1185029Spjd/* 2185029Spjd * CDDL HEADER START 3185029Spjd * 4185029Spjd * The contents of this file are subject to the terms of the 5185029Spjd * Common Development and Distribution License (the "License"). 6185029Spjd * You may not use this file except in compliance with the License. 7185029Spjd * 8185029Spjd * You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE 9185029Spjd * or http://www.opensolaris.org/os/licensing. 10185029Spjd * See the License for the specific language governing permissions 11185029Spjd * and limitations under the License. 12185029Spjd * 13185029Spjd * When distributing Covered Code, include this CDDL HEADER in each 14185029Spjd * file and include the License file at usr/src/OPENSOLARIS.LICENSE. 15185029Spjd * If applicable, add the following below this CDDL HEADER, with the 16185029Spjd * fields enclosed by brackets "[]" replaced with your own identifying 17185029Spjd * information: Portions Copyright [yyyy] [name of copyright owner] 18185029Spjd * 19185029Spjd * CDDL HEADER END 20185029Spjd */ 21185029Spjd 22185029Spjd/* 23219089Spjd * Copyright (c) 2005, 2010, Oracle and/or its affiliates. All rights reserved. 24238422Smm * Copyright (c) 2012 by Delphix. All rights reserved. 25235222Smm * Copyright (c) 2012, Joyent, Inc. All rights reserved. 26230438Spjd * Copyright (c) 2012 Pawel Jakub Dawidek <pawel@dawidek.net>. 27230438Spjd * All rights reserved. 28251646Sdelphij * Copyright (c) 2013 Steven Hartland. All rights reserved. 29185029Spjd */ 30185029Spjd 31185029Spjd#include <assert.h> 32185029Spjd#include <ctype.h> 33185029Spjd#include <errno.h> 34185029Spjd#include <libintl.h> 35185029Spjd#include <stdio.h> 36185029Spjd#include <stdlib.h> 37185029Spjd#include <strings.h> 38185029Spjd#include <unistd.h> 39185029Spjd#include <stddef.h> 40185029Spjd#include <fcntl.h> 41186515Srwatson#include <sys/param.h> 42185029Spjd#include <sys/mount.h> 43219089Spjd#include <pthread.h> 44219089Spjd#include <umem.h> 45235222Smm#include <time.h> 46185029Spjd 47185029Spjd#include <libzfs.h> 48185029Spjd 49185029Spjd#include "zfs_namecheck.h" 50185029Spjd#include "zfs_prop.h" 51219089Spjd#include "zfs_fletcher.h" 52185029Spjd#include "libzfs_impl.h" 53219089Spjd#include <sha2.h> 54219089Spjd#include <sys/zio_checksum.h> 55219089Spjd#include <sys/ddt.h> 56185029Spjd 57248571Smm#ifdef __FreeBSD__ 58248571Smmextern int zfs_ioctl_version; 59248571Smm#endif 60248571Smm 61219089Spjd/* in libzfs_dataset.c */ 62219089Spjdextern void zfs_setprop_error(libzfs_handle_t *, zfs_prop_t, int, char *); 63185029Spjd/* We need to use something for ENODATA. */ 64185029Spjd#define ENODATA EIDRM 65185029Spjd 66228103Smmstatic int zfs_receive_impl(libzfs_handle_t *, const char *, recvflags_t *, 67219089Spjd int, const char *, nvlist_t *, avl_tree_t *, char **, int, uint64_t *); 68185029Spjd 69219089Spjdstatic const zio_cksum_t zero_cksum = { 0 }; 70219089Spjd 71219089Spjdtypedef struct dedup_arg { 72219089Spjd int inputfd; 73219089Spjd int outputfd; 74219089Spjd libzfs_handle_t *dedup_hdl; 75219089Spjd} dedup_arg_t; 76219089Spjd 77235222Smmtypedef struct progress_arg { 78235222Smm zfs_handle_t *pa_zhp; 79235222Smm int pa_fd; 80235222Smm boolean_t pa_parsable; 81235222Smm} progress_arg_t; 82235222Smm 83219089Spjdtypedef struct dataref { 84219089Spjd uint64_t ref_guid; 85219089Spjd uint64_t ref_object; 86219089Spjd uint64_t ref_offset; 87219089Spjd} dataref_t; 88219089Spjd 89219089Spjdtypedef struct dedup_entry { 90219089Spjd struct dedup_entry *dde_next; 91219089Spjd zio_cksum_t dde_chksum; 92219089Spjd uint64_t dde_prop; 93219089Spjd dataref_t dde_ref; 94219089Spjd} dedup_entry_t; 95219089Spjd 96219089Spjd#define MAX_DDT_PHYSMEM_PERCENT 20 97219089Spjd#define SMALLEST_POSSIBLE_MAX_DDT_MB 128 98219089Spjd 99219089Spjdtypedef struct dedup_table { 100219089Spjd dedup_entry_t **dedup_hash_array; 101219089Spjd umem_cache_t *ddecache; 102219089Spjd uint64_t max_ddt_size; /* max dedup table size in bytes */ 103219089Spjd uint64_t cur_ddt_size; /* current dedup table size in bytes */ 104219089Spjd uint64_t ddt_count; 105219089Spjd int numhashbits; 106219089Spjd boolean_t ddt_full; 107219089Spjd} dedup_table_t; 108219089Spjd 109219089Spjdstatic int 110219089Spjdhigh_order_bit(uint64_t n) 111219089Spjd{ 112219089Spjd int count; 113219089Spjd 114219089Spjd for (count = 0; n != 0; count++) 115219089Spjd n >>= 1; 116219089Spjd return (count); 117219089Spjd} 118219089Spjd 119219089Spjdstatic size_t 120219089Spjdssread(void *buf, size_t len, FILE *stream) 121219089Spjd{ 122219089Spjd size_t outlen; 123219089Spjd 124219089Spjd if ((outlen = fread(buf, len, 1, stream)) == 0) 125219089Spjd return (0); 126219089Spjd 127219089Spjd return (outlen); 128219089Spjd} 129219089Spjd 130219089Spjdstatic void 131219089Spjdddt_hash_append(libzfs_handle_t *hdl, dedup_table_t *ddt, dedup_entry_t **ddepp, 132219089Spjd zio_cksum_t *cs, uint64_t prop, dataref_t *dr) 133219089Spjd{ 134219089Spjd dedup_entry_t *dde; 135219089Spjd 136219089Spjd if (ddt->cur_ddt_size >= ddt->max_ddt_size) { 137219089Spjd if (ddt->ddt_full == B_FALSE) { 138219089Spjd zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, 139219089Spjd "Dedup table full. Deduplication will continue " 140219089Spjd "with existing table entries")); 141219089Spjd ddt->ddt_full = B_TRUE; 142219089Spjd } 143219089Spjd return; 144219089Spjd } 145219089Spjd 146219089Spjd if ((dde = umem_cache_alloc(ddt->ddecache, UMEM_DEFAULT)) 147219089Spjd != NULL) { 148219089Spjd assert(*ddepp == NULL); 149219089Spjd dde->dde_next = NULL; 150219089Spjd dde->dde_chksum = *cs; 151219089Spjd dde->dde_prop = prop; 152219089Spjd dde->dde_ref = *dr; 153219089Spjd *ddepp = dde; 154219089Spjd ddt->cur_ddt_size += sizeof (dedup_entry_t); 155219089Spjd ddt->ddt_count++; 156219089Spjd } 157219089Spjd} 158219089Spjd 159185029Spjd/* 160219089Spjd * Using the specified dedup table, do a lookup for an entry with 161219089Spjd * the checksum cs. If found, return the block's reference info 162219089Spjd * in *dr. Otherwise, insert a new entry in the dedup table, using 163219089Spjd * the reference information specified by *dr. 164219089Spjd * 165219089Spjd * return value: true - entry was found 166219089Spjd * false - entry was not found 167219089Spjd */ 168219089Spjdstatic boolean_t 169219089Spjdddt_update(libzfs_handle_t *hdl, dedup_table_t *ddt, zio_cksum_t *cs, 170219089Spjd uint64_t prop, dataref_t *dr) 171219089Spjd{ 172219089Spjd uint32_t hashcode; 173219089Spjd dedup_entry_t **ddepp; 174219089Spjd 175219089Spjd hashcode = BF64_GET(cs->zc_word[0], 0, ddt->numhashbits); 176219089Spjd 177219089Spjd for (ddepp = &(ddt->dedup_hash_array[hashcode]); *ddepp != NULL; 178219089Spjd ddepp = &((*ddepp)->dde_next)) { 179219089Spjd if (ZIO_CHECKSUM_EQUAL(((*ddepp)->dde_chksum), *cs) && 180219089Spjd (*ddepp)->dde_prop == prop) { 181219089Spjd *dr = (*ddepp)->dde_ref; 182219089Spjd return (B_TRUE); 183219089Spjd } 184219089Spjd } 185219089Spjd ddt_hash_append(hdl, ddt, ddepp, cs, prop, dr); 186219089Spjd return (B_FALSE); 187219089Spjd} 188219089Spjd 189219089Spjdstatic int 190219089Spjdcksum_and_write(const void *buf, uint64_t len, zio_cksum_t *zc, int outfd) 191219089Spjd{ 192219089Spjd fletcher_4_incremental_native(buf, len, zc); 193219089Spjd return (write(outfd, buf, len)); 194219089Spjd} 195219089Spjd 196219089Spjd/* 197219089Spjd * This function is started in a separate thread when the dedup option 198219089Spjd * has been requested. The main send thread determines the list of 199219089Spjd * snapshots to be included in the send stream and makes the ioctl calls 200219089Spjd * for each one. But instead of having the ioctl send the output to the 201219089Spjd * the output fd specified by the caller of zfs_send()), the 202219089Spjd * ioctl is told to direct the output to a pipe, which is read by the 203219089Spjd * alternate thread running THIS function. This function does the 204219089Spjd * dedup'ing by: 205219089Spjd * 1. building a dedup table (the DDT) 206219089Spjd * 2. doing checksums on each data block and inserting a record in the DDT 207219089Spjd * 3. looking for matching checksums, and 208219089Spjd * 4. sending a DRR_WRITE_BYREF record instead of a write record whenever 209219089Spjd * a duplicate block is found. 210219089Spjd * The output of this function then goes to the output fd requested 211219089Spjd * by the caller of zfs_send(). 212219089Spjd */ 213219089Spjdstatic void * 214219089Spjdcksummer(void *arg) 215219089Spjd{ 216219089Spjd dedup_arg_t *dda = arg; 217219089Spjd char *buf = malloc(1<<20); 218219089Spjd dmu_replay_record_t thedrr; 219219089Spjd dmu_replay_record_t *drr = &thedrr; 220219089Spjd struct drr_begin *drrb = &thedrr.drr_u.drr_begin; 221219089Spjd struct drr_end *drre = &thedrr.drr_u.drr_end; 222219089Spjd struct drr_object *drro = &thedrr.drr_u.drr_object; 223219089Spjd struct drr_write *drrw = &thedrr.drr_u.drr_write; 224219089Spjd struct drr_spill *drrs = &thedrr.drr_u.drr_spill; 225219089Spjd FILE *ofp; 226219089Spjd int outfd; 227219089Spjd dmu_replay_record_t wbr_drr = {0}; 228219089Spjd struct drr_write_byref *wbr_drrr = &wbr_drr.drr_u.drr_write_byref; 229219089Spjd dedup_table_t ddt; 230219089Spjd zio_cksum_t stream_cksum; 231219089Spjd uint64_t physmem = sysconf(_SC_PHYS_PAGES) * sysconf(_SC_PAGESIZE); 232219089Spjd uint64_t numbuckets; 233219089Spjd 234219089Spjd ddt.max_ddt_size = 235219089Spjd MAX((physmem * MAX_DDT_PHYSMEM_PERCENT)/100, 236219089Spjd SMALLEST_POSSIBLE_MAX_DDT_MB<<20); 237219089Spjd 238219089Spjd numbuckets = ddt.max_ddt_size/(sizeof (dedup_entry_t)); 239219089Spjd 240219089Spjd /* 241219089Spjd * numbuckets must be a power of 2. Increase number to 242219089Spjd * a power of 2 if necessary. 243219089Spjd */ 244219089Spjd if (!ISP2(numbuckets)) 245219089Spjd numbuckets = 1 << high_order_bit(numbuckets); 246219089Spjd 247219089Spjd ddt.dedup_hash_array = calloc(numbuckets, sizeof (dedup_entry_t *)); 248219089Spjd ddt.ddecache = umem_cache_create("dde", sizeof (dedup_entry_t), 0, 249219089Spjd NULL, NULL, NULL, NULL, NULL, 0); 250219089Spjd ddt.cur_ddt_size = numbuckets * sizeof (dedup_entry_t *); 251219089Spjd ddt.numhashbits = high_order_bit(numbuckets) - 1; 252219089Spjd ddt.ddt_full = B_FALSE; 253219089Spjd 254219089Spjd /* Initialize the write-by-reference block. */ 255219089Spjd wbr_drr.drr_type = DRR_WRITE_BYREF; 256219089Spjd wbr_drr.drr_payloadlen = 0; 257219089Spjd 258219089Spjd outfd = dda->outputfd; 259219089Spjd ofp = fdopen(dda->inputfd, "r"); 260219089Spjd while (ssread(drr, sizeof (dmu_replay_record_t), ofp) != 0) { 261219089Spjd 262219089Spjd switch (drr->drr_type) { 263219089Spjd case DRR_BEGIN: 264219089Spjd { 265219089Spjd int fflags; 266219089Spjd ZIO_SET_CHECKSUM(&stream_cksum, 0, 0, 0, 0); 267219089Spjd 268219089Spjd /* set the DEDUP feature flag for this stream */ 269219089Spjd fflags = DMU_GET_FEATUREFLAGS(drrb->drr_versioninfo); 270219089Spjd fflags |= (DMU_BACKUP_FEATURE_DEDUP | 271219089Spjd DMU_BACKUP_FEATURE_DEDUPPROPS); 272219089Spjd DMU_SET_FEATUREFLAGS(drrb->drr_versioninfo, fflags); 273219089Spjd 274219089Spjd if (cksum_and_write(drr, sizeof (dmu_replay_record_t), 275219089Spjd &stream_cksum, outfd) == -1) 276219089Spjd goto out; 277219089Spjd if (DMU_GET_STREAM_HDRTYPE(drrb->drr_versioninfo) == 278219089Spjd DMU_COMPOUNDSTREAM && drr->drr_payloadlen != 0) { 279219089Spjd int sz = drr->drr_payloadlen; 280219089Spjd 281219089Spjd if (sz > 1<<20) { 282219089Spjd free(buf); 283219089Spjd buf = malloc(sz); 284219089Spjd } 285219089Spjd (void) ssread(buf, sz, ofp); 286219089Spjd if (ferror(stdin)) 287219089Spjd perror("fread"); 288219089Spjd if (cksum_and_write(buf, sz, &stream_cksum, 289219089Spjd outfd) == -1) 290219089Spjd goto out; 291219089Spjd } 292219089Spjd break; 293219089Spjd } 294219089Spjd 295219089Spjd case DRR_END: 296219089Spjd { 297219089Spjd /* use the recalculated checksum */ 298219089Spjd ZIO_SET_CHECKSUM(&drre->drr_checksum, 299219089Spjd stream_cksum.zc_word[0], stream_cksum.zc_word[1], 300219089Spjd stream_cksum.zc_word[2], stream_cksum.zc_word[3]); 301219089Spjd if ((write(outfd, drr, 302219089Spjd sizeof (dmu_replay_record_t))) == -1) 303219089Spjd goto out; 304219089Spjd break; 305219089Spjd } 306219089Spjd 307219089Spjd case DRR_OBJECT: 308219089Spjd { 309219089Spjd if (cksum_and_write(drr, sizeof (dmu_replay_record_t), 310219089Spjd &stream_cksum, outfd) == -1) 311219089Spjd goto out; 312219089Spjd if (drro->drr_bonuslen > 0) { 313219089Spjd (void) ssread(buf, 314219089Spjd P2ROUNDUP((uint64_t)drro->drr_bonuslen, 8), 315219089Spjd ofp); 316219089Spjd if (cksum_and_write(buf, 317219089Spjd P2ROUNDUP((uint64_t)drro->drr_bonuslen, 8), 318219089Spjd &stream_cksum, outfd) == -1) 319219089Spjd goto out; 320219089Spjd } 321219089Spjd break; 322219089Spjd } 323219089Spjd 324219089Spjd case DRR_SPILL: 325219089Spjd { 326219089Spjd if (cksum_and_write(drr, sizeof (dmu_replay_record_t), 327219089Spjd &stream_cksum, outfd) == -1) 328219089Spjd goto out; 329219089Spjd (void) ssread(buf, drrs->drr_length, ofp); 330219089Spjd if (cksum_and_write(buf, drrs->drr_length, 331219089Spjd &stream_cksum, outfd) == -1) 332219089Spjd goto out; 333219089Spjd break; 334219089Spjd } 335219089Spjd 336219089Spjd case DRR_FREEOBJECTS: 337219089Spjd { 338219089Spjd if (cksum_and_write(drr, sizeof (dmu_replay_record_t), 339219089Spjd &stream_cksum, outfd) == -1) 340219089Spjd goto out; 341219089Spjd break; 342219089Spjd } 343219089Spjd 344219089Spjd case DRR_WRITE: 345219089Spjd { 346219089Spjd dataref_t dataref; 347219089Spjd 348219089Spjd (void) ssread(buf, drrw->drr_length, ofp); 349219089Spjd 350219089Spjd /* 351219089Spjd * Use the existing checksum if it's dedup-capable, 352219089Spjd * else calculate a SHA256 checksum for it. 353219089Spjd */ 354219089Spjd 355219089Spjd if (ZIO_CHECKSUM_EQUAL(drrw->drr_key.ddk_cksum, 356219089Spjd zero_cksum) || 357219089Spjd !DRR_IS_DEDUP_CAPABLE(drrw->drr_checksumflags)) { 358219089Spjd SHA256_CTX ctx; 359219089Spjd zio_cksum_t tmpsha256; 360219089Spjd 361219089Spjd SHA256Init(&ctx); 362219089Spjd SHA256Update(&ctx, buf, drrw->drr_length); 363219089Spjd SHA256Final(&tmpsha256, &ctx); 364219089Spjd drrw->drr_key.ddk_cksum.zc_word[0] = 365219089Spjd BE_64(tmpsha256.zc_word[0]); 366219089Spjd drrw->drr_key.ddk_cksum.zc_word[1] = 367219089Spjd BE_64(tmpsha256.zc_word[1]); 368219089Spjd drrw->drr_key.ddk_cksum.zc_word[2] = 369219089Spjd BE_64(tmpsha256.zc_word[2]); 370219089Spjd drrw->drr_key.ddk_cksum.zc_word[3] = 371219089Spjd BE_64(tmpsha256.zc_word[3]); 372219089Spjd drrw->drr_checksumtype = ZIO_CHECKSUM_SHA256; 373219089Spjd drrw->drr_checksumflags = DRR_CHECKSUM_DEDUP; 374219089Spjd } 375219089Spjd 376219089Spjd dataref.ref_guid = drrw->drr_toguid; 377219089Spjd dataref.ref_object = drrw->drr_object; 378219089Spjd dataref.ref_offset = drrw->drr_offset; 379219089Spjd 380219089Spjd if (ddt_update(dda->dedup_hdl, &ddt, 381219089Spjd &drrw->drr_key.ddk_cksum, drrw->drr_key.ddk_prop, 382219089Spjd &dataref)) { 383219089Spjd /* block already present in stream */ 384219089Spjd wbr_drrr->drr_object = drrw->drr_object; 385219089Spjd wbr_drrr->drr_offset = drrw->drr_offset; 386219089Spjd wbr_drrr->drr_length = drrw->drr_length; 387219089Spjd wbr_drrr->drr_toguid = drrw->drr_toguid; 388219089Spjd wbr_drrr->drr_refguid = dataref.ref_guid; 389219089Spjd wbr_drrr->drr_refobject = 390219089Spjd dataref.ref_object; 391219089Spjd wbr_drrr->drr_refoffset = 392219089Spjd dataref.ref_offset; 393219089Spjd 394219089Spjd wbr_drrr->drr_checksumtype = 395219089Spjd drrw->drr_checksumtype; 396219089Spjd wbr_drrr->drr_checksumflags = 397219089Spjd drrw->drr_checksumtype; 398219089Spjd wbr_drrr->drr_key.ddk_cksum = 399219089Spjd drrw->drr_key.ddk_cksum; 400219089Spjd wbr_drrr->drr_key.ddk_prop = 401219089Spjd drrw->drr_key.ddk_prop; 402219089Spjd 403219089Spjd if (cksum_and_write(&wbr_drr, 404219089Spjd sizeof (dmu_replay_record_t), &stream_cksum, 405219089Spjd outfd) == -1) 406219089Spjd goto out; 407219089Spjd } else { 408219089Spjd /* block not previously seen */ 409219089Spjd if (cksum_and_write(drr, 410219089Spjd sizeof (dmu_replay_record_t), &stream_cksum, 411219089Spjd outfd) == -1) 412219089Spjd goto out; 413219089Spjd if (cksum_and_write(buf, 414219089Spjd drrw->drr_length, 415219089Spjd &stream_cksum, outfd) == -1) 416219089Spjd goto out; 417219089Spjd } 418219089Spjd break; 419219089Spjd } 420219089Spjd 421219089Spjd case DRR_FREE: 422219089Spjd { 423219089Spjd if (cksum_and_write(drr, sizeof (dmu_replay_record_t), 424219089Spjd &stream_cksum, outfd) == -1) 425219089Spjd goto out; 426219089Spjd break; 427219089Spjd } 428219089Spjd 429219089Spjd default: 430219089Spjd (void) printf("INVALID record type 0x%x\n", 431219089Spjd drr->drr_type); 432219089Spjd /* should never happen, so assert */ 433219089Spjd assert(B_FALSE); 434219089Spjd } 435219089Spjd } 436219089Spjdout: 437219089Spjd umem_cache_destroy(ddt.ddecache); 438219089Spjd free(ddt.dedup_hash_array); 439219089Spjd free(buf); 440219089Spjd (void) fclose(ofp); 441219089Spjd 442219089Spjd return (NULL); 443219089Spjd} 444219089Spjd 445219089Spjd/* 446185029Spjd * Routines for dealing with the AVL tree of fs-nvlists 447185029Spjd */ 448185029Spjdtypedef struct fsavl_node { 449185029Spjd avl_node_t fn_node; 450185029Spjd nvlist_t *fn_nvfs; 451185029Spjd char *fn_snapname; 452185029Spjd uint64_t fn_guid; 453185029Spjd} fsavl_node_t; 454185029Spjd 455185029Spjdstatic int 456185029Spjdfsavl_compare(const void *arg1, const void *arg2) 457185029Spjd{ 458185029Spjd const fsavl_node_t *fn1 = arg1; 459185029Spjd const fsavl_node_t *fn2 = arg2; 460185029Spjd 461185029Spjd if (fn1->fn_guid > fn2->fn_guid) 462185029Spjd return (+1); 463185029Spjd else if (fn1->fn_guid < fn2->fn_guid) 464185029Spjd return (-1); 465185029Spjd else 466185029Spjd return (0); 467185029Spjd} 468185029Spjd 469185029Spjd/* 470185029Spjd * Given the GUID of a snapshot, find its containing filesystem and 471185029Spjd * (optionally) name. 472185029Spjd */ 473185029Spjdstatic nvlist_t * 474185029Spjdfsavl_find(avl_tree_t *avl, uint64_t snapguid, char **snapname) 475185029Spjd{ 476185029Spjd fsavl_node_t fn_find; 477185029Spjd fsavl_node_t *fn; 478185029Spjd 479185029Spjd fn_find.fn_guid = snapguid; 480185029Spjd 481185029Spjd fn = avl_find(avl, &fn_find, NULL); 482185029Spjd if (fn) { 483185029Spjd if (snapname) 484185029Spjd *snapname = fn->fn_snapname; 485185029Spjd return (fn->fn_nvfs); 486185029Spjd } 487185029Spjd return (NULL); 488185029Spjd} 489185029Spjd 490185029Spjdstatic void 491185029Spjdfsavl_destroy(avl_tree_t *avl) 492185029Spjd{ 493185029Spjd fsavl_node_t *fn; 494185029Spjd void *cookie; 495185029Spjd 496185029Spjd if (avl == NULL) 497185029Spjd return; 498185029Spjd 499185029Spjd cookie = NULL; 500185029Spjd while ((fn = avl_destroy_nodes(avl, &cookie)) != NULL) 501185029Spjd free(fn); 502185029Spjd avl_destroy(avl); 503185029Spjd free(avl); 504185029Spjd} 505185029Spjd 506219089Spjd/* 507219089Spjd * Given an nvlist, produce an avl tree of snapshots, ordered by guid 508219089Spjd */ 509185029Spjdstatic avl_tree_t * 510185029Spjdfsavl_create(nvlist_t *fss) 511185029Spjd{ 512185029Spjd avl_tree_t *fsavl; 513185029Spjd nvpair_t *fselem = NULL; 514185029Spjd 515185029Spjd if ((fsavl = malloc(sizeof (avl_tree_t))) == NULL) 516185029Spjd return (NULL); 517185029Spjd 518185029Spjd avl_create(fsavl, fsavl_compare, sizeof (fsavl_node_t), 519185029Spjd offsetof(fsavl_node_t, fn_node)); 520185029Spjd 521185029Spjd while ((fselem = nvlist_next_nvpair(fss, fselem)) != NULL) { 522185029Spjd nvlist_t *nvfs, *snaps; 523185029Spjd nvpair_t *snapelem = NULL; 524185029Spjd 525185029Spjd VERIFY(0 == nvpair_value_nvlist(fselem, &nvfs)); 526185029Spjd VERIFY(0 == nvlist_lookup_nvlist(nvfs, "snaps", &snaps)); 527185029Spjd 528185029Spjd while ((snapelem = 529185029Spjd nvlist_next_nvpair(snaps, snapelem)) != NULL) { 530185029Spjd fsavl_node_t *fn; 531185029Spjd uint64_t guid; 532185029Spjd 533185029Spjd VERIFY(0 == nvpair_value_uint64(snapelem, &guid)); 534185029Spjd if ((fn = malloc(sizeof (fsavl_node_t))) == NULL) { 535185029Spjd fsavl_destroy(fsavl); 536185029Spjd return (NULL); 537185029Spjd } 538185029Spjd fn->fn_nvfs = nvfs; 539185029Spjd fn->fn_snapname = nvpair_name(snapelem); 540185029Spjd fn->fn_guid = guid; 541185029Spjd 542185029Spjd /* 543185029Spjd * Note: if there are multiple snaps with the 544185029Spjd * same GUID, we ignore all but one. 545185029Spjd */ 546185029Spjd if (avl_find(fsavl, fn, NULL) == NULL) 547185029Spjd avl_add(fsavl, fn); 548185029Spjd else 549185029Spjd free(fn); 550185029Spjd } 551185029Spjd } 552185029Spjd 553185029Spjd return (fsavl); 554185029Spjd} 555185029Spjd 556185029Spjd/* 557185029Spjd * Routines for dealing with the giant nvlist of fs-nvlists, etc. 558185029Spjd */ 559185029Spjdtypedef struct send_data { 560185029Spjd uint64_t parent_fromsnap_guid; 561185029Spjd nvlist_t *parent_snaps; 562185029Spjd nvlist_t *fss; 563185029Spjd nvlist_t *snapprops; 564185029Spjd const char *fromsnap; 565185029Spjd const char *tosnap; 566219089Spjd boolean_t recursive; 567185029Spjd 568185029Spjd /* 569185029Spjd * The header nvlist is of the following format: 570185029Spjd * { 571185029Spjd * "tosnap" -> string 572185029Spjd * "fromsnap" -> string (if incremental) 573185029Spjd * "fss" -> { 574185029Spjd * id -> { 575185029Spjd * 576185029Spjd * "name" -> string (full name; for debugging) 577185029Spjd * "parentfromsnap" -> number (guid of fromsnap in parent) 578185029Spjd * 579185029Spjd * "props" -> { name -> value (only if set here) } 580185029Spjd * "snaps" -> { name (lastname) -> number (guid) } 581185029Spjd * "snapprops" -> { name (lastname) -> { name -> value } } 582185029Spjd * 583185029Spjd * "origin" -> number (guid) (if clone) 584185029Spjd * "sent" -> boolean (not on-disk) 585185029Spjd * } 586185029Spjd * } 587185029Spjd * } 588185029Spjd * 589185029Spjd */ 590185029Spjd} send_data_t; 591185029Spjd 592185029Spjdstatic void send_iterate_prop(zfs_handle_t *zhp, nvlist_t *nv); 593185029Spjd 594185029Spjdstatic int 595185029Spjdsend_iterate_snap(zfs_handle_t *zhp, void *arg) 596185029Spjd{ 597185029Spjd send_data_t *sd = arg; 598185029Spjd uint64_t guid = zhp->zfs_dmustats.dds_guid; 599185029Spjd char *snapname; 600185029Spjd nvlist_t *nv; 601185029Spjd 602185029Spjd snapname = strrchr(zhp->zfs_name, '@')+1; 603185029Spjd 604185029Spjd VERIFY(0 == nvlist_add_uint64(sd->parent_snaps, snapname, guid)); 605185029Spjd /* 606185029Spjd * NB: if there is no fromsnap here (it's a newly created fs in 607185029Spjd * an incremental replication), we will substitute the tosnap. 608185029Spjd */ 609185029Spjd if ((sd->fromsnap && strcmp(snapname, sd->fromsnap) == 0) || 610185029Spjd (sd->parent_fromsnap_guid == 0 && sd->tosnap && 611185029Spjd strcmp(snapname, sd->tosnap) == 0)) { 612185029Spjd sd->parent_fromsnap_guid = guid; 613185029Spjd } 614185029Spjd 615185029Spjd VERIFY(0 == nvlist_alloc(&nv, NV_UNIQUE_NAME, 0)); 616185029Spjd send_iterate_prop(zhp, nv); 617185029Spjd VERIFY(0 == nvlist_add_nvlist(sd->snapprops, snapname, nv)); 618185029Spjd nvlist_free(nv); 619185029Spjd 620185029Spjd zfs_close(zhp); 621185029Spjd return (0); 622185029Spjd} 623185029Spjd 624185029Spjdstatic void 625185029Spjdsend_iterate_prop(zfs_handle_t *zhp, nvlist_t *nv) 626185029Spjd{ 627185029Spjd nvpair_t *elem = NULL; 628185029Spjd 629185029Spjd while ((elem = nvlist_next_nvpair(zhp->zfs_props, elem)) != NULL) { 630185029Spjd char *propname = nvpair_name(elem); 631185029Spjd zfs_prop_t prop = zfs_name_to_prop(propname); 632185029Spjd nvlist_t *propnv; 633185029Spjd 634219089Spjd if (!zfs_prop_user(propname)) { 635219089Spjd /* 636219089Spjd * Realistically, this should never happen. However, 637219089Spjd * we want the ability to add DSL properties without 638219089Spjd * needing to make incompatible version changes. We 639219089Spjd * need to ignore unknown properties to allow older 640219089Spjd * software to still send datasets containing these 641219089Spjd * properties, with the unknown properties elided. 642219089Spjd */ 643219089Spjd if (prop == ZPROP_INVAL) 644219089Spjd continue; 645209962Smm 646219089Spjd if (zfs_prop_readonly(prop)) 647219089Spjd continue; 648219089Spjd } 649185029Spjd 650185029Spjd verify(nvpair_value_nvlist(elem, &propnv) == 0); 651219089Spjd if (prop == ZFS_PROP_QUOTA || prop == ZFS_PROP_RESERVATION || 652219089Spjd prop == ZFS_PROP_REFQUOTA || 653219089Spjd prop == ZFS_PROP_REFRESERVATION) { 654219089Spjd char *source; 655185029Spjd uint64_t value; 656185029Spjd verify(nvlist_lookup_uint64(propnv, 657185029Spjd ZPROP_VALUE, &value) == 0); 658185029Spjd if (zhp->zfs_type == ZFS_TYPE_SNAPSHOT) 659185029Spjd continue; 660219089Spjd /* 661219089Spjd * May have no source before SPA_VERSION_RECVD_PROPS, 662219089Spjd * but is still modifiable. 663219089Spjd */ 664219089Spjd if (nvlist_lookup_string(propnv, 665219089Spjd ZPROP_SOURCE, &source) == 0) { 666219089Spjd if ((strcmp(source, zhp->zfs_name) != 0) && 667219089Spjd (strcmp(source, 668219089Spjd ZPROP_SOURCE_VAL_RECVD) != 0)) 669219089Spjd continue; 670219089Spjd } 671185029Spjd } else { 672185029Spjd char *source; 673185029Spjd if (nvlist_lookup_string(propnv, 674185029Spjd ZPROP_SOURCE, &source) != 0) 675185029Spjd continue; 676219089Spjd if ((strcmp(source, zhp->zfs_name) != 0) && 677219089Spjd (strcmp(source, ZPROP_SOURCE_VAL_RECVD) != 0)) 678185029Spjd continue; 679185029Spjd } 680185029Spjd 681185029Spjd if (zfs_prop_user(propname) || 682185029Spjd zfs_prop_get_type(prop) == PROP_TYPE_STRING) { 683185029Spjd char *value; 684185029Spjd verify(nvlist_lookup_string(propnv, 685185029Spjd ZPROP_VALUE, &value) == 0); 686185029Spjd VERIFY(0 == nvlist_add_string(nv, propname, value)); 687185029Spjd } else { 688185029Spjd uint64_t value; 689185029Spjd verify(nvlist_lookup_uint64(propnv, 690185029Spjd ZPROP_VALUE, &value) == 0); 691185029Spjd VERIFY(0 == nvlist_add_uint64(nv, propname, value)); 692185029Spjd } 693185029Spjd } 694185029Spjd} 695185029Spjd 696219089Spjd/* 697219089Spjd * recursively generate nvlists describing datasets. See comment 698219089Spjd * for the data structure send_data_t above for description of contents 699219089Spjd * of the nvlist. 700219089Spjd */ 701185029Spjdstatic int 702185029Spjdsend_iterate_fs(zfs_handle_t *zhp, void *arg) 703185029Spjd{ 704185029Spjd send_data_t *sd = arg; 705185029Spjd nvlist_t *nvfs, *nv; 706219089Spjd int rv = 0; 707185029Spjd uint64_t parent_fromsnap_guid_save = sd->parent_fromsnap_guid; 708185029Spjd uint64_t guid = zhp->zfs_dmustats.dds_guid; 709185029Spjd char guidstring[64]; 710185029Spjd 711185029Spjd VERIFY(0 == nvlist_alloc(&nvfs, NV_UNIQUE_NAME, 0)); 712185029Spjd VERIFY(0 == nvlist_add_string(nvfs, "name", zhp->zfs_name)); 713185029Spjd VERIFY(0 == nvlist_add_uint64(nvfs, "parentfromsnap", 714185029Spjd sd->parent_fromsnap_guid)); 715185029Spjd 716185029Spjd if (zhp->zfs_dmustats.dds_origin[0]) { 717185029Spjd zfs_handle_t *origin = zfs_open(zhp->zfs_hdl, 718185029Spjd zhp->zfs_dmustats.dds_origin, ZFS_TYPE_SNAPSHOT); 719185029Spjd if (origin == NULL) 720185029Spjd return (-1); 721185029Spjd VERIFY(0 == nvlist_add_uint64(nvfs, "origin", 722185029Spjd origin->zfs_dmustats.dds_guid)); 723185029Spjd } 724185029Spjd 725185029Spjd /* iterate over props */ 726185029Spjd VERIFY(0 == nvlist_alloc(&nv, NV_UNIQUE_NAME, 0)); 727185029Spjd send_iterate_prop(zhp, nv); 728185029Spjd VERIFY(0 == nvlist_add_nvlist(nvfs, "props", nv)); 729185029Spjd nvlist_free(nv); 730185029Spjd 731185029Spjd /* iterate over snaps, and set sd->parent_fromsnap_guid */ 732185029Spjd sd->parent_fromsnap_guid = 0; 733185029Spjd VERIFY(0 == nvlist_alloc(&sd->parent_snaps, NV_UNIQUE_NAME, 0)); 734185029Spjd VERIFY(0 == nvlist_alloc(&sd->snapprops, NV_UNIQUE_NAME, 0)); 735244194Ssmh (void) zfs_iter_snapshots_sorted(zhp, send_iterate_snap, sd); 736185029Spjd VERIFY(0 == nvlist_add_nvlist(nvfs, "snaps", sd->parent_snaps)); 737185029Spjd VERIFY(0 == nvlist_add_nvlist(nvfs, "snapprops", sd->snapprops)); 738185029Spjd nvlist_free(sd->parent_snaps); 739185029Spjd nvlist_free(sd->snapprops); 740185029Spjd 741185029Spjd /* add this fs to nvlist */ 742185029Spjd (void) snprintf(guidstring, sizeof (guidstring), 743185029Spjd "0x%llx", (longlong_t)guid); 744185029Spjd VERIFY(0 == nvlist_add_nvlist(sd->fss, guidstring, nvfs)); 745185029Spjd nvlist_free(nvfs); 746185029Spjd 747185029Spjd /* iterate over children */ 748219089Spjd if (sd->recursive) 749219089Spjd rv = zfs_iter_filesystems(zhp, send_iterate_fs, sd); 750185029Spjd 751185029Spjd sd->parent_fromsnap_guid = parent_fromsnap_guid_save; 752185029Spjd 753185029Spjd zfs_close(zhp); 754185029Spjd return (rv); 755185029Spjd} 756185029Spjd 757185029Spjdstatic int 758185029Spjdgather_nvlist(libzfs_handle_t *hdl, const char *fsname, const char *fromsnap, 759219089Spjd const char *tosnap, boolean_t recursive, nvlist_t **nvlp, avl_tree_t **avlp) 760185029Spjd{ 761185029Spjd zfs_handle_t *zhp; 762185029Spjd send_data_t sd = { 0 }; 763185029Spjd int error; 764185029Spjd 765185029Spjd zhp = zfs_open(hdl, fsname, ZFS_TYPE_FILESYSTEM | ZFS_TYPE_VOLUME); 766185029Spjd if (zhp == NULL) 767185029Spjd return (EZFS_BADTYPE); 768185029Spjd 769185029Spjd VERIFY(0 == nvlist_alloc(&sd.fss, NV_UNIQUE_NAME, 0)); 770185029Spjd sd.fromsnap = fromsnap; 771185029Spjd sd.tosnap = tosnap; 772219089Spjd sd.recursive = recursive; 773185029Spjd 774185029Spjd if ((error = send_iterate_fs(zhp, &sd)) != 0) { 775185029Spjd nvlist_free(sd.fss); 776185029Spjd if (avlp != NULL) 777185029Spjd *avlp = NULL; 778185029Spjd *nvlp = NULL; 779185029Spjd return (error); 780185029Spjd } 781185029Spjd 782185029Spjd if (avlp != NULL && (*avlp = fsavl_create(sd.fss)) == NULL) { 783185029Spjd nvlist_free(sd.fss); 784185029Spjd *nvlp = NULL; 785185029Spjd return (EZFS_NOMEM); 786185029Spjd } 787185029Spjd 788185029Spjd *nvlp = sd.fss; 789185029Spjd return (0); 790185029Spjd} 791185029Spjd 792185029Spjd/* 793185029Spjd * Routines specific to "zfs send" 794185029Spjd */ 795185029Spjdtypedef struct send_dump_data { 796185029Spjd /* these are all just the short snapname (the part after the @) */ 797185029Spjd const char *fromsnap; 798185029Spjd const char *tosnap; 799219089Spjd char prevsnap[ZFS_MAXNAMELEN]; 800219089Spjd uint64_t prevsnap_obj; 801185029Spjd boolean_t seenfrom, seento, replicate, doall, fromorigin; 802235222Smm boolean_t verbose, dryrun, parsable, progress; 803185029Spjd int outfd; 804185029Spjd boolean_t err; 805185029Spjd nvlist_t *fss; 806251646Sdelphij nvlist_t *snapholds; 807185029Spjd avl_tree_t *fsavl; 808219089Spjd snapfilter_cb_t *filter_cb; 809219089Spjd void *filter_cb_arg; 810219089Spjd nvlist_t *debugnv; 811219089Spjd char holdtag[ZFS_MAXNAMELEN]; 812219089Spjd int cleanup_fd; 813228103Smm uint64_t size; 814185029Spjd} send_dump_data_t; 815185029Spjd 816228103Smmstatic int 817228103Smmestimate_ioctl(zfs_handle_t *zhp, uint64_t fromsnap_obj, 818228103Smm boolean_t fromorigin, uint64_t *sizep) 819228103Smm{ 820228103Smm zfs_cmd_t zc = { 0 }; 821228103Smm libzfs_handle_t *hdl = zhp->zfs_hdl; 822228103Smm 823228103Smm assert(zhp->zfs_type == ZFS_TYPE_SNAPSHOT); 824228103Smm assert(fromsnap_obj == 0 || !fromorigin); 825228103Smm 826228103Smm (void) strlcpy(zc.zc_name, zhp->zfs_name, sizeof (zc.zc_name)); 827228103Smm zc.zc_obj = fromorigin; 828228103Smm zc.zc_sendobj = zfs_prop_get_int(zhp, ZFS_PROP_OBJSETID); 829228103Smm zc.zc_fromobj = fromsnap_obj; 830228103Smm zc.zc_guid = 1; /* estimate flag */ 831228103Smm 832228103Smm if (zfs_ioctl(zhp->zfs_hdl, ZFS_IOC_SEND, &zc) != 0) { 833228103Smm char errbuf[1024]; 834228103Smm (void) snprintf(errbuf, sizeof (errbuf), dgettext(TEXT_DOMAIN, 835228103Smm "warning: cannot estimate space for '%s'"), zhp->zfs_name); 836228103Smm 837228103Smm switch (errno) { 838228103Smm case EXDEV: 839228103Smm zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, 840228103Smm "not an earlier snapshot from the same fs")); 841228103Smm return (zfs_error(hdl, EZFS_CROSSTARGET, errbuf)); 842228103Smm 843228103Smm case ENOENT: 844228103Smm if (zfs_dataset_exists(hdl, zc.zc_name, 845228103Smm ZFS_TYPE_SNAPSHOT)) { 846228103Smm zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, 847228103Smm "incremental source (@%s) does not exist"), 848228103Smm zc.zc_value); 849228103Smm } 850228103Smm return (zfs_error(hdl, EZFS_NOENT, errbuf)); 851228103Smm 852228103Smm case EDQUOT: 853228103Smm case EFBIG: 854228103Smm case EIO: 855228103Smm case ENOLINK: 856228103Smm case ENOSPC: 857228103Smm case ENXIO: 858228103Smm case EPIPE: 859228103Smm case ERANGE: 860228103Smm case EFAULT: 861228103Smm case EROFS: 862228103Smm zfs_error_aux(hdl, strerror(errno)); 863228103Smm return (zfs_error(hdl, EZFS_BADBACKUP, errbuf)); 864228103Smm 865228103Smm default: 866228103Smm return (zfs_standard_error(hdl, errno, errbuf)); 867228103Smm } 868228103Smm } 869228103Smm 870228103Smm *sizep = zc.zc_objset_type; 871228103Smm 872228103Smm return (0); 873228103Smm} 874228103Smm 875185029Spjd/* 876185029Spjd * Dumps a backup of the given snapshot (incremental from fromsnap if it's not 877185029Spjd * NULL) to the file descriptor specified by outfd. 878185029Spjd */ 879185029Spjdstatic int 880219089Spjddump_ioctl(zfs_handle_t *zhp, const char *fromsnap, uint64_t fromsnap_obj, 881219089Spjd boolean_t fromorigin, int outfd, nvlist_t *debugnv) 882185029Spjd{ 883185029Spjd zfs_cmd_t zc = { 0 }; 884185029Spjd libzfs_handle_t *hdl = zhp->zfs_hdl; 885219089Spjd nvlist_t *thisdbg; 886185029Spjd 887185029Spjd assert(zhp->zfs_type == ZFS_TYPE_SNAPSHOT); 888219089Spjd assert(fromsnap_obj == 0 || !fromorigin); 889185029Spjd 890185029Spjd (void) strlcpy(zc.zc_name, zhp->zfs_name, sizeof (zc.zc_name)); 891185029Spjd zc.zc_cookie = outfd; 892185029Spjd zc.zc_obj = fromorigin; 893219089Spjd zc.zc_sendobj = zfs_prop_get_int(zhp, ZFS_PROP_OBJSETID); 894219089Spjd zc.zc_fromobj = fromsnap_obj; 895185029Spjd 896219089Spjd VERIFY(0 == nvlist_alloc(&thisdbg, NV_UNIQUE_NAME, 0)); 897219089Spjd if (fromsnap && fromsnap[0] != '\0') { 898219089Spjd VERIFY(0 == nvlist_add_string(thisdbg, 899219089Spjd "fromsnap", fromsnap)); 900219089Spjd } 901219089Spjd 902228103Smm if (zfs_ioctl(zhp->zfs_hdl, ZFS_IOC_SEND, &zc) != 0) { 903185029Spjd char errbuf[1024]; 904185029Spjd (void) snprintf(errbuf, sizeof (errbuf), dgettext(TEXT_DOMAIN, 905185029Spjd "warning: cannot send '%s'"), zhp->zfs_name); 906185029Spjd 907219089Spjd VERIFY(0 == nvlist_add_uint64(thisdbg, "error", errno)); 908219089Spjd if (debugnv) { 909219089Spjd VERIFY(0 == nvlist_add_nvlist(debugnv, 910219089Spjd zhp->zfs_name, thisdbg)); 911219089Spjd } 912219089Spjd nvlist_free(thisdbg); 913219089Spjd 914185029Spjd switch (errno) { 915185029Spjd case EXDEV: 916185029Spjd zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, 917185029Spjd "not an earlier snapshot from the same fs")); 918185029Spjd return (zfs_error(hdl, EZFS_CROSSTARGET, errbuf)); 919185029Spjd 920185029Spjd case ENOENT: 921185029Spjd if (zfs_dataset_exists(hdl, zc.zc_name, 922185029Spjd ZFS_TYPE_SNAPSHOT)) { 923185029Spjd zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, 924185029Spjd "incremental source (@%s) does not exist"), 925185029Spjd zc.zc_value); 926185029Spjd } 927185029Spjd return (zfs_error(hdl, EZFS_NOENT, errbuf)); 928185029Spjd 929185029Spjd case EDQUOT: 930185029Spjd case EFBIG: 931185029Spjd case EIO: 932185029Spjd case ENOLINK: 933185029Spjd case ENOSPC: 934228103Smm#ifdef sun 935228103Smm case ENOSTR: 936228103Smm#endif 937185029Spjd case ENXIO: 938185029Spjd case EPIPE: 939185029Spjd case ERANGE: 940185029Spjd case EFAULT: 941185029Spjd case EROFS: 942185029Spjd zfs_error_aux(hdl, strerror(errno)); 943185029Spjd return (zfs_error(hdl, EZFS_BADBACKUP, errbuf)); 944185029Spjd 945185029Spjd default: 946185029Spjd return (zfs_standard_error(hdl, errno, errbuf)); 947185029Spjd } 948185029Spjd } 949185029Spjd 950219089Spjd if (debugnv) 951219089Spjd VERIFY(0 == nvlist_add_nvlist(debugnv, zhp->zfs_name, thisdbg)); 952219089Spjd nvlist_free(thisdbg); 953219089Spjd 954185029Spjd return (0); 955185029Spjd} 956185029Spjd 957251646Sdelphijstatic void 958251646Sdelphijgather_holds(zfs_handle_t *zhp, send_dump_data_t *sdd) 959219089Spjd{ 960219089Spjd assert(zhp->zfs_type == ZFS_TYPE_SNAPSHOT); 961219089Spjd 962219089Spjd /* 963251646Sdelphij * zfs_send() only sets snapholds for sends that need them, 964219089Spjd * e.g. replication and doall. 965219089Spjd */ 966251646Sdelphij if (sdd->snapholds == NULL) 967251646Sdelphij return; 968219089Spjd 969251646Sdelphij fnvlist_add_string(sdd->snapholds, zhp->zfs_name, sdd->holdtag); 970219089Spjd} 971219089Spjd 972235222Smmstatic void * 973235222Smmsend_progress_thread(void *arg) 974235222Smm{ 975235222Smm progress_arg_t *pa = arg; 976235222Smm 977235222Smm zfs_cmd_t zc = { 0 }; 978235222Smm zfs_handle_t *zhp = pa->pa_zhp; 979235222Smm libzfs_handle_t *hdl = zhp->zfs_hdl; 980235222Smm unsigned long long bytes; 981235222Smm char buf[16]; 982235222Smm 983235222Smm time_t t; 984235222Smm struct tm *tm; 985235222Smm 986235222Smm assert(zhp->zfs_type == ZFS_TYPE_SNAPSHOT); 987235222Smm (void) strlcpy(zc.zc_name, zhp->zfs_name, sizeof (zc.zc_name)); 988235222Smm 989235222Smm if (!pa->pa_parsable) 990235222Smm (void) fprintf(stderr, "TIME SENT SNAPSHOT\n"); 991235222Smm 992235222Smm /* 993235222Smm * Print the progress from ZFS_IOC_SEND_PROGRESS every second. 994235222Smm */ 995235222Smm for (;;) { 996235222Smm (void) sleep(1); 997235222Smm 998235222Smm zc.zc_cookie = pa->pa_fd; 999235222Smm if (zfs_ioctl(hdl, ZFS_IOC_SEND_PROGRESS, &zc) != 0) 1000235222Smm return ((void *)-1); 1001235222Smm 1002235222Smm (void) time(&t); 1003235222Smm tm = localtime(&t); 1004235222Smm bytes = zc.zc_cookie; 1005235222Smm 1006235222Smm if (pa->pa_parsable) { 1007235222Smm (void) fprintf(stderr, "%02d:%02d:%02d\t%llu\t%s\n", 1008235222Smm tm->tm_hour, tm->tm_min, tm->tm_sec, 1009235222Smm bytes, zhp->zfs_name); 1010235222Smm } else { 1011235222Smm zfs_nicenum(bytes, buf, sizeof (buf)); 1012235222Smm (void) fprintf(stderr, "%02d:%02d:%02d %5s %s\n", 1013235222Smm tm->tm_hour, tm->tm_min, tm->tm_sec, 1014235222Smm buf, zhp->zfs_name); 1015235222Smm } 1016235222Smm } 1017235222Smm} 1018235222Smm 1019219089Spjdstatic int 1020185029Spjddump_snapshot(zfs_handle_t *zhp, void *arg) 1021185029Spjd{ 1022185029Spjd send_dump_data_t *sdd = arg; 1023235222Smm progress_arg_t pa = { 0 }; 1024235222Smm pthread_t tid; 1025219089Spjd char *thissnap; 1026185029Spjd int err; 1027228103Smm boolean_t isfromsnap, istosnap, fromorigin; 1028219089Spjd boolean_t exclude = B_FALSE; 1029185029Spjd 1030251646Sdelphij err = 0; 1031185029Spjd thissnap = strchr(zhp->zfs_name, '@') + 1; 1032219089Spjd isfromsnap = (sdd->fromsnap != NULL && 1033219089Spjd strcmp(sdd->fromsnap, thissnap) == 0); 1034185029Spjd 1035219089Spjd if (!sdd->seenfrom && isfromsnap) { 1036251646Sdelphij gather_holds(zhp, sdd); 1037251646Sdelphij sdd->seenfrom = B_TRUE; 1038251646Sdelphij (void) strcpy(sdd->prevsnap, thissnap); 1039251646Sdelphij sdd->prevsnap_obj = zfs_prop_get_int(zhp, ZFS_PROP_OBJSETID); 1040185029Spjd zfs_close(zhp); 1041251646Sdelphij return (0); 1042185029Spjd } 1043185029Spjd 1044185029Spjd if (sdd->seento || !sdd->seenfrom) { 1045185029Spjd zfs_close(zhp); 1046185029Spjd return (0); 1047185029Spjd } 1048185029Spjd 1049219089Spjd istosnap = (strcmp(sdd->tosnap, thissnap) == 0); 1050219089Spjd if (istosnap) 1051219089Spjd sdd->seento = B_TRUE; 1052219089Spjd 1053219089Spjd if (!sdd->doall && !isfromsnap && !istosnap) { 1054219089Spjd if (sdd->replicate) { 1055219089Spjd char *snapname; 1056219089Spjd nvlist_t *snapprops; 1057219089Spjd /* 1058219089Spjd * Filter out all intermediate snapshots except origin 1059219089Spjd * snapshots needed to replicate clones. 1060219089Spjd */ 1061219089Spjd nvlist_t *nvfs = fsavl_find(sdd->fsavl, 1062219089Spjd zhp->zfs_dmustats.dds_guid, &snapname); 1063219089Spjd 1064219089Spjd VERIFY(0 == nvlist_lookup_nvlist(nvfs, 1065219089Spjd "snapprops", &snapprops)); 1066219089Spjd VERIFY(0 == nvlist_lookup_nvlist(snapprops, 1067219089Spjd thissnap, &snapprops)); 1068219089Spjd exclude = !nvlist_exists(snapprops, "is_clone_origin"); 1069219089Spjd } else { 1070219089Spjd exclude = B_TRUE; 1071219089Spjd } 1072219089Spjd } 1073219089Spjd 1074219089Spjd /* 1075219089Spjd * If a filter function exists, call it to determine whether 1076219089Spjd * this snapshot will be sent. 1077219089Spjd */ 1078219089Spjd if (exclude || (sdd->filter_cb != NULL && 1079219089Spjd sdd->filter_cb(zhp, sdd->filter_cb_arg) == B_FALSE)) { 1080219089Spjd /* 1081219089Spjd * This snapshot is filtered out. Don't send it, and don't 1082219089Spjd * set prevsnap_obj, so it will be as if this snapshot didn't 1083219089Spjd * exist, and the next accepted snapshot will be sent as 1084219089Spjd * an incremental from the last accepted one, or as the 1085219089Spjd * first (and full) snapshot in the case of a replication, 1086219089Spjd * non-incremental send. 1087219089Spjd */ 1088219089Spjd zfs_close(zhp); 1089219089Spjd return (0); 1090219089Spjd } 1091219089Spjd 1092251646Sdelphij gather_holds(zhp, sdd); 1093228103Smm fromorigin = sdd->prevsnap[0] == '\0' && 1094228103Smm (sdd->fromorigin || sdd->replicate); 1095228103Smm 1096185029Spjd if (sdd->verbose) { 1097228103Smm uint64_t size; 1098228103Smm err = estimate_ioctl(zhp, sdd->prevsnap_obj, 1099228103Smm fromorigin, &size); 1100228103Smm 1101228103Smm if (sdd->parsable) { 1102228103Smm if (sdd->prevsnap[0] != '\0') { 1103228103Smm (void) fprintf(stderr, "incremental\t%s\t%s", 1104228103Smm sdd->prevsnap, zhp->zfs_name); 1105228103Smm } else { 1106228103Smm (void) fprintf(stderr, "full\t%s", 1107228103Smm zhp->zfs_name); 1108228103Smm } 1109228103Smm } else { 1110228103Smm (void) fprintf(stderr, dgettext(TEXT_DOMAIN, 1111228103Smm "send from @%s to %s"), 1112228103Smm sdd->prevsnap, zhp->zfs_name); 1113228103Smm } 1114228103Smm if (err == 0) { 1115228103Smm if (sdd->parsable) { 1116228103Smm (void) fprintf(stderr, "\t%llu\n", 1117228103Smm (longlong_t)size); 1118228103Smm } else { 1119228103Smm char buf[16]; 1120228103Smm zfs_nicenum(size, buf, sizeof (buf)); 1121228103Smm (void) fprintf(stderr, dgettext(TEXT_DOMAIN, 1122228103Smm " estimated size is %s\n"), buf); 1123228103Smm } 1124228103Smm sdd->size += size; 1125228103Smm } else { 1126228103Smm (void) fprintf(stderr, "\n"); 1127228103Smm } 1128185029Spjd } 1129185029Spjd 1130228103Smm if (!sdd->dryrun) { 1131235222Smm /* 1132235222Smm * If progress reporting is requested, spawn a new thread to 1133235222Smm * poll ZFS_IOC_SEND_PROGRESS at a regular interval. 1134235222Smm */ 1135235222Smm if (sdd->progress) { 1136235222Smm pa.pa_zhp = zhp; 1137235222Smm pa.pa_fd = sdd->outfd; 1138235222Smm pa.pa_parsable = sdd->parsable; 1139235222Smm 1140235222Smm if (err = pthread_create(&tid, NULL, 1141235222Smm send_progress_thread, &pa)) { 1142235222Smm zfs_close(zhp); 1143235222Smm return (err); 1144235222Smm } 1145235222Smm } 1146235222Smm 1147228103Smm err = dump_ioctl(zhp, sdd->prevsnap, sdd->prevsnap_obj, 1148228103Smm fromorigin, sdd->outfd, sdd->debugnv); 1149235222Smm 1150235222Smm if (sdd->progress) { 1151235222Smm (void) pthread_cancel(tid); 1152235222Smm (void) pthread_join(tid, NULL); 1153235222Smm } 1154228103Smm } 1155185029Spjd 1156219089Spjd (void) strcpy(sdd->prevsnap, thissnap); 1157219089Spjd sdd->prevsnap_obj = zfs_prop_get_int(zhp, ZFS_PROP_OBJSETID); 1158185029Spjd zfs_close(zhp); 1159185029Spjd return (err); 1160185029Spjd} 1161185029Spjd 1162185029Spjdstatic int 1163185029Spjddump_filesystem(zfs_handle_t *zhp, void *arg) 1164185029Spjd{ 1165185029Spjd int rv = 0; 1166185029Spjd send_dump_data_t *sdd = arg; 1167185029Spjd boolean_t missingfrom = B_FALSE; 1168185029Spjd zfs_cmd_t zc = { 0 }; 1169185029Spjd 1170185029Spjd (void) snprintf(zc.zc_name, sizeof (zc.zc_name), "%s@%s", 1171185029Spjd zhp->zfs_name, sdd->tosnap); 1172185029Spjd if (ioctl(zhp->zfs_hdl->libzfs_fd, ZFS_IOC_OBJSET_STATS, &zc) != 0) { 1173228103Smm (void) fprintf(stderr, dgettext(TEXT_DOMAIN, 1174228103Smm "WARNING: could not send %s@%s: does not exist\n"), 1175185029Spjd zhp->zfs_name, sdd->tosnap); 1176185029Spjd sdd->err = B_TRUE; 1177185029Spjd return (0); 1178185029Spjd } 1179185029Spjd 1180185029Spjd if (sdd->replicate && sdd->fromsnap) { 1181185029Spjd /* 1182185029Spjd * If this fs does not have fromsnap, and we're doing 1183185029Spjd * recursive, we need to send a full stream from the 1184185029Spjd * beginning (or an incremental from the origin if this 1185185029Spjd * is a clone). If we're doing non-recursive, then let 1186185029Spjd * them get the error. 1187185029Spjd */ 1188185029Spjd (void) snprintf(zc.zc_name, sizeof (zc.zc_name), "%s@%s", 1189185029Spjd zhp->zfs_name, sdd->fromsnap); 1190185029Spjd if (ioctl(zhp->zfs_hdl->libzfs_fd, 1191185029Spjd ZFS_IOC_OBJSET_STATS, &zc) != 0) { 1192185029Spjd missingfrom = B_TRUE; 1193185029Spjd } 1194185029Spjd } 1195185029Spjd 1196219089Spjd sdd->seenfrom = sdd->seento = sdd->prevsnap[0] = 0; 1197219089Spjd sdd->prevsnap_obj = 0; 1198219089Spjd if (sdd->fromsnap == NULL || missingfrom) 1199219089Spjd sdd->seenfrom = B_TRUE; 1200185029Spjd 1201219089Spjd rv = zfs_iter_snapshots_sorted(zhp, dump_snapshot, arg); 1202219089Spjd if (!sdd->seenfrom) { 1203228103Smm (void) fprintf(stderr, dgettext(TEXT_DOMAIN, 1204219089Spjd "WARNING: could not send %s@%s:\n" 1205228103Smm "incremental source (%s@%s) does not exist\n"), 1206219089Spjd zhp->zfs_name, sdd->tosnap, 1207219089Spjd zhp->zfs_name, sdd->fromsnap); 1208219089Spjd sdd->err = B_TRUE; 1209219089Spjd } else if (!sdd->seento) { 1210219089Spjd if (sdd->fromsnap) { 1211228103Smm (void) fprintf(stderr, dgettext(TEXT_DOMAIN, 1212185029Spjd "WARNING: could not send %s@%s:\n" 1213219089Spjd "incremental source (%s@%s) " 1214228103Smm "is not earlier than it\n"), 1215185029Spjd zhp->zfs_name, sdd->tosnap, 1216185029Spjd zhp->zfs_name, sdd->fromsnap); 1217185029Spjd } else { 1218228103Smm (void) fprintf(stderr, dgettext(TEXT_DOMAIN, 1219228103Smm "WARNING: " 1220228103Smm "could not send %s@%s: does not exist\n"), 1221219089Spjd zhp->zfs_name, sdd->tosnap); 1222185029Spjd } 1223219089Spjd sdd->err = B_TRUE; 1224185029Spjd } 1225185029Spjd 1226185029Spjd return (rv); 1227185029Spjd} 1228185029Spjd 1229185029Spjdstatic int 1230185029Spjddump_filesystems(zfs_handle_t *rzhp, void *arg) 1231185029Spjd{ 1232185029Spjd send_dump_data_t *sdd = arg; 1233185029Spjd nvpair_t *fspair; 1234185029Spjd boolean_t needagain, progress; 1235185029Spjd 1236185029Spjd if (!sdd->replicate) 1237185029Spjd return (dump_filesystem(rzhp, sdd)); 1238185029Spjd 1239219089Spjd /* Mark the clone origin snapshots. */ 1240219089Spjd for (fspair = nvlist_next_nvpair(sdd->fss, NULL); fspair; 1241219089Spjd fspair = nvlist_next_nvpair(sdd->fss, fspair)) { 1242219089Spjd nvlist_t *nvfs; 1243219089Spjd uint64_t origin_guid = 0; 1244219089Spjd 1245219089Spjd VERIFY(0 == nvpair_value_nvlist(fspair, &nvfs)); 1246219089Spjd (void) nvlist_lookup_uint64(nvfs, "origin", &origin_guid); 1247219089Spjd if (origin_guid != 0) { 1248219089Spjd char *snapname; 1249219089Spjd nvlist_t *origin_nv = fsavl_find(sdd->fsavl, 1250219089Spjd origin_guid, &snapname); 1251219089Spjd if (origin_nv != NULL) { 1252219089Spjd nvlist_t *snapprops; 1253219089Spjd VERIFY(0 == nvlist_lookup_nvlist(origin_nv, 1254219089Spjd "snapprops", &snapprops)); 1255219089Spjd VERIFY(0 == nvlist_lookup_nvlist(snapprops, 1256219089Spjd snapname, &snapprops)); 1257219089Spjd VERIFY(0 == nvlist_add_boolean( 1258219089Spjd snapprops, "is_clone_origin")); 1259219089Spjd } 1260219089Spjd } 1261219089Spjd } 1262185029Spjdagain: 1263185029Spjd needagain = progress = B_FALSE; 1264185029Spjd for (fspair = nvlist_next_nvpair(sdd->fss, NULL); fspair; 1265185029Spjd fspair = nvlist_next_nvpair(sdd->fss, fspair)) { 1266228103Smm nvlist_t *fslist, *parent_nv; 1267185029Spjd char *fsname; 1268185029Spjd zfs_handle_t *zhp; 1269185029Spjd int err; 1270185029Spjd uint64_t origin_guid = 0; 1271228103Smm uint64_t parent_guid = 0; 1272185029Spjd 1273185029Spjd VERIFY(nvpair_value_nvlist(fspair, &fslist) == 0); 1274185029Spjd if (nvlist_lookup_boolean(fslist, "sent") == 0) 1275185029Spjd continue; 1276185029Spjd 1277185029Spjd VERIFY(nvlist_lookup_string(fslist, "name", &fsname) == 0); 1278185029Spjd (void) nvlist_lookup_uint64(fslist, "origin", &origin_guid); 1279228103Smm (void) nvlist_lookup_uint64(fslist, "parentfromsnap", 1280228103Smm &parent_guid); 1281185029Spjd 1282228103Smm if (parent_guid != 0) { 1283228103Smm parent_nv = fsavl_find(sdd->fsavl, parent_guid, NULL); 1284228103Smm if (!nvlist_exists(parent_nv, "sent")) { 1285228103Smm /* parent has not been sent; skip this one */ 1286228103Smm needagain = B_TRUE; 1287228103Smm continue; 1288228103Smm } 1289228103Smm } 1290228103Smm 1291219089Spjd if (origin_guid != 0) { 1292219089Spjd nvlist_t *origin_nv = fsavl_find(sdd->fsavl, 1293219089Spjd origin_guid, NULL); 1294219089Spjd if (origin_nv != NULL && 1295228103Smm !nvlist_exists(origin_nv, "sent")) { 1296219089Spjd /* 1297219089Spjd * origin has not been sent yet; 1298219089Spjd * skip this clone. 1299219089Spjd */ 1300219089Spjd needagain = B_TRUE; 1301219089Spjd continue; 1302219089Spjd } 1303185029Spjd } 1304185029Spjd 1305185029Spjd zhp = zfs_open(rzhp->zfs_hdl, fsname, ZFS_TYPE_DATASET); 1306185029Spjd if (zhp == NULL) 1307185029Spjd return (-1); 1308185029Spjd err = dump_filesystem(zhp, sdd); 1309185029Spjd VERIFY(nvlist_add_boolean(fslist, "sent") == 0); 1310185029Spjd progress = B_TRUE; 1311185029Spjd zfs_close(zhp); 1312185029Spjd if (err) 1313185029Spjd return (err); 1314185029Spjd } 1315185029Spjd if (needagain) { 1316185029Spjd assert(progress); 1317185029Spjd goto again; 1318185029Spjd } 1319228103Smm 1320228103Smm /* clean out the sent flags in case we reuse this fss */ 1321228103Smm for (fspair = nvlist_next_nvpair(sdd->fss, NULL); fspair; 1322228103Smm fspair = nvlist_next_nvpair(sdd->fss, fspair)) { 1323228103Smm nvlist_t *fslist; 1324228103Smm 1325228103Smm VERIFY(nvpair_value_nvlist(fspair, &fslist) == 0); 1326228103Smm (void) nvlist_remove_all(fslist, "sent"); 1327228103Smm } 1328228103Smm 1329185029Spjd return (0); 1330185029Spjd} 1331185029Spjd 1332185029Spjd/* 1333219089Spjd * Generate a send stream for the dataset identified by the argument zhp. 1334219089Spjd * 1335219089Spjd * The content of the send stream is the snapshot identified by 1336219089Spjd * 'tosnap'. Incremental streams are requested in two ways: 1337219089Spjd * - from the snapshot identified by "fromsnap" (if non-null) or 1338219089Spjd * - from the origin of the dataset identified by zhp, which must 1339219089Spjd * be a clone. In this case, "fromsnap" is null and "fromorigin" 1340219089Spjd * is TRUE. 1341219089Spjd * 1342219089Spjd * The send stream is recursive (i.e. dumps a hierarchy of snapshots) and 1343219089Spjd * uses a special header (with a hdrtype field of DMU_COMPOUNDSTREAM) 1344219089Spjd * if "replicate" is set. If "doall" is set, dump all the intermediate 1345219089Spjd * snapshots. The DMU_COMPOUNDSTREAM header is used in the "doall" 1346219089Spjd * case too. If "props" is set, send properties. 1347185029Spjd */ 1348185029Spjdint 1349185029Spjdzfs_send(zfs_handle_t *zhp, const char *fromsnap, const char *tosnap, 1350228103Smm sendflags_t *flags, int outfd, snapfilter_cb_t filter_func, 1351219089Spjd void *cb_arg, nvlist_t **debugnvp) 1352185029Spjd{ 1353185029Spjd char errbuf[1024]; 1354185029Spjd send_dump_data_t sdd = { 0 }; 1355228103Smm int err = 0; 1356185029Spjd nvlist_t *fss = NULL; 1357185029Spjd avl_tree_t *fsavl = NULL; 1358219089Spjd static uint64_t holdseq; 1359219089Spjd int spa_version; 1360251646Sdelphij pthread_t tid = 0; 1361219089Spjd int pipefd[2]; 1362219089Spjd dedup_arg_t dda = { 0 }; 1363219089Spjd int featureflags = 0; 1364185029Spjd 1365185029Spjd (void) snprintf(errbuf, sizeof (errbuf), dgettext(TEXT_DOMAIN, 1366185029Spjd "cannot send '%s'"), zhp->zfs_name); 1367185029Spjd 1368185029Spjd if (fromsnap && fromsnap[0] == '\0') { 1369185029Spjd zfs_error_aux(zhp->zfs_hdl, dgettext(TEXT_DOMAIN, 1370185029Spjd "zero-length incremental source")); 1371185029Spjd return (zfs_error(zhp->zfs_hdl, EZFS_NOENT, errbuf)); 1372185029Spjd } 1373185029Spjd 1374219089Spjd if (zhp->zfs_type == ZFS_TYPE_FILESYSTEM) { 1375219089Spjd uint64_t version; 1376219089Spjd version = zfs_prop_get_int(zhp, ZFS_PROP_VERSION); 1377219089Spjd if (version >= ZPL_VERSION_SA) { 1378219089Spjd featureflags |= DMU_BACKUP_FEATURE_SA_SPILL; 1379219089Spjd } 1380219089Spjd } 1381219089Spjd 1382228103Smm if (flags->dedup && !flags->dryrun) { 1383219089Spjd featureflags |= (DMU_BACKUP_FEATURE_DEDUP | 1384219089Spjd DMU_BACKUP_FEATURE_DEDUPPROPS); 1385219089Spjd if (err = pipe(pipefd)) { 1386219089Spjd zfs_error_aux(zhp->zfs_hdl, strerror(errno)); 1387219089Spjd return (zfs_error(zhp->zfs_hdl, EZFS_PIPEFAILED, 1388219089Spjd errbuf)); 1389219089Spjd } 1390219089Spjd dda.outputfd = outfd; 1391219089Spjd dda.inputfd = pipefd[1]; 1392219089Spjd dda.dedup_hdl = zhp->zfs_hdl; 1393219089Spjd if (err = pthread_create(&tid, NULL, cksummer, &dda)) { 1394219089Spjd (void) close(pipefd[0]); 1395219089Spjd (void) close(pipefd[1]); 1396219089Spjd zfs_error_aux(zhp->zfs_hdl, strerror(errno)); 1397219089Spjd return (zfs_error(zhp->zfs_hdl, 1398219089Spjd EZFS_THREADCREATEFAILED, errbuf)); 1399219089Spjd } 1400219089Spjd } 1401219089Spjd 1402228103Smm if (flags->replicate || flags->doall || flags->props) { 1403185029Spjd dmu_replay_record_t drr = { 0 }; 1404185029Spjd char *packbuf = NULL; 1405185029Spjd size_t buflen = 0; 1406185029Spjd zio_cksum_t zc = { 0 }; 1407185029Spjd 1408228103Smm if (flags->replicate || flags->props) { 1409185029Spjd nvlist_t *hdrnv; 1410185029Spjd 1411185029Spjd VERIFY(0 == nvlist_alloc(&hdrnv, NV_UNIQUE_NAME, 0)); 1412185029Spjd if (fromsnap) { 1413185029Spjd VERIFY(0 == nvlist_add_string(hdrnv, 1414185029Spjd "fromsnap", fromsnap)); 1415185029Spjd } 1416185029Spjd VERIFY(0 == nvlist_add_string(hdrnv, "tosnap", tosnap)); 1417228103Smm if (!flags->replicate) { 1418219089Spjd VERIFY(0 == nvlist_add_boolean(hdrnv, 1419219089Spjd "not_recursive")); 1420219089Spjd } 1421185029Spjd 1422185029Spjd err = gather_nvlist(zhp->zfs_hdl, zhp->zfs_name, 1423228103Smm fromsnap, tosnap, flags->replicate, &fss, &fsavl); 1424185029Spjd if (err) 1425219089Spjd goto err_out; 1426185029Spjd VERIFY(0 == nvlist_add_nvlist(hdrnv, "fss", fss)); 1427185029Spjd err = nvlist_pack(hdrnv, &packbuf, &buflen, 1428185029Spjd NV_ENCODE_XDR, 0); 1429219089Spjd if (debugnvp) 1430219089Spjd *debugnvp = hdrnv; 1431219089Spjd else 1432219089Spjd nvlist_free(hdrnv); 1433251646Sdelphij if (err) 1434219089Spjd goto stderr_out; 1435185029Spjd } 1436185029Spjd 1437228103Smm if (!flags->dryrun) { 1438228103Smm /* write first begin record */ 1439228103Smm drr.drr_type = DRR_BEGIN; 1440228103Smm drr.drr_u.drr_begin.drr_magic = DMU_BACKUP_MAGIC; 1441228103Smm DMU_SET_STREAM_HDRTYPE(drr.drr_u.drr_begin. 1442228103Smm drr_versioninfo, DMU_COMPOUNDSTREAM); 1443228103Smm DMU_SET_FEATUREFLAGS(drr.drr_u.drr_begin. 1444228103Smm drr_versioninfo, featureflags); 1445228103Smm (void) snprintf(drr.drr_u.drr_begin.drr_toname, 1446228103Smm sizeof (drr.drr_u.drr_begin.drr_toname), 1447228103Smm "%s@%s", zhp->zfs_name, tosnap); 1448228103Smm drr.drr_payloadlen = buflen; 1449228103Smm err = cksum_and_write(&drr, sizeof (drr), &zc, outfd); 1450185029Spjd 1451228103Smm /* write header nvlist */ 1452228103Smm if (err != -1 && packbuf != NULL) { 1453228103Smm err = cksum_and_write(packbuf, buflen, &zc, 1454228103Smm outfd); 1455228103Smm } 1456228103Smm free(packbuf); 1457228103Smm if (err == -1) { 1458228103Smm err = errno; 1459228103Smm goto stderr_out; 1460228103Smm } 1461185029Spjd 1462228103Smm /* write end record */ 1463185029Spjd bzero(&drr, sizeof (drr)); 1464185029Spjd drr.drr_type = DRR_END; 1465185029Spjd drr.drr_u.drr_end.drr_checksum = zc; 1466185029Spjd err = write(outfd, &drr, sizeof (drr)); 1467185029Spjd if (err == -1) { 1468219089Spjd err = errno; 1469219089Spjd goto stderr_out; 1470185029Spjd } 1471228103Smm 1472228103Smm err = 0; 1473185029Spjd } 1474185029Spjd } 1475185029Spjd 1476185029Spjd /* dump each stream */ 1477185029Spjd sdd.fromsnap = fromsnap; 1478185029Spjd sdd.tosnap = tosnap; 1479251646Sdelphij if (tid != 0) 1480219089Spjd sdd.outfd = pipefd[0]; 1481219089Spjd else 1482219089Spjd sdd.outfd = outfd; 1483228103Smm sdd.replicate = flags->replicate; 1484228103Smm sdd.doall = flags->doall; 1485228103Smm sdd.fromorigin = flags->fromorigin; 1486185029Spjd sdd.fss = fss; 1487185029Spjd sdd.fsavl = fsavl; 1488228103Smm sdd.verbose = flags->verbose; 1489228103Smm sdd.parsable = flags->parsable; 1490235222Smm sdd.progress = flags->progress; 1491228103Smm sdd.dryrun = flags->dryrun; 1492219089Spjd sdd.filter_cb = filter_func; 1493219089Spjd sdd.filter_cb_arg = cb_arg; 1494219089Spjd if (debugnvp) 1495219089Spjd sdd.debugnv = *debugnvp; 1496238422Smm 1497238422Smm /* 1498238422Smm * Some flags require that we place user holds on the datasets that are 1499238422Smm * being sent so they don't get destroyed during the send. We can skip 1500238422Smm * this step if the pool is imported read-only since the datasets cannot 1501238422Smm * be destroyed. 1502238422Smm */ 1503238422Smm if (!flags->dryrun && !zpool_get_prop_int(zfs_get_pool_handle(zhp), 1504238422Smm ZPOOL_PROP_READONLY, NULL) && 1505238422Smm zfs_spa_version(zhp, &spa_version) == 0 && 1506238422Smm spa_version >= SPA_VERSION_USERREFS && 1507238422Smm (flags->doall || flags->replicate)) { 1508219089Spjd ++holdseq; 1509219089Spjd (void) snprintf(sdd.holdtag, sizeof (sdd.holdtag), 1510219089Spjd ".send-%d-%llu", getpid(), (u_longlong_t)holdseq); 1511219089Spjd sdd.cleanup_fd = open(ZFS_DEV, O_RDWR|O_EXCL); 1512219089Spjd if (sdd.cleanup_fd < 0) { 1513219089Spjd err = errno; 1514219089Spjd goto stderr_out; 1515219089Spjd } 1516251646Sdelphij sdd.snapholds = fnvlist_alloc(); 1517219089Spjd } else { 1518219089Spjd sdd.cleanup_fd = -1; 1519251646Sdelphij sdd.snapholds = NULL; 1520219089Spjd } 1521251646Sdelphij if (flags->verbose || sdd.snapholds != NULL) { 1522228103Smm /* 1523228103Smm * Do a verbose no-op dry run to get all the verbose output 1524251646Sdelphij * or to gather snapshot hold's before generating any data, 1525251646Sdelphij * then do a non-verbose real run to generate the streams. 1526228103Smm */ 1527228103Smm sdd.dryrun = B_TRUE; 1528228103Smm err = dump_filesystems(zhp, &sdd); 1529251646Sdelphij 1530251646Sdelphij if (err != 0) 1531251646Sdelphij goto stderr_out; 1532251646Sdelphij 1533251646Sdelphij if (flags->verbose) { 1534251646Sdelphij if (flags->parsable) { 1535251646Sdelphij (void) fprintf(stderr, "size\t%llu\n", 1536251646Sdelphij (longlong_t)sdd.size); 1537251646Sdelphij } else { 1538251646Sdelphij char buf[16]; 1539251646Sdelphij zfs_nicenum(sdd.size, buf, sizeof (buf)); 1540251646Sdelphij (void) fprintf(stderr, dgettext(TEXT_DOMAIN, 1541251646Sdelphij "total estimated size is %s\n"), buf); 1542251646Sdelphij } 1543251646Sdelphij } 1544251646Sdelphij 1545251646Sdelphij /* Ensure no snaps found is treated as an error. */ 1546251646Sdelphij if (!sdd.seento) { 1547251646Sdelphij err = ENOENT; 1548251646Sdelphij goto err_out; 1549251646Sdelphij } 1550251646Sdelphij 1551251646Sdelphij /* Skip the second run if dryrun was requested. */ 1552251646Sdelphij if (flags->dryrun) 1553251646Sdelphij goto err_out; 1554251646Sdelphij 1555251646Sdelphij if (sdd.snapholds != NULL) { 1556251646Sdelphij err = zfs_hold_nvl(zhp, sdd.cleanup_fd, sdd.snapholds); 1557251646Sdelphij if (err != 0) 1558251646Sdelphij goto stderr_out; 1559251646Sdelphij 1560251646Sdelphij fnvlist_free(sdd.snapholds); 1561251646Sdelphij sdd.snapholds = NULL; 1562251646Sdelphij } 1563251646Sdelphij 1564251646Sdelphij sdd.dryrun = B_FALSE; 1565228103Smm sdd.verbose = B_FALSE; 1566228103Smm } 1567251646Sdelphij 1568185029Spjd err = dump_filesystems(zhp, &sdd); 1569185029Spjd fsavl_destroy(fsavl); 1570185029Spjd nvlist_free(fss); 1571185029Spjd 1572251646Sdelphij /* Ensure no snaps found is treated as an error. */ 1573251646Sdelphij if (err == 0 && !sdd.seento) 1574251646Sdelphij err = ENOENT; 1575251646Sdelphij 1576251646Sdelphij if (tid != 0) { 1577251646Sdelphij if (err != 0) 1578251646Sdelphij (void) pthread_cancel(tid); 1579253818Ssmh (void) close(pipefd[0]); 1580251646Sdelphij (void) pthread_join(tid, NULL); 1581219089Spjd } 1582219089Spjd 1583219089Spjd if (sdd.cleanup_fd != -1) { 1584219089Spjd VERIFY(0 == close(sdd.cleanup_fd)); 1585219089Spjd sdd.cleanup_fd = -1; 1586219089Spjd } 1587219089Spjd 1588228103Smm if (!flags->dryrun && (flags->replicate || flags->doall || 1589228103Smm flags->props)) { 1590185029Spjd /* 1591185029Spjd * write final end record. NB: want to do this even if 1592185029Spjd * there was some error, because it might not be totally 1593185029Spjd * failed. 1594185029Spjd */ 1595185029Spjd dmu_replay_record_t drr = { 0 }; 1596185029Spjd drr.drr_type = DRR_END; 1597185029Spjd if (write(outfd, &drr, sizeof (drr)) == -1) { 1598185029Spjd return (zfs_standard_error(zhp->zfs_hdl, 1599185029Spjd errno, errbuf)); 1600185029Spjd } 1601185029Spjd } 1602185029Spjd 1603185029Spjd return (err || sdd.err); 1604219089Spjd 1605219089Spjdstderr_out: 1606219089Spjd err = zfs_standard_error(zhp->zfs_hdl, err, errbuf); 1607219089Spjderr_out: 1608251646Sdelphij fsavl_destroy(fsavl); 1609251646Sdelphij nvlist_free(fss); 1610251646Sdelphij fnvlist_free(sdd.snapholds); 1611251646Sdelphij 1612219089Spjd if (sdd.cleanup_fd != -1) 1613219089Spjd VERIFY(0 == close(sdd.cleanup_fd)); 1614251646Sdelphij if (tid != 0) { 1615219089Spjd (void) pthread_cancel(tid); 1616253818Ssmh (void) close(pipefd[0]); 1617219089Spjd (void) pthread_join(tid, NULL); 1618219089Spjd } 1619219089Spjd return (err); 1620185029Spjd} 1621185029Spjd 1622185029Spjd/* 1623185029Spjd * Routines specific to "zfs recv" 1624185029Spjd */ 1625185029Spjd 1626185029Spjdstatic int 1627185029Spjdrecv_read(libzfs_handle_t *hdl, int fd, void *buf, int ilen, 1628185029Spjd boolean_t byteswap, zio_cksum_t *zc) 1629185029Spjd{ 1630185029Spjd char *cp = buf; 1631185029Spjd int rv; 1632185029Spjd int len = ilen; 1633185029Spjd 1634185029Spjd do { 1635185029Spjd rv = read(fd, cp, len); 1636185029Spjd cp += rv; 1637185029Spjd len -= rv; 1638185029Spjd } while (rv > 0); 1639185029Spjd 1640185029Spjd if (rv < 0 || len != 0) { 1641185029Spjd zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, 1642185029Spjd "failed to read from stream")); 1643185029Spjd return (zfs_error(hdl, EZFS_BADSTREAM, dgettext(TEXT_DOMAIN, 1644185029Spjd "cannot receive"))); 1645185029Spjd } 1646185029Spjd 1647185029Spjd if (zc) { 1648185029Spjd if (byteswap) 1649185029Spjd fletcher_4_incremental_byteswap(buf, ilen, zc); 1650185029Spjd else 1651185029Spjd fletcher_4_incremental_native(buf, ilen, zc); 1652185029Spjd } 1653185029Spjd return (0); 1654185029Spjd} 1655185029Spjd 1656185029Spjdstatic int 1657185029Spjdrecv_read_nvlist(libzfs_handle_t *hdl, int fd, int len, nvlist_t **nvp, 1658185029Spjd boolean_t byteswap, zio_cksum_t *zc) 1659185029Spjd{ 1660185029Spjd char *buf; 1661185029Spjd int err; 1662185029Spjd 1663185029Spjd buf = zfs_alloc(hdl, len); 1664185029Spjd if (buf == NULL) 1665185029Spjd return (ENOMEM); 1666185029Spjd 1667185029Spjd err = recv_read(hdl, fd, buf, len, byteswap, zc); 1668185029Spjd if (err != 0) { 1669185029Spjd free(buf); 1670185029Spjd return (err); 1671185029Spjd } 1672185029Spjd 1673185029Spjd err = nvlist_unpack(buf, len, nvp, 0); 1674185029Spjd free(buf); 1675185029Spjd if (err != 0) { 1676185029Spjd zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, "invalid " 1677185029Spjd "stream (malformed nvlist)")); 1678185029Spjd return (EINVAL); 1679185029Spjd } 1680185029Spjd return (0); 1681185029Spjd} 1682185029Spjd 1683185029Spjdstatic int 1684185029Spjdrecv_rename(libzfs_handle_t *hdl, const char *name, const char *tryname, 1685228103Smm int baselen, char *newname, recvflags_t *flags) 1686185029Spjd{ 1687185029Spjd static int seq; 1688185029Spjd zfs_cmd_t zc = { 0 }; 1689185029Spjd int err; 1690185029Spjd prop_changelist_t *clp; 1691185029Spjd zfs_handle_t *zhp; 1692185029Spjd 1693185029Spjd zhp = zfs_open(hdl, name, ZFS_TYPE_DATASET); 1694185029Spjd if (zhp == NULL) 1695185029Spjd return (-1); 1696185029Spjd clp = changelist_gather(zhp, ZFS_PROP_NAME, 0, 1697228103Smm flags->force ? MS_FORCE : 0); 1698185029Spjd zfs_close(zhp); 1699185029Spjd if (clp == NULL) 1700185029Spjd return (-1); 1701185029Spjd err = changelist_prefix(clp); 1702185029Spjd if (err) 1703185029Spjd return (err); 1704185029Spjd 1705219089Spjd zc.zc_objset_type = DMU_OST_ZFS; 1706219089Spjd (void) strlcpy(zc.zc_name, name, sizeof (zc.zc_name)); 1707219089Spjd 1708185029Spjd if (tryname) { 1709185029Spjd (void) strcpy(newname, tryname); 1710185029Spjd 1711185029Spjd (void) strlcpy(zc.zc_value, tryname, sizeof (zc.zc_value)); 1712185029Spjd 1713228103Smm if (flags->verbose) { 1714185029Spjd (void) printf("attempting rename %s to %s\n", 1715185029Spjd zc.zc_name, zc.zc_value); 1716185029Spjd } 1717185029Spjd err = ioctl(hdl->libzfs_fd, ZFS_IOC_RENAME, &zc); 1718185029Spjd if (err == 0) 1719185029Spjd changelist_rename(clp, name, tryname); 1720185029Spjd } else { 1721185029Spjd err = ENOENT; 1722185029Spjd } 1723185029Spjd 1724248571Smm if (err != 0 && strncmp(name + baselen, "recv-", 5) != 0) { 1725185029Spjd seq++; 1726185029Spjd 1727248571Smm (void) snprintf(newname, ZFS_MAXNAMELEN, "%.*srecv-%u-%u", 1728248571Smm baselen, name, getpid(), seq); 1729185029Spjd (void) strlcpy(zc.zc_value, newname, sizeof (zc.zc_value)); 1730185029Spjd 1731228103Smm if (flags->verbose) { 1732185029Spjd (void) printf("failed - trying rename %s to %s\n", 1733185029Spjd zc.zc_name, zc.zc_value); 1734185029Spjd } 1735185029Spjd err = ioctl(hdl->libzfs_fd, ZFS_IOC_RENAME, &zc); 1736185029Spjd if (err == 0) 1737185029Spjd changelist_rename(clp, name, newname); 1738228103Smm if (err && flags->verbose) { 1739185029Spjd (void) printf("failed (%u) - " 1740185029Spjd "will try again on next pass\n", errno); 1741185029Spjd } 1742185029Spjd err = EAGAIN; 1743228103Smm } else if (flags->verbose) { 1744185029Spjd if (err == 0) 1745185029Spjd (void) printf("success\n"); 1746185029Spjd else 1747185029Spjd (void) printf("failed (%u)\n", errno); 1748185029Spjd } 1749185029Spjd 1750185029Spjd (void) changelist_postfix(clp); 1751185029Spjd changelist_free(clp); 1752185029Spjd 1753185029Spjd return (err); 1754185029Spjd} 1755185029Spjd 1756185029Spjdstatic int 1757185029Spjdrecv_destroy(libzfs_handle_t *hdl, const char *name, int baselen, 1758228103Smm char *newname, recvflags_t *flags) 1759185029Spjd{ 1760185029Spjd zfs_cmd_t zc = { 0 }; 1761185029Spjd int err = 0; 1762185029Spjd prop_changelist_t *clp; 1763185029Spjd zfs_handle_t *zhp; 1764219089Spjd boolean_t defer = B_FALSE; 1765219089Spjd int spa_version; 1766185029Spjd 1767185029Spjd zhp = zfs_open(hdl, name, ZFS_TYPE_DATASET); 1768185029Spjd if (zhp == NULL) 1769185029Spjd return (-1); 1770185029Spjd clp = changelist_gather(zhp, ZFS_PROP_NAME, 0, 1771228103Smm flags->force ? MS_FORCE : 0); 1772219089Spjd if (zfs_get_type(zhp) == ZFS_TYPE_SNAPSHOT && 1773219089Spjd zfs_spa_version(zhp, &spa_version) == 0 && 1774219089Spjd spa_version >= SPA_VERSION_USERREFS) 1775219089Spjd defer = B_TRUE; 1776185029Spjd zfs_close(zhp); 1777185029Spjd if (clp == NULL) 1778185029Spjd return (-1); 1779185029Spjd err = changelist_prefix(clp); 1780185029Spjd if (err) 1781185029Spjd return (err); 1782185029Spjd 1783185029Spjd zc.zc_objset_type = DMU_OST_ZFS; 1784219089Spjd zc.zc_defer_destroy = defer; 1785185029Spjd (void) strlcpy(zc.zc_name, name, sizeof (zc.zc_name)); 1786185029Spjd 1787228103Smm if (flags->verbose) 1788185029Spjd (void) printf("attempting destroy %s\n", zc.zc_name); 1789185029Spjd err = ioctl(hdl->libzfs_fd, ZFS_IOC_DESTROY, &zc); 1790185029Spjd if (err == 0) { 1791228103Smm if (flags->verbose) 1792185029Spjd (void) printf("success\n"); 1793185029Spjd changelist_remove(clp, zc.zc_name); 1794185029Spjd } 1795185029Spjd 1796185029Spjd (void) changelist_postfix(clp); 1797185029Spjd changelist_free(clp); 1798185029Spjd 1799219089Spjd /* 1800219089Spjd * Deferred destroy might destroy the snapshot or only mark it to be 1801219089Spjd * destroyed later, and it returns success in either case. 1802219089Spjd */ 1803219089Spjd if (err != 0 || (defer && zfs_dataset_exists(hdl, name, 1804219089Spjd ZFS_TYPE_SNAPSHOT))) { 1805185029Spjd err = recv_rename(hdl, name, NULL, baselen, newname, flags); 1806219089Spjd } 1807185029Spjd 1808185029Spjd return (err); 1809185029Spjd} 1810185029Spjd 1811185029Spjdtypedef struct guid_to_name_data { 1812185029Spjd uint64_t guid; 1813185029Spjd char *name; 1814228103Smm char *skip; 1815185029Spjd} guid_to_name_data_t; 1816185029Spjd 1817185029Spjdstatic int 1818185029Spjdguid_to_name_cb(zfs_handle_t *zhp, void *arg) 1819185029Spjd{ 1820185029Spjd guid_to_name_data_t *gtnd = arg; 1821185029Spjd int err; 1822185029Spjd 1823228103Smm if (gtnd->skip != NULL && 1824228103Smm strcmp(zhp->zfs_name, gtnd->skip) == 0) { 1825228103Smm return (0); 1826228103Smm } 1827228103Smm 1828185029Spjd if (zhp->zfs_dmustats.dds_guid == gtnd->guid) { 1829185029Spjd (void) strcpy(gtnd->name, zhp->zfs_name); 1830219089Spjd zfs_close(zhp); 1831185029Spjd return (EEXIST); 1832185029Spjd } 1833228103Smm 1834185029Spjd err = zfs_iter_children(zhp, guid_to_name_cb, gtnd); 1835185029Spjd zfs_close(zhp); 1836185029Spjd return (err); 1837185029Spjd} 1838185029Spjd 1839228103Smm/* 1840228103Smm * Attempt to find the local dataset associated with this guid. In the case of 1841228103Smm * multiple matches, we attempt to find the "best" match by searching 1842228103Smm * progressively larger portions of the hierarchy. This allows one to send a 1843228103Smm * tree of datasets individually and guarantee that we will find the source 1844228103Smm * guid within that hierarchy, even if there are multiple matches elsewhere. 1845228103Smm */ 1846185029Spjdstatic int 1847185029Spjdguid_to_name(libzfs_handle_t *hdl, const char *parent, uint64_t guid, 1848185029Spjd char *name) 1849185029Spjd{ 1850185029Spjd /* exhaustive search all local snapshots */ 1851228103Smm char pname[ZFS_MAXNAMELEN]; 1852185029Spjd guid_to_name_data_t gtnd; 1853185029Spjd int err = 0; 1854185029Spjd zfs_handle_t *zhp; 1855185029Spjd char *cp; 1856185029Spjd 1857185029Spjd gtnd.guid = guid; 1858185029Spjd gtnd.name = name; 1859228103Smm gtnd.skip = NULL; 1860185029Spjd 1861228103Smm (void) strlcpy(pname, parent, sizeof (pname)); 1862185029Spjd 1863228103Smm /* 1864228103Smm * Search progressively larger portions of the hierarchy. This will 1865228103Smm * select the "most local" version of the origin snapshot in the case 1866228103Smm * that there are multiple matching snapshots in the system. 1867228103Smm */ 1868228103Smm while ((cp = strrchr(pname, '/')) != NULL) { 1869228103Smm 1870228103Smm /* Chop off the last component and open the parent */ 1871185029Spjd *cp = '\0'; 1872228103Smm zhp = make_dataset_handle(hdl, pname); 1873185029Spjd 1874228103Smm if (zhp == NULL) 1875228103Smm continue; 1876228103Smm 1877185029Spjd err = zfs_iter_children(zhp, guid_to_name_cb, >nd); 1878185029Spjd zfs_close(zhp); 1879228103Smm if (err == EEXIST) 1880228103Smm return (0); 1881228103Smm 1882228103Smm /* 1883228103Smm * Remember the dataset that we already searched, so we 1884228103Smm * skip it next time through. 1885228103Smm */ 1886228103Smm gtnd.skip = pname; 1887185029Spjd } 1888185029Spjd 1889228103Smm return (ENOENT); 1890185029Spjd} 1891185029Spjd 1892185029Spjd/* 1893228103Smm * Return +1 if guid1 is before guid2, 0 if they are the same, and -1 if 1894228103Smm * guid1 is after guid2. 1895185029Spjd */ 1896185029Spjdstatic int 1897185029Spjdcreated_before(libzfs_handle_t *hdl, avl_tree_t *avl, 1898185029Spjd uint64_t guid1, uint64_t guid2) 1899185029Spjd{ 1900185029Spjd nvlist_t *nvfs; 1901185029Spjd char *fsname, *snapname; 1902185029Spjd char buf[ZFS_MAXNAMELEN]; 1903185029Spjd int rv; 1904228103Smm zfs_handle_t *guid1hdl, *guid2hdl; 1905228103Smm uint64_t create1, create2; 1906185029Spjd 1907185029Spjd if (guid2 == 0) 1908185029Spjd return (0); 1909185029Spjd if (guid1 == 0) 1910185029Spjd return (1); 1911185029Spjd 1912185029Spjd nvfs = fsavl_find(avl, guid1, &snapname); 1913185029Spjd VERIFY(0 == nvlist_lookup_string(nvfs, "name", &fsname)); 1914185029Spjd (void) snprintf(buf, sizeof (buf), "%s@%s", fsname, snapname); 1915228103Smm guid1hdl = zfs_open(hdl, buf, ZFS_TYPE_SNAPSHOT); 1916228103Smm if (guid1hdl == NULL) 1917185029Spjd return (-1); 1918185029Spjd 1919185029Spjd nvfs = fsavl_find(avl, guid2, &snapname); 1920185029Spjd VERIFY(0 == nvlist_lookup_string(nvfs, "name", &fsname)); 1921185029Spjd (void) snprintf(buf, sizeof (buf), "%s@%s", fsname, snapname); 1922228103Smm guid2hdl = zfs_open(hdl, buf, ZFS_TYPE_SNAPSHOT); 1923228103Smm if (guid2hdl == NULL) { 1924228103Smm zfs_close(guid1hdl); 1925185029Spjd return (-1); 1926185029Spjd } 1927185029Spjd 1928228103Smm create1 = zfs_prop_get_int(guid1hdl, ZFS_PROP_CREATETXG); 1929228103Smm create2 = zfs_prop_get_int(guid2hdl, ZFS_PROP_CREATETXG); 1930185029Spjd 1931228103Smm if (create1 < create2) 1932228103Smm rv = -1; 1933228103Smm else if (create1 > create2) 1934228103Smm rv = +1; 1935228103Smm else 1936228103Smm rv = 0; 1937185029Spjd 1938228103Smm zfs_close(guid1hdl); 1939228103Smm zfs_close(guid2hdl); 1940228103Smm 1941185029Spjd return (rv); 1942185029Spjd} 1943185029Spjd 1944185029Spjdstatic int 1945185029Spjdrecv_incremental_replication(libzfs_handle_t *hdl, const char *tofs, 1946228103Smm recvflags_t *flags, nvlist_t *stream_nv, avl_tree_t *stream_avl, 1947219089Spjd nvlist_t *renamed) 1948185029Spjd{ 1949244194Ssmh nvlist_t *local_nv, *deleted = NULL; 1950185029Spjd avl_tree_t *local_avl; 1951185029Spjd nvpair_t *fselem, *nextfselem; 1952219089Spjd char *fromsnap; 1953185029Spjd char newname[ZFS_MAXNAMELEN]; 1954244194Ssmh char guidname[32]; 1955185029Spjd int error; 1956219089Spjd boolean_t needagain, progress, recursive; 1957209962Smm char *s1, *s2; 1958185029Spjd 1959185029Spjd VERIFY(0 == nvlist_lookup_string(stream_nv, "fromsnap", &fromsnap)); 1960185029Spjd 1961219089Spjd recursive = (nvlist_lookup_boolean(stream_nv, "not_recursive") == 1962219089Spjd ENOENT); 1963219089Spjd 1964228103Smm if (flags->dryrun) 1965185029Spjd return (0); 1966185029Spjd 1967185029Spjdagain: 1968185029Spjd needagain = progress = B_FALSE; 1969185029Spjd 1970244194Ssmh VERIFY(0 == nvlist_alloc(&deleted, NV_UNIQUE_NAME, 0)); 1971244194Ssmh 1972185029Spjd if ((error = gather_nvlist(hdl, tofs, fromsnap, NULL, 1973219089Spjd recursive, &local_nv, &local_avl)) != 0) 1974185029Spjd return (error); 1975185029Spjd 1976185029Spjd /* 1977185029Spjd * Process deletes and renames 1978185029Spjd */ 1979185029Spjd for (fselem = nvlist_next_nvpair(local_nv, NULL); 1980185029Spjd fselem; fselem = nextfselem) { 1981185029Spjd nvlist_t *nvfs, *snaps; 1982185029Spjd nvlist_t *stream_nvfs = NULL; 1983185029Spjd nvpair_t *snapelem, *nextsnapelem; 1984185029Spjd uint64_t fromguid = 0; 1985185029Spjd uint64_t originguid = 0; 1986185029Spjd uint64_t stream_originguid = 0; 1987185029Spjd uint64_t parent_fromsnap_guid, stream_parent_fromsnap_guid; 1988219089Spjd char *fsname, *stream_fsname; 1989185029Spjd 1990185029Spjd nextfselem = nvlist_next_nvpair(local_nv, fselem); 1991185029Spjd 1992185029Spjd VERIFY(0 == nvpair_value_nvlist(fselem, &nvfs)); 1993185029Spjd VERIFY(0 == nvlist_lookup_nvlist(nvfs, "snaps", &snaps)); 1994185029Spjd VERIFY(0 == nvlist_lookup_string(nvfs, "name", &fsname)); 1995185029Spjd VERIFY(0 == nvlist_lookup_uint64(nvfs, "parentfromsnap", 1996185029Spjd &parent_fromsnap_guid)); 1997185029Spjd (void) nvlist_lookup_uint64(nvfs, "origin", &originguid); 1998185029Spjd 1999185029Spjd /* 2000185029Spjd * First find the stream's fs, so we can check for 2001185029Spjd * a different origin (due to "zfs promote") 2002185029Spjd */ 2003185029Spjd for (snapelem = nvlist_next_nvpair(snaps, NULL); 2004185029Spjd snapelem; snapelem = nvlist_next_nvpair(snaps, snapelem)) { 2005185029Spjd uint64_t thisguid; 2006185029Spjd 2007185029Spjd VERIFY(0 == nvpair_value_uint64(snapelem, &thisguid)); 2008185029Spjd stream_nvfs = fsavl_find(stream_avl, thisguid, NULL); 2009185029Spjd 2010185029Spjd if (stream_nvfs != NULL) 2011185029Spjd break; 2012185029Spjd } 2013185029Spjd 2014185029Spjd /* check for promote */ 2015185029Spjd (void) nvlist_lookup_uint64(stream_nvfs, "origin", 2016185029Spjd &stream_originguid); 2017185029Spjd if (stream_nvfs && originguid != stream_originguid) { 2018185029Spjd switch (created_before(hdl, local_avl, 2019185029Spjd stream_originguid, originguid)) { 2020185029Spjd case 1: { 2021185029Spjd /* promote it! */ 2022185029Spjd zfs_cmd_t zc = { 0 }; 2023185029Spjd nvlist_t *origin_nvfs; 2024185029Spjd char *origin_fsname; 2025185029Spjd 2026228103Smm if (flags->verbose) 2027185029Spjd (void) printf("promoting %s\n", fsname); 2028185029Spjd 2029185029Spjd origin_nvfs = fsavl_find(local_avl, originguid, 2030185029Spjd NULL); 2031185029Spjd VERIFY(0 == nvlist_lookup_string(origin_nvfs, 2032185029Spjd "name", &origin_fsname)); 2033185029Spjd (void) strlcpy(zc.zc_value, origin_fsname, 2034185029Spjd sizeof (zc.zc_value)); 2035185029Spjd (void) strlcpy(zc.zc_name, fsname, 2036185029Spjd sizeof (zc.zc_name)); 2037185029Spjd error = zfs_ioctl(hdl, ZFS_IOC_PROMOTE, &zc); 2038185029Spjd if (error == 0) 2039185029Spjd progress = B_TRUE; 2040185029Spjd break; 2041185029Spjd } 2042185029Spjd default: 2043185029Spjd break; 2044185029Spjd case -1: 2045185029Spjd fsavl_destroy(local_avl); 2046185029Spjd nvlist_free(local_nv); 2047185029Spjd return (-1); 2048185029Spjd } 2049185029Spjd /* 2050185029Spjd * We had/have the wrong origin, therefore our 2051185029Spjd * list of snapshots is wrong. Need to handle 2052185029Spjd * them on the next pass. 2053185029Spjd */ 2054185029Spjd needagain = B_TRUE; 2055185029Spjd continue; 2056185029Spjd } 2057185029Spjd 2058185029Spjd for (snapelem = nvlist_next_nvpair(snaps, NULL); 2059185029Spjd snapelem; snapelem = nextsnapelem) { 2060185029Spjd uint64_t thisguid; 2061185029Spjd char *stream_snapname; 2062185029Spjd nvlist_t *found, *props; 2063185029Spjd 2064185029Spjd nextsnapelem = nvlist_next_nvpair(snaps, snapelem); 2065185029Spjd 2066185029Spjd VERIFY(0 == nvpair_value_uint64(snapelem, &thisguid)); 2067185029Spjd found = fsavl_find(stream_avl, thisguid, 2068185029Spjd &stream_snapname); 2069185029Spjd 2070185029Spjd /* check for delete */ 2071185029Spjd if (found == NULL) { 2072185029Spjd char name[ZFS_MAXNAMELEN]; 2073185029Spjd 2074228103Smm if (!flags->force) 2075185029Spjd continue; 2076185029Spjd 2077185029Spjd (void) snprintf(name, sizeof (name), "%s@%s", 2078185029Spjd fsname, nvpair_name(snapelem)); 2079185029Spjd 2080185029Spjd error = recv_destroy(hdl, name, 2081185029Spjd strlen(fsname)+1, newname, flags); 2082185029Spjd if (error) 2083185029Spjd needagain = B_TRUE; 2084185029Spjd else 2085185029Spjd progress = B_TRUE; 2086244194Ssmh sprintf(guidname, "%lu", thisguid); 2087244194Ssmh nvlist_add_boolean(deleted, guidname); 2088185029Spjd continue; 2089185029Spjd } 2090185029Spjd 2091185029Spjd stream_nvfs = found; 2092185029Spjd 2093185029Spjd if (0 == nvlist_lookup_nvlist(stream_nvfs, "snapprops", 2094185029Spjd &props) && 0 == nvlist_lookup_nvlist(props, 2095185029Spjd stream_snapname, &props)) { 2096185029Spjd zfs_cmd_t zc = { 0 }; 2097185029Spjd 2098219089Spjd zc.zc_cookie = B_TRUE; /* received */ 2099185029Spjd (void) snprintf(zc.zc_name, sizeof (zc.zc_name), 2100185029Spjd "%s@%s", fsname, nvpair_name(snapelem)); 2101185029Spjd if (zcmd_write_src_nvlist(hdl, &zc, 2102185029Spjd props) == 0) { 2103185029Spjd (void) zfs_ioctl(hdl, 2104185029Spjd ZFS_IOC_SET_PROP, &zc); 2105185029Spjd zcmd_free_nvlists(&zc); 2106185029Spjd } 2107185029Spjd } 2108185029Spjd 2109185029Spjd /* check for different snapname */ 2110185029Spjd if (strcmp(nvpair_name(snapelem), 2111185029Spjd stream_snapname) != 0) { 2112185029Spjd char name[ZFS_MAXNAMELEN]; 2113185029Spjd char tryname[ZFS_MAXNAMELEN]; 2114185029Spjd 2115185029Spjd (void) snprintf(name, sizeof (name), "%s@%s", 2116185029Spjd fsname, nvpair_name(snapelem)); 2117185029Spjd (void) snprintf(tryname, sizeof (name), "%s@%s", 2118185029Spjd fsname, stream_snapname); 2119185029Spjd 2120185029Spjd error = recv_rename(hdl, name, tryname, 2121185029Spjd strlen(fsname)+1, newname, flags); 2122185029Spjd if (error) 2123185029Spjd needagain = B_TRUE; 2124185029Spjd else 2125185029Spjd progress = B_TRUE; 2126185029Spjd } 2127185029Spjd 2128185029Spjd if (strcmp(stream_snapname, fromsnap) == 0) 2129185029Spjd fromguid = thisguid; 2130185029Spjd } 2131185029Spjd 2132185029Spjd /* check for delete */ 2133185029Spjd if (stream_nvfs == NULL) { 2134228103Smm if (!flags->force) 2135185029Spjd continue; 2136185029Spjd 2137185029Spjd error = recv_destroy(hdl, fsname, strlen(tofs)+1, 2138185029Spjd newname, flags); 2139185029Spjd if (error) 2140185029Spjd needagain = B_TRUE; 2141185029Spjd else 2142185029Spjd progress = B_TRUE; 2143244194Ssmh sprintf(guidname, "%lu", parent_fromsnap_guid); 2144244194Ssmh nvlist_add_boolean(deleted, guidname); 2145185029Spjd continue; 2146185029Spjd } 2147185029Spjd 2148219089Spjd if (fromguid == 0) { 2149228103Smm if (flags->verbose) { 2150219089Spjd (void) printf("local fs %s does not have " 2151219089Spjd "fromsnap (%s in stream); must have " 2152219089Spjd "been deleted locally; ignoring\n", 2153219089Spjd fsname, fromsnap); 2154219089Spjd } 2155185029Spjd continue; 2156185029Spjd } 2157185029Spjd 2158185029Spjd VERIFY(0 == nvlist_lookup_string(stream_nvfs, 2159185029Spjd "name", &stream_fsname)); 2160185029Spjd VERIFY(0 == nvlist_lookup_uint64(stream_nvfs, 2161185029Spjd "parentfromsnap", &stream_parent_fromsnap_guid)); 2162185029Spjd 2163209962Smm s1 = strrchr(fsname, '/'); 2164209962Smm s2 = strrchr(stream_fsname, '/'); 2165209962Smm 2166219089Spjd /* 2167244194Ssmh * Check if we're going to rename based on parent guid change 2168244194Ssmh * and the current parent guid was also deleted. If it was then 2169244194Ssmh * rename will fail and is likely unneeded, so avoid this and 2170244194Ssmh * force an early retry to determine the new 2171244194Ssmh * parent_fromsnap_guid. 2172244194Ssmh */ 2173244194Ssmh if (stream_parent_fromsnap_guid != 0 && 2174244194Ssmh parent_fromsnap_guid != 0 && 2175244194Ssmh stream_parent_fromsnap_guid != parent_fromsnap_guid) { 2176244194Ssmh sprintf(guidname, "%lu", parent_fromsnap_guid); 2177244194Ssmh if (nvlist_exists(deleted, guidname)) { 2178244194Ssmh progress = B_TRUE; 2179244194Ssmh needagain = B_TRUE; 2180244194Ssmh goto doagain; 2181244194Ssmh } 2182244194Ssmh } 2183244194Ssmh 2184244194Ssmh /* 2185219089Spjd * Check for rename. If the exact receive path is specified, it 2186219089Spjd * does not count as a rename, but we still need to check the 2187219089Spjd * datasets beneath it. 2188219089Spjd */ 2189185029Spjd if ((stream_parent_fromsnap_guid != 0 && 2190219089Spjd parent_fromsnap_guid != 0 && 2191185029Spjd stream_parent_fromsnap_guid != parent_fromsnap_guid) || 2192228103Smm ((flags->isprefix || strcmp(tofs, fsname) != 0) && 2193219089Spjd (s1 != NULL) && (s2 != NULL) && strcmp(s1, s2) != 0)) { 2194185029Spjd nvlist_t *parent; 2195185029Spjd char tryname[ZFS_MAXNAMELEN]; 2196185029Spjd 2197185029Spjd parent = fsavl_find(local_avl, 2198185029Spjd stream_parent_fromsnap_guid, NULL); 2199185029Spjd /* 2200185029Spjd * NB: parent might not be found if we used the 2201185029Spjd * tosnap for stream_parent_fromsnap_guid, 2202185029Spjd * because the parent is a newly-created fs; 2203185029Spjd * we'll be able to rename it after we recv the 2204185029Spjd * new fs. 2205185029Spjd */ 2206185029Spjd if (parent != NULL) { 2207185029Spjd char *pname; 2208185029Spjd 2209185029Spjd VERIFY(0 == nvlist_lookup_string(parent, "name", 2210185029Spjd &pname)); 2211185029Spjd (void) snprintf(tryname, sizeof (tryname), 2212219089Spjd "%s%s", pname, strrchr(stream_fsname, '/')); 2213185029Spjd } else { 2214185029Spjd tryname[0] = '\0'; 2215228103Smm if (flags->verbose) { 2216185029Spjd (void) printf("local fs %s new parent " 2217185029Spjd "not found\n", fsname); 2218185029Spjd } 2219185029Spjd } 2220185029Spjd 2221219089Spjd newname[0] = '\0'; 2222219089Spjd 2223185029Spjd error = recv_rename(hdl, fsname, tryname, 2224185029Spjd strlen(tofs)+1, newname, flags); 2225219089Spjd 2226219089Spjd if (renamed != NULL && newname[0] != '\0') { 2227219089Spjd VERIFY(0 == nvlist_add_boolean(renamed, 2228219089Spjd newname)); 2229219089Spjd } 2230219089Spjd 2231185029Spjd if (error) 2232185029Spjd needagain = B_TRUE; 2233185029Spjd else 2234185029Spjd progress = B_TRUE; 2235185029Spjd } 2236185029Spjd } 2237185029Spjd 2238244194Ssmhdoagain: 2239185029Spjd fsavl_destroy(local_avl); 2240185029Spjd nvlist_free(local_nv); 2241244194Ssmh nvlist_free(deleted); 2242185029Spjd 2243185029Spjd if (needagain && progress) { 2244185029Spjd /* do another pass to fix up temporary names */ 2245228103Smm if (flags->verbose) 2246185029Spjd (void) printf("another pass:\n"); 2247185029Spjd goto again; 2248185029Spjd } 2249185029Spjd 2250185029Spjd return (needagain); 2251185029Spjd} 2252185029Spjd 2253185029Spjdstatic int 2254185029Spjdzfs_receive_package(libzfs_handle_t *hdl, int fd, const char *destname, 2255228103Smm recvflags_t *flags, dmu_replay_record_t *drr, zio_cksum_t *zc, 2256219089Spjd char **top_zfs, int cleanup_fd, uint64_t *action_handlep) 2257185029Spjd{ 2258185029Spjd nvlist_t *stream_nv = NULL; 2259185029Spjd avl_tree_t *stream_avl = NULL; 2260185029Spjd char *fromsnap = NULL; 2261219089Spjd char *cp; 2262185029Spjd char tofs[ZFS_MAXNAMELEN]; 2263219089Spjd char sendfs[ZFS_MAXNAMELEN]; 2264185029Spjd char errbuf[1024]; 2265185029Spjd dmu_replay_record_t drre; 2266185029Spjd int error; 2267185029Spjd boolean_t anyerr = B_FALSE; 2268185029Spjd boolean_t softerr = B_FALSE; 2269219089Spjd boolean_t recursive; 2270185029Spjd 2271185029Spjd (void) snprintf(errbuf, sizeof (errbuf), dgettext(TEXT_DOMAIN, 2272185029Spjd "cannot receive")); 2273185029Spjd 2274185029Spjd assert(drr->drr_type == DRR_BEGIN); 2275185029Spjd assert(drr->drr_u.drr_begin.drr_magic == DMU_BACKUP_MAGIC); 2276219089Spjd assert(DMU_GET_STREAM_HDRTYPE(drr->drr_u.drr_begin.drr_versioninfo) == 2277219089Spjd DMU_COMPOUNDSTREAM); 2278185029Spjd 2279185029Spjd /* 2280185029Spjd * Read in the nvlist from the stream. 2281185029Spjd */ 2282185029Spjd if (drr->drr_payloadlen != 0) { 2283185029Spjd error = recv_read_nvlist(hdl, fd, drr->drr_payloadlen, 2284228103Smm &stream_nv, flags->byteswap, zc); 2285185029Spjd if (error) { 2286185029Spjd error = zfs_error(hdl, EZFS_BADSTREAM, errbuf); 2287185029Spjd goto out; 2288185029Spjd } 2289185029Spjd } 2290185029Spjd 2291219089Spjd recursive = (nvlist_lookup_boolean(stream_nv, "not_recursive") == 2292219089Spjd ENOENT); 2293219089Spjd 2294219089Spjd if (recursive && strchr(destname, '@')) { 2295219089Spjd zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, 2296219089Spjd "cannot specify snapshot name for multi-snapshot stream")); 2297219089Spjd error = zfs_error(hdl, EZFS_BADSTREAM, errbuf); 2298219089Spjd goto out; 2299219089Spjd } 2300219089Spjd 2301185029Spjd /* 2302185029Spjd * Read in the end record and verify checksum. 2303185029Spjd */ 2304185029Spjd if (0 != (error = recv_read(hdl, fd, &drre, sizeof (drre), 2305228103Smm flags->byteswap, NULL))) 2306185029Spjd goto out; 2307228103Smm if (flags->byteswap) { 2308185029Spjd drre.drr_type = BSWAP_32(drre.drr_type); 2309185029Spjd drre.drr_u.drr_end.drr_checksum.zc_word[0] = 2310185029Spjd BSWAP_64(drre.drr_u.drr_end.drr_checksum.zc_word[0]); 2311185029Spjd drre.drr_u.drr_end.drr_checksum.zc_word[1] = 2312185029Spjd BSWAP_64(drre.drr_u.drr_end.drr_checksum.zc_word[1]); 2313185029Spjd drre.drr_u.drr_end.drr_checksum.zc_word[2] = 2314185029Spjd BSWAP_64(drre.drr_u.drr_end.drr_checksum.zc_word[2]); 2315185029Spjd drre.drr_u.drr_end.drr_checksum.zc_word[3] = 2316185029Spjd BSWAP_64(drre.drr_u.drr_end.drr_checksum.zc_word[3]); 2317185029Spjd } 2318185029Spjd if (drre.drr_type != DRR_END) { 2319185029Spjd error = zfs_error(hdl, EZFS_BADSTREAM, errbuf); 2320185029Spjd goto out; 2321185029Spjd } 2322185029Spjd if (!ZIO_CHECKSUM_EQUAL(drre.drr_u.drr_end.drr_checksum, *zc)) { 2323185029Spjd zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, 2324185029Spjd "incorrect header checksum")); 2325185029Spjd error = zfs_error(hdl, EZFS_BADSTREAM, errbuf); 2326185029Spjd goto out; 2327185029Spjd } 2328185029Spjd 2329185029Spjd (void) nvlist_lookup_string(stream_nv, "fromsnap", &fromsnap); 2330185029Spjd 2331185029Spjd if (drr->drr_payloadlen != 0) { 2332185029Spjd nvlist_t *stream_fss; 2333185029Spjd 2334185029Spjd VERIFY(0 == nvlist_lookup_nvlist(stream_nv, "fss", 2335185029Spjd &stream_fss)); 2336185029Spjd if ((stream_avl = fsavl_create(stream_fss)) == NULL) { 2337185029Spjd zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, 2338185029Spjd "couldn't allocate avl tree")); 2339185029Spjd error = zfs_error(hdl, EZFS_NOMEM, errbuf); 2340185029Spjd goto out; 2341185029Spjd } 2342185029Spjd 2343185029Spjd if (fromsnap != NULL) { 2344219089Spjd nvlist_t *renamed = NULL; 2345219089Spjd nvpair_t *pair = NULL; 2346219089Spjd 2347185029Spjd (void) strlcpy(tofs, destname, ZFS_MAXNAMELEN); 2348228103Smm if (flags->isprefix) { 2349219089Spjd struct drr_begin *drrb = &drr->drr_u.drr_begin; 2350219089Spjd int i; 2351219089Spjd 2352228103Smm if (flags->istail) { 2353219089Spjd cp = strrchr(drrb->drr_toname, '/'); 2354219089Spjd if (cp == NULL) { 2355219089Spjd (void) strlcat(tofs, "/", 2356219089Spjd ZFS_MAXNAMELEN); 2357219089Spjd i = 0; 2358219089Spjd } else { 2359219089Spjd i = (cp - drrb->drr_toname); 2360219089Spjd } 2361219089Spjd } else { 2362219089Spjd i = strcspn(drrb->drr_toname, "/@"); 2363219089Spjd } 2364185029Spjd /* zfs_receive_one() will create_parents() */ 2365219089Spjd (void) strlcat(tofs, &drrb->drr_toname[i], 2366185029Spjd ZFS_MAXNAMELEN); 2367185029Spjd *strchr(tofs, '@') = '\0'; 2368185029Spjd } 2369219089Spjd 2370228103Smm if (recursive && !flags->dryrun && !flags->nomount) { 2371219089Spjd VERIFY(0 == nvlist_alloc(&renamed, 2372219089Spjd NV_UNIQUE_NAME, 0)); 2373219089Spjd } 2374219089Spjd 2375219089Spjd softerr = recv_incremental_replication(hdl, tofs, flags, 2376219089Spjd stream_nv, stream_avl, renamed); 2377219089Spjd 2378219089Spjd /* Unmount renamed filesystems before receiving. */ 2379219089Spjd while ((pair = nvlist_next_nvpair(renamed, 2380219089Spjd pair)) != NULL) { 2381219089Spjd zfs_handle_t *zhp; 2382219089Spjd prop_changelist_t *clp = NULL; 2383219089Spjd 2384219089Spjd zhp = zfs_open(hdl, nvpair_name(pair), 2385219089Spjd ZFS_TYPE_FILESYSTEM); 2386219089Spjd if (zhp != NULL) { 2387219089Spjd clp = changelist_gather(zhp, 2388219089Spjd ZFS_PROP_MOUNTPOINT, 0, 0); 2389219089Spjd zfs_close(zhp); 2390219089Spjd if (clp != NULL) { 2391219089Spjd softerr |= 2392219089Spjd changelist_prefix(clp); 2393219089Spjd changelist_free(clp); 2394219089Spjd } 2395219089Spjd } 2396219089Spjd } 2397219089Spjd 2398219089Spjd nvlist_free(renamed); 2399185029Spjd } 2400185029Spjd } 2401185029Spjd 2402219089Spjd /* 2403219089Spjd * Get the fs specified by the first path in the stream (the top level 2404219089Spjd * specified by 'zfs send') and pass it to each invocation of 2405219089Spjd * zfs_receive_one(). 2406219089Spjd */ 2407219089Spjd (void) strlcpy(sendfs, drr->drr_u.drr_begin.drr_toname, 2408219089Spjd ZFS_MAXNAMELEN); 2409219089Spjd if ((cp = strchr(sendfs, '@')) != NULL) 2410219089Spjd *cp = '\0'; 2411185029Spjd 2412185029Spjd /* Finally, receive each contained stream */ 2413185029Spjd do { 2414185029Spjd /* 2415185029Spjd * we should figure out if it has a recoverable 2416185029Spjd * error, in which case do a recv_skip() and drive on. 2417185029Spjd * Note, if we fail due to already having this guid, 2418185029Spjd * zfs_receive_one() will take care of it (ie, 2419185029Spjd * recv_skip() and return 0). 2420185029Spjd */ 2421185029Spjd error = zfs_receive_impl(hdl, destname, flags, fd, 2422219089Spjd sendfs, stream_nv, stream_avl, top_zfs, cleanup_fd, 2423219089Spjd action_handlep); 2424185029Spjd if (error == ENODATA) { 2425185029Spjd error = 0; 2426185029Spjd break; 2427185029Spjd } 2428185029Spjd anyerr |= error; 2429185029Spjd } while (error == 0); 2430185029Spjd 2431185029Spjd if (drr->drr_payloadlen != 0 && fromsnap != NULL) { 2432185029Spjd /* 2433185029Spjd * Now that we have the fs's they sent us, try the 2434185029Spjd * renames again. 2435185029Spjd */ 2436185029Spjd softerr = recv_incremental_replication(hdl, tofs, flags, 2437219089Spjd stream_nv, stream_avl, NULL); 2438185029Spjd } 2439185029Spjd 2440185029Spjdout: 2441185029Spjd fsavl_destroy(stream_avl); 2442185029Spjd if (stream_nv) 2443185029Spjd nvlist_free(stream_nv); 2444185029Spjd if (softerr) 2445185029Spjd error = -2; 2446185029Spjd if (anyerr) 2447185029Spjd error = -1; 2448185029Spjd return (error); 2449185029Spjd} 2450185029Spjd 2451219089Spjdstatic void 2452219089Spjdtrunc_prop_errs(int truncated) 2453219089Spjd{ 2454219089Spjd ASSERT(truncated != 0); 2455219089Spjd 2456219089Spjd if (truncated == 1) 2457219089Spjd (void) fprintf(stderr, dgettext(TEXT_DOMAIN, 2458219089Spjd "1 more property could not be set\n")); 2459219089Spjd else 2460219089Spjd (void) fprintf(stderr, dgettext(TEXT_DOMAIN, 2461219089Spjd "%d more properties could not be set\n"), truncated); 2462219089Spjd} 2463219089Spjd 2464185029Spjdstatic int 2465185029Spjdrecv_skip(libzfs_handle_t *hdl, int fd, boolean_t byteswap) 2466185029Spjd{ 2467185029Spjd dmu_replay_record_t *drr; 2468185029Spjd void *buf = malloc(1<<20); 2469219089Spjd char errbuf[1024]; 2470185029Spjd 2471219089Spjd (void) snprintf(errbuf, sizeof (errbuf), dgettext(TEXT_DOMAIN, 2472219089Spjd "cannot receive:")); 2473219089Spjd 2474185029Spjd /* XXX would be great to use lseek if possible... */ 2475185029Spjd drr = buf; 2476185029Spjd 2477185029Spjd while (recv_read(hdl, fd, drr, sizeof (dmu_replay_record_t), 2478185029Spjd byteswap, NULL) == 0) { 2479185029Spjd if (byteswap) 2480185029Spjd drr->drr_type = BSWAP_32(drr->drr_type); 2481185029Spjd 2482185029Spjd switch (drr->drr_type) { 2483185029Spjd case DRR_BEGIN: 2484185029Spjd /* NB: not to be used on v2 stream packages */ 2485219089Spjd if (drr->drr_payloadlen != 0) { 2486219089Spjd zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, 2487219089Spjd "invalid substream header")); 2488219089Spjd return (zfs_error(hdl, EZFS_BADSTREAM, errbuf)); 2489219089Spjd } 2490185029Spjd break; 2491185029Spjd 2492185029Spjd case DRR_END: 2493185029Spjd free(buf); 2494185029Spjd return (0); 2495185029Spjd 2496185029Spjd case DRR_OBJECT: 2497185029Spjd if (byteswap) { 2498185029Spjd drr->drr_u.drr_object.drr_bonuslen = 2499185029Spjd BSWAP_32(drr->drr_u.drr_object. 2500185029Spjd drr_bonuslen); 2501185029Spjd } 2502185029Spjd (void) recv_read(hdl, fd, buf, 2503185029Spjd P2ROUNDUP(drr->drr_u.drr_object.drr_bonuslen, 8), 2504185029Spjd B_FALSE, NULL); 2505185029Spjd break; 2506185029Spjd 2507185029Spjd case DRR_WRITE: 2508185029Spjd if (byteswap) { 2509185029Spjd drr->drr_u.drr_write.drr_length = 2510185029Spjd BSWAP_64(drr->drr_u.drr_write.drr_length); 2511185029Spjd } 2512185029Spjd (void) recv_read(hdl, fd, buf, 2513185029Spjd drr->drr_u.drr_write.drr_length, B_FALSE, NULL); 2514185029Spjd break; 2515219089Spjd case DRR_SPILL: 2516219089Spjd if (byteswap) { 2517219089Spjd drr->drr_u.drr_write.drr_length = 2518219089Spjd BSWAP_64(drr->drr_u.drr_spill.drr_length); 2519219089Spjd } 2520219089Spjd (void) recv_read(hdl, fd, buf, 2521219089Spjd drr->drr_u.drr_spill.drr_length, B_FALSE, NULL); 2522219089Spjd break; 2523219089Spjd case DRR_WRITE_BYREF: 2524185029Spjd case DRR_FREEOBJECTS: 2525185029Spjd case DRR_FREE: 2526185029Spjd break; 2527185029Spjd 2528185029Spjd default: 2529219089Spjd zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, 2530219089Spjd "invalid record type")); 2531219089Spjd return (zfs_error(hdl, EZFS_BADSTREAM, errbuf)); 2532185029Spjd } 2533185029Spjd } 2534185029Spjd 2535185029Spjd free(buf); 2536185029Spjd return (-1); 2537185029Spjd} 2538185029Spjd 2539185029Spjd/* 2540185029Spjd * Restores a backup of tosnap from the file descriptor specified by infd. 2541185029Spjd */ 2542185029Spjdstatic int 2543185029Spjdzfs_receive_one(libzfs_handle_t *hdl, int infd, const char *tosnap, 2544228103Smm recvflags_t *flags, dmu_replay_record_t *drr, 2545219089Spjd dmu_replay_record_t *drr_noswap, const char *sendfs, 2546219089Spjd nvlist_t *stream_nv, avl_tree_t *stream_avl, char **top_zfs, int cleanup_fd, 2547219089Spjd uint64_t *action_handlep) 2548185029Spjd{ 2549185029Spjd zfs_cmd_t zc = { 0 }; 2550185029Spjd time_t begin_time; 2551219089Spjd int ioctl_err, ioctl_errno, err; 2552185029Spjd char *cp; 2553185029Spjd struct drr_begin *drrb = &drr->drr_u.drr_begin; 2554185029Spjd char errbuf[1024]; 2555219089Spjd char prop_errbuf[1024]; 2556219089Spjd const char *chopprefix; 2557185029Spjd boolean_t newfs = B_FALSE; 2558185029Spjd boolean_t stream_wantsnewfs; 2559185029Spjd uint64_t parent_snapguid = 0; 2560185029Spjd prop_changelist_t *clp = NULL; 2561185029Spjd nvlist_t *snapprops_nvlist = NULL; 2562219089Spjd zprop_errflags_t prop_errflags; 2563219089Spjd boolean_t recursive; 2564185029Spjd 2565185029Spjd begin_time = time(NULL); 2566185029Spjd 2567185029Spjd (void) snprintf(errbuf, sizeof (errbuf), dgettext(TEXT_DOMAIN, 2568185029Spjd "cannot receive")); 2569185029Spjd 2570219089Spjd recursive = (nvlist_lookup_boolean(stream_nv, "not_recursive") == 2571219089Spjd ENOENT); 2572219089Spjd 2573185029Spjd if (stream_avl != NULL) { 2574185029Spjd char *snapname; 2575185029Spjd nvlist_t *fs = fsavl_find(stream_avl, drrb->drr_toguid, 2576185029Spjd &snapname); 2577185029Spjd nvlist_t *props; 2578185029Spjd int ret; 2579185029Spjd 2580185029Spjd (void) nvlist_lookup_uint64(fs, "parentfromsnap", 2581185029Spjd &parent_snapguid); 2582185029Spjd err = nvlist_lookup_nvlist(fs, "props", &props); 2583185029Spjd if (err) 2584185029Spjd VERIFY(0 == nvlist_alloc(&props, NV_UNIQUE_NAME, 0)); 2585185029Spjd 2586228103Smm if (flags->canmountoff) { 2587185029Spjd VERIFY(0 == nvlist_add_uint64(props, 2588185029Spjd zfs_prop_to_name(ZFS_PROP_CANMOUNT), 0)); 2589185029Spjd } 2590185029Spjd ret = zcmd_write_src_nvlist(hdl, &zc, props); 2591185029Spjd if (err) 2592185029Spjd nvlist_free(props); 2593185029Spjd 2594185029Spjd if (0 == nvlist_lookup_nvlist(fs, "snapprops", &props)) { 2595185029Spjd VERIFY(0 == nvlist_lookup_nvlist(props, 2596185029Spjd snapname, &snapprops_nvlist)); 2597185029Spjd } 2598185029Spjd 2599185029Spjd if (ret != 0) 2600185029Spjd return (-1); 2601185029Spjd } 2602185029Spjd 2603219089Spjd cp = NULL; 2604219089Spjd 2605185029Spjd /* 2606185029Spjd * Determine how much of the snapshot name stored in the stream 2607185029Spjd * we are going to tack on to the name they specified on the 2608185029Spjd * command line, and how much we are going to chop off. 2609185029Spjd * 2610185029Spjd * If they specified a snapshot, chop the entire name stored in 2611185029Spjd * the stream. 2612185029Spjd */ 2613228103Smm if (flags->istail) { 2614185029Spjd /* 2615219089Spjd * A filesystem was specified with -e. We want to tack on only 2616219089Spjd * the tail of the sent snapshot path. 2617185029Spjd */ 2618185029Spjd if (strchr(tosnap, '@')) { 2619185029Spjd zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, "invalid " 2620219089Spjd "argument - snapshot not allowed with -e")); 2621219089Spjd return (zfs_error(hdl, EZFS_INVALIDNAME, errbuf)); 2622219089Spjd } 2623219089Spjd 2624219089Spjd chopprefix = strrchr(sendfs, '/'); 2625219089Spjd 2626219089Spjd if (chopprefix == NULL) { 2627219089Spjd /* 2628219089Spjd * The tail is the poolname, so we need to 2629219089Spjd * prepend a path separator. 2630219089Spjd */ 2631219089Spjd int len = strlen(drrb->drr_toname); 2632219089Spjd cp = malloc(len + 2); 2633219089Spjd cp[0] = '/'; 2634219089Spjd (void) strcpy(&cp[1], drrb->drr_toname); 2635219089Spjd chopprefix = cp; 2636219089Spjd } else { 2637219089Spjd chopprefix = drrb->drr_toname + (chopprefix - sendfs); 2638219089Spjd } 2639228103Smm } else if (flags->isprefix) { 2640219089Spjd /* 2641219089Spjd * A filesystem was specified with -d. We want to tack on 2642219089Spjd * everything but the first element of the sent snapshot path 2643219089Spjd * (all but the pool name). 2644219089Spjd */ 2645219089Spjd if (strchr(tosnap, '@')) { 2646219089Spjd zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, "invalid " 2647185029Spjd "argument - snapshot not allowed with -d")); 2648185029Spjd return (zfs_error(hdl, EZFS_INVALIDNAME, errbuf)); 2649185029Spjd } 2650219089Spjd 2651219089Spjd chopprefix = strchr(drrb->drr_toname, '/'); 2652219089Spjd if (chopprefix == NULL) 2653219089Spjd chopprefix = strchr(drrb->drr_toname, '@'); 2654185029Spjd } else if (strchr(tosnap, '@') == NULL) { 2655185029Spjd /* 2656219089Spjd * If a filesystem was specified without -d or -e, we want to 2657219089Spjd * tack on everything after the fs specified by 'zfs send'. 2658185029Spjd */ 2659219089Spjd chopprefix = drrb->drr_toname + strlen(sendfs); 2660219089Spjd } else { 2661219089Spjd /* A snapshot was specified as an exact path (no -d or -e). */ 2662219089Spjd if (recursive) { 2663219089Spjd zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, 2664219089Spjd "cannot specify snapshot name for multi-snapshot " 2665219089Spjd "stream")); 2666219089Spjd return (zfs_error(hdl, EZFS_BADSTREAM, errbuf)); 2667219089Spjd } 2668219089Spjd chopprefix = drrb->drr_toname + strlen(drrb->drr_toname); 2669185029Spjd } 2670185029Spjd 2671219089Spjd ASSERT(strstr(drrb->drr_toname, sendfs) == drrb->drr_toname); 2672219089Spjd ASSERT(chopprefix > drrb->drr_toname); 2673219089Spjd ASSERT(chopprefix <= drrb->drr_toname + strlen(drrb->drr_toname)); 2674219089Spjd ASSERT(chopprefix[0] == '/' || chopprefix[0] == '@' || 2675219089Spjd chopprefix[0] == '\0'); 2676219089Spjd 2677185029Spjd /* 2678185029Spjd * Determine name of destination snapshot, store in zc_value. 2679185029Spjd */ 2680185029Spjd (void) strcpy(zc.zc_value, tosnap); 2681219089Spjd (void) strncat(zc.zc_value, chopprefix, sizeof (zc.zc_value)); 2682248571Smm#ifdef __FreeBSD__ 2683248571Smm if (zfs_ioctl_version == ZFS_IOCVER_UNDEF) 2684248571Smm zfs_ioctl_version = get_zfs_ioctl_version(); 2685248571Smm /* 2686248571Smm * For forward compatibility hide tosnap in zc_value 2687248571Smm */ 2688248571Smm if (zfs_ioctl_version < ZFS_IOCVER_LZC) 2689248571Smm (void) strcpy(zc.zc_value + strlen(zc.zc_value) + 1, tosnap); 2690248571Smm#endif 2691219089Spjd free(cp); 2692185029Spjd if (!zfs_name_valid(zc.zc_value, ZFS_TYPE_SNAPSHOT)) { 2693185029Spjd zcmd_free_nvlists(&zc); 2694185029Spjd return (zfs_error(hdl, EZFS_INVALIDNAME, errbuf)); 2695185029Spjd } 2696185029Spjd 2697185029Spjd /* 2698185029Spjd * Determine the name of the origin snapshot, store in zc_string. 2699185029Spjd */ 2700185029Spjd if (drrb->drr_flags & DRR_FLAG_CLONE) { 2701228103Smm if (guid_to_name(hdl, zc.zc_value, 2702185029Spjd drrb->drr_fromguid, zc.zc_string) != 0) { 2703185029Spjd zcmd_free_nvlists(&zc); 2704185029Spjd zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, 2705185029Spjd "local origin for clone %s does not exist"), 2706185029Spjd zc.zc_value); 2707185029Spjd return (zfs_error(hdl, EZFS_NOENT, errbuf)); 2708185029Spjd } 2709228103Smm if (flags->verbose) 2710185029Spjd (void) printf("found clone origin %s\n", zc.zc_string); 2711185029Spjd } 2712185029Spjd 2713185029Spjd stream_wantsnewfs = (drrb->drr_fromguid == 0 || 2714185029Spjd (drrb->drr_flags & DRR_FLAG_CLONE)); 2715185029Spjd 2716185029Spjd if (stream_wantsnewfs) { 2717185029Spjd /* 2718185029Spjd * if the parent fs does not exist, look for it based on 2719185029Spjd * the parent snap GUID 2720185029Spjd */ 2721185029Spjd (void) snprintf(errbuf, sizeof (errbuf), dgettext(TEXT_DOMAIN, 2722185029Spjd "cannot receive new filesystem stream")); 2723185029Spjd 2724185029Spjd (void) strcpy(zc.zc_name, zc.zc_value); 2725185029Spjd cp = strrchr(zc.zc_name, '/'); 2726185029Spjd if (cp) 2727185029Spjd *cp = '\0'; 2728185029Spjd if (cp && 2729185029Spjd !zfs_dataset_exists(hdl, zc.zc_name, ZFS_TYPE_DATASET)) { 2730185029Spjd char suffix[ZFS_MAXNAMELEN]; 2731185029Spjd (void) strcpy(suffix, strrchr(zc.zc_value, '/')); 2732228103Smm if (guid_to_name(hdl, zc.zc_name, parent_snapguid, 2733185029Spjd zc.zc_value) == 0) { 2734185029Spjd *strchr(zc.zc_value, '@') = '\0'; 2735185029Spjd (void) strcat(zc.zc_value, suffix); 2736185029Spjd } 2737185029Spjd } 2738185029Spjd } else { 2739185029Spjd /* 2740185029Spjd * if the fs does not exist, look for it based on the 2741185029Spjd * fromsnap GUID 2742185029Spjd */ 2743185029Spjd (void) snprintf(errbuf, sizeof (errbuf), dgettext(TEXT_DOMAIN, 2744185029Spjd "cannot receive incremental stream")); 2745185029Spjd 2746185029Spjd (void) strcpy(zc.zc_name, zc.zc_value); 2747185029Spjd *strchr(zc.zc_name, '@') = '\0'; 2748185029Spjd 2749219089Spjd /* 2750219089Spjd * If the exact receive path was specified and this is the 2751219089Spjd * topmost path in the stream, then if the fs does not exist we 2752219089Spjd * should look no further. 2753219089Spjd */ 2754228103Smm if ((flags->isprefix || (*(chopprefix = drrb->drr_toname + 2755219089Spjd strlen(sendfs)) != '\0' && *chopprefix != '@')) && 2756219089Spjd !zfs_dataset_exists(hdl, zc.zc_name, ZFS_TYPE_DATASET)) { 2757185029Spjd char snap[ZFS_MAXNAMELEN]; 2758185029Spjd (void) strcpy(snap, strchr(zc.zc_value, '@')); 2759228103Smm if (guid_to_name(hdl, zc.zc_name, drrb->drr_fromguid, 2760185029Spjd zc.zc_value) == 0) { 2761185029Spjd *strchr(zc.zc_value, '@') = '\0'; 2762185029Spjd (void) strcat(zc.zc_value, snap); 2763185029Spjd } 2764185029Spjd } 2765185029Spjd } 2766185029Spjd 2767185029Spjd (void) strcpy(zc.zc_name, zc.zc_value); 2768185029Spjd *strchr(zc.zc_name, '@') = '\0'; 2769185029Spjd 2770185029Spjd if (zfs_dataset_exists(hdl, zc.zc_name, ZFS_TYPE_DATASET)) { 2771185029Spjd zfs_handle_t *zhp; 2772219089Spjd 2773185029Spjd /* 2774185029Spjd * Destination fs exists. Therefore this should either 2775185029Spjd * be an incremental, or the stream specifies a new fs 2776185029Spjd * (full stream or clone) and they want us to blow it 2777185029Spjd * away (and have therefore specified -F and removed any 2778185029Spjd * snapshots). 2779185029Spjd */ 2780185029Spjd if (stream_wantsnewfs) { 2781228103Smm if (!flags->force) { 2782185029Spjd zcmd_free_nvlists(&zc); 2783185029Spjd zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, 2784185029Spjd "destination '%s' exists\n" 2785185029Spjd "must specify -F to overwrite it"), 2786185029Spjd zc.zc_name); 2787185029Spjd return (zfs_error(hdl, EZFS_EXISTS, errbuf)); 2788185029Spjd } 2789185029Spjd if (ioctl(hdl->libzfs_fd, ZFS_IOC_SNAPSHOT_LIST_NEXT, 2790185029Spjd &zc) == 0) { 2791185029Spjd zcmd_free_nvlists(&zc); 2792185029Spjd zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, 2793185029Spjd "destination has snapshots (eg. %s)\n" 2794185029Spjd "must destroy them to overwrite it"), 2795185029Spjd zc.zc_name); 2796185029Spjd return (zfs_error(hdl, EZFS_EXISTS, errbuf)); 2797185029Spjd } 2798185029Spjd } 2799185029Spjd 2800185029Spjd if ((zhp = zfs_open(hdl, zc.zc_name, 2801185029Spjd ZFS_TYPE_FILESYSTEM | ZFS_TYPE_VOLUME)) == NULL) { 2802185029Spjd zcmd_free_nvlists(&zc); 2803185029Spjd return (-1); 2804185029Spjd } 2805185029Spjd 2806185029Spjd if (stream_wantsnewfs && 2807185029Spjd zhp->zfs_dmustats.dds_origin[0]) { 2808185029Spjd zcmd_free_nvlists(&zc); 2809185029Spjd zfs_close(zhp); 2810185029Spjd zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, 2811185029Spjd "destination '%s' is a clone\n" 2812185029Spjd "must destroy it to overwrite it"), 2813185029Spjd zc.zc_name); 2814185029Spjd return (zfs_error(hdl, EZFS_EXISTS, errbuf)); 2815185029Spjd } 2816185029Spjd 2817228103Smm if (!flags->dryrun && zhp->zfs_type == ZFS_TYPE_FILESYSTEM && 2818185029Spjd stream_wantsnewfs) { 2819185029Spjd /* We can't do online recv in this case */ 2820185029Spjd clp = changelist_gather(zhp, ZFS_PROP_NAME, 0, 0); 2821185029Spjd if (clp == NULL) { 2822219089Spjd zfs_close(zhp); 2823185029Spjd zcmd_free_nvlists(&zc); 2824185029Spjd return (-1); 2825185029Spjd } 2826185029Spjd if (changelist_prefix(clp) != 0) { 2827185029Spjd changelist_free(clp); 2828219089Spjd zfs_close(zhp); 2829185029Spjd zcmd_free_nvlists(&zc); 2830185029Spjd return (-1); 2831185029Spjd } 2832185029Spjd } 2833185029Spjd zfs_close(zhp); 2834185029Spjd } else { 2835185029Spjd /* 2836185029Spjd * Destination filesystem does not exist. Therefore we better 2837185029Spjd * be creating a new filesystem (either from a full backup, or 2838185029Spjd * a clone). It would therefore be invalid if the user 2839185029Spjd * specified only the pool name (i.e. if the destination name 2840185029Spjd * contained no slash character). 2841185029Spjd */ 2842185029Spjd if (!stream_wantsnewfs || 2843185029Spjd (cp = strrchr(zc.zc_name, '/')) == NULL) { 2844185029Spjd zcmd_free_nvlists(&zc); 2845185029Spjd zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, 2846185029Spjd "destination '%s' does not exist"), zc.zc_name); 2847185029Spjd return (zfs_error(hdl, EZFS_NOENT, errbuf)); 2848185029Spjd } 2849185029Spjd 2850185029Spjd /* 2851185029Spjd * Trim off the final dataset component so we perform the 2852185029Spjd * recvbackup ioctl to the filesystems's parent. 2853185029Spjd */ 2854185029Spjd *cp = '\0'; 2855185029Spjd 2856228103Smm if (flags->isprefix && !flags->istail && !flags->dryrun && 2857185029Spjd create_parents(hdl, zc.zc_value, strlen(tosnap)) != 0) { 2858185029Spjd zcmd_free_nvlists(&zc); 2859185029Spjd return (zfs_error(hdl, EZFS_BADRESTORE, errbuf)); 2860185029Spjd } 2861185029Spjd 2862185029Spjd newfs = B_TRUE; 2863185029Spjd } 2864185029Spjd 2865185029Spjd zc.zc_begin_record = drr_noswap->drr_u.drr_begin; 2866185029Spjd zc.zc_cookie = infd; 2867228103Smm zc.zc_guid = flags->force; 2868228103Smm if (flags->verbose) { 2869185029Spjd (void) printf("%s %s stream of %s into %s\n", 2870228103Smm flags->dryrun ? "would receive" : "receiving", 2871185029Spjd drrb->drr_fromguid ? "incremental" : "full", 2872185029Spjd drrb->drr_toname, zc.zc_value); 2873185029Spjd (void) fflush(stdout); 2874185029Spjd } 2875185029Spjd 2876228103Smm if (flags->dryrun) { 2877185029Spjd zcmd_free_nvlists(&zc); 2878228103Smm return (recv_skip(hdl, infd, flags->byteswap)); 2879185029Spjd } 2880185029Spjd 2881219089Spjd zc.zc_nvlist_dst = (uint64_t)(uintptr_t)prop_errbuf; 2882219089Spjd zc.zc_nvlist_dst_size = sizeof (prop_errbuf); 2883219089Spjd zc.zc_cleanup_fd = cleanup_fd; 2884219089Spjd zc.zc_action_handle = *action_handlep; 2885219089Spjd 2886185029Spjd err = ioctl_err = zfs_ioctl(hdl, ZFS_IOC_RECV, &zc); 2887185029Spjd ioctl_errno = errno; 2888219089Spjd prop_errflags = (zprop_errflags_t)zc.zc_obj; 2889219089Spjd 2890219089Spjd if (err == 0) { 2891219089Spjd nvlist_t *prop_errors; 2892219089Spjd VERIFY(0 == nvlist_unpack((void *)(uintptr_t)zc.zc_nvlist_dst, 2893219089Spjd zc.zc_nvlist_dst_size, &prop_errors, 0)); 2894219089Spjd 2895219089Spjd nvpair_t *prop_err = NULL; 2896219089Spjd 2897219089Spjd while ((prop_err = nvlist_next_nvpair(prop_errors, 2898219089Spjd prop_err)) != NULL) { 2899219089Spjd char tbuf[1024]; 2900219089Spjd zfs_prop_t prop; 2901219089Spjd int intval; 2902219089Spjd 2903219089Spjd prop = zfs_name_to_prop(nvpair_name(prop_err)); 2904219089Spjd (void) nvpair_value_int32(prop_err, &intval); 2905219089Spjd if (strcmp(nvpair_name(prop_err), 2906219089Spjd ZPROP_N_MORE_ERRORS) == 0) { 2907219089Spjd trunc_prop_errs(intval); 2908219089Spjd break; 2909219089Spjd } else { 2910219089Spjd (void) snprintf(tbuf, sizeof (tbuf), 2911219089Spjd dgettext(TEXT_DOMAIN, 2912219089Spjd "cannot receive %s property on %s"), 2913219089Spjd nvpair_name(prop_err), zc.zc_name); 2914219089Spjd zfs_setprop_error(hdl, prop, intval, tbuf); 2915219089Spjd } 2916219089Spjd } 2917219089Spjd nvlist_free(prop_errors); 2918219089Spjd } 2919219089Spjd 2920219089Spjd zc.zc_nvlist_dst = 0; 2921219089Spjd zc.zc_nvlist_dst_size = 0; 2922185029Spjd zcmd_free_nvlists(&zc); 2923185029Spjd 2924185029Spjd if (err == 0 && snapprops_nvlist) { 2925185029Spjd zfs_cmd_t zc2 = { 0 }; 2926185029Spjd 2927185029Spjd (void) strcpy(zc2.zc_name, zc.zc_value); 2928219089Spjd zc2.zc_cookie = B_TRUE; /* received */ 2929185029Spjd if (zcmd_write_src_nvlist(hdl, &zc2, snapprops_nvlist) == 0) { 2930185029Spjd (void) zfs_ioctl(hdl, ZFS_IOC_SET_PROP, &zc2); 2931185029Spjd zcmd_free_nvlists(&zc2); 2932185029Spjd } 2933185029Spjd } 2934185029Spjd 2935219089Spjd if (err && (ioctl_errno == ENOENT || ioctl_errno == EEXIST)) { 2936185029Spjd /* 2937185029Spjd * It may be that this snapshot already exists, 2938185029Spjd * in which case we want to consume & ignore it 2939185029Spjd * rather than failing. 2940185029Spjd */ 2941185029Spjd avl_tree_t *local_avl; 2942185029Spjd nvlist_t *local_nv, *fs; 2943219089Spjd cp = strchr(zc.zc_value, '@'); 2944185029Spjd 2945185029Spjd /* 2946185029Spjd * XXX Do this faster by just iterating over snaps in 2947185029Spjd * this fs. Also if zc_value does not exist, we will 2948185029Spjd * get a strange "does not exist" error message. 2949185029Spjd */ 2950185029Spjd *cp = '\0'; 2951219089Spjd if (gather_nvlist(hdl, zc.zc_value, NULL, NULL, B_FALSE, 2952185029Spjd &local_nv, &local_avl) == 0) { 2953185029Spjd *cp = '@'; 2954185029Spjd fs = fsavl_find(local_avl, drrb->drr_toguid, NULL); 2955185029Spjd fsavl_destroy(local_avl); 2956185029Spjd nvlist_free(local_nv); 2957185029Spjd 2958185029Spjd if (fs != NULL) { 2959228103Smm if (flags->verbose) { 2960185029Spjd (void) printf("snap %s already exists; " 2961185029Spjd "ignoring\n", zc.zc_value); 2962185029Spjd } 2963219089Spjd err = ioctl_err = recv_skip(hdl, infd, 2964228103Smm flags->byteswap); 2965185029Spjd } 2966185029Spjd } 2967185029Spjd *cp = '@'; 2968185029Spjd } 2969185029Spjd 2970185029Spjd if (ioctl_err != 0) { 2971185029Spjd switch (ioctl_errno) { 2972185029Spjd case ENODEV: 2973185029Spjd cp = strchr(zc.zc_value, '@'); 2974185029Spjd *cp = '\0'; 2975185029Spjd zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, 2976185029Spjd "most recent snapshot of %s does not\n" 2977185029Spjd "match incremental source"), zc.zc_value); 2978185029Spjd (void) zfs_error(hdl, EZFS_BADRESTORE, errbuf); 2979185029Spjd *cp = '@'; 2980185029Spjd break; 2981185029Spjd case ETXTBSY: 2982185029Spjd zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, 2983185029Spjd "destination %s has been modified\n" 2984185029Spjd "since most recent snapshot"), zc.zc_name); 2985185029Spjd (void) zfs_error(hdl, EZFS_BADRESTORE, errbuf); 2986185029Spjd break; 2987185029Spjd case EEXIST: 2988185029Spjd cp = strchr(zc.zc_value, '@'); 2989185029Spjd if (newfs) { 2990185029Spjd /* it's the containing fs that exists */ 2991185029Spjd *cp = '\0'; 2992185029Spjd } 2993185029Spjd zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, 2994185029Spjd "destination already exists")); 2995185029Spjd (void) zfs_error_fmt(hdl, EZFS_EXISTS, 2996185029Spjd dgettext(TEXT_DOMAIN, "cannot restore to %s"), 2997185029Spjd zc.zc_value); 2998185029Spjd *cp = '@'; 2999185029Spjd break; 3000185029Spjd case EINVAL: 3001185029Spjd (void) zfs_error(hdl, EZFS_BADSTREAM, errbuf); 3002185029Spjd break; 3003185029Spjd case ECKSUM: 3004185029Spjd zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, 3005185029Spjd "invalid stream (checksum mismatch)")); 3006185029Spjd (void) zfs_error(hdl, EZFS_BADSTREAM, errbuf); 3007185029Spjd break; 3008219089Spjd case ENOTSUP: 3009219089Spjd zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, 3010219089Spjd "pool must be upgraded to receive this stream.")); 3011219089Spjd (void) zfs_error(hdl, EZFS_BADVERSION, errbuf); 3012219089Spjd break; 3013219089Spjd case EDQUOT: 3014219089Spjd zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, 3015219089Spjd "destination %s space quota exceeded"), zc.zc_name); 3016228103Smm (void) zfs_error(hdl, EZFS_NOSPC, errbuf); 3017219089Spjd break; 3018185029Spjd default: 3019185029Spjd (void) zfs_standard_error(hdl, ioctl_errno, errbuf); 3020185029Spjd } 3021185029Spjd } 3022185029Spjd 3023185029Spjd /* 3024219089Spjd * Mount the target filesystem (if created). Also mount any 3025219089Spjd * children of the target filesystem if we did a replication 3026219089Spjd * receive (indicated by stream_avl being non-NULL). 3027185029Spjd */ 3028185029Spjd cp = strchr(zc.zc_value, '@'); 3029185029Spjd if (cp && (ioctl_err == 0 || !newfs)) { 3030185029Spjd zfs_handle_t *h; 3031185029Spjd 3032185029Spjd *cp = '\0'; 3033185029Spjd h = zfs_open(hdl, zc.zc_value, 3034185029Spjd ZFS_TYPE_FILESYSTEM | ZFS_TYPE_VOLUME); 3035185029Spjd if (h != NULL) { 3036185029Spjd if (h->zfs_type == ZFS_TYPE_VOLUME) { 3037185029Spjd *cp = '@'; 3038219089Spjd } else if (newfs || stream_avl) { 3039185029Spjd /* 3040185029Spjd * Track the first/top of hierarchy fs, 3041185029Spjd * for mounting and sharing later. 3042185029Spjd */ 3043185029Spjd if (top_zfs && *top_zfs == NULL) 3044185029Spjd *top_zfs = zfs_strdup(hdl, zc.zc_value); 3045185029Spjd } 3046185029Spjd zfs_close(h); 3047185029Spjd } 3048185029Spjd *cp = '@'; 3049185029Spjd } 3050185029Spjd 3051185029Spjd if (clp) { 3052185029Spjd err |= changelist_postfix(clp); 3053185029Spjd changelist_free(clp); 3054185029Spjd } 3055185029Spjd 3056219089Spjd if (prop_errflags & ZPROP_ERR_NOCLEAR) { 3057219089Spjd (void) fprintf(stderr, dgettext(TEXT_DOMAIN, "Warning: " 3058219089Spjd "failed to clear unreceived properties on %s"), 3059219089Spjd zc.zc_name); 3060219089Spjd (void) fprintf(stderr, "\n"); 3061219089Spjd } 3062219089Spjd if (prop_errflags & ZPROP_ERR_NORESTORE) { 3063219089Spjd (void) fprintf(stderr, dgettext(TEXT_DOMAIN, "Warning: " 3064219089Spjd "failed to restore original properties on %s"), 3065219089Spjd zc.zc_name); 3066219089Spjd (void) fprintf(stderr, "\n"); 3067219089Spjd } 3068219089Spjd 3069185029Spjd if (err || ioctl_err) 3070185029Spjd return (-1); 3071185029Spjd 3072219089Spjd *action_handlep = zc.zc_action_handle; 3073219089Spjd 3074228103Smm if (flags->verbose) { 3075185029Spjd char buf1[64]; 3076185029Spjd char buf2[64]; 3077185029Spjd uint64_t bytes = zc.zc_cookie; 3078185029Spjd time_t delta = time(NULL) - begin_time; 3079185029Spjd if (delta == 0) 3080185029Spjd delta = 1; 3081185029Spjd zfs_nicenum(bytes, buf1, sizeof (buf1)); 3082185029Spjd zfs_nicenum(bytes/delta, buf2, sizeof (buf1)); 3083185029Spjd 3084185029Spjd (void) printf("received %sB stream in %lu seconds (%sB/sec)\n", 3085185029Spjd buf1, delta, buf2); 3086185029Spjd } 3087185029Spjd 3088185029Spjd return (0); 3089185029Spjd} 3090185029Spjd 3091185029Spjdstatic int 3092228103Smmzfs_receive_impl(libzfs_handle_t *hdl, const char *tosnap, recvflags_t *flags, 3093219089Spjd int infd, const char *sendfs, nvlist_t *stream_nv, avl_tree_t *stream_avl, 3094219089Spjd char **top_zfs, int cleanup_fd, uint64_t *action_handlep) 3095185029Spjd{ 3096185029Spjd int err; 3097185029Spjd dmu_replay_record_t drr, drr_noswap; 3098185029Spjd struct drr_begin *drrb = &drr.drr_u.drr_begin; 3099185029Spjd char errbuf[1024]; 3100185029Spjd zio_cksum_t zcksum = { 0 }; 3101219089Spjd uint64_t featureflags; 3102219089Spjd int hdrtype; 3103185029Spjd 3104185029Spjd (void) snprintf(errbuf, sizeof (errbuf), dgettext(TEXT_DOMAIN, 3105185029Spjd "cannot receive")); 3106185029Spjd 3107228103Smm if (flags->isprefix && 3108185029Spjd !zfs_dataset_exists(hdl, tosnap, ZFS_TYPE_DATASET)) { 3109185029Spjd zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, "specified fs " 3110185029Spjd "(%s) does not exist"), tosnap); 3111185029Spjd return (zfs_error(hdl, EZFS_NOENT, errbuf)); 3112185029Spjd } 3113185029Spjd 3114185029Spjd /* read in the BEGIN record */ 3115185029Spjd if (0 != (err = recv_read(hdl, infd, &drr, sizeof (drr), B_FALSE, 3116185029Spjd &zcksum))) 3117185029Spjd return (err); 3118185029Spjd 3119185029Spjd if (drr.drr_type == DRR_END || drr.drr_type == BSWAP_32(DRR_END)) { 3120185029Spjd /* It's the double end record at the end of a package */ 3121185029Spjd return (ENODATA); 3122185029Spjd } 3123185029Spjd 3124185029Spjd /* the kernel needs the non-byteswapped begin record */ 3125185029Spjd drr_noswap = drr; 3126185029Spjd 3127228103Smm flags->byteswap = B_FALSE; 3128185029Spjd if (drrb->drr_magic == BSWAP_64(DMU_BACKUP_MAGIC)) { 3129185029Spjd /* 3130185029Spjd * We computed the checksum in the wrong byteorder in 3131185029Spjd * recv_read() above; do it again correctly. 3132185029Spjd */ 3133185029Spjd bzero(&zcksum, sizeof (zio_cksum_t)); 3134185029Spjd fletcher_4_incremental_byteswap(&drr, sizeof (drr), &zcksum); 3135228103Smm flags->byteswap = B_TRUE; 3136185029Spjd 3137185029Spjd drr.drr_type = BSWAP_32(drr.drr_type); 3138185029Spjd drr.drr_payloadlen = BSWAP_32(drr.drr_payloadlen); 3139185029Spjd drrb->drr_magic = BSWAP_64(drrb->drr_magic); 3140219089Spjd drrb->drr_versioninfo = BSWAP_64(drrb->drr_versioninfo); 3141185029Spjd drrb->drr_creation_time = BSWAP_64(drrb->drr_creation_time); 3142185029Spjd drrb->drr_type = BSWAP_32(drrb->drr_type); 3143185029Spjd drrb->drr_flags = BSWAP_32(drrb->drr_flags); 3144185029Spjd drrb->drr_toguid = BSWAP_64(drrb->drr_toguid); 3145185029Spjd drrb->drr_fromguid = BSWAP_64(drrb->drr_fromguid); 3146185029Spjd } 3147185029Spjd 3148185029Spjd if (drrb->drr_magic != DMU_BACKUP_MAGIC || drr.drr_type != DRR_BEGIN) { 3149185029Spjd zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, "invalid " 3150185029Spjd "stream (bad magic number)")); 3151185029Spjd return (zfs_error(hdl, EZFS_BADSTREAM, errbuf)); 3152185029Spjd } 3153185029Spjd 3154219089Spjd featureflags = DMU_GET_FEATUREFLAGS(drrb->drr_versioninfo); 3155219089Spjd hdrtype = DMU_GET_STREAM_HDRTYPE(drrb->drr_versioninfo); 3156219089Spjd 3157219089Spjd if (!DMU_STREAM_SUPPORTED(featureflags) || 3158219089Spjd (hdrtype != DMU_SUBSTREAM && hdrtype != DMU_COMPOUNDSTREAM)) { 3159219089Spjd zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, 3160219089Spjd "stream has unsupported feature, feature flags = %lx"), 3161219089Spjd featureflags); 3162219089Spjd return (zfs_error(hdl, EZFS_BADSTREAM, errbuf)); 3163219089Spjd } 3164219089Spjd 3165185029Spjd if (strchr(drrb->drr_toname, '@') == NULL) { 3166185029Spjd zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, "invalid " 3167185029Spjd "stream (bad snapshot name)")); 3168185029Spjd return (zfs_error(hdl, EZFS_BADSTREAM, errbuf)); 3169185029Spjd } 3170185029Spjd 3171219089Spjd if (DMU_GET_STREAM_HDRTYPE(drrb->drr_versioninfo) == DMU_SUBSTREAM) { 3172219089Spjd char nonpackage_sendfs[ZFS_MAXNAMELEN]; 3173219089Spjd if (sendfs == NULL) { 3174219089Spjd /* 3175219089Spjd * We were not called from zfs_receive_package(). Get 3176219089Spjd * the fs specified by 'zfs send'. 3177219089Spjd */ 3178219089Spjd char *cp; 3179219089Spjd (void) strlcpy(nonpackage_sendfs, 3180219089Spjd drr.drr_u.drr_begin.drr_toname, ZFS_MAXNAMELEN); 3181219089Spjd if ((cp = strchr(nonpackage_sendfs, '@')) != NULL) 3182219089Spjd *cp = '\0'; 3183219089Spjd sendfs = nonpackage_sendfs; 3184219089Spjd } 3185185029Spjd return (zfs_receive_one(hdl, infd, tosnap, flags, 3186219089Spjd &drr, &drr_noswap, sendfs, stream_nv, stream_avl, 3187219089Spjd top_zfs, cleanup_fd, action_handlep)); 3188219089Spjd } else { 3189219089Spjd assert(DMU_GET_STREAM_HDRTYPE(drrb->drr_versioninfo) == 3190219089Spjd DMU_COMPOUNDSTREAM); 3191185029Spjd return (zfs_receive_package(hdl, infd, tosnap, flags, 3192219089Spjd &drr, &zcksum, top_zfs, cleanup_fd, action_handlep)); 3193185029Spjd } 3194185029Spjd} 3195185029Spjd 3196185029Spjd/* 3197185029Spjd * Restores a backup of tosnap from the file descriptor specified by infd. 3198185029Spjd * Return 0 on total success, -2 if some things couldn't be 3199185029Spjd * destroyed/renamed/promoted, -1 if some things couldn't be received. 3200185029Spjd * (-1 will override -2). 3201185029Spjd */ 3202185029Spjdint 3203228103Smmzfs_receive(libzfs_handle_t *hdl, const char *tosnap, recvflags_t *flags, 3204185029Spjd int infd, avl_tree_t *stream_avl) 3205185029Spjd{ 3206185029Spjd char *top_zfs = NULL; 3207185029Spjd int err; 3208219089Spjd int cleanup_fd; 3209219089Spjd uint64_t action_handle = 0; 3210185029Spjd 3211219089Spjd cleanup_fd = open(ZFS_DEV, O_RDWR|O_EXCL); 3212219089Spjd VERIFY(cleanup_fd >= 0); 3213185029Spjd 3214219089Spjd err = zfs_receive_impl(hdl, tosnap, flags, infd, NULL, NULL, 3215219089Spjd stream_avl, &top_zfs, cleanup_fd, &action_handle); 3216219089Spjd 3217219089Spjd VERIFY(0 == close(cleanup_fd)); 3218219089Spjd 3219228103Smm if (err == 0 && !flags->nomount && top_zfs) { 3220185029Spjd zfs_handle_t *zhp; 3221185029Spjd prop_changelist_t *clp; 3222185029Spjd 3223185029Spjd zhp = zfs_open(hdl, top_zfs, ZFS_TYPE_FILESYSTEM); 3224185029Spjd if (zhp != NULL) { 3225185029Spjd clp = changelist_gather(zhp, ZFS_PROP_MOUNTPOINT, 3226185029Spjd CL_GATHER_MOUNT_ALWAYS, 0); 3227185029Spjd zfs_close(zhp); 3228185029Spjd if (clp != NULL) { 3229185029Spjd /* mount and share received datasets */ 3230185029Spjd err = changelist_postfix(clp); 3231185029Spjd changelist_free(clp); 3232185029Spjd } 3233185029Spjd } 3234185029Spjd if (zhp == NULL || clp == NULL || err) 3235185029Spjd err = -1; 3236185029Spjd } 3237185029Spjd if (top_zfs) 3238185029Spjd free(top_zfs); 3239185029Spjd 3240185029Spjd return (err); 3241185029Spjd} 3242