?  search:
summary | log | graphiclog1 | graphiclog2 | commit | commitdiff | tree | refs | edit | fork
blame | history | raw | HEAD
loader: updates from review
[unleashed.git] / usr / src / boot / sys / boot / zfs / zfsimpl.c
blobd015617f3e7e98d552e8132f2e92857ff1156686
1 /*-
2 * Copyright (c) 2007 Doug Rabson
3 * All rights reserved.
4 *
5 * Redistribution and use in source and binary forms, with or without
6 * modification, are permitted provided that the following conditions
7 * are met:
8 * 1. Redistributions of source code must retain the above copyright
9 * notice, this list of conditions and the following disclaimer.
10 * 2. Redistributions in binary form must reproduce the above copyright
11 * notice, this list of conditions and the following disclaimer in the
12 * documentation and/or other materials provided with the distribution.
13 *
14 * THIS SOFTWARE IS PROVIDED BY THE AUTHOR AND CONTRIBUTORS ``AS IS'' AND
15 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
16 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
17 * ARE DISCLAIMED. IN NO EVENT SHALL THE AUTHOR OR CONTRIBUTORS BE LIABLE
18 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
19 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
20 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
21 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
22 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
23 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
24 * SUCH DAMAGE.
25 */
26
27 #include <sys/cdefs.h>
28
29 /*
30 * Stand-alone ZFS file reader.
31 */
32
33 #include <sys/stat.h>
34 #include <sys/stdint.h>
35
36 #include "zfsimpl.h"
37 #include "zfssubr.c"
38
39
40 struct zfsmount {
41 const spa_t *spa;
42 objset_phys_t objset;
43 uint64_t rootobj;
44 };
45
46 /*
47 * List of all vdevs, chained through v_alllink.
48 */
49 static vdev_list_t zfs_vdevs;
50
51 /*
52 * List of ZFS features supported for read
53 */
54 static const char *features_for_read[] = {
55 "org.illumos:lz4_compress",
56 "com.delphix:hole_birth",
57 "com.delphix:extensible_dataset",
58 "com.delphix:embedded_data",
59 "org.open-zfs:large_blocks",
60 "org.illumos:sha512",
61 NULL
62 };
63
64 /*
65 * List of all pools, chained through spa_link.
66 */
67 static spa_list_t zfs_pools;
68
69 static uint64_t zfs_crc64_table[256];
70 static const dnode_phys_t *dnode_cache_obj = 0;
71 static uint64_t dnode_cache_bn;
72 static char *dnode_cache_buf;
73 static char *zap_scratch;
74 static char *zfs_temp_buf, *zfs_temp_end, *zfs_temp_ptr;
75
76 #define TEMP_SIZE (1024 * 1024)
77
78 static int zio_read(const spa_t *spa, const blkptr_t *bp, void *buf);
79 static int zfs_get_root(const spa_t *spa, uint64_t *objid);
80 static int zfs_rlookup(const spa_t *spa, uint64_t objnum, char *result);
81
82 static void
83 zfs_init(void)
84 {
85 STAILQ_INIT(&zfs_vdevs);
86 STAILQ_INIT(&zfs_pools);
87
88 zfs_temp_buf = malloc(TEMP_SIZE);
89 zfs_temp_end = zfs_temp_buf + TEMP_SIZE;
90 zfs_temp_ptr = zfs_temp_buf;
91 dnode_cache_buf = malloc(SPA_MAXBLOCKSIZE);
92 zap_scratch = malloc(SPA_MAXBLOCKSIZE);
93
94 zfs_init_crc();
95 }
96
97 static void *
98 zfs_alloc(size_t size)
99 {
100 char *ptr;
101
102 if (zfs_temp_ptr + size > zfs_temp_end) {
103 printf("ZFS: out of temporary buffer space\n");
104 for (;;) ;
105 }
106 ptr = zfs_temp_ptr;
107 zfs_temp_ptr += size;
108
109 return (ptr);
110 }
111
112 static void
113 zfs_free(void *ptr, size_t size)
114 {
115
116 zfs_temp_ptr -= size;
117 if (zfs_temp_ptr != ptr) {
118 printf("ZFS: zfs_alloc()/zfs_free() mismatch\n");
119 for (;;) ;
120 }
121 }
122
123 static int
124 xdr_int(const unsigned char **xdr, int *ip)
125 {
126 *ip = ((*xdr)[0] << 24)
127 | ((*xdr)[1] << 16)
128 | ((*xdr)[2] << 8)
129 | ((*xdr)[3] << 0);
130 (*xdr) += 4;
131 return (0);
132 }
133
134 static int
135 xdr_u_int(const unsigned char **xdr, u_int *ip)
136 {
137 *ip = ((*xdr)[0] << 24)
138 | ((*xdr)[1] << 16)
139 | ((*xdr)[2] << 8)
140 | ((*xdr)[3] << 0);
141 (*xdr) += 4;
142 return (0);
143 }
144
145 static int
146 xdr_uint64_t(const unsigned char **xdr, uint64_t *lp)
147 {
148 u_int hi, lo;
149
150 xdr_u_int(xdr, &hi);
151 xdr_u_int(xdr, &lo);
152 *lp = (((uint64_t) hi) << 32) | lo;
153 return (0);
154 }
155
156 static int
157 nvlist_find(const unsigned char *nvlist, const char *name, int type,
158 int* elementsp, void *valuep)
159 {
160 const unsigned char *p, *pair;
161 int junk;
162 int encoded_size, decoded_size;
163
164 p = nvlist;
165 xdr_int(&p, &junk);
166 xdr_int(&p, &junk);
167
168 pair = p;
169 xdr_int(&p, &encoded_size);
170 xdr_int(&p, &decoded_size);
171 while (encoded_size && decoded_size) {
172 int namelen, pairtype, elements;
173 const char *pairname;
174
175 xdr_int(&p, &namelen);
176 pairname = (const char*) p;
177 p += roundup(namelen, 4);
178 xdr_int(&p, &pairtype);
179
180 if (!memcmp(name, pairname, namelen) && type == pairtype) {
181 xdr_int(&p, &elements);
182 if (elementsp)
183 *elementsp = elements;
184 if (type == DATA_TYPE_UINT64) {
185 xdr_uint64_t(&p, (uint64_t *) valuep);
186 return (0);
187 } else if (type == DATA_TYPE_STRING) {
188 int len;
189 xdr_int(&p, &len);
190 (*(const char**) valuep) = (const char*) p;
191 return (0);
192 } else if (type == DATA_TYPE_NVLIST
193 || type == DATA_TYPE_NVLIST_ARRAY) {
194 (*(const unsigned char**) valuep) =
195 (const unsigned char*) p;
196 return (0);
197 } else {
198 return (EIO);
199 }
200 } else {
201 /*
202 * Not the pair we are looking for, skip to the next one.
203 */
204 p = pair + encoded_size;
205 }
206
207 pair = p;
208 xdr_int(&p, &encoded_size);
209 xdr_int(&p, &decoded_size);
210 }
211
212 return (EIO);
213 }
214
215 static int
216 nvlist_check_features_for_read(const unsigned char *nvlist)
217 {
218 const unsigned char *p, *pair;
219 int junk;
220 int encoded_size, decoded_size;
221 int rc;
222
223 rc = 0;
224
225 p = nvlist;
226 xdr_int(&p, &junk);
227 xdr_int(&p, &junk);
228
229 pair = p;
230 xdr_int(&p, &encoded_size);
231 xdr_int(&p, &decoded_size);
232 while (encoded_size && decoded_size) {
233 int namelen, pairtype;
234 const char *pairname;
235 int i, found;
236
237 found = 0;
238
239 xdr_int(&p, &namelen);
240 pairname = (const char*) p;
241 p += roundup(namelen, 4);
242 xdr_int(&p, &pairtype);
243
244 for (i = 0; features_for_read[i] != NULL; i++) {
245 if (!memcmp(pairname, features_for_read[i], namelen)) {
246 found = 1;
247 break;
248 }
249 }
250
251 if (!found) {
252 printf("ZFS: unsupported feature: %s\n", pairname);
253 rc = EIO;
254 }
255
256 p = pair + encoded_size;
257
258 pair = p;
259 xdr_int(&p, &encoded_size);
260 xdr_int(&p, &decoded_size);
261 }
262
263 return (rc);
264 }
265
266 /*
267 * Return the next nvlist in an nvlist array.
268 */
269 static const unsigned char *
270 nvlist_next(const unsigned char *nvlist)
271 {
272 const unsigned char *p, *pair;
273 int junk;
274 int encoded_size, decoded_size;
275
276 p = nvlist;
277 xdr_int(&p, &junk);
278 xdr_int(&p, &junk);
279
280 pair = p;
281 xdr_int(&p, &encoded_size);
282 xdr_int(&p, &decoded_size);
283 while (encoded_size && decoded_size) {
284 p = pair + encoded_size;
285
286 pair = p;
287 xdr_int(&p, &encoded_size);
288 xdr_int(&p, &decoded_size);
289 }
290
291 return p;
292 }
293
294 #ifdef TEST
295
296 static const unsigned char *
297 nvlist_print(const unsigned char *nvlist, unsigned int indent)
298 {
299 static const char* typenames[] = {
300 "DATA_TYPE_UNKNOWN",
301 "DATA_TYPE_BOOLEAN",
302 "DATA_TYPE_BYTE",
303 "DATA_TYPE_INT16",
304 "DATA_TYPE_UINT16",
305 "DATA_TYPE_INT32",
306 "DATA_TYPE_UINT32",
307 "DATA_TYPE_INT64",
308 "DATA_TYPE_UINT64",
309 "DATA_TYPE_STRING",
310 "DATA_TYPE_BYTE_ARRAY",
311 "DATA_TYPE_INT16_ARRAY",
312 "DATA_TYPE_UINT16_ARRAY",
313 "DATA_TYPE_INT32_ARRAY",
314 "DATA_TYPE_UINT32_ARRAY",
315 "DATA_TYPE_INT64_ARRAY",
316 "DATA_TYPE_UINT64_ARRAY",
317 "DATA_TYPE_STRING_ARRAY",
318 "DATA_TYPE_HRTIME",
319 "DATA_TYPE_NVLIST",
320 "DATA_TYPE_NVLIST_ARRAY",
321 "DATA_TYPE_BOOLEAN_VALUE",
322 "DATA_TYPE_INT8",
323 "DATA_TYPE_UINT8",
324 "DATA_TYPE_BOOLEAN_ARRAY",
325 "DATA_TYPE_INT8_ARRAY",
326 "DATA_TYPE_UINT8_ARRAY"
327 };
328
329 unsigned int i, j;
330 const unsigned char *p, *pair;
331 int junk;
332 int encoded_size, decoded_size;
333
334 p = nvlist;
335 xdr_int(&p, &junk);
336 xdr_int(&p, &junk);
337
338 pair = p;
339 xdr_int(&p, &encoded_size);
340 xdr_int(&p, &decoded_size);
341 while (encoded_size && decoded_size) {
342 int namelen, pairtype, elements;
343 const char *pairname;
344
345 xdr_int(&p, &namelen);
346 pairname = (const char*) p;
347 p += roundup(namelen, 4);
348 xdr_int(&p, &pairtype);
349
350 for (i = 0; i < indent; i++)
351 printf(" ");
352 printf("%s %s", typenames[pairtype], pairname);
353
354 xdr_int(&p, &elements);
355 switch (pairtype) {
356 case DATA_TYPE_UINT64: {
357 uint64_t val;
358 xdr_uint64_t(&p, &val);
359 printf(" = 0x%jx\n", (uintmax_t)val);
360 break;
361 }
362
363 case DATA_TYPE_STRING: {
364 int len;
365 xdr_int(&p, &len);
366 printf(" = \"%s\"\n", p);
367 break;
368 }
369
370 case DATA_TYPE_NVLIST:
371 printf("\n");
372 nvlist_print(p, indent + 1);
373 break;
374
375 case DATA_TYPE_NVLIST_ARRAY:
376 for (j = 0; j < elements; j++) {
377 printf("[%d]\n", j);
378 p = nvlist_print(p, indent + 1);
379 if (j != elements - 1) {
380 for (i = 0; i < indent; i++)
381 printf(" ");
382 printf("%s %s", typenames[pairtype], pairname);
383 }
384 }
385 break;
386
387 default:
388 printf("\n");
389 }
390
391 p = pair + encoded_size;
392
393 pair = p;
394 xdr_int(&p, &encoded_size);
395 xdr_int(&p, &decoded_size);
396 }
397
398 return p;
399 }
400
401 #endif
402
403 static int
404 vdev_read_phys(vdev_t *vdev, const blkptr_t *bp, void *buf,
405 off_t offset, size_t size)
406 {
407 size_t psize;
408 int rc;
409
410 if (!vdev->v_phys_read)
411 return (EIO);
412
413 if (bp) {
414 psize = BP_GET_PSIZE(bp);
415 } else {
416 psize = size;
417 }
418
419 /*printf("ZFS: reading %d bytes at 0x%jx to %p\n", psize, (uintmax_t)offset, buf);*/
420 rc = vdev->v_phys_read(vdev, vdev->v_read_priv, offset, buf, psize);
421 if (rc)
422 return (rc);
423 if (bp && zio_checksum_verify(bp, buf))
424 return (EIO);
425
426 return (0);
427 }
428
429 static int
430 vdev_disk_read(vdev_t *vdev, const blkptr_t *bp, void *buf,
431 off_t offset, size_t bytes)
432 {
433
434 return (vdev_read_phys(vdev, bp, buf,
435 offset + VDEV_LABEL_START_SIZE, bytes));
436 }
437
438
439 static int
440 vdev_mirror_read(vdev_t *vdev, const blkptr_t *bp, void *buf,
441 off_t offset, size_t bytes)
442 {
443 vdev_t *kid;
444 int rc;
445
446 rc = EIO;
447 STAILQ_FOREACH(kid, &vdev->v_children, v_childlink) {
448 if (kid->v_state != VDEV_STATE_HEALTHY)
449 continue;
450 rc = kid->v_read(kid, bp, buf, offset, bytes);
451 if (!rc)
452 return (0);
453 }
454
455 return (rc);
456 }
457
458 static int
459 vdev_replacing_read(vdev_t *vdev, const blkptr_t *bp, void *buf,
460 off_t offset, size_t bytes)
461 {
462 vdev_t *kid;
463
464 /*
465 * Here we should have two kids:
466 * First one which is the one we are replacing and we can trust
467 * only this one to have valid data, but it might not be present.
468 * Second one is that one we are replacing with. It is most likely
469 * healthy, but we can't trust it has needed data, so we won't use it.
470 */
471 kid = STAILQ_FIRST(&vdev->v_children);
472 if (kid == NULL)
473 return (EIO);
474 if (kid->v_state != VDEV_STATE_HEALTHY)
475 return (EIO);
476 return (kid->v_read(kid, bp, buf, offset, bytes));
477 }
478
479 static vdev_t *
480 vdev_find(uint64_t guid)
481 {
482 vdev_t *vdev;
483
484 STAILQ_FOREACH(vdev, &zfs_vdevs, v_alllink)
485 if (vdev->v_guid == guid)
486 return (vdev);
487
488 return (0);
489 }
490
491 static vdev_t *
492 vdev_create(uint64_t guid, vdev_read_t *vdev_read)
493 {
494 vdev_t *vdev;
495
496 vdev = malloc(sizeof(vdev_t));
497 memset(vdev, 0, sizeof(vdev_t));
498 STAILQ_INIT(&vdev->v_children);
499 vdev->v_guid = guid;
500 vdev->v_state = VDEV_STATE_OFFLINE;
501 vdev->v_read = vdev_read;
502 vdev->v_phys_read = 0;
503 vdev->v_read_priv = 0;
504 STAILQ_INSERT_TAIL(&zfs_vdevs, vdev, v_alllink);
505
506 return (vdev);
507 }
508
509 static int
510 vdev_init_from_nvlist(const unsigned char *nvlist, vdev_t *pvdev,
511 vdev_t **vdevp, int is_newer)
512 {
513 int rc;
514 uint64_t guid, id, ashift, nparity;
515 const char *type;
516 const char *path;
517 vdev_t *vdev, *kid;
518 const unsigned char *kids;
519 int nkids, i, is_new;
520 uint64_t is_offline, is_faulted, is_degraded, is_removed, isnt_present;
521
522 if (nvlist_find(nvlist, ZPOOL_CONFIG_GUID,
523 DATA_TYPE_UINT64, 0, &guid)
524 || nvlist_find(nvlist, ZPOOL_CONFIG_ID,
525 DATA_TYPE_UINT64, 0, &id)
526 || nvlist_find(nvlist, ZPOOL_CONFIG_TYPE,
527 DATA_TYPE_STRING, 0, &type)) {
528 printf("ZFS: can't find vdev details\n");
529 return (ENOENT);
530 }
531
532 if (strcmp(type, VDEV_TYPE_MIRROR)
533 && strcmp(type, VDEV_TYPE_DISK)
534 #ifdef ZFS_TEST
535 && strcmp(type, VDEV_TYPE_FILE)
536 #endif
537 && strcmp(type, VDEV_TYPE_RAIDZ)
538 && strcmp(type, VDEV_TYPE_REPLACING)) {
539 printf("ZFS: can only boot from disk, mirror, raidz1, raidz2 and raidz3 vdevs\n");
540 return (EIO);
541 }
542
543 is_offline = is_removed = is_faulted = is_degraded = isnt_present = 0;
544
545 nvlist_find(nvlist, ZPOOL_CONFIG_OFFLINE, DATA_TYPE_UINT64, 0,
546 &is_offline);
547 nvlist_find(nvlist, ZPOOL_CONFIG_REMOVED, DATA_TYPE_UINT64, 0,
548 &is_removed);
549 nvlist_find(nvlist, ZPOOL_CONFIG_FAULTED, DATA_TYPE_UINT64, 0,
550 &is_faulted);
551 nvlist_find(nvlist, ZPOOL_CONFIG_DEGRADED, DATA_TYPE_UINT64, 0,
552 &is_degraded);
553 nvlist_find(nvlist, ZPOOL_CONFIG_NOT_PRESENT, DATA_TYPE_UINT64, 0,
554 &isnt_present);
555
556 vdev = vdev_find(guid);
557 if (!vdev) {
558 is_new = 1;
559
560 if (!strcmp(type, VDEV_TYPE_MIRROR))
561 vdev = vdev_create(guid, vdev_mirror_read);
562 else if (!strcmp(type, VDEV_TYPE_RAIDZ))
563 vdev = vdev_create(guid, vdev_raidz_read);
564 else if (!strcmp(type, VDEV_TYPE_REPLACING))
565 vdev = vdev_create(guid, vdev_replacing_read);
566 else
567 vdev = vdev_create(guid, vdev_disk_read);
568
569 vdev->v_id = id;
570 vdev->v_top = pvdev != NULL ? pvdev : vdev;
571 if (nvlist_find(nvlist, ZPOOL_CONFIG_ASHIFT,
572 DATA_TYPE_UINT64, 0, &ashift) == 0)
573 vdev->v_ashift = ashift;
574 else
575 vdev->v_ashift = 0;
576 if (nvlist_find(nvlist, ZPOOL_CONFIG_NPARITY,
577 DATA_TYPE_UINT64, 0, &nparity) == 0)
578 vdev->v_nparity = nparity;
579 else
580 vdev->v_nparity = 0;
581 if (nvlist_find(nvlist, ZPOOL_CONFIG_PATH,
582 DATA_TYPE_STRING, 0, &path) == 0) {
583 if (strncmp(path, "/dev/dsk/", 9) == 0)
584 path += 9;
585 vdev->v_name = strdup(path);
586 if (nvlist_find(nvlist, ZPOOL_CONFIG_PHYS_PATH,
587 DATA_TYPE_STRING, 0, &path) == 0)
588 vdev->v_phys_path = strdup(path);
589 else
590 vdev->v_phys_path = NULL;
591 if (nvlist_find(nvlist, ZPOOL_CONFIG_DEVID,
592 DATA_TYPE_STRING, 0, &path) == 0)
593 vdev->v_devid = strdup(path);
594 else
595 vdev->v_devid = NULL;
596 } else {
597 if (!strcmp(type, "raidz")) {
598 if (vdev->v_nparity == 1)
599 vdev->v_name = "raidz1";
600 else if (vdev->v_nparity == 2)
601 vdev->v_name = "raidz2";
602 else if (vdev->v_nparity == 3)
603 vdev->v_name = "raidz3";
604 else {
605 printf("ZFS: can only boot from disk, mirror, raidz1, raidz2 and raidz3 vdevs\n");
606 return (EIO);
607 }
608 } else {
609 vdev->v_name = strdup(type);
610 }
611 }
612 } else {
613 is_new = 0;
614 }
615
616 if (is_new || is_newer) {
617 /*
618 * This is either new vdev or we've already seen this vdev,
619 * but from an older vdev label, so let's refresh its state
620 * from the newer label.
621 */
622 if (is_offline)
623 vdev->v_state = VDEV_STATE_OFFLINE;
624 else if (is_removed)
625 vdev->v_state = VDEV_STATE_REMOVED;
626 else if (is_faulted)
627 vdev->v_state = VDEV_STATE_FAULTED;
628 else if (is_degraded)
629 vdev->v_state = VDEV_STATE_DEGRADED;
630 else if (isnt_present)
631 vdev->v_state = VDEV_STATE_CANT_OPEN;
632 }
633
634 rc = nvlist_find(nvlist, ZPOOL_CONFIG_CHILDREN,
635 DATA_TYPE_NVLIST_ARRAY, &nkids, &kids);
636 /*
637 * Its ok if we don't have any kids.
638 */
639 if (rc == 0) {
640 vdev->v_nchildren = nkids;
641 for (i = 0; i < nkids; i++) {
642 rc = vdev_init_from_nvlist(kids, vdev, &kid, is_newer);
643 if (rc)
644 return (rc);
645 if (is_new)
646 STAILQ_INSERT_TAIL(&vdev->v_children, kid,
647 v_childlink);
648 kids = nvlist_next(kids);
649 }
650 } else {
651 vdev->v_nchildren = 0;
652 }
653
654 if (vdevp)
655 *vdevp = vdev;
656 return (0);
657 }
658
659 static void
660 vdev_set_state(vdev_t *vdev)
661 {
662 vdev_t *kid;
663 int good_kids;
664 int bad_kids;
665
666 /*
667 * A mirror or raidz is healthy if all its kids are healthy. A
668 * mirror is degraded if any of its kids is healthy; a raidz
669 * is degraded if at most nparity kids are offline.
670 */
671 if (STAILQ_FIRST(&vdev->v_children)) {
672 good_kids = 0;
673 bad_kids = 0;
674 STAILQ_FOREACH(kid, &vdev->v_children, v_childlink) {
675 if (kid->v_state == VDEV_STATE_HEALTHY)
676 good_kids++;
677 else
678 bad_kids++;
679 }
680 if (bad_kids == 0) {
681 vdev->v_state = VDEV_STATE_HEALTHY;
682 } else {
683 if (vdev->v_read == vdev_mirror_read) {
684 if (good_kids) {
685 vdev->v_state = VDEV_STATE_DEGRADED;
686 } else {
687 vdev->v_state = VDEV_STATE_OFFLINE;
688 }
689 } else if (vdev->v_read == vdev_raidz_read) {
690 if (bad_kids > vdev->v_nparity) {
691 vdev->v_state = VDEV_STATE_OFFLINE;
692 } else {
693 vdev->v_state = VDEV_STATE_DEGRADED;
694 }
695 }
696 }
697 }
698 }
699
700 static spa_t *
701 spa_find_by_guid(uint64_t guid)
702 {
703 spa_t *spa;
704
705 STAILQ_FOREACH(spa, &zfs_pools, spa_link)
706 if (spa->spa_guid == guid)
707 return (spa);
708
709 return (0);
710 }
711
712 static spa_t *
713 spa_find_by_name(const char *name)
714 {
715 spa_t *spa;
716
717 STAILQ_FOREACH(spa, &zfs_pools, spa_link)
718 if (!strcmp(spa->spa_name, name))
719 return (spa);
720
721 return (0);
722 }
723
724 spa_t *
725 spa_get_primary(void)
726 {
727 return (STAILQ_FIRST(&zfs_pools));
728 }
729
730 vdev_t *
731 spa_get_primary_vdev(const spa_t *spa)
732 {
733 vdev_t *vdev;
734 vdev_t *kid;
735
736 if (spa == NULL)
737 spa = spa_get_primary();
738 if (spa == NULL)
739 return (NULL);
740 vdev = STAILQ_FIRST(&spa->spa_vdevs);
741 if (vdev == NULL)
742 return (NULL);
743 for (kid = STAILQ_FIRST(&vdev->v_children); kid != NULL;
744 kid = STAILQ_FIRST(&vdev->v_children))
745 vdev = kid;
746 return (vdev);
747 }
748
749 static spa_t *
750 spa_create(uint64_t guid)
751 {
752 spa_t *spa;
753
754 spa = malloc(sizeof(spa_t));
755 memset(spa, 0, sizeof(spa_t));
756 STAILQ_INIT(&spa->spa_vdevs);
757 spa->spa_guid = guid;
758 STAILQ_INSERT_TAIL(&zfs_pools, spa, spa_link);
759
760 return (spa);
761 }
762
763 static const char *
764 state_name(vdev_state_t state)
765 {
766 static const char* names[] = {
767 "UNKNOWN",
768 "CLOSED",
769 "OFFLINE",
770 "REMOVED",
771 "CANT_OPEN",
772 "FAULTED",
773 "DEGRADED",
774 "ONLINE"
775 };
776 return names[state];
777 }
778
779 static int
780 pager_printf(const char *fmt, ...)
781 {
782 char line[80];
783 va_list args;
784
785 va_start(args, fmt);
786 vsnprintf(line, sizeof (line), fmt, args);
787 va_end(args);
788 return (pager_output(line));
789 }
790
791 #define STATUS_FORMAT " %s %s\n"
792
793 static int
794 print_state(int indent, const char *name, vdev_state_t state)
795 {
796 int i;
797 char buf[512];
798
799 buf[0] = 0;
800 for (i = 0; i < indent; i++)
801 strcat(buf, " ");
802 strcat(buf, name);
803 return (pager_printf(STATUS_FORMAT, buf, state_name(state)));
804 }
805
806 static int
807 vdev_status(vdev_t *vdev, int indent)
808 {
809 vdev_t *kid;
810 int ret;
811 ret = print_state(indent, vdev->v_name, vdev->v_state);
812 if (ret != 0)
813 return (ret);
814
815 STAILQ_FOREACH(kid, &vdev->v_children, v_childlink) {
816 ret = vdev_status(kid, indent + 1);
817 if (ret != 0)
818 return (ret);
819 }
820 return (ret);
821 }
822
823 static int
824 spa_status(spa_t *spa)
825 {
826 static char bootfs[ZFS_MAXNAMELEN];
827 uint64_t rootid;
828 vdev_t *vdev;
829 int good_kids, bad_kids, degraded_kids, ret;
830 vdev_state_t state;
831
832 ret = pager_printf(" pool: %s\n", spa->spa_name);
833 if (ret != 0)
834 return (ret);
835
836 if (zfs_get_root(spa, &rootid) == 0 &&
837 zfs_rlookup(spa, rootid, bootfs) == 0) {
838 if (bootfs[0] == '\0')
839 ret = pager_printf("bootfs: %s\n", spa->spa_name);
840 else
841 ret = pager_printf("bootfs: %s/%s\n", spa->spa_name,
842 bootfs);
843 if (ret != 0)
844 return (ret);
845 }
846 ret = pager_printf("config:\n\n");
847 if (ret != 0)
848 return (ret);
849 ret = pager_printf(STATUS_FORMAT, "NAME", "STATE");
850 if (ret != 0)
851 return (ret);
852
853 good_kids = 0;
854 degraded_kids = 0;
855 bad_kids = 0;
856 STAILQ_FOREACH(vdev, &spa->spa_vdevs, v_childlink) {
857 if (vdev->v_state == VDEV_STATE_HEALTHY)
858 good_kids++;
859 else if (vdev->v_state == VDEV_STATE_DEGRADED)
860 degraded_kids++;
861 else
862 bad_kids++;
863 }
864
865 state = VDEV_STATE_CLOSED;
866 if (good_kids > 0 && (degraded_kids + bad_kids) == 0)
867 state = VDEV_STATE_HEALTHY;
868 else if ((good_kids + degraded_kids) > 0)
869 state = VDEV_STATE_DEGRADED;
870
871 ret = print_state(0, spa->spa_name, state);
872 if (ret != 0)
873 return (ret);
874 STAILQ_FOREACH(vdev, &spa->spa_vdevs, v_childlink) {
875 ret = vdev_status(vdev, 1);
876 if (ret != 0)
877 return (ret);
878 }
879 return (ret);
880 }
881
882 int
883 spa_all_status(void)
884 {
885 spa_t *spa;
886 int first = 1, ret = 0;
887
888 STAILQ_FOREACH(spa, &zfs_pools, spa_link) {
889 if (!first) {
890 ret = pager_printf("\n");
891 if (ret != 0)
892 return (ret);
893 }
894 first = 0;
895 ret = spa_status(spa);
896 if (ret != 0)
897 return (ret);
898 }
899 return (ret);
900 }
901
902 static int
903 vdev_probe(vdev_phys_read_t *phys_read, void *read_priv, spa_t **spap)
904 {
905 vdev_t vtmp;
906 vdev_phys_t *vdev_label = (vdev_phys_t *) zap_scratch;
907 spa_t *spa;
908 vdev_t *vdev, *top_vdev, *pool_vdev;
909 off_t off;
910 blkptr_t bp;
911 const unsigned char *nvlist;
912 uint64_t val;
913 uint64_t guid;
914 uint64_t pool_txg, pool_guid;
915 uint64_t is_log;
916 const char *pool_name;
917 const unsigned char *vdevs;
918 const unsigned char *features;
919 int i, rc, is_newer;
920 char *upbuf;
921 const struct uberblock *up;
922
923 /*
924 * Load the vdev label and figure out which
925 * uberblock is most current.
926 */
927 memset(&vtmp, 0, sizeof(vtmp));
928 vtmp.v_phys_read = phys_read;
929 vtmp.v_read_priv = read_priv;
930 off = offsetof(vdev_label_t, vl_vdev_phys);
931 BP_ZERO(&bp);
932 BP_SET_LSIZE(&bp, sizeof(vdev_phys_t));
933 BP_SET_PSIZE(&bp, sizeof(vdev_phys_t));
934 BP_SET_CHECKSUM(&bp, ZIO_CHECKSUM_LABEL);
935 BP_SET_COMPRESS(&bp, ZIO_COMPRESS_OFF);
936 DVA_SET_OFFSET(BP_IDENTITY(&bp), off);
937 ZIO_SET_CHECKSUM(&bp.blk_cksum, off, 0, 0, 0);
938 if (vdev_read_phys(&vtmp, &bp, vdev_label, off, 0))
939 return (EIO);
940
941 if (vdev_label->vp_nvlist[0] != NV_ENCODE_XDR) {
942 return (EIO);
943 }
944
945 nvlist = (const unsigned char *) vdev_label->vp_nvlist + 4;
946
947 if (nvlist_find(nvlist,
948 ZPOOL_CONFIG_VERSION,
949 DATA_TYPE_UINT64, 0, &val)) {
950 return (EIO);
951 }
952
953 if (!SPA_VERSION_IS_SUPPORTED(val)) {
954 printf("ZFS: unsupported ZFS version %u (should be %u)\n",
955 (unsigned) val, (unsigned) SPA_VERSION);
956 return (EIO);
957 }
958
959 /* Check ZFS features for read */
960 if (nvlist_find(nvlist,
961 ZPOOL_CONFIG_FEATURES_FOR_READ,
962 DATA_TYPE_NVLIST, 0, &features) == 0
963 && nvlist_check_features_for_read(features) != 0)
964 return (EIO);
965
966 if (nvlist_find(nvlist,
967 ZPOOL_CONFIG_POOL_STATE,
968 DATA_TYPE_UINT64, 0, &val)) {
969 return (EIO);
970 }
971
972 if (val == POOL_STATE_DESTROYED) {
973 /* We don't boot only from destroyed pools. */
974 return (EIO);
975 }
976
977 if (nvlist_find(nvlist,
978 ZPOOL_CONFIG_POOL_TXG,
979 DATA_TYPE_UINT64, 0, &pool_txg)
980 || nvlist_find(nvlist,
981 ZPOOL_CONFIG_POOL_GUID,
982 DATA_TYPE_UINT64, 0, &pool_guid)
983 || nvlist_find(nvlist,
984 ZPOOL_CONFIG_POOL_NAME,
985 DATA_TYPE_STRING, 0, &pool_name)) {
986 /*
987 * Cache and spare devices end up here - just ignore
988 * them.
989 */
990 /*printf("ZFS: can't find pool details\n");*/
991 return (EIO);
992 }
993
994 is_log = 0;
995 (void) nvlist_find(nvlist, ZPOOL_CONFIG_IS_LOG, DATA_TYPE_UINT64, 0,
996 &is_log);
997 if (is_log)
998 return (EIO);
999
1000 /*
1001 * Create the pool if this is the first time we've seen it.
1002 */
1003 spa = spa_find_by_guid(pool_guid);
1004 if (!spa) {
1005 spa = spa_create(pool_guid);
1006 spa->spa_name = strdup(pool_name);
1007 }
1008 if (pool_txg > spa->spa_txg) {
1009 spa->spa_txg = pool_txg;
1010 is_newer = 1;
1011 } else
1012 is_newer = 0;
1013
1014 /*
1015 * Get the vdev tree and create our in-core copy of it.
1016 * If we already have a vdev with this guid, this must
1017 * be some kind of alias (overlapping slices, dangerously dedicated
1018 * disks etc).
1019 */
1020 if (nvlist_find(nvlist,
1021 ZPOOL_CONFIG_GUID,
1022 DATA_TYPE_UINT64, 0, &guid)) {
1023 return (EIO);
1024 }
1025 vdev = vdev_find(guid);
1026 if (vdev && vdev->v_phys_read) /* Has this vdev already been inited? */
1027 return (EIO);
1028
1029 if (nvlist_find(nvlist,
1030 ZPOOL_CONFIG_VDEV_TREE,
1031 DATA_TYPE_NVLIST, 0, &vdevs)) {
1032 return (EIO);
1033 }
1034
1035 rc = vdev_init_from_nvlist(vdevs, NULL, &top_vdev, is_newer);
1036 if (rc)
1037 return (rc);
1038
1039 /*
1040 * Add the toplevel vdev to the pool if its not already there.
1041 */
1042 STAILQ_FOREACH(pool_vdev, &spa->spa_vdevs, v_childlink)
1043 if (top_vdev == pool_vdev)
1044 break;
1045 if (!pool_vdev && top_vdev)
1046 STAILQ_INSERT_TAIL(&spa->spa_vdevs, top_vdev, v_childlink);
1047
1048 /*
1049 * We should already have created an incomplete vdev for this
1050 * vdev. Find it and initialise it with our read proc.
1051 */
1052 vdev = vdev_find(guid);
1053 if (vdev) {
1054 vdev->v_phys_read = phys_read;
1055 vdev->v_read_priv = read_priv;
1056 vdev->v_state = VDEV_STATE_HEALTHY;
1057 } else {
1058 printf("ZFS: inconsistent nvlist contents\n");
1059 return (EIO);
1060 }
1061
1062 /*
1063 * Re-evaluate top-level vdev state.
1064 */
1065 vdev_set_state(top_vdev);
1066
1067 /*
1068 * Ok, we are happy with the pool so far. Lets find
1069 * the best uberblock and then we can actually access
1070 * the contents of the pool.
1071 */
1072 upbuf = zfs_alloc(VDEV_UBERBLOCK_SIZE(vdev));
1073 up = (const struct uberblock *)upbuf;
1074 for (i = 0;
1075 i < VDEV_UBERBLOCK_COUNT(vdev);
1076 i++) {
1077 off = VDEV_UBERBLOCK_OFFSET(vdev, i);
1078 BP_ZERO(&bp);
1079 DVA_SET_OFFSET(&bp.blk_dva[0], off);
1080 BP_SET_LSIZE(&bp, VDEV_UBERBLOCK_SIZE(vdev));
1081 BP_SET_PSIZE(&bp, VDEV_UBERBLOCK_SIZE(vdev));
1082 BP_SET_CHECKSUM(&bp, ZIO_CHECKSUM_LABEL);
1083 BP_SET_COMPRESS(&bp, ZIO_COMPRESS_OFF);
1084 ZIO_SET_CHECKSUM(&bp.blk_cksum, off, 0, 0, 0);
1085
1086 if (vdev_read_phys(vdev, &bp, upbuf, off, 0))
1087 continue;
1088
1089 if (up->ub_magic != UBERBLOCK_MAGIC)
1090 continue;
1091 if (up->ub_txg < spa->spa_txg)
1092 continue;
1093 if (up->ub_txg > spa->spa_uberblock.ub_txg) {
1094 spa->spa_uberblock = *up;
1095 } else if (up->ub_txg == spa->spa_uberblock.ub_txg) {
1096 if (up->ub_timestamp > spa->spa_uberblock.ub_timestamp)
1097 spa->spa_uberblock = *up;
1098 }
1099 }
1100 zfs_free(upbuf, VDEV_UBERBLOCK_SIZE(vdev));
1101
1102 if (spap)
1103 *spap = spa;
1104 return (0);
1105 }
1106
1107 static int
1108 ilog2(int n)
1109 {
1110 int v;
1111
1112 for (v = 0; v < 32; v++)
1113 if (n == (1 << v))
1114 return v;
1115 return -1;
1116 }
1117
1118 static int
1119 zio_read_gang(const spa_t *spa, const blkptr_t *bp, void *buf)
1120 {
1121 blkptr_t gbh_bp;
1122 zio_gbh_phys_t zio_gb;
1123 char *pbuf;
1124 int i;
1125
1126 /* Artificial BP for gang block header. */
1127 gbh_bp = *bp;
1128 BP_SET_PSIZE(&gbh_bp, SPA_GANGBLOCKSIZE);
1129 BP_SET_LSIZE(&gbh_bp, SPA_GANGBLOCKSIZE);
1130 BP_SET_CHECKSUM(&gbh_bp, ZIO_CHECKSUM_GANG_HEADER);
1131 BP_SET_COMPRESS(&gbh_bp, ZIO_COMPRESS_OFF);
1132 for (i = 0; i < SPA_DVAS_PER_BP; i++)
1133 DVA_SET_GANG(&gbh_bp.blk_dva[i], 0);
1134
1135 /* Read gang header block using the artificial BP. */
1136 if (zio_read(spa, &gbh_bp, &zio_gb))
1137 return (EIO);
1138
1139 pbuf = buf;
1140 for (i = 0; i < SPA_GBH_NBLKPTRS; i++) {
1141 blkptr_t *gbp = &zio_gb.zg_blkptr[i];
1142
1143 if (BP_IS_HOLE(gbp))
1144 continue;
1145 if (zio_read(spa, gbp, pbuf))
1146 return (EIO);
1147 pbuf += BP_GET_PSIZE(gbp);
1148 }
1149
1150 if (zio_checksum_verify(bp, buf))
1151 return (EIO);
1152 return (0);
1153 }
1154
1155 static int
1156 zio_read(const spa_t *spa, const blkptr_t *bp, void *buf)
1157 {
1158 int cpfunc = BP_GET_COMPRESS(bp);
1159 uint64_t align, size;
1160 void *pbuf;
1161 int i, error;
1162
1163 /*
1164 * Process data embedded in block pointer
1165 */
1166 if (BP_IS_EMBEDDED(bp)) {
1167 ASSERT(BPE_GET_ETYPE(bp) == BP_EMBEDDED_TYPE_DATA);
1168
1169 size = BPE_GET_PSIZE(bp);
1170 ASSERT(size <= BPE_PAYLOAD_SIZE);
1171
1172 if (cpfunc != ZIO_COMPRESS_OFF)
1173 pbuf = zfs_alloc(size);
1174 else
1175 pbuf = buf;
1176
1177 decode_embedded_bp_compressed(bp, pbuf);
1178 error = 0;
1179
1180 if (cpfunc != ZIO_COMPRESS_OFF) {
1181 error = zio_decompress_data(cpfunc, pbuf,
1182 size, buf, BP_GET_LSIZE(bp));
1183 zfs_free(pbuf, size);
1184 }
1185 if (error != 0)
1186 printf("ZFS: i/o error - unable to decompress block pointer data, error %d\n",
1187 error);
1188 return (error);
1189 }
1190
1191 error = EIO;
1192
1193 for (i = 0; i < SPA_DVAS_PER_BP; i++) {
1194 const dva_t *dva = &bp->blk_dva[i];
1195 vdev_t *vdev;
1196 int vdevid;
1197 off_t offset;
1198
1199 if (!dva->dva_word[0] && !dva->dva_word[1])
1200 continue;
1201
1202 vdevid = DVA_GET_VDEV(dva);
1203 offset = DVA_GET_OFFSET(dva);
1204 STAILQ_FOREACH(vdev, &spa->spa_vdevs, v_childlink) {
1205 if (vdev->v_id == vdevid)
1206 break;
1207 }
1208 if (!vdev || !vdev->v_read)
1209 continue;
1210
1211 size = BP_GET_PSIZE(bp);
1212 if (vdev->v_read == vdev_raidz_read) {
1213 align = 1ULL << vdev->v_top->v_ashift;
1214 if (P2PHASE(size, align) != 0)
1215 size = P2ROUNDUP(size, align);
1216 }
1217 if (size != BP_GET_PSIZE(bp) || cpfunc != ZIO_COMPRESS_OFF)
1218 pbuf = zfs_alloc(size);
1219 else
1220 pbuf = buf;
1221
1222 if (DVA_GET_GANG(dva))
1223 error = zio_read_gang(spa, bp, pbuf);
1224 else
1225 error = vdev->v_read(vdev, bp, pbuf, offset, size);
1226 if (error == 0) {
1227 if (cpfunc != ZIO_COMPRESS_OFF)
1228 error = zio_decompress_data(cpfunc, pbuf,
1229 BP_GET_PSIZE(bp), buf, BP_GET_LSIZE(bp));
1230 else if (size != BP_GET_PSIZE(bp))
1231 bcopy(pbuf, buf, BP_GET_PSIZE(bp));
1232 }
1233 if (buf != pbuf)
1234 zfs_free(pbuf, size);
1235 if (error == 0)
1236 break;
1237 }
1238 if (error != 0)
1239 printf("ZFS: i/o error - all block copies unavailable\n");
1240 return (error);
1241 }
1242
1243 static int
1244 dnode_read(const spa_t *spa, const dnode_phys_t *dnode, off_t offset, void *buf, size_t buflen)
1245 {
1246 int ibshift = dnode->dn_indblkshift - SPA_BLKPTRSHIFT;
1247 int bsize = dnode->dn_datablkszsec << SPA_MINBLOCKSHIFT;
1248 int nlevels = dnode->dn_nlevels;
1249 int i, rc;
1250
1251 if (bsize > SPA_MAXBLOCKSIZE) {
1252 printf("ZFS: I/O error - blocks larger than %llu are not "
1253 "supported\n", SPA_MAXBLOCKSIZE);
1254 return (EIO);
1255 }
1256
1257 /*
1258 * Note: bsize may not be a power of two here so we need to do an
1259 * actual divide rather than a bitshift.
1260 */
1261 while (buflen > 0) {
1262 uint64_t bn = offset / bsize;
1263 int boff = offset % bsize;
1264 int ibn;
1265 const blkptr_t *indbp;
1266 blkptr_t bp;
1267
1268 if (bn > dnode->dn_maxblkid) {
1269 printf("warning: zfs bug: bn %llx > dn_maxblkid %llx\n",
1270 (unsigned long long)bn,
1271 (unsigned long long)dnode->dn_maxblkid);
1272 /*
1273 * zfs bug, will not return error
1274 * return (EIO);
1275 */
1276 }
1277
1278 if (dnode == dnode_cache_obj && bn == dnode_cache_bn)
1279 goto cached;
1280
1281 indbp = dnode->dn_blkptr;
1282 for (i = 0; i < nlevels; i++) {
1283 /*
1284 * Copy the bp from the indirect array so that
1285 * we can re-use the scratch buffer for multi-level
1286 * objects.
1287 */
1288 ibn = bn >> ((nlevels - i - 1) * ibshift);
1289 ibn &= ((1 << ibshift) - 1);
1290 bp = indbp[ibn];
1291 if (BP_IS_HOLE(&bp)) {
1292 memset(dnode_cache_buf, 0, bsize);
1293 break;
1294 }
1295 rc = zio_read(spa, &bp, dnode_cache_buf);
1296 if (rc)
1297 return (rc);
1298 indbp = (const blkptr_t *) dnode_cache_buf;
1299 }
1300 dnode_cache_obj = dnode;
1301 dnode_cache_bn = bn;
1302 cached:
1303
1304 /*
1305 * The buffer contains our data block. Copy what we
1306 * need from it and loop.
1307 */
1308 i = bsize - boff;
1309 if (i > buflen) i = buflen;
1310 memcpy(buf, &dnode_cache_buf[boff], i);
1311 buf = ((char*) buf) + i;
1312 offset += i;
1313 buflen -= i;
1314 }
1315
1316 return (0);
1317 }
1318
1319 /*
1320 * Lookup a value in a microzap directory. Assumes that the zap
1321 * scratch buffer contains the directory contents.
1322 */
1323 static int
1324 mzap_lookup(const dnode_phys_t *dnode, const char *name, uint64_t *value)
1325 {
1326 const mzap_phys_t *mz;
1327 const mzap_ent_phys_t *mze;
1328 size_t size;
1329 int chunks, i;
1330
1331 /*
1332 * Microzap objects use exactly one block. Read the whole
1333 * thing.
1334 */
1335 size = dnode->dn_datablkszsec * 512;
1336
1337 mz = (const mzap_phys_t *) zap_scratch;
1338 chunks = size / MZAP_ENT_LEN - 1;
1339
1340 for (i = 0; i < chunks; i++) {
1341 mze = &mz->mz_chunk[i];
1342 if (!strcmp(mze->mze_name, name)) {
1343 *value = mze->mze_value;
1344 return (0);
1345 }
1346 }
1347
1348 return (ENOENT);
1349 }
1350
1351 /*
1352 * Compare a name with a zap leaf entry. Return non-zero if the name
1353 * matches.
1354 */
1355 static int
1356 fzap_name_equal(const zap_leaf_t *zl, const zap_leaf_chunk_t *zc, const char *name)
1357 {
1358 size_t namelen;
1359 const zap_leaf_chunk_t *nc;
1360 const char *p;
1361
1362 namelen = zc->l_entry.le_name_numints;
1363
1364 nc = &ZAP_LEAF_CHUNK(zl, zc->l_entry.le_name_chunk);
1365 p = name;
1366 while (namelen > 0) {
1367 size_t len;
1368 len = namelen;
1369 if (len > ZAP_LEAF_ARRAY_BYTES)
1370 len = ZAP_LEAF_ARRAY_BYTES;
1371 if (memcmp(p, nc->l_array.la_array, len))
1372 return (0);
1373 p += len;
1374 namelen -= len;
1375 nc = &ZAP_LEAF_CHUNK(zl, nc->l_array.la_next);
1376 }
1377
1378 return 1;
1379 }
1380
1381 /*
1382 * Extract a uint64_t value from a zap leaf entry.
1383 */
1384 static uint64_t
1385 fzap_leaf_value(const zap_leaf_t *zl, const zap_leaf_chunk_t *zc)
1386 {
1387 const zap_leaf_chunk_t *vc;
1388 int i;
1389 uint64_t value;
1390 const uint8_t *p;
1391
1392 vc = &ZAP_LEAF_CHUNK(zl, zc->l_entry.le_value_chunk);
1393 for (i = 0, value = 0, p = vc->l_array.la_array; i < 8; i++) {
1394 value = (value << 8) | p[i];
1395 }
1396
1397 return value;
1398 }
1399
1400 /*
1401 * Lookup a value in a fatzap directory. Assumes that the zap scratch
1402 * buffer contains the directory header.
1403 */
1404 static int
1405 fzap_lookup(const spa_t *spa, const dnode_phys_t *dnode, const char *name, uint64_t *value)
1406 {
1407 int bsize = dnode->dn_datablkszsec << SPA_MINBLOCKSHIFT;
1408 zap_phys_t zh = *(zap_phys_t *) zap_scratch;
1409 fat_zap_t z;
1410 uint64_t *ptrtbl;
1411 uint64_t hash;
1412 int rc;
1413
1414 if (zh.zap_magic != ZAP_MAGIC)
1415 return (EIO);
1416
1417 z.zap_block_shift = ilog2(bsize);
1418 z.zap_phys = (zap_phys_t *) zap_scratch;
1419
1420 /*
1421 * Figure out where the pointer table is and read it in if necessary.
1422 */
1423 if (zh.zap_ptrtbl.zt_blk) {
1424 rc = dnode_read(spa, dnode, zh.zap_ptrtbl.zt_blk * bsize,
1425 zap_scratch, bsize);
1426 if (rc)
1427 return (rc);
1428 ptrtbl = (uint64_t *) zap_scratch;
1429 } else {
1430 ptrtbl = &ZAP_EMBEDDED_PTRTBL_ENT(&z, 0);
1431 }
1432
1433 hash = zap_hash(zh.zap_salt, name);
1434
1435 zap_leaf_t zl;
1436 zl.l_bs = z.zap_block_shift;
1437
1438 off_t off = ptrtbl[hash >> (64 - zh.zap_ptrtbl.zt_shift)] << zl.l_bs;
1439 zap_leaf_chunk_t *zc;
1440
1441 rc = dnode_read(spa, dnode, off, zap_scratch, bsize);
1442 if (rc)
1443 return (rc);
1444
1445 zl.l_phys = (zap_leaf_phys_t *) zap_scratch;
1446
1447 /*
1448 * Make sure this chunk matches our hash.
1449 */
1450 if (zl.l_phys->l_hdr.lh_prefix_len > 0
1451 && zl.l_phys->l_hdr.lh_prefix
1452 != hash >> (64 - zl.l_phys->l_hdr.lh_prefix_len))
1453 return (ENOENT);
1454
1455 /*
1456 * Hash within the chunk to find our entry.
1457 */
1458 int shift = (64 - ZAP_LEAF_HASH_SHIFT(&zl) - zl.l_phys->l_hdr.lh_prefix_len);
1459 int h = (hash >> shift) & ((1 << ZAP_LEAF_HASH_SHIFT(&zl)) - 1);
1460 h = zl.l_phys->l_hash[h];
1461 if (h == 0xffff)
1462 return (ENOENT);
1463 zc = &ZAP_LEAF_CHUNK(&zl, h);
1464 while (zc->l_entry.le_hash != hash) {
1465 if (zc->l_entry.le_next == 0xffff) {
1466 zc = 0;
1467 break;
1468 }
1469 zc = &ZAP_LEAF_CHUNK(&zl, zc->l_entry.le_next);
1470 }
1471 if (fzap_name_equal(&zl, zc, name)) {
1472 if (zc->l_entry.le_value_intlen * zc->l_entry.le_value_numints > 8)
1473 return (E2BIG);
1474 *value = fzap_leaf_value(&zl, zc);
1475 return (0);
1476 }
1477
1478 return (ENOENT);
1479 }
1480
1481 /*
1482 * Lookup a name in a zap object and return its value as a uint64_t.
1483 */
1484 static int
1485 zap_lookup(const spa_t *spa, const dnode_phys_t *dnode, const char *name, uint64_t *value)
1486 {
1487 int rc;
1488 uint64_t zap_type;
1489 size_t size = dnode->dn_datablkszsec << SPA_MINBLOCKSHIFT;
1490
1491 rc = dnode_read(spa, dnode, 0, zap_scratch, size);
1492 if (rc)
1493 return (rc);
1494
1495 zap_type = *(uint64_t *) zap_scratch;
1496 if (zap_type == ZBT_MICRO)
1497 return mzap_lookup(dnode, name, value);
1498 else if (zap_type == ZBT_HEADER)
1499 return fzap_lookup(spa, dnode, name, value);
1500 printf("ZFS: invalid zap_type=%d\n", (int)zap_type);
1501 return (EIO);
1502 }
1503
1504 /*
1505 * List a microzap directory. Assumes that the zap scratch buffer contains
1506 * the directory contents.
1507 */
1508 static int
1509 mzap_list(const dnode_phys_t *dnode, int (*callback)(const char *, uint64_t))
1510 {
1511 const mzap_phys_t *mz;
1512 const mzap_ent_phys_t *mze;
1513 size_t size;
1514 int chunks, i, rc;
1515
1516 /*
1517 * Microzap objects use exactly one block. Read the whole
1518 * thing.
1519 */
1520 size = dnode->dn_datablkszsec * 512;
1521 mz = (const mzap_phys_t *) zap_scratch;
1522 chunks = size / MZAP_ENT_LEN - 1;
1523
1524 for (i = 0; i < chunks; i++) {
1525 mze = &mz->mz_chunk[i];
1526 if (mze->mze_name[0]) {
1527 rc = callback(mze->mze_name, mze->mze_value);
1528 if (rc != 0)
1529 return (rc);
1530 }
1531 }
1532
1533 return (0);
1534 }
1535
1536 /*
1537 * List a fatzap directory. Assumes that the zap scratch buffer contains
1538 * the directory header.
1539 */
1540 static int
1541 fzap_list(const spa_t *spa, const dnode_phys_t *dnode, int (*callback)(const char *, uint64_t))
1542 {
1543 int bsize = dnode->dn_datablkszsec << SPA_MINBLOCKSHIFT;
1544 zap_phys_t zh = *(zap_phys_t *) zap_scratch;
1545 fat_zap_t z;
1546 int i, j, rc;
1547
1548 if (zh.zap_magic != ZAP_MAGIC)
1549 return (EIO);
1550
1551 z.zap_block_shift = ilog2(bsize);
1552 z.zap_phys = (zap_phys_t *) zap_scratch;
1553
1554 /*
1555 * This assumes that the leaf blocks start at block 1. The
1556 * documentation isn't exactly clear on this.
1557 */
1558 zap_leaf_t zl;
1559 zl.l_bs = z.zap_block_shift;
1560 for (i = 0; i < zh.zap_num_leafs; i++) {
1561 off_t off = (i + 1) << zl.l_bs;
1562 char name[256], *p;
1563 uint64_t value;
1564
1565 if (dnode_read(spa, dnode, off, zap_scratch, bsize))
1566 return (EIO);
1567
1568 zl.l_phys = (zap_leaf_phys_t *) zap_scratch;
1569
1570 for (j = 0; j < ZAP_LEAF_NUMCHUNKS(&zl); j++) {
1571 zap_leaf_chunk_t *zc, *nc;
1572 int namelen;
1573
1574 zc = &ZAP_LEAF_CHUNK(&zl, j);
1575 if (zc->l_entry.le_type != ZAP_CHUNK_ENTRY)
1576 continue;
1577 namelen = zc->l_entry.le_name_numints;
1578 if (namelen > sizeof(name))
1579 namelen = sizeof(name);
1580
1581 /*
1582 * Paste the name back together.
1583 */
1584 nc = &ZAP_LEAF_CHUNK(&zl, zc->l_entry.le_name_chunk);
1585 p = name;
1586 while (namelen > 0) {
1587 int len;
1588 len = namelen;
1589 if (len > ZAP_LEAF_ARRAY_BYTES)
1590 len = ZAP_LEAF_ARRAY_BYTES;
1591 memcpy(p, nc->l_array.la_array, len);
1592 p += len;
1593 namelen -= len;
1594 nc = &ZAP_LEAF_CHUNK(&zl, nc->l_array.la_next);
1595 }
1596
1597 /*
1598 * Assume the first eight bytes of the value are
1599 * a uint64_t.
1600 */
1601 value = fzap_leaf_value(&zl, zc);
1602
1603 //printf("%s 0x%jx\n", name, (uintmax_t)value);
1604 rc = callback((const char *)name, value);
1605 if (rc != 0)
1606 return (rc);
1607 }
1608 }
1609
1610 return (0);
1611 }
1612
1613 static int zfs_printf(const char *name, uint64_t value __unused)
1614 {
1615
1616 printf("%s\n", name);
1617
1618 return (0);
1619 }
1620
1621 /*
1622 * List a zap directory.
1623 */
1624 static int
1625 zap_list(const spa_t *spa, const dnode_phys_t *dnode)
1626 {
1627 uint64_t zap_type;
1628 size_t size = dnode->dn_datablkszsec * 512;
1629
1630 if (dnode_read(spa, dnode, 0, zap_scratch, size))
1631 return (EIO);
1632
1633 zap_type = *(uint64_t *) zap_scratch;
1634 if (zap_type == ZBT_MICRO)
1635 return mzap_list(dnode, zfs_printf);
1636 else
1637 return fzap_list(spa, dnode, zfs_printf);
1638 }
1639
1640 static int
1641 objset_get_dnode(const spa_t *spa, const objset_phys_t *os, uint64_t objnum, dnode_phys_t *dnode)
1642 {
1643 off_t offset;
1644
1645 offset = objnum * sizeof(dnode_phys_t);
1646 return dnode_read(spa, &os->os_meta_dnode, offset,
1647 dnode, sizeof(dnode_phys_t));
1648 }
1649
1650 static int
1651 mzap_rlookup(const spa_t *spa, const dnode_phys_t *dnode, char *name, uint64_t value)
1652 {
1653 const mzap_phys_t *mz;
1654 const mzap_ent_phys_t *mze;
1655 size_t size;
1656 int chunks, i;
1657
1658 /*
1659 * Microzap objects use exactly one block. Read the whole
1660 * thing.
1661 */
1662 size = dnode->dn_datablkszsec * 512;
1663
1664 mz = (const mzap_phys_t *) zap_scratch;
1665 chunks = size / MZAP_ENT_LEN - 1;
1666
1667 for (i = 0; i < chunks; i++) {
1668 mze = &mz->mz_chunk[i];
1669 if (value == mze->mze_value) {
1670 strcpy(name, mze->mze_name);
1671 return (0);
1672 }
1673 }
1674
1675 return (ENOENT);
1676 }
1677
1678 static void
1679 fzap_name_copy(const zap_leaf_t *zl, const zap_leaf_chunk_t *zc, char *name)
1680 {
1681 size_t namelen;
1682 const zap_leaf_chunk_t *nc;
1683 char *p;
1684
1685 namelen = zc->l_entry.le_name_numints;
1686
1687 nc = &ZAP_LEAF_CHUNK(zl, zc->l_entry.le_name_chunk);
1688 p = name;
1689 while (namelen > 0) {
1690 size_t len;
1691 len = namelen;
1692 if (len > ZAP_LEAF_ARRAY_BYTES)
1693 len = ZAP_LEAF_ARRAY_BYTES;
1694 memcpy(p, nc->l_array.la_array, len);
1695 p += len;
1696 namelen -= len;
1697 nc = &ZAP_LEAF_CHUNK(zl, nc->l_array.la_next);
1698 }
1699
1700 *p = '\0';
1701 }
1702
1703 static int
1704 fzap_rlookup(const spa_t *spa, const dnode_phys_t *dnode, char *name, uint64_t value)
1705 {
1706 int bsize = dnode->dn_datablkszsec << SPA_MINBLOCKSHIFT;
1707 zap_phys_t zh = *(zap_phys_t *) zap_scratch;
1708 fat_zap_t z;
1709 int i, j;
1710
1711 if (zh.zap_magic != ZAP_MAGIC)
1712 return (EIO);
1713
1714 z.zap_block_shift = ilog2(bsize);
1715 z.zap_phys = (zap_phys_t *) zap_scratch;
1716
1717 /*
1718 * This assumes that the leaf blocks start at block 1. The
1719 * documentation isn't exactly clear on this.
1720 */
1721 zap_leaf_t zl;
1722 zl.l_bs = z.zap_block_shift;
1723 for (i = 0; i < zh.zap_num_leafs; i++) {
1724 off_t off = (i + 1) << zl.l_bs;
1725
1726 if (dnode_read(spa, dnode, off, zap_scratch, bsize))
1727 return (EIO);
1728
1729 zl.l_phys = (zap_leaf_phys_t *) zap_scratch;
1730
1731 for (j = 0; j < ZAP_LEAF_NUMCHUNKS(&zl); j++) {
1732 zap_leaf_chunk_t *zc;
1733
1734 zc = &ZAP_LEAF_CHUNK(&zl, j);
1735 if (zc->l_entry.le_type != ZAP_CHUNK_ENTRY)
1736 continue;
1737 if (zc->l_entry.le_value_intlen != 8 ||
1738 zc->l_entry.le_value_numints != 1)
1739 continue;
1740
1741 if (fzap_leaf_value(&zl, zc) == value) {
1742 fzap_name_copy(&zl, zc, name);
1743 return (0);
1744 }
1745 }
1746 }
1747
1748 return (ENOENT);
1749 }
1750
1751 static int
1752 zap_rlookup(const spa_t *spa, const dnode_phys_t *dnode, char *name, uint64_t value)
1753 {
1754 int rc;
1755 uint64_t zap_type;
1756 size_t size = dnode->dn_datablkszsec * 512;
1757
1758 rc = dnode_read(spa, dnode, 0, zap_scratch, size);
1759 if (rc)
1760 return (rc);
1761
1762 zap_type = *(uint64_t *) zap_scratch;
1763 if (zap_type == ZBT_MICRO)
1764 return mzap_rlookup(spa, dnode, name, value);
1765 else
1766 return fzap_rlookup(spa, dnode, name, value);
1767 }
1768
1769 static int
1770 zfs_rlookup(const spa_t *spa, uint64_t objnum, char *result)
1771 {
1772 char name[256];
1773 char component[256];
1774 uint64_t dir_obj, parent_obj, child_dir_zapobj;
1775 dnode_phys_t child_dir_zap, dataset, dir, parent;
1776 dsl_dir_phys_t *dd;
1777 dsl_dataset_phys_t *ds;
1778 char *p;
1779 int len;
1780
1781 p = &name[sizeof(name) - 1];
1782 *p = '\0';
1783
1784 if (objset_get_dnode(spa, &spa->spa_mos, objnum, &dataset)) {
1785 printf("ZFS: can't find dataset %ju\n", (uintmax_t)objnum);
1786 return (EIO);
1787 }
1788 ds = (dsl_dataset_phys_t *)&dataset.dn_bonus;
1789 dir_obj = ds->ds_dir_obj;
1790
1791 for (;;) {
1792 if (objset_get_dnode(spa, &spa->spa_mos, dir_obj, &dir) != 0)
1793 return (EIO);
1794 dd = (dsl_dir_phys_t *)&dir.dn_bonus;
1795
1796 /* Actual loop condition. */
1797 parent_obj = dd->dd_parent_obj;
1798 if (parent_obj == 0)
1799 break;
1800
1801 if (objset_get_dnode(spa, &spa->spa_mos, parent_obj, &parent) != 0)
1802 return (EIO);
1803 dd = (dsl_dir_phys_t *)&parent.dn_bonus;
1804 child_dir_zapobj = dd->dd_child_dir_zapobj;
1805 if (objset_get_dnode(spa, &spa->spa_mos, child_dir_zapobj, &child_dir_zap) != 0)
1806 return (EIO);
1807 if (zap_rlookup(spa, &child_dir_zap, component, dir_obj) != 0)
1808 return (EIO);
1809
1810 len = strlen(component);
1811 p -= len;
1812 memcpy(p, component, len);
1813 --p;
1814 *p = '/';
1815
1816 /* Actual loop iteration. */
1817 dir_obj = parent_obj;
1818 }
1819
1820 if (*p != '\0')
1821 ++p;
1822 strcpy(result, p);
1823
1824 return (0);
1825 }
1826
1827 static int
1828 zfs_lookup_dataset(const spa_t *spa, const char *name, uint64_t *objnum)
1829 {
1830 char element[256];
1831 uint64_t dir_obj, child_dir_zapobj;
1832 dnode_phys_t child_dir_zap, dir;
1833 dsl_dir_phys_t *dd;
1834 const char *p, *q;
1835
1836 if (objset_get_dnode(spa, &spa->spa_mos, DMU_POOL_DIRECTORY_OBJECT, &dir))
1837 return (EIO);
1838 if (zap_lookup(spa, &dir, DMU_POOL_ROOT_DATASET, &dir_obj))
1839 return (EIO);
1840
1841 p = name;
1842 for (;;) {
1843 if (objset_get_dnode(spa, &spa->spa_mos, dir_obj, &dir))
1844 return (EIO);
1845 dd = (dsl_dir_phys_t *)&dir.dn_bonus;
1846
1847 while (*p == '/')
1848 p++;
1849 /* Actual loop condition #1. */
1850 if (*p == '\0')
1851 break;
1852
1853 q = strchr(p, '/');
1854 if (q) {
1855 memcpy(element, p, q - p);
1856 element[q - p] = '\0';
1857 p = q + 1;
1858 } else {
1859 strcpy(element, p);
1860 p += strlen(p);
1861 }
1862
1863 child_dir_zapobj = dd->dd_child_dir_zapobj;
1864 if (objset_get_dnode(spa, &spa->spa_mos, child_dir_zapobj, &child_dir_zap) != 0)
1865 return (EIO);
1866
1867 /* Actual loop condition #2. */
1868 if (zap_lookup(spa, &child_dir_zap, element, &dir_obj) != 0)
1869 return (ENOENT);
1870 }
1871
1872 *objnum = dd->dd_head_dataset_obj;
1873 return (0);
1874 }
1875
1876 #pragma GCC diagnostic ignored "-Wstrict-aliasing"
1877 static int
1878 zfs_list_dataset(const spa_t *spa, uint64_t objnum/*, int pos, char *entry*/)
1879 {
1880 uint64_t dir_obj, child_dir_zapobj;
1881 dnode_phys_t child_dir_zap, dir, dataset;
1882 dsl_dataset_phys_t *ds;
1883 dsl_dir_phys_t *dd;
1884
1885 if (objset_get_dnode(spa, &spa->spa_mos, objnum, &dataset)) {
1886 printf("ZFS: can't find dataset %ju\n", (uintmax_t)objnum);
1887 return (EIO);
1888 }
1889 ds = (dsl_dataset_phys_t *) &dataset.dn_bonus;
1890 dir_obj = ds->ds_dir_obj;
1891
1892 if (objset_get_dnode(spa, &spa->spa_mos, dir_obj, &dir)) {
1893 printf("ZFS: can't find dirobj %ju\n", (uintmax_t)dir_obj);
1894 return (EIO);
1895 }
1896 dd = (dsl_dir_phys_t *)&dir.dn_bonus;
1897
1898 child_dir_zapobj = dd->dd_child_dir_zapobj;
1899 if (objset_get_dnode(spa, &spa->spa_mos, child_dir_zapobj, &child_dir_zap) != 0) {
1900 printf("ZFS: can't find child zap %ju\n", (uintmax_t)dir_obj);
1901 return (EIO);
1902 }
1903
1904 return (zap_list(spa, &child_dir_zap) != 0);
1905 }
1906
1907 int
1908 zfs_callback_dataset(const spa_t *spa, uint64_t objnum, int (*callback)(const char *, uint64_t))
1909 {
1910 uint64_t dir_obj, child_dir_zapobj, zap_type;
1911 dnode_phys_t child_dir_zap, dir, dataset;
1912 dsl_dataset_phys_t *ds;
1913 dsl_dir_phys_t *dd;
1914 int err;
1915
1916 err = objset_get_dnode(spa, &spa->spa_mos, objnum, &dataset);
1917 if (err != 0) {
1918 printf("ZFS: can't find dataset %ju\n", (uintmax_t)objnum);
1919 return (err);
1920 }
1921 ds = (dsl_dataset_phys_t *) &dataset.dn_bonus;
1922 dir_obj = ds->ds_dir_obj;
1923
1924 err = objset_get_dnode(spa, &spa->spa_mos, dir_obj, &dir);
1925 if (err != 0) {
1926 printf("ZFS: can't find dirobj %ju\n", (uintmax_t)dir_obj);
1927 return (err);
1928 }
1929 dd = (dsl_dir_phys_t *)&dir.dn_bonus;
1930
1931 child_dir_zapobj = dd->dd_child_dir_zapobj;
1932 err = objset_get_dnode(spa, &spa->spa_mos, child_dir_zapobj, &child_dir_zap);
1933 if (err != 0) {
1934 printf("ZFS: can't find child zap %ju\n", (uintmax_t)dir_obj);
1935 return (err);
1936 }
1937
1938 err = dnode_read(spa, &child_dir_zap, 0, zap_scratch, child_dir_zap.dn_datablkszsec * 512);
1939 if (err != 0)
1940 return (err);
1941
1942 zap_type = *(uint64_t *) zap_scratch;
1943 if (zap_type == ZBT_MICRO)
1944 return mzap_list(&child_dir_zap, callback);
1945 else
1946 return fzap_list(spa, &child_dir_zap, callback);
1947 }
1948
1949 /*
1950 * Find the object set given the object number of its dataset object
1951 * and return its details in *objset
1952 */
1953 static int
1954 zfs_mount_dataset(const spa_t *spa, uint64_t objnum, objset_phys_t *objset)
1955 {
1956 dnode_phys_t dataset;
1957 dsl_dataset_phys_t *ds;
1958
1959 if (objset_get_dnode(spa, &spa->spa_mos, objnum, &dataset)) {
1960 printf("ZFS: can't find dataset %ju\n", (uintmax_t)objnum);
1961 return (EIO);
1962 }
1963
1964 ds = (dsl_dataset_phys_t *) &dataset.dn_bonus;
1965 if (zio_read(spa, &ds->ds_bp, objset)) {
1966 printf("ZFS: can't read object set for dataset %ju\n",
1967 (uintmax_t)objnum);
1968 return (EIO);
1969 }
1970
1971 return (0);
1972 }
1973
1974 /*
1975 * Find the object set pointed to by the BOOTFS property or the root
1976 * dataset if there is none and return its details in *objset
1977 */
1978 static int
1979 zfs_get_root(const spa_t *spa, uint64_t *objid)
1980 {
1981 dnode_phys_t dir, propdir;
1982 uint64_t props, bootfs, root;
1983
1984 *objid = 0;
1985
1986 /*
1987 * Start with the MOS directory object.
1988 */
1989 if (objset_get_dnode(spa, &spa->spa_mos, DMU_POOL_DIRECTORY_OBJECT, &dir)) {
1990 printf("ZFS: can't read MOS object directory\n");
1991 return (EIO);
1992 }
1993
1994 /*
1995 * Lookup the pool_props and see if we can find a bootfs.
1996 */
1997 if (zap_lookup(spa, &dir, DMU_POOL_PROPS, &props) == 0
1998 && objset_get_dnode(spa, &spa->spa_mos, props, &propdir) == 0
1999 && zap_lookup(spa, &propdir, "bootfs", &bootfs) == 0
2000 && bootfs != 0)
2001 {
2002 *objid = bootfs;
2003 return (0);
2004 }
2005 /*
2006 * Lookup the root dataset directory
2007 */
2008 if (zap_lookup(spa, &dir, DMU_POOL_ROOT_DATASET, &root)
2009 || objset_get_dnode(spa, &spa->spa_mos, root, &dir)) {
2010 printf("ZFS: can't find root dsl_dir\n");
2011 return (EIO);
2012 }
2013
2014 /*
2015 * Use the information from the dataset directory's bonus buffer
2016 * to find the dataset object and from that the object set itself.
2017 */
2018 dsl_dir_phys_t *dd = (dsl_dir_phys_t *) &dir.dn_bonus;
2019 *objid = dd->dd_head_dataset_obj;
2020 return (0);
2021 }
2022
2023 static int
2024 zfs_mount(const spa_t *spa, uint64_t rootobj, struct zfsmount *mnt)
2025 {
2026
2027 mnt->spa = spa;
2028
2029 /*
2030 * Find the root object set if not explicitly provided
2031 */
2032 if (rootobj == 0 && zfs_get_root(spa, &rootobj)) {
2033 printf("ZFS: can't find root filesystem\n");
2034 return (EIO);
2035 }
2036
2037 if (zfs_mount_dataset(spa, rootobj, &mnt->objset)) {
2038 printf("ZFS: can't open root filesystem\n");
2039 return (EIO);
2040 }
2041
2042 mnt->rootobj = rootobj;
2043
2044 return (0);
2045 }
2046
2047 /*
2048 * callback function for feature name checks.
2049 */
2050 static int
2051 check_feature(const char *name, uint64_t value)
2052 {
2053 int i;
2054
2055 if (value == 0)
2056 return (0);
2057 if (name[0] == '\0')
2058 return (0);
2059
2060 for (i = 0; features_for_read[i] != NULL; i++) {
2061 if (strcmp(name, features_for_read[i]) == 0)
2062 return (0);
2063 }
2064 printf("ZFS: unsupported feature: %s\n", name);
2065 return (EIO);
2066 }
2067
2068 /*
2069 * Checks whether the MOS features that are active are supported.
2070 */
2071 static int
2072 check_mos_features(const spa_t *spa)
2073 {
2074 dnode_phys_t dir;
2075 uint64_t objnum, zap_type;
2076 size_t size;
2077 int rc;
2078
2079 if ((rc = objset_get_dnode(spa, &spa->spa_mos, DMU_OT_OBJECT_DIRECTORY,
2080 &dir)) != 0)
2081 return (rc);
2082 if ((rc = zap_lookup(spa, &dir, DMU_POOL_FEATURES_FOR_READ, &objnum)) != 0)
2083 return (rc);
2084
2085 if ((rc = objset_get_dnode(spa, &spa->spa_mos, objnum, &dir)) != 0)
2086 return (rc);
2087
2088 if (dir.dn_type != DMU_OTN_ZAP_METADATA)
2089 return (EIO);
2090
2091 size = dir.dn_datablkszsec * 512;
2092 if (dnode_read(spa, &dir, 0, zap_scratch, size))
2093 return (EIO);
2094
2095 zap_type = *(uint64_t *) zap_scratch;
2096 if (zap_type == ZBT_MICRO)
2097 rc = mzap_list(&dir, check_feature);
2098 else
2099 rc = fzap_list(spa, &dir, check_feature);
2100
2101 return (rc);
2102 }
2103
2104 static int
2105 zfs_spa_init(spa_t *spa)
2106 {
2107 int rc;
2108
2109 if (zio_read(spa, &spa->spa_uberblock.ub_rootbp, &spa->spa_mos)) {
2110 printf("ZFS: can't read MOS of pool %s\n", spa->spa_name);
2111 return (EIO);
2112 }
2113 if (spa->spa_mos.os_type != DMU_OST_META) {
2114 printf("ZFS: corrupted MOS of pool %s\n", spa->spa_name);
2115 return (EIO);
2116 }
2117
2118 rc = check_mos_features(spa);
2119 if (rc != 0) {
2120 printf("ZFS: pool %s is not supported\n", spa->spa_name);
2121 }
2122
2123 return (rc);
2124 }
2125
2126 static int
2127 zfs_dnode_stat(const spa_t *spa, dnode_phys_t *dn, struct stat *sb)
2128 {
2129
2130 if (dn->dn_bonustype != DMU_OT_SA) {
2131 znode_phys_t *zp = (znode_phys_t *)dn->dn_bonus;
2132
2133 sb->st_mode = zp->zp_mode;
2134 sb->st_uid = zp->zp_uid;
2135 sb->st_gid = zp->zp_gid;
2136 sb->st_size = zp->zp_size;
2137 } else {
2138 sa_hdr_phys_t *sahdrp;
2139 int hdrsize;
2140 size_t size = 0;
2141 void *buf = NULL;
2142
2143 if (dn->dn_bonuslen != 0)
2144 sahdrp = (sa_hdr_phys_t *)DN_BONUS(dn);
2145 else {
2146 if ((dn->dn_flags & DNODE_FLAG_SPILL_BLKPTR) != 0) {
2147 blkptr_t *bp = &dn->dn_spill;
2148 int error;
2149
2150 size = BP_GET_LSIZE(bp);
2151 buf = zfs_alloc(size);
2152 error = zio_read(spa, bp, buf);
2153 if (error != 0) {
2154 zfs_free(buf, size);
2155 return (error);
2156 }
2157 sahdrp = buf;
2158 } else {
2159 return (EIO);
2160 }
2161 }
2162 hdrsize = SA_HDR_SIZE(sahdrp);
2163 sb->st_mode = *(uint64_t *)((char *)sahdrp + hdrsize +
2164 SA_MODE_OFFSET);
2165 sb->st_uid = *(uint64_t *)((char *)sahdrp + hdrsize +
2166 SA_UID_OFFSET);
2167 sb->st_gid = *(uint64_t *)((char *)sahdrp + hdrsize +
2168 SA_GID_OFFSET);
2169 sb->st_size = *(uint64_t *)((char *)sahdrp + hdrsize +
2170 SA_SIZE_OFFSET);
2171 if (buf != NULL)
2172 zfs_free(buf, size);
2173 }
2174
2175 return (0);
2176 }
2177
2178 /*
2179 * Lookup a file and return its dnode.
2180 */
2181 static int
2182 zfs_lookup(const struct zfsmount *mnt, const char *upath, dnode_phys_t *dnode)
2183 {
2184 int rc;
2185 uint64_t objnum, rootnum, parentnum;
2186 const spa_t *spa;
2187 dnode_phys_t dn;
2188 const char *p, *q;
2189 char element[256];
2190 char path[1024];
2191 int symlinks_followed = 0;
2192 struct stat sb;
2193
2194 spa = mnt->spa;
2195 if (mnt->objset.os_type != DMU_OST_ZFS) {
2196 printf("ZFS: unexpected object set type %ju\n",
2197 (uintmax_t)mnt->objset.os_type);
2198 return (EIO);
2199 }
2200
2201 /*
2202 * Get the root directory dnode.
2203 */
2204 rc = objset_get_dnode(spa, &mnt->objset, MASTER_NODE_OBJ, &dn);
2205 if (rc)
2206 return (rc);
2207
2208 rc = zap_lookup(spa, &dn, ZFS_ROOT_OBJ, &rootnum);
2209 if (rc)
2210 return (rc);
2211
2212 rc = objset_get_dnode(spa, &mnt->objset, rootnum, &dn);
2213 if (rc)
2214 return (rc);
2215
2216 objnum = rootnum;
2217 p = upath;
2218 while (p && *p) {
2219 while (*p == '/')
2220 p++;
2221 if (!*p)
2222 break;
2223 q = strchr(p, '/');
2224 if (q) {
2225 memcpy(element, p, q - p);
2226 element[q - p] = 0;
2227 p = q;
2228 } else {
2229 strcpy(element, p);
2230 p = 0;
2231 }
2232
2233 rc = zfs_dnode_stat(spa, &dn, &sb);
2234 if (rc)
2235 return (rc);
2236 if (!S_ISDIR(sb.st_mode))
2237 return (ENOTDIR);
2238
2239 parentnum = objnum;
2240 rc = zap_lookup(spa, &dn, element, &objnum);
2241 if (rc)
2242 return (rc);
2243 objnum = ZFS_DIRENT_OBJ(objnum);
2244
2245 rc = objset_get_dnode(spa, &mnt->objset, objnum, &dn);
2246 if (rc)
2247 return (rc);
2248
2249 /*
2250 * Check for symlink.
2251 */
2252 rc = zfs_dnode_stat(spa, &dn, &sb);
2253 if (rc)
2254 return (rc);
2255 if (S_ISLNK(sb.st_mode)) {
2256 if (symlinks_followed > 10)
2257 return (EMLINK);
2258 symlinks_followed++;
2259
2260 /*
2261 * Read the link value and copy the tail of our
2262 * current path onto the end.
2263 */
2264 if (p)
2265 strcpy(&path[sb.st_size], p);
2266 else
2267 path[sb.st_size] = 0;
2268 /*
2269 * Second test is purely to silence bogus compiler
2270 * warning about accessing past the end of dn_bonus.
2271 */
2272 if (sb.st_size + sizeof(znode_phys_t) <=
2273 dn.dn_bonuslen && sizeof(znode_phys_t) <=
2274 sizeof(dn.dn_bonus)) {
2275 memcpy(path, &dn.dn_bonus[sizeof(znode_phys_t)],
2276 sb.st_size);
2277 } else {
2278 rc = dnode_read(spa, &dn, 0, path, sb.st_size);
2279 if (rc)
2280 return (rc);
2281 }
2282
2283 /*
2284 * Restart with the new path, starting either at
2285 * the root or at the parent depending whether or
2286 * not the link is relative.
2287 */
2288 p = path;
2289 if (*p == '/')
2290 objnum = rootnum;
2291 else
2292 objnum = parentnum;
2293 objset_get_dnode(spa, &mnt->objset, objnum, &dn);
2294 }
2295 }
2296
2297 *dnode = dn;
2298 return (0);
2299 }
Unleashed OS