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) 2013 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 extern int64_t mdb_gethrtime(void);
50 #else
51 #define ZFS_OBJ_NAME "libzpool.so.1"
52 #endif
53
54 #define ZFS_STRUCT "struct " ZFS_OBJ_NAME "`"
55
56 #ifndef _KERNEL
57 int aok;
58 #endif
59
60 static int
61 getmember(uintptr_t addr, const char *type, mdb_ctf_id_t *idp,
62 const char *member, int len, void *buf)
63 {
64 mdb_ctf_id_t id;
65 ulong_t off;
66 char name[64];
67
68 if (idp == NULL) {
69 if (mdb_ctf_lookup_by_name(type, &id) == -1) {
70 mdb_warn("couldn't find type %s", type);
71 return (DCMD_ERR);
72 }
73 idp = &id;
74 } else {
75 type = name;
76 mdb_ctf_type_name(*idp, name, sizeof (name));
77 }
78
79 if (mdb_ctf_offsetof(*idp, member, &off) == -1) {
80 mdb_warn("couldn't find member %s of type %s\n", member, type);
81 return (DCMD_ERR);
82 }
83 if (off % 8 != 0) {
84 mdb_warn("member %s of type %s is unsupported bitfield",
85 member, type);
86 return (DCMD_ERR);
87 }
88 off /= 8;
89
90 if (mdb_vread(buf, len, addr + off) == -1) {
91 mdb_warn("failed to read %s from %s at %p",
92 member, type, addr + off);
93 return (DCMD_ERR);
94 }
95 /* mdb_warn("read %s from %s at %p+%llx\n", member, type, addr, off); */
96
97 return (0);
98 }
99
100 #define GETMEMB(addr, structname, member, dest) \
101 getmember(addr, ZFS_STRUCT structname, NULL, #member, \
102 sizeof (dest), &(dest))
103
104 #define GETMEMBID(addr, ctfid, member, dest) \
105 getmember(addr, NULL, ctfid, #member, sizeof (dest), &(dest))
106
107 static boolean_t
108 strisprint(const char *cp)
109 {
110 for (; *cp; cp++) {
111 if (!isprint(*cp))
112 return (B_FALSE);
113 }
114 return (B_TRUE);
115 }
116
117 static int verbose;
118
119 static int
120 freelist_walk_init(mdb_walk_state_t *wsp)
121 {
122 if (wsp->walk_addr == NULL) {
123 mdb_warn("must supply starting address\n");
124 return (WALK_ERR);
125 }
126
127 wsp->walk_data = 0; /* Index into the freelist */
128 return (WALK_NEXT);
129 }
130
131 static int
132 freelist_walk_step(mdb_walk_state_t *wsp)
133 {
134 uint64_t entry;
135 uintptr_t number = (uintptr_t)wsp->walk_data;
136 char *ddata[] = { "ALLOC", "FREE", "CONDENSE", "INVALID",
137 "INVALID", "INVALID", "INVALID", "INVALID" };
138 int mapshift = SPA_MINBLOCKSHIFT;
139
140 if (mdb_vread(&entry, sizeof (entry), wsp->walk_addr) == -1) {
141 mdb_warn("failed to read freelist entry %p", wsp->walk_addr);
142 return (WALK_DONE);
143 }
144 wsp->walk_addr += sizeof (entry);
145 wsp->walk_data = (void *)(number + 1);
146
147 if (SM_DEBUG_DECODE(entry)) {
148 mdb_printf("DEBUG: %3u %10s: txg=%llu pass=%llu\n",
149 number,
150 ddata[SM_DEBUG_ACTION_DECODE(entry)],
151 SM_DEBUG_TXG_DECODE(entry),
152 SM_DEBUG_SYNCPASS_DECODE(entry));
153 } else {
154 mdb_printf("Entry: %3u offsets=%08llx-%08llx type=%c "
155 "size=%06llx", number,
156 SM_OFFSET_DECODE(entry) << mapshift,
157 (SM_OFFSET_DECODE(entry) + SM_RUN_DECODE(entry)) <<
158 mapshift,
159 SM_TYPE_DECODE(entry) == SM_ALLOC ? 'A' : 'F',
160 SM_RUN_DECODE(entry) << mapshift);
161 if (verbose)
162 mdb_printf(" (raw=%012llx)\n", entry);
163 mdb_printf("\n");
164 }
165 return (WALK_NEXT);
166 }
167
168 static int
169 mdb_dsl_dir_name(uintptr_t addr, char *buf)
170 {
171 static int gotid;
172 static mdb_ctf_id_t dd_id;
173 uintptr_t dd_parent;
174 char dd_myname[MAXNAMELEN];
175
176 if (!gotid) {
177 if (mdb_ctf_lookup_by_name(ZFS_STRUCT "dsl_dir",
178 &dd_id) == -1) {
179 mdb_warn("couldn't find struct dsl_dir");
180 return (DCMD_ERR);
181 }
182 gotid = TRUE;
183 }
184 if (GETMEMBID(addr, &dd_id, dd_parent, dd_parent) ||
185 GETMEMBID(addr, &dd_id, dd_myname, dd_myname)) {
186 return (DCMD_ERR);
187 }
188
189 if (dd_parent) {
190 if (mdb_dsl_dir_name(dd_parent, buf))
191 return (DCMD_ERR);
192 strcat(buf, "/");
193 }
194
195 if (dd_myname[0])
196 strcat(buf, dd_myname);
197 else
198 strcat(buf, "???");
199
200 return (0);
201 }
202
203 static int
204 objset_name(uintptr_t addr, char *buf)
205 {
206 static int gotid;
207 static mdb_ctf_id_t os_id, ds_id;
208 uintptr_t os_dsl_dataset;
209 char ds_snapname[MAXNAMELEN];
210 uintptr_t ds_dir;
211
212 buf[0] = '\0';
213
214 if (!gotid) {
215 if (mdb_ctf_lookup_by_name(ZFS_STRUCT "objset",
216 &os_id) == -1) {
217 mdb_warn("couldn't find struct objset");
218 return (DCMD_ERR);
219 }
220 if (mdb_ctf_lookup_by_name(ZFS_STRUCT "dsl_dataset",
221 &ds_id) == -1) {
222 mdb_warn("couldn't find struct dsl_dataset");
223 return (DCMD_ERR);
224 }
225
226 gotid = TRUE;
227 }
228
229 if (GETMEMBID(addr, &os_id, os_dsl_dataset, os_dsl_dataset))
230 return (DCMD_ERR);
231
232 if (os_dsl_dataset == 0) {
233 strcat(buf, "mos");
234 return (0);
235 }
236
237 if (GETMEMBID(os_dsl_dataset, &ds_id, ds_snapname, ds_snapname) ||
238 GETMEMBID(os_dsl_dataset, &ds_id, ds_dir, ds_dir)) {
239 return (DCMD_ERR);
240 }
241
242 if (ds_dir && mdb_dsl_dir_name(ds_dir, buf))
243 return (DCMD_ERR);
244
245 if (ds_snapname[0]) {
246 strcat(buf, "@");
247 strcat(buf, ds_snapname);
248 }
249 return (0);
250 }
251
252 static void
253 enum_lookup(char *out, size_t size, mdb_ctf_id_t id, int val,
254 const char *prefix)
255 {
256 const char *cp;
257 size_t len = strlen(prefix);
258
259 if ((cp = mdb_ctf_enum_name(id, val)) != NULL) {
260 if (strncmp(cp, prefix, len) == 0)
261 cp += len;
262 (void) strncpy(out, cp, size);
263 } else {
264 mdb_snprintf(out, size, "? (%d)", val);
265 }
266 }
267
268 /* ARGSUSED */
269 static int
270 zfs_params(uintptr_t addr, uint_t flags, int argc, const mdb_arg_t *argv)
271 {
272 /*
273 * This table can be approximately generated by running:
274 * egrep "^[a-z0-9_]+ [a-z0-9_]+( =.*)?;" *.c | cut -d ' ' -f 2
275 */
276 static const char *params[] = {
277 "arc_reduce_dnlc_percent",
278 "zfs_arc_max",
279 "zfs_arc_min",
280 "arc_shrink_shift",
281 "zfs_mdcomp_disable",
282 "zfs_prefetch_disable",
283 "zfetch_max_streams",
284 "zfetch_min_sec_reap",
285 "zfetch_block_cap",
286 "zfetch_array_rd_sz",
287 "zfs_default_bs",
288 "zfs_default_ibs",
289 "metaslab_aliquot",
290 "reference_tracking_enable",
291 "reference_history",
292 "spa_max_replication_override",
293 "spa_mode_global",
294 "zfs_flags",
295 "zfs_txg_synctime_ms",
296 "zfs_txg_timeout",
297 "zfs_write_limit_min",
298 "zfs_write_limit_max",
299 "zfs_write_limit_shift",
300 "zfs_write_limit_override",
301 "zfs_no_write_throttle",
302 "zfs_vdev_cache_max",
303 "zfs_vdev_cache_size",
304 "zfs_vdev_cache_bshift",
305 "vdev_mirror_shift",
306 "zfs_vdev_max_pending",
307 "zfs_vdev_min_pending",
308 "zfs_scrub_limit",
309 "zfs_no_scrub_io",
310 "zfs_no_scrub_prefetch",
311 "zfs_vdev_time_shift",
312 "zfs_vdev_ramp_rate",
313 "zfs_vdev_aggregation_limit",
314 "fzap_default_block_shift",
315 "zfs_immediate_write_sz",
316 "zfs_read_chunk_size",
317 "zfs_nocacheflush",
318 "zil_replay_disable",
319 "metaslab_gang_bang",
320 "metaslab_df_alloc_threshold",
321 "metaslab_df_free_pct",
322 "zio_injection_enabled",
323 "zvol_immediate_write_sz",
324 };
325
326 for (int i = 0; i < sizeof (params) / sizeof (params[0]); i++) {
327 int sz;
328 uint64_t val64;
329 uint32_t *val32p = (uint32_t *)&val64;
330
331 sz = mdb_readvar(&val64, params[i]);
332 if (sz == 4) {
333 mdb_printf("%s = 0x%x\n", params[i], *val32p);
334 } else if (sz == 8) {
335 mdb_printf("%s = 0x%llx\n", params[i], val64);
336 } else {
337 mdb_warn("variable %s not found", params[i]);
338 }
339 }
340
341 return (DCMD_OK);
342 }
343
344 /* ARGSUSED */
345 static int
346 blkptr(uintptr_t addr, uint_t flags, int argc, const mdb_arg_t *argv)
347 {
348 mdb_ctf_id_t type_enum, checksum_enum, compress_enum;
349 char type[80], checksum[80], compress[80];
350 blkptr_t blk, *bp = &blk;
351 char buf[BP_SPRINTF_LEN];
352
353 if (mdb_vread(&blk, sizeof (blkptr_t), addr) == -1) {
354 mdb_warn("failed to read blkptr_t");
355 return (DCMD_ERR);
356 }
357
358 if (mdb_ctf_lookup_by_name("enum dmu_object_type", &type_enum) == -1 ||
359 mdb_ctf_lookup_by_name("enum zio_checksum", &checksum_enum) == -1 ||
360 mdb_ctf_lookup_by_name("enum zio_compress", &compress_enum) == -1) {
361 mdb_warn("Could not find blkptr enumerated types");
362 return (DCMD_ERR);
363 }
364
365 enum_lookup(type, sizeof (type), type_enum,
366 BP_GET_TYPE(bp), "DMU_OT_");
367 enum_lookup(checksum, sizeof (checksum), checksum_enum,
368 BP_GET_CHECKSUM(bp), "ZIO_CHECKSUM_");
369 enum_lookup(compress, sizeof (compress), compress_enum,
370 BP_GET_COMPRESS(bp), "ZIO_COMPRESS_");
371
372 SPRINTF_BLKPTR(mdb_snprintf, '\n', buf, bp, type, checksum, compress);
373
374 mdb_printf("%s\n", buf);
375
376 return (DCMD_OK);
377 }
378
379 typedef struct mdb_dmu_buf_impl {
380 struct {
381 uint64_t db_object;
382 } db;
383 void *db_objset;
384 uint64_t db_level;
385 uint64_t db_blkid;
386 struct {
387 uint64_t rc_count;
388 } db_holds;
389 } mdb_dmu_buf_impl_t;
390
391 /* ARGSUSED */
392 static int
393 dbuf(uintptr_t addr, uint_t flags, int argc, const mdb_arg_t *argv)
394 {
395 mdb_dmu_buf_impl_t db;
396 char objectname[32];
397 char blkidname[32];
398 char path[MAXNAMELEN];
399
400 if (DCMD_HDRSPEC(flags))
401 mdb_printf(" addr object lvl blkid holds os\n");
402
403 if (mdb_ctf_vread(&db, ZFS_STRUCT "dmu_buf_impl", "mdb_dmu_buf_impl_t",
404 addr, 0) == -1)
405 return (DCMD_ERR);
406
407 if (db.db.db_object == DMU_META_DNODE_OBJECT)
408 (void) strcpy(objectname, "mdn");
409 else
410 (void) mdb_snprintf(objectname, sizeof (objectname), "%llx",
411 (u_longlong_t)db.db.db_object);
412
413 if (db.db_blkid == DMU_BONUS_BLKID)
414 (void) strcpy(blkidname, "bonus");
415 else
416 (void) mdb_snprintf(blkidname, sizeof (blkidname), "%llx",
417 (u_longlong_t)db.db_blkid);
418
419 if (objset_name((uintptr_t)db.db_objset, path)) {
420 return (DCMD_ERR);
421 }
422
423 mdb_printf("%p %8s %1u %9s %2llu %s\n", addr,
424 objectname, (int)db.db_level, blkidname,
425 db.db_holds.rc_count, path);
426
427 return (DCMD_OK);
428 }
429
430 /* ARGSUSED */
431 static int
432 dbuf_stats(uintptr_t addr, uint_t flags, int argc, const mdb_arg_t *argv)
433 {
434 #define HISTOSZ 32
435 uintptr_t dbp;
436 dmu_buf_impl_t db;
437 dbuf_hash_table_t ht;
438 uint64_t bucket, ndbufs;
439 uint64_t histo[HISTOSZ];
440 uint64_t histo2[HISTOSZ];
441 int i, maxidx;
442
443 if (mdb_readvar(&ht, "dbuf_hash_table") == -1) {
444 mdb_warn("failed to read 'dbuf_hash_table'");
445 return (DCMD_ERR);
446 }
447
448 for (i = 0; i < HISTOSZ; i++) {
449 histo[i] = 0;
450 histo2[i] = 0;
451 }
452
453 ndbufs = 0;
454 for (bucket = 0; bucket < ht.hash_table_mask+1; bucket++) {
455 int len;
456
457 if (mdb_vread(&dbp, sizeof (void *),
458 (uintptr_t)(ht.hash_table+bucket)) == -1) {
459 mdb_warn("failed to read hash bucket %u at %p",
460 bucket, ht.hash_table+bucket);
461 return (DCMD_ERR);
462 }
463
464 len = 0;
465 while (dbp != 0) {
466 if (mdb_vread(&db, sizeof (dmu_buf_impl_t),
467 dbp) == -1) {
468 mdb_warn("failed to read dbuf at %p", dbp);
469 return (DCMD_ERR);
470 }
471 dbp = (uintptr_t)db.db_hash_next;
472 for (i = MIN(len, HISTOSZ - 1); i >= 0; i--)
473 histo2[i]++;
474 len++;
475 ndbufs++;
476 }
477
478 if (len >= HISTOSZ)
479 len = HISTOSZ-1;
480 histo[len]++;
481 }
482
483 mdb_printf("hash table has %llu buckets, %llu dbufs "
484 "(avg %llu buckets/dbuf)\n",
485 ht.hash_table_mask+1, ndbufs,
486 (ht.hash_table_mask+1)/ndbufs);
487
488 mdb_printf("\n");
489 maxidx = 0;
490 for (i = 0; i < HISTOSZ; i++)
491 if (histo[i] > 0)
492 maxidx = i;
493 mdb_printf("hash chain length number of buckets\n");
494 for (i = 0; i <= maxidx; i++)
495 mdb_printf("%u %llu\n", i, histo[i]);
496
497 mdb_printf("\n");
498 maxidx = 0;
499 for (i = 0; i < HISTOSZ; i++)
500 if (histo2[i] > 0)
501 maxidx = i;
502 mdb_printf("hash chain depth number of dbufs\n");
503 for (i = 0; i <= maxidx; i++)
504 mdb_printf("%u or more %llu %llu%%\n",
505 i, histo2[i], histo2[i]*100/ndbufs);
506
507
508 return (DCMD_OK);
509 }
510
511 #define CHAIN_END 0xffff
512 /*
513 * ::zap_leaf [-v]
514 *
515 * Print a zap_leaf_phys_t, assumed to be 16k
516 */
517 /* ARGSUSED */
518 static int
519 zap_leaf(uintptr_t addr, uint_t flags, int argc, const mdb_arg_t *argv)
520 {
521 char buf[16*1024];
522 int verbose = B_FALSE;
523 int four = B_FALSE;
524 zap_leaf_t l = { 0 };
525 zap_leaf_phys_t *zlp = (void *)buf;
526 int i;
527
528 if (mdb_getopts(argc, argv,
529 'v', MDB_OPT_SETBITS, TRUE, &verbose,
530 '4', MDB_OPT_SETBITS, TRUE, &four,
531 NULL) != argc)
532 return (DCMD_USAGE);
533
534 l.l_phys = zlp;
535 l.l_bs = 14; /* assume 16k blocks */
536 if (four)
537 l.l_bs = 12;
538
539 if (!(flags & DCMD_ADDRSPEC)) {
540 return (DCMD_USAGE);
541 }
542
543 if (mdb_vread(buf, sizeof (buf), addr) == -1) {
544 mdb_warn("failed to read zap_leaf_phys_t at %p", addr);
545 return (DCMD_ERR);
546 }
547
548 if (zlp->l_hdr.lh_block_type != ZBT_LEAF ||
549 zlp->l_hdr.lh_magic != ZAP_LEAF_MAGIC) {
550 mdb_warn("This does not appear to be a zap_leaf_phys_t");
551 return (DCMD_ERR);
552 }
553
554 mdb_printf("zap_leaf_phys_t at %p:\n", addr);
555 mdb_printf(" lh_prefix_len = %u\n", zlp->l_hdr.lh_prefix_len);
556 mdb_printf(" lh_prefix = %llx\n", zlp->l_hdr.lh_prefix);
557 mdb_printf(" lh_nentries = %u\n", zlp->l_hdr.lh_nentries);
558 mdb_printf(" lh_nfree = %u\n", zlp->l_hdr.lh_nfree,
559 zlp->l_hdr.lh_nfree * 100 / (ZAP_LEAF_NUMCHUNKS(&l)));
560 mdb_printf(" lh_freelist = %u\n", zlp->l_hdr.lh_freelist);
561 mdb_printf(" lh_flags = %x (%s)\n", zlp->l_hdr.lh_flags,
562 zlp->l_hdr.lh_flags & ZLF_ENTRIES_CDSORTED ?
563 "ENTRIES_CDSORTED" : "");
564
565 if (verbose) {
566 mdb_printf(" hash table:\n");
567 for (i = 0; i < ZAP_LEAF_HASH_NUMENTRIES(&l); i++) {
568 if (zlp->l_hash[i] != CHAIN_END)
569 mdb_printf(" %u: %u\n", i, zlp->l_hash[i]);
570 }
571 }
572
573 mdb_printf(" chunks:\n");
574 for (i = 0; i < ZAP_LEAF_NUMCHUNKS(&l); i++) {
575 /* LINTED: alignment */
576 zap_leaf_chunk_t *zlc = &ZAP_LEAF_CHUNK(&l, i);
577 switch (zlc->l_entry.le_type) {
578 case ZAP_CHUNK_FREE:
579 if (verbose) {
580 mdb_printf(" %u: free; lf_next = %u\n",
581 i, zlc->l_free.lf_next);
582 }
583 break;
584 case ZAP_CHUNK_ENTRY:
585 mdb_printf(" %u: entry\n", i);
586 if (verbose) {
587 mdb_printf(" le_next = %u\n",
588 zlc->l_entry.le_next);
589 }
590 mdb_printf(" le_name_chunk = %u\n",
591 zlc->l_entry.le_name_chunk);
592 mdb_printf(" le_name_numints = %u\n",
593 zlc->l_entry.le_name_numints);
594 mdb_printf(" le_value_chunk = %u\n",
595 zlc->l_entry.le_value_chunk);
596 mdb_printf(" le_value_intlen = %u\n",
597 zlc->l_entry.le_value_intlen);
598 mdb_printf(" le_value_numints = %u\n",
599 zlc->l_entry.le_value_numints);
600 mdb_printf(" le_cd = %u\n",
601 zlc->l_entry.le_cd);
602 mdb_printf(" le_hash = %llx\n",
603 zlc->l_entry.le_hash);
604 break;
605 case ZAP_CHUNK_ARRAY:
606 mdb_printf(" %u: array", i);
607 if (strisprint((char *)zlc->l_array.la_array))
608 mdb_printf(" \"%s\"", zlc->l_array.la_array);
609 mdb_printf("\n");
610 if (verbose) {
611 int j;
612 mdb_printf(" ");
613 for (j = 0; j < ZAP_LEAF_ARRAY_BYTES; j++) {
614 mdb_printf("%02x ",
615 zlc->l_array.la_array[j]);
616 }
617 mdb_printf("\n");
618 }
619 if (zlc->l_array.la_next != CHAIN_END) {
620 mdb_printf(" lf_next = %u\n",
621 zlc->l_array.la_next);
622 }
623 break;
624 default:
625 mdb_printf(" %u: undefined type %u\n",
626 zlc->l_entry.le_type);
627 }
628 }
629
630 return (DCMD_OK);
631 }
632
633 typedef struct dbufs_data {
634 mdb_ctf_id_t id;
635 uint64_t objset;
636 uint64_t object;
637 uint64_t level;
638 uint64_t blkid;
639 char *osname;
640 } dbufs_data_t;
641
642 #define DBUFS_UNSET (0xbaddcafedeadbeefULL)
643
644 /* ARGSUSED */
645 static int
646 dbufs_cb(uintptr_t addr, const void *unknown, void *arg)
647 {
648 dbufs_data_t *data = arg;
649 uintptr_t objset;
650 dmu_buf_t db;
651 uint8_t level;
652 uint64_t blkid;
653 char osname[MAXNAMELEN];
654
655 if (GETMEMBID(addr, &data->id, db_objset, objset) ||
656 GETMEMBID(addr, &data->id, db, db) ||
657 GETMEMBID(addr, &data->id, db_level, level) ||
658 GETMEMBID(addr, &data->id, db_blkid, blkid)) {
659 return (WALK_ERR);
660 }
661
662 if ((data->objset == DBUFS_UNSET || data->objset == objset) &&
663 (data->osname == NULL || (objset_name(objset, osname) == 0 &&
664 strcmp(data->osname, osname) == 0)) &&
665 (data->object == DBUFS_UNSET || data->object == db.db_object) &&
666 (data->level == DBUFS_UNSET || data->level == level) &&
667 (data->blkid == DBUFS_UNSET || data->blkid == blkid)) {
668 mdb_printf("%#lr\n", addr);
669 }
670 return (WALK_NEXT);
671 }
672
673 /* ARGSUSED */
674 static int
675 dbufs(uintptr_t addr, uint_t flags, int argc, const mdb_arg_t *argv)
676 {
677 dbufs_data_t data;
678 char *object = NULL;
679 char *blkid = NULL;
680
681 data.objset = data.object = data.level = data.blkid = DBUFS_UNSET;
682 data.osname = NULL;
683
684 if (mdb_getopts(argc, argv,
685 'O', MDB_OPT_UINT64, &data.objset,
686 'n', MDB_OPT_STR, &data.osname,
687 'o', MDB_OPT_STR, &object,
688 'l', MDB_OPT_UINT64, &data.level,
689 'b', MDB_OPT_STR, &blkid) != argc) {
690 return (DCMD_USAGE);
691 }
692
693 if (object) {
694 if (strcmp(object, "mdn") == 0) {
695 data.object = DMU_META_DNODE_OBJECT;
696 } else {
697 data.object = mdb_strtoull(object);
698 }
699 }
700
701 if (blkid) {
702 if (strcmp(blkid, "bonus") == 0) {
703 data.blkid = DMU_BONUS_BLKID;
704 } else {
705 data.blkid = mdb_strtoull(blkid);
706 }
707 }
708
709 if (mdb_ctf_lookup_by_name(ZFS_STRUCT "dmu_buf_impl", &data.id) == -1) {
710 mdb_warn("couldn't find struct dmu_buf_impl_t");
711 return (DCMD_ERR);
712 }
713
714 if (mdb_walk("dmu_buf_impl_t", dbufs_cb, &data) != 0) {
715 mdb_warn("can't walk dbufs");
716 return (DCMD_ERR);
717 }
718
719 return (DCMD_OK);
720 }
721
722 typedef struct abuf_find_data {
723 dva_t dva;
724 mdb_ctf_id_t id;
725 } abuf_find_data_t;
726
727 /* ARGSUSED */
728 static int
729 abuf_find_cb(uintptr_t addr, const void *unknown, void *arg)
730 {
731 abuf_find_data_t *data = arg;
732 dva_t dva;
733
734 if (GETMEMBID(addr, &data->id, b_dva, dva)) {
735 return (WALK_ERR);
736 }
737
738 if (dva.dva_word[0] == data->dva.dva_word[0] &&
739 dva.dva_word[1] == data->dva.dva_word[1]) {
740 mdb_printf("%#lr\n", addr);
741 }
742 return (WALK_NEXT);
743 }
744
745 /* ARGSUSED */
746 static int
747 abuf_find(uintptr_t addr, uint_t flags, int argc, const mdb_arg_t *argv)
748 {
749 abuf_find_data_t data;
750 GElf_Sym sym;
751 int i;
752 const char *syms[] = {
753 "ARC_mru",
754 "ARC_mru_ghost",
755 "ARC_mfu",
756 "ARC_mfu_ghost",
757 };
758
759 if (argc != 2)
760 return (DCMD_USAGE);
761
762 for (i = 0; i < 2; i ++) {
763 switch (argv[i].a_type) {
764 case MDB_TYPE_STRING:
765 data.dva.dva_word[i] = mdb_strtoull(argv[i].a_un.a_str);
766 break;
767 case MDB_TYPE_IMMEDIATE:
768 data.dva.dva_word[i] = argv[i].a_un.a_val;
769 break;
770 default:
771 return (DCMD_USAGE);
772 }
773 }
774
775 if (mdb_ctf_lookup_by_name(ZFS_STRUCT "arc_buf_hdr", &data.id) == -1) {
776 mdb_warn("couldn't find struct arc_buf_hdr");
777 return (DCMD_ERR);
778 }
779
780 for (i = 0; i < sizeof (syms) / sizeof (syms[0]); i++) {
781 if (mdb_lookup_by_name(syms[i], &sym)) {
782 mdb_warn("can't find symbol %s", syms[i]);
783 return (DCMD_ERR);
784 }
785
786 if (mdb_pwalk("list", abuf_find_cb, &data, sym.st_value) != 0) {
787 mdb_warn("can't walk %s", syms[i]);
788 return (DCMD_ERR);
789 }
790 }
791
792 return (DCMD_OK);
793 }
794
795
796 typedef struct dbgmsg_arg {
797 boolean_t da_verbose;
798 boolean_t da_address;
799 } dbgmsg_arg_t;
800
801 /* ARGSUSED */
802 static int
803 dbgmsg_cb(uintptr_t addr, const void *unknown, void *arg)
804 {
805 static mdb_ctf_id_t id;
806 static boolean_t gotid;
807 static ulong_t off;
808
809 dbgmsg_arg_t *da = arg;
810 time_t timestamp;
811 char buf[1024];
812
813 if (!gotid) {
814 if (mdb_ctf_lookup_by_name(ZFS_STRUCT "zfs_dbgmsg", &id) ==
815 -1) {
816 mdb_warn("couldn't find struct zfs_dbgmsg");
817 return (WALK_ERR);
818 }
819 gotid = TRUE;
820 if (mdb_ctf_offsetof(id, "zdm_msg", &off) == -1) {
821 mdb_warn("couldn't find zdm_msg");
822 return (WALK_ERR);
823 }
824 off /= 8;
825 }
826
827
828 if (GETMEMBID(addr, &id, zdm_timestamp, timestamp)) {
829 return (WALK_ERR);
830 }
831
832 if (mdb_readstr(buf, sizeof (buf), addr + off) == -1) {
833 mdb_warn("failed to read zdm_msg at %p\n", addr + off);
834 return (DCMD_ERR);
835 }
836
837 if (da->da_address)
838 mdb_printf("%p ", addr);
839 if (da->da_verbose)
840 mdb_printf("%Y ", timestamp);
841
842 mdb_printf("%s\n", buf);
843
844 if (da->da_verbose)
845 (void) mdb_call_dcmd("whatis", addr, DCMD_ADDRSPEC, 0, NULL);
846
847 return (WALK_NEXT);
848 }
849
850 /* ARGSUSED */
851 static int
852 dbgmsg(uintptr_t addr, uint_t flags, int argc, const mdb_arg_t *argv)
853 {
854 GElf_Sym sym;
855 dbgmsg_arg_t da = { 0 };
856
857 if (mdb_getopts(argc, argv,
858 'v', MDB_OPT_SETBITS, B_TRUE, &da.da_verbose,
859 'a', MDB_OPT_SETBITS, B_TRUE, &da.da_address,
860 NULL) != argc)
861 return (DCMD_USAGE);
862
863 if (mdb_lookup_by_obj(ZFS_OBJ_NAME, "zfs_dbgmsgs", &sym)) {
864 mdb_warn("can't find zfs_dbgmsgs");
865 return (DCMD_ERR);
866 }
867
868 if (mdb_pwalk("list", dbgmsg_cb, &da, sym.st_value) != 0) {
869 mdb_warn("can't walk zfs_dbgmsgs");
870 return (DCMD_ERR);
871 }
872
873 return (DCMD_OK);
874 }
875
876 /*ARGSUSED*/
877 static int
878 arc_print(uintptr_t addr, uint_t flags, int argc, const mdb_arg_t *argv)
879 {
880 kstat_named_t *stats;
881 GElf_Sym sym;
882 int nstats, i;
883 uint_t opt_a = FALSE;
884 uint_t opt_b = FALSE;
885 uint_t shift = 0;
886 const char *suffix;
887
888 static const char *bytestats[] = {
889 "p", "c", "c_min", "c_max", "size", "duplicate_buffers_size",
890 "arc_meta_used", "arc_meta_limit", "arc_meta_max",
891 NULL
892 };
893
894 static const char *extras[] = {
895 "arc_no_grow", "arc_tempreserve",
896 NULL
897 };
898
899 if (mdb_lookup_by_name("arc_stats", &sym) == -1) {
900 mdb_warn("failed to find 'arc_stats'");
901 return (DCMD_ERR);
902 }
903
904 stats = mdb_zalloc(sym.st_size, UM_SLEEP | UM_GC);
905
906 if (mdb_vread(stats, sym.st_size, sym.st_value) == -1) {
907 mdb_warn("couldn't read 'arc_stats' at %p", sym.st_value);
908 return (DCMD_ERR);
909 }
910
911 nstats = sym.st_size / sizeof (kstat_named_t);
912
913 /* NB: -a / opt_a are ignored for backwards compatability */
914 if (mdb_getopts(argc, argv,
915 'a', MDB_OPT_SETBITS, TRUE, &opt_a,
916 'b', MDB_OPT_SETBITS, TRUE, &opt_b,
917 'k', MDB_OPT_SETBITS, 10, &shift,
918 'm', MDB_OPT_SETBITS, 20, &shift,
919 'g', MDB_OPT_SETBITS, 30, &shift,
920 NULL) != argc)
921 return (DCMD_USAGE);
922
923 if (!opt_b && !shift)
924 shift = 20;
925
926 switch (shift) {
927 case 0:
928 suffix = "B";
929 break;
930 case 10:
931 suffix = "KB";
932 break;
933 case 20:
934 suffix = "MB";
935 break;
936 case 30:
937 suffix = "GB";
938 break;
939 default:
940 suffix = "XX";
941 }
942
943 for (i = 0; i < nstats; i++) {
944 int j;
945 boolean_t bytes = B_FALSE;
946
947 for (j = 0; bytestats[j]; j++) {
948 if (strcmp(stats[i].name, bytestats[j]) == 0) {
949 bytes = B_TRUE;
950 break;
951 }
952 }
953
954 if (bytes) {
955 mdb_printf("%-25s = %9llu %s\n", stats[i].name,
956 stats[i].value.ui64 >> shift, suffix);
957 } else {
958 mdb_printf("%-25s = %9llu\n", stats[i].name,
959 stats[i].value.ui64);
960 }
961 }
962
963 for (i = 0; extras[i]; i++) {
964 uint64_t buf;
965
966 if (mdb_lookup_by_name(extras[i], &sym) == -1) {
967 mdb_warn("failed to find '%s'", extras[i]);
968 return (DCMD_ERR);
969 }
970
971 if (sym.st_size != sizeof (uint64_t) &&
972 sym.st_size != sizeof (uint32_t)) {
973 mdb_warn("expected scalar for variable '%s'\n",
974 extras[i]);
975 return (DCMD_ERR);
976 }
977
978 if (mdb_vread(&buf, sym.st_size, sym.st_value) == -1) {
979 mdb_warn("couldn't read '%s'", extras[i]);
980 return (DCMD_ERR);
981 }
982
983 mdb_printf("%-25s = ", extras[i]);
984
985 /* NB: all the 64-bit extras happen to be byte counts */
986 if (sym.st_size == sizeof (uint64_t))
987 mdb_printf("%9llu %s\n", buf >> shift, suffix);
988
989 if (sym.st_size == sizeof (uint32_t))
990 mdb_printf("%9d\n", *((uint32_t *)&buf));
991 }
992 return (DCMD_OK);
993 }
994
995 /*
996 * ::spa
997 *
998 * -c Print configuration information as well
999 * -v Print vdev state
1000 * -e Print vdev error stats
1001 *
1002 * Print a summarized spa_t. When given no arguments, prints out a table of all
1003 * active pools on the system.
1004 */
1005 /* ARGSUSED */
1006 static int
1007 spa_print(uintptr_t addr, uint_t flags, int argc, const mdb_arg_t *argv)
1008 {
1009 spa_t spa;
1010 const char *statetab[] = { "ACTIVE", "EXPORTED", "DESTROYED",
1011 "SPARE", "L2CACHE", "UNINIT", "UNAVAIL", "POTENTIAL" };
1012 const char *state;
1013 int config = FALSE;
1014 int vdevs = FALSE;
1015 int errors = FALSE;
1016
1017 if (mdb_getopts(argc, argv,
1018 'c', MDB_OPT_SETBITS, TRUE, &config,
1019 'v', MDB_OPT_SETBITS, TRUE, &vdevs,
1020 'e', MDB_OPT_SETBITS, TRUE, &errors,
1021 NULL) != argc)
1022 return (DCMD_USAGE);
1023
1024 if (!(flags & DCMD_ADDRSPEC)) {
1025 if (mdb_walk_dcmd("spa", "spa", argc, argv) == -1) {
1026 mdb_warn("can't walk spa");
1027 return (DCMD_ERR);
1028 }
1029
1030 return (DCMD_OK);
1031 }
1032
1033 if (flags & DCMD_PIPE_OUT) {
1034 mdb_printf("%#lr\n", addr);
1035 return (DCMD_OK);
1036 }
1037
1038 if (DCMD_HDRSPEC(flags))
1039 mdb_printf("%<u>%-?s %9s %-*s%</u>\n", "ADDR", "STATE",
1040 sizeof (uintptr_t) == 4 ? 60 : 52, "NAME");
1041
1042 if (mdb_vread(&spa, sizeof (spa), addr) == -1) {
1043 mdb_warn("failed to read spa_t at %p", addr);
1044 return (DCMD_ERR);
1045 }
1046
1047 if (spa.spa_state < 0 || spa.spa_state > POOL_STATE_UNAVAIL)
1048 state = "UNKNOWN";
1049 else
1050 state = statetab[spa.spa_state];
1051
1052 mdb_printf("%0?p %9s %s\n", addr, state, spa.spa_name);
1053
1054 if (config) {
1055 mdb_printf("\n");
1056 mdb_inc_indent(4);
1057 if (mdb_call_dcmd("spa_config", addr, flags, 0,
1058 NULL) != DCMD_OK)
1059 return (DCMD_ERR);
1060 mdb_dec_indent(4);
1061 }
1062
1063 if (vdevs || errors) {
1064 mdb_arg_t v;
1065
1066 v.a_type = MDB_TYPE_STRING;
1067 v.a_un.a_str = "-e";
1068
1069 mdb_printf("\n");
1070 mdb_inc_indent(4);
1071 if (mdb_call_dcmd("spa_vdevs", addr, flags, errors ? 1 : 0,
1072 &v) != DCMD_OK)
1073 return (DCMD_ERR);
1074 mdb_dec_indent(4);
1075 }
1076
1077 return (DCMD_OK);
1078 }
1079
1080 typedef struct mdb_spa_config_spa {
1081 nvlist_t *spa_config;
1082 } mdb_spa_config_spa_t;
1083
1084 /*
1085 * ::spa_config
1086 *
1087 * Given a spa_t, print the configuration information stored in spa_config.
1088 * Since it's just an nvlist, format it as an indented list of name=value pairs.
1089 * We simply read the value of spa_config and pass off to ::nvlist.
1090 */
1091 /* ARGSUSED */
1092 static int
1093 spa_print_config(uintptr_t addr, uint_t flags, int argc, const mdb_arg_t *argv)
1094 {
1095 mdb_spa_config_spa_t spa;
1096
1097 if (argc != 0 || !(flags & DCMD_ADDRSPEC))
1098 return (DCMD_USAGE);
1099
1100 if (mdb_ctf_vread(&spa, ZFS_STRUCT "spa", "mdb_spa_config_spa_t",
1101 addr, 0) == -1)
1102 return (DCMD_ERR);
1103
1104 if (spa.spa_config == NULL) {
1105 mdb_printf("(none)\n");
1106 return (DCMD_OK);
1107 }
1108
1109 return (mdb_call_dcmd("nvlist", (uintptr_t)spa.spa_config, flags,
1110 0, NULL));
1111 }
1112
1113 /*
1114 * ::vdev
1115 *
1116 * Print out a summarized vdev_t, in the following form:
1117 *
1118 * ADDR STATE AUX DESC
1119 * fffffffbcde23df0 HEALTHY - /dev/dsk/c0t0d0
1120 *
1121 * If '-r' is specified, recursively visit all children.
1122 *
1123 * With '-e', the statistics associated with the vdev are printed as well.
1124 */
1125 static int
1126 do_print_vdev(uintptr_t addr, int flags, int depth, int stats,
1127 int recursive)
1128 {
1129 vdev_t vdev;
1130 char desc[MAXNAMELEN];
1131 int c, children;
1132 uintptr_t *child;
1133 const char *state, *aux;
1134
1135 if (mdb_vread(&vdev, sizeof (vdev), (uintptr_t)addr) == -1) {
1136 mdb_warn("failed to read vdev_t at %p\n", (uintptr_t)addr);
1137 return (DCMD_ERR);
1138 }
1139
1140 if (flags & DCMD_PIPE_OUT) {
1141 mdb_printf("%#lr", addr);
1142 } else {
1143 if (vdev.vdev_path != NULL) {
1144 if (mdb_readstr(desc, sizeof (desc),
1145 (uintptr_t)vdev.vdev_path) == -1) {
1146 mdb_warn("failed to read vdev_path at %p\n",
1147 vdev.vdev_path);
1148 return (DCMD_ERR);
1149 }
1150 } else if (vdev.vdev_ops != NULL) {
1151 vdev_ops_t ops;
1152 if (mdb_vread(&ops, sizeof (ops),
1153 (uintptr_t)vdev.vdev_ops) == -1) {
1154 mdb_warn("failed to read vdev_ops at %p\n",
1155 vdev.vdev_ops);
1156 return (DCMD_ERR);
1157 }
1158 (void) strcpy(desc, ops.vdev_op_type);
1159 } else {
1160 (void) strcpy(desc, "<unknown>");
1161 }
1162
1163 if (depth == 0 && DCMD_HDRSPEC(flags))
1164 mdb_printf("%<u>%-?s %-9s %-12s %-*s%</u>\n",
1165 "ADDR", "STATE", "AUX",
1166 sizeof (uintptr_t) == 4 ? 43 : 35,
1167 "DESCRIPTION");
1168
1169 mdb_printf("%0?p ", addr);
1170
1171 switch (vdev.vdev_state) {
1172 case VDEV_STATE_CLOSED:
1173 state = "CLOSED";
1174 break;
1175 case VDEV_STATE_OFFLINE:
1176 state = "OFFLINE";
1177 break;
1178 case VDEV_STATE_CANT_OPEN:
1179 state = "CANT_OPEN";
1180 break;
1181 case VDEV_STATE_DEGRADED:
1182 state = "DEGRADED";
1183 break;
1184 case VDEV_STATE_HEALTHY:
1185 state = "HEALTHY";
1186 break;
1187 case VDEV_STATE_REMOVED:
1188 state = "REMOVED";
1189 break;
1190 case VDEV_STATE_FAULTED:
1191 state = "FAULTED";
1192 break;
1193 default:
1194 state = "UNKNOWN";
1195 break;
1196 }
1197
1198 switch (vdev.vdev_stat.vs_aux) {
1199 case VDEV_AUX_NONE:
1200 aux = "-";
1201 break;
1202 case VDEV_AUX_OPEN_FAILED:
1203 aux = "OPEN_FAILED";
1204 break;
1205 case VDEV_AUX_CORRUPT_DATA:
1206 aux = "CORRUPT_DATA";
1207 break;
1208 case VDEV_AUX_NO_REPLICAS:
1209 aux = "NO_REPLICAS";
1210 break;
1211 case VDEV_AUX_BAD_GUID_SUM:
1212 aux = "BAD_GUID_SUM";
1213 break;
1214 case VDEV_AUX_TOO_SMALL:
1215 aux = "TOO_SMALL";
1216 break;
1217 case VDEV_AUX_BAD_LABEL:
1218 aux = "BAD_LABEL";
1219 break;
1220 case VDEV_AUX_VERSION_NEWER:
1221 aux = "VERS_NEWER";
1222 break;
1223 case VDEV_AUX_VERSION_OLDER:
1224 aux = "VERS_OLDER";
1225 break;
1226 case VDEV_AUX_UNSUP_FEAT:
1227 aux = "UNSUP_FEAT";
1228 break;
1229 case VDEV_AUX_SPARED:
1230 aux = "SPARED";
1231 break;
1232 case VDEV_AUX_ERR_EXCEEDED:
1233 aux = "ERR_EXCEEDED";
1234 break;
1235 case VDEV_AUX_IO_FAILURE:
1236 aux = "IO_FAILURE";
1237 break;
1238 case VDEV_AUX_BAD_LOG:
1239 aux = "BAD_LOG";
1240 break;
1241 case VDEV_AUX_EXTERNAL:
1242 aux = "EXTERNAL";
1243 break;
1244 case VDEV_AUX_SPLIT_POOL:
1245 aux = "SPLIT_POOL";
1246 break;
1247 default:
1248 aux = "UNKNOWN";
1249 break;
1250 }
1251
1252 mdb_printf("%-9s %-12s %*s%s\n", state, aux, depth, "", desc);
1253
1254 if (stats) {
1255 vdev_stat_t *vs = &vdev.vdev_stat;
1256 int i;
1257
1258 mdb_inc_indent(4);
1259 mdb_printf("\n");
1260 mdb_printf("%<u> %12s %12s %12s %12s "
1261 "%12s%</u>\n", "READ", "WRITE", "FREE", "CLAIM",
1262 "IOCTL");
1263 mdb_printf("OPS ");
1264 for (i = 1; i < ZIO_TYPES; i++)
1265 mdb_printf("%11#llx%s", vs->vs_ops[i],
1266 i == ZIO_TYPES - 1 ? "" : " ");
1267 mdb_printf("\n");
1268 mdb_printf("BYTES ");
1269 for (i = 1; i < ZIO_TYPES; i++)
1270 mdb_printf("%11#llx%s", vs->vs_bytes[i],
1271 i == ZIO_TYPES - 1 ? "" : " ");
1272
1273
1274 mdb_printf("\n");
1275 mdb_printf("EREAD %10#llx\n", vs->vs_read_errors);
1276 mdb_printf("EWRITE %10#llx\n", vs->vs_write_errors);
1277 mdb_printf("ECKSUM %10#llx\n",
1278 vs->vs_checksum_errors);
1279 mdb_dec_indent(4);
1280 }
1281
1282 if (stats)
1283 mdb_printf("\n");
1284 }
1285
1286 children = vdev.vdev_children;
1287
1288 if (children == 0 || !recursive)
1289 return (DCMD_OK);
1290
1291 child = mdb_alloc(children * sizeof (void *), UM_SLEEP | UM_GC);
1292 if (mdb_vread(child, children * sizeof (void *),
1293 (uintptr_t)vdev.vdev_child) == -1) {
1294 mdb_warn("failed to read vdev children at %p", vdev.vdev_child);
1295 return (DCMD_ERR);
1296 }
1297
1298 for (c = 0; c < children; c++) {
1299 if (do_print_vdev(child[c], flags, depth + 2, stats,
1300 recursive))
1301 return (DCMD_ERR);
1302 }
1303
1304 return (DCMD_OK);
1305 }
1306
1307 static int
1308 vdev_print(uintptr_t addr, uint_t flags, int argc, const mdb_arg_t *argv)
1309 {
1310 int recursive = FALSE;
1311 int stats = FALSE;
1312 uint64_t depth = 0;
1313
1314 if (mdb_getopts(argc, argv,
1315 'r', MDB_OPT_SETBITS, TRUE, &recursive,
1316 'e', MDB_OPT_SETBITS, TRUE, &stats,
1317 'd', MDB_OPT_UINT64, &depth,
1318 NULL) != argc)
1319 return (DCMD_USAGE);
1320
1321 if (!(flags & DCMD_ADDRSPEC)) {
1322 mdb_warn("no vdev_t address given\n");
1323 return (DCMD_ERR);
1324 }
1325
1326 return (do_print_vdev(addr, flags, (int)depth, stats, recursive));
1327 }
1328
1329 typedef struct metaslab_walk_data {
1330 uint64_t mw_numvdevs;
1331 uintptr_t *mw_vdevs;
1332 int mw_curvdev;
1333 uint64_t mw_nummss;
1334 uintptr_t *mw_mss;
1335 int mw_curms;
1336 } metaslab_walk_data_t;
1337
1338 static int
1339 metaslab_walk_step(mdb_walk_state_t *wsp)
1340 {
1341 metaslab_walk_data_t *mw = wsp->walk_data;
1342 metaslab_t ms;
1343 uintptr_t msp;
1344
1345 if (mw->mw_curvdev >= mw->mw_numvdevs)
1346 return (WALK_DONE);
1347
1348 if (mw->mw_mss == NULL) {
1349 uintptr_t mssp;
1350 uintptr_t vdevp;
1351
1352 ASSERT(mw->mw_curms == 0);
1353 ASSERT(mw->mw_nummss == 0);
1354
1355 vdevp = mw->mw_vdevs[mw->mw_curvdev];
1356 if (GETMEMB(vdevp, "vdev", vdev_ms, mssp) ||
1357 GETMEMB(vdevp, "vdev", vdev_ms_count, mw->mw_nummss)) {
1358 return (WALK_ERR);
1359 }
1360
1361 mw->mw_mss = mdb_alloc(mw->mw_nummss * sizeof (void*),
1362 UM_SLEEP | UM_GC);
1363 if (mdb_vread(mw->mw_mss, mw->mw_nummss * sizeof (void*),
1364 mssp) == -1) {
1365 mdb_warn("failed to read vdev_ms at %p", mssp);
1366 return (WALK_ERR);
1367 }
1368 }
1369
1370 if (mw->mw_curms >= mw->mw_nummss) {
1371 mw->mw_mss = NULL;
1372 mw->mw_curms = 0;
1373 mw->mw_nummss = 0;
1374 mw->mw_curvdev++;
1375 return (WALK_NEXT);
1376 }
1377
1378 msp = mw->mw_mss[mw->mw_curms];
1379 if (mdb_vread(&ms, sizeof (metaslab_t), msp) == -1) {
1380 mdb_warn("failed to read metaslab_t at %p", msp);
1381 return (WALK_ERR);
1382 }
1383
1384 mw->mw_curms++;
1385
1386 return (wsp->walk_callback(msp, &ms, wsp->walk_cbdata));
1387 }
1388
1389 /* ARGSUSED */
1390 static int
1391 metaslab_walk_init(mdb_walk_state_t *wsp)
1392 {
1393 metaslab_walk_data_t *mw;
1394 uintptr_t root_vdevp;
1395 uintptr_t childp;
1396
1397 if (wsp->walk_addr == NULL) {
1398 mdb_warn("must supply address of spa_t\n");
1399 return (WALK_ERR);
1400 }
1401
1402 mw = mdb_zalloc(sizeof (metaslab_walk_data_t), UM_SLEEP | UM_GC);
1403
1404 if (GETMEMB(wsp->walk_addr, "spa", spa_root_vdev, root_vdevp) ||
1405 GETMEMB(root_vdevp, "vdev", vdev_children, mw->mw_numvdevs) ||
1406 GETMEMB(root_vdevp, "vdev", vdev_child, childp)) {
1407 return (DCMD_ERR);
1408 }
1409
1410 mw->mw_vdevs = mdb_alloc(mw->mw_numvdevs * sizeof (void *),
1411 UM_SLEEP | UM_GC);
1412 if (mdb_vread(mw->mw_vdevs, mw->mw_numvdevs * sizeof (void *),
1413 childp) == -1) {
1414 mdb_warn("failed to read root vdev children at %p", childp);
1415 return (DCMD_ERR);
1416 }
1417
1418 wsp->walk_data = mw;
1419
1420 return (WALK_NEXT);
1421 }
1422
1423 typedef struct mdb_spa {
1424 uintptr_t spa_dsl_pool;
1425 uintptr_t spa_root_vdev;
1426 } mdb_spa_t;
1427
1428 struct mdb_dsl_dir_phys;
1429 typedef struct mdb_dsl_dir_dbuf {
1430 uint8_t dddb_pad[offsetof(dmu_buf_t, db_data)];
1431 uintptr_t dddb_data;
1432 } mdb_dsl_dir_dbuf_t;
1433
1434 typedef struct mdb_dsl_dir {
1435 union {
1436 dmu_buf_t *dd_dmu_db;
1437 mdb_dsl_dir_dbuf_t *dd_db;
1438 } dd_db_u;
1439 int64_t dd_space_towrite[TXG_SIZE];
1440 } mdb_dsl_dir_t;
1441
1442 typedef struct mdb_dsl_dir_phys {
1443 uint64_t dd_used_bytes;
1444 uint64_t dd_compressed_bytes;
1445 uint64_t dd_uncompressed_bytes;
1446 } mdb_dsl_dir_phys_t;
1447
1448 typedef struct mdb_vdev {
1449 uintptr_t vdev_parent;
1450 uintptr_t vdev_ms;
1451 uint64_t vdev_ms_count;
1452 vdev_stat_t vdev_stat;
1453 } mdb_vdev_t;
1454
1455 typedef struct mdb_metaslab {
1456 space_map_t ms_allocmap[TXG_SIZE];
1457 space_map_t ms_freemap[TXG_SIZE];
1458 space_map_t ms_map;
1459 space_map_obj_t ms_smo;
1460 space_map_obj_t ms_smo_syncing;
1461 } mdb_metaslab_t;
1462
1463 typedef struct space_data {
1464 uint64_t ms_allocmap[TXG_SIZE];
1465 uint64_t ms_freemap[TXG_SIZE];
1466 uint64_t ms_map;
1467 uint64_t avail;
1468 uint64_t nowavail;
1469 } space_data_t;
1470
1471 /* ARGSUSED */
1472 static int
1473 space_cb(uintptr_t addr, const void *unknown, void *arg)
1474 {
1475 space_data_t *sd = arg;
1476 mdb_metaslab_t ms;
1477
1478 if (GETMEMB(addr, "metaslab", ms_allocmap, ms.ms_allocmap) ||
1479 GETMEMB(addr, "metaslab", ms_freemap, ms.ms_freemap) ||
1480 GETMEMB(addr, "metaslab", ms_map, ms.ms_map) ||
1481 GETMEMB(addr, "metaslab", ms_smo, ms.ms_smo) ||
1482 GETMEMB(addr, "metaslab", ms_smo_syncing, ms.ms_smo_syncing)) {
1483 return (WALK_ERR);
1484 }
1485
1486 sd->ms_allocmap[0] += ms.ms_allocmap[0].sm_space;
1487 sd->ms_allocmap[1] += ms.ms_allocmap[1].sm_space;
1488 sd->ms_allocmap[2] += ms.ms_allocmap[2].sm_space;
1489 sd->ms_allocmap[3] += ms.ms_allocmap[3].sm_space;
1490 sd->ms_freemap[0] += ms.ms_freemap[0].sm_space;
1491 sd->ms_freemap[1] += ms.ms_freemap[1].sm_space;
1492 sd->ms_freemap[2] += ms.ms_freemap[2].sm_space;
1493 sd->ms_freemap[3] += ms.ms_freemap[3].sm_space;
1494 sd->ms_map += ms.ms_map.sm_space;
1495 sd->avail += ms.ms_map.sm_size - ms.ms_smo.smo_alloc;
1496 sd->nowavail += ms.ms_map.sm_size - ms.ms_smo_syncing.smo_alloc;
1497
1498 return (WALK_NEXT);
1499 }
1500
1501 /*
1502 * ::spa_space [-b]
1503 *
1504 * Given a spa_t, print out it's on-disk space usage and in-core
1505 * estimates of future usage. If -b is given, print space in bytes.
1506 * Otherwise print in megabytes.
1507 */
1508 /* ARGSUSED */
1509 static int
1510 spa_space(uintptr_t addr, uint_t flags, int argc, const mdb_arg_t *argv)
1511 {
1512 mdb_spa_t spa;
1513 uintptr_t dp_root_dir;
1514 mdb_dsl_dir_t dd;
1515 mdb_dsl_dir_phys_t dsp;
1516 uint64_t children;
1517 uintptr_t childaddr;
1518 space_data_t sd;
1519 int shift = 20;
1520 char *suffix = "M";
1521 int bytes = B_FALSE;
1522
1523 if (mdb_getopts(argc, argv, 'b', MDB_OPT_SETBITS, TRUE, &bytes, NULL) !=
1524 argc)
1525 return (DCMD_USAGE);
1526 if (!(flags & DCMD_ADDRSPEC))
1527 return (DCMD_USAGE);
1528
1529 if (bytes) {
1530 shift = 0;
1531 suffix = "";
1532 }
1533
1534 if (GETMEMB(addr, "spa", spa_dsl_pool, spa.spa_dsl_pool) ||
1535 GETMEMB(addr, "spa", spa_root_vdev, spa.spa_root_vdev) ||
1536 GETMEMB(spa.spa_root_vdev, "vdev", vdev_children, children) ||
1537 GETMEMB(spa.spa_root_vdev, "vdev", vdev_child, childaddr) ||
1538 GETMEMB(spa.spa_dsl_pool, "dsl_pool",
1539 dp_root_dir, dp_root_dir) ||
1540 GETMEMB(dp_root_dir, "dsl_dir", dd_phys, dd.dd_phys) ||
1541 GETMEMB(dp_root_dir, "dsl_dir",
1542 dd_space_towrite, dd.dd_space_towrite) ||
1543 GETMEMB(dd.dd_phys, "dsl_dir_phys",
1544 dd_used_bytes, dsp.dd_used_bytes) ||
1545 GETMEMB(dd.dd_phys, "dsl_dir_phys",
1546 dd_compressed_bytes, dsp.dd_compressed_bytes) ||
1547 GETMEMB(dd.dd_phys, "dsl_dir_phys",
1548 dd_uncompressed_bytes, dsp.dd_uncompressed_bytes)) {
1549 return (DCMD_ERR);
1550 }
1551
1552 mdb_printf("dd_space_towrite = %llu%s %llu%s %llu%s %llu%s\n",
1553 dd.dd_space_towrite[0] >> shift, suffix,
1554 dd.dd_space_towrite[1] >> shift, suffix,
1555 dd.dd_space_towrite[2] >> shift, suffix,
1556 dd.dd_space_towrite[3] >> shift, suffix);
1557
1558 mdb_printf("dd_phys.dd_used_bytes = %llu%s\n",
1559 dsp.dd_used_bytes >> shift, suffix);
1560 mdb_printf("dd_phys.dd_compressed_bytes = %llu%s\n",
1561 dsp.dd_compressed_bytes >> shift, suffix);
1562 mdb_printf("dd_phys.dd_uncompressed_bytes = %llu%s\n",
1563 dsp.dd_uncompressed_bytes >> shift, suffix);
1564
1565 bzero(&sd, sizeof (sd));
1566 if (mdb_pwalk("metaslab", space_cb, &sd, addr) != 0) {
1567 mdb_warn("can't walk metaslabs");
1568 return (DCMD_ERR);
1569 }
1570
1571 mdb_printf("ms_allocmap = %llu%s %llu%s %llu%s %llu%s\n",
1572 sd.ms_allocmap[0] >> shift, suffix,
1573 sd.ms_allocmap[1] >> shift, suffix,
1574 sd.ms_allocmap[2] >> shift, suffix,
1575 sd.ms_allocmap[3] >> shift, suffix);
1576 mdb_printf("ms_freemap = %llu%s %llu%s %llu%s %llu%s\n",
1577 sd.ms_freemap[0] >> shift, suffix,
1578 sd.ms_freemap[1] >> shift, suffix,
1579 sd.ms_freemap[2] >> shift, suffix,
1580 sd.ms_freemap[3] >> shift, suffix);
1581 mdb_printf("ms_map = %llu%s\n", sd.ms_map >> shift, suffix);
1582 mdb_printf("last synced avail = %llu%s\n", sd.avail >> shift, suffix);
1583 mdb_printf("current syncing avail = %llu%s\n",
1584 sd.nowavail >> shift, suffix);
1585
1586 return (DCMD_OK);
1587 }
1588
1589 /*
1590 * ::spa_verify
1591 *
1592 * Given a spa_t, verify that that the pool is self-consistent.
1593 * Currently, it only checks to make sure that the vdev tree exists.
1594 */
1595 /* ARGSUSED */
1596 static int
1597 spa_verify(uintptr_t addr, uint_t flags, int argc, const mdb_arg_t *argv)
1598 {
1599 spa_t spa;
1600
1601 if (argc != 0 || !(flags & DCMD_ADDRSPEC))
1602 return (DCMD_USAGE);
1603
1604 if (mdb_vread(&spa, sizeof (spa), addr) == -1) {
1605 mdb_warn("failed to read spa_t at %p", addr);
1606 return (DCMD_ERR);
1607 }
1608
1609 if (spa.spa_root_vdev == NULL) {
1610 mdb_printf("no vdev tree present\n");
1611 return (DCMD_OK);
1612 }
1613
1614 return (DCMD_OK);
1615 }
1616
1617 static int
1618 spa_print_aux(spa_aux_vdev_t *sav, uint_t flags, mdb_arg_t *v,
1619 const char *name)
1620 {
1621 uintptr_t *aux;
1622 size_t len;
1623 int ret, i;
1624
1625 /*
1626 * Iterate over aux vdevs and print those out as well. This is a
1627 * little annoying because we don't have a root vdev to pass to ::vdev.
1628 * Instead, we print a single line and then call it for each child
1629 * vdev.
1630 */
1631 if (sav->sav_count != 0) {
1632 v[1].a_type = MDB_TYPE_STRING;
1633 v[1].a_un.a_str = "-d";
1634 v[2].a_type = MDB_TYPE_IMMEDIATE;
1635 v[2].a_un.a_val = 2;
1636
1637 len = sav->sav_count * sizeof (uintptr_t);
1638 aux = mdb_alloc(len, UM_SLEEP);
1639 if (mdb_vread(aux, len,
1640 (uintptr_t)sav->sav_vdevs) == -1) {
1641 mdb_free(aux, len);
1642 mdb_warn("failed to read l2cache vdevs at %p",
1643 sav->sav_vdevs);
1644 return (DCMD_ERR);
1645 }
1646
1647 mdb_printf("%-?s %-9s %-12s %s\n", "-", "-", "-", name);
1648
1649 for (i = 0; i < sav->sav_count; i++) {
1650 ret = mdb_call_dcmd("vdev", aux[i], flags, 3, v);
1651 if (ret != DCMD_OK) {
1652 mdb_free(aux, len);
1653 return (ret);
1654 }
1655 }
1656
1657 mdb_free(aux, len);
1658 }
1659
1660 return (0);
1661 }
1662
1663 /*
1664 * ::spa_vdevs
1665 *
1666 * -e Include error stats
1667 *
1668 * Print out a summarized list of vdevs for the given spa_t.
1669 * This is accomplished by invoking "::vdev -re" on the root vdev, as well as
1670 * iterating over the cache devices.
1671 */
1672 /* ARGSUSED */
1673 static int
1674 spa_vdevs(uintptr_t addr, uint_t flags, int argc, const mdb_arg_t *argv)
1675 {
1676 spa_t spa;
1677 mdb_arg_t v[3];
1678 int errors = FALSE;
1679 int ret;
1680
1681 if (mdb_getopts(argc, argv,
1682 'e', MDB_OPT_SETBITS, TRUE, &errors,
1683 NULL) != argc)
1684 return (DCMD_USAGE);
1685
1686 if (!(flags & DCMD_ADDRSPEC))
1687 return (DCMD_USAGE);
1688
1689 if (mdb_vread(&spa, sizeof (spa), addr) == -1) {
1690 mdb_warn("failed to read spa_t at %p", addr);
1691 return (DCMD_ERR);
1692 }
1693
1694 /*
1695 * Unitialized spa_t structures can have a NULL root vdev.
1696 */
1697 if (spa.spa_root_vdev == NULL) {
1698 mdb_printf("no associated vdevs\n");
1699 return (DCMD_OK);
1700 }
1701
1702 v[0].a_type = MDB_TYPE_STRING;
1703 v[0].a_un.a_str = errors ? "-re" : "-r";
1704
1705 ret = mdb_call_dcmd("vdev", (uintptr_t)spa.spa_root_vdev,
1706 flags, 1, v);
1707 if (ret != DCMD_OK)
1708 return (ret);
1709
1710 if (spa_print_aux(&spa.spa_l2cache, flags, v, "cache") != 0 ||
1711 spa_print_aux(&spa.spa_spares, flags, v, "spares") != 0)
1712 return (DCMD_ERR);
1713
1714 return (DCMD_OK);
1715 }
1716
1717 /*
1718 * ::zio
1719 *
1720 * Print a summary of zio_t and all its children. This is intended to display a
1721 * zio tree, and hence we only pick the most important pieces of information for
1722 * the main summary. More detailed information can always be found by doing a
1723 * '::print zio' on the underlying zio_t. The columns we display are:
1724 *
1725 * ADDRESS TYPE STAGE WAITER TIME_ELAPSED
1726 *
1727 * The 'address' column is indented by one space for each depth level as we
1728 * descend down the tree.
1729 */
1730
1731 #define ZIO_MAXINDENT 7
1732 #define ZIO_MAXWIDTH (sizeof (uintptr_t) * 2 + ZIO_MAXINDENT)
1733 #define ZIO_WALK_SELF 0
1734 #define ZIO_WALK_CHILD 1
1735 #define ZIO_WALK_PARENT 2
1736
1737 typedef struct zio_print_args {
1738 int zpa_current_depth;
1739 int zpa_min_depth;
1740 int zpa_max_depth;
1741 int zpa_type;
1742 uint_t zpa_flags;
1743 } zio_print_args_t;
1744
1745 typedef struct mdb_zio {
1746 enum zio_type io_type;
1747 enum zio_stage io_stage;
1748 void *io_waiter;
1749 void *io_spa;
1750 struct {
1751 struct {
1752 void *list_next;
1753 } list_head;
1754 } io_parent_list;
1755 int io_error;
1756 } mdb_zio_t;
1757
1758 typedef struct mdb_zio_timestamp {
1759 hrtime_t io_timestamp;
1760 } mdb_zio_timestamp_t;
1761
1762 static int zio_child_cb(uintptr_t addr, const void *unknown, void *arg);
1763
1764 static int
1765 zio_print_cb(uintptr_t addr, zio_print_args_t *zpa)
1766 {
1767 mdb_ctf_id_t type_enum, stage_enum;
1768 int indent = zpa->zpa_current_depth;
1769 const char *type, *stage;
1770 uintptr_t laddr;
1771 mdb_zio_t zio;
1772 mdb_zio_timestamp_t zio_timestamp = { 0 };
1773
1774 if (mdb_ctf_vread(&zio, ZFS_STRUCT "zio", "mdb_zio_t", addr, 0) == -1)
1775 return (WALK_ERR);
1776 (void) mdb_ctf_vread(&zio_timestamp, ZFS_STRUCT "zio",
1777 "mdb_zio_timestamp_t", addr, MDB_CTF_VREAD_QUIET);
1778
1779 if (indent > ZIO_MAXINDENT)
1780 indent = ZIO_MAXINDENT;
1781
1782 if (mdb_ctf_lookup_by_name("enum zio_type", &type_enum) == -1 ||
1783 mdb_ctf_lookup_by_name("enum zio_stage", &stage_enum) == -1) {
1784 mdb_warn("failed to lookup zio enums");
1785 return (WALK_ERR);
1786 }
1787
1788 if ((type = mdb_ctf_enum_name(type_enum, zio.io_type)) != NULL)
1789 type += sizeof ("ZIO_TYPE_") - 1;
1790 else
1791 type = "?";
1792
1793 if (zio.io_error == 0) {
1794 stage = mdb_ctf_enum_name(stage_enum, zio.io_stage);
1795 if (stage != NULL)
1796 stage += sizeof ("ZIO_STAGE_") - 1;
1797 else
1798 stage = "?";
1799 } else {
1800 stage = "FAILED";
1801 }
1802
1803 if (zpa->zpa_current_depth >= zpa->zpa_min_depth) {
1804 if (zpa->zpa_flags & DCMD_PIPE_OUT) {
1805 mdb_printf("%?p\n", addr);
1806 } else {
1807 mdb_printf("%*s%-*p %-5s %-16s ", indent, "",
1808 ZIO_MAXWIDTH - indent, addr, type, stage);
1809 if (zio.io_waiter)
1810 mdb_printf("%-16p ", zio.io_waiter);
1811 else
1812 mdb_printf("%-16s ", "-");
1813 #ifdef _KERNEL
1814 if (zio_timestamp.io_timestamp != 0) {
1815 mdb_printf("%llums", (mdb_gethrtime() -
1816 zio_timestamp.io_timestamp) /
1817 1000000);
1818 } else {
1819 mdb_printf("%-12s ", "-");
1820 }
1821 #else
1822 mdb_printf("%-12s ", "-");
1823 #endif
1824 mdb_printf("\n");
1825 }
1826 }
1827
1828 if (zpa->zpa_current_depth >= zpa->zpa_max_depth)
1829 return (WALK_NEXT);
1830
1831 if (zpa->zpa_type == ZIO_WALK_PARENT)
1832 laddr = addr + mdb_ctf_offsetof_by_name(ZFS_STRUCT "zio",
1833 "io_parent_list");
1834 else
1835 laddr = addr + mdb_ctf_offsetof_by_name(ZFS_STRUCT "zio",
1836 "io_child_list");
1837
1838 zpa->zpa_current_depth++;
1839 if (mdb_pwalk("list", zio_child_cb, zpa, laddr) != 0) {
1840 mdb_warn("failed to walk zio_t children at %p\n", laddr);
1841 return (WALK_ERR);
1842 }
1843 zpa->zpa_current_depth--;
1844
1845 return (WALK_NEXT);
1846 }
1847
1848 /* ARGSUSED */
1849 static int
1850 zio_child_cb(uintptr_t addr, const void *unknown, void *arg)
1851 {
1852 zio_link_t zl;
1853 uintptr_t ziop;
1854 zio_print_args_t *zpa = arg;
1855
1856 if (mdb_vread(&zl, sizeof (zl), addr) == -1) {
1857 mdb_warn("failed to read zio_link_t at %p", addr);
1858 return (WALK_ERR);
1859 }
1860
1861 if (zpa->zpa_type == ZIO_WALK_PARENT)
1862 ziop = (uintptr_t)zl.zl_parent;
1863 else
1864 ziop = (uintptr_t)zl.zl_child;
1865
1866 return (zio_print_cb(ziop, arg));
1867 }
1868
1869 /* ARGSUSED */
1870 static int
1871 zio_print(uintptr_t addr, uint_t flags, int argc, const mdb_arg_t *argv)
1872 {
1873 zio_print_args_t zpa = { 0 };
1874
1875 if (!(flags & DCMD_ADDRSPEC))
1876 return (DCMD_USAGE);
1877
1878 if (mdb_getopts(argc, argv,
1879 'r', MDB_OPT_SETBITS, INT_MAX, &zpa.zpa_max_depth,
1880 'c', MDB_OPT_SETBITS, ZIO_WALK_CHILD, &zpa.zpa_type,
1881 'p', MDB_OPT_SETBITS, ZIO_WALK_PARENT, &zpa.zpa_type,
1882 NULL) != argc)
1883 return (DCMD_USAGE);
1884
1885 zpa.zpa_flags = flags;
1886 if (zpa.zpa_max_depth != 0) {
1887 if (zpa.zpa_type == ZIO_WALK_SELF)
1888 zpa.zpa_type = ZIO_WALK_CHILD;
1889 } else if (zpa.zpa_type != ZIO_WALK_SELF) {
1890 zpa.zpa_min_depth = 1;
1891 zpa.zpa_max_depth = 1;
1892 }
1893
1894 if (!(flags & DCMD_PIPE_OUT) && DCMD_HDRSPEC(flags)) {
1895 mdb_printf("%<u>%-*s %-5s %-16s %-16s %-12s%</u>\n",
1896 ZIO_MAXWIDTH, "ADDRESS", "TYPE", "STAGE", "WAITER",
1897 "TIME_ELAPSED");
1898 }
1899
1900 if (zio_print_cb(addr, &zpa) != WALK_NEXT)
1901 return (DCMD_ERR);
1902
1903 return (DCMD_OK);
1904 }
1905
1906 /*
1907 * [addr]::zio_state
1908 *
1909 * Print a summary of all zio_t structures on the system, or for a particular
1910 * pool. This is equivalent to '::walk zio_root | ::zio'.
1911 */
1912 /*ARGSUSED*/
1913 static int
1914 zio_state(uintptr_t addr, uint_t flags, int argc, const mdb_arg_t *argv)
1915 {
1916 /*
1917 * MDB will remember the last address of the pipeline, so if we don't
1918 * zero this we'll end up trying to walk zio structures for a
1919 * non-existent spa_t.
1920 */
1921 if (!(flags & DCMD_ADDRSPEC))
1922 addr = 0;
1923
1924 return (mdb_pwalk_dcmd("zio_root", "zio", argc, argv, addr));
1925 }
1926
1927 typedef struct txg_list_walk_data {
1928 uintptr_t lw_head[TXG_SIZE];
1929 int lw_txgoff;
1930 int lw_maxoff;
1931 size_t lw_offset;
1932 void *lw_obj;
1933 } txg_list_walk_data_t;
1934
1935 static int
1936 txg_list_walk_init_common(mdb_walk_state_t *wsp, int txg, int maxoff)
1937 {
1938 txg_list_walk_data_t *lwd;
1939 txg_list_t list;
1940 int i;
1941
1942 lwd = mdb_alloc(sizeof (txg_list_walk_data_t), UM_SLEEP | UM_GC);
1943 if (mdb_vread(&list, sizeof (txg_list_t), wsp->walk_addr) == -1) {
1944 mdb_warn("failed to read txg_list_t at %#lx", wsp->walk_addr);
1945 return (WALK_ERR);
1946 }
1947
1948 for (i = 0; i < TXG_SIZE; i++)
1949 lwd->lw_head[i] = (uintptr_t)list.tl_head[i];
1950 lwd->lw_offset = list.tl_offset;
1951 lwd->lw_obj = mdb_alloc(lwd->lw_offset + sizeof (txg_node_t),
1952 UM_SLEEP | UM_GC);
1953 lwd->lw_txgoff = txg;
1954 lwd->lw_maxoff = maxoff;
1955
1956 wsp->walk_addr = lwd->lw_head[lwd->lw_txgoff];
1957 wsp->walk_data = lwd;
1958
1959 return (WALK_NEXT);
1960 }
1961
1962 static int
1963 txg_list_walk_init(mdb_walk_state_t *wsp)
1964 {
1965 return (txg_list_walk_init_common(wsp, 0, TXG_SIZE-1));
1966 }
1967
1968 static int
1969 txg_list0_walk_init(mdb_walk_state_t *wsp)
1970 {
1971 return (txg_list_walk_init_common(wsp, 0, 0));
1972 }
1973
1974 static int
1975 txg_list1_walk_init(mdb_walk_state_t *wsp)
1976 {
1977 return (txg_list_walk_init_common(wsp, 1, 1));
1978 }
1979
1980 static int
1981 txg_list2_walk_init(mdb_walk_state_t *wsp)
1982 {
1983 return (txg_list_walk_init_common(wsp, 2, 2));
1984 }
1985
1986 static int
1987 txg_list3_walk_init(mdb_walk_state_t *wsp)
1988 {
1989 return (txg_list_walk_init_common(wsp, 3, 3));
1990 }
1991
1992 static int
1993 txg_list_walk_step(mdb_walk_state_t *wsp)
1994 {
1995 txg_list_walk_data_t *lwd = wsp->walk_data;
1996 uintptr_t addr;
1997 txg_node_t *node;
1998 int status;
1999
2000 while (wsp->walk_addr == NULL && lwd->lw_txgoff < lwd->lw_maxoff) {
2001 lwd->lw_txgoff++;
2002 wsp->walk_addr = lwd->lw_head[lwd->lw_txgoff];
2003 }
2004
2005 if (wsp->walk_addr == NULL)
2006 return (WALK_DONE);
2007
2008 addr = wsp->walk_addr - lwd->lw_offset;
2009
2010 if (mdb_vread(lwd->lw_obj,
2011 lwd->lw_offset + sizeof (txg_node_t), addr) == -1) {
2012 mdb_warn("failed to read list element at %#lx", addr);
2013 return (WALK_ERR);
2014 }
2015
2016 status = wsp->walk_callback(addr, lwd->lw_obj, wsp->walk_cbdata);
2017 node = (txg_node_t *)((uintptr_t)lwd->lw_obj + lwd->lw_offset);
2018 wsp->walk_addr = (uintptr_t)node->tn_next[lwd->lw_txgoff];
2019
2020 return (status);
2021 }
2022
2023 /*
2024 * ::walk spa
2025 *
2026 * Walk all named spa_t structures in the namespace. This is nothing more than
2027 * a layered avl walk.
2028 */
2029 static int
2030 spa_walk_init(mdb_walk_state_t *wsp)
2031 {
2032 GElf_Sym sym;
2033
2034 if (wsp->walk_addr != NULL) {
2035 mdb_warn("spa walk only supports global walks\n");
2036 return (WALK_ERR);
2037 }
2038
2039 if (mdb_lookup_by_obj(ZFS_OBJ_NAME, "spa_namespace_avl", &sym) == -1) {
2040 mdb_warn("failed to find symbol 'spa_namespace_avl'");
2041 return (WALK_ERR);
2042 }
2043
2044 wsp->walk_addr = (uintptr_t)sym.st_value;
2045
2046 if (mdb_layered_walk("avl", wsp) == -1) {
2047 mdb_warn("failed to walk 'avl'\n");
2048 return (WALK_ERR);
2049 }
2050
2051 return (WALK_NEXT);
2052 }
2053
2054 static int
2055 spa_walk_step(mdb_walk_state_t *wsp)
2056 {
2057 spa_t spa;
2058
2059 if (mdb_vread(&spa, sizeof (spa), wsp->walk_addr) == -1) {
2060 mdb_warn("failed to read spa_t at %p", wsp->walk_addr);
2061 return (WALK_ERR);
2062 }
2063
2064 return (wsp->walk_callback(wsp->walk_addr, &spa, wsp->walk_cbdata));
2065 }
2066
2067 /*
2068 * [addr]::walk zio
2069 *
2070 * Walk all active zio_t structures on the system. This is simply a layered
2071 * walk on top of ::walk zio_cache, with the optional ability to limit the
2072 * structures to a particular pool.
2073 */
2074 static int
2075 zio_walk_init(mdb_walk_state_t *wsp)
2076 {
2077 wsp->walk_data = (void *)wsp->walk_addr;
2078
2079 if (mdb_layered_walk("zio_cache", wsp) == -1) {
2080 mdb_warn("failed to walk 'zio_cache'\n");
2081 return (WALK_ERR);
2082 }
2083
2084 return (WALK_NEXT);
2085 }
2086
2087 static int
2088 zio_walk_step(mdb_walk_state_t *wsp)
2089 {
2090 mdb_zio_t zio;
2091
2092 if (mdb_ctf_vread(&zio, ZFS_STRUCT "zio", "mdb_zio_t",
2093 wsp->walk_addr, 0) == -1)
2094 return (WALK_ERR);
2095
2096 if (wsp->walk_data != NULL && wsp->walk_data != zio.io_spa)
2097 return (WALK_NEXT);
2098
2099 return (wsp->walk_callback(wsp->walk_addr, &zio, wsp->walk_cbdata));
2100 }
2101
2102 /*
2103 * [addr]::walk zio_root
2104 *
2105 * Walk only root zio_t structures, optionally for a particular spa_t.
2106 */
2107 static int
2108 zio_walk_root_step(mdb_walk_state_t *wsp)
2109 {
2110 mdb_zio_t zio;
2111
2112 if (mdb_ctf_vread(&zio, ZFS_STRUCT "zio", "mdb_zio_t",
2113 wsp->walk_addr, 0) == -1)
2114 return (WALK_ERR);
2115
2116 if (wsp->walk_data != NULL && wsp->walk_data != zio.io_spa)
2117 return (WALK_NEXT);
2118
2119 /* If the parent list is not empty, ignore */
2120 if ((uintptr_t)zio.io_parent_list.list_head.list_next !=
2121 wsp->walk_addr +
2122 mdb_ctf_offsetof_by_name(ZFS_STRUCT "zio", "io_parent_list") +
2123 mdb_ctf_offsetof_by_name("struct list", "list_head"))
2124 return (WALK_NEXT);
2125
2126 return (wsp->walk_callback(wsp->walk_addr, &zio, wsp->walk_cbdata));
2127 }
2128
2129 #define NICENUM_BUFLEN 6
2130
2131 static int
2132 snprintfrac(char *buf, int len,
2133 uint64_t numerator, uint64_t denom, int frac_digits)
2134 {
2135 int mul = 1;
2136 int whole, frac, i;
2137
2138 for (i = frac_digits; i; i--)
2139 mul *= 10;
2140 whole = numerator / denom;
2141 frac = mul * numerator / denom - mul * whole;
2142 return (mdb_snprintf(buf, len, "%u.%0*u", whole, frac_digits, frac));
2143 }
2144
2145 static void
2146 mdb_nicenum(uint64_t num, char *buf)
2147 {
2148 uint64_t n = num;
2149 int index = 0;
2150 char *u;
2151
2152 while (n >= 1024) {
2153 n = (n + (1024 / 2)) / 1024; /* Round up or down */
2154 index++;
2155 }
2156
2157 u = &" \0K\0M\0G\0T\0P\0E\0"[index*2];
2158
2159 if (index == 0) {
2160 (void) mdb_snprintf(buf, NICENUM_BUFLEN, "%llu",
2161 (u_longlong_t)n);
2162 } else if (n < 10 && (num & (num - 1)) != 0) {
2163 (void) snprintfrac(buf, NICENUM_BUFLEN,
2164 num, 1ULL << 10 * index, 2);
2165 strcat(buf, u);
2166 } else if (n < 100 && (num & (num - 1)) != 0) {
2167 (void) snprintfrac(buf, NICENUM_BUFLEN,
2168 num, 1ULL << 10 * index, 1);
2169 strcat(buf, u);
2170 } else {
2171 (void) mdb_snprintf(buf, NICENUM_BUFLEN, "%llu%s",
2172 (u_longlong_t)n, u);
2173 }
2174 }
2175
2176 /*
2177 * ::zfs_blkstats
2178 *
2179 * -v print verbose per-level information
2180 *
2181 */
2182 static int
2183 zfs_blkstats(uintptr_t addr, uint_t flags, int argc, const mdb_arg_t *argv)
2184 {
2185 boolean_t verbose = B_FALSE;
2186 zfs_all_blkstats_t stats;
2187 dmu_object_type_t t;
2188 zfs_blkstat_t *tzb;
2189 uint64_t ditto;
2190 dmu_object_type_info_t dmu_ot[DMU_OT_NUMTYPES + 10];
2191 /* +10 in case it grew */
2192
2193 if (mdb_readvar(&dmu_ot, "dmu_ot") == -1) {
2194 mdb_warn("failed to read 'dmu_ot'");
2195 return (DCMD_ERR);
2196 }
2197
2198 if (mdb_getopts(argc, argv,
2199 'v', MDB_OPT_SETBITS, TRUE, &verbose,
2200 NULL) != argc)
2201 return (DCMD_USAGE);
2202
2203 if (!(flags & DCMD_ADDRSPEC))
2204 return (DCMD_USAGE);
2205
2206 if (GETMEMB(addr, "spa", spa_dsl_pool, addr) ||
2207 GETMEMB(addr, "dsl_pool", dp_blkstats, addr) ||
2208 mdb_vread(&stats, sizeof (zfs_all_blkstats_t), addr) == -1) {
2209 mdb_warn("failed to read data at %p;", addr);
2210 mdb_printf("maybe no stats? run \"zpool scrub\" first.");
2211 return (DCMD_ERR);
2212 }
2213
2214 tzb = &stats.zab_type[DN_MAX_LEVELS][DMU_OT_TOTAL];
2215 if (tzb->zb_gangs != 0) {
2216 mdb_printf("Ganged blocks: %llu\n",
2217 (longlong_t)tzb->zb_gangs);
2218 }
2219
2220 ditto = tzb->zb_ditto_2_of_2_samevdev + tzb->zb_ditto_2_of_3_samevdev +
2221 tzb->zb_ditto_3_of_3_samevdev;
2222 if (ditto != 0) {
2223 mdb_printf("Dittoed blocks on same vdev: %llu\n",
2224 (longlong_t)ditto);
2225 }
2226
2227 mdb_printf("\nBlocks\tLSIZE\tPSIZE\tASIZE"
2228 "\t avg\t comp\t%%Total\tType\n");
2229
2230 for (t = 0; t <= DMU_OT_TOTAL; t++) {
2231 char csize[NICENUM_BUFLEN], lsize[NICENUM_BUFLEN];
2232 char psize[NICENUM_BUFLEN], asize[NICENUM_BUFLEN];
2233 char avg[NICENUM_BUFLEN];
2234 char comp[NICENUM_BUFLEN], pct[NICENUM_BUFLEN];
2235 char typename[64];
2236 int l;
2237
2238
2239 if (t == DMU_OT_DEFERRED)
2240 strcpy(typename, "deferred free");
2241 else if (t == DMU_OT_OTHER)
2242 strcpy(typename, "other");
2243 else if (t == DMU_OT_TOTAL)
2244 strcpy(typename, "Total");
2245 else if (mdb_readstr(typename, sizeof (typename),
2246 (uintptr_t)dmu_ot[t].ot_name) == -1) {
2247 mdb_warn("failed to read type name");
2248 return (DCMD_ERR);
2249 }
2250
2251 if (stats.zab_type[DN_MAX_LEVELS][t].zb_asize == 0)
2252 continue;
2253
2254 for (l = -1; l < DN_MAX_LEVELS; l++) {
2255 int level = (l == -1 ? DN_MAX_LEVELS : l);
2256 zfs_blkstat_t *zb = &stats.zab_type[level][t];
2257
2258 if (zb->zb_asize == 0)
2259 continue;
2260
2261 /*
2262 * Don't print each level unless requested.
2263 */
2264 if (!verbose && level != DN_MAX_LEVELS)
2265 continue;
2266
2267 /*
2268 * If all the space is level 0, don't print the
2269 * level 0 separately.
2270 */
2271 if (level == 0 && zb->zb_asize ==
2272 stats.zab_type[DN_MAX_LEVELS][t].zb_asize)
2273 continue;
2274
2275 mdb_nicenum(zb->zb_count, csize);
2276 mdb_nicenum(zb->zb_lsize, lsize);
2277 mdb_nicenum(zb->zb_psize, psize);
2278 mdb_nicenum(zb->zb_asize, asize);
2279 mdb_nicenum(zb->zb_asize / zb->zb_count, avg);
2280 (void) snprintfrac(comp, NICENUM_BUFLEN,
2281 zb->zb_lsize, zb->zb_psize, 2);
2282 (void) snprintfrac(pct, NICENUM_BUFLEN,
2283 100 * zb->zb_asize, tzb->zb_asize, 2);
2284
2285 mdb_printf("%6s\t%5s\t%5s\t%5s\t%5s"
2286 "\t%5s\t%6s\t",
2287 csize, lsize, psize, asize, avg, comp, pct);
2288
2289 if (level == DN_MAX_LEVELS)
2290 mdb_printf("%s\n", typename);
2291 else
2292 mdb_printf(" L%d %s\n",
2293 level, typename);
2294 }
2295 }
2296
2297 return (DCMD_OK);
2298 }
2299
2300 /* ARGSUSED */
2301 static int
2302 reference_cb(uintptr_t addr, const void *ignored, void *arg)
2303 {
2304 static int gotid;
2305 static mdb_ctf_id_t ref_id;
2306 uintptr_t ref_holder;
2307 uintptr_t ref_removed;
2308 uint64_t ref_number;
2309 boolean_t holder_is_str = B_FALSE;
2310 char holder_str[128];
2311 boolean_t removed = (boolean_t)arg;
2312
2313 if (!gotid) {
2314 if (mdb_ctf_lookup_by_name("struct reference", &ref_id) == -1) {
2315 mdb_warn("couldn't find struct reference");
2316 return (WALK_ERR);
2317 }
2318 gotid = TRUE;
2319 }
2320
2321 if (GETMEMBID(addr, &ref_id, ref_holder, ref_holder) ||
2322 GETMEMBID(addr, &ref_id, ref_removed, ref_removed) ||
2323 GETMEMBID(addr, &ref_id, ref_number, ref_number))
2324 return (WALK_ERR);
2325
2326 if (mdb_readstr(holder_str, sizeof (holder_str), ref_holder) != -1)
2327 holder_is_str = strisprint(holder_str);
2328
2329 if (removed)
2330 mdb_printf("removed ");
2331 mdb_printf("reference ");
2332 if (ref_number != 1)
2333 mdb_printf("with count=%llu ", ref_number);
2334 mdb_printf("with tag %p", (void*)ref_holder);
2335 if (holder_is_str)
2336 mdb_printf(" \"%s\"", holder_str);
2337 mdb_printf(", held at:\n");
2338
2339 (void) mdb_call_dcmd("whatis", addr, DCMD_ADDRSPEC, 0, NULL);
2340
2341 if (removed) {
2342 mdb_printf("removed at:\n");
2343 (void) mdb_call_dcmd("whatis", ref_removed,
2344 DCMD_ADDRSPEC, 0, NULL);
2345 }
2346
2347 mdb_printf("\n");
2348
2349 return (WALK_NEXT);
2350 }
2351
2352 /* ARGSUSED */
2353 static int
2354 refcount(uintptr_t addr, uint_t flags, int argc, const mdb_arg_t *argv)
2355 {
2356 uint64_t rc_count, rc_removed_count;
2357 uintptr_t rc_list, rc_removed;
2358 static int gotid;
2359 static mdb_ctf_id_t rc_id;
2360 ulong_t off;
2361 boolean_t released = B_FALSE;
2362
2363 if (!(flags & DCMD_ADDRSPEC))
2364 return (DCMD_USAGE);
2365
2366 if (mdb_getopts(argc, argv,
2367 'r', MDB_OPT_SETBITS, B_TRUE, &released,
2368 NULL) != argc)
2369 return (DCMD_USAGE);
2370
2371 if (!gotid) {
2372 /*
2373 * The refcount structure is different when compiled debug
2374 * vs nondebug. Therefore, we want to make sure we get the
2375 * refcount definition from the ZFS module, in case it has
2376 * been compiled debug but genunix is nondebug.
2377 */
2378 if (mdb_ctf_lookup_by_name(ZFS_STRUCT "refcount",
2379 &rc_id) == -1) {
2380 mdb_warn("couldn't find struct refcount");
2381 return (DCMD_ERR);
2382 }
2383 gotid = TRUE;
2384 }
2385
2386 if (GETMEMBID(addr, &rc_id, rc_count, rc_count))
2387 return (DCMD_ERR);
2388
2389 if (GETMEMBID(addr, &rc_id, rc_removed_count, rc_removed_count)) {
2390 mdb_printf("refcount_t at %p has %llu current holds\n",
2391 addr, (longlong_t)rc_count);
2392 return (DCMD_OK);
2393 }
2394
2395 mdb_printf("refcount_t at %p has %llu current holds, "
2396 "%llu recently released holds\n",
2397 addr, (longlong_t)rc_count, (longlong_t)rc_removed_count);
2398
2399 if (rc_count > 0)
2400 mdb_printf("current holds:\n");
2401 if (mdb_ctf_offsetof(rc_id, "rc_list", &off) == -1)
2402 return (DCMD_ERR);
2403 rc_list = addr + off/NBBY;
2404 mdb_pwalk("list", reference_cb, (void*)B_FALSE, rc_list);
2405
2406 if (released) {
2407 if (rc_removed_count > 0)
2408 mdb_printf("released holds:\n");
2409 if (mdb_ctf_offsetof(rc_id, "rc_removed", &off) == -1)
2410 return (DCMD_ERR);
2411 rc_removed = addr + off/NBBY;
2412 mdb_pwalk("list", reference_cb, (void*)B_TRUE, rc_removed);
2413 }
2414
2415 return (DCMD_OK);
2416 }
2417
2418 /* ARGSUSED */
2419 static int
2420 sa_attr_table(uintptr_t addr, uint_t flags, int argc, const mdb_arg_t *argv)
2421 {
2422 sa_attr_table_t *table;
2423 sa_os_t sa_os;
2424 char *name;
2425 int i;
2426
2427 if (mdb_vread(&sa_os, sizeof (sa_os_t), addr) == -1) {
2428 mdb_warn("failed to read sa_os at %p", addr);
2429 return (DCMD_ERR);
2430 }
2431
2432 table = mdb_alloc(sizeof (sa_attr_table_t) * sa_os.sa_num_attrs,
2433 UM_SLEEP | UM_GC);
2434 name = mdb_alloc(MAXPATHLEN, UM_SLEEP | UM_GC);
2435
2436 if (mdb_vread(table, sizeof (sa_attr_table_t) * sa_os.sa_num_attrs,
2437 (uintptr_t)sa_os.sa_attr_table) == -1) {
2438 mdb_warn("failed to read sa_os at %p", addr);
2439 return (DCMD_ERR);
2440 }
2441
2442 mdb_printf("%<u>%-10s %-10s %-10s %-10s %s%</u>\n",
2443 "ATTR ID", "REGISTERED", "LENGTH", "BSWAP", "NAME");
2444 for (i = 0; i != sa_os.sa_num_attrs; i++) {
2445 mdb_readstr(name, MAXPATHLEN, (uintptr_t)table[i].sa_name);
2446 mdb_printf("%5x %8x %8x %8x %-s\n",
2447 (int)table[i].sa_attr, (int)table[i].sa_registered,
2448 (int)table[i].sa_length, table[i].sa_byteswap, name);
2449 }
2450
2451 return (DCMD_OK);
2452 }
2453
2454 static int
2455 sa_get_off_table(uintptr_t addr, uint32_t **off_tab, int attr_count)
2456 {
2457 uintptr_t idx_table;
2458
2459 if (GETMEMB(addr, "sa_idx_tab", sa_idx_tab, idx_table)) {
2460 mdb_printf("can't find offset table in sa_idx_tab\n");
2461 return (-1);
2462 }
2463
2464 *off_tab = mdb_alloc(attr_count * sizeof (uint32_t),
2465 UM_SLEEP | UM_GC);
2466
2467 if (mdb_vread(*off_tab,
2468 attr_count * sizeof (uint32_t), idx_table) == -1) {
2469 mdb_warn("failed to attribute offset table %p", idx_table);
2470 return (-1);
2471 }
2472
2473 return (DCMD_OK);
2474 }
2475
2476 /*ARGSUSED*/
2477 static int
2478 sa_attr_print(uintptr_t addr, uint_t flags, int argc, const mdb_arg_t *argv)
2479 {
2480 uint32_t *offset_tab;
2481 int attr_count;
2482 uint64_t attr_id;
2483 uintptr_t attr_addr;
2484 uintptr_t bonus_tab, spill_tab;
2485 uintptr_t db_bonus, db_spill;
2486 uintptr_t os, os_sa;
2487 uintptr_t db_data;
2488
2489 if (argc != 1)
2490 return (DCMD_USAGE);
2491
2492 if (argv[0].a_type == MDB_TYPE_STRING)
2493 attr_id = mdb_strtoull(argv[0].a_un.a_str);
2494 else
2495 return (DCMD_USAGE);
2496
2497 if (GETMEMB(addr, "sa_handle", sa_bonus_tab, bonus_tab) ||
2498 GETMEMB(addr, "sa_handle", sa_spill_tab, spill_tab) ||
2499 GETMEMB(addr, "sa_handle", sa_os, os) ||
2500 GETMEMB(addr, "sa_handle", sa_bonus, db_bonus) ||
2501 GETMEMB(addr, "sa_handle", sa_spill, db_spill)) {
2502 mdb_printf("Can't find necessary information in sa_handle "
2503 "in sa_handle\n");
2504 return (DCMD_ERR);
2505 }
2506
2507 if (GETMEMB(os, "objset", os_sa, os_sa)) {
2508 mdb_printf("Can't find os_sa in objset\n");
2509 return (DCMD_ERR);
2510 }
2511
2512 if (GETMEMB(os_sa, "sa_os", sa_num_attrs, attr_count)) {
2513 mdb_printf("Can't find sa_num_attrs\n");
2514 return (DCMD_ERR);
2515 }
2516
2517 if (attr_id > attr_count) {
2518 mdb_printf("attribute id number is out of range\n");
2519 return (DCMD_ERR);
2520 }
2521
2522 if (bonus_tab) {
2523 if (sa_get_off_table(bonus_tab, &offset_tab,
2524 attr_count) == -1) {
2525 return (DCMD_ERR);
2526 }
2527
2528 if (GETMEMB(db_bonus, "dmu_buf", db_data, db_data)) {
2529 mdb_printf("can't find db_data in bonus dbuf\n");
2530 return (DCMD_ERR);
2531 }
2532 }
2533
2534 if (bonus_tab && !TOC_ATTR_PRESENT(offset_tab[attr_id]) &&
2535 spill_tab == NULL) {
2536 mdb_printf("Attribute does not exist\n");
2537 return (DCMD_ERR);
2538 } else if (!TOC_ATTR_PRESENT(offset_tab[attr_id]) && spill_tab) {
2539 if (sa_get_off_table(spill_tab, &offset_tab,
2540 attr_count) == -1) {
2541 return (DCMD_ERR);
2542 }
2543 if (GETMEMB(db_spill, "dmu_buf", db_data, db_data)) {
2544 mdb_printf("can't find db_data in spill dbuf\n");
2545 return (DCMD_ERR);
2546 }
2547 if (!TOC_ATTR_PRESENT(offset_tab[attr_id])) {
2548 mdb_printf("Attribute does not exist\n");
2549 return (DCMD_ERR);
2550 }
2551 }
2552 attr_addr = db_data + TOC_OFF(offset_tab[attr_id]);
2553 mdb_printf("%p\n", attr_addr);
2554 return (DCMD_OK);
2555 }
2556
2557 /* ARGSUSED */
2558 static int
2559 zfs_ace_print_common(uintptr_t addr, uint_t flags,
2560 uint64_t id, uint32_t access_mask, uint16_t ace_flags,
2561 uint16_t ace_type, int verbose)
2562 {
2563 if (DCMD_HDRSPEC(flags) && !verbose)
2564 mdb_printf("%<u>%-?s %-8s %-8s %-8s %s%</u>\n",
2565 "ADDR", "FLAGS", "MASK", "TYPE", "ID");
2566
2567 if (!verbose) {
2568 mdb_printf("%0?p %-8x %-8x %-8x %-llx\n", addr,
2569 ace_flags, access_mask, ace_type, id);
2570 return (DCMD_OK);
2571 }
2572
2573 switch (ace_flags & ACE_TYPE_FLAGS) {
2574 case ACE_OWNER:
2575 mdb_printf("owner@:");
2576 break;
2577 case (ACE_IDENTIFIER_GROUP | ACE_GROUP):
2578 mdb_printf("group@:");
2579 break;
2580 case ACE_EVERYONE:
2581 mdb_printf("everyone@:");
2582 break;
2583 case ACE_IDENTIFIER_GROUP:
2584 mdb_printf("group:%llx:", (u_longlong_t)id);
2585 break;
2586 case 0: /* User entry */
2587 mdb_printf("user:%llx:", (u_longlong_t)id);
2588 break;
2589 }
2590
2591 /* print out permission mask */
2592 if (access_mask & ACE_READ_DATA)
2593 mdb_printf("r");
2594 else
2595 mdb_printf("-");
2596 if (access_mask & ACE_WRITE_DATA)
2597 mdb_printf("w");
2598 else
2599 mdb_printf("-");
2600 if (access_mask & ACE_EXECUTE)
2601 mdb_printf("x");
2602 else
2603 mdb_printf("-");
2604 if (access_mask & ACE_APPEND_DATA)
2605 mdb_printf("p");
2606 else
2607 mdb_printf("-");
2608 if (access_mask & ACE_DELETE)
2609 mdb_printf("d");
2610 else
2611 mdb_printf("-");
2612 if (access_mask & ACE_DELETE_CHILD)
2613 mdb_printf("D");
2614 else
2615 mdb_printf("-");
2616 if (access_mask & ACE_READ_ATTRIBUTES)
2617 mdb_printf("a");
2618 else
2619 mdb_printf("-");
2620 if (access_mask & ACE_WRITE_ATTRIBUTES)
2621 mdb_printf("A");
2622 else
2623 mdb_printf("-");
2624 if (access_mask & ACE_READ_NAMED_ATTRS)
2625 mdb_printf("R");
2626 else
2627 mdb_printf("-");
2628 if (access_mask & ACE_WRITE_NAMED_ATTRS)
2629 mdb_printf("W");
2630 else
2631 mdb_printf("-");
2632 if (access_mask & ACE_READ_ACL)
2633 mdb_printf("c");
2634 else
2635 mdb_printf("-");
2636 if (access_mask & ACE_WRITE_ACL)
2637 mdb_printf("C");
2638 else
2639 mdb_printf("-");
2640 if (access_mask & ACE_WRITE_OWNER)
2641 mdb_printf("o");
2642 else
2643 mdb_printf("-");
2644 if (access_mask & ACE_SYNCHRONIZE)
2645 mdb_printf("s");
2646 else
2647 mdb_printf("-");
2648
2649 mdb_printf(":");
2650
2651 /* Print out inheritance flags */
2652 if (ace_flags & ACE_FILE_INHERIT_ACE)
2653 mdb_printf("f");
2654 else
2655 mdb_printf("-");
2656 if (ace_flags & ACE_DIRECTORY_INHERIT_ACE)
2657 mdb_printf("d");
2658 else
2659 mdb_printf("-");
2660 if (ace_flags & ACE_INHERIT_ONLY_ACE)
2661 mdb_printf("i");
2662 else
2663 mdb_printf("-");
2664 if (ace_flags & ACE_NO_PROPAGATE_INHERIT_ACE)
2665 mdb_printf("n");
2666 else
2667 mdb_printf("-");
2668 if (ace_flags & ACE_SUCCESSFUL_ACCESS_ACE_FLAG)
2669 mdb_printf("S");
2670 else
2671 mdb_printf("-");
2672 if (ace_flags & ACE_FAILED_ACCESS_ACE_FLAG)
2673 mdb_printf("F");
2674 else
2675 mdb_printf("-");
2676 if (ace_flags & ACE_INHERITED_ACE)
2677 mdb_printf("I");
2678 else
2679 mdb_printf("-");
2680
2681 switch (ace_type) {
2682 case ACE_ACCESS_ALLOWED_ACE_TYPE:
2683 mdb_printf(":allow\n");
2684 break;
2685 case ACE_ACCESS_DENIED_ACE_TYPE:
2686 mdb_printf(":deny\n");
2687 break;
2688 case ACE_SYSTEM_AUDIT_ACE_TYPE:
2689 mdb_printf(":audit\n");
2690 break;
2691 case ACE_SYSTEM_ALARM_ACE_TYPE:
2692 mdb_printf(":alarm\n");
2693 break;
2694 default:
2695 mdb_printf(":?\n");
2696 }
2697 return (DCMD_OK);
2698 }
2699
2700 /* ARGSUSED */
2701 static int
2702 zfs_ace_print(uintptr_t addr, uint_t flags, int argc, const mdb_arg_t *argv)
2703 {
2704 zfs_ace_t zace;
2705 int verbose = FALSE;
2706 uint64_t id;
2707
2708 if (!(flags & DCMD_ADDRSPEC))
2709 return (DCMD_USAGE);
2710
2711 if (mdb_getopts(argc, argv,
2712 'v', MDB_OPT_SETBITS, TRUE, &verbose, TRUE, NULL) != argc)
2713 return (DCMD_USAGE);
2714
2715 if (mdb_vread(&zace, sizeof (zfs_ace_t), addr) == -1) {
2716 mdb_warn("failed to read zfs_ace_t");
2717 return (DCMD_ERR);
2718 }
2719
2720 if ((zace.z_hdr.z_flags & ACE_TYPE_FLAGS) == 0 ||
2721 (zace.z_hdr.z_flags & ACE_TYPE_FLAGS) == ACE_IDENTIFIER_GROUP)
2722 id = zace.z_fuid;
2723 else
2724 id = -1;
2725
2726 return (zfs_ace_print_common(addr, flags, id, zace.z_hdr.z_access_mask,
2727 zace.z_hdr.z_flags, zace.z_hdr.z_type, verbose));
2728 }
2729
2730 /* ARGSUSED */
2731 static int
2732 zfs_ace0_print(uintptr_t addr, uint_t flags, int argc, const mdb_arg_t *argv)
2733 {
2734 ace_t ace;
2735 uint64_t id;
2736 int verbose = FALSE;
2737
2738 if (!(flags & DCMD_ADDRSPEC))
2739 return (DCMD_USAGE);
2740
2741 if (mdb_getopts(argc, argv,
2742 'v', MDB_OPT_SETBITS, TRUE, &verbose, TRUE, NULL) != argc)
2743 return (DCMD_USAGE);
2744
2745 if (mdb_vread(&ace, sizeof (ace_t), addr) == -1) {
2746 mdb_warn("failed to read ace_t");
2747 return (DCMD_ERR);
2748 }
2749
2750 if ((ace.a_flags & ACE_TYPE_FLAGS) == 0 ||
2751 (ace.a_flags & ACE_TYPE_FLAGS) == ACE_IDENTIFIER_GROUP)
2752 id = ace.a_who;
2753 else
2754 id = -1;
2755
2756 return (zfs_ace_print_common(addr, flags, id, ace.a_access_mask,
2757 ace.a_flags, ace.a_type, verbose));
2758 }
2759
2760 typedef struct acl_dump_args {
2761 int a_argc;
2762 const mdb_arg_t *a_argv;
2763 uint16_t a_version;
2764 int a_flags;
2765 } acl_dump_args_t;
2766
2767 /* ARGSUSED */
2768 static int
2769 acl_aces_cb(uintptr_t addr, const void *unknown, void *arg)
2770 {
2771 acl_dump_args_t *acl_args = (acl_dump_args_t *)arg;
2772
2773 if (acl_args->a_version == 1) {
2774 if (mdb_call_dcmd("zfs_ace", addr,
2775 DCMD_ADDRSPEC|acl_args->a_flags, acl_args->a_argc,
2776 acl_args->a_argv) != DCMD_OK) {
2777 return (WALK_ERR);
2778 }
2779 } else {
2780 if (mdb_call_dcmd("zfs_ace0", addr,
2781 DCMD_ADDRSPEC|acl_args->a_flags, acl_args->a_argc,
2782 acl_args->a_argv) != DCMD_OK) {
2783 return (WALK_ERR);
2784 }
2785 }
2786 acl_args->a_flags = DCMD_LOOP;
2787 return (WALK_NEXT);
2788 }
2789
2790 /* ARGSUSED */
2791 static int
2792 acl_cb(uintptr_t addr, const void *unknown, void *arg)
2793 {
2794 acl_dump_args_t *acl_args = (acl_dump_args_t *)arg;
2795
2796 if (acl_args->a_version == 1) {
2797 if (mdb_pwalk("zfs_acl_node_aces", acl_aces_cb,
2798 arg, addr) != 0) {
2799 mdb_warn("can't walk ACEs");
2800 return (DCMD_ERR);
2801 }
2802 } else {
2803 if (mdb_pwalk("zfs_acl_node_aces0", acl_aces_cb,
2804 arg, addr) != 0) {
2805 mdb_warn("can't walk ACEs");
2806 return (DCMD_ERR);
2807 }
2808 }
2809 return (WALK_NEXT);
2810 }
2811
2812 /* ARGSUSED */
2813 static int
2814 zfs_acl_dump(uintptr_t addr, uint_t flags, int argc, const mdb_arg_t *argv)
2815 {
2816 zfs_acl_t zacl;
2817 int verbose = FALSE;
2818 acl_dump_args_t acl_args;
2819
2820 if (!(flags & DCMD_ADDRSPEC))
2821 return (DCMD_USAGE);
2822
2823 if (mdb_getopts(argc, argv,
2824 'v', MDB_OPT_SETBITS, TRUE, &verbose, TRUE, NULL) != argc)
2825 return (DCMD_USAGE);
2826
2827 if (mdb_vread(&zacl, sizeof (zfs_acl_t), addr) == -1) {
2828 mdb_warn("failed to read zfs_acl_t");
2829 return (DCMD_ERR);
2830 }
2831
2832 acl_args.a_argc = argc;
2833 acl_args.a_argv = argv;
2834 acl_args.a_version = zacl.z_version;
2835 acl_args.a_flags = DCMD_LOOPFIRST;
2836
2837 if (mdb_pwalk("zfs_acl_node", acl_cb, &acl_args, addr) != 0) {
2838 mdb_warn("can't walk ACL");
2839 return (DCMD_ERR);
2840 }
2841
2842 return (DCMD_OK);
2843 }
2844
2845 /* ARGSUSED */
2846 static int
2847 zfs_acl_node_walk_init(mdb_walk_state_t *wsp)
2848 {
2849 if (wsp->walk_addr == NULL) {
2850 mdb_warn("must supply address of zfs_acl_node_t\n");
2851 return (WALK_ERR);
2852 }
2853
2854 wsp->walk_addr +=
2855 mdb_ctf_offsetof_by_name(ZFS_STRUCT "zfs_acl", "z_acl");
2856
2857 if (mdb_layered_walk("list", wsp) == -1) {
2858 mdb_warn("failed to walk 'list'\n");
2859 return (WALK_ERR);
2860 }
2861
2862 return (WALK_NEXT);
2863 }
2864
2865 static int
2866 zfs_acl_node_walk_step(mdb_walk_state_t *wsp)
2867 {
2868 zfs_acl_node_t aclnode;
2869
2870 if (mdb_vread(&aclnode, sizeof (zfs_acl_node_t),
2871 wsp->walk_addr) == -1) {
2872 mdb_warn("failed to read zfs_acl_node at %p", wsp->walk_addr);
2873 return (WALK_ERR);
2874 }
2875
2876 return (wsp->walk_callback(wsp->walk_addr, &aclnode, wsp->walk_cbdata));
2877 }
2878
2879 typedef struct ace_walk_data {
2880 int ace_count;
2881 int ace_version;
2882 } ace_walk_data_t;
2883
2884 static int
2885 zfs_aces_walk_init_common(mdb_walk_state_t *wsp, int version,
2886 int ace_count, uintptr_t ace_data)
2887 {
2888 ace_walk_data_t *ace_walk_data;
2889
2890 if (wsp->walk_addr == NULL) {
2891 mdb_warn("must supply address of zfs_acl_node_t\n");
2892 return (WALK_ERR);
2893 }
2894
2895 ace_walk_data = mdb_alloc(sizeof (ace_walk_data_t), UM_SLEEP | UM_GC);
2896
2897 ace_walk_data->ace_count = ace_count;
2898 ace_walk_data->ace_version = version;
2899
2900 wsp->walk_addr = ace_data;
2901 wsp->walk_data = ace_walk_data;
2902
2903 return (WALK_NEXT);
2904 }
2905
2906 static int
2907 zfs_acl_node_aces_walk_init_common(mdb_walk_state_t *wsp, int version)
2908 {
2909 static int gotid;
2910 static mdb_ctf_id_t acl_id;
2911 int z_ace_count;
2912 uintptr_t z_acldata;
2913
2914 if (!gotid) {
2915 if (mdb_ctf_lookup_by_name("struct zfs_acl_node",
2916 &acl_id) == -1) {
2917 mdb_warn("couldn't find struct zfs_acl_node");
2918 return (DCMD_ERR);
2919 }
2920 gotid = TRUE;
2921 }
2922
2923 if (GETMEMBID(wsp->walk_addr, &acl_id, z_ace_count, z_ace_count)) {
2924 return (DCMD_ERR);
2925 }
2926 if (GETMEMBID(wsp->walk_addr, &acl_id, z_acldata, z_acldata)) {
2927 return (DCMD_ERR);
2928 }
2929
2930 return (zfs_aces_walk_init_common(wsp, version,
2931 z_ace_count, z_acldata));
2932 }
2933
2934 /* ARGSUSED */
2935 static int
2936 zfs_acl_node_aces_walk_init(mdb_walk_state_t *wsp)
2937 {
2938 return (zfs_acl_node_aces_walk_init_common(wsp, 1));
2939 }
2940
2941 /* ARGSUSED */
2942 static int
2943 zfs_acl_node_aces0_walk_init(mdb_walk_state_t *wsp)
2944 {
2945 return (zfs_acl_node_aces_walk_init_common(wsp, 0));
2946 }
2947
2948 static int
2949 zfs_aces_walk_step(mdb_walk_state_t *wsp)
2950 {
2951 ace_walk_data_t *ace_data = wsp->walk_data;
2952 zfs_ace_t zace;
2953 ace_t *acep;
2954 int status;
2955 int entry_type;
2956 int allow_type;
2957 uintptr_t ptr;
2958
2959 if (ace_data->ace_count == 0)
2960 return (WALK_DONE);
2961
2962 if (mdb_vread(&zace, sizeof (zfs_ace_t), wsp->walk_addr) == -1) {
2963 mdb_warn("failed to read zfs_ace_t at %#lx",
2964 wsp->walk_addr);
2965 return (WALK_ERR);
2966 }
2967
2968 switch (ace_data->ace_version) {
2969 case 0:
2970 acep = (ace_t *)&zace;
2971 entry_type = acep->a_flags & ACE_TYPE_FLAGS;
2972 allow_type = acep->a_type;
2973 break;
2974 case 1:
2975 entry_type = zace.z_hdr.z_flags & ACE_TYPE_FLAGS;
2976 allow_type = zace.z_hdr.z_type;
2977 break;
2978 default:
2979 return (WALK_ERR);
2980 }
2981
2982 ptr = (uintptr_t)wsp->walk_addr;
2983 switch (entry_type) {
2984 case ACE_OWNER:
2985 case ACE_EVERYONE:
2986 case (ACE_IDENTIFIER_GROUP | ACE_GROUP):
2987 ptr += ace_data->ace_version == 0 ?
2988 sizeof (ace_t) : sizeof (zfs_ace_hdr_t);
2989 break;
2990 case ACE_IDENTIFIER_GROUP:
2991 default:
2992 switch (allow_type) {
2993 case ACE_ACCESS_ALLOWED_OBJECT_ACE_TYPE:
2994 case ACE_ACCESS_DENIED_OBJECT_ACE_TYPE:
2995 case ACE_SYSTEM_AUDIT_OBJECT_ACE_TYPE:
2996 case ACE_SYSTEM_ALARM_OBJECT_ACE_TYPE:
2997 ptr += ace_data->ace_version == 0 ?
2998 sizeof (ace_t) : sizeof (zfs_object_ace_t);
2999 break;
3000 default:
3001 ptr += ace_data->ace_version == 0 ?
3002 sizeof (ace_t) : sizeof (zfs_ace_t);
3003 break;
3004 }
3005 }
3006
3007 ace_data->ace_count--;
3008 status = wsp->walk_callback(wsp->walk_addr,
3009 (void *)(uintptr_t)&zace, wsp->walk_cbdata);
3010
3011 wsp->walk_addr = ptr;
3012 return (status);
3013 }
3014
3015 typedef struct mdb_zfs_rrwlock {
3016 kthread_t *rr_writer;
3017 boolean_t rr_writer_wanted;
3018 } mdb_zfs_rrwlock_t;
3019
3020 static uint_t rrw_key;
3021
3022 /* ARGSUSED */
3023 static int
3024 rrwlock(uintptr_t addr, uint_t flags, int argc, const mdb_arg_t *argv)
3025 {
3026 mdb_zfs_rrwlock_t rrw;
3027
3028 if (rrw_key == 0) {
3029 if (mdb_ctf_readsym(&rrw_key, "uint_t", "rrw_tsd_key", 0) == -1)
3030 return (DCMD_ERR);
3031 }
3032
3033 if (mdb_ctf_vread(&rrw, "rrwlock_t", "mdb_zfs_rrwlock_t", addr,
3034 0) == -1)
3035 return (DCMD_ERR);
3036
3037 if (rrw.rr_writer != NULL) {
3038 mdb_printf("write lock held by thread %p\n", rrw.rr_writer);
3039 return (DCMD_OK);
3040 }
3041
3042 if (rrw.rr_writer_wanted) {
3043 mdb_printf("writer wanted\n");
3044 }
3045
3046 mdb_printf("anonymous references:\n");
3047 (void) mdb_call_dcmd("refcount", addr +
3048 mdb_ctf_offsetof_by_name(ZFS_STRUCT "rrwlock", "rr_anon_rcount"),
3049 DCMD_ADDRSPEC, 0, NULL);
3050
3051 mdb_printf("linked references:\n");
3052 (void) mdb_call_dcmd("refcount", addr +
3053 mdb_ctf_offsetof_by_name(ZFS_STRUCT "rrwlock", "rr_linked_rcount"),
3054 DCMD_ADDRSPEC, 0, NULL);
3055
3056 /*
3057 * XXX This should find references from
3058 * "::walk thread | ::tsd -v <rrw_key>", but there is no support
3059 * for programmatic consumption of dcmds, so this would be
3060 * difficult, potentially requiring reimplementing ::tsd (both
3061 * user and kernel versions) in this MDB module.
3062 */
3063
3064 return (DCMD_OK);
3065 }
3066
3067 /*
3068 * MDB module linkage information:
3069 *
3070 * We declare a list of structures describing our dcmds, and a function
3071 * named _mdb_init to return a pointer to our module information.
3072 */
3073
3074 static const mdb_dcmd_t dcmds[] = {
3075 { "arc", "[-bkmg]", "print ARC variables", arc_print },
3076 { "blkptr", ":", "print blkptr_t", blkptr },
3077 { "dbuf", ":", "print dmu_buf_impl_t", dbuf },
3078 { "dbuf_stats", ":", "dbuf stats", dbuf_stats },
3079 { "dbufs",
3080 "\t[-O objset_t*] [-n objset_name | \"mos\"] "
3081 "[-o object | \"mdn\"] \n"
3082 "\t[-l level] [-b blkid | \"bonus\"]",
3083 "find dmu_buf_impl_t's that match specified criteria", dbufs },
3084 { "abuf_find", "dva_word[0] dva_word[1]",
3085 "find arc_buf_hdr_t of a specified DVA",
3086 abuf_find },
3087 { "spa", "?[-cv]", "spa_t summary", spa_print },
3088 { "spa_config", ":", "print spa_t configuration", spa_print_config },
3089 { "spa_verify", ":", "verify spa_t consistency", spa_verify },
3090 { "spa_space", ":[-b]", "print spa_t on-disk space usage", spa_space },
3091 { "spa_vdevs", ":", "given a spa_t, print vdev summary", spa_vdevs },
3092 { "vdev", ":[-re]\n"
3093 "\t-r display recursively\n"
3094 "\t-e print statistics",
3095 "vdev_t summary", vdev_print },
3096 { "zio", ":[cpr]\n"
3097 "\t-c display children\n"
3098 "\t-p display parents\n"
3099 "\t-r display recursively",
3100 "zio_t summary", zio_print },
3101 { "zio_state", "?", "print out all zio_t structures on system or "
3102 "for a particular pool", zio_state },
3103 { "zfs_blkstats", ":[-v]",
3104 "given a spa_t, print block type stats from last scrub",
3105 zfs_blkstats },
3106 { "zfs_params", "", "print zfs tunable parameters", zfs_params },
3107 { "refcount", ":[-r]\n"
3108 "\t-r display recently removed references",
3109 "print refcount_t holders", refcount },
3110 { "zap_leaf", "", "print zap_leaf_phys_t", zap_leaf },
3111 { "zfs_aces", ":[-v]", "print all ACEs from a zfs_acl_t",
3112 zfs_acl_dump },
3113 { "zfs_ace", ":[-v]", "print zfs_ace", zfs_ace_print },
3114 { "zfs_ace0", ":[-v]", "print zfs_ace0", zfs_ace0_print },
3115 { "sa_attr_table", ":", "print SA attribute table from sa_os_t",
3116 sa_attr_table},
3117 { "sa_attr", ": attr_id",
3118 "print SA attribute address when given sa_handle_t", sa_attr_print},
3119 { "zfs_dbgmsg", ":[-va]",
3120 "print zfs debug log", dbgmsg},
3121 { "rrwlock", ":",
3122 "print rrwlock_t, including readers", rrwlock},
3123 { NULL }
3124 };
3125
3126 static const mdb_walker_t walkers[] = {
3127 { "zms_freelist", "walk ZFS metaslab freelist",
3128 freelist_walk_init, freelist_walk_step, NULL },
3129 { "txg_list", "given any txg_list_t *, walk all entries in all txgs",
3130 txg_list_walk_init, txg_list_walk_step, NULL },
3131 { "txg_list0", "given any txg_list_t *, walk all entries in txg 0",
3132 txg_list0_walk_init, txg_list_walk_step, NULL },
3133 { "txg_list1", "given any txg_list_t *, walk all entries in txg 1",
3134 txg_list1_walk_init, txg_list_walk_step, NULL },
3135 { "txg_list2", "given any txg_list_t *, walk all entries in txg 2",
3136 txg_list2_walk_init, txg_list_walk_step, NULL },
3137 { "txg_list3", "given any txg_list_t *, walk all entries in txg 3",
3138 txg_list3_walk_init, txg_list_walk_step, NULL },
3139 { "zio", "walk all zio structures, optionally for a particular spa_t",
3140 zio_walk_init, zio_walk_step, NULL },
3141 { "zio_root",
3142 "walk all root zio_t structures, optionally for a particular spa_t",
3143 zio_walk_init, zio_walk_root_step, NULL },
3144 { "spa", "walk all spa_t entries in the namespace",
3145 spa_walk_init, spa_walk_step, NULL },
3146 { "metaslab", "given a spa_t *, walk all metaslab_t structures",
3147 metaslab_walk_init, metaslab_walk_step, NULL },
3148 { "zfs_acl_node", "given a zfs_acl_t, walk all zfs_acl_nodes",
3149 zfs_acl_node_walk_init, zfs_acl_node_walk_step, NULL },
3150 { "zfs_acl_node_aces", "given a zfs_acl_node_t, walk all ACEs",
3151 zfs_acl_node_aces_walk_init, zfs_aces_walk_step, NULL },
3152 { "zfs_acl_node_aces0",
3153 "given a zfs_acl_node_t, walk all ACEs as ace_t",
3154 zfs_acl_node_aces0_walk_init, zfs_aces_walk_step, NULL },
3155 { NULL }
3156 };
3157
3158 static const mdb_modinfo_t modinfo = {
3159 MDB_API_VERSION, dcmds, walkers
3160 };
3161
3162 const mdb_modinfo_t *
3163 _mdb_init(void)
3164 {
3165 return (&modinfo);
3166 }