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) 2012 Cyril Plisko. All rights reserved.
24 * Copyright (c) 2013, 2015 by Delphix. 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>
35 #include <sys/fcntl.h>
37 #include <sys/fs/zfs.h>
38 #include <sys/zfs_znode.h>
39 #include <sys/zfs_dir.h>
40 #include <sys/zfs_acl.h>
41 #include <sys/zfs_fuid.h>
42 #include <sys/zfs_vnops.h>
45 #include <sys/byteorder.h>
49 #include <sys/atomic.h>
54 * Functions to replay ZFS intent log (ZIL) records
55 * The functions are called through a function vector (zfs_replay_vector)
56 * which is indexed by the transaction type.
60 zfs_init_vattr(vattr_t
*vap
, uint64_t mask
, uint64_t mode
,
61 uint64_t uid
, uint64_t gid
, uint64_t rdev
, uint64_t nodeid
)
63 bzero(vap
, sizeof (*vap
));
64 vap
->va_mask
= (uint_t
)mask
;
65 vap
->va_type
= IFTOVT(mode
);
67 vap
->va_uid
= (uid_t
)(IS_EPHEMERAL(uid
)) ? -1 : uid
;
68 vap
->va_gid
= (gid_t
)(IS_EPHEMERAL(gid
)) ? -1 : gid
;
70 vap
->va_nodeid
= nodeid
;
75 zfs_replay_error(zfsvfs_t
*zfsvfs
, lr_t
*lr
, 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
|= ATTR_XVATTR
;
91 if ((xoap
= xva_getxoptattr(xvap
)) == NULL
) {
92 xvap
->xva_vattr
.va_mask
&= ~ATTR_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(zfsvfs_t
*zfsvfs
,
269 lr_acl_create_t
*lracl
, boolean_t byteswap
)
271 char *name
= NULL
; /* location determined later */
272 lr_create_t
*lr
= (lr_create_t
*)lracl
;
274 struct inode
*ip
= NULL
;
277 vsecattr_t vsec
= { 0 };
287 txtype
= (lr
->lr_common
.lrc_txtype
& ~TX_CI
);
289 byteswap_uint64_array(lracl
, sizeof (*lracl
));
290 if (txtype
== TX_CREATE_ACL_ATTR
||
291 txtype
== TX_MKDIR_ACL_ATTR
) {
292 lrattr
= (lr_attr_t
*)(caddr_t
)(lracl
+ 1);
293 zfs_replay_swap_attrs(lrattr
);
294 xvatlen
= ZIL_XVAT_SIZE(lrattr
->lr_attr_masksize
);
297 aclstart
= (caddr_t
)(lracl
+ 1) + xvatlen
;
298 zfs_ace_byteswap(aclstart
, lracl
->lr_acl_bytes
, B_FALSE
);
300 if (lracl
->lr_fuidcnt
) {
301 byteswap_uint64_array((caddr_t
)aclstart
+
302 ZIL_ACE_LENGTH(lracl
->lr_acl_bytes
),
303 lracl
->lr_fuidcnt
* sizeof (uint64_t));
307 if ((error
= zfs_zget(zfsvfs
, lr
->lr_doid
, &dzp
)) != 0)
310 objid
= LR_FOID_GET_OBJ(lr
->lr_foid
);
311 dnodesize
= LR_FOID_GET_SLOTS(lr
->lr_foid
) << DNODE_SHIFT
;
314 zfs_init_vattr(&xva
.xva_vattr
, ATTR_MODE
| ATTR_UID
| ATTR_GID
,
315 lr
->lr_mode
, lr
->lr_uid
, lr
->lr_gid
, lr
->lr_rdev
, objid
);
318 * All forms of zfs create (create, mkdir, mkxattrdir, symlink)
319 * eventually end up in zfs_mknode(), which assigns the object's
320 * creation time, generation number, and dnode size. The generic
321 * zfs_create() has no concept of these attributes, so we smuggle
322 * the values inside the vattr's otherwise unused va_ctime,
323 * va_nblocks, and va_fsid fields.
325 ZFS_TIME_DECODE(&xva
.xva_vattr
.va_ctime
, lr
->lr_crtime
);
326 xva
.xva_vattr
.va_nblocks
= lr
->lr_gen
;
327 xva
.xva_vattr
.va_fsid
= dnodesize
;
329 error
= dmu_object_info(zfsvfs
->z_os
, lr
->lr_foid
, NULL
);
333 if (lr
->lr_common
.lrc_txtype
& TX_CI
)
337 aclstart
= (caddr_t
)(lracl
+ 1);
338 fuidstart
= (caddr_t
)aclstart
+
339 ZIL_ACE_LENGTH(lracl
->lr_acl_bytes
);
340 zfsvfs
->z_fuid_replay
= zfs_replay_fuids(fuidstart
,
341 (void *)&name
, lracl
->lr_fuidcnt
, lracl
->lr_domcnt
,
342 lr
->lr_uid
, lr
->lr_gid
);
344 case TX_CREATE_ACL_ATTR
:
346 lrattr
= (lr_attr_t
*)(caddr_t
)(lracl
+ 1);
347 xvatlen
= ZIL_XVAT_SIZE(lrattr
->lr_attr_masksize
);
348 xva
.xva_vattr
.va_mask
|= ATTR_XVATTR
;
349 zfs_replay_xvattr(lrattr
, &xva
);
351 vsec
.vsa_mask
= VSA_ACE
| VSA_ACE_ACLFLAGS
;
352 vsec
.vsa_aclentp
= (caddr_t
)(lracl
+ 1) + xvatlen
;
353 vsec
.vsa_aclcnt
= lracl
->lr_aclcnt
;
354 vsec
.vsa_aclentsz
= lracl
->lr_acl_bytes
;
355 vsec
.vsa_aclflags
= lracl
->lr_acl_flags
;
356 if (zfsvfs
->z_fuid_replay
== NULL
) {
357 fuidstart
= (caddr_t
)(lracl
+ 1) + xvatlen
+
358 ZIL_ACE_LENGTH(lracl
->lr_acl_bytes
);
359 zfsvfs
->z_fuid_replay
=
360 zfs_replay_fuids(fuidstart
,
361 (void *)&name
, lracl
->lr_fuidcnt
, lracl
->lr_domcnt
,
362 lr
->lr_uid
, lr
->lr_gid
);
365 error
= zfs_create(ZTOI(dzp
), name
, &xva
.xva_vattr
,
366 0, 0, &ip
, kcred
, vflg
, &vsec
);
369 aclstart
= (caddr_t
)(lracl
+ 1);
370 fuidstart
= (caddr_t
)aclstart
+
371 ZIL_ACE_LENGTH(lracl
->lr_acl_bytes
);
372 zfsvfs
->z_fuid_replay
= zfs_replay_fuids(fuidstart
,
373 (void *)&name
, lracl
->lr_fuidcnt
, lracl
->lr_domcnt
,
374 lr
->lr_uid
, lr
->lr_gid
);
376 case TX_MKDIR_ACL_ATTR
:
378 lrattr
= (lr_attr_t
*)(caddr_t
)(lracl
+ 1);
379 xvatlen
= ZIL_XVAT_SIZE(lrattr
->lr_attr_masksize
);
380 zfs_replay_xvattr(lrattr
, &xva
);
382 vsec
.vsa_mask
= VSA_ACE
| VSA_ACE_ACLFLAGS
;
383 vsec
.vsa_aclentp
= (caddr_t
)(lracl
+ 1) + xvatlen
;
384 vsec
.vsa_aclcnt
= lracl
->lr_aclcnt
;
385 vsec
.vsa_aclentsz
= lracl
->lr_acl_bytes
;
386 vsec
.vsa_aclflags
= lracl
->lr_acl_flags
;
387 if (zfsvfs
->z_fuid_replay
== NULL
) {
388 fuidstart
= (caddr_t
)(lracl
+ 1) + xvatlen
+
389 ZIL_ACE_LENGTH(lracl
->lr_acl_bytes
);
390 zfsvfs
->z_fuid_replay
=
391 zfs_replay_fuids(fuidstart
,
392 (void *)&name
, lracl
->lr_fuidcnt
, lracl
->lr_domcnt
,
393 lr
->lr_uid
, lr
->lr_gid
);
395 error
= zfs_mkdir(ZTOI(dzp
), name
, &xva
.xva_vattr
,
396 &ip
, kcred
, vflg
, &vsec
);
399 error
= SET_ERROR(ENOTSUP
);
403 if (error
== 0 && ip
!= NULL
)
408 if (zfsvfs
->z_fuid_replay
)
409 zfs_fuid_info_free(zfsvfs
->z_fuid_replay
);
410 zfsvfs
->z_fuid_replay
= NULL
;
416 zfs_replay_create(zfsvfs_t
*zfsvfs
, lr_create_t
*lr
, boolean_t byteswap
)
418 char *name
= NULL
; /* location determined later */
419 char *link
; /* symlink content follows name */
421 struct inode
*ip
= NULL
;
424 size_t lrsize
= sizeof (lr_create_t
);
433 txtype
= (lr
->lr_common
.lrc_txtype
& ~TX_CI
);
435 byteswap_uint64_array(lr
, sizeof (*lr
));
436 if (txtype
== TX_CREATE_ATTR
|| txtype
== TX_MKDIR_ATTR
)
437 zfs_replay_swap_attrs((lr_attr_t
*)(lr
+ 1));
441 if ((error
= zfs_zget(zfsvfs
, lr
->lr_doid
, &dzp
)) != 0)
444 objid
= LR_FOID_GET_OBJ(lr
->lr_foid
);
445 dnodesize
= LR_FOID_GET_SLOTS(lr
->lr_foid
) << DNODE_SHIFT
;
448 zfs_init_vattr(&xva
.xva_vattr
, ATTR_MODE
| ATTR_UID
| ATTR_GID
,
449 lr
->lr_mode
, lr
->lr_uid
, lr
->lr_gid
, lr
->lr_rdev
, objid
);
452 * All forms of zfs create (create, mkdir, mkxattrdir, symlink)
453 * eventually end up in zfs_mknode(), which assigns the object's
454 * creation time, generation number, and dnode slot count. The
455 * generic zfs_create() has no concept of these attributes, so
456 * we smuggle the values inside * the vattr's otherwise unused
457 * va_ctime, va_nblocks, and va_nlink fields.
459 ZFS_TIME_DECODE(&xva
.xva_vattr
.va_ctime
, lr
->lr_crtime
);
460 xva
.xva_vattr
.va_nblocks
= lr
->lr_gen
;
461 xva
.xva_vattr
.va_fsid
= dnodesize
;
463 error
= dmu_object_info(zfsvfs
->z_os
, objid
, NULL
);
467 if (lr
->lr_common
.lrc_txtype
& TX_CI
)
471 * Symlinks don't have fuid info, and CIFS never creates
474 * The _ATTR versions will grab the fuid info in their subcases.
476 if ((int)lr
->lr_common
.lrc_txtype
!= TX_SYMLINK
&&
477 (int)lr
->lr_common
.lrc_txtype
!= TX_MKDIR_ATTR
&&
478 (int)lr
->lr_common
.lrc_txtype
!= TX_CREATE_ATTR
) {
480 zfsvfs
->z_fuid_replay
=
481 zfs_replay_fuid_domain(start
, &start
,
482 lr
->lr_uid
, lr
->lr_gid
);
487 lrattr
= (lr_attr_t
*)(caddr_t
)(lr
+ 1);
488 xvatlen
= ZIL_XVAT_SIZE(lrattr
->lr_attr_masksize
);
489 zfs_replay_xvattr((lr_attr_t
*)((caddr_t
)lr
+ lrsize
), &xva
);
490 start
= (caddr_t
)(lr
+ 1) + xvatlen
;
491 zfsvfs
->z_fuid_replay
=
492 zfs_replay_fuid_domain(start
, &start
,
493 lr
->lr_uid
, lr
->lr_gid
);
494 name
= (char *)start
;
499 name
= (char *)start
;
501 error
= zfs_create(ZTOI(dzp
), name
, &xva
.xva_vattr
,
502 0, 0, &ip
, kcred
, vflg
, NULL
);
505 lrattr
= (lr_attr_t
*)(caddr_t
)(lr
+ 1);
506 xvatlen
= ZIL_XVAT_SIZE(lrattr
->lr_attr_masksize
);
507 zfs_replay_xvattr((lr_attr_t
*)((caddr_t
)lr
+ lrsize
), &xva
);
508 start
= (caddr_t
)(lr
+ 1) + xvatlen
;
509 zfsvfs
->z_fuid_replay
=
510 zfs_replay_fuid_domain(start
, &start
,
511 lr
->lr_uid
, lr
->lr_gid
);
512 name
= (char *)start
;
517 name
= (char *)(lr
+ 1);
519 error
= zfs_mkdir(ZTOI(dzp
), name
, &xva
.xva_vattr
,
520 &ip
, kcred
, vflg
, NULL
);
523 error
= zfs_make_xattrdir(dzp
, &xva
.xva_vattr
, &ip
, kcred
);
526 name
= (char *)(lr
+ 1);
527 link
= name
+ strlen(name
) + 1;
528 error
= zfs_symlink(ZTOI(dzp
), name
, &xva
.xva_vattr
,
529 link
, &ip
, kcred
, vflg
);
532 error
= SET_ERROR(ENOTSUP
);
536 if (error
== 0 && ip
!= NULL
)
541 if (zfsvfs
->z_fuid_replay
)
542 zfs_fuid_info_free(zfsvfs
->z_fuid_replay
);
543 zfsvfs
->z_fuid_replay
= NULL
;
548 zfs_replay_remove(zfsvfs_t
*zfsvfs
, lr_remove_t
*lr
, boolean_t byteswap
)
550 char *name
= (char *)(lr
+ 1); /* name follows lr_remove_t */
556 byteswap_uint64_array(lr
, sizeof (*lr
));
558 if ((error
= zfs_zget(zfsvfs
, lr
->lr_doid
, &dzp
)) != 0)
561 if (lr
->lr_common
.lrc_txtype
& TX_CI
)
564 switch ((int)lr
->lr_common
.lrc_txtype
) {
566 error
= zfs_remove(ZTOI(dzp
), name
, kcred
, vflg
);
569 error
= zfs_rmdir(ZTOI(dzp
), name
, NULL
, kcred
, vflg
);
572 error
= SET_ERROR(ENOTSUP
);
581 zfs_replay_link(zfsvfs_t
*zfsvfs
, lr_link_t
*lr
, boolean_t byteswap
)
583 char *name
= (char *)(lr
+ 1); /* name follows lr_link_t */
589 byteswap_uint64_array(lr
, sizeof (*lr
));
591 if ((error
= zfs_zget(zfsvfs
, lr
->lr_doid
, &dzp
)) != 0)
594 if ((error
= zfs_zget(zfsvfs
, lr
->lr_link_obj
, &zp
)) != 0) {
599 if (lr
->lr_common
.lrc_txtype
& TX_CI
)
602 error
= zfs_link(ZTOI(dzp
), ZTOI(zp
), name
, kcred
, vflg
);
611 zfs_replay_rename(zfsvfs_t
*zfsvfs
, lr_rename_t
*lr
, boolean_t byteswap
)
613 char *sname
= (char *)(lr
+ 1); /* sname and tname follow lr_rename_t */
614 char *tname
= sname
+ strlen(sname
) + 1;
615 znode_t
*sdzp
, *tdzp
;
620 byteswap_uint64_array(lr
, sizeof (*lr
));
622 if ((error
= zfs_zget(zfsvfs
, lr
->lr_sdoid
, &sdzp
)) != 0)
625 if ((error
= zfs_zget(zfsvfs
, lr
->lr_tdoid
, &tdzp
)) != 0) {
630 if (lr
->lr_common
.lrc_txtype
& TX_CI
)
633 error
= zfs_rename(ZTOI(sdzp
), sname
, ZTOI(tdzp
), tname
, kcred
, vflg
);
642 zfs_replay_write(zfsvfs_t
*zfsvfs
, lr_write_t
*lr
, boolean_t byteswap
)
644 char *data
= (char *)(lr
+ 1); /* data follows lr_write_t */
647 uint64_t eod
, offset
, length
;
650 byteswap_uint64_array(lr
, sizeof (*lr
));
652 if ((error
= zfs_zget(zfsvfs
, lr
->lr_foid
, &zp
)) != 0) {
654 * As we can log writes out of order, it's possible the
655 * file has been removed. In this case just drop the write
656 * and return success.
663 offset
= lr
->lr_offset
;
664 length
= lr
->lr_length
;
665 eod
= offset
+ length
; /* end of data for this write */
668 * This may be a write from a dmu_sync() for a whole block,
669 * and may extend beyond the current end of the file.
670 * We can't just replay what was written for this TX_WRITE as
671 * a future TX_WRITE2 may extend the eof and the data for that
672 * write needs to be there. So we write the whole block and
673 * reduce the eof. This needs to be done within the single dmu
674 * transaction created within vn_rdwr -> zfs_write. So a possible
675 * new end of file is passed through in zfsvfs->z_replay_eof
678 zfsvfs
->z_replay_eof
= 0; /* 0 means don't change end of file */
680 /* If it's a dmu_sync() block, write the whole block */
681 if (lr
->lr_common
.lrc_reclen
== sizeof (lr_write_t
)) {
682 uint64_t blocksize
= BP_GET_LSIZE(&lr
->lr_blkptr
);
683 if (length
< blocksize
) {
684 offset
-= offset
% blocksize
;
687 if (zp
->z_size
< eod
)
688 zfsvfs
->z_replay_eof
= eod
;
691 written
= zpl_write_common(ZTOI(zp
), data
, length
, &offset
,
692 UIO_SYSSPACE
, 0, kcred
);
695 else if (written
< length
)
696 error
= SET_ERROR(EIO
); /* short write */
699 zfsvfs
->z_replay_eof
= 0; /* safety */
705 * TX_WRITE2 are only generated when dmu_sync() returns EALREADY
706 * meaning the pool block is already being synced. So now that we always write
707 * out full blocks, all we have to do is expand the eof if
711 zfs_replay_write2(zfsvfs_t
*zfsvfs
, lr_write_t
*lr
, boolean_t byteswap
)
718 byteswap_uint64_array(lr
, sizeof (*lr
));
720 if ((error
= zfs_zget(zfsvfs
, lr
->lr_foid
, &zp
)) != 0)
724 end
= lr
->lr_offset
+ lr
->lr_length
;
725 if (end
> zp
->z_size
) {
726 dmu_tx_t
*tx
= dmu_tx_create(zfsvfs
->z_os
);
729 dmu_tx_hold_sa(tx
, zp
->z_sa_hdl
, B_FALSE
);
730 error
= dmu_tx_assign(tx
, TXG_WAIT
);
733 if (error
== ERESTART
) {
741 (void) sa_update(zp
->z_sa_hdl
, SA_ZPL_SIZE(zfsvfs
),
742 (void *)&zp
->z_size
, sizeof (uint64_t), tx
);
744 /* Ensure the replayed seq is updated */
745 (void) zil_replaying(zfsvfs
->z_log
, tx
);
756 zfs_replay_truncate(zfsvfs_t
*zfsvfs
, lr_truncate_t
*lr
, boolean_t byteswap
)
763 byteswap_uint64_array(lr
, sizeof (*lr
));
765 if ((error
= zfs_zget(zfsvfs
, lr
->lr_foid
, &zp
)) != 0)
768 bzero(&fl
, sizeof (fl
));
771 fl
.l_start
= lr
->lr_offset
;
772 fl
.l_len
= lr
->lr_length
;
774 error
= zfs_space(ZTOI(zp
), F_FREESP
, &fl
, FWRITE
| FOFFMAX
,
775 lr
->lr_offset
, kcred
);
783 zfs_replay_setattr(zfsvfs_t
*zfsvfs
, lr_setattr_t
*lr
, boolean_t byteswap
)
787 vattr_t
*vap
= &xva
.xva_vattr
;
793 byteswap_uint64_array(lr
, sizeof (*lr
));
795 if ((lr
->lr_mask
& ATTR_XVATTR
) &&
796 zfsvfs
->z_version
>= ZPL_VERSION_INITIAL
)
797 zfs_replay_swap_attrs((lr_attr_t
*)(lr
+ 1));
800 if ((error
= zfs_zget(zfsvfs
, lr
->lr_foid
, &zp
)) != 0)
803 zfs_init_vattr(vap
, lr
->lr_mask
, lr
->lr_mode
,
804 lr
->lr_uid
, lr
->lr_gid
, 0, lr
->lr_foid
);
806 vap
->va_size
= lr
->lr_size
;
807 ZFS_TIME_DECODE(&vap
->va_atime
, lr
->lr_atime
);
808 ZFS_TIME_DECODE(&vap
->va_mtime
, lr
->lr_mtime
);
809 gethrestime(&vap
->va_ctime
);
810 vap
->va_mask
|= ATTR_CTIME
;
813 * Fill in xvattr_t portions if necessary.
816 start
= (lr_setattr_t
*)(lr
+ 1);
817 if (vap
->va_mask
& ATTR_XVATTR
) {
818 zfs_replay_xvattr((lr_attr_t
*)start
, &xva
);
819 start
= (caddr_t
)start
+
820 ZIL_XVAT_SIZE(((lr_attr_t
*)start
)->lr_attr_masksize
);
822 xva
.xva_vattr
.va_mask
&= ~ATTR_XVATTR
;
824 zfsvfs
->z_fuid_replay
= zfs_replay_fuid_domain(start
, &start
,
825 lr
->lr_uid
, lr
->lr_gid
);
827 error
= zfs_setattr(ZTOI(zp
), vap
, 0, kcred
);
829 zfs_fuid_info_free(zfsvfs
->z_fuid_replay
);
830 zfsvfs
->z_fuid_replay
= NULL
;
837 zfs_replay_acl_v0(zfsvfs_t
*zfsvfs
, lr_acl_v0_t
*lr
, boolean_t byteswap
)
839 ace_t
*ace
= (ace_t
*)(lr
+ 1); /* ace array follows lr_acl_t */
845 byteswap_uint64_array(lr
, sizeof (*lr
));
846 zfs_oldace_byteswap(ace
, lr
->lr_aclcnt
);
849 if ((error
= zfs_zget(zfsvfs
, lr
->lr_foid
, &zp
)) != 0)
852 bzero(&vsa
, sizeof (vsa
));
853 vsa
.vsa_mask
= VSA_ACE
| VSA_ACECNT
;
854 vsa
.vsa_aclcnt
= lr
->lr_aclcnt
;
855 vsa
.vsa_aclentsz
= sizeof (ace_t
) * vsa
.vsa_aclcnt
;
856 vsa
.vsa_aclflags
= 0;
857 vsa
.vsa_aclentp
= ace
;
859 error
= zfs_setsecattr(ZTOI(zp
), &vsa
, 0, kcred
);
867 * Replaying ACLs is complicated by FUID support.
868 * The log record may contain some optional data
869 * to be used for replaying FUID's. These pieces
870 * are the actual FUIDs that were created initially.
871 * The FUID table index may no longer be valid and
872 * during zfs_create() a new index may be assigned.
873 * Because of this the log will contain the original
874 * domain+rid in order to create a new FUID.
876 * The individual ACEs may contain an ephemeral uid/gid which is no
877 * longer valid and will need to be replaced with an actual FUID.
881 zfs_replay_acl(zfsvfs_t
*zfsvfs
, lr_acl_t
*lr
, boolean_t byteswap
)
883 ace_t
*ace
= (ace_t
*)(lr
+ 1);
889 byteswap_uint64_array(lr
, sizeof (*lr
));
890 zfs_ace_byteswap(ace
, lr
->lr_acl_bytes
, B_FALSE
);
891 if (lr
->lr_fuidcnt
) {
892 byteswap_uint64_array((caddr_t
)ace
+
893 ZIL_ACE_LENGTH(lr
->lr_acl_bytes
),
894 lr
->lr_fuidcnt
* sizeof (uint64_t));
898 if ((error
= zfs_zget(zfsvfs
, lr
->lr_foid
, &zp
)) != 0)
901 bzero(&vsa
, sizeof (vsa
));
902 vsa
.vsa_mask
= VSA_ACE
| VSA_ACECNT
| VSA_ACE_ACLFLAGS
;
903 vsa
.vsa_aclcnt
= lr
->lr_aclcnt
;
904 vsa
.vsa_aclentp
= ace
;
905 vsa
.vsa_aclentsz
= lr
->lr_acl_bytes
;
906 vsa
.vsa_aclflags
= lr
->lr_acl_flags
;
908 if (lr
->lr_fuidcnt
) {
909 void *fuidstart
= (caddr_t
)ace
+
910 ZIL_ACE_LENGTH(lr
->lr_acl_bytes
);
912 zfsvfs
->z_fuid_replay
=
913 zfs_replay_fuids(fuidstart
, &fuidstart
,
914 lr
->lr_fuidcnt
, lr
->lr_domcnt
, 0, 0);
917 error
= zfs_setsecattr(ZTOI(zp
), &vsa
, 0, kcred
);
919 if (zfsvfs
->z_fuid_replay
)
920 zfs_fuid_info_free(zfsvfs
->z_fuid_replay
);
922 zfsvfs
->z_fuid_replay
= NULL
;
929 * Callback vectors for replaying records
931 zil_replay_func_t zfs_replay_vector
[TX_MAX_TYPE
] = {
932 (zil_replay_func_t
)zfs_replay_error
, /* no such type */
933 (zil_replay_func_t
)zfs_replay_create
, /* TX_CREATE */
934 (zil_replay_func_t
)zfs_replay_create
, /* TX_MKDIR */
935 (zil_replay_func_t
)zfs_replay_create
, /* TX_MKXATTR */
936 (zil_replay_func_t
)zfs_replay_create
, /* TX_SYMLINK */
937 (zil_replay_func_t
)zfs_replay_remove
, /* TX_REMOVE */
938 (zil_replay_func_t
)zfs_replay_remove
, /* TX_RMDIR */
939 (zil_replay_func_t
)zfs_replay_link
, /* TX_LINK */
940 (zil_replay_func_t
)zfs_replay_rename
, /* TX_RENAME */
941 (zil_replay_func_t
)zfs_replay_write
, /* TX_WRITE */
942 (zil_replay_func_t
)zfs_replay_truncate
, /* TX_TRUNCATE */
943 (zil_replay_func_t
)zfs_replay_setattr
, /* TX_SETATTR */
944 (zil_replay_func_t
)zfs_replay_acl_v0
, /* TX_ACL_V0 */
945 (zil_replay_func_t
)zfs_replay_acl
, /* TX_ACL */
946 (zil_replay_func_t
)zfs_replay_create_acl
, /* TX_CREATE_ACL */
947 (zil_replay_func_t
)zfs_replay_create
, /* TX_CREATE_ATTR */
948 (zil_replay_func_t
)zfs_replay_create_acl
, /* TX_CREATE_ACL_ATTR */
949 (zil_replay_func_t
)zfs_replay_create_acl
, /* TX_MKDIR_ACL */
950 (zil_replay_func_t
)zfs_replay_create
, /* TX_MKDIR_ATTR */
951 (zil_replay_func_t
)zfs_replay_create_acl
, /* TX_MKDIR_ACL_ATTR */
952 (zil_replay_func_t
)zfs_replay_write2
, /* TX_WRITE2 */