]> git.proxmox.com Git - mirror_zfs.git/blob - zfs/zcmd/zdb/zdb.c
projects / mirror_zfs.git / blob
? search:
summary | shortlog | log | commit | commitdiff | tree
blame | history | raw | HEAD
Rebase to OpenSolaris b103, in the process we are removing any code which did not...
[mirror_zfs.git] / zfs / zcmd / zdb / zdb.c
1 /*
2 * CDDL HEADER START
3 *
4 * The contents of this file are subject to the terms of the
5 * Common Development and Distribution License (the "License").
6 * You may not use this file except in compliance with the License.
7 *
8 * You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE
9 * or http://www.opensolaris.org/os/licensing.
10 * See the License for the specific language governing permissions
11 * and limitations under the License.
12 *
13 * When distributing Covered Code, include this CDDL HEADER in each
14 * file and include the License file at usr/src/OPENSOLARIS.LICENSE.
15 * If applicable, add the following below this CDDL HEADER, with the
16 * fields enclosed by brackets "[]" replaced with your own identifying
17 * information: Portions Copyright [yyyy] [name of copyright owner]
18 *
19 * CDDL HEADER END
20 */
21 /*
22 * Copyright 2008 Sun Microsystems, Inc. All rights reserved.
23 * Use is subject to license terms.
24 */
25
26 #include <stdio.h>
27 #include <stdio_ext.h>
28 #include <stdlib.h>
29 #include <ctype.h>
30 #include <sys/zfs_context.h>
31 #include <sys/spa.h>
32 #include <sys/spa_impl.h>
33 #include <sys/dmu.h>
34 #include <sys/zap.h>
35 #include <sys/fs/zfs.h>
36 #include <sys/zfs_znode.h>
37 #include <sys/vdev.h>
38 #include <sys/vdev_impl.h>
39 #include <sys/metaslab_impl.h>
40 #include <sys/dmu_objset.h>
41 #include <sys/dsl_dir.h>
42 #include <sys/dsl_dataset.h>
43 #include <sys/dsl_pool.h>
44 #include <sys/dbuf.h>
45 #include <sys/zil.h>
46 #include <sys/zil_impl.h>
47 #include <sys/stat.h>
48 #include <sys/resource.h>
49 #include <sys/dmu_traverse.h>
50 #include <sys/zio_checksum.h>
51 #include <sys/zio_compress.h>
52 #include <sys/zfs_fuid.h>
53 #include <sys/arc.h>
54 #undef ZFS_MAXNAMELEN
55 #undef verify
56 #include <libzfs.h>
57
58 const char cmdname[] = "zdb";
59 uint8_t dump_opt[256];
60
61 typedef void object_viewer_t(objset_t *, uint64_t, void *data, size_t size);
62
63 extern void dump_intent_log(zilog_t *);
64 uint64_t *zopt_object = NULL;
65 int zopt_objects = 0;
66 libzfs_handle_t *g_zfs;
67 boolean_t zdb_sig_user_data = B_TRUE;
68 int zdb_sig_cksumalg = ZIO_CHECKSUM_SHA256;
69
70 /*
71 * These libumem hooks provide a reasonable set of defaults for the allocator's
72 * debugging facilities.
73 */
74 const char *
75 _umem_debug_init()
76 {
77 return ("default,verbose"); /* $UMEM_DEBUG setting */
78 }
79
80 const char *
81 _umem_logging_init(void)
82 {
83 return ("fail,contents"); /* $UMEM_LOGGING setting */
84 }
85
86 static void
87 usage(void)
88 {
89 (void) fprintf(stderr,
90 "Usage: %s [-udibcsv] [-U cachefile_path] "
91 "[-S user:cksumalg] "
92 "dataset [object...]\n"
93 " %s -C [pool]\n"
94 " %s -l dev\n"
95 " %s -R pool:vdev:offset:size:flags\n"
96 " %s [-p path_to_vdev_dir]\n"
97 " %s -e pool | GUID | devid ...\n",
98 cmdname, cmdname, cmdname, cmdname, cmdname, cmdname);
99
100 (void) fprintf(stderr, " -u uberblock\n");
101 (void) fprintf(stderr, " -d datasets\n");
102 (void) fprintf(stderr, " -C cached pool configuration\n");
103 (void) fprintf(stderr, " -i intent logs\n");
104 (void) fprintf(stderr, " -b block statistics\n");
105 (void) fprintf(stderr, " -c checksum all data blocks\n");
106 (void) fprintf(stderr, " -s report stats on zdb's I/O\n");
107 (void) fprintf(stderr, " -S <user|all>:<cksum_alg|all> -- "
108 "dump blkptr signatures\n");
109 (void) fprintf(stderr, " -v verbose (applies to all others)\n");
110 (void) fprintf(stderr, " -l dump label contents\n");
111 (void) fprintf(stderr, " -U cachefile_path -- use alternate "
112 "cachefile\n");
113 (void) fprintf(stderr, " -R read and display block from a "
114 "device\n");
115 (void) fprintf(stderr, " -e Pool is exported/destroyed/"
116 "has altroot\n");
117 (void) fprintf(stderr, " -p <Path to vdev dir> (use with -e)\n");
118 (void) fprintf(stderr, "Specify an option more than once (e.g. -bb) "
119 "to make only that option verbose\n");
120 (void) fprintf(stderr, "Default is to dump everything non-verbosely\n");
121 exit(1);
122 }
123
124 static void
125 fatal(const char *fmt, ...)
126 {
127 va_list ap;
128
129 va_start(ap, fmt);
130 (void) fprintf(stderr, "%s: ", cmdname);
131 (void) vfprintf(stderr, fmt, ap);
132 va_end(ap);
133 (void) fprintf(stderr, "\n");
134
135 abort();
136 }
137
138 static void
139 dump_nvlist(nvlist_t *list, int indent)
140 {
141 nvpair_t *elem = NULL;
142
143 while ((elem = nvlist_next_nvpair(list, elem)) != NULL) {
144 switch (nvpair_type(elem)) {
145 case DATA_TYPE_STRING:
146 {
147 char *value;
148
149 VERIFY(nvpair_value_string(elem, &value) == 0);
150 (void) printf("%*s%s='%s'\n", indent, "",
151 nvpair_name(elem), value);
152 }
153 break;
154
155 case DATA_TYPE_UINT64:
156 {
157 uint64_t value;
158
159 VERIFY(nvpair_value_uint64(elem, &value) == 0);
160 (void) printf("%*s%s=%llu\n", indent, "",
161 nvpair_name(elem), (u_longlong_t)value);
162 }
163 break;
164
165 case DATA_TYPE_NVLIST:
166 {
167 nvlist_t *value;
168
169 VERIFY(nvpair_value_nvlist(elem, &value) == 0);
170 (void) printf("%*s%s\n", indent, "",
171 nvpair_name(elem));
172 dump_nvlist(value, indent + 4);
173 }
174 break;
175
176 case DATA_TYPE_NVLIST_ARRAY:
177 {
178 nvlist_t **value;
179 uint_t c, count;
180
181 VERIFY(nvpair_value_nvlist_array(elem, &value,
182 &count) == 0);
183
184 for (c = 0; c < count; c++) {
185 (void) printf("%*s%s[%u]\n", indent, "",
186 nvpair_name(elem), c);
187 dump_nvlist(value[c], indent + 8);
188 }
189 }
190 break;
191
192 default:
193
194 (void) printf("bad config type %d for %s\n",
195 nvpair_type(elem), nvpair_name(elem));
196 }
197 }
198 }
199
200 /* ARGSUSED */
201 static void
202 dump_packed_nvlist(objset_t *os, uint64_t object, void *data, size_t size)
203 {
204 nvlist_t *nv;
205 size_t nvsize = *(uint64_t *)data;
206 char *packed = umem_alloc(nvsize, UMEM_NOFAIL);
207
208 VERIFY(0 == dmu_read(os, object, 0, nvsize, packed));
209
210 VERIFY(nvlist_unpack(packed, nvsize, &nv, 0) == 0);
211
212 umem_free(packed, nvsize);
213
214 dump_nvlist(nv, 8);
215
216 nvlist_free(nv);
217 }
218
219 const char dump_zap_stars[] = "****************************************";
220 const int dump_zap_width = sizeof (dump_zap_stars) - 1;
221
222 static void
223 dump_zap_histogram(uint64_t histo[ZAP_HISTOGRAM_SIZE])
224 {
225 int i;
226 int minidx = ZAP_HISTOGRAM_SIZE - 1;
227 int maxidx = 0;
228 uint64_t max = 0;
229
230 for (i = 0; i < ZAP_HISTOGRAM_SIZE; i++) {
231 if (histo[i] > max)
232 max = histo[i];
233 if (histo[i] > 0 && i > maxidx)
234 maxidx = i;
235 if (histo[i] > 0 && i < minidx)
236 minidx = i;
237 }
238
239 if (max < dump_zap_width)
240 max = dump_zap_width;
241
242 for (i = minidx; i <= maxidx; i++)
243 (void) printf("\t\t\t%u: %6llu %s\n", i, (u_longlong_t)histo[i],
244 &dump_zap_stars[(max - histo[i]) * dump_zap_width / max]);
245 }
246
247 static void
248 dump_zap_stats(objset_t *os, uint64_t object)
249 {
250 int error;
251 zap_stats_t zs;
252
253 error = zap_get_stats(os, object, &zs);
254 if (error)
255 return;
256
257 if (zs.zs_ptrtbl_len == 0) {
258 ASSERT(zs.zs_num_blocks == 1);
259 (void) printf("\tmicrozap: %llu bytes, %llu entries\n",
260 (u_longlong_t)zs.zs_blocksize,
261 (u_longlong_t)zs.zs_num_entries);
262 return;
263 }
264
265 (void) printf("\tFat ZAP stats:\n");
266
267 (void) printf("\t\tPointer table:\n");
268 (void) printf("\t\t\t%llu elements\n",
269 (u_longlong_t)zs.zs_ptrtbl_len);
270 (void) printf("\t\t\tzt_blk: %llu\n",
271 (u_longlong_t)zs.zs_ptrtbl_zt_blk);
272 (void) printf("\t\t\tzt_numblks: %llu\n",
273 (u_longlong_t)zs.zs_ptrtbl_zt_numblks);
274 (void) printf("\t\t\tzt_shift: %llu\n",
275 (u_longlong_t)zs.zs_ptrtbl_zt_shift);
276 (void) printf("\t\t\tzt_blks_copied: %llu\n",
277 (u_longlong_t)zs.zs_ptrtbl_blks_copied);
278 (void) printf("\t\t\tzt_nextblk: %llu\n",
279 (u_longlong_t)zs.zs_ptrtbl_nextblk);
280
281 (void) printf("\t\tZAP entries: %llu\n",
282 (u_longlong_t)zs.zs_num_entries);
283 (void) printf("\t\tLeaf blocks: %llu\n",
284 (u_longlong_t)zs.zs_num_leafs);
285 (void) printf("\t\tTotal blocks: %llu\n",
286 (u_longlong_t)zs.zs_num_blocks);
287 (void) printf("\t\tzap_block_type: 0x%llx\n",
288 (u_longlong_t)zs.zs_block_type);
289 (void) printf("\t\tzap_magic: 0x%llx\n",
290 (u_longlong_t)zs.zs_magic);
291 (void) printf("\t\tzap_salt: 0x%llx\n",
292 (u_longlong_t)zs.zs_salt);
293
294 (void) printf("\t\tLeafs with 2^n pointers:\n");
295 dump_zap_histogram(zs.zs_leafs_with_2n_pointers);
296
297 (void) printf("\t\tBlocks with n*5 entries:\n");
298 dump_zap_histogram(zs.zs_blocks_with_n5_entries);
299
300 (void) printf("\t\tBlocks n/10 full:\n");
301 dump_zap_histogram(zs.zs_blocks_n_tenths_full);
302
303 (void) printf("\t\tEntries with n chunks:\n");
304 dump_zap_histogram(zs.zs_entries_using_n_chunks);
305
306 (void) printf("\t\tBuckets with n entries:\n");
307 dump_zap_histogram(zs.zs_buckets_with_n_entries);
308 }
309
310 /*ARGSUSED*/
311 static void
312 dump_none(objset_t *os, uint64_t object, void *data, size_t size)
313 {
314 }
315
316 /*ARGSUSED*/
317 void
318 dump_uint8(objset_t *os, uint64_t object, void *data, size_t size)
319 {
320 }
321
322 /*ARGSUSED*/
323 static void
324 dump_uint64(objset_t *os, uint64_t object, void *data, size_t size)
325 {
326 }
327
328 /*ARGSUSED*/
329 static void
330 dump_zap(objset_t *os, uint64_t object, void *data, size_t size)
331 {
332 zap_cursor_t zc;
333 zap_attribute_t attr;
334 void *prop;
335 int i;
336
337 dump_zap_stats(os, object);
338 (void) printf("\n");
339
340 for (zap_cursor_init(&zc, os, object);
341 zap_cursor_retrieve(&zc, &attr) == 0;
342 zap_cursor_advance(&zc)) {
343 (void) printf("\t\t%s = ", attr.za_name);
344 if (attr.za_num_integers == 0) {
345 (void) printf("\n");
346 continue;
347 }
348 prop = umem_zalloc(attr.za_num_integers *
349 attr.za_integer_length, UMEM_NOFAIL);
350 (void) zap_lookup(os, object, attr.za_name,
351 attr.za_integer_length, attr.za_num_integers, prop);
352 if (attr.za_integer_length == 1) {
353 (void) printf("%s", (char *)prop);
354 } else {
355 for (i = 0; i < attr.za_num_integers; i++) {
356 switch (attr.za_integer_length) {
357 case 2:
358 (void) printf("%u ",
359 ((uint16_t *)prop)[i]);
360 break;
361 case 4:
362 (void) printf("%u ",
363 ((uint32_t *)prop)[i]);
364 break;
365 case 8:
366 (void) printf("%lld ",
367 (u_longlong_t)((int64_t *)prop)[i]);
368 break;
369 }
370 }
371 }
372 (void) printf("\n");
373 umem_free(prop, attr.za_num_integers * attr.za_integer_length);
374 }
375 zap_cursor_fini(&zc);
376 }
377
378 /*ARGSUSED*/
379 static void
380 dump_zpldir(objset_t *os, uint64_t object, void *data, size_t size)
381 {
382 zap_cursor_t zc;
383 zap_attribute_t attr;
384 const char *typenames[] = {
385 /* 0 */ "not specified",
386 /* 1 */ "FIFO",
387 /* 2 */ "Character Device",
388 /* 3 */ "3 (invalid)",
389 /* 4 */ "Directory",
390 /* 5 */ "5 (invalid)",
391 /* 6 */ "Block Device",
392 /* 7 */ "7 (invalid)",
393 /* 8 */ "Regular File",
394 /* 9 */ "9 (invalid)",
395 /* 10 */ "Symbolic Link",
396 /* 11 */ "11 (invalid)",
397 /* 12 */ "Socket",
398 /* 13 */ "Door",
399 /* 14 */ "Event Port",
400 /* 15 */ "15 (invalid)",
401 };
402
403 dump_zap_stats(os, object);
404 (void) printf("\n");
405
406 for (zap_cursor_init(&zc, os, object);
407 zap_cursor_retrieve(&zc, &attr) == 0;
408 zap_cursor_advance(&zc)) {
409 (void) printf("\t\t%s = %lld (type: %s)\n",
410 attr.za_name, ZFS_DIRENT_OBJ(attr.za_first_integer),
411 typenames[ZFS_DIRENT_TYPE(attr.za_first_integer)]);
412 }
413 zap_cursor_fini(&zc);
414 }
415
416 static void
417 dump_spacemap(objset_t *os, space_map_obj_t *smo, space_map_t *sm)
418 {
419 uint64_t alloc, offset, entry;
420 uint8_t mapshift = sm->sm_shift;
421 uint64_t mapstart = sm->sm_start;
422 char *ddata[] = { "ALLOC", "FREE", "CONDENSE", "INVALID",
423 "INVALID", "INVALID", "INVALID", "INVALID" };
424
425 if (smo->smo_object == 0)
426 return;
427
428 /*
429 * Print out the freelist entries in both encoded and decoded form.
430 */
431 alloc = 0;
432 for (offset = 0; offset < smo->smo_objsize; offset += sizeof (entry)) {
433 VERIFY(0 == dmu_read(os, smo->smo_object, offset,
434 sizeof (entry), &entry));
435 if (SM_DEBUG_DECODE(entry)) {
436 (void) printf("\t\t[%4llu] %s: txg %llu, pass %llu\n",
437 (u_longlong_t)(offset / sizeof (entry)),
438 ddata[SM_DEBUG_ACTION_DECODE(entry)],
439 (u_longlong_t)SM_DEBUG_TXG_DECODE(entry),
440 (u_longlong_t)SM_DEBUG_SYNCPASS_DECODE(entry));
441 } else {
442 (void) printf("\t\t[%4llu] %c range:"
443 " %08llx-%08llx size: %06llx\n",
444 (u_longlong_t)(offset / sizeof (entry)),
445 SM_TYPE_DECODE(entry) == SM_ALLOC ? 'A' : 'F',
446 (u_longlong_t)((SM_OFFSET_DECODE(entry) <<
447 mapshift) + mapstart),
448 (u_longlong_t)((SM_OFFSET_DECODE(entry) <<
449 mapshift) + mapstart + (SM_RUN_DECODE(entry) <<
450 mapshift)),
451 (u_longlong_t)(SM_RUN_DECODE(entry) << mapshift));
452 if (SM_TYPE_DECODE(entry) == SM_ALLOC)
453 alloc += SM_RUN_DECODE(entry) << mapshift;
454 else
455 alloc -= SM_RUN_DECODE(entry) << mapshift;
456 }
457 }
458 if (alloc != smo->smo_alloc) {
459 (void) printf("space_map_object alloc (%llu) INCONSISTENT "
460 "with space map summary (%llu)\n",
461 (u_longlong_t)smo->smo_alloc, (u_longlong_t)alloc);
462 }
463 }
464
465 static void
466 dump_metaslab(metaslab_t *msp)
467 {
468 char freebuf[5];
469 space_map_obj_t *smo = &msp->ms_smo;
470 vdev_t *vd = msp->ms_group->mg_vd;
471 spa_t *spa = vd->vdev_spa;
472
473 nicenum(msp->ms_map.sm_size - smo->smo_alloc, freebuf);
474
475 if (dump_opt['d'] <= 5) {
476 (void) printf("\t%10llx %10llu %5s\n",
477 (u_longlong_t)msp->ms_map.sm_start,
478 (u_longlong_t)smo->smo_object,
479 freebuf);
480 return;
481 }
482
483 (void) printf(
484 "\tvdev %llu offset %08llx spacemap %4llu free %5s\n",
485 (u_longlong_t)vd->vdev_id, (u_longlong_t)msp->ms_map.sm_start,
486 (u_longlong_t)smo->smo_object, freebuf);
487
488 ASSERT(msp->ms_map.sm_size == (1ULL << vd->vdev_ms_shift));
489
490 dump_spacemap(spa->spa_meta_objset, smo, &msp->ms_map);
491 }
492
493 static void
494 dump_metaslabs(spa_t *spa)
495 {
496 vdev_t *rvd = spa->spa_root_vdev;
497 vdev_t *vd;
498 int c, m;
499
500 (void) printf("\nMetaslabs:\n");
501
502 for (c = 0; c < rvd->vdev_children; c++) {
503 vd = rvd->vdev_child[c];
504
505 (void) printf("\n vdev %llu\n\n", (u_longlong_t)vd->vdev_id);
506
507 if (dump_opt['d'] <= 5) {
508 (void) printf("\t%10s %10s %5s\n",
509 "offset", "spacemap", "free");
510 (void) printf("\t%10s %10s %5s\n",
511 "------", "--------", "----");
512 }
513 for (m = 0; m < vd->vdev_ms_count; m++)
514 dump_metaslab(vd->vdev_ms[m]);
515 (void) printf("\n");
516 }
517 }
518
519 static void
520 dump_dtl(vdev_t *vd, int indent)
521 {
522 avl_tree_t *t = &vd->vdev_dtl_map.sm_root;
523 space_seg_t *ss;
524 vdev_t *pvd;
525 int c;
526
527 if (indent == 0)
528 (void) printf("\nDirty time logs:\n\n");
529
530 (void) printf("\t%*s%s\n", indent, "",
531 vd->vdev_path ? vd->vdev_path :
532 vd->vdev_parent ? vd->vdev_ops->vdev_op_type :
533 spa_name(vd->vdev_spa));
534
535 for (ss = avl_first(t); ss; ss = AVL_NEXT(t, ss)) {
536 /*
537 * Everything in this DTL must appear in all parent DTL unions.
538 */
539 for (pvd = vd; pvd; pvd = pvd->vdev_parent)
540 ASSERT(vdev_dtl_contains(&pvd->vdev_dtl_map,
541 ss->ss_start, ss->ss_end - ss->ss_start));
542 (void) printf("\t%*soutage [%llu,%llu] length %llu\n",
543 indent, "",
544 (u_longlong_t)ss->ss_start,
545 (u_longlong_t)ss->ss_end - 1,
546 (u_longlong_t)(ss->ss_end - ss->ss_start));
547 }
548
549 (void) printf("\n");
550
551 if (dump_opt['d'] > 5 && vd->vdev_children == 0) {
552 dump_spacemap(vd->vdev_spa->spa_meta_objset, &vd->vdev_dtl,
553 &vd->vdev_dtl_map);
554 (void) printf("\n");
555 }
556
557 for (c = 0; c < vd->vdev_children; c++)
558 dump_dtl(vd->vdev_child[c], indent + 4);
559 }
560
561 /*ARGSUSED*/
562 static void
563 dump_dnode(objset_t *os, uint64_t object, void *data, size_t size)
564 {
565 }
566
567 static uint64_t
568 blkid2offset(const dnode_phys_t *dnp, int level, uint64_t blkid)
569 {
570 if (level < 0)
571 return (blkid);
572
573 return ((blkid << (level * (dnp->dn_indblkshift - SPA_BLKPTRSHIFT))) *
574 dnp->dn_datablkszsec << SPA_MINBLOCKSHIFT);
575 }
576
577 static void
578 sprintf_blkptr_compact(char *blkbuf, blkptr_t *bp, int alldvas)
579 {
580 dva_t *dva = bp->blk_dva;
581 int ndvas = alldvas ? BP_GET_NDVAS(bp) : 1;
582 int i;
583
584 blkbuf[0] = '\0';
585
586 for (i = 0; i < ndvas; i++)
587 (void) sprintf(blkbuf + strlen(blkbuf), "%llu:%llx:%llx ",
588 (u_longlong_t)DVA_GET_VDEV(&dva[i]),
589 (u_longlong_t)DVA_GET_OFFSET(&dva[i]),
590 (u_longlong_t)DVA_GET_ASIZE(&dva[i]));
591
592 (void) sprintf(blkbuf + strlen(blkbuf), "%llxL/%llxP F=%llu B=%llu",
593 (u_longlong_t)BP_GET_LSIZE(bp),
594 (u_longlong_t)BP_GET_PSIZE(bp),
595 (u_longlong_t)bp->blk_fill,
596 (u_longlong_t)bp->blk_birth);
597 }
598
599 static void
600 print_indirect(blkptr_t *bp, const zbookmark_t *zb,
601 const dnode_phys_t *dnp)
602 {
603 char blkbuf[BP_SPRINTF_LEN];
604 int l;
605
606 ASSERT3U(BP_GET_TYPE(bp), ==, dnp->dn_type);
607 ASSERT3U(BP_GET_LEVEL(bp), ==, zb->zb_level);
608
609 (void) printf("%16llx ",
610 (u_longlong_t)blkid2offset(dnp, zb->zb_level, zb->zb_blkid));
611
612 ASSERT(zb->zb_level >= 0);
613
614 for (l = dnp->dn_nlevels - 1; l >= -1; l--) {
615 if (l == zb->zb_level) {
616 (void) printf("L%llx", (u_longlong_t)zb->zb_level);
617 } else {
618 (void) printf(" ");
619 }
620 }
621
622 sprintf_blkptr_compact(blkbuf, bp, dump_opt['d'] > 5 ? 1 : 0);
623 (void) printf("%s\n", blkbuf);
624 }
625
626 #define SET_BOOKMARK(zb, objset, object, level, blkid) \
627 { \
628 (zb)->zb_objset = objset; \
629 (zb)->zb_object = object; \
630 (zb)->zb_level = level; \
631 (zb)->zb_blkid = blkid; \
632 }
633
634 static int
635 visit_indirect(spa_t *spa, const dnode_phys_t *dnp,
636 blkptr_t *bp, const zbookmark_t *zb)
637 {
638 int err;
639
640 if (bp->blk_birth == 0)
641 return (0);
642
643 print_indirect(bp, zb, dnp);
644
645 if (BP_GET_LEVEL(bp) > 0) {
646 uint32_t flags = ARC_WAIT;
647 int i;
648 blkptr_t *cbp;
649 int epb = BP_GET_LSIZE(bp) >> SPA_BLKPTRSHIFT;
650 arc_buf_t *buf;
651 uint64_t fill = 0;
652
653 err = arc_read_nolock(NULL, spa, bp, arc_getbuf_func, &buf,
654 ZIO_PRIORITY_ASYNC_READ, ZIO_FLAG_CANFAIL, &flags, zb);
655 if (err)
656 return (err);
657
658 /* recursively visit blocks below this */
659 cbp = buf->b_data;
660 for (i = 0; i < epb; i++, cbp++) {
661 zbookmark_t czb;
662
663 SET_BOOKMARK(&czb, zb->zb_objset, zb->zb_object,
664 zb->zb_level - 1,
665 zb->zb_blkid * epb + i);
666 err = visit_indirect(spa, dnp, cbp, &czb);
667 if (err)
668 break;
669 fill += cbp->blk_fill;
670 }
671 ASSERT3U(fill, ==, bp->blk_fill);
672 (void) arc_buf_remove_ref(buf, &buf);
673 }
674
675 return (err);
676 }
677
678 /*ARGSUSED*/
679 static void
680 dump_indirect(dnode_t *dn)
681 {
682 dnode_phys_t *dnp = dn->dn_phys;
683 int j;
684 zbookmark_t czb;
685
686 (void) printf("Indirect blocks:\n");
687
688 SET_BOOKMARK(&czb, dmu_objset_id(&dn->dn_objset->os),
689 dn->dn_object, dnp->dn_nlevels - 1, 0);
690 for (j = 0; j < dnp->dn_nblkptr; j++) {
691 czb.zb_blkid = j;
692 (void) visit_indirect(dmu_objset_spa(&dn->dn_objset->os), dnp,
693 &dnp->dn_blkptr[j], &czb);
694 }
695
696 (void) printf("\n");
697 }
698
699 /*ARGSUSED*/
700 static void
701 dump_dsl_dir(objset_t *os, uint64_t object, void *data, size_t size)
702 {
703 dsl_dir_phys_t *dd = data;
704 time_t crtime;
705 char nice[6];
706
707 if (dd == NULL)
708 return;
709
710 ASSERT3U(size, >=, sizeof (dsl_dir_phys_t));
711
712 crtime = dd->dd_creation_time;
713 (void) printf("\t\tcreation_time = %s", ctime(&crtime));
714 (void) printf("\t\thead_dataset_obj = %llu\n",
715 (u_longlong_t)dd->dd_head_dataset_obj);
716 (void) printf("\t\tparent_dir_obj = %llu\n",
717 (u_longlong_t)dd->dd_parent_obj);
718 (void) printf("\t\torigin_obj = %llu\n",
719 (u_longlong_t)dd->dd_origin_obj);
720 (void) printf("\t\tchild_dir_zapobj = %llu\n",
721 (u_longlong_t)dd->dd_child_dir_zapobj);
722 nicenum(dd->dd_used_bytes, nice);
723 (void) printf("\t\tused_bytes = %s\n", nice);
724 nicenum(dd->dd_compressed_bytes, nice);
725 (void) printf("\t\tcompressed_bytes = %s\n", nice);
726 nicenum(dd->dd_uncompressed_bytes, nice);
727 (void) printf("\t\tuncompressed_bytes = %s\n", nice);
728 nicenum(dd->dd_quota, nice);
729 (void) printf("\t\tquota = %s\n", nice);
730 nicenum(dd->dd_reserved, nice);
731 (void) printf("\t\treserved = %s\n", nice);
732 (void) printf("\t\tprops_zapobj = %llu\n",
733 (u_longlong_t)dd->dd_props_zapobj);
734 (void) printf("\t\tdeleg_zapobj = %llu\n",
735 (u_longlong_t)dd->dd_deleg_zapobj);
736 (void) printf("\t\tflags = %llx\n",
737 (u_longlong_t)dd->dd_flags);
738
739 #define DO(which) \
740 nicenum(dd->dd_used_breakdown[DD_USED_ ## which], nice); \
741 (void) printf("\t\tused_breakdown[" #which "] = %s\n", nice)
742 DO(HEAD);
743 DO(SNAP);
744 DO(CHILD);
745 DO(CHILD_RSRV);
746 DO(REFRSRV);
747 #undef DO
748 }
749
750 /*ARGSUSED*/
751 static void
752 dump_dsl_dataset(objset_t *os, uint64_t object, void *data, size_t size)
753 {
754 dsl_dataset_phys_t *ds = data;
755 time_t crtime;
756 char used[6], compressed[6], uncompressed[6], unique[6];
757 char blkbuf[BP_SPRINTF_LEN];
758
759 if (ds == NULL)
760 return;
761
762 ASSERT(size == sizeof (*ds));
763 crtime = ds->ds_creation_time;
764 nicenum(ds->ds_used_bytes, used);
765 nicenum(ds->ds_compressed_bytes, compressed);
766 nicenum(ds->ds_uncompressed_bytes, uncompressed);
767 nicenum(ds->ds_unique_bytes, unique);
768 sprintf_blkptr(blkbuf, BP_SPRINTF_LEN, &ds->ds_bp);
769
770 (void) printf("\t\tdir_obj = %llu\n",
771 (u_longlong_t)ds->ds_dir_obj);
772 (void) printf("\t\tprev_snap_obj = %llu\n",
773 (u_longlong_t)ds->ds_prev_snap_obj);
774 (void) printf("\t\tprev_snap_txg = %llu\n",
775 (u_longlong_t)ds->ds_prev_snap_txg);
776 (void) printf("\t\tnext_snap_obj = %llu\n",
777 (u_longlong_t)ds->ds_next_snap_obj);
778 (void) printf("\t\tsnapnames_zapobj = %llu\n",
779 (u_longlong_t)ds->ds_snapnames_zapobj);
780 (void) printf("\t\tnum_children = %llu\n",
781 (u_longlong_t)ds->ds_num_children);
782 (void) printf("\t\tcreation_time = %s", ctime(&crtime));
783 (void) printf("\t\tcreation_txg = %llu\n",
784 (u_longlong_t)ds->ds_creation_txg);
785 (void) printf("\t\tdeadlist_obj = %llu\n",
786 (u_longlong_t)ds->ds_deadlist_obj);
787 (void) printf("\t\tused_bytes = %s\n", used);
788 (void) printf("\t\tcompressed_bytes = %s\n", compressed);
789 (void) printf("\t\tuncompressed_bytes = %s\n", uncompressed);
790 (void) printf("\t\tunique = %s\n", unique);
791 (void) printf("\t\tfsid_guid = %llu\n",
792 (u_longlong_t)ds->ds_fsid_guid);
793 (void) printf("\t\tguid = %llu\n",
794 (u_longlong_t)ds->ds_guid);
795 (void) printf("\t\tflags = %llx\n",
796 (u_longlong_t)ds->ds_flags);
797 (void) printf("\t\tnext_clones_obj = %llu\n",
798 (u_longlong_t)ds->ds_next_clones_obj);
799 (void) printf("\t\tprops_obj = %llu\n",
800 (u_longlong_t)ds->ds_props_obj);
801 (void) printf("\t\tbp = %s\n", blkbuf);
802 }
803
804 static void
805 dump_bplist(objset_t *mos, uint64_t object, char *name)
806 {
807 bplist_t bpl = { 0 };
808 blkptr_t blk, *bp = &blk;
809 uint64_t itor = 0;
810 char bytes[6];
811 char comp[6];
812 char uncomp[6];
813
814 if (dump_opt['d'] < 3)
815 return;
816
817 mutex_init(&bpl.bpl_lock, NULL, MUTEX_DEFAULT, NULL);
818 VERIFY(0 == bplist_open(&bpl, mos, object));
819 if (bplist_empty(&bpl)) {
820 bplist_close(&bpl);
821 mutex_destroy(&bpl.bpl_lock);
822 return;
823 }
824
825 nicenum(bpl.bpl_phys->bpl_bytes, bytes);
826 if (bpl.bpl_dbuf->db_size == sizeof (bplist_phys_t)) {
827 nicenum(bpl.bpl_phys->bpl_comp, comp);
828 nicenum(bpl.bpl_phys->bpl_uncomp, uncomp);
829 (void) printf("\n %s: %llu entries, %s (%s/%s comp)\n",
830 name, (u_longlong_t)bpl.bpl_phys->bpl_entries,
831 bytes, comp, uncomp);
832 } else {
833 (void) printf("\n %s: %llu entries, %s\n",
834 name, (u_longlong_t)bpl.bpl_phys->bpl_entries, bytes);
835 }
836
837 if (dump_opt['d'] < 5) {
838 bplist_close(&bpl);
839 mutex_destroy(&bpl.bpl_lock);
840 return;
841 }
842
843 (void) printf("\n");
844
845 while (bplist_iterate(&bpl, &itor, bp) == 0) {
846 char blkbuf[BP_SPRINTF_LEN];
847
848 ASSERT(bp->blk_birth != 0);
849 sprintf_blkptr_compact(blkbuf, bp, dump_opt['d'] > 5 ? 1 : 0);
850 (void) printf("\tItem %3llu: %s\n",
851 (u_longlong_t)itor - 1, blkbuf);
852 }
853
854 bplist_close(&bpl);
855 mutex_destroy(&bpl.bpl_lock);
856 }
857
858 static avl_tree_t idx_tree;
859 static avl_tree_t domain_tree;
860 static boolean_t fuid_table_loaded;
861
862 static void
863 fuid_table_destroy()
864 {
865 if (fuid_table_loaded) {
866 zfs_fuid_table_destroy(&idx_tree, &domain_tree);
867 fuid_table_loaded = B_FALSE;
868 }
869 }
870
871 /*
872 * print uid or gid information.
873 * For normal POSIX id just the id is printed in decimal format.
874 * For CIFS files with FUID the fuid is printed in hex followed by
875 * the doman-rid string.
876 */
877 static void
878 print_idstr(uint64_t id, const char *id_type)
879 {
880 if (FUID_INDEX(id)) {
881 char *domain;
882
883 domain = zfs_fuid_idx_domain(&idx_tree, FUID_INDEX(id));
884 (void) printf("\t%s %llx [%s-%d]\n", id_type,
885 (u_longlong_t)id, domain, (int)FUID_RID(id));
886 } else {
887 (void) printf("\t%s %llu\n", id_type, (u_longlong_t)id);
888 }
889
890 }
891
892 static void
893 dump_uidgid(objset_t *os, znode_phys_t *zp)
894 {
895 uint32_t uid_idx, gid_idx;
896
897 uid_idx = FUID_INDEX(zp->zp_uid);
898 gid_idx = FUID_INDEX(zp->zp_gid);
899
900 /* Load domain table, if not already loaded */
901 if (!fuid_table_loaded && (uid_idx || gid_idx)) {
902 uint64_t fuid_obj;
903
904 /* first find the fuid object. It lives in the master node */
905 VERIFY(zap_lookup(os, MASTER_NODE_OBJ, ZFS_FUID_TABLES,
906 8, 1, &fuid_obj) == 0);
907 (void) zfs_fuid_table_load(os, fuid_obj,
908 &idx_tree, &domain_tree);
909 fuid_table_loaded = B_TRUE;
910 }
911
912 print_idstr(zp->zp_uid, "uid");
913 print_idstr(zp->zp_gid, "gid");
914 }
915
916 /*ARGSUSED*/
917 static void
918 dump_znode(objset_t *os, uint64_t object, void *data, size_t size)
919 {
920 znode_phys_t *zp = data;
921 time_t z_crtime, z_atime, z_mtime, z_ctime;
922 char path[MAXPATHLEN * 2]; /* allow for xattr and failure prefix */
923 int error;
924
925 ASSERT(size >= sizeof (znode_phys_t));
926
927 error = zfs_obj_to_path(os, object, path, sizeof (path));
928 if (error != 0) {
929 (void) snprintf(path, sizeof (path), "\?\?\?<object#%llu>",
930 (u_longlong_t)object);
931 }
932
933 if (dump_opt['d'] < 3) {
934 (void) printf("\t%s\n", path);
935 return;
936 }
937
938 z_crtime = (time_t)zp->zp_crtime[0];
939 z_atime = (time_t)zp->zp_atime[0];
940 z_mtime = (time_t)zp->zp_mtime[0];
941 z_ctime = (time_t)zp->zp_ctime[0];
942
943 (void) printf("\tpath %s\n", path);
944 dump_uidgid(os, zp);
945 (void) printf("\tatime %s", ctime(&z_atime));
946 (void) printf("\tmtime %s", ctime(&z_mtime));
947 (void) printf("\tctime %s", ctime(&z_ctime));
948 (void) printf("\tcrtime %s", ctime(&z_crtime));
949 (void) printf("\tgen %llu\n", (u_longlong_t)zp->zp_gen);
950 (void) printf("\tmode %llo\n", (u_longlong_t)zp->zp_mode);
951 (void) printf("\tsize %llu\n", (u_longlong_t)zp->zp_size);
952 (void) printf("\tparent %llu\n", (u_longlong_t)zp->zp_parent);
953 (void) printf("\tlinks %llu\n", (u_longlong_t)zp->zp_links);
954 (void) printf("\txattr %llu\n", (u_longlong_t)zp->zp_xattr);
955 (void) printf("\trdev 0x%016llx\n", (u_longlong_t)zp->zp_rdev);
956 }
957
958 /*ARGSUSED*/
959 static void
960 dump_acl(objset_t *os, uint64_t object, void *data, size_t size)
961 {
962 }
963
964 /*ARGSUSED*/
965 static void
966 dump_dmu_objset(objset_t *os, uint64_t object, void *data, size_t size)
967 {
968 }
969
970 static object_viewer_t *object_viewer[DMU_OT_NUMTYPES] = {
971 dump_none, /* unallocated */
972 dump_zap, /* object directory */
973 dump_uint64, /* object array */
974 dump_none, /* packed nvlist */
975 dump_packed_nvlist, /* packed nvlist size */
976 dump_none, /* bplist */
977 dump_none, /* bplist header */
978 dump_none, /* SPA space map header */
979 dump_none, /* SPA space map */
980 dump_none, /* ZIL intent log */
981 dump_dnode, /* DMU dnode */
982 dump_dmu_objset, /* DMU objset */
983 dump_dsl_dir, /* DSL directory */
984 dump_zap, /* DSL directory child map */
985 dump_zap, /* DSL dataset snap map */
986 dump_zap, /* DSL props */
987 dump_dsl_dataset, /* DSL dataset */
988 dump_znode, /* ZFS znode */
989 dump_acl, /* ZFS V0 ACL */
990 dump_uint8, /* ZFS plain file */
991 dump_zpldir, /* ZFS directory */
992 dump_zap, /* ZFS master node */
993 dump_zap, /* ZFS delete queue */
994 dump_uint8, /* zvol object */
995 dump_zap, /* zvol prop */
996 dump_uint8, /* other uint8[] */
997 dump_uint64, /* other uint64[] */
998 dump_zap, /* other ZAP */
999 dump_zap, /* persistent error log */
1000 dump_uint8, /* SPA history */
1001 dump_uint64, /* SPA history offsets */
1002 dump_zap, /* Pool properties */
1003 dump_zap, /* DSL permissions */
1004 dump_acl, /* ZFS ACL */
1005 dump_uint8, /* ZFS SYSACL */
1006 dump_none, /* FUID nvlist */
1007 dump_packed_nvlist, /* FUID nvlist size */
1008 dump_zap, /* DSL dataset next clones */
1009 dump_zap, /* DSL scrub queue */
1010 };
1011
1012 static void
1013 dump_object(objset_t *os, uint64_t object, int verbosity, int *print_header)
1014 {
1015 dmu_buf_t *db = NULL;
1016 dmu_object_info_t doi;
1017 dnode_t *dn;
1018 void *bonus = NULL;
1019 size_t bsize = 0;
1020 char iblk[6], dblk[6], lsize[6], asize[6], bonus_size[6], segsize[6];
1021 char aux[50];
1022 int error;
1023
1024 if (*print_header) {
1025 (void) printf("\n Object lvl iblk dblk lsize"
1026 " asize type\n");
1027 *print_header = 0;
1028 }
1029
1030 if (object == 0) {
1031 dn = os->os->os_meta_dnode;
1032 } else {
1033 error = dmu_bonus_hold(os, object, FTAG, &db);
1034 if (error)
1035 fatal("dmu_bonus_hold(%llu) failed, errno %u",
1036 object, error);
1037 bonus = db->db_data;
1038 bsize = db->db_size;
1039 dn = ((dmu_buf_impl_t *)db)->db_dnode;
1040 }
1041 dmu_object_info_from_dnode(dn, &doi);
1042
1043 nicenum(doi.doi_metadata_block_size, iblk);
1044 nicenum(doi.doi_data_block_size, dblk);
1045 nicenum(doi.doi_data_block_size * (doi.doi_max_block_offset + 1),
1046 lsize);
1047 nicenum(doi.doi_physical_blks << 9, asize);
1048 nicenum(doi.doi_bonus_size, bonus_size);
1049
1050 aux[0] = '\0';
1051
1052 if (doi.doi_checksum != ZIO_CHECKSUM_INHERIT || verbosity >= 6) {
1053 (void) snprintf(aux + strlen(aux), sizeof (aux), " (K=%s)",
1054 zio_checksum_table[doi.doi_checksum].ci_name);
1055 }
1056
1057 if (doi.doi_compress != ZIO_COMPRESS_INHERIT || verbosity >= 6) {
1058 (void) snprintf(aux + strlen(aux), sizeof (aux), " (Z=%s)",
1059 zio_compress_table[doi.doi_compress].ci_name);
1060 }
1061
1062 (void) printf("%10lld %3u %5s %5s %5s %5s %s%s\n",
1063 (u_longlong_t)object, doi.doi_indirection, iblk, dblk, lsize,
1064 asize, dmu_ot[doi.doi_type].ot_name, aux);
1065
1066 if (doi.doi_bonus_type != DMU_OT_NONE && verbosity > 3) {
1067 (void) printf("%10s %3s %5s %5s %5s %5s %s\n",
1068 "", "", "", "", bonus_size, "bonus",
1069 dmu_ot[doi.doi_bonus_type].ot_name);
1070 }
1071
1072 if (verbosity >= 4) {
1073 object_viewer[doi.doi_bonus_type](os, object, bonus, bsize);
1074 object_viewer[doi.doi_type](os, object, NULL, 0);
1075 *print_header = 1;
1076 }
1077
1078 if (verbosity >= 5)
1079 dump_indirect(dn);
1080
1081 if (verbosity >= 5) {
1082 /*
1083 * Report the list of segments that comprise the object.
1084 */
1085 uint64_t start = 0;
1086 uint64_t end;
1087 uint64_t blkfill = 1;
1088 int minlvl = 1;
1089
1090 if (dn->dn_type == DMU_OT_DNODE) {
1091 minlvl = 0;
1092 blkfill = DNODES_PER_BLOCK;
1093 }
1094
1095 for (;;) {
1096 error = dnode_next_offset(dn,
1097 0, &start, minlvl, blkfill, 0);
1098 if (error)
1099 break;
1100 end = start;
1101 error = dnode_next_offset(dn,
1102 DNODE_FIND_HOLE, &end, minlvl, blkfill, 0);
1103 nicenum(end - start, segsize);
1104 (void) printf("\t\tsegment [%016llx, %016llx)"
1105 " size %5s\n", (u_longlong_t)start,
1106 (u_longlong_t)end, segsize);
1107 if (error)
1108 break;
1109 start = end;
1110 }
1111 }
1112
1113 if (db != NULL)
1114 dmu_buf_rele(db, FTAG);
1115 }
1116
1117 static char *objset_types[DMU_OST_NUMTYPES] = {
1118 "NONE", "META", "ZPL", "ZVOL", "OTHER", "ANY" };
1119
1120 static void
1121 dump_dir(objset_t *os)
1122 {
1123 dmu_objset_stats_t dds;
1124 uint64_t object, object_count;
1125 uint64_t refdbytes, usedobjs, scratch;
1126 char numbuf[8];
1127 char blkbuf[BP_SPRINTF_LEN];
1128 char osname[MAXNAMELEN];
1129 char *type = "UNKNOWN";
1130 int verbosity = dump_opt['d'];
1131 int print_header = 1;
1132 int i, error;
1133
1134 dmu_objset_fast_stat(os, &dds);
1135
1136 if (dds.dds_type < DMU_OST_NUMTYPES)
1137 type = objset_types[dds.dds_type];
1138
1139 if (dds.dds_type == DMU_OST_META) {
1140 dds.dds_creation_txg = TXG_INITIAL;
1141 usedobjs = os->os->os_rootbp->blk_fill;
1142 refdbytes = os->os->os_spa->spa_dsl_pool->
1143 dp_mos_dir->dd_phys->dd_used_bytes;
1144 } else {
1145 dmu_objset_space(os, &refdbytes, &scratch, &usedobjs, &scratch);
1146 }
1147
1148 ASSERT3U(usedobjs, ==, os->os->os_rootbp->blk_fill);
1149
1150 nicenum(refdbytes, numbuf);
1151
1152 if (verbosity >= 4) {
1153 (void) strcpy(blkbuf, ", rootbp ");
1154 sprintf_blkptr(blkbuf + strlen(blkbuf),
1155 BP_SPRINTF_LEN - strlen(blkbuf), os->os->os_rootbp);
1156 } else {
1157 blkbuf[0] = '\0';
1158 }
1159
1160 dmu_objset_name(os, osname);
1161
1162 (void) printf("Dataset %s [%s], ID %llu, cr_txg %llu, "
1163 "%s, %llu objects%s\n",
1164 osname, type, (u_longlong_t)dmu_objset_id(os),
1165 (u_longlong_t)dds.dds_creation_txg,
1166 numbuf, (u_longlong_t)usedobjs, blkbuf);
1167
1168 dump_intent_log(dmu_objset_zil(os));
1169
1170 if (dmu_objset_ds(os) != NULL)
1171 dump_bplist(dmu_objset_pool(os)->dp_meta_objset,
1172 dmu_objset_ds(os)->ds_phys->ds_deadlist_obj, "Deadlist");
1173
1174 if (verbosity < 2)
1175 return;
1176
1177 if (os->os->os_rootbp->blk_birth == 0)
1178 return;
1179
1180 if (zopt_objects != 0) {
1181 for (i = 0; i < zopt_objects; i++)
1182 dump_object(os, zopt_object[i], verbosity,
1183 &print_header);
1184 (void) printf("\n");
1185 return;
1186 }
1187
1188 dump_object(os, 0, verbosity, &print_header);
1189 object_count = 1;
1190
1191 object = 0;
1192 while ((error = dmu_object_next(os, &object, B_FALSE, 0)) == 0) {
1193 dump_object(os, object, verbosity, &print_header);
1194 object_count++;
1195 }
1196
1197 ASSERT3U(object_count, ==, usedobjs);
1198
1199 (void) printf("\n");
1200
1201 if (error != ESRCH)
1202 fatal("dmu_object_next() = %d", error);
1203 }
1204
1205 static void
1206 dump_uberblock(uberblock_t *ub)
1207 {
1208 time_t timestamp = ub->ub_timestamp;
1209
1210 (void) printf("Uberblock\n\n");
1211 (void) printf("\tmagic = %016llx\n", (u_longlong_t)ub->ub_magic);
1212 (void) printf("\tversion = %llu\n", (u_longlong_t)ub->ub_version);
1213 (void) printf("\ttxg = %llu\n", (u_longlong_t)ub->ub_txg);
1214 (void) printf("\tguid_sum = %llu\n", (u_longlong_t)ub->ub_guid_sum);
1215 (void) printf("\ttimestamp = %llu UTC = %s",
1216 (u_longlong_t)ub->ub_timestamp, asctime(localtime(&timestamp)));
1217 if (dump_opt['u'] >= 3) {
1218 char blkbuf[BP_SPRINTF_LEN];
1219 sprintf_blkptr(blkbuf, BP_SPRINTF_LEN, &ub->ub_rootbp);
1220 (void) printf("\trootbp = %s\n", blkbuf);
1221 }
1222 (void) printf("\n");
1223 }
1224
1225 static void
1226 dump_config(const char *pool)
1227 {
1228 spa_t *spa = NULL;
1229
1230 mutex_enter(&spa_namespace_lock);
1231 while ((spa = spa_next(spa)) != NULL) {
1232 if (pool == NULL)
1233 (void) printf("%s\n", spa_name(spa));
1234 if (pool == NULL || strcmp(pool, spa_name(spa)) == 0)
1235 dump_nvlist(spa->spa_config, 4);
1236 }
1237 mutex_exit(&spa_namespace_lock);
1238 }
1239
1240 static void
1241 dump_cachefile(const char *cachefile)
1242 {
1243 int fd;
1244 struct stat64 statbuf;
1245 char *buf;
1246 nvlist_t *config;
1247
1248 if ((fd = open64(cachefile, O_RDONLY)) < 0) {
1249 (void) printf("cannot open '%s': %s\n", cachefile,
1250 strerror(errno));
1251 exit(1);
1252 }
1253
1254 if (fstat64(fd, &statbuf) != 0) {
1255 (void) printf("failed to stat '%s': %s\n", cachefile,
1256 strerror(errno));
1257 exit(1);
1258 }
1259
1260 if ((buf = malloc(statbuf.st_size)) == NULL) {
1261 (void) fprintf(stderr, "failed to allocate %llu bytes\n",
1262 (u_longlong_t)statbuf.st_size);
1263 exit(1);
1264 }
1265
1266 if (read(fd, buf, statbuf.st_size) != statbuf.st_size) {
1267 (void) fprintf(stderr, "failed to read %llu bytes\n",
1268 (u_longlong_t)statbuf.st_size);
1269 exit(1);
1270 }
1271
1272 (void) close(fd);
1273
1274 if (nvlist_unpack(buf, statbuf.st_size, &config, 0) != 0) {
1275 (void) fprintf(stderr, "failed to unpack nvlist\n");
1276 exit(1);
1277 }
1278
1279 free(buf);
1280
1281 dump_nvlist(config, 0);
1282
1283 nvlist_free(config);
1284 }
1285
1286 static void
1287 dump_label(const char *dev)
1288 {
1289 int fd;
1290 vdev_label_t label;
1291 char *buf = label.vl_vdev_phys.vp_nvlist;
1292 size_t buflen = sizeof (label.vl_vdev_phys.vp_nvlist);
1293 struct stat64 statbuf;
1294 uint64_t psize;
1295 int l;
1296
1297 if ((fd = open64(dev, O_RDONLY)) < 0) {
1298 (void) printf("cannot open '%s': %s\n", dev, strerror(errno));
1299 exit(1);
1300 }
1301
1302 if (fstat64(fd, &statbuf) != 0) {
1303 (void) printf("failed to stat '%s': %s\n", dev,
1304 strerror(errno));
1305 exit(1);
1306 }
1307
1308 psize = statbuf.st_size;
1309 psize = P2ALIGN(psize, (uint64_t)sizeof (vdev_label_t));
1310
1311 for (l = 0; l < VDEV_LABELS; l++) {
1312
1313 nvlist_t *config = NULL;
1314
1315 (void) printf("--------------------------------------------\n");
1316 (void) printf("LABEL %d\n", l);
1317 (void) printf("--------------------------------------------\n");
1318
1319 if (pread64(fd, &label, sizeof (label),
1320 vdev_label_offset(psize, l, 0)) != sizeof (label)) {
1321 (void) printf("failed to read label %d\n", l);
1322 continue;
1323 }
1324
1325 if (nvlist_unpack(buf, buflen, &config, 0) != 0) {
1326 (void) printf("failed to unpack label %d\n", l);
1327 continue;
1328 }
1329 dump_nvlist(config, 4);
1330 nvlist_free(config);
1331 }
1332 }
1333
1334 /*ARGSUSED*/
1335 static int
1336 dump_one_dir(char *dsname, void *arg)
1337 {
1338 int error;
1339 objset_t *os;
1340
1341 error = dmu_objset_open(dsname, DMU_OST_ANY,
1342 DS_MODE_USER | DS_MODE_READONLY, &os);
1343 if (error) {
1344 (void) printf("Could not open %s\n", dsname);
1345 return (0);
1346 }
1347 dump_dir(os);
1348 dmu_objset_close(os);
1349 fuid_table_destroy();
1350 return (0);
1351 }
1352
1353 static void
1354 zdb_leak(space_map_t *sm, uint64_t start, uint64_t size)
1355 {
1356 vdev_t *vd = sm->sm_ppd;
1357
1358 (void) printf("leaked space: vdev %llu, offset 0x%llx, size %llu\n",
1359 (u_longlong_t)vd->vdev_id, (u_longlong_t)start, (u_longlong_t)size);
1360 }
1361
1362 /* ARGSUSED */
1363 static void
1364 zdb_space_map_load(space_map_t *sm)
1365 {
1366 }
1367
1368 static void
1369 zdb_space_map_unload(space_map_t *sm)
1370 {
1371 space_map_vacate(sm, zdb_leak, sm);
1372 }
1373
1374 /* ARGSUSED */
1375 static void
1376 zdb_space_map_claim(space_map_t *sm, uint64_t start, uint64_t size)
1377 {
1378 }
1379
1380 static space_map_ops_t zdb_space_map_ops = {
1381 zdb_space_map_load,
1382 zdb_space_map_unload,
1383 NULL, /* alloc */
1384 zdb_space_map_claim,
1385 NULL /* free */
1386 };
1387
1388 static void
1389 zdb_leak_init(spa_t *spa)
1390 {
1391 vdev_t *rvd = spa->spa_root_vdev;
1392
1393 for (int c = 0; c < rvd->vdev_children; c++) {
1394 vdev_t *vd = rvd->vdev_child[c];
1395 for (int m = 0; m < vd->vdev_ms_count; m++) {
1396 metaslab_t *msp = vd->vdev_ms[m];
1397 mutex_enter(&msp->ms_lock);
1398 VERIFY(space_map_load(&msp->ms_map, &zdb_space_map_ops,
1399 SM_ALLOC, &msp->ms_smo, spa->spa_meta_objset) == 0);
1400 msp->ms_map.sm_ppd = vd;
1401 mutex_exit(&msp->ms_lock);
1402 }
1403 }
1404 }
1405
1406 static void
1407 zdb_leak_fini(spa_t *spa)
1408 {
1409 vdev_t *rvd = spa->spa_root_vdev;
1410
1411 for (int c = 0; c < rvd->vdev_children; c++) {
1412 vdev_t *vd = rvd->vdev_child[c];
1413 for (int m = 0; m < vd->vdev_ms_count; m++) {
1414 metaslab_t *msp = vd->vdev_ms[m];
1415 mutex_enter(&msp->ms_lock);
1416 space_map_unload(&msp->ms_map);
1417 mutex_exit(&msp->ms_lock);
1418 }
1419 }
1420 }
1421
1422 /*
1423 * Verify that the sum of the sizes of all blocks in the pool adds up
1424 * to the SPA's sa_alloc total.
1425 */
1426 typedef struct zdb_blkstats {
1427 uint64_t zb_asize;
1428 uint64_t zb_lsize;
1429 uint64_t zb_psize;
1430 uint64_t zb_count;
1431 } zdb_blkstats_t;
1432
1433 #define DMU_OT_DEFERRED DMU_OT_NONE
1434 #define DMU_OT_TOTAL DMU_OT_NUMTYPES
1435
1436 #define ZB_TOTAL DN_MAX_LEVELS
1437
1438 typedef struct zdb_cb {
1439 zdb_blkstats_t zcb_type[ZB_TOTAL + 1][DMU_OT_TOTAL + 1];
1440 uint64_t zcb_errors[256];
1441 int zcb_readfails;
1442 int zcb_haderrors;
1443 } zdb_cb_t;
1444
1445 static void
1446 zdb_count_block(spa_t *spa, zdb_cb_t *zcb, blkptr_t *bp, dmu_object_type_t type)
1447 {
1448 for (int i = 0; i < 4; i++) {
1449 int l = (i < 2) ? BP_GET_LEVEL(bp) : ZB_TOTAL;
1450 int t = (i & 1) ? type : DMU_OT_TOTAL;
1451 zdb_blkstats_t *zb = &zcb->zcb_type[l][t];
1452
1453 zb->zb_asize += BP_GET_ASIZE(bp);
1454 zb->zb_lsize += BP_GET_LSIZE(bp);
1455 zb->zb_psize += BP_GET_PSIZE(bp);
1456 zb->zb_count++;
1457 }
1458
1459 if (dump_opt['S']) {
1460 boolean_t print_sig;
1461
1462 print_sig = !zdb_sig_user_data || (BP_GET_LEVEL(bp) == 0 &&
1463 BP_GET_TYPE(bp) == DMU_OT_PLAIN_FILE_CONTENTS);
1464
1465 if (BP_GET_CHECKSUM(bp) < zdb_sig_cksumalg)
1466 print_sig = B_FALSE;
1467
1468 if (print_sig) {
1469 (void) printf("%llu\t%lld\t%lld\t%s\t%s\t%s\t"
1470 "%llx:%llx:%llx:%llx\n",
1471 (u_longlong_t)BP_GET_LEVEL(bp),
1472 (longlong_t)BP_GET_PSIZE(bp),
1473 (longlong_t)BP_GET_NDVAS(bp),
1474 dmu_ot[BP_GET_TYPE(bp)].ot_name,
1475 zio_checksum_table[BP_GET_CHECKSUM(bp)].ci_name,
1476 zio_compress_table[BP_GET_COMPRESS(bp)].ci_name,
1477 (u_longlong_t)bp->blk_cksum.zc_word[0],
1478 (u_longlong_t)bp->blk_cksum.zc_word[1],
1479 (u_longlong_t)bp->blk_cksum.zc_word[2],
1480 (u_longlong_t)bp->blk_cksum.zc_word[3]);
1481 }
1482 }
1483
1484 VERIFY(zio_wait(zio_claim(NULL, spa, spa_first_txg(spa), bp,
1485 NULL, NULL, ZIO_FLAG_MUSTSUCCEED)) == 0);
1486 }
1487
1488 static int
1489 zdb_blkptr_cb(spa_t *spa, blkptr_t *bp, const zbookmark_t *zb,
1490 const dnode_phys_t *dnp, void *arg)
1491 {
1492 zdb_cb_t *zcb = arg;
1493 char blkbuf[BP_SPRINTF_LEN];
1494
1495 if (bp == NULL)
1496 return (0);
1497
1498 zdb_count_block(spa, zcb, bp, BP_GET_TYPE(bp));
1499
1500 if (dump_opt['c'] || dump_opt['S']) {
1501 int ioerr, size;
1502 void *data;
1503
1504 size = BP_GET_LSIZE(bp);
1505 data = malloc(size);
1506 ioerr = zio_wait(zio_read(NULL, spa, bp, data, size,
1507 NULL, NULL, ZIO_PRIORITY_ASYNC_READ,
1508 ZIO_FLAG_CANFAIL | ZIO_FLAG_SCRUB, zb));
1509 free(data);
1510
1511 /* We expect io errors on intent log */
1512 if (ioerr && BP_GET_TYPE(bp) != DMU_OT_INTENT_LOG) {
1513 zcb->zcb_haderrors = 1;
1514 zcb->zcb_errors[ioerr]++;
1515
1516 if (dump_opt['b'] >= 2)
1517 sprintf_blkptr(blkbuf, BP_SPRINTF_LEN, bp);
1518 else
1519 blkbuf[0] = '\0';
1520
1521 if (!dump_opt['S']) {
1522 (void) printf("zdb_blkptr_cb: "
1523 "Got error %d reading "
1524 "<%llu, %llu, %lld, %llx> %s -- skipping\n",
1525 ioerr,
1526 (u_longlong_t)zb->zb_objset,
1527 (u_longlong_t)zb->zb_object,
1528 (u_longlong_t)zb->zb_level,
1529 (u_longlong_t)zb->zb_blkid,
1530 blkbuf);
1531 }
1532 }
1533 }
1534
1535 zcb->zcb_readfails = 0;
1536
1537 if (dump_opt['b'] >= 4) {
1538 sprintf_blkptr(blkbuf, BP_SPRINTF_LEN, bp);
1539 (void) printf("objset %llu object %llu offset 0x%llx %s\n",
1540 (u_longlong_t)zb->zb_objset,
1541 (u_longlong_t)zb->zb_object,
1542 (u_longlong_t)blkid2offset(dnp, zb->zb_level, zb->zb_blkid),
1543 blkbuf);
1544 }
1545
1546 return (0);
1547 }
1548
1549 static int
1550 dump_block_stats(spa_t *spa)
1551 {
1552 zdb_cb_t zcb = { 0 };
1553 zdb_blkstats_t *zb, *tzb;
1554 uint64_t alloc, space, logalloc;
1555 vdev_t *rvd = spa->spa_root_vdev;
1556 int leaks = 0;
1557 int c, e;
1558
1559 if (!dump_opt['S']) {
1560 (void) printf("\nTraversing all blocks to %sverify"
1561 " nothing leaked ...\n",
1562 dump_opt['c'] ? "verify checksums and " : "");
1563 }
1564
1565 /*
1566 * Load all space maps as SM_ALLOC maps, then traverse the pool
1567 * claiming each block we discover. If the pool is perfectly
1568 * consistent, the space maps will be empty when we're done.
1569 * Anything left over is a leak; any block we can't claim (because
1570 * it's not part of any space map) is a double allocation,
1571 * reference to a freed block, or an unclaimed log block.
1572 */
1573 zdb_leak_init(spa);
1574
1575 /*
1576 * If there's a deferred-free bplist, process that first.
1577 */
1578 if (spa->spa_sync_bplist_obj != 0) {
1579 bplist_t *bpl = &spa->spa_sync_bplist;
1580 blkptr_t blk;
1581 uint64_t itor = 0;
1582
1583 VERIFY(0 == bplist_open(bpl, spa->spa_meta_objset,
1584 spa->spa_sync_bplist_obj));
1585
1586 while (bplist_iterate(bpl, &itor, &blk) == 0) {
1587 if (dump_opt['b'] >= 4) {
1588 char blkbuf[BP_SPRINTF_LEN];
1589 sprintf_blkptr(blkbuf, BP_SPRINTF_LEN, &blk);
1590 (void) printf("[%s] %s\n",
1591 "deferred free", blkbuf);
1592 }
1593 zdb_count_block(spa, &zcb, &blk, DMU_OT_DEFERRED);
1594 }
1595
1596 bplist_close(bpl);
1597 }
1598
1599 zcb.zcb_haderrors |= traverse_pool(spa, zdb_blkptr_cb, &zcb);
1600
1601 if (zcb.zcb_haderrors && !dump_opt['S']) {
1602 (void) printf("\nError counts:\n\n");
1603 (void) printf("\t%5s %s\n", "errno", "count");
1604 for (e = 0; e < 256; e++) {
1605 if (zcb.zcb_errors[e] != 0) {
1606 (void) printf("\t%5d %llu\n",
1607 e, (u_longlong_t)zcb.zcb_errors[e]);
1608 }
1609 }
1610 }
1611
1612 /*
1613 * Report any leaked segments.
1614 */
1615 zdb_leak_fini(spa);
1616
1617 /*
1618 * If we're interested in printing out the blkptr signatures,
1619 * return now as we don't print out anything else (including
1620 * errors and leaks).
1621 */
1622 if (dump_opt['S'])
1623 return (zcb.zcb_haderrors ? 3 : 0);
1624
1625 alloc = spa_get_alloc(spa);
1626 space = spa_get_space(spa);
1627
1628 /*
1629 * Log blocks allocated from a separate log device don't count
1630 * as part of the normal pool space; factor them in here.
1631 */
1632 logalloc = 0;
1633
1634 for (c = 0; c < rvd->vdev_children; c++)
1635 if (rvd->vdev_child[c]->vdev_islog)
1636 logalloc += rvd->vdev_child[c]->vdev_stat.vs_alloc;
1637
1638 tzb = &zcb.zcb_type[ZB_TOTAL][DMU_OT_TOTAL];
1639
1640 if (tzb->zb_asize == alloc + logalloc) {
1641 (void) printf("\n\tNo leaks (block sum matches space"
1642 " maps exactly)\n");
1643 } else {
1644 (void) printf("block traversal size %llu != alloc %llu "
1645 "(leaked %lld)\n",
1646 (u_longlong_t)tzb->zb_asize,
1647 (u_longlong_t)alloc + logalloc,
1648 (u_longlong_t)(alloc + logalloc - tzb->zb_asize));
1649 leaks = 1;
1650 }
1651
1652 if (tzb->zb_count == 0)
1653 return (2);
1654
1655 (void) printf("\n");
1656 (void) printf("\tbp count: %10llu\n",
1657 (u_longlong_t)tzb->zb_count);
1658 (void) printf("\tbp logical: %10llu\t avg: %6llu\n",
1659 (u_longlong_t)tzb->zb_lsize,
1660 (u_longlong_t)(tzb->zb_lsize / tzb->zb_count));
1661 (void) printf("\tbp physical: %10llu\t avg:"
1662 " %6llu\tcompression: %6.2f\n",
1663 (u_longlong_t)tzb->zb_psize,
1664 (u_longlong_t)(tzb->zb_psize / tzb->zb_count),
1665 (double)tzb->zb_lsize / tzb->zb_psize);
1666 (void) printf("\tbp allocated: %10llu\t avg:"
1667 " %6llu\tcompression: %6.2f\n",
1668 (u_longlong_t)tzb->zb_asize,
1669 (u_longlong_t)(tzb->zb_asize / tzb->zb_count),
1670 (double)tzb->zb_lsize / tzb->zb_asize);
1671 (void) printf("\tSPA allocated: %10llu\tused: %5.2f%%\n",
1672 (u_longlong_t)alloc, 100.0 * alloc / space);
1673
1674 if (dump_opt['b'] >= 2) {
1675 int l, t, level;
1676 (void) printf("\nBlocks\tLSIZE\tPSIZE\tASIZE"
1677 "\t avg\t comp\t%%Total\tType\n");
1678
1679 for (t = 0; t <= DMU_OT_NUMTYPES; t++) {
1680 char csize[6], lsize[6], psize[6], asize[6], avg[6];
1681 char *typename;
1682
1683 typename = t == DMU_OT_DEFERRED ? "deferred free" :
1684 t == DMU_OT_TOTAL ? "Total" : dmu_ot[t].ot_name;
1685
1686 if (zcb.zcb_type[ZB_TOTAL][t].zb_asize == 0) {
1687 (void) printf("%6s\t%5s\t%5s\t%5s"
1688 "\t%5s\t%5s\t%6s\t%s\n",
1689 "-",
1690 "-",
1691 "-",
1692 "-",
1693 "-",
1694 "-",
1695 "-",
1696 typename);
1697 continue;
1698 }
1699
1700 for (l = ZB_TOTAL - 1; l >= -1; l--) {
1701 level = (l == -1 ? ZB_TOTAL : l);
1702 zb = &zcb.zcb_type[level][t];
1703
1704 if (zb->zb_asize == 0)
1705 continue;
1706
1707 if (dump_opt['b'] < 3 && level != ZB_TOTAL)
1708 continue;
1709
1710 if (level == 0 && zb->zb_asize ==
1711 zcb.zcb_type[ZB_TOTAL][t].zb_asize)
1712 continue;
1713
1714 nicenum(zb->zb_count, csize);
1715 nicenum(zb->zb_lsize, lsize);
1716 nicenum(zb->zb_psize, psize);
1717 nicenum(zb->zb_asize, asize);
1718 nicenum(zb->zb_asize / zb->zb_count, avg);
1719
1720 (void) printf("%6s\t%5s\t%5s\t%5s\t%5s"
1721 "\t%5.2f\t%6.2f\t",
1722 csize, lsize, psize, asize, avg,
1723 (double)zb->zb_lsize / zb->zb_psize,
1724 100.0 * zb->zb_asize / tzb->zb_asize);
1725
1726 if (level == ZB_TOTAL)
1727 (void) printf("%s\n", typename);
1728 else
1729 (void) printf(" L%d %s\n",
1730 level, typename);
1731 }
1732 }
1733 }
1734
1735 (void) printf("\n");
1736
1737 if (leaks)
1738 return (2);
1739
1740 if (zcb.zcb_haderrors)
1741 return (3);
1742
1743 return (0);
1744 }
1745
1746 static void
1747 dump_zpool(spa_t *spa)
1748 {
1749 dsl_pool_t *dp = spa_get_dsl(spa);
1750 int rc = 0;
1751
1752 if (dump_opt['u'])
1753 dump_uberblock(&spa->spa_uberblock);
1754
1755 if (dump_opt['d'] || dump_opt['i']) {
1756 dump_dir(dp->dp_meta_objset);
1757 if (dump_opt['d'] >= 3) {
1758 dump_bplist(dp->dp_meta_objset,
1759 spa->spa_sync_bplist_obj, "Deferred frees");
1760 dump_dtl(spa->spa_root_vdev, 0);
1761 dump_metaslabs(spa);
1762 }
1763 (void) dmu_objset_find(spa_name(spa), dump_one_dir, NULL,
1764 DS_FIND_SNAPSHOTS | DS_FIND_CHILDREN);
1765 }
1766
1767 if (dump_opt['b'] || dump_opt['c'] || dump_opt['S'])
1768 rc = dump_block_stats(spa);
1769
1770 if (dump_opt['s'])
1771 show_pool_stats(spa);
1772
1773 if (rc != 0)
1774 exit(rc);
1775 }
1776
1777 #define ZDB_FLAG_CHECKSUM 0x0001
1778 #define ZDB_FLAG_DECOMPRESS 0x0002
1779 #define ZDB_FLAG_BSWAP 0x0004
1780 #define ZDB_FLAG_GBH 0x0008
1781 #define ZDB_FLAG_INDIRECT 0x0010
1782 #define ZDB_FLAG_PHYS 0x0020
1783 #define ZDB_FLAG_RAW 0x0040
1784 #define ZDB_FLAG_PRINT_BLKPTR 0x0080
1785
1786 int flagbits[256];
1787
1788 static void
1789 zdb_print_blkptr(blkptr_t *bp, int flags)
1790 {
1791 dva_t *dva = bp->blk_dva;
1792 int d;
1793
1794 if (flags & ZDB_FLAG_BSWAP)
1795 byteswap_uint64_array((void *)bp, sizeof (blkptr_t));
1796 /*
1797 * Super-ick warning: This code is also duplicated in
1798 * cmd/mdb/common/modules/zfs/zfs.c . Yeah, I hate code
1799 * replication, too.
1800 */
1801 for (d = 0; d < BP_GET_NDVAS(bp); d++) {
1802 (void) printf("\tDVA[%d]: vdev_id %lld / %llx\n", d,
1803 (longlong_t)DVA_GET_VDEV(&dva[d]),
1804 (longlong_t)DVA_GET_OFFSET(&dva[d]));
1805 (void) printf("\tDVA[%d]: GANG: %-5s GRID: %04llx\t"
1806 "ASIZE: %llx\n", d,
1807 DVA_GET_GANG(&dva[d]) ? "TRUE" : "FALSE",
1808 (longlong_t)DVA_GET_GRID(&dva[d]),
1809 (longlong_t)DVA_GET_ASIZE(&dva[d]));
1810 (void) printf("\tDVA[%d]: :%llu:%llx:%llx:%s%s%s%s\n", d,
1811 (u_longlong_t)DVA_GET_VDEV(&dva[d]),
1812 (longlong_t)DVA_GET_OFFSET(&dva[d]),
1813 (longlong_t)BP_GET_PSIZE(bp),
1814 BP_SHOULD_BYTESWAP(bp) ? "e" : "",
1815 !DVA_GET_GANG(&dva[d]) && BP_GET_LEVEL(bp) != 0 ?
1816 "d" : "",
1817 DVA_GET_GANG(&dva[d]) ? "g" : "",
1818 BP_GET_COMPRESS(bp) != 0 ? "d" : "");
1819 }
1820 (void) printf("\tLSIZE: %-16llx\t\tPSIZE: %llx\n",
1821 (longlong_t)BP_GET_LSIZE(bp), (longlong_t)BP_GET_PSIZE(bp));
1822 (void) printf("\tENDIAN: %6s\t\t\t\t\tTYPE: %s\n",
1823 BP_GET_BYTEORDER(bp) ? "LITTLE" : "BIG",
1824 dmu_ot[BP_GET_TYPE(bp)].ot_name);
1825 (void) printf("\tBIRTH: %-16llx LEVEL: %-2llu\tFILL: %llx\n",
1826 (u_longlong_t)bp->blk_birth, (u_longlong_t)BP_GET_LEVEL(bp),
1827 (u_longlong_t)bp->blk_fill);
1828 (void) printf("\tCKFUNC: %-16s\t\tCOMP: %s\n",
1829 zio_checksum_table[BP_GET_CHECKSUM(bp)].ci_name,
1830 zio_compress_table[BP_GET_COMPRESS(bp)].ci_name);
1831 (void) printf("\tCKSUM: %llx:%llx:%llx:%llx\n",
1832 (u_longlong_t)bp->blk_cksum.zc_word[0],
1833 (u_longlong_t)bp->blk_cksum.zc_word[1],
1834 (u_longlong_t)bp->blk_cksum.zc_word[2],
1835 (u_longlong_t)bp->blk_cksum.zc_word[3]);
1836 }
1837
1838 static void
1839 zdb_dump_indirect(blkptr_t *bp, int nbps, int flags)
1840 {
1841 int i;
1842
1843 for (i = 0; i < nbps; i++)
1844 zdb_print_blkptr(&bp[i], flags);
1845 }
1846
1847 static void
1848 zdb_dump_gbh(void *buf, int flags)
1849 {
1850 zdb_dump_indirect((blkptr_t *)buf, SPA_GBH_NBLKPTRS, flags);
1851 }
1852
1853 static void
1854 zdb_dump_block_raw(void *buf, uint64_t size, int flags)
1855 {
1856 if (flags & ZDB_FLAG_BSWAP)
1857 byteswap_uint64_array(buf, size);
1858 (void) write(2, buf, size);
1859 }
1860
1861 static void
1862 zdb_dump_block(char *label, void *buf, uint64_t size, int flags)
1863 {
1864 uint64_t *d = (uint64_t *)buf;
1865 int nwords = size / sizeof (uint64_t);
1866 int do_bswap = !!(flags & ZDB_FLAG_BSWAP);
1867 int i, j;
1868 char *hdr, *c;
1869
1870
1871 if (do_bswap)
1872 hdr = " 7 6 5 4 3 2 1 0 f e d c b a 9 8";
1873 else
1874 hdr = " 0 1 2 3 4 5 6 7 8 9 a b c d e f";
1875
1876 (void) printf("\n%s\n%6s %s 0123456789abcdef\n", label, "", hdr);
1877
1878 for (i = 0; i < nwords; i += 2) {
1879 (void) printf("%06llx: %016llx %016llx ",
1880 (u_longlong_t)(i * sizeof (uint64_t)),
1881 (u_longlong_t)(do_bswap ? BSWAP_64(d[i]) : d[i]),
1882 (u_longlong_t)(do_bswap ? BSWAP_64(d[i + 1]) : d[i + 1]));
1883
1884 c = (char *)&d[i];
1885 for (j = 0; j < 2 * sizeof (uint64_t); j++)
1886 (void) printf("%c", isprint(c[j]) ? c[j] : '.');
1887 (void) printf("\n");
1888 }
1889 }
1890
1891 /*
1892 * There are two acceptable formats:
1893 * leaf_name - For example: c1t0d0 or /tmp/ztest.0a
1894 * child[.child]* - For example: 0.1.1
1895 *
1896 * The second form can be used to specify arbitrary vdevs anywhere
1897 * in the heirarchy. For example, in a pool with a mirror of
1898 * RAID-Zs, you can specify either RAID-Z vdev with 0.0 or 0.1 .
1899 */
1900 static vdev_t *
1901 zdb_vdev_lookup(vdev_t *vdev, char *path)
1902 {
1903 char *s, *p, *q;
1904 int i;
1905
1906 if (vdev == NULL)
1907 return (NULL);
1908
1909 /* First, assume the x.x.x.x format */
1910 i = (int)strtoul(path, &s, 10);
1911 if (s == path || (s && *s != '.' && *s != '\0'))
1912 goto name;
1913 if (i < 0 || i >= vdev->vdev_children)
1914 return (NULL);
1915
1916 vdev = vdev->vdev_child[i];
1917 if (*s == '\0')
1918 return (vdev);
1919 return (zdb_vdev_lookup(vdev, s+1));
1920
1921 name:
1922 for (i = 0; i < vdev->vdev_children; i++) {
1923 vdev_t *vc = vdev->vdev_child[i];
1924
1925 if (vc->vdev_path == NULL) {
1926 vc = zdb_vdev_lookup(vc, path);
1927 if (vc == NULL)
1928 continue;
1929 else
1930 return (vc);
1931 }
1932
1933 p = strrchr(vc->vdev_path, '/');
1934 p = p ? p + 1 : vc->vdev_path;
1935 q = &vc->vdev_path[strlen(vc->vdev_path) - 2];
1936
1937 if (strcmp(vc->vdev_path, path) == 0)
1938 return (vc);
1939 if (strcmp(p, path) == 0)
1940 return (vc);
1941 if (strcmp(q, "s0") == 0 && strncmp(p, path, q - p) == 0)
1942 return (vc);
1943 }
1944
1945 return (NULL);
1946 }
1947
1948 /*
1949 * Read a block from a pool and print it out. The syntax of the
1950 * block descriptor is:
1951 *
1952 * pool:vdev_specifier:offset:size[:flags]
1953 *
1954 * pool - The name of the pool you wish to read from
1955 * vdev_specifier - Which vdev (see comment for zdb_vdev_lookup)
1956 * offset - offset, in hex, in bytes
1957 * size - Amount of data to read, in hex, in bytes
1958 * flags - A string of characters specifying options
1959 * b: Decode a blkptr at given offset within block
1960 * *c: Calculate and display checksums
1961 * *d: Decompress data before dumping
1962 * e: Byteswap data before dumping
1963 * *g: Display data as a gang block header
1964 * *i: Display as an indirect block
1965 * p: Do I/O to physical offset
1966 * r: Dump raw data to stdout
1967 *
1968 * * = not yet implemented
1969 */
1970 static void
1971 zdb_read_block(char *thing, spa_t **spap)
1972 {
1973 spa_t *spa = *spap;
1974 int flags = 0;
1975 uint64_t offset = 0, size = 0, blkptr_offset = 0;
1976 zio_t *zio;
1977 vdev_t *vd;
1978 void *buf;
1979 char *s, *p, *dup, *pool, *vdev, *flagstr;
1980 int i, error, zio_flags;
1981
1982 dup = strdup(thing);
1983 s = strtok(dup, ":");
1984 pool = s ? s : "";
1985 s = strtok(NULL, ":");
1986 vdev = s ? s : "";
1987 s = strtok(NULL, ":");
1988 offset = strtoull(s ? s : "", NULL, 16);
1989 s = strtok(NULL, ":");
1990 size = strtoull(s ? s : "", NULL, 16);
1991 s = strtok(NULL, ":");
1992 flagstr = s ? s : "";
1993
1994 s = NULL;
1995 if (size == 0)
1996 s = "size must not be zero";
1997 if (!IS_P2ALIGNED(size, DEV_BSIZE))
1998 s = "size must be a multiple of sector size";
1999 if (!IS_P2ALIGNED(offset, DEV_BSIZE))
2000 s = "offset must be a multiple of sector size";
2001 if (s) {
2002 (void) printf("Invalid block specifier: %s - %s\n", thing, s);
2003 free(dup);
2004 return;
2005 }
2006
2007 for (s = strtok(flagstr, ":"); s; s = strtok(NULL, ":")) {
2008 for (i = 0; flagstr[i]; i++) {
2009 int bit = flagbits[(uchar_t)flagstr[i]];
2010
2011 if (bit == 0) {
2012 (void) printf("***Invalid flag: %c\n",
2013 flagstr[i]);
2014 continue;
2015 }
2016 flags |= bit;
2017
2018 /* If it's not something with an argument, keep going */
2019 if ((bit & (ZDB_FLAG_CHECKSUM | ZDB_FLAG_DECOMPRESS |
2020 ZDB_FLAG_PRINT_BLKPTR)) == 0)
2021 continue;
2022
2023 p = &flagstr[i + 1];
2024 if (bit == ZDB_FLAG_PRINT_BLKPTR)
2025 blkptr_offset = strtoull(p, &p, 16);
2026 if (*p != ':' && *p != '\0') {
2027 (void) printf("***Invalid flag arg: '%s'\n", s);
2028 free(dup);
2029 return;
2030 }
2031 }
2032 }
2033
2034 if (spa == NULL || strcmp(spa_name(spa), pool) != 0) {
2035 if (spa)
2036 spa_close(spa, (void *)zdb_read_block);
2037 error = spa_open(pool, spap, (void *)zdb_read_block);
2038 if (error)
2039 fatal("Failed to open pool '%s': %s",
2040 pool, strerror(error));
2041 spa = *spap;
2042 }
2043
2044 vd = zdb_vdev_lookup(spa->spa_root_vdev, vdev);
2045 if (vd == NULL) {
2046 (void) printf("***Invalid vdev: %s\n", vdev);
2047 free(dup);
2048 return;
2049 } else {
2050 if (vd->vdev_path)
2051 (void) printf("Found vdev: %s\n", vd->vdev_path);
2052 else
2053 (void) printf("Found vdev type: %s\n",
2054 vd->vdev_ops->vdev_op_type);
2055 }
2056
2057 buf = umem_alloc(size, UMEM_NOFAIL);
2058
2059 zio_flags = ZIO_FLAG_DONT_CACHE | ZIO_FLAG_DONT_QUEUE |
2060 ZIO_FLAG_DONT_PROPAGATE | ZIO_FLAG_DONT_RETRY;
2061
2062 spa_config_enter(spa, SCL_STATE, FTAG, RW_READER);
2063 zio = zio_root(spa, NULL, NULL, 0);
2064 /* XXX todo - cons up a BP so RAID-Z will be happy */
2065 zio_nowait(zio_vdev_child_io(zio, NULL, vd, offset, buf, size,
2066 ZIO_TYPE_READ, ZIO_PRIORITY_SYNC_READ, zio_flags, NULL, NULL));
2067 error = zio_wait(zio);
2068 spa_config_exit(spa, SCL_STATE, FTAG);
2069
2070 if (error) {
2071 (void) printf("Read of %s failed, error: %d\n", thing, error);
2072 goto out;
2073 }
2074
2075 if (flags & ZDB_FLAG_PRINT_BLKPTR)
2076 zdb_print_blkptr((blkptr_t *)(void *)
2077 ((uintptr_t)buf + (uintptr_t)blkptr_offset), flags);
2078 else if (flags & ZDB_FLAG_RAW)
2079 zdb_dump_block_raw(buf, size, flags);
2080 else if (flags & ZDB_FLAG_INDIRECT)
2081 zdb_dump_indirect((blkptr_t *)buf, size / sizeof (blkptr_t),
2082 flags);
2083 else if (flags & ZDB_FLAG_GBH)
2084 zdb_dump_gbh(buf, flags);
2085 else
2086 zdb_dump_block(thing, buf, size, flags);
2087
2088 out:
2089 umem_free(buf, size);
2090 free(dup);
2091 }
2092
2093 static boolean_t
2094 nvlist_string_match(nvlist_t *config, char *name, char *tgt)
2095 {
2096 char *s;
2097
2098 if (nvlist_lookup_string(config, name, &s) != 0)
2099 return (B_FALSE);
2100
2101 return (strcmp(s, tgt) == 0);
2102 }
2103
2104 static boolean_t
2105 nvlist_uint64_match(nvlist_t *config, char *name, uint64_t tgt)
2106 {
2107 uint64_t val;
2108
2109 if (nvlist_lookup_uint64(config, name, &val) != 0)
2110 return (B_FALSE);
2111
2112 return (val == tgt);
2113 }
2114
2115 static boolean_t
2116 vdev_child_guid_match(nvlist_t *vdev, uint64_t guid)
2117 {
2118 nvlist_t **child;
2119 uint_t c, children;
2120
2121 verify(nvlist_lookup_nvlist_array(vdev, ZPOOL_CONFIG_CHILDREN,
2122 &child, &children) == 0);
2123 for (c = 0; c < children; ++c)
2124 if (nvlist_uint64_match(child[c], ZPOOL_CONFIG_GUID, guid))
2125 return (B_TRUE);
2126 return (B_FALSE);
2127 }
2128
2129 static boolean_t
2130 vdev_child_string_match(nvlist_t *vdev, char *tgt)
2131 {
2132 nvlist_t **child;
2133 uint_t c, children;
2134
2135 verify(nvlist_lookup_nvlist_array(vdev, ZPOOL_CONFIG_CHILDREN,
2136 &child, &children) == 0);
2137 for (c = 0; c < children; ++c) {
2138 if (nvlist_string_match(child[c], ZPOOL_CONFIG_PATH, tgt) ||
2139 nvlist_string_match(child[c], ZPOOL_CONFIG_DEVID, tgt))
2140 return (B_TRUE);
2141 }
2142 return (B_FALSE);
2143 }
2144
2145 static boolean_t
2146 vdev_guid_match(nvlist_t *config, uint64_t guid)
2147 {
2148 nvlist_t *nvroot;
2149
2150 verify(nvlist_lookup_nvlist(config, ZPOOL_CONFIG_VDEV_TREE,
2151 &nvroot) == 0);
2152
2153 return (nvlist_uint64_match(nvroot, ZPOOL_CONFIG_GUID, guid) ||
2154 vdev_child_guid_match(nvroot, guid));
2155 }
2156
2157 static boolean_t
2158 vdev_string_match(nvlist_t *config, char *tgt)
2159 {
2160 nvlist_t *nvroot;
2161
2162 verify(nvlist_lookup_nvlist(config, ZPOOL_CONFIG_VDEV_TREE,
2163 &nvroot) == 0);
2164
2165 return (vdev_child_string_match(nvroot, tgt));
2166 }
2167
2168 static boolean_t
2169 pool_match(nvlist_t *config, char *tgt)
2170 {
2171 uint64_t guid = strtoull(tgt, NULL, 0);
2172
2173 if (guid != 0) {
2174 return (
2175 nvlist_uint64_match(config, ZPOOL_CONFIG_POOL_GUID, guid) ||
2176 vdev_guid_match(config, guid));
2177 } else {
2178 return (
2179 nvlist_string_match(config, ZPOOL_CONFIG_POOL_NAME, tgt) ||
2180 vdev_string_match(config, tgt));
2181 }
2182 }
2183
2184 static int
2185 find_exported_zpool(char *pool_id, nvlist_t **configp, char *vdev_dir)
2186 {
2187 nvlist_t *pools;
2188 int error = ENOENT;
2189 nvlist_t *match = NULL;
2190
2191 if (vdev_dir != NULL)
2192 pools = zpool_find_import_activeok(g_zfs, 1, &vdev_dir);
2193 else
2194 pools = zpool_find_import_activeok(g_zfs, 0, NULL);
2195
2196 if (pools != NULL) {
2197 nvpair_t *elem = NULL;
2198
2199 while ((elem = nvlist_next_nvpair(pools, elem)) != NULL) {
2200 verify(nvpair_value_nvlist(elem, configp) == 0);
2201 if (pool_match(*configp, pool_id)) {
2202 if (match != NULL) {
2203 (void) fatal(
2204 "More than one matching pool - "
2205 "specify guid/devid/device path.");
2206 } else {
2207 match = *configp;
2208 error = 0;
2209 }
2210 }
2211 }
2212 }
2213
2214 *configp = error ? NULL : match;
2215
2216 return (error);
2217 }
2218
2219 int
2220 main(int argc, char **argv)
2221 {
2222 int i, c;
2223 struct rlimit rl = { 1024, 1024 };
2224 spa_t *spa;
2225 objset_t *os = NULL;
2226 char *endstr;
2227 int dump_all = 1;
2228 int verbose = 0;
2229 int error;
2230 int exported = 0;
2231 char *vdev_dir = NULL;
2232
2233 (void) setrlimit(RLIMIT_NOFILE, &rl);
2234 (void) enable_extended_FILE_stdio(-1, -1);
2235
2236 dprintf_setup(&argc, argv);
2237
2238 while ((c = getopt(argc, argv, "udibcsvCS:U:lRep:")) != -1) {
2239 switch (c) {
2240 case 'u':
2241 case 'd':
2242 case 'i':
2243 case 'b':
2244 case 'c':
2245 case 's':
2246 case 'C':
2247 case 'l':
2248 case 'R':
2249 dump_opt[c]++;
2250 dump_all = 0;
2251 break;
2252 case 'v':
2253 verbose++;
2254 break;
2255 case 'U':
2256 spa_config_path = optarg;
2257 break;
2258 case 'e':
2259 exported = 1;
2260 break;
2261 case 'p':
2262 vdev_dir = optarg;
2263 break;
2264 case 'S':
2265 dump_opt[c]++;
2266 dump_all = 0;
2267 zdb_sig_user_data = (strncmp(optarg, "user:", 5) == 0);
2268 if (!zdb_sig_user_data && strncmp(optarg, "all:", 4))
2269 usage();
2270 endstr = strchr(optarg, ':') + 1;
2271 if (strcmp(endstr, "fletcher2") == 0)
2272 zdb_sig_cksumalg = ZIO_CHECKSUM_FLETCHER_2;
2273 else if (strcmp(endstr, "fletcher4") == 0)
2274 zdb_sig_cksumalg = ZIO_CHECKSUM_FLETCHER_4;
2275 else if (strcmp(endstr, "sha256") == 0)
2276 zdb_sig_cksumalg = ZIO_CHECKSUM_SHA256;
2277 else if (strcmp(endstr, "all") == 0)
2278 zdb_sig_cksumalg = ZIO_CHECKSUM_FLETCHER_2;
2279 else
2280 usage();
2281 break;
2282 default:
2283 usage();
2284 break;
2285 }
2286 }
2287
2288 if (vdev_dir != NULL && exported == 0) {
2289 (void) fprintf(stderr, "-p option requires use of -e\n");
2290 usage();
2291 }
2292
2293 kernel_init(FREAD);
2294 g_zfs = libzfs_init();
2295 ASSERT(g_zfs != NULL);
2296
2297 for (c = 0; c < 256; c++) {
2298 if (dump_all && c != 'l' && c != 'R')
2299 dump_opt[c] = 1;
2300 if (dump_opt[c])
2301 dump_opt[c] += verbose;
2302 }
2303
2304 argc -= optind;
2305 argv += optind;
2306
2307 if (argc < 1) {
2308 if (dump_opt['C']) {
2309 dump_cachefile(spa_config_path);
2310 return (0);
2311 }
2312 usage();
2313 }
2314
2315 if (dump_opt['l']) {
2316 dump_label(argv[0]);
2317 return (0);
2318 }
2319
2320 if (dump_opt['R']) {
2321 flagbits['b'] = ZDB_FLAG_PRINT_BLKPTR;
2322 flagbits['c'] = ZDB_FLAG_CHECKSUM;
2323 flagbits['d'] = ZDB_FLAG_DECOMPRESS;
2324 flagbits['e'] = ZDB_FLAG_BSWAP;
2325 flagbits['g'] = ZDB_FLAG_GBH;
2326 flagbits['i'] = ZDB_FLAG_INDIRECT;
2327 flagbits['p'] = ZDB_FLAG_PHYS;
2328 flagbits['r'] = ZDB_FLAG_RAW;
2329
2330 spa = NULL;
2331 while (argv[0]) {
2332 zdb_read_block(argv[0], &spa);
2333 argv++;
2334 argc--;
2335 }
2336 if (spa)
2337 spa_close(spa, (void *)zdb_read_block);
2338 return (0);
2339 }
2340
2341 if (dump_opt['C'])
2342 dump_config(argv[0]);
2343
2344 error = 0;
2345 if (exported) {
2346 /*
2347 * Check to see if the name refers to an exported zpool
2348 */
2349 char *slash;
2350 nvlist_t *exported_conf = NULL;
2351
2352 if ((slash = strchr(argv[0], '/')) != NULL)
2353 *slash = '\0';
2354
2355 error = find_exported_zpool(argv[0], &exported_conf, vdev_dir);
2356 if (error == 0) {
2357 nvlist_t *nvl = NULL;
2358
2359 if (vdev_dir != NULL) {
2360 if (nvlist_alloc(&nvl, NV_UNIQUE_NAME, 0) != 0)
2361 error = ENOMEM;
2362 else if (nvlist_add_string(nvl,
2363 zpool_prop_to_name(ZPOOL_PROP_ALTROOT),
2364 vdev_dir) != 0)
2365 error = ENOMEM;
2366 }
2367
2368 if (error == 0)
2369 error = spa_import_faulted(argv[0],
2370 exported_conf, nvl);
2371
2372 nvlist_free(nvl);
2373 }
2374
2375 if (slash != NULL)
2376 *slash = '/';
2377 }
2378
2379 if (error == 0) {
2380 if (strchr(argv[0], '/') != NULL) {
2381 error = dmu_objset_open(argv[0], DMU_OST_ANY,
2382 DS_MODE_USER | DS_MODE_READONLY, &os);
2383 } else {
2384 error = spa_open(argv[0], &spa, FTAG);
2385 }
2386 }
2387
2388 if (error)
2389 fatal("can't open %s: %s", argv[0], strerror(error));
2390
2391 argv++;
2392 if (--argc > 0) {
2393 zopt_objects = argc;
2394 zopt_object = calloc(zopt_objects, sizeof (uint64_t));
2395 for (i = 0; i < zopt_objects; i++) {
2396 errno = 0;
2397 zopt_object[i] = strtoull(argv[i], NULL, 0);
2398 if (zopt_object[i] == 0 && errno != 0)
2399 fatal("bad object number %s: %s",
2400 argv[i], strerror(errno));
2401 }
2402 }
2403
2404 if (os != NULL) {
2405 dump_dir(os);
2406 dmu_objset_close(os);
2407 } else {
2408 dump_zpool(spa);
2409 spa_close(spa, FTAG);
2410 }
2411
2412 fuid_table_destroy();
2413
2414 libzfs_fini(g_zfs);
2415 kernel_fini();
2416
2417 return (0);
2418 }
ZFS On Linux mirror for PVE