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