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]
23 * Copyright (c) 2005, 2010, Oracle and/or its affiliates. All rights reserved.
24 * Copyright (c) 2011, 2015 by Delphix. All rights reserved.
25 * Copyright (c) 2012, Joyent, Inc. All rights reserved.
26 * Copyright (c) 2012 Pawel Jakub Dawidek. All rights reserved.
27 * Copyright (c) 2013 Steven Hartland. All rights reserved.
28 * Copyright 2015, OmniTI Computer Consulting, Inc. All rights reserved.
29 * Copyright (c) 2014 Integros [integros.com]
30 * Copyright 2016 Igor Kozhukhov <ikozhukhov@gmail.com>
43 #include <sys/param.h>
44 #include <sys/mount.h>
50 #include <libzfs_core.h>
52 #include "zfs_namecheck.h"
54 #include "zfs_fletcher.h"
55 #include "libzfs_impl.h"
58 #include <sys/zio_checksum.h>
62 extern int zfs_ioctl_version;
65 /* in libzfs_dataset.c */
66 extern void zfs_setprop_error(libzfs_handle_t *, zfs_prop_t, int, char *);
67 /* We need to use something for ENODATA. */
70 static int zfs_receive_impl(libzfs_handle_t *, const char *, const char *,
71 recvflags_t *, int, const char *, nvlist_t *, avl_tree_t *, char **, int,
72 uint64_t *, const char *);
73 static int guid_to_name(libzfs_handle_t *, const char *,
74 uint64_t, boolean_t, char *);
76 static const zio_cksum_t zero_cksum = { 0 };
78 typedef struct dedup_arg {
81 libzfs_handle_t *dedup_hdl;
84 typedef struct progress_arg {
87 boolean_t pa_parsable;
92 typedef struct dataref {
98 typedef struct dedup_entry {
99 struct dedup_entry *dde_next;
100 zio_cksum_t dde_chksum;
105 #define MAX_DDT_PHYSMEM_PERCENT 20
106 #define SMALLEST_POSSIBLE_MAX_DDT_MB 128
108 typedef struct dedup_table {
109 dedup_entry_t **dedup_hash_array;
110 umem_cache_t *ddecache;
111 uint64_t max_ddt_size; /* max dedup table size in bytes */
112 uint64_t cur_ddt_size; /* current dedup table size in bytes */
119 high_order_bit(uint64_t n)
123 for (count = 0; n != 0; count++)
129 ssread(void *buf, size_t len, FILE *stream)
133 if ((outlen = fread(buf, len, 1, stream)) == 0)
140 ddt_hash_append(libzfs_handle_t *hdl, dedup_table_t *ddt, dedup_entry_t **ddepp,
141 zio_cksum_t *cs, uint64_t prop, dataref_t *dr)
145 if (ddt->cur_ddt_size >= ddt->max_ddt_size) {
146 if (ddt->ddt_full == B_FALSE) {
147 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
148 "Dedup table full. Deduplication will continue "
149 "with existing table entries"));
150 ddt->ddt_full = B_TRUE;
155 if ((dde = umem_cache_alloc(ddt->ddecache, UMEM_DEFAULT))
157 assert(*ddepp == NULL);
158 dde->dde_next = NULL;
159 dde->dde_chksum = *cs;
160 dde->dde_prop = prop;
163 ddt->cur_ddt_size += sizeof (dedup_entry_t);
169 * Using the specified dedup table, do a lookup for an entry with
170 * the checksum cs. If found, return the block's reference info
171 * in *dr. Otherwise, insert a new entry in the dedup table, using
172 * the reference information specified by *dr.
174 * return value: true - entry was found
175 * false - entry was not found
178 ddt_update(libzfs_handle_t *hdl, dedup_table_t *ddt, zio_cksum_t *cs,
179 uint64_t prop, dataref_t *dr)
182 dedup_entry_t **ddepp;
184 hashcode = BF64_GET(cs->zc_word[0], 0, ddt->numhashbits);
186 for (ddepp = &(ddt->dedup_hash_array[hashcode]); *ddepp != NULL;
187 ddepp = &((*ddepp)->dde_next)) {
188 if (ZIO_CHECKSUM_EQUAL(((*ddepp)->dde_chksum), *cs) &&
189 (*ddepp)->dde_prop == prop) {
190 *dr = (*ddepp)->dde_ref;
194 ddt_hash_append(hdl, ddt, ddepp, cs, prop, dr);
199 dump_record(dmu_replay_record_t *drr, void *payload, int payload_len,
200 zio_cksum_t *zc, int outfd)
202 ASSERT3U(offsetof(dmu_replay_record_t, drr_u.drr_checksum.drr_checksum),
203 ==, sizeof (dmu_replay_record_t) - sizeof (zio_cksum_t));
204 (void) fletcher_4_incremental_native(drr,
205 offsetof(dmu_replay_record_t, drr_u.drr_checksum.drr_checksum), zc);
206 if (drr->drr_type != DRR_BEGIN) {
207 ASSERT(ZIO_CHECKSUM_IS_ZERO(&drr->drr_u.
208 drr_checksum.drr_checksum));
209 drr->drr_u.drr_checksum.drr_checksum = *zc;
211 (void) fletcher_4_incremental_native(
212 &drr->drr_u.drr_checksum.drr_checksum, sizeof (zio_cksum_t), zc);
213 if (write(outfd, drr, sizeof (*drr)) == -1)
215 if (payload_len != 0) {
216 (void) fletcher_4_incremental_native(payload, payload_len, zc);
217 if (write(outfd, payload, payload_len) == -1)
224 * This function is started in a separate thread when the dedup option
225 * has been requested. The main send thread determines the list of
226 * snapshots to be included in the send stream and makes the ioctl calls
227 * for each one. But instead of having the ioctl send the output to the
228 * the output fd specified by the caller of zfs_send()), the
229 * ioctl is told to direct the output to a pipe, which is read by the
230 * alternate thread running THIS function. This function does the
232 * 1. building a dedup table (the DDT)
233 * 2. doing checksums on each data block and inserting a record in the DDT
234 * 3. looking for matching checksums, and
235 * 4. sending a DRR_WRITE_BYREF record instead of a write record whenever
236 * a duplicate block is found.
237 * The output of this function then goes to the output fd requested
238 * by the caller of zfs_send().
243 dedup_arg_t *dda = arg;
244 char *buf = zfs_alloc(dda->dedup_hdl, SPA_MAXBLOCKSIZE);
245 dmu_replay_record_t thedrr;
246 dmu_replay_record_t *drr = &thedrr;
250 zio_cksum_t stream_cksum;
251 uint64_t physmem = sysconf(_SC_PHYS_PAGES) * sysconf(_SC_PAGESIZE);
255 MAX((physmem * MAX_DDT_PHYSMEM_PERCENT) / 100,
256 SMALLEST_POSSIBLE_MAX_DDT_MB << 20);
258 numbuckets = ddt.max_ddt_size / (sizeof (dedup_entry_t));
261 * numbuckets must be a power of 2. Increase number to
262 * a power of 2 if necessary.
264 if (!ISP2(numbuckets))
265 numbuckets = 1 << high_order_bit(numbuckets);
267 ddt.dedup_hash_array = calloc(numbuckets, sizeof (dedup_entry_t *));
268 ddt.ddecache = umem_cache_create("dde", sizeof (dedup_entry_t), 0,
269 NULL, NULL, NULL, NULL, NULL, 0);
270 ddt.cur_ddt_size = numbuckets * sizeof (dedup_entry_t *);
271 ddt.numhashbits = high_order_bit(numbuckets) - 1;
272 ddt.ddt_full = B_FALSE;
274 outfd = dda->outputfd;
275 ofp = fdopen(dda->inputfd, "r");
276 while (ssread(drr, sizeof (*drr), ofp) != 0) {
279 * kernel filled in checksum, we are going to write same
280 * record, but need to regenerate checksum.
282 if (drr->drr_type != DRR_BEGIN) {
283 bzero(&drr->drr_u.drr_checksum.drr_checksum,
284 sizeof (drr->drr_u.drr_checksum.drr_checksum));
287 switch (drr->drr_type) {
290 struct drr_begin *drrb = &drr->drr_u.drr_begin;
293 ZIO_SET_CHECKSUM(&stream_cksum, 0, 0, 0, 0);
295 ASSERT3U(drrb->drr_magic, ==, DMU_BACKUP_MAGIC);
297 /* set the DEDUP feature flag for this stream */
298 fflags = DMU_GET_FEATUREFLAGS(drrb->drr_versioninfo);
299 fflags |= (DMU_BACKUP_FEATURE_DEDUP |
300 DMU_BACKUP_FEATURE_DEDUPPROPS);
301 DMU_SET_FEATUREFLAGS(drrb->drr_versioninfo, fflags);
303 if (drr->drr_payloadlen != 0) {
304 sz = drr->drr_payloadlen;
306 if (sz > SPA_MAXBLOCKSIZE) {
307 buf = zfs_realloc(dda->dedup_hdl, buf,
308 SPA_MAXBLOCKSIZE, sz);
310 (void) ssread(buf, sz, ofp);
314 if (dump_record(drr, buf, sz, &stream_cksum,
322 struct drr_end *drre = &drr->drr_u.drr_end;
323 /* use the recalculated checksum */
324 drre->drr_checksum = stream_cksum;
325 if (dump_record(drr, NULL, 0, &stream_cksum,
333 struct drr_object *drro = &drr->drr_u.drr_object;
334 if (drro->drr_bonuslen > 0) {
336 P2ROUNDUP((uint64_t)drro->drr_bonuslen, 8),
339 if (dump_record(drr, buf,
340 P2ROUNDUP((uint64_t)drro->drr_bonuslen, 8),
341 &stream_cksum, outfd) != 0)
348 struct drr_spill *drrs = &drr->drr_u.drr_spill;
349 (void) ssread(buf, drrs->drr_length, ofp);
350 if (dump_record(drr, buf, drrs->drr_length,
351 &stream_cksum, outfd) != 0)
356 case DRR_FREEOBJECTS:
358 if (dump_record(drr, NULL, 0, &stream_cksum,
366 struct drr_write *drrw = &drr->drr_u.drr_write;
368 uint64_t payload_size;
370 payload_size = DRR_WRITE_PAYLOAD_SIZE(drrw);
371 (void) ssread(buf, payload_size, ofp);
374 * Use the existing checksum if it's dedup-capable,
375 * else calculate a SHA256 checksum for it.
378 if (ZIO_CHECKSUM_EQUAL(drrw->drr_key.ddk_cksum,
380 !DRR_IS_DEDUP_CAPABLE(drrw->drr_checksumflags)) {
382 zio_cksum_t tmpsha256;
385 SHA256Update(&ctx, buf, payload_size);
386 SHA256Final(&tmpsha256, &ctx);
387 drrw->drr_key.ddk_cksum.zc_word[0] =
388 BE_64(tmpsha256.zc_word[0]);
389 drrw->drr_key.ddk_cksum.zc_word[1] =
390 BE_64(tmpsha256.zc_word[1]);
391 drrw->drr_key.ddk_cksum.zc_word[2] =
392 BE_64(tmpsha256.zc_word[2]);
393 drrw->drr_key.ddk_cksum.zc_word[3] =
394 BE_64(tmpsha256.zc_word[3]);
395 drrw->drr_checksumtype = ZIO_CHECKSUM_SHA256;
396 drrw->drr_checksumflags = DRR_CHECKSUM_DEDUP;
399 dataref.ref_guid = drrw->drr_toguid;
400 dataref.ref_object = drrw->drr_object;
401 dataref.ref_offset = drrw->drr_offset;
403 if (ddt_update(dda->dedup_hdl, &ddt,
404 &drrw->drr_key.ddk_cksum, drrw->drr_key.ddk_prop,
406 dmu_replay_record_t wbr_drr = {0};
407 struct drr_write_byref *wbr_drrr =
408 &wbr_drr.drr_u.drr_write_byref;
410 /* block already present in stream */
411 wbr_drr.drr_type = DRR_WRITE_BYREF;
413 wbr_drrr->drr_object = drrw->drr_object;
414 wbr_drrr->drr_offset = drrw->drr_offset;
415 wbr_drrr->drr_length = drrw->drr_logical_size;
416 wbr_drrr->drr_toguid = drrw->drr_toguid;
417 wbr_drrr->drr_refguid = dataref.ref_guid;
418 wbr_drrr->drr_refobject =
420 wbr_drrr->drr_refoffset =
423 wbr_drrr->drr_checksumtype =
424 drrw->drr_checksumtype;
425 wbr_drrr->drr_checksumflags =
426 drrw->drr_checksumtype;
427 wbr_drrr->drr_key.ddk_cksum =
428 drrw->drr_key.ddk_cksum;
429 wbr_drrr->drr_key.ddk_prop =
430 drrw->drr_key.ddk_prop;
432 if (dump_record(&wbr_drr, NULL, 0,
433 &stream_cksum, outfd) != 0)
436 /* block not previously seen */
437 if (dump_record(drr, buf, payload_size,
438 &stream_cksum, outfd) != 0)
444 case DRR_WRITE_EMBEDDED:
446 struct drr_write_embedded *drrwe =
447 &drr->drr_u.drr_write_embedded;
449 P2ROUNDUP((uint64_t)drrwe->drr_psize, 8), ofp);
450 if (dump_record(drr, buf,
451 P2ROUNDUP((uint64_t)drrwe->drr_psize, 8),
452 &stream_cksum, outfd) != 0)
459 if (dump_record(drr, NULL, 0, &stream_cksum,
466 (void) fprintf(stderr, "INVALID record type 0x%x\n",
468 /* should never happen, so assert */
473 umem_cache_destroy(ddt.ddecache);
474 free(ddt.dedup_hash_array);
482 * Routines for dealing with the AVL tree of fs-nvlists
484 typedef struct fsavl_node {
492 fsavl_compare(const void *arg1, const void *arg2)
494 const fsavl_node_t *fn1 = (const fsavl_node_t *)arg1;
495 const fsavl_node_t *fn2 = (const fsavl_node_t *)arg2;
497 return (AVL_CMP(fn1->fn_guid, fn2->fn_guid));
501 * Given the GUID of a snapshot, find its containing filesystem and
505 fsavl_find(avl_tree_t *avl, uint64_t snapguid, char **snapname)
507 fsavl_node_t fn_find;
510 fn_find.fn_guid = snapguid;
512 fn = avl_find(avl, &fn_find, NULL);
515 *snapname = fn->fn_snapname;
516 return (fn->fn_nvfs);
522 fsavl_destroy(avl_tree_t *avl)
531 while ((fn = avl_destroy_nodes(avl, &cookie)) != NULL)
538 * Given an nvlist, produce an avl tree of snapshots, ordered by guid
541 fsavl_create(nvlist_t *fss)
544 nvpair_t *fselem = NULL;
546 if ((fsavl = malloc(sizeof (avl_tree_t))) == NULL)
549 avl_create(fsavl, fsavl_compare, sizeof (fsavl_node_t),
550 offsetof(fsavl_node_t, fn_node));
552 while ((fselem = nvlist_next_nvpair(fss, fselem)) != NULL) {
553 nvlist_t *nvfs, *snaps;
554 nvpair_t *snapelem = NULL;
556 VERIFY(0 == nvpair_value_nvlist(fselem, &nvfs));
557 VERIFY(0 == nvlist_lookup_nvlist(nvfs, "snaps", &snaps));
560 nvlist_next_nvpair(snaps, snapelem)) != NULL) {
564 VERIFY(0 == nvpair_value_uint64(snapelem, &guid));
565 if ((fn = malloc(sizeof (fsavl_node_t))) == NULL) {
566 fsavl_destroy(fsavl);
570 fn->fn_snapname = nvpair_name(snapelem);
574 * Note: if there are multiple snaps with the
575 * same GUID, we ignore all but one.
577 if (avl_find(fsavl, fn, NULL) == NULL)
588 * Routines for dealing with the giant nvlist of fs-nvlists, etc.
590 typedef struct send_data {
592 * assigned inside every recursive call,
593 * restored from *_save on return:
595 * guid of fromsnap snapshot in parent dataset
596 * txg of fromsnap snapshot in current dataset
597 * txg of tosnap snapshot in current dataset
600 uint64_t parent_fromsnap_guid;
601 uint64_t fromsnap_txg;
604 /* the nvlists get accumulated during depth-first traversal */
605 nvlist_t *parent_snaps;
609 /* send-receive configuration, does not change during traversal */
611 const char *fromsnap;
617 * The header nvlist is of the following format:
620 * "fromsnap" -> string (if incremental)
624 * "name" -> string (full name; for debugging)
625 * "parentfromsnap" -> number (guid of fromsnap in parent)
627 * "props" -> { name -> value (only if set here) }
628 * "snaps" -> { name (lastname) -> number (guid) }
629 * "snapprops" -> { name (lastname) -> { name -> value } }
631 * "origin" -> number (guid) (if clone)
632 * "sent" -> boolean (not on-disk)
640 static void send_iterate_prop(zfs_handle_t *zhp, nvlist_t *nv);
643 send_iterate_snap(zfs_handle_t *zhp, void *arg)
645 send_data_t *sd = arg;
646 uint64_t guid = zhp->zfs_dmustats.dds_guid;
647 uint64_t txg = zhp->zfs_dmustats.dds_creation_txg;
651 snapname = strrchr(zhp->zfs_name, '@')+1;
653 if (sd->tosnap_txg != 0 && txg > sd->tosnap_txg) {
655 (void) fprintf(stderr, dgettext(TEXT_DOMAIN,
656 "skipping snapshot %s because it was created "
657 "after the destination snapshot (%s)\n"),
658 zhp->zfs_name, sd->tosnap);
664 VERIFY(0 == nvlist_add_uint64(sd->parent_snaps, snapname, guid));
666 * NB: if there is no fromsnap here (it's a newly created fs in
667 * an incremental replication), we will substitute the tosnap.
669 if ((sd->fromsnap && strcmp(snapname, sd->fromsnap) == 0) ||
670 (sd->parent_fromsnap_guid == 0 && sd->tosnap &&
671 strcmp(snapname, sd->tosnap) == 0)) {
672 sd->parent_fromsnap_guid = guid;
675 VERIFY(0 == nvlist_alloc(&nv, NV_UNIQUE_NAME, 0));
676 send_iterate_prop(zhp, nv);
677 VERIFY(0 == nvlist_add_nvlist(sd->snapprops, snapname, nv));
685 send_iterate_prop(zfs_handle_t *zhp, nvlist_t *nv)
687 nvpair_t *elem = NULL;
689 while ((elem = nvlist_next_nvpair(zhp->zfs_props, elem)) != NULL) {
690 char *propname = nvpair_name(elem);
691 zfs_prop_t prop = zfs_name_to_prop(propname);
694 if (!zfs_prop_user(propname)) {
696 * Realistically, this should never happen. However,
697 * we want the ability to add DSL properties without
698 * needing to make incompatible version changes. We
699 * need to ignore unknown properties to allow older
700 * software to still send datasets containing these
701 * properties, with the unknown properties elided.
703 if (prop == ZPROP_INVAL)
706 if (zfs_prop_readonly(prop))
710 verify(nvpair_value_nvlist(elem, &propnv) == 0);
711 if (prop == ZFS_PROP_QUOTA || prop == ZFS_PROP_RESERVATION ||
712 prop == ZFS_PROP_REFQUOTA ||
713 prop == ZFS_PROP_REFRESERVATION) {
716 verify(nvlist_lookup_uint64(propnv,
717 ZPROP_VALUE, &value) == 0);
718 if (zhp->zfs_type == ZFS_TYPE_SNAPSHOT)
721 * May have no source before SPA_VERSION_RECVD_PROPS,
722 * but is still modifiable.
724 if (nvlist_lookup_string(propnv,
725 ZPROP_SOURCE, &source) == 0) {
726 if ((strcmp(source, zhp->zfs_name) != 0) &&
728 ZPROP_SOURCE_VAL_RECVD) != 0))
733 if (nvlist_lookup_string(propnv,
734 ZPROP_SOURCE, &source) != 0)
736 if ((strcmp(source, zhp->zfs_name) != 0) &&
737 (strcmp(source, ZPROP_SOURCE_VAL_RECVD) != 0))
741 if (zfs_prop_user(propname) ||
742 zfs_prop_get_type(prop) == PROP_TYPE_STRING) {
744 verify(nvlist_lookup_string(propnv,
745 ZPROP_VALUE, &value) == 0);
746 VERIFY(0 == nvlist_add_string(nv, propname, value));
749 verify(nvlist_lookup_uint64(propnv,
750 ZPROP_VALUE, &value) == 0);
751 VERIFY(0 == nvlist_add_uint64(nv, propname, value));
757 * returns snapshot creation txg
758 * and returns 0 if the snapshot does not exist
761 get_snap_txg(libzfs_handle_t *hdl, const char *fs, const char *snap)
763 char name[ZFS_MAX_DATASET_NAME_LEN];
766 if (fs == NULL || fs[0] == '\0' || snap == NULL || snap[0] == '\0')
769 (void) snprintf(name, sizeof (name), "%s@%s", fs, snap);
770 if (zfs_dataset_exists(hdl, name, ZFS_TYPE_SNAPSHOT)) {
771 zfs_handle_t *zhp = zfs_open(hdl, name, ZFS_TYPE_SNAPSHOT);
773 txg = zfs_prop_get_int(zhp, ZFS_PROP_CREATETXG);
782 * recursively generate nvlists describing datasets. See comment
783 * for the data structure send_data_t above for description of contents
787 send_iterate_fs(zfs_handle_t *zhp, void *arg)
789 send_data_t *sd = arg;
792 uint64_t parent_fromsnap_guid_save = sd->parent_fromsnap_guid;
793 uint64_t fromsnap_txg_save = sd->fromsnap_txg;
794 uint64_t tosnap_txg_save = sd->tosnap_txg;
795 uint64_t txg = zhp->zfs_dmustats.dds_creation_txg;
796 uint64_t guid = zhp->zfs_dmustats.dds_guid;
797 uint64_t fromsnap_txg, tosnap_txg;
800 fromsnap_txg = get_snap_txg(zhp->zfs_hdl, zhp->zfs_name, sd->fromsnap);
801 if (fromsnap_txg != 0)
802 sd->fromsnap_txg = fromsnap_txg;
804 tosnap_txg = get_snap_txg(zhp->zfs_hdl, zhp->zfs_name, sd->tosnap);
806 sd->tosnap_txg = tosnap_txg;
809 * on the send side, if the current dataset does not have tosnap,
810 * perform two additional checks:
812 * - skip sending the current dataset if it was created later than
814 * - return error if the current dataset was created earlier than
817 if (sd->tosnap != NULL && tosnap_txg == 0) {
818 if (sd->tosnap_txg != 0 && txg > sd->tosnap_txg) {
820 (void) fprintf(stderr, dgettext(TEXT_DOMAIN,
821 "skipping dataset %s: snapshot %s does "
822 "not exist\n"), zhp->zfs_name, sd->tosnap);
825 (void) fprintf(stderr, dgettext(TEXT_DOMAIN,
826 "cannot send %s@%s%s: snapshot %s@%s does not "
827 "exist\n"), sd->fsname, sd->tosnap, sd->recursive ?
828 dgettext(TEXT_DOMAIN, " recursively") : "",
829 zhp->zfs_name, sd->tosnap);
835 VERIFY(0 == nvlist_alloc(&nvfs, NV_UNIQUE_NAME, 0));
836 VERIFY(0 == nvlist_add_string(nvfs, "name", zhp->zfs_name));
837 VERIFY(0 == nvlist_add_uint64(nvfs, "parentfromsnap",
838 sd->parent_fromsnap_guid));
840 if (zhp->zfs_dmustats.dds_origin[0]) {
841 zfs_handle_t *origin = zfs_open(zhp->zfs_hdl,
842 zhp->zfs_dmustats.dds_origin, ZFS_TYPE_SNAPSHOT);
843 if (origin == NULL) {
847 VERIFY(0 == nvlist_add_uint64(nvfs, "origin",
848 origin->zfs_dmustats.dds_guid));
851 /* iterate over props */
852 VERIFY(0 == nvlist_alloc(&nv, NV_UNIQUE_NAME, 0));
853 send_iterate_prop(zhp, nv);
854 VERIFY(0 == nvlist_add_nvlist(nvfs, "props", nv));
857 /* iterate over snaps, and set sd->parent_fromsnap_guid */
858 sd->parent_fromsnap_guid = 0;
859 VERIFY(0 == nvlist_alloc(&sd->parent_snaps, NV_UNIQUE_NAME, 0));
860 VERIFY(0 == nvlist_alloc(&sd->snapprops, NV_UNIQUE_NAME, 0));
861 (void) zfs_iter_snapshots_sorted(zhp, send_iterate_snap, sd);
862 VERIFY(0 == nvlist_add_nvlist(nvfs, "snaps", sd->parent_snaps));
863 VERIFY(0 == nvlist_add_nvlist(nvfs, "snapprops", sd->snapprops));
864 nvlist_free(sd->parent_snaps);
865 nvlist_free(sd->snapprops);
867 /* add this fs to nvlist */
868 (void) snprintf(guidstring, sizeof (guidstring),
869 "0x%llx", (longlong_t)guid);
870 VERIFY(0 == nvlist_add_nvlist(sd->fss, guidstring, nvfs));
873 /* iterate over children */
875 rv = zfs_iter_filesystems(zhp, send_iterate_fs, sd);
878 sd->parent_fromsnap_guid = parent_fromsnap_guid_save;
879 sd->fromsnap_txg = fromsnap_txg_save;
880 sd->tosnap_txg = tosnap_txg_save;
887 gather_nvlist(libzfs_handle_t *hdl, const char *fsname, const char *fromsnap,
888 const char *tosnap, boolean_t recursive, boolean_t verbose,
889 nvlist_t **nvlp, avl_tree_t **avlp)
892 send_data_t sd = { 0 };
895 zhp = zfs_open(hdl, fsname, ZFS_TYPE_FILESYSTEM | ZFS_TYPE_VOLUME);
897 return (EZFS_BADTYPE);
899 VERIFY(0 == nvlist_alloc(&sd.fss, NV_UNIQUE_NAME, 0));
901 sd.fromsnap = fromsnap;
903 sd.recursive = recursive;
904 sd.verbose = verbose;
906 if ((error = send_iterate_fs(zhp, &sd)) != 0) {
914 if (avlp != NULL && (*avlp = fsavl_create(sd.fss)) == NULL) {
925 * Routines specific to "zfs send"
927 typedef struct send_dump_data {
928 /* these are all just the short snapname (the part after the @) */
929 const char *fromsnap;
931 char prevsnap[ZFS_MAX_DATASET_NAME_LEN];
932 uint64_t prevsnap_obj;
933 boolean_t seenfrom, seento, replicate, doall, fromorigin;
934 boolean_t verbose, dryrun, parsable, progress, embed_data, std_out;
935 boolean_t progressastitle;
936 boolean_t large_block, compress;
942 snapfilter_cb_t *filter_cb;
945 char holdtag[ZFS_MAX_DATASET_NAME_LEN];
951 zfs_send_space(zfs_handle_t *zhp, const char *snapname, const char *from,
952 enum lzc_send_flags flags, uint64_t *spacep)
954 libzfs_handle_t *hdl = zhp->zfs_hdl;
957 assert(snapname != NULL);
958 error = lzc_send_space(snapname, from, flags, spacep);
962 (void) snprintf(errbuf, sizeof (errbuf), dgettext(TEXT_DOMAIN,
963 "warning: cannot estimate space for '%s'"), snapname);
967 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
968 "not an earlier snapshot from the same fs"));
969 return (zfs_error(hdl, EZFS_CROSSTARGET, errbuf));
972 if (zfs_dataset_exists(hdl, snapname,
973 ZFS_TYPE_SNAPSHOT)) {
974 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
975 "incremental source (%s) does not exist"),
978 return (zfs_error(hdl, EZFS_NOENT, errbuf));
991 zfs_error_aux(hdl, strerror(error));
992 return (zfs_error(hdl, EZFS_BADBACKUP, errbuf));
995 return (zfs_standard_error(hdl, error, errbuf));
1003 * Dumps a backup of the given snapshot (incremental from fromsnap if it's not
1004 * NULL) to the file descriptor specified by outfd.
1007 dump_ioctl(zfs_handle_t *zhp, const char *fromsnap, uint64_t fromsnap_obj,
1008 boolean_t fromorigin, int outfd, enum lzc_send_flags flags,
1011 zfs_cmd_t zc = { 0 };
1012 libzfs_handle_t *hdl = zhp->zfs_hdl;
1015 assert(zhp->zfs_type == ZFS_TYPE_SNAPSHOT);
1016 assert(fromsnap_obj == 0 || !fromorigin);
1018 (void) strlcpy(zc.zc_name, zhp->zfs_name, sizeof (zc.zc_name));
1019 zc.zc_cookie = outfd;
1020 zc.zc_obj = fromorigin;
1021 zc.zc_sendobj = zfs_prop_get_int(zhp, ZFS_PROP_OBJSETID);
1022 zc.zc_fromobj = fromsnap_obj;
1023 zc.zc_flags = flags;
1025 VERIFY(0 == nvlist_alloc(&thisdbg, NV_UNIQUE_NAME, 0));
1026 if (fromsnap && fromsnap[0] != '\0') {
1027 VERIFY(0 == nvlist_add_string(thisdbg,
1028 "fromsnap", fromsnap));
1031 if (zfs_ioctl(zhp->zfs_hdl, ZFS_IOC_SEND, &zc) != 0) {
1033 (void) snprintf(errbuf, sizeof (errbuf), dgettext(TEXT_DOMAIN,
1034 "warning: cannot send '%s'"), zhp->zfs_name);
1036 VERIFY(0 == nvlist_add_uint64(thisdbg, "error", errno));
1038 VERIFY(0 == nvlist_add_nvlist(debugnv,
1039 zhp->zfs_name, thisdbg));
1041 nvlist_free(thisdbg);
1045 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
1046 "not an earlier snapshot from the same fs"));
1047 return (zfs_error(hdl, EZFS_CROSSTARGET, errbuf));
1050 if (zfs_dataset_exists(hdl, zc.zc_name,
1051 ZFS_TYPE_SNAPSHOT)) {
1052 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
1053 "incremental source (@%s) does not exist"),
1056 return (zfs_error(hdl, EZFS_NOENT, errbuf));
1071 zfs_error_aux(hdl, strerror(errno));
1072 return (zfs_error(hdl, EZFS_BADBACKUP, errbuf));
1075 return (zfs_standard_error(hdl, errno, errbuf));
1080 VERIFY(0 == nvlist_add_nvlist(debugnv, zhp->zfs_name, thisdbg));
1081 nvlist_free(thisdbg);
1087 gather_holds(zfs_handle_t *zhp, send_dump_data_t *sdd)
1089 assert(zhp->zfs_type == ZFS_TYPE_SNAPSHOT);
1092 * zfs_send() only sets snapholds for sends that need them,
1093 * e.g. replication and doall.
1095 if (sdd->snapholds == NULL)
1098 fnvlist_add_string(sdd->snapholds, zhp->zfs_name, sdd->holdtag);
1102 send_progress_thread(void *arg)
1104 progress_arg_t *pa = arg;
1105 zfs_cmd_t zc = { 0 };
1106 zfs_handle_t *zhp = pa->pa_zhp;
1107 libzfs_handle_t *hdl = zhp->zfs_hdl;
1108 unsigned long long bytes, total;
1113 (void) strlcpy(zc.zc_name, zhp->zfs_name, sizeof (zc.zc_name));
1115 if (!pa->pa_parsable && !pa->pa_astitle)
1116 (void) fprintf(stderr, "TIME SENT SNAPSHOT\n");
1119 * Print the progress from ZFS_IOC_SEND_PROGRESS every second.
1124 zc.zc_cookie = pa->pa_fd;
1125 if (zfs_ioctl(hdl, ZFS_IOC_SEND_PROGRESS, &zc) != 0)
1126 return ((void *)-1);
1130 bytes = zc.zc_cookie;
1132 if (pa->pa_astitle) {
1134 if (pa->pa_size > bytes)
1135 pct = 100 * bytes / pa->pa_size;
1139 setproctitle("sending %s (%d%%: %llu/%llu)",
1140 zhp->zfs_name, pct, bytes, pa->pa_size);
1141 } else if (pa->pa_parsable) {
1142 (void) fprintf(stderr, "%02d:%02d:%02d\t%llu\t%s\n",
1143 tm->tm_hour, tm->tm_min, tm->tm_sec,
1144 bytes, zhp->zfs_name);
1146 zfs_nicenum(bytes, buf, sizeof (buf));
1147 (void) fprintf(stderr, "%02d:%02d:%02d %5s %s\n",
1148 tm->tm_hour, tm->tm_min, tm->tm_sec,
1149 buf, zhp->zfs_name);
1155 send_print_verbose(FILE *fout, const char *tosnap, const char *fromsnap,
1156 uint64_t size, boolean_t parsable)
1159 if (fromsnap != NULL) {
1160 (void) fprintf(fout, "incremental\t%s\t%s",
1163 (void) fprintf(fout, "full\t%s",
1167 if (fromsnap != NULL) {
1168 if (strchr(fromsnap, '@') == NULL &&
1169 strchr(fromsnap, '#') == NULL) {
1170 (void) fprintf(fout, dgettext(TEXT_DOMAIN,
1171 "send from @%s to %s"),
1174 (void) fprintf(fout, dgettext(TEXT_DOMAIN,
1175 "send from %s to %s"),
1179 (void) fprintf(fout, dgettext(TEXT_DOMAIN,
1186 (void) fprintf(fout, "\t%llu",
1188 } else if (size != 0) {
1190 zfs_nicenum(size, buf, sizeof (buf));
1191 (void) fprintf(fout, dgettext(TEXT_DOMAIN,
1192 " estimated size is %s"), buf);
1194 (void) fprintf(fout, "\n");
1198 dump_snapshot(zfs_handle_t *zhp, void *arg)
1200 send_dump_data_t *sdd = arg;
1201 progress_arg_t pa = { 0 };
1204 enum lzc_send_flags flags = 0;
1206 boolean_t isfromsnap, istosnap, fromorigin;
1207 boolean_t exclude = B_FALSE;
1208 FILE *fout = sdd->std_out ? stdout : stderr;
1212 thissnap = strchr(zhp->zfs_name, '@') + 1;
1213 isfromsnap = (sdd->fromsnap != NULL &&
1214 strcmp(sdd->fromsnap, thissnap) == 0);
1216 if (!sdd->seenfrom && isfromsnap) {
1217 gather_holds(zhp, sdd);
1218 sdd->seenfrom = B_TRUE;
1219 (void) strcpy(sdd->prevsnap, thissnap);
1220 sdd->prevsnap_obj = zfs_prop_get_int(zhp, ZFS_PROP_OBJSETID);
1225 if (sdd->seento || !sdd->seenfrom) {
1230 istosnap = (strcmp(sdd->tosnap, thissnap) == 0);
1232 sdd->seento = B_TRUE;
1234 if (sdd->large_block)
1235 flags |= LZC_SEND_FLAG_LARGE_BLOCK;
1236 if (sdd->embed_data)
1237 flags |= LZC_SEND_FLAG_EMBED_DATA;
1239 flags |= LZC_SEND_FLAG_COMPRESS;
1241 if (!sdd->doall && !isfromsnap && !istosnap) {
1242 if (sdd->replicate) {
1244 nvlist_t *snapprops;
1246 * Filter out all intermediate snapshots except origin
1247 * snapshots needed to replicate clones.
1249 nvlist_t *nvfs = fsavl_find(sdd->fsavl,
1250 zhp->zfs_dmustats.dds_guid, &snapname);
1252 VERIFY(0 == nvlist_lookup_nvlist(nvfs,
1253 "snapprops", &snapprops));
1254 VERIFY(0 == nvlist_lookup_nvlist(snapprops,
1255 thissnap, &snapprops));
1256 exclude = !nvlist_exists(snapprops, "is_clone_origin");
1263 * If a filter function exists, call it to determine whether
1264 * this snapshot will be sent.
1266 if (exclude || (sdd->filter_cb != NULL &&
1267 sdd->filter_cb(zhp, sdd->filter_cb_arg) == B_FALSE)) {
1269 * This snapshot is filtered out. Don't send it, and don't
1270 * set prevsnap_obj, so it will be as if this snapshot didn't
1271 * exist, and the next accepted snapshot will be sent as
1272 * an incremental from the last accepted one, or as the
1273 * first (and full) snapshot in the case of a replication,
1274 * non-incremental send.
1280 gather_holds(zhp, sdd);
1281 fromorigin = sdd->prevsnap[0] == '\0' &&
1282 (sdd->fromorigin || sdd->replicate);
1284 if (sdd->verbose || sdd->progress) {
1285 char fromds[ZFS_MAX_DATASET_NAME_LEN];
1287 if (sdd->prevsnap[0] != '\0') {
1288 (void) strlcpy(fromds, zhp->zfs_name, sizeof (fromds));
1289 *(strchr(fromds, '@') + 1) = '\0';
1290 (void) strlcat(fromds, sdd->prevsnap, sizeof (fromds));
1292 if (zfs_send_space(zhp, zhp->zfs_name,
1293 sdd->prevsnap[0] ? fromds : NULL, flags, &size) != 0) {
1294 size = 0; /* cannot estimate send space */
1296 send_print_verbose(fout, zhp->zfs_name,
1297 sdd->prevsnap[0] ? sdd->prevsnap : NULL,
1298 size, sdd->parsable);
1305 * If progress reporting is requested, spawn a new thread to
1306 * poll ZFS_IOC_SEND_PROGRESS at a regular interval.
1308 if (sdd->progress) {
1310 pa.pa_fd = sdd->outfd;
1311 pa.pa_parsable = sdd->parsable;
1312 pa.pa_size = sdd->size;
1313 pa.pa_astitle = sdd->progressastitle;
1315 if ((err = pthread_create(&tid, NULL,
1316 send_progress_thread, &pa)) != 0) {
1322 err = dump_ioctl(zhp, sdd->prevsnap, sdd->prevsnap_obj,
1323 fromorigin, sdd->outfd, flags, sdd->debugnv);
1325 if (sdd->progress) {
1326 (void) pthread_cancel(tid);
1327 (void) pthread_join(tid, NULL);
1331 (void) strcpy(sdd->prevsnap, thissnap);
1332 sdd->prevsnap_obj = zfs_prop_get_int(zhp, ZFS_PROP_OBJSETID);
1338 dump_filesystem(zfs_handle_t *zhp, void *arg)
1341 send_dump_data_t *sdd = arg;
1342 boolean_t missingfrom = B_FALSE;
1343 zfs_cmd_t zc = { 0 };
1345 (void) snprintf(zc.zc_name, sizeof (zc.zc_name), "%s@%s",
1346 zhp->zfs_name, sdd->tosnap);
1347 if (ioctl(zhp->zfs_hdl->libzfs_fd, ZFS_IOC_OBJSET_STATS, &zc) != 0) {
1348 (void) fprintf(stderr, dgettext(TEXT_DOMAIN,
1349 "WARNING: could not send %s@%s: does not exist\n"),
1350 zhp->zfs_name, sdd->tosnap);
1355 if (sdd->replicate && sdd->fromsnap) {
1357 * If this fs does not have fromsnap, and we're doing
1358 * recursive, we need to send a full stream from the
1359 * beginning (or an incremental from the origin if this
1360 * is a clone). If we're doing non-recursive, then let
1361 * them get the error.
1363 (void) snprintf(zc.zc_name, sizeof (zc.zc_name), "%s@%s",
1364 zhp->zfs_name, sdd->fromsnap);
1365 if (ioctl(zhp->zfs_hdl->libzfs_fd,
1366 ZFS_IOC_OBJSET_STATS, &zc) != 0) {
1367 missingfrom = B_TRUE;
1371 sdd->seenfrom = sdd->seento = sdd->prevsnap[0] = 0;
1372 sdd->prevsnap_obj = 0;
1373 if (sdd->fromsnap == NULL || missingfrom)
1374 sdd->seenfrom = B_TRUE;
1376 rv = zfs_iter_snapshots_sorted(zhp, dump_snapshot, arg);
1377 if (!sdd->seenfrom) {
1378 (void) fprintf(stderr, dgettext(TEXT_DOMAIN,
1379 "WARNING: could not send %s@%s:\n"
1380 "incremental source (%s@%s) does not exist\n"),
1381 zhp->zfs_name, sdd->tosnap,
1382 zhp->zfs_name, sdd->fromsnap);
1384 } else if (!sdd->seento) {
1385 if (sdd->fromsnap) {
1386 (void) fprintf(stderr, dgettext(TEXT_DOMAIN,
1387 "WARNING: could not send %s@%s:\n"
1388 "incremental source (%s@%s) "
1389 "is not earlier than it\n"),
1390 zhp->zfs_name, sdd->tosnap,
1391 zhp->zfs_name, sdd->fromsnap);
1393 (void) fprintf(stderr, dgettext(TEXT_DOMAIN,
1395 "could not send %s@%s: does not exist\n"),
1396 zhp->zfs_name, sdd->tosnap);
1405 dump_filesystems(zfs_handle_t *rzhp, void *arg)
1407 send_dump_data_t *sdd = arg;
1409 boolean_t needagain, progress;
1411 if (!sdd->replicate)
1412 return (dump_filesystem(rzhp, sdd));
1414 /* Mark the clone origin snapshots. */
1415 for (fspair = nvlist_next_nvpair(sdd->fss, NULL); fspair;
1416 fspair = nvlist_next_nvpair(sdd->fss, fspair)) {
1418 uint64_t origin_guid = 0;
1420 VERIFY(0 == nvpair_value_nvlist(fspair, &nvfs));
1421 (void) nvlist_lookup_uint64(nvfs, "origin", &origin_guid);
1422 if (origin_guid != 0) {
1424 nvlist_t *origin_nv = fsavl_find(sdd->fsavl,
1425 origin_guid, &snapname);
1426 if (origin_nv != NULL) {
1427 nvlist_t *snapprops;
1428 VERIFY(0 == nvlist_lookup_nvlist(origin_nv,
1429 "snapprops", &snapprops));
1430 VERIFY(0 == nvlist_lookup_nvlist(snapprops,
1431 snapname, &snapprops));
1432 VERIFY(0 == nvlist_add_boolean(
1433 snapprops, "is_clone_origin"));
1438 needagain = progress = B_FALSE;
1439 for (fspair = nvlist_next_nvpair(sdd->fss, NULL); fspair;
1440 fspair = nvlist_next_nvpair(sdd->fss, fspair)) {
1441 nvlist_t *fslist, *parent_nv;
1445 uint64_t origin_guid = 0;
1446 uint64_t parent_guid = 0;
1448 VERIFY(nvpair_value_nvlist(fspair, &fslist) == 0);
1449 if (nvlist_lookup_boolean(fslist, "sent") == 0)
1452 VERIFY(nvlist_lookup_string(fslist, "name", &fsname) == 0);
1453 (void) nvlist_lookup_uint64(fslist, "origin", &origin_guid);
1454 (void) nvlist_lookup_uint64(fslist, "parentfromsnap",
1457 if (parent_guid != 0) {
1458 parent_nv = fsavl_find(sdd->fsavl, parent_guid, NULL);
1459 if (!nvlist_exists(parent_nv, "sent")) {
1460 /* parent has not been sent; skip this one */
1466 if (origin_guid != 0) {
1467 nvlist_t *origin_nv = fsavl_find(sdd->fsavl,
1469 if (origin_nv != NULL &&
1470 !nvlist_exists(origin_nv, "sent")) {
1472 * origin has not been sent yet;
1480 zhp = zfs_open(rzhp->zfs_hdl, fsname, ZFS_TYPE_DATASET);
1483 err = dump_filesystem(zhp, sdd);
1484 VERIFY(nvlist_add_boolean(fslist, "sent") == 0);
1495 /* clean out the sent flags in case we reuse this fss */
1496 for (fspair = nvlist_next_nvpair(sdd->fss, NULL); fspair;
1497 fspair = nvlist_next_nvpair(sdd->fss, fspair)) {
1500 VERIFY(nvpair_value_nvlist(fspair, &fslist) == 0);
1501 (void) nvlist_remove_all(fslist, "sent");
1508 zfs_send_resume_token_to_nvlist(libzfs_handle_t *hdl, const char *token)
1510 unsigned int version;
1512 unsigned long long checksum, packed_len;
1515 * Decode token header, which is:
1516 * <token version>-<checksum of payload>-<uncompressed payload length>
1517 * Note that the only supported token version is 1.
1519 nread = sscanf(token, "%u-%llx-%llx-",
1520 &version, &checksum, &packed_len);
1522 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
1523 "resume token is corrupt (invalid format)"));
1527 if (version != ZFS_SEND_RESUME_TOKEN_VERSION) {
1528 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
1529 "resume token is corrupt (invalid version %u)"),
1534 /* convert hexadecimal representation to binary */
1535 token = strrchr(token, '-') + 1;
1536 int len = strlen(token) / 2;
1537 unsigned char *compressed = zfs_alloc(hdl, len);
1538 for (int i = 0; i < len; i++) {
1539 nread = sscanf(token + i * 2, "%2hhx", compressed + i);
1542 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
1543 "resume token is corrupt "
1544 "(payload is not hex-encoded)"));
1549 /* verify checksum */
1551 fletcher_4_native(compressed, len, NULL, &cksum);
1552 if (cksum.zc_word[0] != checksum) {
1554 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
1555 "resume token is corrupt (incorrect checksum)"));
1560 void *packed = zfs_alloc(hdl, packed_len);
1561 uLongf packed_len_long = packed_len;
1562 if (uncompress(packed, &packed_len_long, compressed, len) != Z_OK ||
1563 packed_len_long != packed_len) {
1566 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
1567 "resume token is corrupt (decompression failed)"));
1573 int error = nvlist_unpack(packed, packed_len, &nv, KM_SLEEP);
1577 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
1578 "resume token is corrupt (nvlist_unpack failed)"));
1585 zfs_send_resume(libzfs_handle_t *hdl, sendflags_t *flags, int outfd,
1586 const char *resume_token)
1590 char *fromname = NULL;
1591 uint64_t resumeobj, resumeoff, toguid, fromguid, bytes;
1594 char name[ZFS_MAX_DATASET_NAME_LEN];
1595 enum lzc_send_flags lzc_flags = 0;
1597 FILE *fout = (flags->verbose && flags->dryrun) ? stdout : stderr;
1599 (void) snprintf(errbuf, sizeof (errbuf), dgettext(TEXT_DOMAIN,
1600 "cannot resume send"));
1602 nvlist_t *resume_nvl =
1603 zfs_send_resume_token_to_nvlist(hdl, resume_token);
1604 if (resume_nvl == NULL) {
1606 * zfs_error_aux has already been set by
1607 * zfs_send_resume_token_to_nvlist
1609 return (zfs_error(hdl, EZFS_FAULT, errbuf));
1611 if (flags->verbose) {
1612 (void) fprintf(fout, dgettext(TEXT_DOMAIN,
1613 "resume token contents:\n"));
1614 nvlist_print(fout, resume_nvl);
1617 if (nvlist_lookup_string(resume_nvl, "toname", &toname) != 0 ||
1618 nvlist_lookup_uint64(resume_nvl, "object", &resumeobj) != 0 ||
1619 nvlist_lookup_uint64(resume_nvl, "offset", &resumeoff) != 0 ||
1620 nvlist_lookup_uint64(resume_nvl, "bytes", &bytes) != 0 ||
1621 nvlist_lookup_uint64(resume_nvl, "toguid", &toguid) != 0) {
1622 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
1623 "resume token is corrupt"));
1624 return (zfs_error(hdl, EZFS_FAULT, errbuf));
1627 (void) nvlist_lookup_uint64(resume_nvl, "fromguid", &fromguid);
1629 if (flags->largeblock || nvlist_exists(resume_nvl, "largeblockok"))
1630 lzc_flags |= LZC_SEND_FLAG_LARGE_BLOCK;
1631 if (flags->embed_data || nvlist_exists(resume_nvl, "embedok"))
1632 lzc_flags |= LZC_SEND_FLAG_EMBED_DATA;
1633 if (flags->compress || nvlist_exists(resume_nvl, "compressok"))
1634 lzc_flags |= LZC_SEND_FLAG_COMPRESS;
1636 if (guid_to_name(hdl, toname, toguid, B_FALSE, name) != 0) {
1637 if (zfs_dataset_exists(hdl, toname, ZFS_TYPE_DATASET)) {
1638 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
1639 "'%s' is no longer the same snapshot used in "
1640 "the initial send"), toname);
1642 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
1643 "'%s' used in the initial send no longer exists"),
1646 return (zfs_error(hdl, EZFS_BADPATH, errbuf));
1648 zhp = zfs_open(hdl, name, ZFS_TYPE_DATASET);
1650 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
1651 "unable to access '%s'"), name);
1652 return (zfs_error(hdl, EZFS_BADPATH, errbuf));
1655 if (fromguid != 0) {
1656 if (guid_to_name(hdl, toname, fromguid, B_TRUE, name) != 0) {
1657 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
1658 "incremental source %#llx no longer exists"),
1659 (longlong_t)fromguid);
1660 return (zfs_error(hdl, EZFS_BADPATH, errbuf));
1665 if (flags->progress || flags->verbose) {
1666 error = lzc_send_space(zhp->zfs_name, fromname,
1669 size = MAX(0, (int64_t)(size - bytes));
1671 if (flags->verbose) {
1672 send_print_verbose(fout, zhp->zfs_name, fromname,
1673 size, flags->parsable);
1676 if (!flags->dryrun) {
1677 progress_arg_t pa = { 0 };
1680 * If progress reporting is requested, spawn a new thread to
1681 * poll ZFS_IOC_SEND_PROGRESS at a regular interval.
1683 if (flags->progress) {
1686 pa.pa_parsable = flags->parsable;
1688 pa.pa_astitle = flags->progressastitle;
1690 error = pthread_create(&tid, NULL,
1691 send_progress_thread, &pa);
1698 error = lzc_send_resume(zhp->zfs_name, fromname, outfd,
1699 lzc_flags, resumeobj, resumeoff);
1701 if (flags->progress) {
1702 (void) pthread_cancel(tid);
1703 (void) pthread_join(tid, NULL);
1707 (void) snprintf(errbuf, sizeof (errbuf), dgettext(TEXT_DOMAIN,
1708 "warning: cannot send '%s'"), zhp->zfs_name);
1730 zfs_error_aux(hdl, strerror(errno));
1731 return (zfs_error(hdl, EZFS_BADBACKUP, errbuf));
1734 return (zfs_standard_error(hdl, errno, errbuf));
1745 * Generate a send stream for the dataset identified by the argument zhp.
1747 * The content of the send stream is the snapshot identified by
1748 * 'tosnap'. Incremental streams are requested in two ways:
1749 * - from the snapshot identified by "fromsnap" (if non-null) or
1750 * - from the origin of the dataset identified by zhp, which must
1751 * be a clone. In this case, "fromsnap" is null and "fromorigin"
1754 * The send stream is recursive (i.e. dumps a hierarchy of snapshots) and
1755 * uses a special header (with a hdrtype field of DMU_COMPOUNDSTREAM)
1756 * if "replicate" is set. If "doall" is set, dump all the intermediate
1757 * snapshots. The DMU_COMPOUNDSTREAM header is used in the "doall"
1758 * case too. If "props" is set, send properties.
1761 zfs_send(zfs_handle_t *zhp, const char *fromsnap, const char *tosnap,
1762 sendflags_t *flags, int outfd, snapfilter_cb_t filter_func,
1763 void *cb_arg, nvlist_t **debugnvp)
1766 send_dump_data_t sdd = { 0 };
1768 nvlist_t *fss = NULL;
1769 avl_tree_t *fsavl = NULL;
1770 static uint64_t holdseq;
1774 dedup_arg_t dda = { 0 };
1775 int featureflags = 0;
1778 (void) snprintf(errbuf, sizeof (errbuf), dgettext(TEXT_DOMAIN,
1779 "cannot send '%s'"), zhp->zfs_name);
1781 if (fromsnap && fromsnap[0] == '\0') {
1782 zfs_error_aux(zhp->zfs_hdl, dgettext(TEXT_DOMAIN,
1783 "zero-length incremental source"));
1784 return (zfs_error(zhp->zfs_hdl, EZFS_NOENT, errbuf));
1787 if (zhp->zfs_type == ZFS_TYPE_FILESYSTEM) {
1789 version = zfs_prop_get_int(zhp, ZFS_PROP_VERSION);
1790 if (version >= ZPL_VERSION_SA) {
1791 featureflags |= DMU_BACKUP_FEATURE_SA_SPILL;
1795 if (flags->dedup && !flags->dryrun) {
1796 featureflags |= (DMU_BACKUP_FEATURE_DEDUP |
1797 DMU_BACKUP_FEATURE_DEDUPPROPS);
1798 if ((err = pipe(pipefd)) != 0) {
1799 zfs_error_aux(zhp->zfs_hdl, strerror(errno));
1800 return (zfs_error(zhp->zfs_hdl, EZFS_PIPEFAILED,
1803 dda.outputfd = outfd;
1804 dda.inputfd = pipefd[1];
1805 dda.dedup_hdl = zhp->zfs_hdl;
1806 if ((err = pthread_create(&tid, NULL, cksummer, &dda)) != 0) {
1807 (void) close(pipefd[0]);
1808 (void) close(pipefd[1]);
1809 zfs_error_aux(zhp->zfs_hdl, strerror(errno));
1810 return (zfs_error(zhp->zfs_hdl,
1811 EZFS_THREADCREATEFAILED, errbuf));
1815 if (flags->replicate || flags->doall || flags->props) {
1816 dmu_replay_record_t drr = { 0 };
1817 char *packbuf = NULL;
1819 zio_cksum_t zc = { 0 };
1821 if (flags->replicate || flags->props) {
1824 VERIFY(0 == nvlist_alloc(&hdrnv, NV_UNIQUE_NAME, 0));
1826 VERIFY(0 == nvlist_add_string(hdrnv,
1827 "fromsnap", fromsnap));
1829 VERIFY(0 == nvlist_add_string(hdrnv, "tosnap", tosnap));
1830 if (!flags->replicate) {
1831 VERIFY(0 == nvlist_add_boolean(hdrnv,
1835 err = gather_nvlist(zhp->zfs_hdl, zhp->zfs_name,
1836 fromsnap, tosnap, flags->replicate, flags->verbose,
1840 VERIFY(0 == nvlist_add_nvlist(hdrnv, "fss", fss));
1841 err = nvlist_pack(hdrnv, &packbuf, &buflen,
1851 if (!flags->dryrun) {
1852 /* write first begin record */
1853 drr.drr_type = DRR_BEGIN;
1854 drr.drr_u.drr_begin.drr_magic = DMU_BACKUP_MAGIC;
1855 DMU_SET_STREAM_HDRTYPE(drr.drr_u.drr_begin.
1856 drr_versioninfo, DMU_COMPOUNDSTREAM);
1857 DMU_SET_FEATUREFLAGS(drr.drr_u.drr_begin.
1858 drr_versioninfo, featureflags);
1859 (void) snprintf(drr.drr_u.drr_begin.drr_toname,
1860 sizeof (drr.drr_u.drr_begin.drr_toname),
1861 "%s@%s", zhp->zfs_name, tosnap);
1862 drr.drr_payloadlen = buflen;
1864 err = dump_record(&drr, packbuf, buflen, &zc, outfd);
1869 /* write end record */
1870 bzero(&drr, sizeof (drr));
1871 drr.drr_type = DRR_END;
1872 drr.drr_u.drr_end.drr_checksum = zc;
1873 err = write(outfd, &drr, sizeof (drr));
1883 /* dump each stream */
1884 sdd.fromsnap = fromsnap;
1885 sdd.tosnap = tosnap;
1887 sdd.outfd = pipefd[0];
1890 sdd.replicate = flags->replicate;
1891 sdd.doall = flags->doall;
1892 sdd.fromorigin = flags->fromorigin;
1895 sdd.verbose = flags->verbose;
1896 sdd.parsable = flags->parsable;
1897 sdd.progress = flags->progress;
1898 sdd.progressastitle = flags->progressastitle;
1899 sdd.dryrun = flags->dryrun;
1900 sdd.large_block = flags->largeblock;
1901 sdd.embed_data = flags->embed_data;
1902 sdd.compress = flags->compress;
1903 sdd.filter_cb = filter_func;
1904 sdd.filter_cb_arg = cb_arg;
1906 sdd.debugnv = *debugnvp;
1907 if (sdd.verbose && sdd.dryrun)
1908 sdd.std_out = B_TRUE;
1909 fout = sdd.std_out ? stdout : stderr;
1912 * Some flags require that we place user holds on the datasets that are
1913 * being sent so they don't get destroyed during the send. We can skip
1914 * this step if the pool is imported read-only since the datasets cannot
1917 if (!flags->dryrun && !zpool_get_prop_int(zfs_get_pool_handle(zhp),
1918 ZPOOL_PROP_READONLY, NULL) &&
1919 zfs_spa_version(zhp, &spa_version) == 0 &&
1920 spa_version >= SPA_VERSION_USERREFS &&
1921 (flags->doall || flags->replicate)) {
1923 (void) snprintf(sdd.holdtag, sizeof (sdd.holdtag),
1924 ".send-%d-%llu", getpid(), (u_longlong_t)holdseq);
1925 sdd.cleanup_fd = open(ZFS_DEV, O_RDWR|O_EXCL);
1926 if (sdd.cleanup_fd < 0) {
1930 sdd.snapholds = fnvlist_alloc();
1932 sdd.cleanup_fd = -1;
1933 sdd.snapholds = NULL;
1935 if (flags->progress || flags->verbose || sdd.snapholds != NULL) {
1937 * Do a verbose no-op dry run to get all the verbose output
1938 * or to gather snapshot hold's before generating any data,
1939 * then do a non-verbose real run to generate the streams.
1941 sdd.dryrun = B_TRUE;
1942 err = dump_filesystems(zhp, &sdd);
1947 if (flags->verbose) {
1948 if (flags->parsable) {
1949 (void) fprintf(fout, "size\t%llu\n",
1950 (longlong_t)sdd.size);
1953 zfs_nicenum(sdd.size, buf, sizeof (buf));
1954 (void) fprintf(fout, dgettext(TEXT_DOMAIN,
1955 "total estimated size is %s\n"), buf);
1959 /* Ensure no snaps found is treated as an error. */
1965 /* Skip the second run if dryrun was requested. */
1969 if (sdd.snapholds != NULL) {
1970 err = zfs_hold_nvl(zhp, sdd.cleanup_fd, sdd.snapholds);
1974 fnvlist_free(sdd.snapholds);
1975 sdd.snapholds = NULL;
1978 sdd.dryrun = B_FALSE;
1979 sdd.verbose = B_FALSE;
1982 err = dump_filesystems(zhp, &sdd);
1983 fsavl_destroy(fsavl);
1986 /* Ensure no snaps found is treated as an error. */
1987 if (err == 0 && !sdd.seento)
1992 (void) pthread_cancel(tid);
1993 (void) close(pipefd[0]);
1994 (void) pthread_join(tid, NULL);
1997 if (sdd.cleanup_fd != -1) {
1998 VERIFY(0 == close(sdd.cleanup_fd));
1999 sdd.cleanup_fd = -1;
2002 if (!flags->dryrun && (flags->replicate || flags->doall ||
2005 * write final end record. NB: want to do this even if
2006 * there was some error, because it might not be totally
2009 dmu_replay_record_t drr = { 0 };
2010 drr.drr_type = DRR_END;
2011 if (write(outfd, &drr, sizeof (drr)) == -1) {
2012 return (zfs_standard_error(zhp->zfs_hdl,
2017 return (err || sdd.err);
2020 err = zfs_standard_error(zhp->zfs_hdl, err, errbuf);
2022 fsavl_destroy(fsavl);
2024 fnvlist_free(sdd.snapholds);
2026 if (sdd.cleanup_fd != -1)
2027 VERIFY(0 == close(sdd.cleanup_fd));
2029 (void) pthread_cancel(tid);
2030 (void) close(pipefd[0]);
2031 (void) pthread_join(tid, NULL);
2037 zfs_send_one(zfs_handle_t *zhp, const char *from, int fd, sendflags_t flags)
2040 libzfs_handle_t *hdl = zhp->zfs_hdl;
2041 enum lzc_send_flags lzc_flags = 0;
2042 FILE *fout = (flags.verbose && flags.dryrun) ? stdout : stderr;
2045 if (flags.largeblock)
2046 lzc_flags |= LZC_SEND_FLAG_LARGE_BLOCK;
2047 if (flags.embed_data)
2048 lzc_flags |= LZC_SEND_FLAG_EMBED_DATA;
2050 lzc_flags |= LZC_SEND_FLAG_COMPRESS;
2052 if (flags.verbose) {
2054 err = lzc_send_space(zhp->zfs_name, from, lzc_flags, &size);
2056 send_print_verbose(fout, zhp->zfs_name, from, size,
2058 if (flags.parsable) {
2059 (void) fprintf(fout, "size\t%llu\n",
2063 zfs_nicenum(size, buf, sizeof (buf));
2064 (void) fprintf(fout, dgettext(TEXT_DOMAIN,
2065 "total estimated size is %s\n"), buf);
2068 (void) fprintf(stderr, "Cannot estimate send size: "
2069 "%s\n", strerror(errno));
2076 (void) snprintf(errbuf, sizeof (errbuf), dgettext(TEXT_DOMAIN,
2077 "warning: cannot send '%s'"), zhp->zfs_name);
2079 err = lzc_send(zhp->zfs_name, from, fd, lzc_flags);
2083 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
2084 "not an earlier snapshot from the same fs"));
2085 return (zfs_error(hdl, EZFS_CROSSTARGET, errbuf));
2089 if (lzc_exists(zhp->zfs_name)) {
2090 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
2091 "incremental source (%s) does not exist"),
2094 return (zfs_error(hdl, EZFS_NOENT, errbuf));
2097 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
2098 "target is busy; if a filesystem, "
2099 "it must not be mounted"));
2100 return (zfs_error(hdl, EZFS_BUSY, errbuf));
2115 zfs_error_aux(hdl, strerror(errno));
2116 return (zfs_error(hdl, EZFS_BADBACKUP, errbuf));
2119 return (zfs_standard_error(hdl, errno, errbuf));
2126 * Routines specific to "zfs recv"
2130 recv_read(libzfs_handle_t *hdl, int fd, void *buf, int ilen,
2131 boolean_t byteswap, zio_cksum_t *zc)
2137 assert(ilen <= SPA_MAXBLOCKSIZE);
2140 rv = read(fd, cp, len);
2145 if (rv < 0 || len != 0) {
2146 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
2147 "failed to read from stream"));
2148 return (zfs_error(hdl, EZFS_BADSTREAM, dgettext(TEXT_DOMAIN,
2149 "cannot receive")));
2154 (void) fletcher_4_incremental_byteswap(buf, ilen, zc);
2156 (void) fletcher_4_incremental_native(buf, ilen, zc);
2162 recv_read_nvlist(libzfs_handle_t *hdl, int fd, int len, nvlist_t **nvp,
2163 boolean_t byteswap, zio_cksum_t *zc)
2168 buf = zfs_alloc(hdl, len);
2172 err = recv_read(hdl, fd, buf, len, byteswap, zc);
2178 err = nvlist_unpack(buf, len, nvp, 0);
2181 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, "invalid "
2182 "stream (malformed nvlist)"));
2189 recv_rename(libzfs_handle_t *hdl, const char *name, const char *tryname,
2190 int baselen, char *newname, recvflags_t *flags)
2194 prop_changelist_t *clp;
2197 zhp = zfs_open(hdl, name, ZFS_TYPE_DATASET);
2200 clp = changelist_gather(zhp, ZFS_PROP_NAME, 0,
2201 flags->force ? MS_FORCE : 0);
2205 err = changelist_prefix(clp);
2210 (void) strcpy(newname, tryname);
2211 if (flags->verbose) {
2212 (void) printf("attempting rename %s to %s\n",
2215 err = lzc_rename(name, newname);
2217 changelist_rename(clp, name, tryname);
2222 if (err != 0 && strncmp(name + baselen, "recv-", 5) != 0) {
2225 (void) snprintf(newname, ZFS_MAX_DATASET_NAME_LEN,
2226 "%.*srecv-%u-%u", baselen, name, getpid(), seq);
2227 if (flags->verbose) {
2228 (void) printf("failed - trying rename %s to %s\n",
2231 err = lzc_rename(name, newname);
2233 changelist_rename(clp, name, newname);
2234 if (err && flags->verbose) {
2235 (void) printf("failed (%u) - "
2236 "will try again on next pass\n", errno);
2239 } else if (flags->verbose) {
2241 (void) printf("success\n");
2243 (void) printf("failed (%u)\n", errno);
2246 (void) changelist_postfix(clp);
2247 changelist_free(clp);
2253 recv_destroy(libzfs_handle_t *hdl, const char *name, int baselen,
2254 char *newname, recvflags_t *flags)
2257 prop_changelist_t *clp;
2259 boolean_t defer = B_FALSE;
2262 zhp = zfs_open(hdl, name, ZFS_TYPE_DATASET);
2265 clp = changelist_gather(zhp, ZFS_PROP_NAME, 0,
2266 flags->force ? MS_FORCE : 0);
2267 if (zfs_get_type(zhp) == ZFS_TYPE_SNAPSHOT &&
2268 zfs_spa_version(zhp, &spa_version) == 0 &&
2269 spa_version >= SPA_VERSION_USERREFS)
2274 err = changelist_prefix(clp);
2279 (void) printf("attempting destroy %s\n", name);
2280 if (zhp->zfs_type == ZFS_TYPE_SNAPSHOT) {
2281 nvlist_t *nv = fnvlist_alloc();
2282 fnvlist_add_boolean(nv, name);
2283 err = lzc_destroy_snaps(nv, defer, NULL);
2286 err = lzc_destroy(name);
2290 (void) printf("success\n");
2291 changelist_remove(clp, name);
2294 (void) changelist_postfix(clp);
2295 changelist_free(clp);
2298 * Deferred destroy might destroy the snapshot or only mark it to be
2299 * destroyed later, and it returns success in either case.
2301 if (err != 0 || (defer && zfs_dataset_exists(hdl, name,
2302 ZFS_TYPE_SNAPSHOT))) {
2303 err = recv_rename(hdl, name, NULL, baselen, newname, flags);
2309 typedef struct guid_to_name_data {
2311 boolean_t bookmark_ok;
2314 } guid_to_name_data_t;
2317 guid_to_name_cb(zfs_handle_t *zhp, void *arg)
2319 guid_to_name_data_t *gtnd = arg;
2323 if (gtnd->skip != NULL &&
2324 (slash = strrchr(zhp->zfs_name, '/')) != NULL &&
2325 strcmp(slash + 1, gtnd->skip) == 0) {
2330 if (zfs_prop_get_int(zhp, ZFS_PROP_GUID) == gtnd->guid) {
2331 (void) strcpy(gtnd->name, zhp->zfs_name);
2336 err = zfs_iter_children(zhp, guid_to_name_cb, gtnd);
2337 if (err != EEXIST && gtnd->bookmark_ok)
2338 err = zfs_iter_bookmarks(zhp, guid_to_name_cb, gtnd);
2344 * Attempt to find the local dataset associated with this guid. In the case of
2345 * multiple matches, we attempt to find the "best" match by searching
2346 * progressively larger portions of the hierarchy. This allows one to send a
2347 * tree of datasets individually and guarantee that we will find the source
2348 * guid within that hierarchy, even if there are multiple matches elsewhere.
2351 guid_to_name(libzfs_handle_t *hdl, const char *parent, uint64_t guid,
2352 boolean_t bookmark_ok, char *name)
2354 char pname[ZFS_MAX_DATASET_NAME_LEN];
2355 guid_to_name_data_t gtnd;
2358 gtnd.bookmark_ok = bookmark_ok;
2363 * Search progressively larger portions of the hierarchy, starting
2364 * with the filesystem specified by 'parent'. This will
2365 * select the "most local" version of the origin snapshot in the case
2366 * that there are multiple matching snapshots in the system.
2368 (void) strlcpy(pname, parent, sizeof (pname));
2369 char *cp = strrchr(pname, '@');
2371 cp = strchr(pname, '\0');
2372 for (; cp != NULL; cp = strrchr(pname, '/')) {
2373 /* Chop off the last component and open the parent */
2375 zfs_handle_t *zhp = make_dataset_handle(hdl, pname);
2379 int err = guid_to_name_cb(zfs_handle_dup(zhp), >nd);
2381 err = zfs_iter_children(zhp, guid_to_name_cb, >nd);
2382 if (err != EEXIST && bookmark_ok)
2383 err = zfs_iter_bookmarks(zhp, guid_to_name_cb, >nd);
2389 * Remember the last portion of the dataset so we skip it next
2390 * time through (as we've already searched that portion of the
2393 gtnd.skip = strrchr(pname, '/') + 1;
2400 * Return +1 if guid1 is before guid2, 0 if they are the same, and -1 if
2401 * guid1 is after guid2.
2404 created_before(libzfs_handle_t *hdl, avl_tree_t *avl,
2405 uint64_t guid1, uint64_t guid2)
2408 char *fsname, *snapname;
2409 char buf[ZFS_MAX_DATASET_NAME_LEN];
2411 zfs_handle_t *guid1hdl, *guid2hdl;
2412 uint64_t create1, create2;
2419 nvfs = fsavl_find(avl, guid1, &snapname);
2420 VERIFY(0 == nvlist_lookup_string(nvfs, "name", &fsname));
2421 (void) snprintf(buf, sizeof (buf), "%s@%s", fsname, snapname);
2422 guid1hdl = zfs_open(hdl, buf, ZFS_TYPE_SNAPSHOT);
2423 if (guid1hdl == NULL)
2426 nvfs = fsavl_find(avl, guid2, &snapname);
2427 VERIFY(0 == nvlist_lookup_string(nvfs, "name", &fsname));
2428 (void) snprintf(buf, sizeof (buf), "%s@%s", fsname, snapname);
2429 guid2hdl = zfs_open(hdl, buf, ZFS_TYPE_SNAPSHOT);
2430 if (guid2hdl == NULL) {
2431 zfs_close(guid1hdl);
2435 create1 = zfs_prop_get_int(guid1hdl, ZFS_PROP_CREATETXG);
2436 create2 = zfs_prop_get_int(guid2hdl, ZFS_PROP_CREATETXG);
2438 if (create1 < create2)
2440 else if (create1 > create2)
2445 zfs_close(guid1hdl);
2446 zfs_close(guid2hdl);
2452 recv_incremental_replication(libzfs_handle_t *hdl, const char *tofs,
2453 recvflags_t *flags, nvlist_t *stream_nv, avl_tree_t *stream_avl,
2456 nvlist_t *local_nv, *deleted = NULL;
2457 avl_tree_t *local_avl;
2458 nvpair_t *fselem, *nextfselem;
2460 char newname[ZFS_MAX_DATASET_NAME_LEN];
2463 boolean_t needagain, progress, recursive;
2466 VERIFY(0 == nvlist_lookup_string(stream_nv, "fromsnap", &fromsnap));
2468 recursive = (nvlist_lookup_boolean(stream_nv, "not_recursive") ==
2475 needagain = progress = B_FALSE;
2477 VERIFY(0 == nvlist_alloc(&deleted, NV_UNIQUE_NAME, 0));
2479 if ((error = gather_nvlist(hdl, tofs, fromsnap, NULL,
2480 recursive, B_FALSE, &local_nv, &local_avl)) != 0)
2484 * Process deletes and renames
2486 for (fselem = nvlist_next_nvpair(local_nv, NULL);
2487 fselem; fselem = nextfselem) {
2488 nvlist_t *nvfs, *snaps;
2489 nvlist_t *stream_nvfs = NULL;
2490 nvpair_t *snapelem, *nextsnapelem;
2491 uint64_t fromguid = 0;
2492 uint64_t originguid = 0;
2493 uint64_t stream_originguid = 0;
2494 uint64_t parent_fromsnap_guid, stream_parent_fromsnap_guid;
2495 char *fsname, *stream_fsname;
2497 nextfselem = nvlist_next_nvpair(local_nv, fselem);
2499 VERIFY(0 == nvpair_value_nvlist(fselem, &nvfs));
2500 VERIFY(0 == nvlist_lookup_nvlist(nvfs, "snaps", &snaps));
2501 VERIFY(0 == nvlist_lookup_string(nvfs, "name", &fsname));
2502 VERIFY(0 == nvlist_lookup_uint64(nvfs, "parentfromsnap",
2503 &parent_fromsnap_guid));
2504 (void) nvlist_lookup_uint64(nvfs, "origin", &originguid);
2507 * First find the stream's fs, so we can check for
2508 * a different origin (due to "zfs promote")
2510 for (snapelem = nvlist_next_nvpair(snaps, NULL);
2511 snapelem; snapelem = nvlist_next_nvpair(snaps, snapelem)) {
2514 VERIFY(0 == nvpair_value_uint64(snapelem, &thisguid));
2515 stream_nvfs = fsavl_find(stream_avl, thisguid, NULL);
2517 if (stream_nvfs != NULL)
2521 /* check for promote */
2522 (void) nvlist_lookup_uint64(stream_nvfs, "origin",
2523 &stream_originguid);
2524 if (stream_nvfs && originguid != stream_originguid) {
2525 switch (created_before(hdl, local_avl,
2526 stream_originguid, originguid)) {
2529 zfs_cmd_t zc = { 0 };
2530 nvlist_t *origin_nvfs;
2531 char *origin_fsname;
2534 (void) printf("promoting %s\n", fsname);
2536 origin_nvfs = fsavl_find(local_avl, originguid,
2538 VERIFY(0 == nvlist_lookup_string(origin_nvfs,
2539 "name", &origin_fsname));
2540 (void) strlcpy(zc.zc_value, origin_fsname,
2541 sizeof (zc.zc_value));
2542 (void) strlcpy(zc.zc_name, fsname,
2543 sizeof (zc.zc_name));
2544 error = zfs_ioctl(hdl, ZFS_IOC_PROMOTE, &zc);
2552 fsavl_destroy(local_avl);
2553 nvlist_free(local_nv);
2557 * We had/have the wrong origin, therefore our
2558 * list of snapshots is wrong. Need to handle
2559 * them on the next pass.
2565 for (snapelem = nvlist_next_nvpair(snaps, NULL);
2566 snapelem; snapelem = nextsnapelem) {
2568 char *stream_snapname;
2569 nvlist_t *found, *props;
2571 nextsnapelem = nvlist_next_nvpair(snaps, snapelem);
2573 VERIFY(0 == nvpair_value_uint64(snapelem, &thisguid));
2574 found = fsavl_find(stream_avl, thisguid,
2577 /* check for delete */
2578 if (found == NULL) {
2579 char name[ZFS_MAX_DATASET_NAME_LEN];
2584 (void) snprintf(name, sizeof (name), "%s@%s",
2585 fsname, nvpair_name(snapelem));
2587 error = recv_destroy(hdl, name,
2588 strlen(fsname)+1, newname, flags);
2593 sprintf(guidname, "%" PRIu64, thisguid);
2594 nvlist_add_boolean(deleted, guidname);
2598 stream_nvfs = found;
2600 if (0 == nvlist_lookup_nvlist(stream_nvfs, "snapprops",
2601 &props) && 0 == nvlist_lookup_nvlist(props,
2602 stream_snapname, &props)) {
2603 zfs_cmd_t zc = { 0 };
2605 zc.zc_cookie = B_TRUE; /* received */
2606 (void) snprintf(zc.zc_name, sizeof (zc.zc_name),
2607 "%s@%s", fsname, nvpair_name(snapelem));
2608 if (zcmd_write_src_nvlist(hdl, &zc,
2610 (void) zfs_ioctl(hdl,
2611 ZFS_IOC_SET_PROP, &zc);
2612 zcmd_free_nvlists(&zc);
2616 /* check for different snapname */
2617 if (strcmp(nvpair_name(snapelem),
2618 stream_snapname) != 0) {
2619 char name[ZFS_MAX_DATASET_NAME_LEN];
2620 char tryname[ZFS_MAX_DATASET_NAME_LEN];
2622 (void) snprintf(name, sizeof (name), "%s@%s",
2623 fsname, nvpair_name(snapelem));
2624 (void) snprintf(tryname, sizeof (name), "%s@%s",
2625 fsname, stream_snapname);
2627 error = recv_rename(hdl, name, tryname,
2628 strlen(fsname)+1, newname, flags);
2635 if (strcmp(stream_snapname, fromsnap) == 0)
2636 fromguid = thisguid;
2639 /* check for delete */
2640 if (stream_nvfs == NULL) {
2644 error = recv_destroy(hdl, fsname, strlen(tofs)+1,
2650 sprintf(guidname, "%" PRIu64, parent_fromsnap_guid);
2651 nvlist_add_boolean(deleted, guidname);
2655 if (fromguid == 0) {
2656 if (flags->verbose) {
2657 (void) printf("local fs %s does not have "
2658 "fromsnap (%s in stream); must have "
2659 "been deleted locally; ignoring\n",
2665 VERIFY(0 == nvlist_lookup_string(stream_nvfs,
2666 "name", &stream_fsname));
2667 VERIFY(0 == nvlist_lookup_uint64(stream_nvfs,
2668 "parentfromsnap", &stream_parent_fromsnap_guid));
2670 s1 = strrchr(fsname, '/');
2671 s2 = strrchr(stream_fsname, '/');
2674 * Check if we're going to rename based on parent guid change
2675 * and the current parent guid was also deleted. If it was then
2676 * rename will fail and is likely unneeded, so avoid this and
2677 * force an early retry to determine the new
2678 * parent_fromsnap_guid.
2680 if (stream_parent_fromsnap_guid != 0 &&
2681 parent_fromsnap_guid != 0 &&
2682 stream_parent_fromsnap_guid != parent_fromsnap_guid) {
2683 sprintf(guidname, "%" PRIu64, parent_fromsnap_guid);
2684 if (nvlist_exists(deleted, guidname)) {
2692 * Check for rename. If the exact receive path is specified, it
2693 * does not count as a rename, but we still need to check the
2694 * datasets beneath it.
2696 if ((stream_parent_fromsnap_guid != 0 &&
2697 parent_fromsnap_guid != 0 &&
2698 stream_parent_fromsnap_guid != parent_fromsnap_guid) ||
2699 ((flags->isprefix || strcmp(tofs, fsname) != 0) &&
2700 (s1 != NULL) && (s2 != NULL) && strcmp(s1, s2) != 0)) {
2702 char tryname[ZFS_MAX_DATASET_NAME_LEN];
2704 parent = fsavl_find(local_avl,
2705 stream_parent_fromsnap_guid, NULL);
2707 * NB: parent might not be found if we used the
2708 * tosnap for stream_parent_fromsnap_guid,
2709 * because the parent is a newly-created fs;
2710 * we'll be able to rename it after we recv the
2713 if (parent != NULL) {
2716 VERIFY(0 == nvlist_lookup_string(parent, "name",
2718 (void) snprintf(tryname, sizeof (tryname),
2719 "%s%s", pname, strrchr(stream_fsname, '/'));
2722 if (flags->verbose) {
2723 (void) printf("local fs %s new parent "
2724 "not found\n", fsname);
2730 error = recv_rename(hdl, fsname, tryname,
2731 strlen(tofs)+1, newname, flags);
2733 if (renamed != NULL && newname[0] != '\0') {
2734 VERIFY(0 == nvlist_add_boolean(renamed,
2746 fsavl_destroy(local_avl);
2747 nvlist_free(local_nv);
2748 nvlist_free(deleted);
2750 if (needagain && progress) {
2751 /* do another pass to fix up temporary names */
2753 (void) printf("another pass:\n");
2761 zfs_receive_package(libzfs_handle_t *hdl, int fd, const char *destname,
2762 recvflags_t *flags, dmu_replay_record_t *drr, zio_cksum_t *zc,
2763 char **top_zfs, int cleanup_fd, uint64_t *action_handlep)
2765 nvlist_t *stream_nv = NULL;
2766 avl_tree_t *stream_avl = NULL;
2767 char *fromsnap = NULL;
2768 char *sendsnap = NULL;
2770 char tofs[ZFS_MAX_DATASET_NAME_LEN];
2771 char sendfs[ZFS_MAX_DATASET_NAME_LEN];
2773 dmu_replay_record_t drre;
2775 boolean_t anyerr = B_FALSE;
2776 boolean_t softerr = B_FALSE;
2777 boolean_t recursive;
2779 (void) snprintf(errbuf, sizeof (errbuf), dgettext(TEXT_DOMAIN,
2782 assert(drr->drr_type == DRR_BEGIN);
2783 assert(drr->drr_u.drr_begin.drr_magic == DMU_BACKUP_MAGIC);
2784 assert(DMU_GET_STREAM_HDRTYPE(drr->drr_u.drr_begin.drr_versioninfo) ==
2785 DMU_COMPOUNDSTREAM);
2788 * Read in the nvlist from the stream.
2790 if (drr->drr_payloadlen != 0) {
2791 error = recv_read_nvlist(hdl, fd, drr->drr_payloadlen,
2792 &stream_nv, flags->byteswap, zc);
2794 error = zfs_error(hdl, EZFS_BADSTREAM, errbuf);
2799 recursive = (nvlist_lookup_boolean(stream_nv, "not_recursive") ==
2802 if (recursive && strchr(destname, '@')) {
2803 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
2804 "cannot specify snapshot name for multi-snapshot stream"));
2805 error = zfs_error(hdl, EZFS_BADSTREAM, errbuf);
2810 * Read in the end record and verify checksum.
2812 if (0 != (error = recv_read(hdl, fd, &drre, sizeof (drre),
2813 flags->byteswap, NULL)))
2815 if (flags->byteswap) {
2816 drre.drr_type = BSWAP_32(drre.drr_type);
2817 drre.drr_u.drr_end.drr_checksum.zc_word[0] =
2818 BSWAP_64(drre.drr_u.drr_end.drr_checksum.zc_word[0]);
2819 drre.drr_u.drr_end.drr_checksum.zc_word[1] =
2820 BSWAP_64(drre.drr_u.drr_end.drr_checksum.zc_word[1]);
2821 drre.drr_u.drr_end.drr_checksum.zc_word[2] =
2822 BSWAP_64(drre.drr_u.drr_end.drr_checksum.zc_word[2]);
2823 drre.drr_u.drr_end.drr_checksum.zc_word[3] =
2824 BSWAP_64(drre.drr_u.drr_end.drr_checksum.zc_word[3]);
2826 if (drre.drr_type != DRR_END) {
2827 error = zfs_error(hdl, EZFS_BADSTREAM, errbuf);
2830 if (!ZIO_CHECKSUM_EQUAL(drre.drr_u.drr_end.drr_checksum, *zc)) {
2831 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
2832 "incorrect header checksum"));
2833 error = zfs_error(hdl, EZFS_BADSTREAM, errbuf);
2837 (void) nvlist_lookup_string(stream_nv, "fromsnap", &fromsnap);
2839 if (drr->drr_payloadlen != 0) {
2840 nvlist_t *stream_fss;
2842 VERIFY(0 == nvlist_lookup_nvlist(stream_nv, "fss",
2844 if ((stream_avl = fsavl_create(stream_fss)) == NULL) {
2845 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
2846 "couldn't allocate avl tree"));
2847 error = zfs_error(hdl, EZFS_NOMEM, errbuf);
2851 if (fromsnap != NULL && recursive) {
2852 nvlist_t *renamed = NULL;
2853 nvpair_t *pair = NULL;
2855 (void) strlcpy(tofs, destname, sizeof (tofs));
2856 if (flags->isprefix) {
2857 struct drr_begin *drrb = &drr->drr_u.drr_begin;
2860 if (flags->istail) {
2861 cp = strrchr(drrb->drr_toname, '/');
2863 (void) strlcat(tofs, "/",
2867 i = (cp - drrb->drr_toname);
2870 i = strcspn(drrb->drr_toname, "/@");
2872 /* zfs_receive_one() will create_parents() */
2873 (void) strlcat(tofs, &drrb->drr_toname[i],
2875 *strchr(tofs, '@') = '\0';
2878 if (!flags->dryrun && !flags->nomount) {
2879 VERIFY(0 == nvlist_alloc(&renamed,
2880 NV_UNIQUE_NAME, 0));
2883 softerr = recv_incremental_replication(hdl, tofs, flags,
2884 stream_nv, stream_avl, renamed);
2886 /* Unmount renamed filesystems before receiving. */
2887 while ((pair = nvlist_next_nvpair(renamed,
2890 prop_changelist_t *clp = NULL;
2892 zhp = zfs_open(hdl, nvpair_name(pair),
2893 ZFS_TYPE_FILESYSTEM);
2895 clp = changelist_gather(zhp,
2896 ZFS_PROP_MOUNTPOINT, 0, 0);
2900 changelist_prefix(clp);
2901 changelist_free(clp);
2906 nvlist_free(renamed);
2911 * Get the fs specified by the first path in the stream (the top level
2912 * specified by 'zfs send') and pass it to each invocation of
2913 * zfs_receive_one().
2915 (void) strlcpy(sendfs, drr->drr_u.drr_begin.drr_toname,
2917 if ((cp = strchr(sendfs, '@')) != NULL) {
2920 * Find the "sendsnap", the final snapshot in a replication
2921 * stream. zfs_receive_one() handles certain errors
2922 * differently, depending on if the contained stream is the
2925 sendsnap = (cp + 1);
2928 /* Finally, receive each contained stream */
2931 * we should figure out if it has a recoverable
2932 * error, in which case do a recv_skip() and drive on.
2933 * Note, if we fail due to already having this guid,
2934 * zfs_receive_one() will take care of it (ie,
2935 * recv_skip() and return 0).
2937 error = zfs_receive_impl(hdl, destname, NULL, flags, fd,
2938 sendfs, stream_nv, stream_avl, top_zfs, cleanup_fd,
2939 action_handlep, sendsnap);
2940 if (error == ENODATA) {
2945 } while (error == 0);
2947 if (drr->drr_payloadlen != 0 && recursive && fromsnap != NULL) {
2949 * Now that we have the fs's they sent us, try the
2952 softerr = recv_incremental_replication(hdl, tofs, flags,
2953 stream_nv, stream_avl, NULL);
2957 fsavl_destroy(stream_avl);
2958 nvlist_free(stream_nv);
2967 trunc_prop_errs(int truncated)
2969 ASSERT(truncated != 0);
2972 (void) fprintf(stderr, dgettext(TEXT_DOMAIN,
2973 "1 more property could not be set\n"));
2975 (void) fprintf(stderr, dgettext(TEXT_DOMAIN,
2976 "%d more properties could not be set\n"), truncated);
2980 recv_skip(libzfs_handle_t *hdl, int fd, boolean_t byteswap)
2982 dmu_replay_record_t *drr;
2983 void *buf = zfs_alloc(hdl, SPA_MAXBLOCKSIZE);
2986 (void) snprintf(errbuf, sizeof (errbuf), dgettext(TEXT_DOMAIN,
2987 "cannot receive:"));
2989 /* XXX would be great to use lseek if possible... */
2992 while (recv_read(hdl, fd, drr, sizeof (dmu_replay_record_t),
2993 byteswap, NULL) == 0) {
2995 drr->drr_type = BSWAP_32(drr->drr_type);
2997 switch (drr->drr_type) {
2999 if (drr->drr_payloadlen != 0) {
3000 (void) recv_read(hdl, fd, buf,
3001 drr->drr_payloadlen, B_FALSE, NULL);
3011 drr->drr_u.drr_object.drr_bonuslen =
3012 BSWAP_32(drr->drr_u.drr_object.
3015 (void) recv_read(hdl, fd, buf,
3016 P2ROUNDUP(drr->drr_u.drr_object.drr_bonuslen, 8),
3022 drr->drr_u.drr_write.drr_logical_size =
3024 drr->drr_u.drr_write.drr_logical_size);
3025 drr->drr_u.drr_write.drr_compressed_size =
3027 drr->drr_u.drr_write.drr_compressed_size);
3029 uint64_t payload_size =
3030 DRR_WRITE_PAYLOAD_SIZE(&drr->drr_u.drr_write);
3031 (void) recv_read(hdl, fd, buf,
3032 payload_size, B_FALSE, NULL);
3036 drr->drr_u.drr_spill.drr_length =
3037 BSWAP_64(drr->drr_u.drr_spill.drr_length);
3039 (void) recv_read(hdl, fd, buf,
3040 drr->drr_u.drr_spill.drr_length, B_FALSE, NULL);
3042 case DRR_WRITE_EMBEDDED:
3044 drr->drr_u.drr_write_embedded.drr_psize =
3045 BSWAP_32(drr->drr_u.drr_write_embedded.
3048 (void) recv_read(hdl, fd, buf,
3049 P2ROUNDUP(drr->drr_u.drr_write_embedded.drr_psize,
3052 case DRR_WRITE_BYREF:
3053 case DRR_FREEOBJECTS:
3058 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
3059 "invalid record type"));
3060 return (zfs_error(hdl, EZFS_BADSTREAM, errbuf));
3069 recv_ecksum_set_aux(libzfs_handle_t *hdl, const char *target_snap,
3070 boolean_t resumable)
3072 char target_fs[ZFS_MAX_DATASET_NAME_LEN];
3074 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
3075 "checksum mismatch or incomplete stream"));
3079 (void) strlcpy(target_fs, target_snap, sizeof (target_fs));
3080 *strchr(target_fs, '@') = '\0';
3081 zfs_handle_t *zhp = zfs_open(hdl, target_fs,
3082 ZFS_TYPE_FILESYSTEM | ZFS_TYPE_VOLUME);
3086 char token_buf[ZFS_MAXPROPLEN];
3087 int error = zfs_prop_get(zhp, ZFS_PROP_RECEIVE_RESUME_TOKEN,
3088 token_buf, sizeof (token_buf),
3089 NULL, NULL, 0, B_TRUE);
3091 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
3092 "checksum mismatch or incomplete stream.\n"
3093 "Partially received snapshot is saved.\n"
3094 "A resuming stream can be generated on the sending "
3095 "system by running:\n"
3103 * Restores a backup of tosnap from the file descriptor specified by infd.
3106 zfs_receive_one(libzfs_handle_t *hdl, int infd, const char *tosnap,
3107 const char *originsnap, recvflags_t *flags, dmu_replay_record_t *drr,
3108 dmu_replay_record_t *drr_noswap, const char *sendfs, nvlist_t *stream_nv,
3109 avl_tree_t *stream_avl, char **top_zfs, int cleanup_fd,
3110 uint64_t *action_handlep, const char *finalsnap)
3112 zfs_cmd_t zc = { 0 };
3114 int ioctl_err, ioctl_errno, err;
3116 struct drr_begin *drrb = &drr->drr_u.drr_begin;
3118 char prop_errbuf[1024];
3119 const char *chopprefix;
3120 boolean_t newfs = B_FALSE;
3121 boolean_t stream_wantsnewfs;
3122 uint64_t parent_snapguid = 0;
3123 prop_changelist_t *clp = NULL;
3124 nvlist_t *snapprops_nvlist = NULL;
3125 zprop_errflags_t prop_errflags;
3126 boolean_t recursive;
3127 char *snapname = NULL;
3129 begin_time = time(NULL);
3131 (void) snprintf(errbuf, sizeof (errbuf), dgettext(TEXT_DOMAIN,
3134 recursive = (nvlist_lookup_boolean(stream_nv, "not_recursive") ==
3137 if (stream_avl != NULL) {
3138 nvlist_t *fs = fsavl_find(stream_avl, drrb->drr_toguid,
3143 (void) nvlist_lookup_uint64(fs, "parentfromsnap",
3145 err = nvlist_lookup_nvlist(fs, "props", &props);
3147 VERIFY(0 == nvlist_alloc(&props, NV_UNIQUE_NAME, 0));
3149 if (flags->canmountoff) {
3150 VERIFY(0 == nvlist_add_uint64(props,
3151 zfs_prop_to_name(ZFS_PROP_CANMOUNT), 0));
3153 ret = zcmd_write_src_nvlist(hdl, &zc, props);
3157 if (0 == nvlist_lookup_nvlist(fs, "snapprops", &props)) {
3158 VERIFY(0 == nvlist_lookup_nvlist(props,
3159 snapname, &snapprops_nvlist));
3169 * Determine how much of the snapshot name stored in the stream
3170 * we are going to tack on to the name they specified on the
3171 * command line, and how much we are going to chop off.
3173 * If they specified a snapshot, chop the entire name stored in
3176 if (flags->istail) {
3178 * A filesystem was specified with -e. We want to tack on only
3179 * the tail of the sent snapshot path.
3181 if (strchr(tosnap, '@')) {
3182 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, "invalid "
3183 "argument - snapshot not allowed with -e"));
3184 return (zfs_error(hdl, EZFS_INVALIDNAME, errbuf));
3187 chopprefix = strrchr(sendfs, '/');
3189 if (chopprefix == NULL) {
3191 * The tail is the poolname, so we need to
3192 * prepend a path separator.
3194 int len = strlen(drrb->drr_toname);
3195 cp = malloc(len + 2);
3197 (void) strcpy(&cp[1], drrb->drr_toname);
3200 chopprefix = drrb->drr_toname + (chopprefix - sendfs);
3202 } else if (flags->isprefix) {
3204 * A filesystem was specified with -d. We want to tack on
3205 * everything but the first element of the sent snapshot path
3206 * (all but the pool name).
3208 if (strchr(tosnap, '@')) {
3209 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, "invalid "
3210 "argument - snapshot not allowed with -d"));
3211 return (zfs_error(hdl, EZFS_INVALIDNAME, errbuf));
3214 chopprefix = strchr(drrb->drr_toname, '/');
3215 if (chopprefix == NULL)
3216 chopprefix = strchr(drrb->drr_toname, '@');
3217 } else if (strchr(tosnap, '@') == NULL) {
3219 * If a filesystem was specified without -d or -e, we want to
3220 * tack on everything after the fs specified by 'zfs send'.
3222 chopprefix = drrb->drr_toname + strlen(sendfs);
3224 /* A snapshot was specified as an exact path (no -d or -e). */
3226 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
3227 "cannot specify snapshot name for multi-snapshot "
3229 return (zfs_error(hdl, EZFS_BADSTREAM, errbuf));
3231 chopprefix = drrb->drr_toname + strlen(drrb->drr_toname);
3234 ASSERT(strstr(drrb->drr_toname, sendfs) == drrb->drr_toname);
3235 ASSERT(chopprefix > drrb->drr_toname);
3236 ASSERT(chopprefix <= drrb->drr_toname + strlen(drrb->drr_toname));
3237 ASSERT(chopprefix[0] == '/' || chopprefix[0] == '@' ||
3238 chopprefix[0] == '\0');
3241 * Determine name of destination snapshot, store in zc_value.
3243 (void) strcpy(zc.zc_value, tosnap);
3244 (void) strncat(zc.zc_value, chopprefix, sizeof (zc.zc_value));
3246 if (zfs_ioctl_version == ZFS_IOCVER_UNDEF)
3247 zfs_ioctl_version = get_zfs_ioctl_version();
3249 * For forward compatibility hide tosnap in zc_value
3251 if (zfs_ioctl_version < ZFS_IOCVER_LZC)
3252 (void) strcpy(zc.zc_value + strlen(zc.zc_value) + 1, tosnap);
3255 if (!zfs_name_valid(zc.zc_value, ZFS_TYPE_SNAPSHOT)) {
3256 zcmd_free_nvlists(&zc);
3257 return (zfs_error(hdl, EZFS_INVALIDNAME, errbuf));
3261 * Determine the name of the origin snapshot, store in zc_string.
3264 (void) strncpy(zc.zc_string, originsnap, sizeof (zc.zc_string));
3266 (void) printf("using provided clone origin %s\n",
3268 } else if (drrb->drr_flags & DRR_FLAG_CLONE) {
3269 if (guid_to_name(hdl, zc.zc_value,
3270 drrb->drr_fromguid, B_FALSE, zc.zc_string) != 0) {
3271 zcmd_free_nvlists(&zc);
3272 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
3273 "local origin for clone %s does not exist"),
3275 return (zfs_error(hdl, EZFS_NOENT, errbuf));
3278 (void) printf("found clone origin %s\n", zc.zc_string);
3281 boolean_t resuming = DMU_GET_FEATUREFLAGS(drrb->drr_versioninfo) &
3282 DMU_BACKUP_FEATURE_RESUMING;
3283 stream_wantsnewfs = (drrb->drr_fromguid == 0 ||
3284 (drrb->drr_flags & DRR_FLAG_CLONE) || originsnap) && !resuming;
3286 if (stream_wantsnewfs) {
3288 * if the parent fs does not exist, look for it based on
3289 * the parent snap GUID
3291 (void) snprintf(errbuf, sizeof (errbuf), dgettext(TEXT_DOMAIN,
3292 "cannot receive new filesystem stream"));
3294 (void) strcpy(zc.zc_name, zc.zc_value);
3295 cp = strrchr(zc.zc_name, '/');
3299 !zfs_dataset_exists(hdl, zc.zc_name, ZFS_TYPE_DATASET)) {
3300 char suffix[ZFS_MAX_DATASET_NAME_LEN];
3301 (void) strcpy(suffix, strrchr(zc.zc_value, '/'));
3302 if (guid_to_name(hdl, zc.zc_name, parent_snapguid,
3303 B_FALSE, zc.zc_value) == 0) {
3304 *strchr(zc.zc_value, '@') = '\0';
3305 (void) strcat(zc.zc_value, suffix);
3310 * If the fs does not exist, look for it based on the
3314 (void) snprintf(errbuf, sizeof (errbuf),
3315 dgettext(TEXT_DOMAIN,
3316 "cannot receive resume stream"));
3318 (void) snprintf(errbuf, sizeof (errbuf),
3319 dgettext(TEXT_DOMAIN,
3320 "cannot receive incremental stream"));
3323 (void) strcpy(zc.zc_name, zc.zc_value);
3324 *strchr(zc.zc_name, '@') = '\0';
3327 * If the exact receive path was specified and this is the
3328 * topmost path in the stream, then if the fs does not exist we
3329 * should look no further.
3331 if ((flags->isprefix || (*(chopprefix = drrb->drr_toname +
3332 strlen(sendfs)) != '\0' && *chopprefix != '@')) &&
3333 !zfs_dataset_exists(hdl, zc.zc_name, ZFS_TYPE_DATASET)) {
3334 char snap[ZFS_MAX_DATASET_NAME_LEN];
3335 (void) strcpy(snap, strchr(zc.zc_value, '@'));
3336 if (guid_to_name(hdl, zc.zc_name, drrb->drr_fromguid,
3337 B_FALSE, zc.zc_value) == 0) {
3338 *strchr(zc.zc_value, '@') = '\0';
3339 (void) strcat(zc.zc_value, snap);
3344 (void) strcpy(zc.zc_name, zc.zc_value);
3345 *strchr(zc.zc_name, '@') = '\0';
3347 if (zfs_dataset_exists(hdl, zc.zc_name, ZFS_TYPE_DATASET)) {
3351 * Destination fs exists. It must be one of these cases:
3352 * - an incremental send stream
3353 * - the stream specifies a new fs (full stream or clone)
3354 * and they want us to blow away the existing fs (and
3355 * have therefore specified -F and removed any snapshots)
3356 * - we are resuming a failed receive.
3358 if (stream_wantsnewfs) {
3359 if (!flags->force) {
3360 zcmd_free_nvlists(&zc);
3361 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
3362 "destination '%s' exists\n"
3363 "must specify -F to overwrite it"),
3365 return (zfs_error(hdl, EZFS_EXISTS, errbuf));
3367 if (ioctl(hdl->libzfs_fd, ZFS_IOC_SNAPSHOT_LIST_NEXT,
3369 zcmd_free_nvlists(&zc);
3370 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
3371 "destination has snapshots (eg. %s)\n"
3372 "must destroy them to overwrite it"),
3374 return (zfs_error(hdl, EZFS_EXISTS, errbuf));
3378 if ((zhp = zfs_open(hdl, zc.zc_name,
3379 ZFS_TYPE_FILESYSTEM | ZFS_TYPE_VOLUME)) == NULL) {
3380 zcmd_free_nvlists(&zc);
3384 if (stream_wantsnewfs &&
3385 zhp->zfs_dmustats.dds_origin[0]) {
3386 zcmd_free_nvlists(&zc);
3388 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
3389 "destination '%s' is a clone\n"
3390 "must destroy it to overwrite it"),
3392 return (zfs_error(hdl, EZFS_EXISTS, errbuf));
3395 if (!flags->dryrun && zhp->zfs_type == ZFS_TYPE_FILESYSTEM &&
3396 stream_wantsnewfs) {
3397 /* We can't do online recv in this case */
3398 clp = changelist_gather(zhp, ZFS_PROP_NAME, 0, 0);
3401 zcmd_free_nvlists(&zc);
3404 if (changelist_prefix(clp) != 0) {
3405 changelist_free(clp);
3407 zcmd_free_nvlists(&zc);
3413 * If we are resuming a newfs, set newfs here so that we will
3414 * mount it if the recv succeeds this time. We can tell
3415 * that it was a newfs on the first recv because the fs
3416 * itself will be inconsistent (if the fs existed when we
3417 * did the first recv, we would have received it into
3420 if (resuming && zfs_prop_get_int(zhp, ZFS_PROP_INCONSISTENT))
3426 * Destination filesystem does not exist. Therefore we better
3427 * be creating a new filesystem (either from a full backup, or
3428 * a clone). It would therefore be invalid if the user
3429 * specified only the pool name (i.e. if the destination name
3430 * contained no slash character).
3432 if (!stream_wantsnewfs ||
3433 (cp = strrchr(zc.zc_name, '/')) == NULL) {
3434 zcmd_free_nvlists(&zc);
3435 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
3436 "destination '%s' does not exist"), zc.zc_name);
3437 return (zfs_error(hdl, EZFS_NOENT, errbuf));
3441 * Trim off the final dataset component so we perform the
3442 * recvbackup ioctl to the filesystems's parent.
3446 if (flags->isprefix && !flags->istail && !flags->dryrun &&
3447 create_parents(hdl, zc.zc_value, strlen(tosnap)) != 0) {
3448 zcmd_free_nvlists(&zc);
3449 return (zfs_error(hdl, EZFS_BADRESTORE, errbuf));
3455 zc.zc_begin_record = *drr_noswap;
3456 zc.zc_cookie = infd;
3457 zc.zc_guid = flags->force;
3458 zc.zc_resumable = flags->resumable;
3459 if (flags->verbose) {
3460 (void) printf("%s %s stream of %s into %s\n",
3461 flags->dryrun ? "would receive" : "receiving",
3462 drrb->drr_fromguid ? "incremental" : "full",
3463 drrb->drr_toname, zc.zc_value);
3464 (void) fflush(stdout);
3467 if (flags->dryrun) {
3468 zcmd_free_nvlists(&zc);
3469 return (recv_skip(hdl, infd, flags->byteswap));
3472 zc.zc_nvlist_dst = (uint64_t)(uintptr_t)prop_errbuf;
3473 zc.zc_nvlist_dst_size = sizeof (prop_errbuf);
3474 zc.zc_cleanup_fd = cleanup_fd;
3475 zc.zc_action_handle = *action_handlep;
3477 err = ioctl_err = zfs_ioctl(hdl, ZFS_IOC_RECV, &zc);
3478 ioctl_errno = errno;
3479 prop_errflags = (zprop_errflags_t)zc.zc_obj;
3482 nvlist_t *prop_errors;
3483 VERIFY(0 == nvlist_unpack((void *)(uintptr_t)zc.zc_nvlist_dst,
3484 zc.zc_nvlist_dst_size, &prop_errors, 0));
3486 nvpair_t *prop_err = NULL;
3488 while ((prop_err = nvlist_next_nvpair(prop_errors,
3489 prop_err)) != NULL) {
3494 prop = zfs_name_to_prop(nvpair_name(prop_err));
3495 (void) nvpair_value_int32(prop_err, &intval);
3496 if (strcmp(nvpair_name(prop_err),
3497 ZPROP_N_MORE_ERRORS) == 0) {
3498 trunc_prop_errs(intval);
3500 } else if (snapname == NULL || finalsnap == NULL ||
3501 strcmp(finalsnap, snapname) == 0 ||
3502 strcmp(nvpair_name(prop_err),
3503 zfs_prop_to_name(ZFS_PROP_REFQUOTA)) != 0) {
3505 * Skip the special case of, for example,
3506 * "refquota", errors on intermediate
3507 * snapshots leading up to a final one.
3508 * That's why we have all of the checks above.
3510 * See zfs_ioctl.c's extract_delay_props() for
3511 * a list of props which can fail on
3512 * intermediate snapshots, but shouldn't
3513 * affect the overall receive.
3515 (void) snprintf(tbuf, sizeof (tbuf),
3516 dgettext(TEXT_DOMAIN,
3517 "cannot receive %s property on %s"),
3518 nvpair_name(prop_err), zc.zc_name);
3519 zfs_setprop_error(hdl, prop, intval, tbuf);
3522 nvlist_free(prop_errors);
3525 zc.zc_nvlist_dst = 0;
3526 zc.zc_nvlist_dst_size = 0;
3527 zcmd_free_nvlists(&zc);
3529 if (err == 0 && snapprops_nvlist) {
3530 zfs_cmd_t zc2 = { 0 };
3532 (void) strcpy(zc2.zc_name, zc.zc_value);
3533 zc2.zc_cookie = B_TRUE; /* received */
3534 if (zcmd_write_src_nvlist(hdl, &zc2, snapprops_nvlist) == 0) {
3535 (void) zfs_ioctl(hdl, ZFS_IOC_SET_PROP, &zc2);
3536 zcmd_free_nvlists(&zc2);
3540 if (err && (ioctl_errno == ENOENT || ioctl_errno == EEXIST)) {
3542 * It may be that this snapshot already exists,
3543 * in which case we want to consume & ignore it
3544 * rather than failing.
3546 avl_tree_t *local_avl;
3547 nvlist_t *local_nv, *fs;
3548 cp = strchr(zc.zc_value, '@');
3551 * XXX Do this faster by just iterating over snaps in
3552 * this fs. Also if zc_value does not exist, we will
3553 * get a strange "does not exist" error message.
3556 if (gather_nvlist(hdl, zc.zc_value, NULL, NULL, B_FALSE,
3557 B_FALSE, &local_nv, &local_avl) == 0) {
3559 fs = fsavl_find(local_avl, drrb->drr_toguid, NULL);
3560 fsavl_destroy(local_avl);
3561 nvlist_free(local_nv);
3564 if (flags->verbose) {
3565 (void) printf("snap %s already exists; "
3566 "ignoring\n", zc.zc_value);
3568 err = ioctl_err = recv_skip(hdl, infd,
3575 if (ioctl_err != 0) {
3576 switch (ioctl_errno) {
3578 cp = strchr(zc.zc_value, '@');
3580 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
3581 "most recent snapshot of %s does not\n"
3582 "match incremental source"), zc.zc_value);
3583 (void) zfs_error(hdl, EZFS_BADRESTORE, errbuf);
3587 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
3588 "destination %s has been modified\n"
3589 "since most recent snapshot"), zc.zc_name);
3590 (void) zfs_error(hdl, EZFS_BADRESTORE, errbuf);
3593 cp = strchr(zc.zc_value, '@');
3595 /* it's the containing fs that exists */
3598 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
3599 "destination already exists"));
3600 (void) zfs_error_fmt(hdl, EZFS_EXISTS,
3601 dgettext(TEXT_DOMAIN, "cannot restore to %s"),
3606 (void) zfs_error(hdl, EZFS_BADSTREAM, errbuf);
3609 recv_ecksum_set_aux(hdl, zc.zc_value, flags->resumable);
3610 (void) zfs_error(hdl, EZFS_BADSTREAM, errbuf);
3613 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
3614 "pool must be upgraded to receive this stream."));
3615 (void) zfs_error(hdl, EZFS_BADVERSION, errbuf);
3618 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
3619 "destination %s space quota exceeded"), zc.zc_name);
3620 (void) zfs_error(hdl, EZFS_NOSPC, errbuf);
3623 (void) zfs_standard_error(hdl, ioctl_errno, errbuf);
3628 * Mount the target filesystem (if created). Also mount any
3629 * children of the target filesystem if we did a replication
3630 * receive (indicated by stream_avl being non-NULL).
3632 cp = strchr(zc.zc_value, '@');
3633 if (cp && (ioctl_err == 0 || !newfs)) {
3637 h = zfs_open(hdl, zc.zc_value,
3638 ZFS_TYPE_FILESYSTEM | ZFS_TYPE_VOLUME);
3640 if (h->zfs_type == ZFS_TYPE_VOLUME) {
3642 } else if (newfs || stream_avl) {
3644 * Track the first/top of hierarchy fs,
3645 * for mounting and sharing later.
3647 if (top_zfs && *top_zfs == NULL)
3648 *top_zfs = zfs_strdup(hdl, zc.zc_value);
3656 if (!flags->nomount)
3657 err |= changelist_postfix(clp);
3658 changelist_free(clp);
3661 if (prop_errflags & ZPROP_ERR_NOCLEAR) {
3662 (void) fprintf(stderr, dgettext(TEXT_DOMAIN, "Warning: "
3663 "failed to clear unreceived properties on %s"),
3665 (void) fprintf(stderr, "\n");
3667 if (prop_errflags & ZPROP_ERR_NORESTORE) {
3668 (void) fprintf(stderr, dgettext(TEXT_DOMAIN, "Warning: "
3669 "failed to restore original properties on %s"),
3671 (void) fprintf(stderr, "\n");
3674 if (err || ioctl_err)
3677 *action_handlep = zc.zc_action_handle;
3679 if (flags->verbose) {
3682 uint64_t bytes = zc.zc_cookie;
3683 time_t delta = time(NULL) - begin_time;
3686 zfs_nicenum(bytes, buf1, sizeof (buf1));
3687 zfs_nicenum(bytes/delta, buf2, sizeof (buf1));
3689 (void) printf("received %sB stream in %lu seconds (%sB/sec)\n",
3697 zfs_receive_impl(libzfs_handle_t *hdl, const char *tosnap,
3698 const char *originsnap, recvflags_t *flags, int infd, const char *sendfs,
3699 nvlist_t *stream_nv, avl_tree_t *stream_avl, char **top_zfs, int cleanup_fd,
3700 uint64_t *action_handlep, const char *finalsnap)
3703 dmu_replay_record_t drr, drr_noswap;
3704 struct drr_begin *drrb = &drr.drr_u.drr_begin;
3706 zio_cksum_t zcksum = { 0 };
3707 uint64_t featureflags;
3710 (void) snprintf(errbuf, sizeof (errbuf), dgettext(TEXT_DOMAIN,
3713 if (flags->isprefix &&
3714 !zfs_dataset_exists(hdl, tosnap, ZFS_TYPE_DATASET)) {
3715 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, "specified fs "
3716 "(%s) does not exist"), tosnap);
3717 return (zfs_error(hdl, EZFS_NOENT, errbuf));
3720 !zfs_dataset_exists(hdl, originsnap, ZFS_TYPE_DATASET)) {
3721 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, "specified origin fs "
3722 "(%s) does not exist"), originsnap);
3723 return (zfs_error(hdl, EZFS_NOENT, errbuf));
3726 /* read in the BEGIN record */
3727 if (0 != (err = recv_read(hdl, infd, &drr, sizeof (drr), B_FALSE,
3731 if (drr.drr_type == DRR_END || drr.drr_type == BSWAP_32(DRR_END)) {
3732 /* It's the double end record at the end of a package */
3736 /* the kernel needs the non-byteswapped begin record */
3739 flags->byteswap = B_FALSE;
3740 if (drrb->drr_magic == BSWAP_64(DMU_BACKUP_MAGIC)) {
3742 * We computed the checksum in the wrong byteorder in
3743 * recv_read() above; do it again correctly.
3745 bzero(&zcksum, sizeof (zio_cksum_t));
3746 (void) fletcher_4_incremental_byteswap(&drr,
3747 sizeof (drr), &zcksum);
3748 flags->byteswap = B_TRUE;
3750 drr.drr_type = BSWAP_32(drr.drr_type);
3751 drr.drr_payloadlen = BSWAP_32(drr.drr_payloadlen);
3752 drrb->drr_magic = BSWAP_64(drrb->drr_magic);
3753 drrb->drr_versioninfo = BSWAP_64(drrb->drr_versioninfo);
3754 drrb->drr_creation_time = BSWAP_64(drrb->drr_creation_time);
3755 drrb->drr_type = BSWAP_32(drrb->drr_type);
3756 drrb->drr_flags = BSWAP_32(drrb->drr_flags);
3757 drrb->drr_toguid = BSWAP_64(drrb->drr_toguid);
3758 drrb->drr_fromguid = BSWAP_64(drrb->drr_fromguid);
3761 if (drrb->drr_magic != DMU_BACKUP_MAGIC || drr.drr_type != DRR_BEGIN) {
3762 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, "invalid "
3763 "stream (bad magic number)"));
3764 return (zfs_error(hdl, EZFS_BADSTREAM, errbuf));
3767 featureflags = DMU_GET_FEATUREFLAGS(drrb->drr_versioninfo);
3768 hdrtype = DMU_GET_STREAM_HDRTYPE(drrb->drr_versioninfo);
3770 if (!DMU_STREAM_SUPPORTED(featureflags) ||
3771 (hdrtype != DMU_SUBSTREAM && hdrtype != DMU_COMPOUNDSTREAM)) {
3772 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
3773 "stream has unsupported feature, feature flags = %lx"),
3775 return (zfs_error(hdl, EZFS_BADSTREAM, errbuf));
3778 if (strchr(drrb->drr_toname, '@') == NULL) {
3779 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, "invalid "
3780 "stream (bad snapshot name)"));
3781 return (zfs_error(hdl, EZFS_BADSTREAM, errbuf));
3784 if (DMU_GET_STREAM_HDRTYPE(drrb->drr_versioninfo) == DMU_SUBSTREAM) {
3785 char nonpackage_sendfs[ZFS_MAX_DATASET_NAME_LEN];
3786 if (sendfs == NULL) {
3788 * We were not called from zfs_receive_package(). Get
3789 * the fs specified by 'zfs send'.
3792 (void) strlcpy(nonpackage_sendfs,
3793 drr.drr_u.drr_begin.drr_toname,
3794 sizeof (nonpackage_sendfs));
3795 if ((cp = strchr(nonpackage_sendfs, '@')) != NULL)
3797 sendfs = nonpackage_sendfs;
3798 VERIFY(finalsnap == NULL);
3800 return (zfs_receive_one(hdl, infd, tosnap, originsnap, flags,
3801 &drr, &drr_noswap, sendfs, stream_nv, stream_avl, top_zfs,
3802 cleanup_fd, action_handlep, finalsnap));
3804 assert(DMU_GET_STREAM_HDRTYPE(drrb->drr_versioninfo) ==
3805 DMU_COMPOUNDSTREAM);
3806 return (zfs_receive_package(hdl, infd, tosnap, flags, &drr,
3807 &zcksum, top_zfs, cleanup_fd, action_handlep));
3812 * Restores a backup of tosnap from the file descriptor specified by infd.
3813 * Return 0 on total success, -2 if some things couldn't be
3814 * destroyed/renamed/promoted, -1 if some things couldn't be received.
3815 * (-1 will override -2, if -1 and the resumable flag was specified the
3816 * transfer can be resumed if the sending side supports it).
3819 zfs_receive(libzfs_handle_t *hdl, const char *tosnap, nvlist_t *props,
3820 recvflags_t *flags, int infd, avl_tree_t *stream_avl)
3822 char *top_zfs = NULL;
3825 uint64_t action_handle = 0;
3826 char *originsnap = NULL;
3828 err = nvlist_lookup_string(props, "origin", &originsnap);
3829 if (err && err != ENOENT)
3833 cleanup_fd = open(ZFS_DEV, O_RDWR|O_EXCL);
3834 VERIFY(cleanup_fd >= 0);
3836 err = zfs_receive_impl(hdl, tosnap, originsnap, flags, infd, NULL, NULL,
3837 stream_avl, &top_zfs, cleanup_fd, &action_handle, NULL);
3839 VERIFY(0 == close(cleanup_fd));
3841 if (err == 0 && !flags->nomount && top_zfs) {
3843 prop_changelist_t *clp;
3845 zhp = zfs_open(hdl, top_zfs, ZFS_TYPE_FILESYSTEM);
3847 clp = changelist_gather(zhp, ZFS_PROP_MOUNTPOINT,
3848 CL_GATHER_MOUNT_ALWAYS, 0);
3851 /* mount and share received datasets */
3852 err = changelist_postfix(clp);
3853 changelist_free(clp);
3856 if (zhp == NULL || clp == NULL || err)