog, dmu_tx_t *tx, int txtype,
534 znode_t *zp, uint64_t off, uint64_t len)
539 if (zil_replaying(zilog, tx) || zp->z_unlinked)
542 itx = zil_itx_create(txtype, sizeof (*lr));
543 lr = (lr_truncate_t *)&itx->itx_lr;
544 lr->lr_foid = zp->z_id;
548 itx->itx_sync = (zp->z_sync_cnt != 0);
549 zil_itx_assign(zilog, itx, tx);
553 * zfs_log_setattr() handles TX_SETATTR transactions.
556 zfs_log_setattr(zilog_t *zilog, dmu_tx_t *tx, int txtype,
557 znode_t *zp, vattr_t *vap, uint_t mask_applied, zfs_fuid_info_t *fuidp)
561 xvattr_t *xvap = (xvattr_t *)vap;
562 size_t recsize = sizeof (lr_setattr_t);
565 if (zil_replaying(zilog, tx) || zp->z_unlinked)
569 * If XVATTR set, then log record size needs to allow
570 * for lr_attr_t + xvattr mask, mapsize and create time
571 * plus actual attribute values
573 if (vap->va_mask & AT_XVATTR)
574 recsize = sizeof (*lr) + ZIL_XVAT_SIZE(xvap->xva_mapsize);
577 recsize += fuidp->z_domain_str_sz;
579 itx = zil_itx_create(txtype, recsize);
580 lr = (lr_setattr_t *)&itx->itx_lr;
581 lr->lr_foid = zp->z_id;
582 lr->lr_mask = (uint64_t)mask_applied;
583 lr->lr_mode = (uint64_t)vap->va_mode;
584 if ((mask_applied & AT_UID) && IS_EPHEMERAL(vap->va_uid))
585 lr->lr_uid = fuidp->z_fuid_owner;
587 lr->lr_uid = (uint64_t)vap->va_uid;
589 if ((mask_applied & AT_GID) && IS_EPHEMERAL(vap->va_gid))
590 lr->lr_gid = fuidp->z_fuid_group;
592 lr->lr_gid = (uint64_t)vap->va_gid;
594 lr->lr_size = (uint64_t)vap->va_size;
595 ZFS_TIME_ENCODE(&vap->va_atime, lr->lr_atime);
596 ZFS_TIME_ENCODE(&vap->va_mtime, lr->lr_mtime);
597 start = (lr_setattr_t *)(lr + 1);
598 if (vap->va_mask & AT_XVATTR) {
599 zfs_log_xvattr((lr_attr_t *)start, xvap);
600 start = (caddr_t)start + ZIL_XVAT_SIZE(xvap->xva_mapsize);
604 * Now stick on domain information if any on end
608 (void) zfs_log_fuid_domains(fuidp, start);
610 itx->itx_sync = (zp->z_sync_cnt != 0);
611 zil_itx_assign(zilog, itx, tx);
615 * zfs_log_acl() handles TX_ACL transactions.
618 zfs_log_acl(zilog_t *zilog, dmu_tx_t *tx, znode_t *zp,
619 vsecattr_t *vsecp, zfs_fuid_info_t *fuidp)
627 size_t aclbytes = vsecp->vsa_aclentsz;
629 if (zil_replaying(zilog, tx) || zp->z_unlinked)
632 txtype = (zp->z_zfsvfs->z_version < ZPL_VERSION_FUID) ?
635 if (txtype == TX_ACL)
636 lrsize = sizeof (*lr);
638 lrsize = sizeof (*lrv0);
641 ((txtype == TX_ACL) ? ZIL_ACE_LENGTH(aclbytes) : aclbytes) +
642 (fuidp ? fuidp->z_domain_str_sz : 0) +
643 sizeof (uint64_t) * (fuidp ? fuidp->z_fuid_cnt : 0);
645 itx = zil_itx_create(txtype, txsize);
647 lr = (lr_acl_t *)&itx->itx_lr;
648 lr->lr_foid = zp->z_id;
649 if (txtype == TX_ACL) {
650 lr->lr_acl_bytes = aclbytes;
651 lr->lr_domcnt = fuidp ? fuidp->z_domain_cnt : 0;
652 lr->lr_fuidcnt = fuidp ? fuidp->z_fuid_cnt : 0;
653 if (vsecp->vsa_mask & VSA_ACE_ACLFLAGS)
654 lr->lr_acl_flags = (uint64_t)vsecp->vsa_aclflags;
656 lr->lr_acl_flags = 0;
658 lr->lr_aclcnt = (uint64_t)vsecp->vsa_aclcnt;
660 if (txtype == TX_ACL_V0) {
661 lrv0 = (lr_acl_v0_t *)lr;
662 bcopy(vsecp->vsa_aclentp, (ace_t *)(lrv0 + 1), aclbytes);
664 void *start = (ace_t *)(lr + 1);
666 bcopy(vsecp->vsa_aclentp, start, aclbytes);
668 start = (caddr_t)start + ZIL_ACE_LENGTH(aclbytes);
671 start = zfs_log_fuid_ids(fuidp, start);
672 (void) zfs_log_fuid_domains(fuidp, start);
676 itx->itx_sync = (zp->z_sync_cnt != 0);
677 zil_itx_assign(zilog, itx, tx);
680 #endif /* HAVE_ZPL */