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 (c) 2005, 2010, Oracle and/or its affiliates. All rights reserved.
23 * Copyright 2011 Nexenta Systems, Inc. All rights reserved.
24 * Copyright (c) 2012 by Delphix. All rights reserved.
25 */
26
27 /* Portions Copyright 2010 Robert Milkowski */
28
29 #include <mdb/mdb_ctf.h>
30 #include <sys/zfs_context.h>
31 #include <sys/mdb_modapi.h>
32 #include <sys/dbuf.h>
33 #include <sys/dmu_objset.h>
34 #include <sys/dsl_dir.h>
35 #include <sys/dsl_pool.h>
36 #include <sys/metaslab_impl.h>
37 #include <sys/space_map.h>
38 #include <sys/list.h>
39 #include <sys/spa_impl.h>
40 #include <sys/vdev_impl.h>
41 #include <sys/zap_leaf.h>
42 #include <sys/zap_impl.h>
43 #include <ctype.h>
44 #include <sys/zfs_acl.h>
45 #include <sys/sa_impl.h>
46
47 #ifdef _KERNEL
48 #define ZFS_OBJ_NAME "zfs"
49 #else
50 #define ZFS_OBJ_NAME "libzpool.so.1"
51 #endif
52
53 #ifndef _KERNEL
54 int aok;
55 #endif
56
57 static int
58 getmember(uintptr_t addr, const char *type, mdb_ctf_id_t *idp,
59 const char *member, int len, void *buf)
60 {
61 mdb_ctf_id_t id;
62 ulong_t off;
63 char name[64];
64
65 if (idp == NULL) {
66 if (mdb_ctf_lookup_by_name(type, &id) == -1) {
67 mdb_warn("couldn't find type %s", type);
68 return (DCMD_ERR);
69 }
70 idp = &id;
71 } else {
72 type = name;
73 mdb_ctf_type_name(*idp, name, sizeof (name));
74 }
75
76 if (mdb_ctf_offsetof(*idp, member, &off) == -1) {
77 mdb_warn("couldn't find member %s of type %s\n", member, type);
78 return (DCMD_ERR);
79 }
80 if (off % 8 != 0) {
81 mdb_warn("member %s of type %s is unsupported bitfield",
82 member, type);
83 return (DCMD_ERR);
84 }
85 off /= 8;
86
87 if (mdb_vread(buf, len, addr + off) == -1) {
88 mdb_warn("failed to read %s from %s at %p",
89 member, type, addr + off);
90 return (DCMD_ERR);
91 }
92 /* mdb_warn("read %s from %s at %p+%llx\n", member, type, addr, off); */
93
94 return (0);
95 }
96
97 #define GETMEMB(addr, type, member, dest) \
98 getmember(addr, #type, NULL, #member, sizeof (dest), &(dest))
99
100 #define GETMEMBID(addr, ctfid, member, dest) \
101 getmember(addr, NULL, ctfid, #member, sizeof (dest), &(dest))
102
103 static int
104 getrefcount(uintptr_t addr, mdb_ctf_id_t *id,
105 const char *member, uint64_t *rc)
106 {
107 static int gotid;
108 static mdb_ctf_id_t rc_id;
109 ulong_t off;
110
111 if (!gotid) {
112 if (mdb_ctf_lookup_by_name("struct refcount", &rc_id) == -1) {
113 mdb_warn("couldn't find struct refcount");
114 return (DCMD_ERR);
115 }
116 gotid = TRUE;
117 }
118
119 if (mdb_ctf_offsetof(*id, member, &off) == -1) {
120 char name[64];
121 mdb_ctf_type_name(*id, name, sizeof (name));
122 mdb_warn("couldn't find member %s of type %s\n", member, name);
123 return (DCMD_ERR);
124 }
125 off /= 8;
126
127 return (GETMEMBID(addr + off, &rc_id, rc_count, *rc));
128 }
129
130 static boolean_t
131 strisprint(const char *cp)
132 {
133 for (; *cp; cp++) {
134 if (!isprint(*cp))
135 return (B_FALSE);
136 }
137 return (B_TRUE);
138 }
139
140 static int verbose;
141
142 static int
143 freelist_walk_init(mdb_walk_state_t *wsp)
144 {
145 if (wsp->walk_addr == NULL) {
146 mdb_warn("must supply starting address\n");
147 return (WALK_ERR);
148 }
149
150 wsp->walk_data = 0; /* Index into the freelist */
151 return (WALK_NEXT);
152 }
153
154 static int
155 freelist_walk_step(mdb_walk_state_t *wsp)
156 {
157 uint64_t entry;
158 uintptr_t number = (uintptr_t)wsp->walk_data;
159 char *ddata[] = { "ALLOC", "FREE", "CONDENSE", "INVALID",
160 "INVALID", "INVALID", "INVALID", "INVALID" };
161 int mapshift = SPA_MINBLOCKSHIFT;
162
163 if (mdb_vread(&entry, sizeof (entry), wsp->walk_addr) == -1) {
164 mdb_warn("failed to read freelist entry %p", wsp->walk_addr);
165 return (WALK_DONE);
166 }
167 wsp->walk_addr += sizeof (entry);
168 wsp->walk_data = (void *)(number + 1);
169
170 if (SM_DEBUG_DECODE(entry)) {
171 mdb_printf("DEBUG: %3u %10s: txg=%llu pass=%llu\n",
172 number,
173 ddata[SM_DEBUG_ACTION_DECODE(entry)],
174 SM_DEBUG_TXG_DECODE(entry),
175 SM_DEBUG_SYNCPASS_DECODE(entry));
176 } else {
177 mdb_printf("Entry: %3u offsets=%08llx-%08llx type=%c "
178 "size=%06llx", number,
179 SM_OFFSET_DECODE(entry) << mapshift,
180 (SM_OFFSET_DECODE(entry) + SM_RUN_DECODE(entry)) <<
181 mapshift,
182 SM_TYPE_DECODE(entry) == SM_ALLOC ? 'A' : 'F',
183 SM_RUN_DECODE(entry) << mapshift);
184 if (verbose)
185 mdb_printf(" (raw=%012llx)\n", entry);
186 mdb_printf("\n");
187 }
188 return (WALK_NEXT);
189 }
190
191
192 static int
193 dataset_name(uintptr_t addr, char *buf)
194 {
195 static int gotid;
196 static mdb_ctf_id_t dd_id;
197 uintptr_t dd_parent;
198 char dd_myname[MAXNAMELEN];
199
200 if (!gotid) {
201 if (mdb_ctf_lookup_by_name("struct dsl_dir",
202 &dd_id) == -1) {
203 mdb_warn("couldn't find struct dsl_dir");
204 return (DCMD_ERR);
205 }
206 gotid = TRUE;
207 }
208 if (GETMEMBID(addr, &dd_id, dd_parent, dd_parent) ||
209 GETMEMBID(addr, &dd_id, dd_myname, dd_myname)) {
210 return (DCMD_ERR);
211 }
212
213 if (dd_parent) {
214 if (dataset_name(dd_parent, buf))
215 return (DCMD_ERR);
216 strcat(buf, "/");
217 }
218
219 if (dd_myname[0])
220 strcat(buf, dd_myname);
221 else
222 strcat(buf, "???");
223
224 return (0);
225 }
226
227 static int
228 objset_name(uintptr_t addr, char *buf)
229 {
230 static int gotid;
231 static mdb_ctf_id_t os_id, ds_id;
232 uintptr_t os_dsl_dataset;
233 char ds_snapname[MAXNAMELEN];
234 uintptr_t ds_dir;
235
236 buf[0] = '\0';
237
238 if (!gotid) {
239 if (mdb_ctf_lookup_by_name("struct objset",
240 &os_id) == -1) {
241 mdb_warn("couldn't find struct objset");
242 return (DCMD_ERR);
243 }
244 if (mdb_ctf_lookup_by_name("struct dsl_dataset",
245 &ds_id) == -1) {
246 mdb_warn("couldn't find struct dsl_dataset");
247 return (DCMD_ERR);
248 }
249
250 gotid = TRUE;
251 }
252
253 if (GETMEMBID(addr, &os_id, os_dsl_dataset, os_dsl_dataset))
254 return (DCMD_ERR);
255
256 if (os_dsl_dataset == 0) {
257 strcat(buf, "mos");
258 return (0);
259 }
260
261 if (GETMEMBID(os_dsl_dataset, &ds_id, ds_snapname, ds_snapname) ||
262 GETMEMBID(os_dsl_dataset, &ds_id, ds_dir, ds_dir)) {
263 return (DCMD_ERR);
264 }
265
266 if (ds_dir && dataset_name(ds_dir, buf))
267 return (DCMD_ERR);
268
269 if (ds_snapname[0]) {
270 strcat(buf, "@");
271 strcat(buf, ds_snapname);
272 }
273 return (0);
274 }
275
276 static void
277 enum_lookup(char *out, size_t size, mdb_ctf_id_t id, int val,
278 const char *prefix)
279 {
280 const char *cp;
281 size_t len = strlen(prefix);
282
283 if ((cp = mdb_ctf_enum_name(id, val)) != NULL) {
284 if (strncmp(cp, prefix, len) == 0)
285 cp += len;
286 (void) strncpy(out, cp, size);
287 } else {
288 mdb_snprintf(out, size, "? (%d)", val);
289 }
290 }
291
292 /* ARGSUSED */
293 static int
294 zfs_params(uintptr_t addr, uint_t flags, int argc, const mdb_arg_t *argv)
295 {
296 /*
297 * This table can be approximately generated by running:
298 * egrep "^[a-z0-9_]+ [a-z0-9_]+( =.*)?;" *.c | cut -d ' ' -f 2
299 */
300 static const char *params[] = {
301 "arc_reduce_dnlc_percent",
302 "zfs_arc_max",
303 "zfs_arc_min",
304 "arc_shrink_shift",
305 "zfs_mdcomp_disable",
306 "zfs_prefetch_disable",
307 "zfetch_max_streams",
308 "zfetch_min_sec_reap",
309 "zfetch_block_cap",
310 "zfetch_array_rd_sz",
311 "zfs_default_bs",
312 "zfs_default_ibs",
313 "metaslab_aliquot",
314 "reference_tracking_enable",
315 "reference_history",
316 "spa_max_replication_override",
317 "spa_mode_global",
318 "zfs_flags",
319 "zfs_txg_synctime_ms",
320 "zfs_txg_timeout",
321 "zfs_write_limit_min",
322 "zfs_write_limit_max",
323 "zfs_write_limit_shift",
324 "zfs_write_limit_override",
325 "zfs_no_write_throttle",
326 "zfs_vdev_cache_max",
327 "zfs_vdev_cache_size",
328 "zfs_vdev_cache_bshift",
329 "vdev_mirror_shift",
330 "zfs_vdev_max_pending",
331 "zfs_vdev_min_pending",
332 "zfs_scrub_limit",
333 "zfs_no_scrub_io",
334 "zfs_no_scrub_prefetch",
335 "zfs_vdev_time_shift",
336 "zfs_vdev_ramp_rate",
337 "zfs_vdev_aggregation_limit",
338 "fzap_default_block_shift",
339 "zfs_immediate_write_sz",
340 "zfs_read_chunk_size",
341 "zfs_nocacheflush",
342 "zil_replay_disable",
343 "metaslab_gang_bang",
344 "metaslab_df_alloc_threshold",
345 "metaslab_df_free_pct",
346 "zio_injection_enabled",
347 "zvol_immediate_write_sz",
348 };
349
350 for (int i = 0; i < sizeof (params) / sizeof (params[0]); i++) {
351 int sz;
352 uint64_t val64;
353 uint32_t *val32p = (uint32_t *)&val64;
354
355 sz = mdb_readvar(&val64, params[i]);
356 if (sz == 4) {
357 mdb_printf("%s = 0x%x\n", params[i], *val32p);
358 } else if (sz == 8) {
359 mdb_printf("%s = 0x%llx\n", params[i], val64);
360 } else {
361 mdb_warn("variable %s not found", params[i]);
362 }
363 }
364
365 return (DCMD_OK);
366 }
367
368 /* ARGSUSED */
369 static int
370 blkptr(uintptr_t addr, uint_t flags, int argc, const mdb_arg_t *argv)
371 {
372 mdb_ctf_id_t type_enum, checksum_enum, compress_enum;
373 char type[80], checksum[80], compress[80];
374 blkptr_t blk, *bp = &blk;
375 char buf[BP_SPRINTF_LEN];
376
377 if (mdb_vread(&blk, sizeof (blkptr_t), addr) == -1) {
378 mdb_warn("failed to read blkptr_t");
379 return (DCMD_ERR);
380 }
381
382 if (mdb_ctf_lookup_by_name("enum dmu_object_type", &type_enum) == -1 ||
383 mdb_ctf_lookup_by_name("enum zio_checksum", &checksum_enum) == -1 ||
384 mdb_ctf_lookup_by_name("enum zio_compress", &compress_enum) == -1) {
385 mdb_warn("Could not find blkptr enumerated types");
386 return (DCMD_ERR);
387 }
388
389 enum_lookup(type, sizeof (type), type_enum,
390 BP_GET_TYPE(bp), "DMU_OT_");
391 enum_lookup(checksum, sizeof (checksum), checksum_enum,
392 BP_GET_CHECKSUM(bp), "ZIO_CHECKSUM_");
393 enum_lookup(compress, sizeof (compress), compress_enum,
394 BP_GET_COMPRESS(bp), "ZIO_COMPRESS_");
395
396 SPRINTF_BLKPTR(mdb_snprintf, '\n', buf, bp, type, checksum, compress);
397
398 mdb_printf("%s\n", buf);
399
400 return (DCMD_OK);
401 }
402
403 /* ARGSUSED */
404 static int
405 dbuf(uintptr_t addr, uint_t flags, int argc, const mdb_arg_t *argv)
406 {
407 mdb_ctf_id_t id;
408 dmu_buf_t db;
409 uintptr_t objset;
410 uint8_t level;
411 uint64_t blkid;
412 uint64_t holds;
413 char objectname[32];
414 char blkidname[32];
415 char path[MAXNAMELEN];
416
417 if (DCMD_HDRSPEC(flags)) {
418 mdb_printf(" addr object lvl blkid holds os\n");
419 }
420
421 if (mdb_ctf_lookup_by_name("struct dmu_buf_impl", &id) == -1) {
422 mdb_warn("couldn't find struct dmu_buf_impl_t");
423 return (DCMD_ERR);
424 }
425
426 if (GETMEMBID(addr, &id, db_objset, objset) ||
427 GETMEMBID(addr, &id, db, db) ||
428 GETMEMBID(addr, &id, db_level, level) ||
429 GETMEMBID(addr, &id, db_blkid, blkid)) {
430 return (WALK_ERR);
431 }
432
433 if (getrefcount(addr, &id, "db_holds", &holds)) {
434 return (WALK_ERR);
435 }
436
437 if (db.db_object == DMU_META_DNODE_OBJECT)
438 (void) strcpy(objectname, "mdn");
439 else
440 (void) mdb_snprintf(objectname, sizeof (objectname), "%llx",
441 (u_longlong_t)db.db_object);
442
443 if (blkid == DMU_BONUS_BLKID)
444 (void) strcpy(blkidname, "bonus");
445 else
446 (void) mdb_snprintf(blkidname, sizeof (blkidname), "%llx",
447 (u_longlong_t)blkid);
448
449 if (objset_name(objset, path)) {
450 return (WALK_ERR);
451 }
452
453 mdb_printf("%p %8s %1u %9s %2llu %s\n",
454 addr, objectname, level, blkidname, holds, path);
455
456 return (DCMD_OK);
457 }
458
459 /* ARGSUSED */
460 static int
461 dbuf_stats(uintptr_t addr, uint_t flags, int argc, const mdb_arg_t *argv)
462 {
463 #define HISTOSZ 32
464 uintptr_t dbp;
465 dmu_buf_impl_t db;
466 dbuf_hash_table_t ht;
467 uint64_t bucket, ndbufs;
468 uint64_t histo[HISTOSZ];
469 uint64_t histo2[HISTOSZ];
470 int i, maxidx;
471
472 if (mdb_readvar(&ht, "dbuf_hash_table") == -1) {
473 mdb_warn("failed to read 'dbuf_hash_table'");
474 return (DCMD_ERR);
475 }
476
477 for (i = 0; i < HISTOSZ; i++) {
478 histo[i] = 0;
479 histo2[i] = 0;
480 }
481
482 ndbufs = 0;
483 for (bucket = 0; bucket < ht.hash_table_mask+1; bucket++) {
484 int len;
485
486 if (mdb_vread(&dbp, sizeof (void *),
487 (uintptr_t)(ht.hash_table+bucket)) == -1) {
488 mdb_warn("failed to read hash bucket %u at %p",
489 bucket, ht.hash_table+bucket);
490 return (DCMD_ERR);
491 }
492
493 len = 0;
494 while (dbp != 0) {
495 if (mdb_vread(&db, sizeof (dmu_buf_impl_t),
496 dbp) == -1) {
497 mdb_warn("failed to read dbuf at %p", dbp);
498 return (DCMD_ERR);
499 }
500 dbp = (uintptr_t)db.db_hash_next;
501 for (i = MIN(len, HISTOSZ - 1); i >= 0; i--)
502 histo2[i]++;
503 len++;
504 ndbufs++;
505 }
506
507 if (len >= HISTOSZ)
508 len = HISTOSZ-1;
509 histo[len]++;
510 }
511
512 mdb_printf("hash table has %llu buckets, %llu dbufs "
513 "(avg %llu buckets/dbuf)\n",
514 ht.hash_table_mask+1, ndbufs,
515 (ht.hash_table_mask+1)/ndbufs);
516
517 mdb_printf("\n");
518 maxidx = 0;
519 for (i = 0; i < HISTOSZ; i++)
520 if (histo[i] > 0)
521 maxidx = i;
522 mdb_printf("hash chain length number of buckets\n");
523 for (i = 0; i <= maxidx; i++)
524 mdb_printf("%u %llu\n", i, histo[i]);
525
526 mdb_printf("\n");
527 maxidx = 0;
528 for (i = 0; i < HISTOSZ; i++)
529 if (histo2[i] > 0)
530 maxidx = i;
531 mdb_printf("hash chain depth number of dbufs\n");
532 for (i = 0; i <= maxidx; i++)
533 mdb_printf("%u or more %llu %llu%%\n",
534 i, histo2[i], histo2[i]*100/ndbufs);
535
536
537 return (DCMD_OK);
538 }
539
540 #define CHAIN_END 0xffff
541 /*
542 * ::zap_leaf [-v]
543 *
544 * Print a zap_leaf_phys_t, assumed to be 16k
545 */
546 /* ARGSUSED */
547 static int
548 zap_leaf(uintptr_t addr, uint_t flags, int argc, const mdb_arg_t *argv)
549 {
550 char buf[16*1024];
551 int verbose = B_FALSE;
552 int four = B_FALSE;
553 zap_leaf_t l;
554 zap_leaf_phys_t *zlp = (void *)buf;
555 int i;
556
557 if (mdb_getopts(argc, argv,
558 'v', MDB_OPT_SETBITS, TRUE, &verbose,
559 '4', MDB_OPT_SETBITS, TRUE, &four,
560 NULL) != argc)
561 return (DCMD_USAGE);
562
563 l.l_phys = zlp;
564 l.l_bs = 14; /* assume 16k blocks */
565 if (four)
566 l.l_bs = 12;
567
568 if (!(flags & DCMD_ADDRSPEC)) {
569 return (DCMD_USAGE);
570 }
571
572 if (mdb_vread(buf, sizeof (buf), addr) == -1) {
573 mdb_warn("failed to read zap_leaf_phys_t at %p", addr);
574 return (DCMD_ERR);
575 }
576
577 if (zlp->l_hdr.lh_block_type != ZBT_LEAF ||
578 zlp->l_hdr.lh_magic != ZAP_LEAF_MAGIC) {
579 mdb_warn("This does not appear to be a zap_leaf_phys_t");
580 return (DCMD_ERR);
581 }
582
583 mdb_printf("zap_leaf_phys_t at %p:\n", addr);
584 mdb_printf(" lh_prefix_len = %u\n", zlp->l_hdr.lh_prefix_len);
585 mdb_printf(" lh_prefix = %llx\n", zlp->l_hdr.lh_prefix);
586 mdb_printf(" lh_nentries = %u\n", zlp->l_hdr.lh_nentries);
587 mdb_printf(" lh_nfree = %u\n", zlp->l_hdr.lh_nfree,
588 zlp->l_hdr.lh_nfree * 100 / (ZAP_LEAF_NUMCHUNKS(&l)));
589 mdb_printf(" lh_freelist = %u\n", zlp->l_hdr.lh_freelist);
590 mdb_printf(" lh_flags = %x (%s)\n", zlp->l_hdr.lh_flags,
591 zlp->l_hdr.lh_flags & ZLF_ENTRIES_CDSORTED ?
592 "ENTRIES_CDSORTED" : "");
593
594 if (verbose) {
595 mdb_printf(" hash table:\n");
596 for (i = 0; i < ZAP_LEAF_HASH_NUMENTRIES(&l); i++) {
597 if (zlp->l_hash[i] != CHAIN_END)
598 mdb_printf(" %u: %u\n", i, zlp->l_hash[i]);
599 }
600 }
601
602 mdb_printf(" chunks:\n");
603 for (i = 0; i < ZAP_LEAF_NUMCHUNKS(&l); i++) {
604 /* LINTED: alignment */
605 zap_leaf_chunk_t *zlc = &ZAP_LEAF_CHUNK(&l, i);
606 switch (zlc->l_entry.le_type) {
607 case ZAP_CHUNK_FREE:
608 if (verbose) {
609 mdb_printf(" %u: free; lf_next = %u\n",
610 i, zlc->l_free.lf_next);
611 }
612 break;
613 case ZAP_CHUNK_ENTRY:
614 mdb_printf(" %u: entry\n", i);
615 if (verbose) {
616 mdb_printf(" le_next = %u\n",
617 zlc->l_entry.le_next);
618 }
619 mdb_printf(" le_name_chunk = %u\n",
620 zlc->l_entry.le_name_chunk);
621 mdb_printf(" le_name_numints = %u\n",
622 zlc->l_entry.le_name_numints);
623 mdb_printf(" le_value_chunk = %u\n",
624 zlc->l_entry.le_value_chunk);
625 mdb_printf(" le_value_intlen = %u\n",
626 zlc->l_entry.le_value_intlen);
627 mdb_printf(" le_value_numints = %u\n",
628 zlc->l_entry.le_value_numints);
629 mdb_printf(" le_cd = %u\n",
630 zlc->l_entry.le_cd);
631 mdb_printf(" le_hash = %llx\n",
632 zlc->l_entry.le_hash);
633 break;
634 case ZAP_CHUNK_ARRAY:
635 mdb_printf(" %u: array", i);
636 if (strisprint((char *)zlc->l_array.la_array))
637 mdb_printf(" \"%s\"", zlc->l_array.la_array);
638 mdb_printf("\n");
639 if (verbose) {
640 int j;
641 mdb_printf(" ");
642 for (j = 0; j < ZAP_LEAF_ARRAY_BYTES; j++) {
643 mdb_printf("%02x ",
644 zlc->l_array.la_array[j]);
645 }
646 mdb_printf("\n");
647 }
648 if (zlc->l_array.la_next != CHAIN_END) {
649 mdb_printf(" lf_next = %u\n",
650 zlc->l_array.la_next);
651 }
652 break;
653 default:
654 mdb_printf(" %u: undefined type %u\n",
655 zlc->l_entry.le_type);
656 }
657 }
658
659 return (DCMD_OK);
660 }
661
662 typedef struct dbufs_data {
663 mdb_ctf_id_t id;
664 uint64_t objset;
665 uint64_t object;
666 uint64_t level;
667 uint64_t blkid;
668 char *osname;
669 } dbufs_data_t;
670
671 #define DBUFS_UNSET (0xbaddcafedeadbeefULL)
672
673 /* ARGSUSED */
674 static int
675 dbufs_cb(uintptr_t addr, const void *unknown, void *arg)
676 {
677 dbufs_data_t *data = arg;
678 uintptr_t objset;
679 dmu_buf_t db;
680 uint8_t level;
681 uint64_t blkid;
682 char osname[MAXNAMELEN];
683
684 if (GETMEMBID(addr, &data->id, db_objset, objset) ||
685 GETMEMBID(addr, &data->id, db, db) ||
686 GETMEMBID(addr, &data->id, db_level, level) ||
687 GETMEMBID(addr, &data->id, db_blkid, blkid)) {
688 return (WALK_ERR);
689 }
690
691 if ((data->objset == DBUFS_UNSET || data->objset == objset) &&
692 (data->osname == NULL || (objset_name(objset, osname) == 0 &&
693 strcmp(data->osname, osname) == 0)) &&
694 (data->object == DBUFS_UNSET || data->object == db.db_object) &&
695 (data->level == DBUFS_UNSET || data->level == level) &&
696 (data->blkid == DBUFS_UNSET || data->blkid == blkid)) {
697 mdb_printf("%#lr\n", addr);
698 }
699 return (WALK_NEXT);
700 }
701
702 /* ARGSUSED */
703 static int
704 dbufs(uintptr_t addr, uint_t flags, int argc, const mdb_arg_t *argv)
705 {
706 dbufs_data_t data;
707 char *object = NULL;
708 char *blkid = NULL;
709
710 data.objset = data.object = data.level = data.blkid = DBUFS_UNSET;
711 data.osname = NULL;
712
713 if (mdb_getopts(argc, argv,
714 'O', MDB_OPT_UINT64, &data.objset,
715 'n', MDB_OPT_STR, &data.osname,
716 'o', MDB_OPT_STR, &object,
717 'l', MDB_OPT_UINT64, &data.level,
718 'b', MDB_OPT_STR, &blkid) != argc) {
719 return (DCMD_USAGE);
720 }
721
722 if (object) {
723 if (strcmp(object, "mdn") == 0) {
724 data.object = DMU_META_DNODE_OBJECT;
725 } else {
726 data.object = mdb_strtoull(object);
727 }
728 }
729
730 if (blkid) {
731 if (strcmp(blkid, "bonus") == 0) {
732 data.blkid = DMU_BONUS_BLKID;
733 } else {
734 data.blkid = mdb_strtoull(blkid);
735 }
736 }
737
738 if (mdb_ctf_lookup_by_name("struct dmu_buf_impl", &data.id) == -1) {
739 mdb_warn("couldn't find struct dmu_buf_impl_t");
740 return (DCMD_ERR);
741 }
742
743 if (mdb_walk("dmu_buf_impl_t", dbufs_cb, &data) != 0) {
744 mdb_warn("can't walk dbufs");
745 return (DCMD_ERR);
746 }
747
748 return (DCMD_OK);
749 }
750
751 typedef struct abuf_find_data {
752 dva_t dva;
753 mdb_ctf_id_t id;
754 } abuf_find_data_t;
755
756 /* ARGSUSED */
757 static int
758 abuf_find_cb(uintptr_t addr, const void *unknown, void *arg)
759 {
760 abuf_find_data_t *data = arg;
761 dva_t dva;
762
763 if (GETMEMBID(addr, &data->id, b_dva, dva)) {
764 return (WALK_ERR);
765 }
766
767 if (dva.dva_word[0] == data->dva.dva_word[0] &&
768 dva.dva_word[1] == data->dva.dva_word[1]) {
769 mdb_printf("%#lr\n", addr);
770 }
771 return (WALK_NEXT);
772 }
773
774 /* ARGSUSED */
775 static int
776 abuf_find(uintptr_t addr, uint_t flags, int argc, const mdb_arg_t *argv)
777 {
778 abuf_find_data_t data;
779 GElf_Sym sym;
780 int i;
781 const char *syms[] = {
782 "ARC_mru",
783 "ARC_mru_ghost",
784 "ARC_mfu",
785 "ARC_mfu_ghost",
786 };
787
788 if (argc != 2)
789 return (DCMD_USAGE);
790
791 for (i = 0; i < 2; i ++) {
792 switch (argv[i].a_type) {
793 case MDB_TYPE_STRING:
794 data.dva.dva_word[i] = mdb_strtoull(argv[i].a_un.a_str);
795 break;
796 case MDB_TYPE_IMMEDIATE:
797 data.dva.dva_word[i] = argv[i].a_un.a_val;
798 break;
799 default:
800 return (DCMD_USAGE);
801 }
802 }
803
804 if (mdb_ctf_lookup_by_name("struct arc_buf_hdr", &data.id) == -1) {
805 mdb_warn("couldn't find struct arc_buf_hdr");
806 return (DCMD_ERR);
807 }
808
809 for (i = 0; i < sizeof (syms) / sizeof (syms[0]); i++) {
810 if (mdb_lookup_by_name(syms[i], &sym)) {
811 mdb_warn("can't find symbol %s", syms[i]);
812 return (DCMD_ERR);
813 }
814
815 if (mdb_pwalk("list", abuf_find_cb, &data, sym.st_value) != 0) {
816 mdb_warn("can't walk %s", syms[i]);
817 return (DCMD_ERR);
818 }
819 }
820
821 return (DCMD_OK);
822 }
823
824 /* ARGSUSED */
825 static int
826 dbgmsg_cb(uintptr_t addr, const void *unknown, void *arg)
827 {
828 static mdb_ctf_id_t id;
829 static boolean_t gotid;
830 static ulong_t off;
831
832 int *verbosep = arg;
833 time_t timestamp;
834 char buf[1024];
835
836 if (!gotid) {
837 if (mdb_ctf_lookup_by_name("struct zfs_dbgmsg", &id) == -1) {
838 mdb_warn("couldn't find struct zfs_dbgmsg");
839 return (WALK_ERR);
840 }
841 gotid = TRUE;
842 if (mdb_ctf_offsetof(id, "zdm_msg", &off) == -1) {
843 mdb_warn("couldn't find zdm_msg");
844 return (WALK_ERR);
845 }
846 off /= 8;
847 }
848
849
850 if (GETMEMBID(addr, &id, zdm_timestamp, timestamp)) {
851 return (WALK_ERR);
852 }
853
854 if (mdb_readstr(buf, sizeof (buf), addr + off) == -1) {
855 mdb_warn("failed to read zdm_msg at %p\n", addr + off);
856 return (DCMD_ERR);
857 }
858
859 if (*verbosep)
860 mdb_printf("%Y ", timestamp);
861
862 mdb_printf("%s\n", buf);
863
864 if (*verbosep)
865 (void) mdb_call_dcmd("whatis", addr, DCMD_ADDRSPEC, 0, NULL);
866
867 return (WALK_NEXT);
868 }
869
870 /* ARGSUSED */
871 static int
872 dbgmsg(uintptr_t addr, uint_t flags, int argc, const mdb_arg_t *argv)
873 {
874 GElf_Sym sym;
875 int verbose = FALSE;
876
877 if (mdb_getopts(argc, argv,
878 'v', MDB_OPT_SETBITS, TRUE, &verbose,
879 NULL) != argc)
880 return (DCMD_USAGE);
881
882 if (mdb_lookup_by_name("zfs_dbgmsgs", &sym)) {
883 mdb_warn("can't find zfs_dbgmsgs");
884 return (DCMD_ERR);
885 }
886
887 if (mdb_pwalk("list", dbgmsg_cb, &verbose, sym.st_value) != 0) {
888 mdb_warn("can't walk zfs_dbgmsgs");
889 return (DCMD_ERR);
890 }
891
892 return (DCMD_OK);
893 }
894
895 /*ARGSUSED*/
896 static int
897 arc_print(uintptr_t addr, uint_t flags, int argc, const mdb_arg_t *argv)
898 {
899 kstat_named_t *stats;
900 GElf_Sym sym;
901 int nstats, i;
902 uint_t opt_a = FALSE;
903 uint_t opt_b = FALSE;
904 uint_t shift = 0;
905 const char *suffix;
906
907 static const char *bytestats[] = {
908 "p", "c", "c_min", "c_max", "size", NULL
909 };
910
911 static const char *extras[] = {
912 "arc_no_grow", "arc_tempreserve",
913 "arc_meta_used", "arc_meta_limit", "arc_meta_max",
914 NULL
915 };
916
917 if (mdb_lookup_by_name("arc_stats", &sym) == -1) {
918 mdb_warn("failed to find 'arc_stats'");
919 return (DCMD_ERR);
920 }
921
922 stats = mdb_zalloc(sym.st_size, UM_SLEEP | UM_GC);
923
924 if (mdb_vread(stats, sym.st_size, sym.st_value) == -1) {
925 mdb_warn("couldn't read 'arc_stats' at %p", sym.st_value);
926 return (DCMD_ERR);
927 }
928
929 nstats = sym.st_size / sizeof (kstat_named_t);
930
931 /* NB: -a / opt_a are ignored for backwards compatability */
932 if (mdb_getopts(argc, argv,
933 'a', MDB_OPT_SETBITS, TRUE, &opt_a,
934 'b', MDB_OPT_SETBITS, TRUE, &opt_b,
935 'k', MDB_OPT_SETBITS, 10, &shift,
936 'm', MDB_OPT_SETBITS, 20, &shift,
937 'g', MDB_OPT_SETBITS, 30, &shift,
938 NULL) != argc)
939 return (DCMD_USAGE);
940
941 if (!opt_b && !shift)
942 shift = 20;
943
944 switch (shift) {
945 case 0:
946 suffix = "B";
947 break;
948 case 10:
949 suffix = "KB";
950 break;
951 case 20:
952 suffix = "MB";
953 break;
954 case 30:
955 suffix = "GB";
956 break;
957 default:
958 suffix = "XX";
959 }
960
961 for (i = 0; i < nstats; i++) {
962 int j;
963 boolean_t bytes = B_FALSE;
964
965 for (j = 0; bytestats[j]; j++) {
966 if (strcmp(stats[i].name, bytestats[j]) == 0) {
967 bytes = B_TRUE;
968 break;
969 }
970 }
971
972 if (bytes) {
973 mdb_printf("%-25s = %9llu %s\n", stats[i].name,
974 stats[i].value.ui64 >> shift, suffix);
975 } else {
976 mdb_printf("%-25s = %9llu\n", stats[i].name,
977 stats[i].value.ui64);
978 }
979 }
980
981 for (i = 0; extras[i]; i++) {
982 uint64_t buf;
983
984 if (mdb_lookup_by_name(extras[i], &sym) == -1) {
985 mdb_warn("failed to find '%s'", extras[i]);
986 return (DCMD_ERR);
987 }
988
989 if (sym.st_size != sizeof (uint64_t) &&
990 sym.st_size != sizeof (uint32_t)) {
991 mdb_warn("expected scalar for variable '%s'\n",
992 extras[i]);
993 return (DCMD_ERR);
994 }
995
996 if (mdb_vread(&buf, sym.st_size, sym.st_value) == -1) {
997 mdb_warn("couldn't read '%s'", extras[i]);
998 return (DCMD_ERR);
999 }
1000
1001 mdb_printf("%-25s = ", extras[i]);
1002
1003 /* NB: all the 64-bit extras happen to be byte counts */
1004 if (sym.st_size == sizeof (uint64_t))
1005 mdb_printf("%9llu %s\n", buf >> shift, suffix);
1006
1007 if (sym.st_size == sizeof (uint32_t))
1008 mdb_printf("%9d\n", *((uint32_t *)&buf));
1009 }
1010 return (DCMD_OK);
1011 }
1012
1013 /*
1014 * ::spa
1015 *
1016 * -c Print configuration information as well
1017 * -v Print vdev state
1018 * -e Print vdev error stats
1019 *
1020 * Print a summarized spa_t. When given no arguments, prints out a table of all
1021 * active pools on the system.
1022 */
1023 /* ARGSUSED */
1024 static int
1025 spa_print(uintptr_t addr, uint_t flags, int argc, const mdb_arg_t *argv)
1026 {
1027 spa_t spa;
1028 const char *statetab[] = { "ACTIVE", "EXPORTED", "DESTROYED",
1029 "SPARE", "L2CACHE", "UNINIT", "UNAVAIL", "POTENTIAL" };
1030 const char *state;
1031 int config = FALSE;
1032 int vdevs = FALSE;
1033 int errors = FALSE;
1034
1035 if (mdb_getopts(argc, argv,
1036 'c', MDB_OPT_SETBITS, TRUE, &config,
1037 'v', MDB_OPT_SETBITS, TRUE, &vdevs,
1038 'e', MDB_OPT_SETBITS, TRUE, &errors,
1039 NULL) != argc)
1040 return (DCMD_USAGE);
1041
1042 if (!(flags & DCMD_ADDRSPEC)) {
1043 if (mdb_walk_dcmd("spa", "spa", argc, argv) == -1) {
1044 mdb_warn("can't walk spa");
1045 return (DCMD_ERR);
1046 }
1047
1048 return (DCMD_OK);
1049 }
1050
1051 if (flags & DCMD_PIPE_OUT) {
1052 mdb_printf("%#lr\n", addr);
1053 return (DCMD_OK);
1054 }
1055
1056 if (DCMD_HDRSPEC(flags))
1057 mdb_printf("%<u>%-?s %9s %-*s%</u>\n", "ADDR", "STATE",
1058 sizeof (uintptr_t) == 4 ? 60 : 52, "NAME");
1059
1060 if (mdb_vread(&spa, sizeof (spa), addr) == -1) {
1061 mdb_warn("failed to read spa_t at %p", addr);
1062 return (DCMD_ERR);
1063 }
1064
1065 if (spa.spa_state < 0 || spa.spa_state > POOL_STATE_UNAVAIL)
1066 state = "UNKNOWN";
1067 else
1068 state = statetab[spa.spa_state];
1069
1070 mdb_printf("%0?p %9s %s\n", addr, state, spa.spa_name);
1071
1072 if (config) {
1073 mdb_printf("\n");
1074 mdb_inc_indent(4);
1075 if (mdb_call_dcmd("spa_config", addr, flags, 0,
1076 NULL) != DCMD_OK)
1077 return (DCMD_ERR);
1078 mdb_dec_indent(4);
1079 }
1080
1081 if (vdevs || errors) {
1082 mdb_arg_t v;
1083
1084 v.a_type = MDB_TYPE_STRING;
1085 v.a_un.a_str = "-e";
1086
1087 mdb_printf("\n");
1088 mdb_inc_indent(4);
1089 if (mdb_call_dcmd("spa_vdevs", addr, flags, errors ? 1 : 0,
1090 &v) != DCMD_OK)
1091 return (DCMD_ERR);
1092 mdb_dec_indent(4);
1093 }
1094
1095 return (DCMD_OK);
1096 }
1097
1098 /*
1099 * ::spa_config
1100 *
1101 * Given a spa_t, print the configuration information stored in spa_config.
1102 * Since it's just an nvlist, format it as an indented list of name=value pairs.
1103 * We simply read the value of spa_config and pass off to ::nvlist.
1104 */
1105 /* ARGSUSED */
1106 static int
1107 spa_print_config(uintptr_t addr, uint_t flags, int argc, const mdb_arg_t *argv)
1108 {
1109 spa_t spa;
1110
1111 if (argc != 0 || !(flags & DCMD_ADDRSPEC))
1112 return (DCMD_USAGE);
1113
1114 if (mdb_vread(&spa, sizeof (spa), addr) == -1) {
1115 mdb_warn("failed to read spa_t at %p", addr);
1116 return (DCMD_ERR);
1117 }
1118
1119 if (spa.spa_config == NULL) {
1120 mdb_printf("(none)\n");
1121 return (DCMD_OK);
1122 }
1123
1124 return (mdb_call_dcmd("nvlist", (uintptr_t)spa.spa_config, flags,
1125 0, NULL));
1126 }
1127
1128 /*
1129 * ::vdev
1130 *
1131 * Print out a summarized vdev_t, in the following form:
1132 *
1133 * ADDR STATE AUX DESC
1134 * fffffffbcde23df0 HEALTHY - /dev/dsk/c0t0d0
1135 *
1136 * If '-r' is specified, recursively visit all children.
1137 *
1138 * With '-e', the statistics associated with the vdev are printed as well.
1139 */
1140 static int
1141 do_print_vdev(uintptr_t addr, int flags, int depth, int stats,
1142 int recursive)
1143 {
1144 vdev_t vdev;
1145 char desc[MAXNAMELEN];
1146 int c, children;
1147 uintptr_t *child;
1148 const char *state, *aux;
1149
1150 if (mdb_vread(&vdev, sizeof (vdev), (uintptr_t)addr) == -1) {
1151 mdb_warn("failed to read vdev_t at %p\n", (uintptr_t)addr);
1152 return (DCMD_ERR);
1153 }
1154
1155 if (flags & DCMD_PIPE_OUT) {
1156 mdb_printf("%#lr", addr);
1157 } else {
1158 if (vdev.vdev_path != NULL) {
1159 if (mdb_readstr(desc, sizeof (desc),
1160 (uintptr_t)vdev.vdev_path) == -1) {
1161 mdb_warn("failed to read vdev_path at %p\n",
1162 vdev.vdev_path);
1163 return (DCMD_ERR);
1164 }
1165 } else if (vdev.vdev_ops != NULL) {
1166 vdev_ops_t ops;
1167 if (mdb_vread(&ops, sizeof (ops),
1168 (uintptr_t)vdev.vdev_ops) == -1) {
1169 mdb_warn("failed to read vdev_ops at %p\n",
1170 vdev.vdev_ops);
1171 return (DCMD_ERR);
1172 }
1173 (void) strcpy(desc, ops.vdev_op_type);
1174 } else {
1175 (void) strcpy(desc, "<unknown>");
1176 }
1177
1178 if (depth == 0 && DCMD_HDRSPEC(flags))
1179 mdb_printf("%<u>%-?s %-9s %-12s %-*s%</u>\n",
1180 "ADDR", "STATE", "AUX",
1181 sizeof (uintptr_t) == 4 ? 43 : 35,
1182 "DESCRIPTION");
1183
1184 mdb_printf("%0?p ", addr);
1185
1186 switch (vdev.vdev_state) {
1187 case VDEV_STATE_CLOSED:
1188 state = "CLOSED";
1189 break;
1190 case VDEV_STATE_OFFLINE:
1191 state = "OFFLINE";
1192 break;
1193 case VDEV_STATE_CANT_OPEN:
1194 state = "CANT_OPEN";
1195 break;
1196 case VDEV_STATE_DEGRADED:
1197 state = "DEGRADED";
1198 break;
1199 case VDEV_STATE_HEALTHY:
1200 state = "HEALTHY";
1201 break;
1202 case VDEV_STATE_REMOVED:
1203 state = "REMOVED";
1204 break;
1205 case VDEV_STATE_FAULTED:
1206 state = "FAULTED";
1207 break;
1208 default:
1209 state = "UNKNOWN";
1210 break;
1211 }
1212
1213 switch (vdev.vdev_stat.vs_aux) {
1214 case VDEV_AUX_NONE:
1215 aux = "-";
1216 break;
1217 case VDEV_AUX_OPEN_FAILED:
1218 aux = "OPEN_FAILED";
1219 break;
1220 case VDEV_AUX_CORRUPT_DATA:
1221 aux = "CORRUPT_DATA";
1222 break;
1223 case VDEV_AUX_NO_REPLICAS:
1224 aux = "NO_REPLICAS";
1225 break;
1226 case VDEV_AUX_BAD_GUID_SUM:
1227 aux = "BAD_GUID_SUM";
1228 break;
1229 case VDEV_AUX_TOO_SMALL:
1230 aux = "TOO_SMALL";
1231 break;
1232 case VDEV_AUX_BAD_LABEL:
1233 aux = "BAD_LABEL";
1234 break;
1235 case VDEV_AUX_VERSION_NEWER:
1236 aux = "VERS_NEWER";
1237 break;
1238 case VDEV_AUX_VERSION_OLDER:
1239 aux = "VERS_OLDER";
1240 break;
1241 case VDEV_AUX_UNSUP_FEAT:
1242 aux = "UNSUP_FEAT";
1243 break;
1244 case VDEV_AUX_SPARED:
1245 aux = "SPARED";
1246 break;
1247 case VDEV_AUX_ERR_EXCEEDED:
1248 aux = "ERR_EXCEEDED";
1249 break;
1250 case VDEV_AUX_IO_FAILURE:
1251 aux = "IO_FAILURE";
1252 break;
1253 case VDEV_AUX_BAD_LOG:
1254 aux = "BAD_LOG";
1255 break;
1256 case VDEV_AUX_EXTERNAL:
1257 aux = "EXTERNAL";
1258 break;
1259 case VDEV_AUX_SPLIT_POOL:
1260 aux = "SPLIT_POOL";
1261 break;
1262 default:
1263 aux = "UNKNOWN";
1264 break;
1265 }
1266
1267 mdb_printf("%-9s %-12s %*s%s\n", state, aux, depth, "", desc);
1268
1269 if (stats) {
1270 vdev_stat_t *vs = &vdev.vdev_stat;
1271 int i;
1272
1273 mdb_inc_indent(4);
1274 mdb_printf("\n");
1275 mdb_printf("%<u> %12s %12s %12s %12s "
1276 "%12s%</u>\n", "READ", "WRITE", "FREE", "CLAIM",
1277 "IOCTL");
1278 mdb_printf("OPS ");
1279 for (i = 1; i < ZIO_TYPES; i++)
1280 mdb_printf("%11#llx%s", vs->vs_ops[i],
1281 i == ZIO_TYPES - 1 ? "" : " ");
1282 mdb_printf("\n");
1283 mdb_printf("BYTES ");
1284 for (i = 1; i < ZIO_TYPES; i++)
1285 mdb_printf("%11#llx%s", vs->vs_bytes[i],
1286 i == ZIO_TYPES - 1 ? "" : " ");
1287
1288
1289 mdb_printf("\n");
1290 mdb_printf("EREAD %10#llx\n", vs->vs_read_errors);
1291 mdb_printf("EWRITE %10#llx\n", vs->vs_write_errors);
1292 mdb_printf("ECKSUM %10#llx\n",
1293 vs->vs_checksum_errors);
1294 mdb_dec_indent(4);
1295 }
1296
1297 if (stats)
1298 mdb_printf("\n");
1299 }
1300
1301 children = vdev.vdev_children;
1302
1303 if (children == 0 || !recursive)
1304 return (DCMD_OK);
1305
1306 child = mdb_alloc(children * sizeof (void *), UM_SLEEP | UM_GC);
1307 if (mdb_vread(child, children * sizeof (void *),
1308 (uintptr_t)vdev.vdev_child) == -1) {
1309 mdb_warn("failed to read vdev children at %p", vdev.vdev_child);
1310 return (DCMD_ERR);
1311 }
1312
1313 for (c = 0; c < children; c++) {
1314 if (do_print_vdev(child[c], flags, depth + 2, stats,
1315 recursive))
1316 return (DCMD_ERR);
1317 }
1318
1319 return (DCMD_OK);
1320 }
1321
1322 static int
1323 vdev_print(uintptr_t addr, uint_t flags, int argc, const mdb_arg_t *argv)
1324 {
1325 int recursive = FALSE;
1326 int stats = FALSE;
1327 uint64_t depth = 0;
1328
1329 if (mdb_getopts(argc, argv,
1330 'r', MDB_OPT_SETBITS, TRUE, &recursive,
1331 'e', MDB_OPT_SETBITS, TRUE, &stats,
1332 'd', MDB_OPT_UINT64, &depth,
1333 NULL) != argc)
1334 return (DCMD_USAGE);
1335
1336 if (!(flags & DCMD_ADDRSPEC)) {
1337 mdb_warn("no vdev_t address given\n");
1338 return (DCMD_ERR);
1339 }
1340
1341 return (do_print_vdev(addr, flags, (int)depth, stats, recursive));
1342 }
1343
1344 typedef struct metaslab_walk_data {
1345 uint64_t mw_numvdevs;
1346 uintptr_t *mw_vdevs;
1347 int mw_curvdev;
1348 uint64_t mw_nummss;
1349 uintptr_t *mw_mss;
1350 int mw_curms;
1351 } metaslab_walk_data_t;
1352
1353 static int
1354 metaslab_walk_step(mdb_walk_state_t *wsp)
1355 {
1356 metaslab_walk_data_t *mw = wsp->walk_data;
1357 metaslab_t ms;
1358 uintptr_t msp;
1359
1360 if (mw->mw_curvdev >= mw->mw_numvdevs)
1361 return (WALK_DONE);
1362
1363 if (mw->mw_mss == NULL) {
1364 uintptr_t mssp;
1365 uintptr_t vdevp;
1366
1367 ASSERT(mw->mw_curms == 0);
1368 ASSERT(mw->mw_nummss == 0);
1369
1370 vdevp = mw->mw_vdevs[mw->mw_curvdev];
1371 if (GETMEMB(vdevp, struct vdev, vdev_ms, mssp) ||
1372 GETMEMB(vdevp, struct vdev, vdev_ms_count, mw->mw_nummss)) {
1373 return (WALK_ERR);
1374 }
1375
1376 mw->mw_mss = mdb_alloc(mw->mw_nummss * sizeof (void*),
1377 UM_SLEEP | UM_GC);
1378 if (mdb_vread(mw->mw_mss, mw->mw_nummss * sizeof (void*),
1379 mssp) == -1) {
1380 mdb_warn("failed to read vdev_ms at %p", mssp);
1381 return (WALK_ERR);
1382 }
1383 }
1384
1385 if (mw->mw_curms >= mw->mw_nummss) {
1386 mw->mw_mss = NULL;
1387 mw->mw_curms = 0;
1388 mw->mw_nummss = 0;
1389 mw->mw_curvdev++;
1390 return (WALK_NEXT);
1391 }
1392
1393 msp = mw->mw_mss[mw->mw_curms];
1394 if (mdb_vread(&ms, sizeof (metaslab_t), msp) == -1) {
1395 mdb_warn("failed to read metaslab_t at %p", msp);
1396 return (WALK_ERR);
1397 }
1398
1399 mw->mw_curms++;
1400
1401 return (wsp->walk_callback(msp, &ms, wsp->walk_cbdata));
1402 }
1403
1404 /* ARGSUSED */
1405 static int
1406 metaslab_walk_init(mdb_walk_state_t *wsp)
1407 {
1408 metaslab_walk_data_t *mw;
1409 uintptr_t root_vdevp;
1410 uintptr_t childp;
1411
1412 if (wsp->walk_addr == NULL) {
1413 mdb_warn("must supply address of spa_t\n");
1414 return (WALK_ERR);
1415 }
1416
1417 mw = mdb_zalloc(sizeof (metaslab_walk_data_t), UM_SLEEP | UM_GC);
1418
1419 if (GETMEMB(wsp->walk_addr, struct spa, spa_root_vdev, root_vdevp) ||
1420 GETMEMB(root_vdevp, struct vdev, vdev_children, mw->mw_numvdevs) ||
1421 GETMEMB(root_vdevp, struct vdev, vdev_child, childp)) {
1422 return (DCMD_ERR);
1423 }
1424
1425 mw->mw_vdevs = mdb_alloc(mw->mw_numvdevs * sizeof (void *),
1426 UM_SLEEP | UM_GC);
1427 if (mdb_vread(mw->mw_vdevs, mw->mw_numvdevs * sizeof (void *),
1428 childp) == -1) {
1429 mdb_warn("failed to read root vdev children at %p", childp);
1430 return (DCMD_ERR);
1431 }
1432
1433 wsp->walk_data = mw;
1434
1435 return (WALK_NEXT);
1436 }
1437
1438 typedef struct mdb_spa {
1439 uintptr_t spa_dsl_pool;
1440 uintptr_t spa_root_vdev;
1441 } mdb_spa_t;
1442
1443 typedef struct mdb_dsl_dir {
1444 uintptr_t dd_phys;
1445 int64_t dd_space_towrite[TXG_SIZE];
1446 } mdb_dsl_dir_t;
1447
1448 typedef struct mdb_dsl_dir_phys {
1449 uint64_t dd_used_bytes;
1450 uint64_t dd_compressed_bytes;
1451 uint64_t dd_uncompressed_bytes;
1452 } mdb_dsl_dir_phys_t;
1453
1454 typedef struct mdb_vdev {
1455 uintptr_t vdev_parent;
1456 uintptr_t vdev_ms;
1457 uint64_t vdev_ms_count;
1458 vdev_stat_t vdev_stat;
1459 } mdb_vdev_t;
1460
1461 typedef struct mdb_metaslab {
1462 space_map_t ms_allocmap[TXG_SIZE];
1463 space_map_t ms_freemap[TXG_SIZE];
1464 space_map_t ms_map;
1465 space_map_obj_t ms_smo;
1466 space_map_obj_t ms_smo_syncing;
1467 } mdb_metaslab_t;
1468
1469 typedef struct space_data {
1470 uint64_t ms_allocmap[TXG_SIZE];
1471 uint64_t ms_freemap[TXG_SIZE];
1472 uint64_t ms_map;
1473 uint64_t avail;
1474 uint64_t nowavail;
1475 } space_data_t;
1476
1477 /* ARGSUSED */
1478 static int
1479 space_cb(uintptr_t addr, const void *unknown, void *arg)
1480 {
1481 space_data_t *sd = arg;
1482 mdb_metaslab_t ms;
1483
1484 if (GETMEMB(addr, struct metaslab, ms_allocmap, ms.ms_allocmap) ||
1485 GETMEMB(addr, struct metaslab, ms_freemap, ms.ms_freemap) ||
1486 GETMEMB(addr, struct metaslab, ms_map, ms.ms_map) ||
1487 GETMEMB(addr, struct metaslab, ms_smo, ms.ms_smo) ||
1488 GETMEMB(addr, struct metaslab, ms_smo_syncing, ms.ms_smo_syncing)) {
1489 return (WALK_ERR);
1490 }
1491
1492 sd->ms_allocmap[0] += ms.ms_allocmap[0].sm_space;
1493 sd->ms_allocmap[1] += ms.ms_allocmap[1].sm_space;
1494 sd->ms_allocmap[2] += ms.ms_allocmap[2].sm_space;
1495 sd->ms_allocmap[3] += ms.ms_allocmap[3].sm_space;
1496 sd->ms_freemap[0] += ms.ms_freemap[0].sm_space;
1497 sd->ms_freemap[1] += ms.ms_freemap[1].sm_space;
1498 sd->ms_freemap[2] += ms.ms_freemap[2].sm_space;
1499 sd->ms_freemap[3] += ms.ms_freemap[3].sm_space;
1500 sd->ms_map += ms.ms_map.sm_space;
1501 sd->avail += ms.ms_map.sm_size - ms.ms_smo.smo_alloc;
1502 sd->nowavail += ms.ms_map.sm_size - ms.ms_smo_syncing.smo_alloc;
1503
1504 return (WALK_NEXT);
1505 }
1506
1507 /*
1508 * ::spa_space [-b]
1509 *
1510 * Given a spa_t, print out it's on-disk space usage and in-core
1511 * estimates of future usage. If -b is given, print space in bytes.
1512 * Otherwise print in megabytes.
1513 */
1514 /* ARGSUSED */
1515 static int
1516 spa_space(uintptr_t addr, uint_t flags, int argc, const mdb_arg_t *argv)
1517 {
1518 mdb_spa_t spa;
1519 uintptr_t dp_root_dir;
1520 mdb_dsl_dir_t dd;
1521 mdb_dsl_dir_phys_t dsp;
1522 uint64_t children;
1523 uintptr_t childaddr;
1524 space_data_t sd;
1525 int shift = 20;
1526 char *suffix = "M";
1527 int bits = FALSE;
1528
1529 if (mdb_getopts(argc, argv, 'b', MDB_OPT_SETBITS, TRUE, &bits, NULL) !=
1530 argc)
1531 return (DCMD_USAGE);
1532 if (!(flags & DCMD_ADDRSPEC))
1533 return (DCMD_USAGE);
1534
1535 if (bits) {
1536 shift = 0;
1537 suffix = "";
1538 }
1539
1540 if (GETMEMB(addr, struct spa, spa_dsl_pool, spa.spa_dsl_pool) ||
1541 GETMEMB(addr, struct spa, spa_root_vdev, spa.spa_root_vdev) ||
1542 GETMEMB(spa.spa_root_vdev, struct vdev, vdev_children, children) ||
1543 GETMEMB(spa.spa_root_vdev, struct vdev, vdev_child, childaddr) ||
1544 GETMEMB(spa.spa_dsl_pool, struct dsl_pool,
1545 dp_root_dir, dp_root_dir) ||
1546 GETMEMB(dp_root_dir, struct dsl_dir, dd_phys, dd.dd_phys) ||
1547 GETMEMB(dp_root_dir, struct dsl_dir,
1548 dd_space_towrite, dd.dd_space_towrite) ||
1549 GETMEMB(dd.dd_phys, struct dsl_dir_phys,
1550 dd_used_bytes, dsp.dd_used_bytes) ||
1551 GETMEMB(dd.dd_phys, struct dsl_dir_phys,
1552 dd_compressed_bytes, dsp.dd_compressed_bytes) ||
1553 GETMEMB(dd.dd_phys, struct dsl_dir_phys,
1554 dd_uncompressed_bytes, dsp.dd_uncompressed_bytes)) {
1555 return (DCMD_ERR);
1556 }
1557
1558 mdb_printf("dd_space_towrite = %llu%s %llu%s %llu%s %llu%s\n",
1559 dd.dd_space_towrite[0] >> shift, suffix,
1560 dd.dd_space_towrite[1] >> shift, suffix,
1561 dd.dd_space_towrite[2] >> shift, suffix,
1562 dd.dd_space_towrite[3] >> shift, suffix);
1563
1564 mdb_printf("dd_phys.dd_used_bytes = %llu%s\n",
1565 dsp.dd_used_bytes >> shift, suffix);
1566 mdb_printf("dd_phys.dd_compressed_bytes = %llu%s\n",
1567 dsp.dd_compressed_bytes >> shift, suffix);
1568 mdb_printf("dd_phys.dd_uncompressed_bytes = %llu%s\n",
1569 dsp.dd_uncompressed_bytes >> shift, suffix);
1570
1571 bzero(&sd, sizeof (sd));
1572 if (mdb_pwalk("metaslab", space_cb, &sd, addr) != 0) {
1573 mdb_warn("can't walk metaslabs");
1574 return (DCMD_ERR);
1575 }
1576
1577 mdb_printf("ms_allocmap = %llu%s %llu%s %llu%s %llu%s\n",
1578 sd.ms_allocmap[0] >> shift, suffix,
1579 sd.ms_allocmap[1] >> shift, suffix,
1580 sd.ms_allocmap[2] >> shift, suffix,
1581 sd.ms_allocmap[3] >> shift, suffix);
1582 mdb_printf("ms_freemap = %llu%s %llu%s %llu%s %llu%s\n",
1583 sd.ms_freemap[0] >> shift, suffix,
1584 sd.ms_freemap[1] >> shift, suffix,
1585 sd.ms_freemap[2] >> shift, suffix,
1586 sd.ms_freemap[3] >> shift, suffix);
1587 mdb_printf("ms_map = %llu%s\n", sd.ms_map >> shift, suffix);
1588 mdb_printf("last synced avail = %llu%s\n", sd.avail >> shift, suffix);
1589 mdb_printf("current syncing avail = %llu%s\n",
1590 sd.nowavail >> shift, suffix);
1591
1592 return (DCMD_OK);
1593 }
1594
1595 /*
1596 * ::spa_verify
1597 *
1598 * Given a spa_t, verify that that the pool is self-consistent.
1599 * Currently, it only checks to make sure that the vdev tree exists.
1600 */
1601 /* ARGSUSED */
1602 static int
1603 spa_verify(uintptr_t addr, uint_t flags, int argc, const mdb_arg_t *argv)
1604 {
1605 spa_t spa;
1606
1607 if (argc != 0 || !(flags & DCMD_ADDRSPEC))
1608 return (DCMD_USAGE);
1609
1610 if (mdb_vread(&spa, sizeof (spa), addr) == -1) {
1611 mdb_warn("failed to read spa_t at %p", addr);
1612 return (DCMD_ERR);
1613 }
1614
1615 if (spa.spa_root_vdev == NULL) {
1616 mdb_printf("no vdev tree present\n");
1617 return (DCMD_OK);
1618 }
1619
1620 return (DCMD_OK);
1621 }
1622
1623 static int
1624 spa_print_aux(spa_aux_vdev_t *sav, uint_t flags, mdb_arg_t *v,
1625 const char *name)
1626 {
1627 uintptr_t *aux;
1628 size_t len;
1629 int ret, i;
1630
1631 /*
1632 * Iterate over aux vdevs and print those out as well. This is a
1633 * little annoying because we don't have a root vdev to pass to ::vdev.
1634 * Instead, we print a single line and then call it for each child
1635 * vdev.
1636 */
1637 if (sav->sav_count != 0) {
1638 v[1].a_type = MDB_TYPE_STRING;
1639 v[1].a_un.a_str = "-d";
1640 v[2].a_type = MDB_TYPE_IMMEDIATE;
1641 v[2].a_un.a_val = 2;
1642
1643 len = sav->sav_count * sizeof (uintptr_t);
1644 aux = mdb_alloc(len, UM_SLEEP);
1645 if (mdb_vread(aux, len,
1646 (uintptr_t)sav->sav_vdevs) == -1) {
1647 mdb_free(aux, len);
1648 mdb_warn("failed to read l2cache vdevs at %p",
1649 sav->sav_vdevs);
1650 return (DCMD_ERR);
1651 }
1652
1653 mdb_printf("%-?s %-9s %-12s %s\n", "-", "-", "-", name);
1654
1655 for (i = 0; i < sav->sav_count; i++) {
1656 ret = mdb_call_dcmd("vdev", aux[i], flags, 3, v);
1657 if (ret != DCMD_OK) {
1658 mdb_free(aux, len);
1659 return (ret);
1660 }
1661 }
1662
1663 mdb_free(aux, len);
1664 }
1665
1666 return (0);
1667 }
1668
1669 /*
1670 * ::spa_vdevs
1671 *
1672 * -e Include error stats
1673 *
1674 * Print out a summarized list of vdevs for the given spa_t.
1675 * This is accomplished by invoking "::vdev -re" on the root vdev, as well as
1676 * iterating over the cache devices.
1677 */
1678 /* ARGSUSED */
1679 static int
1680 spa_vdevs(uintptr_t addr, uint_t flags, int argc, const mdb_arg_t *argv)
1681 {
1682 spa_t spa;
1683 mdb_arg_t v[3];
1684 int errors = FALSE;
1685 int ret;
1686
1687 if (mdb_getopts(argc, argv,
1688 'e', MDB_OPT_SETBITS, TRUE, &errors,
1689 NULL) != argc)
1690 return (DCMD_USAGE);
1691
1692 if (!(flags & DCMD_ADDRSPEC))
1693 return (DCMD_USAGE);
1694
1695 if (mdb_vread(&spa, sizeof (spa), addr) == -1) {
1696 mdb_warn("failed to read spa_t at %p", addr);
1697 return (DCMD_ERR);
1698 }
1699
1700 /*
1701 * Unitialized spa_t structures can have a NULL root vdev.
1702 */
1703 if (spa.spa_root_vdev == NULL) {
1704 mdb_printf("no associated vdevs\n");
1705 return (DCMD_OK);
1706 }
1707
1708 v[0].a_type = MDB_TYPE_STRING;
1709 v[0].a_un.a_str = errors ? "-re" : "-r";
1710
1711 ret = mdb_call_dcmd("vdev", (uintptr_t)spa.spa_root_vdev,
1712 flags, 1, v);
1713 if (ret != DCMD_OK)
1714 return (ret);
1715
1716 if (spa_print_aux(&spa.spa_l2cache, flags, v, "cache") != 0 ||
1717 spa_print_aux(&spa.spa_spares, flags, v, "spares") != 0)
1718 return (DCMD_ERR);
1719
1720 return (DCMD_OK);
1721 }
1722
1723 /*
1724 * ::zio
1725 *
1726 * Print a summary of zio_t and all its children. This is intended to display a
1727 * zio tree, and hence we only pick the most important pieces of information for
1728 * the main summary. More detailed information can always be found by doing a
1729 * '::print zio' on the underlying zio_t. The columns we display are:
1730 *
1731 * ADDRESS TYPE STAGE WAITER
1732 *
1733 * The 'address' column is indented by one space for each depth level as we
1734 * descend down the tree.
1735 */
1736
1737 #define ZIO_MAXINDENT 24
1738 #define ZIO_MAXWIDTH (sizeof (uintptr_t) * 2 + ZIO_MAXINDENT)
1739 #define ZIO_WALK_SELF 0
1740 #define ZIO_WALK_CHILD 1
1741 #define ZIO_WALK_PARENT 2
1742
1743 typedef struct zio_print_args {
1744 int zpa_current_depth;
1745 int zpa_min_depth;
1746 int zpa_max_depth;
1747 int zpa_type;
1748 uint_t zpa_flags;
1749 } zio_print_args_t;
1750
1751 static int zio_child_cb(uintptr_t addr, const void *unknown, void *arg);
1752
1753 static int
1754 zio_print_cb(uintptr_t addr, const void *data, void *priv)
1755 {
1756 const zio_t *zio = data;
1757 zio_print_args_t *zpa = priv;
1758 mdb_ctf_id_t type_enum, stage_enum;
1759 int indent = zpa->zpa_current_depth;
1760 const char *type, *stage;
1761 uintptr_t laddr;
1762
1763 if (indent > ZIO_MAXINDENT)
1764 indent = ZIO_MAXINDENT;
1765
1766 if (mdb_ctf_lookup_by_name("enum zio_type", &type_enum) == -1 ||
1767 mdb_ctf_lookup_by_name("enum zio_stage", &stage_enum) == -1) {
1768 mdb_warn("failed to lookup zio enums");
1769 return (WALK_ERR);
1770 }
1771
1772 if ((type = mdb_ctf_enum_name(type_enum, zio->io_type)) != NULL)
1773 type += sizeof ("ZIO_TYPE_") - 1;
1774 else
1775 type = "?";
1776
1777 if ((stage = mdb_ctf_enum_name(stage_enum, zio->io_stage)) != NULL)
1778 stage += sizeof ("ZIO_STAGE_") - 1;
1779 else
1780 stage = "?";
1781
1782 if (zpa->zpa_current_depth >= zpa->zpa_min_depth) {
1783 if (zpa->zpa_flags & DCMD_PIPE_OUT) {
1784 mdb_printf("%?p\n", addr);
1785 } else {
1786 mdb_printf("%*s%-*p %-5s %-16s ", indent, "",
1787 ZIO_MAXWIDTH - indent, addr, type, stage);
1788 if (zio->io_waiter)
1789 mdb_printf("%?p\n", zio->io_waiter);
1790 else
1791 mdb_printf("-\n");
1792 }
1793 }
1794
1795 if (zpa->zpa_current_depth >= zpa->zpa_max_depth)
1796 return (WALK_NEXT);
1797
1798 if (zpa->zpa_type == ZIO_WALK_PARENT)
1799 laddr = addr + OFFSETOF(zio_t, io_parent_list);
1800 else
1801 laddr = addr + OFFSETOF(zio_t, io_child_list);
1802
1803 zpa->zpa_current_depth++;
1804 if (mdb_pwalk("list", zio_child_cb, zpa, laddr) != 0) {
1805 mdb_warn("failed to walk zio_t children at %p\n", laddr);
1806 return (WALK_ERR);
1807 }
1808 zpa->zpa_current_depth--;
1809
1810 return (WALK_NEXT);
1811 }
1812
1813 /* ARGSUSED */
1814 static int
1815 zio_child_cb(uintptr_t addr, const void *unknown, void *arg)
1816 {
1817 zio_link_t zl;
1818 zio_t zio;
1819 uintptr_t ziop;
1820 zio_print_args_t *zpa = arg;
1821
1822 if (mdb_vread(&zl, sizeof (zl), addr) == -1) {
1823 mdb_warn("failed to read zio_link_t at %p", addr);
1824 return (WALK_ERR);
1825 }
1826
1827 if (zpa->zpa_type == ZIO_WALK_PARENT)
1828 ziop = (uintptr_t)zl.zl_parent;
1829 else
1830 ziop = (uintptr_t)zl.zl_child;
1831
1832 if (mdb_vread(&zio, sizeof (zio_t), ziop) == -1) {
1833 mdb_warn("failed to read zio_t at %p", ziop);
1834 return (WALK_ERR);
1835 }
1836
1837 return (zio_print_cb(ziop, &zio, arg));
1838 }
1839
1840 /* ARGSUSED */
1841 static int
1842 zio_print(uintptr_t addr, uint_t flags, int argc, const mdb_arg_t *argv)
1843 {
1844 zio_t zio;
1845 zio_print_args_t zpa = { 0 };
1846
1847 if (!(flags & DCMD_ADDRSPEC))
1848 return (DCMD_USAGE);
1849
1850 if (mdb_getopts(argc, argv,
1851 'r', MDB_OPT_SETBITS, INT_MAX, &zpa.zpa_max_depth,
1852 'c', MDB_OPT_SETBITS, ZIO_WALK_CHILD, &zpa.zpa_type,
1853 'p', MDB_OPT_SETBITS, ZIO_WALK_PARENT, &zpa.zpa_type,
1854 NULL) != argc)
1855 return (DCMD_USAGE);
1856
1857 zpa.zpa_flags = flags;
1858 if (zpa.zpa_max_depth != 0) {
1859 if (zpa.zpa_type == ZIO_WALK_SELF)
1860 zpa.zpa_type = ZIO_WALK_CHILD;
1861 } else if (zpa.zpa_type != ZIO_WALK_SELF) {
1862 zpa.zpa_min_depth = 1;
1863 zpa.zpa_max_depth = 1;
1864 }
1865
1866 if (mdb_vread(&zio, sizeof (zio_t), addr) == -1) {
1867 mdb_warn("failed to read zio_t at %p", addr);
1868 return (DCMD_ERR);
1869 }
1870
1871 if (!(flags & DCMD_PIPE_OUT) && DCMD_HDRSPEC(flags))
1872 mdb_printf("%<u>%-*s %-5s %-16s %-?s%</u>\n", ZIO_MAXWIDTH,
1873 "ADDRESS", "TYPE", "STAGE", "WAITER");
1874
1875 if (zio_print_cb(addr, &zio, &zpa) != WALK_NEXT)
1876 return (DCMD_ERR);
1877
1878 return (DCMD_OK);
1879 }
1880
1881 /*
1882 * [addr]::zio_state
1883 *
1884 * Print a summary of all zio_t structures on the system, or for a particular
1885 * pool. This is equivalent to '::walk zio_root | ::zio'.
1886 */
1887 /*ARGSUSED*/
1888 static int
1889 zio_state(uintptr_t addr, uint_t flags, int argc, const mdb_arg_t *argv)
1890 {
1891 /*
1892 * MDB will remember the last address of the pipeline, so if we don't
1893 * zero this we'll end up trying to walk zio structures for a
1894 * non-existent spa_t.
1895 */
1896 if (!(flags & DCMD_ADDRSPEC))
1897 addr = 0;
1898
1899 return (mdb_pwalk_dcmd("zio_root", "zio", argc, argv, addr));
1900 }
1901
1902 typedef struct txg_list_walk_data {
1903 uintptr_t lw_head[TXG_SIZE];
1904 int lw_txgoff;
1905 int lw_maxoff;
1906 size_t lw_offset;
1907 void *lw_obj;
1908 } txg_list_walk_data_t;
1909
1910 static int
1911 txg_list_walk_init_common(mdb_walk_state_t *wsp, int txg, int maxoff)
1912 {
1913 txg_list_walk_data_t *lwd;
1914 txg_list_t list;
1915 int i;
1916
1917 lwd = mdb_alloc(sizeof (txg_list_walk_data_t), UM_SLEEP | UM_GC);
1918 if (mdb_vread(&list, sizeof (txg_list_t), wsp->walk_addr) == -1) {
1919 mdb_warn("failed to read txg_list_t at %#lx", wsp->walk_addr);
1920 return (WALK_ERR);
1921 }
1922
1923 for (i = 0; i < TXG_SIZE; i++)
1924 lwd->lw_head[i] = (uintptr_t)list.tl_head[i];
1925 lwd->lw_offset = list.tl_offset;
1926 lwd->lw_obj = mdb_alloc(lwd->lw_offset + sizeof (txg_node_t),
1927 UM_SLEEP | UM_GC);
1928 lwd->lw_txgoff = txg;
1929 lwd->lw_maxoff = maxoff;
1930
1931 wsp->walk_addr = lwd->lw_head[lwd->lw_txgoff];
1932 wsp->walk_data = lwd;
1933
1934 return (WALK_NEXT);
1935 }
1936
1937 static int
1938 txg_list_walk_init(mdb_walk_state_t *wsp)
1939 {
1940 return (txg_list_walk_init_common(wsp, 0, TXG_SIZE-1));
1941 }
1942
1943 static int
1944 txg_list0_walk_init(mdb_walk_state_t *wsp)
1945 {
1946 return (txg_list_walk_init_common(wsp, 0, 0));
1947 }
1948
1949 static int
1950 txg_list1_walk_init(mdb_walk_state_t *wsp)
1951 {
1952 return (txg_list_walk_init_common(wsp, 1, 1));
1953 }
1954
1955 static int
1956 txg_list2_walk_init(mdb_walk_state_t *wsp)
1957 {
1958 return (txg_list_walk_init_common(wsp, 2, 2));
1959 }
1960
1961 static int
1962 txg_list3_walk_init(mdb_walk_state_t *wsp)
1963 {
1964 return (txg_list_walk_init_common(wsp, 3, 3));
1965 }
1966
1967 static int
1968 txg_list_walk_step(mdb_walk_state_t *wsp)
1969 {
1970 txg_list_walk_data_t *lwd = wsp->walk_data;
1971 uintptr_t addr;
1972 txg_node_t *node;
1973 int status;
1974
1975 while (wsp->walk_addr == NULL && lwd->lw_txgoff < lwd->lw_maxoff) {
1976 lwd->lw_txgoff++;
1977 wsp->walk_addr = lwd->lw_head[lwd->lw_txgoff];
1978 }
1979
1980 if (wsp->walk_addr == NULL)
1981 return (WALK_DONE);
1982
1983 addr = wsp->walk_addr - lwd->lw_offset;
1984
1985 if (mdb_vread(lwd->lw_obj,
1986 lwd->lw_offset + sizeof (txg_node_t), addr) == -1) {
1987 mdb_warn("failed to read list element at %#lx", addr);
1988 return (WALK_ERR);
1989 }
1990
1991 status = wsp->walk_callback(addr, lwd->lw_obj, wsp->walk_cbdata);
1992 node = (txg_node_t *)((uintptr_t)lwd->lw_obj + lwd->lw_offset);
1993 wsp->walk_addr = (uintptr_t)node->tn_next[lwd->lw_txgoff];
1994
1995 return (status);
1996 }
1997
1998 /*
1999 * ::walk spa
2000 *
2001 * Walk all named spa_t structures in the namespace. This is nothing more than
2002 * a layered avl walk.
2003 */
2004 static int
2005 spa_walk_init(mdb_walk_state_t *wsp)
2006 {
2007 GElf_Sym sym;
2008
2009 if (wsp->walk_addr != NULL) {
2010 mdb_warn("spa walk only supports global walks\n");
2011 return (WALK_ERR);
2012 }
2013
2014 if (mdb_lookup_by_obj(ZFS_OBJ_NAME, "spa_namespace_avl", &sym) == -1) {
2015 mdb_warn("failed to find symbol 'spa_namespace_avl'");
2016 return (WALK_ERR);
2017 }
2018
2019 wsp->walk_addr = (uintptr_t)sym.st_value;
2020
2021 if (mdb_layered_walk("avl", wsp) == -1) {
2022 mdb_warn("failed to walk 'avl'\n");
2023 return (WALK_ERR);
2024 }
2025
2026 return (WALK_NEXT);
2027 }
2028
2029 static int
2030 spa_walk_step(mdb_walk_state_t *wsp)
2031 {
2032 spa_t spa;
2033
2034 if (mdb_vread(&spa, sizeof (spa), wsp->walk_addr) == -1) {
2035 mdb_warn("failed to read spa_t at %p", wsp->walk_addr);
2036 return (WALK_ERR);
2037 }
2038
2039 return (wsp->walk_callback(wsp->walk_addr, &spa, wsp->walk_cbdata));
2040 }
2041
2042 /*
2043 * [addr]::walk zio
2044 *
2045 * Walk all active zio_t structures on the system. This is simply a layered
2046 * walk on top of ::walk zio_cache, with the optional ability to limit the
2047 * structures to a particular pool.
2048 */
2049 static int
2050 zio_walk_init(mdb_walk_state_t *wsp)
2051 {
2052 wsp->walk_data = (void *)wsp->walk_addr;
2053
2054 if (mdb_layered_walk("zio_cache", wsp) == -1) {
2055 mdb_warn("failed to walk 'zio_cache'\n");
2056 return (WALK_ERR);
2057 }
2058
2059 return (WALK_NEXT);
2060 }
2061
2062 static int
2063 zio_walk_step(mdb_walk_state_t *wsp)
2064 {
2065 zio_t zio;
2066
2067 if (mdb_vread(&zio, sizeof (zio), wsp->walk_addr) == -1) {
2068 mdb_warn("failed to read zio_t at %p", wsp->walk_addr);
2069 return (WALK_ERR);
2070 }
2071
2072 if (wsp->walk_data != NULL && wsp->walk_data != zio.io_spa)
2073 return (WALK_NEXT);
2074
2075 return (wsp->walk_callback(wsp->walk_addr, &zio, wsp->walk_cbdata));
2076 }
2077
2078 /*
2079 * [addr]::walk zio_root
2080 *
2081 * Walk only root zio_t structures, optionally for a particular spa_t.
2082 */
2083 static int
2084 zio_walk_root_step(mdb_walk_state_t *wsp)
2085 {
2086 zio_t zio;
2087
2088 if (mdb_vread(&zio, sizeof (zio), wsp->walk_addr) == -1) {
2089 mdb_warn("failed to read zio_t at %p", wsp->walk_addr);
2090 return (WALK_ERR);
2091 }
2092
2093 if (wsp->walk_data != NULL && wsp->walk_data != zio.io_spa)
2094 return (WALK_NEXT);
2095
2096 /* If the parent list is not empty, ignore */
2097 if (zio.io_parent_list.list_head.list_next !=
2098 &((zio_t *)wsp->walk_addr)->io_parent_list.list_head)
2099 return (WALK_NEXT);
2100
2101 return (wsp->walk_callback(wsp->walk_addr, &zio, wsp->walk_cbdata));
2102 }
2103
2104 #define NICENUM_BUFLEN 6
2105
2106 static int
2107 snprintfrac(char *buf, int len,
2108 uint64_t numerator, uint64_t denom, int frac_digits)
2109 {
2110 int mul = 1;
2111 int whole, frac, i;
2112
2113 for (i = frac_digits; i; i--)
2114 mul *= 10;
2115 whole = numerator / denom;
2116 frac = mul * numerator / denom - mul * whole;
2117 return (mdb_snprintf(buf, len, "%u.%0*u", whole, frac_digits, frac));
2118 }
2119
2120 static void
2121 mdb_nicenum(uint64_t num, char *buf)
2122 {
2123 uint64_t n = num;
2124 int index = 0;
2125 char *u;
2126
2127 while (n >= 1024) {
2128 n = (n + (1024 / 2)) / 1024; /* Round up or down */
2129 index++;
2130 }
2131
2132 u = &" \0K\0M\0G\0T\0P\0E\0"[index*2];
2133
2134 if (index == 0) {
2135 (void) mdb_snprintf(buf, NICENUM_BUFLEN, "%llu",
2136 (u_longlong_t)n);
2137 } else if (n < 10 && (num & (num - 1)) != 0) {
2138 (void) snprintfrac(buf, NICENUM_BUFLEN,
2139 num, 1ULL << 10 * index, 2);
2140 strcat(buf, u);
2141 } else if (n < 100 && (num & (num - 1)) != 0) {
2142 (void) snprintfrac(buf, NICENUM_BUFLEN,
2143 num, 1ULL << 10 * index, 1);
2144 strcat(buf, u);
2145 } else {
2146 (void) mdb_snprintf(buf, NICENUM_BUFLEN, "%llu%s",
2147 (u_longlong_t)n, u);
2148 }
2149 }
2150
2151 /*
2152 * ::zfs_blkstats
2153 *
2154 * -v print verbose per-level information
2155 *
2156 */
2157 static int
2158 zfs_blkstats(uintptr_t addr, uint_t flags, int argc, const mdb_arg_t *argv)
2159 {
2160 boolean_t verbose = B_FALSE;
2161 zfs_all_blkstats_t stats;
2162 dmu_object_type_t t;
2163 zfs_blkstat_t *tzb;
2164 uint64_t ditto;
2165 dmu_object_type_info_t dmu_ot[DMU_OT_NUMTYPES + 10];
2166 /* +10 in case it grew */
2167
2168 if (mdb_readvar(&dmu_ot, "dmu_ot") == -1) {
2169 mdb_warn("failed to read 'dmu_ot'");
2170 return (DCMD_ERR);
2171 }
2172
2173 if (mdb_getopts(argc, argv,
2174 'v', MDB_OPT_SETBITS, TRUE, &verbose,
2175 NULL) != argc)
2176 return (DCMD_USAGE);
2177
2178 if (!(flags & DCMD_ADDRSPEC))
2179 return (DCMD_USAGE);
2180
2181 if (GETMEMB(addr, struct spa, spa_dsl_pool, addr) ||
2182 GETMEMB(addr, struct dsl_pool, dp_blkstats, addr) ||
2183 mdb_vread(&stats, sizeof (zfs_all_blkstats_t), addr) == -1) {
2184 mdb_warn("failed to read data at %p;", addr);
2185 mdb_printf("maybe no stats? run \"zpool scrub\" first.");
2186 return (DCMD_ERR);
2187 }
2188
2189 tzb = &stats.zab_type[DN_MAX_LEVELS][DMU_OT_TOTAL];
2190 if (tzb->zb_gangs != 0) {
2191 mdb_printf("Ganged blocks: %llu\n",
2192 (longlong_t)tzb->zb_gangs);
2193 }
2194
2195 ditto = tzb->zb_ditto_2_of_2_samevdev + tzb->zb_ditto_2_of_3_samevdev +
2196 tzb->zb_ditto_3_of_3_samevdev;
2197 if (ditto != 0) {
2198 mdb_printf("Dittoed blocks on same vdev: %llu\n",
2199 (longlong_t)ditto);
2200 }
2201
2202 mdb_printf("\nBlocks\tLSIZE\tPSIZE\tASIZE"
2203 "\t avg\t comp\t%%Total\tType\n");
2204
2205 for (t = 0; t <= DMU_OT_TOTAL; t++) {
2206 char csize[NICENUM_BUFLEN], lsize[NICENUM_BUFLEN];
2207 char psize[NICENUM_BUFLEN], asize[NICENUM_BUFLEN];
2208 char avg[NICENUM_BUFLEN];
2209 char comp[NICENUM_BUFLEN], pct[NICENUM_BUFLEN];
2210 char typename[64];
2211 int l;
2212
2213
2214 if (t == DMU_OT_DEFERRED)
2215 strcpy(typename, "deferred free");
2216 else if (t == DMU_OT_OTHER)
2217 strcpy(typename, "other");
2218 else if (t == DMU_OT_TOTAL)
2219 strcpy(typename, "Total");
2220 else if (mdb_readstr(typename, sizeof (typename),
2221 (uintptr_t)dmu_ot[t].ot_name) == -1) {
2222 mdb_warn("failed to read type name");
2223 return (DCMD_ERR);
2224 }
2225
2226 if (stats.zab_type[DN_MAX_LEVELS][t].zb_asize == 0)
2227 continue;
2228
2229 for (l = -1; l < DN_MAX_LEVELS; l++) {
2230 int level = (l == -1 ? DN_MAX_LEVELS : l);
2231 zfs_blkstat_t *zb = &stats.zab_type[level][t];
2232
2233 if (zb->zb_asize == 0)
2234 continue;
2235
2236 /*
2237 * Don't print each level unless requested.
2238 */
2239 if (!verbose && level != DN_MAX_LEVELS)
2240 continue;
2241
2242 /*
2243 * If all the space is level 0, don't print the
2244 * level 0 separately.
2245 */
2246 if (level == 0 && zb->zb_asize ==
2247 stats.zab_type[DN_MAX_LEVELS][t].zb_asize)
2248 continue;
2249
2250 mdb_nicenum(zb->zb_count, csize);
2251 mdb_nicenum(zb->zb_lsize, lsize);
2252 mdb_nicenum(zb->zb_psize, psize);
2253 mdb_nicenum(zb->zb_asize, asize);
2254 mdb_nicenum(zb->zb_asize / zb->zb_count, avg);
2255 (void) snprintfrac(comp, NICENUM_BUFLEN,
2256 zb->zb_lsize, zb->zb_psize, 2);
2257 (void) snprintfrac(pct, NICENUM_BUFLEN,
2258 100 * zb->zb_asize, tzb->zb_asize, 2);
2259
2260 mdb_printf("%6s\t%5s\t%5s\t%5s\t%5s"
2261 "\t%5s\t%6s\t",
2262 csize, lsize, psize, asize, avg, comp, pct);
2263
2264 if (level == DN_MAX_LEVELS)
2265 mdb_printf("%s\n", typename);
2266 else
2267 mdb_printf(" L%d %s\n",
2268 level, typename);
2269 }
2270 }
2271
2272 return (DCMD_OK);
2273 }
2274
2275 /* ARGSUSED */
2276 static int
2277 reference_cb(uintptr_t addr, const void *ignored, void *arg)
2278 {
2279 static int gotid;
2280 static mdb_ctf_id_t ref_id;
2281 uintptr_t ref_holder;
2282 uintptr_t ref_removed;
2283 uint64_t ref_number;
2284 boolean_t holder_is_str = B_FALSE;
2285 char holder_str[128];
2286 boolean_t removed = (boolean_t)arg;
2287
2288 if (!gotid) {
2289 if (mdb_ctf_lookup_by_name("struct reference", &ref_id) == -1) {
2290 mdb_warn("couldn't find struct reference");
2291 return (WALK_ERR);
2292 }
2293 gotid = TRUE;
2294 }
2295
2296 if (GETMEMBID(addr, &ref_id, ref_holder, ref_holder) ||
2297 GETMEMBID(addr, &ref_id, ref_removed, ref_removed) ||
2298 GETMEMBID(addr, &ref_id, ref_number, ref_number))
2299 return (WALK_ERR);
2300
2301 if (mdb_readstr(holder_str, sizeof (holder_str), ref_holder) != -1)
2302 holder_is_str = strisprint(holder_str);
2303
2304 if (removed)
2305 mdb_printf("removed ");
2306 mdb_printf("reference ");
2307 if (ref_number != 1)
2308 mdb_printf("with count=%llu ", ref_number);
2309 mdb_printf("with tag %p", (void*)ref_holder);
2310 if (holder_is_str)
2311 mdb_printf(" \"%s\"", holder_str);
2312 mdb_printf(", held at:\n");
2313
2314 (void) mdb_call_dcmd("whatis", addr, DCMD_ADDRSPEC, 0, NULL);
2315
2316 if (removed) {
2317 mdb_printf("removed at:\n");
2318 (void) mdb_call_dcmd("whatis", ref_removed,
2319 DCMD_ADDRSPEC, 0, NULL);
2320 }
2321
2322 mdb_printf("\n");
2323
2324 return (WALK_NEXT);
2325 }
2326
2327 /* ARGSUSED */
2328 static int
2329 refcount(uintptr_t addr, uint_t flags, int argc, const mdb_arg_t *argv)
2330 {
2331 uint64_t rc_count, rc_removed_count;
2332 uintptr_t rc_list, rc_removed;
2333 static int gotid;
2334 static mdb_ctf_id_t rc_id;
2335 ulong_t off;
2336
2337 if (!(flags & DCMD_ADDRSPEC))
2338 return (DCMD_USAGE);
2339
2340 if (!gotid) {
2341 if (mdb_ctf_lookup_by_name("struct refcount", &rc_id) == -1) {
2342 mdb_warn("couldn't find struct refcount");
2343 return (DCMD_ERR);
2344 }
2345 gotid = TRUE;
2346 }
2347
2348 if (GETMEMBID(addr, &rc_id, rc_count, rc_count) ||
2349 GETMEMBID(addr, &rc_id, rc_removed_count, rc_removed_count))
2350 return (DCMD_ERR);
2351
2352 mdb_printf("refcount_t at %p has %llu current holds, "
2353 "%llu recently released holds\n",
2354 addr, (longlong_t)rc_count, (longlong_t)rc_removed_count);
2355
2356 if (rc_count > 0)
2357 mdb_printf("current holds:\n");
2358 if (mdb_ctf_offsetof(rc_id, "rc_list", &off) == -1)
2359 return (DCMD_ERR);
2360 rc_list = addr + off/NBBY;
2361 mdb_pwalk("list", reference_cb, (void*)B_FALSE, rc_list);
2362
2363 if (rc_removed_count > 0)
2364 mdb_printf("released holds:\n");
2365 if (mdb_ctf_offsetof(rc_id, "rc_removed", &off) == -1)
2366 return (DCMD_ERR);
2367 rc_removed = addr + off/NBBY;
2368 mdb_pwalk("list", reference_cb, (void*)B_TRUE, rc_removed);
2369
2370 return (DCMD_OK);
2371 }
2372
2373 /* ARGSUSED */
2374 static int
2375 sa_attr_table(uintptr_t addr, uint_t flags, int argc, const mdb_arg_t *argv)
2376 {
2377 sa_attr_table_t *table;
2378 sa_os_t sa_os;
2379 char *name;
2380 int i;
2381
2382 if (mdb_vread(&sa_os, sizeof (sa_os_t), addr) == -1) {
2383 mdb_warn("failed to read sa_os at %p", addr);
2384 return (DCMD_ERR);
2385 }
2386
2387 table = mdb_alloc(sizeof (sa_attr_table_t) * sa_os.sa_num_attrs,
2388 UM_SLEEP | UM_GC);
2389 name = mdb_alloc(MAXPATHLEN, UM_SLEEP | UM_GC);
2390
2391 if (mdb_vread(table, sizeof (sa_attr_table_t) * sa_os.sa_num_attrs,
2392 (uintptr_t)sa_os.sa_attr_table) == -1) {
2393 mdb_warn("failed to read sa_os at %p", addr);
2394 return (DCMD_ERR);
2395 }
2396
2397 mdb_printf("%<u>%-10s %-10s %-10s %-10s %s%</u>\n",
2398 "ATTR ID", "REGISTERED", "LENGTH", "BSWAP", "NAME");
2399 for (i = 0; i != sa_os.sa_num_attrs; i++) {
2400 mdb_readstr(name, MAXPATHLEN, (uintptr_t)table[i].sa_name);
2401 mdb_printf("%5x %8x %8x %8x %-s\n",
2402 (int)table[i].sa_attr, (int)table[i].sa_registered,
2403 (int)table[i].sa_length, table[i].sa_byteswap, name);
2404 }
2405
2406 return (DCMD_OK);
2407 }
2408
2409 static int
2410 sa_get_off_table(uintptr_t addr, uint32_t **off_tab, int attr_count)
2411 {
2412 uintptr_t idx_table;
2413
2414 if (GETMEMB(addr, struct sa_idx_tab, sa_idx_tab, idx_table)) {
2415 mdb_printf("can't find offset table in sa_idx_tab\n");
2416 return (-1);
2417 }
2418
2419 *off_tab = mdb_alloc(attr_count * sizeof (uint32_t),
2420 UM_SLEEP | UM_GC);
2421
2422 if (mdb_vread(*off_tab,
2423 attr_count * sizeof (uint32_t), idx_table) == -1) {
2424 mdb_warn("failed to attribute offset table %p", idx_table);
2425 return (-1);
2426 }
2427
2428 return (DCMD_OK);
2429 }
2430
2431 /*ARGSUSED*/
2432 static int
2433 sa_attr_print(uintptr_t addr, uint_t flags, int argc, const mdb_arg_t *argv)
2434 {
2435 uint32_t *offset_tab;
2436 int attr_count;
2437 uint64_t attr_id;
2438 uintptr_t attr_addr;
2439 uintptr_t bonus_tab, spill_tab;
2440 uintptr_t db_bonus, db_spill;
2441 uintptr_t os, os_sa;
2442 uintptr_t db_data;
2443
2444 if (argc != 1)
2445 return (DCMD_USAGE);
2446
2447 if (argv[0].a_type == MDB_TYPE_STRING)
2448 attr_id = mdb_strtoull(argv[0].a_un.a_str);
2449 else
2450 return (DCMD_USAGE);
2451
2452 if (GETMEMB(addr, struct sa_handle, sa_bonus_tab, bonus_tab) ||
2453 GETMEMB(addr, struct sa_handle, sa_spill_tab, spill_tab) ||
2454 GETMEMB(addr, struct sa_handle, sa_os, os) ||
2455 GETMEMB(addr, struct sa_handle, sa_bonus, db_bonus) ||
2456 GETMEMB(addr, struct sa_handle, sa_spill, db_spill)) {
2457 mdb_printf("Can't find necessary information in sa_handle "
2458 "in sa_handle\n");
2459 return (DCMD_ERR);
2460 }
2461
2462 if (GETMEMB(os, struct objset, os_sa, os_sa)) {
2463 mdb_printf("Can't find os_sa in objset\n");
2464 return (DCMD_ERR);
2465 }
2466
2467 if (GETMEMB(os_sa, struct sa_os, sa_num_attrs, attr_count)) {
2468 mdb_printf("Can't find sa_num_attrs\n");
2469 return (DCMD_ERR);
2470 }
2471
2472 if (attr_id > attr_count) {
2473 mdb_printf("attribute id number is out of range\n");
2474 return (DCMD_ERR);
2475 }
2476
2477 if (bonus_tab) {
2478 if (sa_get_off_table(bonus_tab, &offset_tab,
2479 attr_count) == -1) {
2480 return (DCMD_ERR);
2481 }
2482
2483 if (GETMEMB(db_bonus, struct dmu_buf, db_data, db_data)) {
2484 mdb_printf("can't find db_data in bonus dbuf\n");
2485 return (DCMD_ERR);
2486 }
2487 }
2488
2489 if (bonus_tab && !TOC_ATTR_PRESENT(offset_tab[attr_id]) &&
2490 spill_tab == NULL) {
2491 mdb_printf("Attribute does not exist\n");
2492 return (DCMD_ERR);
2493 } else if (!TOC_ATTR_PRESENT(offset_tab[attr_id]) && spill_tab) {
2494 if (sa_get_off_table(spill_tab, &offset_tab,
2495 attr_count) == -1) {
2496 return (DCMD_ERR);
2497 }
2498 if (GETMEMB(db_spill, struct dmu_buf, db_data, db_data)) {
2499 mdb_printf("can't find db_data in spill dbuf\n");
2500 return (DCMD_ERR);
2501 }
2502 if (!TOC_ATTR_PRESENT(offset_tab[attr_id])) {
2503 mdb_printf("Attribute does not exist\n");
2504 return (DCMD_ERR);
2505 }
2506 }
2507 attr_addr = db_data + TOC_OFF(offset_tab[attr_id]);
2508 mdb_printf("%p\n", attr_addr);
2509 return (DCMD_OK);
2510 }
2511
2512 /* ARGSUSED */
2513 static int
2514 zfs_ace_print_common(uintptr_t addr, uint_t flags,
2515 uint64_t id, uint32_t access_mask, uint16_t ace_flags,
2516 uint16_t ace_type, int verbose)
2517 {
2518 if (DCMD_HDRSPEC(flags) && !verbose)
2519 mdb_printf("%<u>%-?s %-8s %-8s %-8s %s%</u>\n",
2520 "ADDR", "FLAGS", "MASK", "TYPE", "ID");
2521
2522 if (!verbose) {
2523 mdb_printf("%0?p %-8x %-8x %-8x %-llx\n", addr,
2524 ace_flags, access_mask, ace_type, id);
2525 return (DCMD_OK);
2526 }
2527
2528 switch (ace_flags & ACE_TYPE_FLAGS) {
2529 case ACE_OWNER:
2530 mdb_printf("owner@:");
2531 break;
2532 case (ACE_IDENTIFIER_GROUP | ACE_GROUP):
2533 mdb_printf("group@:");
2534 break;
2535 case ACE_EVERYONE:
2536 mdb_printf("everyone@:");
2537 break;
2538 case ACE_IDENTIFIER_GROUP:
2539 mdb_printf("group:%llx:", (u_longlong_t)id);
2540 break;
2541 case 0: /* User entry */
2542 mdb_printf("user:%llx:", (u_longlong_t)id);
2543 break;
2544 }
2545
2546 /* print out permission mask */
2547 if (access_mask & ACE_READ_DATA)
2548 mdb_printf("r");
2549 else
2550 mdb_printf("-");
2551 if (access_mask & ACE_WRITE_DATA)
2552 mdb_printf("w");
2553 else
2554 mdb_printf("-");
2555 if (access_mask & ACE_EXECUTE)
2556 mdb_printf("x");
2557 else
2558 mdb_printf("-");
2559 if (access_mask & ACE_APPEND_DATA)
2560 mdb_printf("p");
2561 else
2562 mdb_printf("-");
2563 if (access_mask & ACE_DELETE)
2564 mdb_printf("d");
2565 else
2566 mdb_printf("-");
2567 if (access_mask & ACE_DELETE_CHILD)
2568 mdb_printf("D");
2569 else
2570 mdb_printf("-");
2571 if (access_mask & ACE_READ_ATTRIBUTES)
2572 mdb_printf("a");
2573 else
2574 mdb_printf("-");
2575 if (access_mask & ACE_WRITE_ATTRIBUTES)
2576 mdb_printf("A");
2577 else
2578 mdb_printf("-");
2579 if (access_mask & ACE_READ_NAMED_ATTRS)
2580 mdb_printf("R");
2581 else
2582 mdb_printf("-");
2583 if (access_mask & ACE_WRITE_NAMED_ATTRS)
2584 mdb_printf("W");
2585 else
2586 mdb_printf("-");
2587 if (access_mask & ACE_READ_ACL)
2588 mdb_printf("c");
2589 else
2590 mdb_printf("-");
2591 if (access_mask & ACE_WRITE_ACL)
2592 mdb_printf("C");
2593 else
2594 mdb_printf("-");
2595 if (access_mask & ACE_WRITE_OWNER)
2596 mdb_printf("o");
2597 else
2598 mdb_printf("-");
2599 if (access_mask & ACE_SYNCHRONIZE)
2600 mdb_printf("s");
2601 else
2602 mdb_printf("-");
2603
2604 mdb_printf(":");
2605
2606 /* Print out inheritance flags */
2607 if (ace_flags & ACE_FILE_INHERIT_ACE)
2608 mdb_printf("f");
2609 else
2610 mdb_printf("-");
2611 if (ace_flags & ACE_DIRECTORY_INHERIT_ACE)
2612 mdb_printf("d");
2613 else
2614 mdb_printf("-");
2615 if (ace_flags & ACE_INHERIT_ONLY_ACE)
2616 mdb_printf("i");
2617 else
2618 mdb_printf("-");
2619 if (ace_flags & ACE_NO_PROPAGATE_INHERIT_ACE)
2620 mdb_printf("n");
2621 else
2622 mdb_printf("-");
2623 if (ace_flags & ACE_SUCCESSFUL_ACCESS_ACE_FLAG)
2624 mdb_printf("S");
2625 else
2626 mdb_printf("-");
2627 if (ace_flags & ACE_FAILED_ACCESS_ACE_FLAG)
2628 mdb_printf("F");
2629 else
2630 mdb_printf("-");
2631 if (ace_flags & ACE_INHERITED_ACE)
2632 mdb_printf("I");
2633 else
2634 mdb_printf("-");
2635
2636 switch (ace_type) {
2637 case ACE_ACCESS_ALLOWED_ACE_TYPE:
2638 mdb_printf(":allow\n");
2639 break;
2640 case ACE_ACCESS_DENIED_ACE_TYPE:
2641 mdb_printf(":deny\n");
2642 break;
2643 case ACE_SYSTEM_AUDIT_ACE_TYPE:
2644 mdb_printf(":audit\n");
2645 break;
2646 case ACE_SYSTEM_ALARM_ACE_TYPE:
2647 mdb_printf(":alarm\n");
2648 break;
2649 default:
2650 mdb_printf(":?\n");
2651 }
2652 return (DCMD_OK);
2653 }
2654
2655 /* ARGSUSED */
2656 static int
2657 zfs_ace_print(uintptr_t addr, uint_t flags, int argc, const mdb_arg_t *argv)
2658 {
2659 zfs_ace_t zace;
2660 int verbose = FALSE;
2661 uint64_t id;
2662
2663 if (!(flags & DCMD_ADDRSPEC))
2664 return (DCMD_USAGE);
2665
2666 if (mdb_getopts(argc, argv,
2667 'v', MDB_OPT_SETBITS, TRUE, &verbose, TRUE, NULL) != argc)
2668 return (DCMD_USAGE);
2669
2670 if (mdb_vread(&zace, sizeof (zfs_ace_t), addr) == -1) {
2671 mdb_warn("failed to read zfs_ace_t");
2672 return (DCMD_ERR);
2673 }
2674
2675 if ((zace.z_hdr.z_flags & ACE_TYPE_FLAGS) == 0 ||
2676 (zace.z_hdr.z_flags & ACE_TYPE_FLAGS) == ACE_IDENTIFIER_GROUP)
2677 id = zace.z_fuid;
2678 else
2679 id = -1;
2680
2681 return (zfs_ace_print_common(addr, flags, id, zace.z_hdr.z_access_mask,
2682 zace.z_hdr.z_flags, zace.z_hdr.z_type, verbose));
2683 }
2684
2685 /* ARGSUSED */
2686 static int
2687 zfs_ace0_print(uintptr_t addr, uint_t flags, int argc, const mdb_arg_t *argv)
2688 {
2689 ace_t ace;
2690 uint64_t id;
2691 int verbose = FALSE;
2692
2693 if (!(flags & DCMD_ADDRSPEC))
2694 return (DCMD_USAGE);
2695
2696 if (mdb_getopts(argc, argv,
2697 'v', MDB_OPT_SETBITS, TRUE, &verbose, TRUE, NULL) != argc)
2698 return (DCMD_USAGE);
2699
2700 if (mdb_vread(&ace, sizeof (ace_t), addr) == -1) {
2701 mdb_warn("failed to read ace_t");
2702 return (DCMD_ERR);
2703 }
2704
2705 if ((ace.a_flags & ACE_TYPE_FLAGS) == 0 ||
2706 (ace.a_flags & ACE_TYPE_FLAGS) == ACE_IDENTIFIER_GROUP)
2707 id = ace.a_who;
2708 else
2709 id = -1;
2710
2711 return (zfs_ace_print_common(addr, flags, id, ace.a_access_mask,
2712 ace.a_flags, ace.a_type, verbose));
2713 }
2714
2715 typedef struct acl_dump_args {
2716 int a_argc;
2717 const mdb_arg_t *a_argv;
2718 uint16_t a_version;
2719 int a_flags;
2720 } acl_dump_args_t;
2721
2722 /* ARGSUSED */
2723 static int
2724 acl_aces_cb(uintptr_t addr, const void *unknown, void *arg)
2725 {
2726 acl_dump_args_t *acl_args = (acl_dump_args_t *)arg;
2727
2728 if (acl_args->a_version == 1) {
2729 if (mdb_call_dcmd("zfs_ace", addr,
2730 DCMD_ADDRSPEC|acl_args->a_flags, acl_args->a_argc,
2731 acl_args->a_argv) != DCMD_OK) {
2732 return (WALK_ERR);
2733 }
2734 } else {
2735 if (mdb_call_dcmd("zfs_ace0", addr,
2736 DCMD_ADDRSPEC|acl_args->a_flags, acl_args->a_argc,
2737 acl_args->a_argv) != DCMD_OK) {
2738 return (WALK_ERR);
2739 }
2740 }
2741 acl_args->a_flags = DCMD_LOOP;
2742 return (WALK_NEXT);
2743 }
2744
2745 /* ARGSUSED */
2746 static int
2747 acl_cb(uintptr_t addr, const void *unknown, void *arg)
2748 {
2749 acl_dump_args_t *acl_args = (acl_dump_args_t *)arg;
2750
2751 if (acl_args->a_version == 1) {
2752 if (mdb_pwalk("zfs_acl_node_aces", acl_aces_cb,
2753 arg, addr) != 0) {
2754 mdb_warn("can't walk ACEs");
2755 return (DCMD_ERR);
2756 }
2757 } else {
2758 if (mdb_pwalk("zfs_acl_node_aces0", acl_aces_cb,
2759 arg, addr) != 0) {
2760 mdb_warn("can't walk ACEs");
2761 return (DCMD_ERR);
2762 }
2763 }
2764 return (WALK_NEXT);
2765 }
2766
2767 /* ARGSUSED */
2768 static int
2769 zfs_acl_dump(uintptr_t addr, uint_t flags, int argc, const mdb_arg_t *argv)
2770 {
2771 zfs_acl_t zacl;
2772 int verbose = FALSE;
2773 acl_dump_args_t acl_args;
2774
2775 if (!(flags & DCMD_ADDRSPEC))
2776 return (DCMD_USAGE);
2777
2778 if (mdb_getopts(argc, argv,
2779 'v', MDB_OPT_SETBITS, TRUE, &verbose, TRUE, NULL) != argc)
2780 return (DCMD_USAGE);
2781
2782 if (mdb_vread(&zacl, sizeof (zfs_acl_t), addr) == -1) {
2783 mdb_warn("failed to read zfs_acl_t");
2784 return (DCMD_ERR);
2785 }
2786
2787 acl_args.a_argc = argc;
2788 acl_args.a_argv = argv;
2789 acl_args.a_version = zacl.z_version;
2790 acl_args.a_flags = DCMD_LOOPFIRST;
2791
2792 if (mdb_pwalk("zfs_acl_node", acl_cb, &acl_args, addr) != 0) {
2793 mdb_warn("can't walk ACL");
2794 return (DCMD_ERR);
2795 }
2796
2797 return (DCMD_OK);
2798 }
2799
2800 /* ARGSUSED */
2801 static int
2802 zfs_acl_node_walk_init(mdb_walk_state_t *wsp)
2803 {
2804 if (wsp->walk_addr == NULL) {
2805 mdb_warn("must supply address of zfs_acl_node_t\n");
2806 return (WALK_ERR);
2807 }
2808
2809 wsp->walk_addr += OFFSETOF(zfs_acl_t, z_acl);
2810
2811 if (mdb_layered_walk("list", wsp) == -1) {
2812 mdb_warn("failed to walk 'list'\n");
2813 return (WALK_ERR);
2814 }
2815
2816 return (WALK_NEXT);
2817 }
2818
2819 static int
2820 zfs_acl_node_walk_step(mdb_walk_state_t *wsp)
2821 {
2822 zfs_acl_node_t aclnode;
2823
2824 if (mdb_vread(&aclnode, sizeof (zfs_acl_node_t),
2825 wsp->walk_addr) == -1) {
2826 mdb_warn("failed to read zfs_acl_node at %p", wsp->walk_addr);
2827 return (WALK_ERR);
2828 }
2829
2830 return (wsp->walk_callback(wsp->walk_addr, &aclnode, wsp->walk_cbdata));
2831 }
2832
2833 typedef struct ace_walk_data {
2834 int ace_count;
2835 int ace_version;
2836 } ace_walk_data_t;
2837
2838 static int
2839 zfs_aces_walk_init_common(mdb_walk_state_t *wsp, int version,
2840 int ace_count, uintptr_t ace_data)
2841 {
2842 ace_walk_data_t *ace_walk_data;
2843
2844 if (wsp->walk_addr == NULL) {
2845 mdb_warn("must supply address of zfs_acl_node_t\n");
2846 return (WALK_ERR);
2847 }
2848
2849 ace_walk_data = mdb_alloc(sizeof (ace_walk_data_t), UM_SLEEP | UM_GC);
2850
2851 ace_walk_data->ace_count = ace_count;
2852 ace_walk_data->ace_version = version;
2853
2854 wsp->walk_addr = ace_data;
2855 wsp->walk_data = ace_walk_data;
2856
2857 return (WALK_NEXT);
2858 }
2859
2860 static int
2861 zfs_acl_node_aces_walk_init_common(mdb_walk_state_t *wsp, int version)
2862 {
2863 static int gotid;
2864 static mdb_ctf_id_t acl_id;
2865 int z_ace_count;
2866 uintptr_t z_acldata;
2867
2868 if (!gotid) {
2869 if (mdb_ctf_lookup_by_name("struct zfs_acl_node",
2870 &acl_id) == -1) {
2871 mdb_warn("couldn't find struct zfs_acl_node");
2872 return (DCMD_ERR);
2873 }
2874 gotid = TRUE;
2875 }
2876
2877 if (GETMEMBID(wsp->walk_addr, &acl_id, z_ace_count, z_ace_count)) {
2878 return (DCMD_ERR);
2879 }
2880 if (GETMEMBID(wsp->walk_addr, &acl_id, z_acldata, z_acldata)) {
2881 return (DCMD_ERR);
2882 }
2883
2884 return (zfs_aces_walk_init_common(wsp, version,
2885 z_ace_count, z_acldata));
2886 }
2887
2888 /* ARGSUSED */
2889 static int
2890 zfs_acl_node_aces_walk_init(mdb_walk_state_t *wsp)
2891 {
2892 return (zfs_acl_node_aces_walk_init_common(wsp, 1));
2893 }
2894
2895 /* ARGSUSED */
2896 static int
2897 zfs_acl_node_aces0_walk_init(mdb_walk_state_t *wsp)
2898 {
2899 return (zfs_acl_node_aces_walk_init_common(wsp, 0));
2900 }
2901
2902 static int
2903 zfs_aces_walk_step(mdb_walk_state_t *wsp)
2904 {
2905 ace_walk_data_t *ace_data = wsp->walk_data;
2906 zfs_ace_t zace;
2907 ace_t *acep;
2908 int status;
2909 int entry_type;
2910 int allow_type;
2911 uintptr_t ptr;
2912
2913 if (ace_data->ace_count == 0)
2914 return (WALK_DONE);
2915
2916 if (mdb_vread(&zace, sizeof (zfs_ace_t), wsp->walk_addr) == -1) {
2917 mdb_warn("failed to read zfs_ace_t at %#lx",
2918 wsp->walk_addr);
2919 return (WALK_ERR);
2920 }
2921
2922 switch (ace_data->ace_version) {
2923 case 0:
2924 acep = (ace_t *)&zace;
2925 entry_type = acep->a_flags & ACE_TYPE_FLAGS;
2926 allow_type = acep->a_type;
2927 break;
2928 case 1:
2929 entry_type = zace.z_hdr.z_flags & ACE_TYPE_FLAGS;
2930 allow_type = zace.z_hdr.z_type;
2931 break;
2932 default:
2933 return (WALK_ERR);
2934 }
2935
2936 ptr = (uintptr_t)wsp->walk_addr;
2937 switch (entry_type) {
2938 case ACE_OWNER:
2939 case ACE_EVERYONE:
2940 case (ACE_IDENTIFIER_GROUP | ACE_GROUP):
2941 ptr += ace_data->ace_version == 0 ?
2942 sizeof (ace_t) : sizeof (zfs_ace_hdr_t);
2943 break;
2944 case ACE_IDENTIFIER_GROUP:
2945 default:
2946 switch (allow_type) {
2947 case ACE_ACCESS_ALLOWED_OBJECT_ACE_TYPE:
2948 case ACE_ACCESS_DENIED_OBJECT_ACE_TYPE:
2949 case ACE_SYSTEM_AUDIT_OBJECT_ACE_TYPE:
2950 case ACE_SYSTEM_ALARM_OBJECT_ACE_TYPE:
2951 ptr += ace_data->ace_version == 0 ?
2952 sizeof (ace_t) : sizeof (zfs_object_ace_t);
2953 break;
2954 default:
2955 ptr += ace_data->ace_version == 0 ?
2956 sizeof (ace_t) : sizeof (zfs_ace_t);
2957 break;
2958 }
2959 }
2960
2961 ace_data->ace_count--;
2962 status = wsp->walk_callback(wsp->walk_addr,
2963 (void *)(uintptr_t)&zace, wsp->walk_cbdata);
2964
2965 wsp->walk_addr = ptr;
2966 return (status);
2967 }
2968
2969 /*
2970 * MDB module linkage information:
2971 *
2972 * We declare a list of structures describing our dcmds, and a function
2973 * named _mdb_init to return a pointer to our module information.
2974 */
2975
2976 static const mdb_dcmd_t dcmds[] = {
2977 { "arc", "[-bkmg]", "print ARC variables", arc_print },
2978 { "blkptr", ":", "print blkptr_t", blkptr },
2979 { "dbuf", ":", "print dmu_buf_impl_t", dbuf },
2980 { "dbuf_stats", ":", "dbuf stats", dbuf_stats },
2981 { "dbufs",
2982 "\t[-O objset_t*] [-n objset_name | \"mos\"] "
2983 "[-o object | \"mdn\"] \n"
2984 "\t[-l level] [-b blkid | \"bonus\"]",
2985 "find dmu_buf_impl_t's that match specified criteria", dbufs },
2986 { "abuf_find", "dva_word[0] dva_word[1]",
2987 "find arc_buf_hdr_t of a specified DVA",
2988 abuf_find },
2989 { "spa", "?[-cv]", "spa_t summary", spa_print },
2990 { "spa_config", ":", "print spa_t configuration", spa_print_config },
2991 { "spa_verify", ":", "verify spa_t consistency", spa_verify },
2992 { "spa_space", ":[-b]", "print spa_t on-disk space usage", spa_space },
2993 { "spa_vdevs", ":", "given a spa_t, print vdev summary", spa_vdevs },
2994 { "vdev", ":[-re]\n"
2995 "\t-r display recursively\n"
2996 "\t-e print statistics",
2997 "vdev_t summary", vdev_print },
2998 { "zio", ":[cpr]\n"
2999 "\t-c display children\n"
3000 "\t-p display parents\n"
3001 "\t-r display recursively",
3002 "zio_t summary", zio_print },
3003 { "zio_state", "?", "print out all zio_t structures on system or "
3004 "for a particular pool", zio_state },
3005 { "zfs_blkstats", ":[-v]",
3006 "given a spa_t, print block type stats from last scrub",
3007 zfs_blkstats },
3008 { "zfs_params", "", "print zfs tunable parameters", zfs_params },
3009 { "refcount", "", "print refcount_t holders", refcount },
3010 { "zap_leaf", "", "print zap_leaf_phys_t", zap_leaf },
3011 { "zfs_aces", ":[-v]", "print all ACEs from a zfs_acl_t",
3012 zfs_acl_dump },
3013 { "zfs_ace", ":[-v]", "print zfs_ace", zfs_ace_print },
3014 { "zfs_ace0", ":[-v]", "print zfs_ace0", zfs_ace0_print },
3015 { "sa_attr_table", ":", "print SA attribute table from sa_os_t",
3016 sa_attr_table},
3017 { "sa_attr", ": attr_id",
3018 "print SA attribute address when given sa_handle_t", sa_attr_print},
3019 { "zfs_dbgmsg", ":[-v]",
3020 "print zfs debug log", dbgmsg},
3021 { NULL }
3022 };
3023
3024 static const mdb_walker_t walkers[] = {
3025 { "zms_freelist", "walk ZFS metaslab freelist",
3026 freelist_walk_init, freelist_walk_step, NULL },
3027 { "txg_list", "given any txg_list_t *, walk all entries in all txgs",
3028 txg_list_walk_init, txg_list_walk_step, NULL },
3029 { "txg_list0", "given any txg_list_t *, walk all entries in txg 0",
3030 txg_list0_walk_init, txg_list_walk_step, NULL },
3031 { "txg_list1", "given any txg_list_t *, walk all entries in txg 1",
3032 txg_list1_walk_init, txg_list_walk_step, NULL },
3033 { "txg_list2", "given any txg_list_t *, walk all entries in txg 2",
3034 txg_list2_walk_init, txg_list_walk_step, NULL },
3035 { "txg_list3", "given any txg_list_t *, walk all entries in txg 3",
3036 txg_list3_walk_init, txg_list_walk_step, NULL },
3037 { "zio", "walk all zio structures, optionally for a particular spa_t",
3038 zio_walk_init, zio_walk_step, NULL },
3039 { "zio_root", "walk all root zio_t structures, optionally for a "
3040 "particular spa_t",
3041 zio_walk_init, zio_walk_root_step, NULL },
3042 { "spa", "walk all spa_t entries in the namespace",
3043 spa_walk_init, spa_walk_step, NULL },
3044 { "metaslab", "given a spa_t *, walk all metaslab_t structures",
3045 metaslab_walk_init, metaslab_walk_step, NULL },
3046 { "zfs_acl_node", "given a zfs_acl_t, walk all zfs_acl_nodes",
3047 zfs_acl_node_walk_init, zfs_acl_node_walk_step, NULL },
3048 { "zfs_acl_node_aces", "given a zfs_acl_node_t, walk all ACEs",
3049 zfs_acl_node_aces_walk_init, zfs_aces_walk_step, NULL },
3050 { "zfs_acl_node_aces0",
3051 "given a zfs_acl_node_t, walk all ACEs as ace_t",
3052 zfs_acl_node_aces0_walk_init, zfs_aces_walk_step, NULL },
3053 { NULL }
3054 };
3055
3056 static const mdb_modinfo_t modinfo = {
3057 MDB_API_VERSION, dcmds, walkers
3058 };
3059
3060 const mdb_modinfo_t *
3061 _mdb_init(void)
3062 {
3063 return (&modinfo);
3064 }