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;
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 typedef struct mdb_dsl_dir {
1429 uintptr_t dd_phys;
1430 int64_t dd_space_towrite[TXG_SIZE];
1431 } mdb_dsl_dir_t;
1432
1433 typedef struct mdb_dsl_dir_phys {
1434 uint64_t dd_used_bytes;
1435 uint64_t dd_compressed_bytes;
1436 uint64_t dd_uncompressed_bytes;
1437 } mdb_dsl_dir_phys_t;
1438
1439 typedef struct mdb_vdev {
1440 uintptr_t vdev_parent;
1441 uintptr_t vdev_ms;
1442 uint64_t vdev_ms_count;
1443 vdev_stat_t vdev_stat;
1444 } mdb_vdev_t;
1445
1446 typedef struct mdb_metaslab {
1447 space_map_t ms_allocmap[TXG_SIZE];
1448 space_map_t ms_freemap[TXG_SIZE];
1449 space_map_t ms_map;
1450 space_map_obj_t ms_smo;
1451 space_map_obj_t ms_smo_syncing;
1452 } mdb_metaslab_t;
1453
1454 typedef struct space_data {
1455 uint64_t ms_allocmap[TXG_SIZE];
1456 uint64_t ms_freemap[TXG_SIZE];
1457 uint64_t ms_map;
1458 uint64_t avail;
1459 uint64_t nowavail;
1460 } space_data_t;
1461
1462 /* ARGSUSED */
1463 static int
1464 space_cb(uintptr_t addr, const void *unknown, void *arg)
1465 {
1466 space_data_t *sd = arg;
1467 mdb_metaslab_t ms;
1468
1469 if (GETMEMB(addr, "metaslab", ms_allocmap, ms.ms_allocmap) ||
1470 GETMEMB(addr, "metaslab", ms_freemap, ms.ms_freemap) ||
1471 GETMEMB(addr, "metaslab", ms_map, ms.ms_map) ||
1472 GETMEMB(addr, "metaslab", ms_smo, ms.ms_smo) ||
1473 GETMEMB(addr, "metaslab", ms_smo_syncing, ms.ms_smo_syncing)) {
1474 return (WALK_ERR);
1475 }
1476
1477 sd->ms_allocmap[0] += ms.ms_allocmap[0].sm_space;
1478 sd->ms_allocmap[1] += ms.ms_allocmap[1].sm_space;
1479 sd->ms_allocmap[2] += ms.ms_allocmap[2].sm_space;
1480 sd->ms_allocmap[3] += ms.ms_allocmap[3].sm_space;
1481 sd->ms_freemap[0] += ms.ms_freemap[0].sm_space;
1482 sd->ms_freemap[1] += ms.ms_freemap[1].sm_space;
1483 sd->ms_freemap[2] += ms.ms_freemap[2].sm_space;
1484 sd->ms_freemap[3] += ms.ms_freemap[3].sm_space;
1485 sd->ms_map += ms.ms_map.sm_space;
1486 sd->avail += ms.ms_map.sm_size - ms.ms_smo.smo_alloc;
1487 sd->nowavail += ms.ms_map.sm_size - ms.ms_smo_syncing.smo_alloc;
1488
1489 return (WALK_NEXT);
1490 }
1491
1492 /*
1493 * ::spa_space [-b]
1494 *
1495 * Given a spa_t, print out it's on-disk space usage and in-core
1496 * estimates of future usage. If -b is given, print space in bytes.
1497 * Otherwise print in megabytes.
1498 */
1499 /* ARGSUSED */
1500 static int
1501 spa_space(uintptr_t addr, uint_t flags, int argc, const mdb_arg_t *argv)
1502 {
1503 mdb_spa_t spa;
1504 uintptr_t dp_root_dir;
1505 mdb_dsl_dir_t dd;
1506 mdb_dsl_dir_phys_t dsp;
1507 uint64_t children;
1508 uintptr_t childaddr;
1509 space_data_t sd;
1510 int shift = 20;
1511 char *suffix = "M";
1512 int bytes = B_FALSE;
1513
1514 if (mdb_getopts(argc, argv, 'b', MDB_OPT_SETBITS, TRUE, &bytes, NULL) !=
1515 argc)
1516 return (DCMD_USAGE);
1517 if (!(flags & DCMD_ADDRSPEC))
1518 return (DCMD_USAGE);
1519
1520 if (bytes) {
1521 shift = 0;
1522 suffix = "";
1523 }
1524
1525 if (GETMEMB(addr, "spa", spa_dsl_pool, spa.spa_dsl_pool) ||
1526 GETMEMB(addr, "spa", spa_root_vdev, spa.spa_root_vdev) ||
1527 GETMEMB(spa.spa_root_vdev, "vdev", vdev_children, children) ||
1528 GETMEMB(spa.spa_root_vdev, "vdev", vdev_child, childaddr) ||
1529 GETMEMB(spa.spa_dsl_pool, "dsl_pool",
1530 dp_root_dir, dp_root_dir) ||
1531 GETMEMB(dp_root_dir, "dsl_dir", dd_phys, dd.dd_phys) ||
1532 GETMEMB(dp_root_dir, "dsl_dir",
1533 dd_space_towrite, dd.dd_space_towrite) ||
1534 GETMEMB(dd.dd_phys, "dsl_dir_phys",
1535 dd_used_bytes, dsp.dd_used_bytes) ||
1536 GETMEMB(dd.dd_phys, "dsl_dir_phys",
1537 dd_compressed_bytes, dsp.dd_compressed_bytes) ||
1538 GETMEMB(dd.dd_phys, "dsl_dir_phys",
1539 dd_uncompressed_bytes, dsp.dd_uncompressed_bytes)) {
1540 return (DCMD_ERR);
1541 }
1542
1543 mdb_printf("dd_space_towrite = %llu%s %llu%s %llu%s %llu%s\n",
1544 dd.dd_space_towrite[0] >> shift, suffix,
1545 dd.dd_space_towrite[1] >> shift, suffix,
1546 dd.dd_space_towrite[2] >> shift, suffix,
1547 dd.dd_space_towrite[3] >> shift, suffix);
1548
1549 mdb_printf("dd_phys.dd_used_bytes = %llu%s\n",
1550 dsp.dd_used_bytes >> shift, suffix);
1551 mdb_printf("dd_phys.dd_compressed_bytes = %llu%s\n",
1552 dsp.dd_compressed_bytes >> shift, suffix);
1553 mdb_printf("dd_phys.dd_uncompressed_bytes = %llu%s\n",
1554 dsp.dd_uncompressed_bytes >> shift, suffix);
1555
1556 bzero(&sd, sizeof (sd));
1557 if (mdb_pwalk("metaslab", space_cb, &sd, addr) != 0) {
1558 mdb_warn("can't walk metaslabs");
1559 return (DCMD_ERR);
1560 }
1561
1562 mdb_printf("ms_allocmap = %llu%s %llu%s %llu%s %llu%s\n",
1563 sd.ms_allocmap[0] >> shift, suffix,
1564 sd.ms_allocmap[1] >> shift, suffix,
1565 sd.ms_allocmap[2] >> shift, suffix,
1566 sd.ms_allocmap[3] >> shift, suffix);
1567 mdb_printf("ms_freemap = %llu%s %llu%s %llu%s %llu%s\n",
1568 sd.ms_freemap[0] >> shift, suffix,
1569 sd.ms_freemap[1] >> shift, suffix,
1570 sd.ms_freemap[2] >> shift, suffix,
1571 sd.ms_freemap[3] >> shift, suffix);
1572 mdb_printf("ms_map = %llu%s\n", sd.ms_map >> shift, suffix);
1573 mdb_printf("last synced avail = %llu%s\n", sd.avail >> shift, suffix);
1574 mdb_printf("current syncing avail = %llu%s\n",
1575 sd.nowavail >> shift, suffix);
1576
1577 return (DCMD_OK);
1578 }
1579
1580 /*
1581 * ::spa_verify
1582 *
1583 * Given a spa_t, verify that that the pool is self-consistent.
1584 * Currently, it only checks to make sure that the vdev tree exists.
1585 */
1586 /* ARGSUSED */
1587 static int
1588 spa_verify(uintptr_t addr, uint_t flags, int argc, const mdb_arg_t *argv)
1589 {
1590 spa_t spa;
1591
1592 if (argc != 0 || !(flags & DCMD_ADDRSPEC))
1593 return (DCMD_USAGE);
1594
1595 if (mdb_vread(&spa, sizeof (spa), addr) == -1) {
1596 mdb_warn("failed to read spa_t at %p", addr);
1597 return (DCMD_ERR);
1598 }
1599
1600 if (spa.spa_root_vdev == NULL) {
1601 mdb_printf("no vdev tree present\n");
1602 return (DCMD_OK);
1603 }
1604
1605 return (DCMD_OK);
1606 }
1607
1608 static int
1609 spa_print_aux(spa_aux_vdev_t *sav, uint_t flags, mdb_arg_t *v,
1610 const char *name)
1611 {
1612 uintptr_t *aux;
1613 size_t len;
1614 int ret, i;
1615
1616 /*
1617 * Iterate over aux vdevs and print those out as well. This is a
1618 * little annoying because we don't have a root vdev to pass to ::vdev.
1619 * Instead, we print a single line and then call it for each child
1620 * vdev.
1621 */
1622 if (sav->sav_count != 0) {
1623 v[1].a_type = MDB_TYPE_STRING;
1624 v[1].a_un.a_str = "-d";
1625 v[2].a_type = MDB_TYPE_IMMEDIATE;
1626 v[2].a_un.a_val = 2;
1627
1628 len = sav->sav_count * sizeof (uintptr_t);
1629 aux = mdb_alloc(len, UM_SLEEP);
1630 if (mdb_vread(aux, len,
1631 (uintptr_t)sav->sav_vdevs) == -1) {
1632 mdb_free(aux, len);
1633 mdb_warn("failed to read l2cache vdevs at %p",
1634 sav->sav_vdevs);
1635 return (DCMD_ERR);
1636 }
1637
1638 mdb_printf("%-?s %-9s %-12s %s\n", "-", "-", "-", name);
1639
1640 for (i = 0; i < sav->sav_count; i++) {
1641 ret = mdb_call_dcmd("vdev", aux[i], flags, 3, v);
1642 if (ret != DCMD_OK) {
1643 mdb_free(aux, len);
1644 return (ret);
1645 }
1646 }
1647
1648 mdb_free(aux, len);
1649 }
1650
1651 return (0);
1652 }
1653
1654 /*
1655 * ::spa_vdevs
1656 *
1657 * -e Include error stats
1658 *
1659 * Print out a summarized list of vdevs for the given spa_t.
1660 * This is accomplished by invoking "::vdev -re" on the root vdev, as well as
1661 * iterating over the cache devices.
1662 */
1663 /* ARGSUSED */
1664 static int
1665 spa_vdevs(uintptr_t addr, uint_t flags, int argc, const mdb_arg_t *argv)
1666 {
1667 spa_t spa;
1668 mdb_arg_t v[3];
1669 int errors = FALSE;
1670 int ret;
1671
1672 if (mdb_getopts(argc, argv,
1673 'e', MDB_OPT_SETBITS, TRUE, &errors,
1674 NULL) != argc)
1675 return (DCMD_USAGE);
1676
1677 if (!(flags & DCMD_ADDRSPEC))
1678 return (DCMD_USAGE);
1679
1680 if (mdb_vread(&spa, sizeof (spa), addr) == -1) {
1681 mdb_warn("failed to read spa_t at %p", addr);
1682 return (DCMD_ERR);
1683 }
1684
1685 /*
1686 * Unitialized spa_t structures can have a NULL root vdev.
1687 */
1688 if (spa.spa_root_vdev == NULL) {
1689 mdb_printf("no associated vdevs\n");
1690 return (DCMD_OK);
1691 }
1692
1693 v[0].a_type = MDB_TYPE_STRING;
1694 v[0].a_un.a_str = errors ? "-re" : "-r";
1695
1696 ret = mdb_call_dcmd("vdev", (uintptr_t)spa.spa_root_vdev,
1697 flags, 1, v);
1698 if (ret != DCMD_OK)
1699 return (ret);
1700
1701 if (spa_print_aux(&spa.spa_l2cache, flags, v, "cache") != 0 ||
1702 spa_print_aux(&spa.spa_spares, flags, v, "spares") != 0)
1703 return (DCMD_ERR);
1704
1705 return (DCMD_OK);
1706 }
1707
1708 /*
1709 * ::zio
1710 *
1711 * Print a summary of zio_t and all its children. This is intended to display a
1712 * zio tree, and hence we only pick the most important pieces of information for
1713 * the main summary. More detailed information can always be found by doing a
1714 * '::print zio' on the underlying zio_t. The columns we display are:
1715 *
1716 * ADDRESS TYPE STAGE WAITER TIME_ELAPSED
1717 *
1718 * The 'address' column is indented by one space for each depth level as we
1719 * descend down the tree.
1720 */
1721
1722 #define ZIO_MAXINDENT 7
1723 #define ZIO_MAXWIDTH (sizeof (uintptr_t) * 2 + ZIO_MAXINDENT)
1724 #define ZIO_WALK_SELF 0
1725 #define ZIO_WALK_CHILD 1
1726 #define ZIO_WALK_PARENT 2
1727
1728 typedef struct zio_print_args {
1729 int zpa_current_depth;
1730 int zpa_min_depth;
1731 int zpa_max_depth;
1732 int zpa_type;
1733 uint_t zpa_flags;
1734 } zio_print_args_t;
1735
1736 typedef struct mdb_zio {
1737 enum zio_type io_type;
1738 enum zio_stage io_stage;
1739 void *io_waiter;
1740 void *io_spa;
1741 struct {
1742 struct {
1743 void *list_next;
1744 } list_head;
1745 } io_parent_list;
1746 int io_error;
1747 } mdb_zio_t;
1748
1749 typedef struct mdb_zio_timestamp {
1750 hrtime_t io_timestamp;
1751 } mdb_zio_timestamp_t;
1752
1753 static int zio_child_cb(uintptr_t addr, const void *unknown, void *arg);
1754
1755 static int
1756 zio_print_cb(uintptr_t addr, zio_print_args_t *zpa)
1757 {
1758 mdb_ctf_id_t type_enum, stage_enum;
1759 int indent = zpa->zpa_current_depth;
1760 const char *type, *stage;
1761 uintptr_t laddr;
1762 mdb_zio_t zio;
1763 mdb_zio_timestamp_t zio_timestamp = { 0 };
1764
1765 if (mdb_ctf_vread(&zio, ZFS_STRUCT "zio", "mdb_zio_t", addr, 0) == -1)
1766 return (WALK_ERR);
1767 (void) mdb_ctf_vread(&zio_timestamp, ZFS_STRUCT "zio",
1768 "mdb_zio_timestamp_t", addr, MDB_CTF_VREAD_QUIET);
1769
1770 if (indent > ZIO_MAXINDENT)
1771 indent = ZIO_MAXINDENT;
1772
1773 if (mdb_ctf_lookup_by_name("enum zio_type", &type_enum) == -1 ||
1774 mdb_ctf_lookup_by_name("enum zio_stage", &stage_enum) == -1) {
1775 mdb_warn("failed to lookup zio enums");
1776 return (WALK_ERR);
1777 }
1778
1779 if ((type = mdb_ctf_enum_name(type_enum, zio.io_type)) != NULL)
1780 type += sizeof ("ZIO_TYPE_") - 1;
1781 else
1782 type = "?";
1783
1784 if (zio.io_error == 0) {
1785 stage = mdb_ctf_enum_name(stage_enum, zio.io_stage);
1786 if (stage != NULL)
1787 stage += sizeof ("ZIO_STAGE_") - 1;
1788 else
1789 stage = "?";
1790 } else {
1791 stage = "FAILED";
1792 }
1793
1794 if (zpa->zpa_current_depth >= zpa->zpa_min_depth) {
1795 if (zpa->zpa_flags & DCMD_PIPE_OUT) {
1796 mdb_printf("%?p\n", addr);
1797 } else {
1798 mdb_printf("%*s%-*p %-5s %-16s ", indent, "",
1799 ZIO_MAXWIDTH - indent, addr, type, stage);
1800 if (zio.io_waiter)
1801 mdb_printf("%-16p ", zio.io_waiter);
1802 else
1803 mdb_printf("%-16s ", "-");
1804 #ifdef _KERNEL
1805 if (zio_timestamp.io_timestamp != 0) {
1806 mdb_printf("%llums", (mdb_gethrtime() -
1807 zio_timestamp.io_timestamp) /
1808 1000000);
1809 } else {
1810 mdb_printf("%-12s ", "-");
1811 }
1812 #else
1813 mdb_printf("%-12s ", "-");
1814 #endif
1815 mdb_printf("\n");
1816 }
1817 }
1818
1819 if (zpa->zpa_current_depth >= zpa->zpa_max_depth)
1820 return (WALK_NEXT);
1821
1822 if (zpa->zpa_type == ZIO_WALK_PARENT)
1823 laddr = addr + mdb_ctf_offsetof_by_name(ZFS_STRUCT "zio",
1824 "io_parent_list");
1825 else
1826 laddr = addr + mdb_ctf_offsetof_by_name(ZFS_STRUCT "zio",
1827 "io_child_list");
1828
1829 zpa->zpa_current_depth++;
1830 if (mdb_pwalk("list", zio_child_cb, zpa, laddr) != 0) {
1831 mdb_warn("failed to walk zio_t children at %p\n", laddr);
1832 return (WALK_ERR);
1833 }
1834 zpa->zpa_current_depth--;
1835
1836 return (WALK_NEXT);
1837 }
1838
1839 /* ARGSUSED */
1840 static int
1841 zio_child_cb(uintptr_t addr, const void *unknown, void *arg)
1842 {
1843 zio_link_t zl;
1844 uintptr_t ziop;
1845 zio_print_args_t *zpa = arg;
1846
1847 if (mdb_vread(&zl, sizeof (zl), addr) == -1) {
1848 mdb_warn("failed to read zio_link_t at %p", addr);
1849 return (WALK_ERR);
1850 }
1851
1852 if (zpa->zpa_type == ZIO_WALK_PARENT)
1853 ziop = (uintptr_t)zl.zl_parent;
1854 else
1855 ziop = (uintptr_t)zl.zl_child;
1856
1857 return (zio_print_cb(ziop, arg));
1858 }
1859
1860 /* ARGSUSED */
1861 static int
1862 zio_print(uintptr_t addr, uint_t flags, int argc, const mdb_arg_t *argv)
1863 {
1864 zio_print_args_t zpa = { 0 };
1865
1866 if (!(flags & DCMD_ADDRSPEC))
1867 return (DCMD_USAGE);
1868
1869 if (mdb_getopts(argc, argv,
1870 'r', MDB_OPT_SETBITS, INT_MAX, &zpa.zpa_max_depth,
1871 'c', MDB_OPT_SETBITS, ZIO_WALK_CHILD, &zpa.zpa_type,
1872 'p', MDB_OPT_SETBITS, ZIO_WALK_PARENT, &zpa.zpa_type,
1873 NULL) != argc)
1874 return (DCMD_USAGE);
1875
1876 zpa.zpa_flags = flags;
1877 if (zpa.zpa_max_depth != 0) {
1878 if (zpa.zpa_type == ZIO_WALK_SELF)
1879 zpa.zpa_type = ZIO_WALK_CHILD;
1880 } else if (zpa.zpa_type != ZIO_WALK_SELF) {
1881 zpa.zpa_min_depth = 1;
1882 zpa.zpa_max_depth = 1;
1883 }
1884
1885 if (!(flags & DCMD_PIPE_OUT) && DCMD_HDRSPEC(flags)) {
1886 mdb_printf("%<u>%-*s %-5s %-16s %-16s %-12s%</u>\n",
1887 ZIO_MAXWIDTH, "ADDRESS", "TYPE", "STAGE", "WAITER",
1888 "TIME_ELAPSED");
1889 }
1890
1891 if (zio_print_cb(addr, &zpa) != WALK_NEXT)
1892 return (DCMD_ERR);
1893
1894 return (DCMD_OK);
1895 }
1896
1897 /*
1898 * [addr]::zio_state
1899 *
1900 * Print a summary of all zio_t structures on the system, or for a particular
1901 * pool. This is equivalent to '::walk zio_root | ::zio'.
1902 */
1903 /*ARGSUSED*/
1904 static int
1905 zio_state(uintptr_t addr, uint_t flags, int argc, const mdb_arg_t *argv)
1906 {
1907 /*
1908 * MDB will remember the last address of the pipeline, so if we don't
1909 * zero this we'll end up trying to walk zio structures for a
1910 * non-existent spa_t.
1911 */
1912 if (!(flags & DCMD_ADDRSPEC))
1913 addr = 0;
1914
1915 return (mdb_pwalk_dcmd("zio_root", "zio", argc, argv, addr));
1916 }
1917
1918 typedef struct txg_list_walk_data {
1919 uintptr_t lw_head[TXG_SIZE];
1920 int lw_txgoff;
1921 int lw_maxoff;
1922 size_t lw_offset;
1923 void *lw_obj;
1924 } txg_list_walk_data_t;
1925
1926 static int
1927 txg_list_walk_init_common(mdb_walk_state_t *wsp, int txg, int maxoff)
1928 {
1929 txg_list_walk_data_t *lwd;
1930 txg_list_t list;
1931 int i;
1932
1933 lwd = mdb_alloc(sizeof (txg_list_walk_data_t), UM_SLEEP | UM_GC);
1934 if (mdb_vread(&list, sizeof (txg_list_t), wsp->walk_addr) == -1) {
1935 mdb_warn("failed to read txg_list_t at %#lx", wsp->walk_addr);
1936 return (WALK_ERR);
1937 }
1938
1939 for (i = 0; i < TXG_SIZE; i++)
1940 lwd->lw_head[i] = (uintptr_t)list.tl_head[i];
1941 lwd->lw_offset = list.tl_offset;
1942 lwd->lw_obj = mdb_alloc(lwd->lw_offset + sizeof (txg_node_t),
1943 UM_SLEEP | UM_GC);
1944 lwd->lw_txgoff = txg;
1945 lwd->lw_maxoff = maxoff;
1946
1947 wsp->walk_addr = lwd->lw_head[lwd->lw_txgoff];
1948 wsp->walk_data = lwd;
1949
1950 return (WALK_NEXT);
1951 }
1952
1953 static int
1954 txg_list_walk_init(mdb_walk_state_t *wsp)
1955 {
1956 return (txg_list_walk_init_common(wsp, 0, TXG_SIZE-1));
1957 }
1958
1959 static int
1960 txg_list0_walk_init(mdb_walk_state_t *wsp)
1961 {
1962 return (txg_list_walk_init_common(wsp, 0, 0));
1963 }
1964
1965 static int
1966 txg_list1_walk_init(mdb_walk_state_t *wsp)
1967 {
1968 return (txg_list_walk_init_common(wsp, 1, 1));
1969 }
1970
1971 static int
1972 txg_list2_walk_init(mdb_walk_state_t *wsp)
1973 {
1974 return (txg_list_walk_init_common(wsp, 2, 2));
1975 }
1976
1977 static int
1978 txg_list3_walk_init(mdb_walk_state_t *wsp)
1979 {
1980 return (txg_list_walk_init_common(wsp, 3, 3));
1981 }
1982
1983 static int
1984 txg_list_walk_step(mdb_walk_state_t *wsp)
1985 {
1986 txg_list_walk_data_t *lwd = wsp->walk_data;
1987 uintptr_t addr;
1988 txg_node_t *node;
1989 int status;
1990
1991 while (wsp->walk_addr == NULL && lwd->lw_txgoff < lwd->lw_maxoff) {
1992 lwd->lw_txgoff++;
1993 wsp->walk_addr = lwd->lw_head[lwd->lw_txgoff];
1994 }
1995
1996 if (wsp->walk_addr == NULL)
1997 return (WALK_DONE);
1998
1999 addr = wsp->walk_addr - lwd->lw_offset;
2000
2001 if (mdb_vread(lwd->lw_obj,
2002 lwd->lw_offset + sizeof (txg_node_t), addr) == -1) {
2003 mdb_warn("failed to read list element at %#lx", addr);
2004 return (WALK_ERR);
2005 }
2006
2007 status = wsp->walk_callback(addr, lwd->lw_obj, wsp->walk_cbdata);
2008 node = (txg_node_t *)((uintptr_t)lwd->lw_obj + lwd->lw_offset);
2009 wsp->walk_addr = (uintptr_t)node->tn_next[lwd->lw_txgoff];
2010
2011 return (status);
2012 }
2013
2014 /*
2015 * ::walk spa
2016 *
2017 * Walk all named spa_t structures in the namespace. This is nothing more than
2018 * a layered avl walk.
2019 */
2020 static int
2021 spa_walk_init(mdb_walk_state_t *wsp)
2022 {
2023 GElf_Sym sym;
2024
2025 if (wsp->walk_addr != NULL) {
2026 mdb_warn("spa walk only supports global walks\n");
2027 return (WALK_ERR);
2028 }
2029
2030 if (mdb_lookup_by_obj(ZFS_OBJ_NAME, "spa_namespace_avl", &sym) == -1) {
2031 mdb_warn("failed to find symbol 'spa_namespace_avl'");
2032 return (WALK_ERR);
2033 }
2034
2035 wsp->walk_addr = (uintptr_t)sym.st_value;
2036
2037 if (mdb_layered_walk("avl", wsp) == -1) {
2038 mdb_warn("failed to walk 'avl'\n");
2039 return (WALK_ERR);
2040 }
2041
2042 return (WALK_NEXT);
2043 }
2044
2045 static int
2046 spa_walk_step(mdb_walk_state_t *wsp)
2047 {
2048 spa_t spa;
2049
2050 if (mdb_vread(&spa, sizeof (spa), wsp->walk_addr) == -1) {
2051 mdb_warn("failed to read spa_t at %p", wsp->walk_addr);
2052 return (WALK_ERR);
2053 }
2054
2055 return (wsp->walk_callback(wsp->walk_addr, &spa, wsp->walk_cbdata));
2056 }
2057
2058 /*
2059 * [addr]::walk zio
2060 *
2061 * Walk all active zio_t structures on the system. This is simply a layered
2062 * walk on top of ::walk zio_cache, with the optional ability to limit the
2063 * structures to a particular pool.
2064 */
2065 static int
2066 zio_walk_init(mdb_walk_state_t *wsp)
2067 {
2068 wsp->walk_data = (void *)wsp->walk_addr;
2069
2070 if (mdb_layered_walk("zio_cache", wsp) == -1) {
2071 mdb_warn("failed to walk 'zio_cache'\n");
2072 return (WALK_ERR);
2073 }
2074
2075 return (WALK_NEXT);
2076 }
2077
2078 static int
2079 zio_walk_step(mdb_walk_state_t *wsp)
2080 {
2081 mdb_zio_t zio;
2082
2083 if (mdb_ctf_vread(&zio, ZFS_STRUCT "zio", "mdb_zio_t",
2084 wsp->walk_addr, 0) == -1)
2085 return (WALK_ERR);
2086
2087 if (wsp->walk_data != NULL && wsp->walk_data != zio.io_spa)
2088 return (WALK_NEXT);
2089
2090 return (wsp->walk_callback(wsp->walk_addr, &zio, wsp->walk_cbdata));
2091 }
2092
2093 /*
2094 * [addr]::walk zio_root
2095 *
2096 * Walk only root zio_t structures, optionally for a particular spa_t.
2097 */
2098 static int
2099 zio_walk_root_step(mdb_walk_state_t *wsp)
2100 {
2101 mdb_zio_t zio;
2102
2103 if (mdb_ctf_vread(&zio, ZFS_STRUCT "zio", "mdb_zio_t",
2104 wsp->walk_addr, 0) == -1)
2105 return (WALK_ERR);
2106
2107 if (wsp->walk_data != NULL && wsp->walk_data != zio.io_spa)
2108 return (WALK_NEXT);
2109
2110 /* If the parent list is not empty, ignore */
2111 if ((uintptr_t)zio.io_parent_list.list_head.list_next !=
2112 wsp->walk_addr +
2113 mdb_ctf_offsetof_by_name(ZFS_STRUCT "zio", "io_parent_list") +
2114 mdb_ctf_offsetof_by_name("struct list", "list_head"))
2115 return (WALK_NEXT);
2116
2117 return (wsp->walk_callback(wsp->walk_addr, &zio, wsp->walk_cbdata));
2118 }
2119
2120 #define NICENUM_BUFLEN 6
2121
2122 static int
2123 snprintfrac(char *buf, int len,
2124 uint64_t numerator, uint64_t denom, int frac_digits)
2125 {
2126 int mul = 1;
2127 int whole, frac, i;
2128
2129 for (i = frac_digits; i; i--)
2130 mul *= 10;
2131 whole = numerator / denom;
2132 frac = mul * numerator / denom - mul * whole;
2133 return (mdb_snprintf(buf, len, "%u.%0*u", whole, frac_digits, frac));
2134 }
2135
2136 static void
2137 mdb_nicenum(uint64_t num, char *buf)
2138 {
2139 uint64_t n = num;
2140 int index = 0;
2141 char *u;
2142
2143 while (n >= 1024) {
2144 n = (n + (1024 / 2)) / 1024; /* Round up or down */
2145 index++;
2146 }
2147
2148 u = &" \0K\0M\0G\0T\0P\0E\0"[index*2];
2149
2150 if (index == 0) {
2151 (void) mdb_snprintf(buf, NICENUM_BUFLEN, "%llu",
2152 (u_longlong_t)n);
2153 } else if (n < 10 && (num & (num - 1)) != 0) {
2154 (void) snprintfrac(buf, NICENUM_BUFLEN,
2155 num, 1ULL << 10 * index, 2);
2156 strcat(buf, u);
2157 } else if (n < 100 && (num & (num - 1)) != 0) {
2158 (void) snprintfrac(buf, NICENUM_BUFLEN,
2159 num, 1ULL << 10 * index, 1);
2160 strcat(buf, u);
2161 } else {
2162 (void) mdb_snprintf(buf, NICENUM_BUFLEN, "%llu%s",
2163 (u_longlong_t)n, u);
2164 }
2165 }
2166
2167 /*
2168 * ::zfs_blkstats
2169 *
2170 * -v print verbose per-level information
2171 *
2172 */
2173 static int
2174 zfs_blkstats(uintptr_t addr, uint_t flags, int argc, const mdb_arg_t *argv)
2175 {
2176 boolean_t verbose = B_FALSE;
2177 zfs_all_blkstats_t stats;
2178 dmu_object_type_t t;
2179 zfs_blkstat_t *tzb;
2180 uint64_t ditto;
2181 dmu_object_type_info_t dmu_ot[DMU_OT_NUMTYPES + 10];
2182 /* +10 in case it grew */
2183
2184 if (mdb_readvar(&dmu_ot, "dmu_ot") == -1) {
2185 mdb_warn("failed to read 'dmu_ot'");
2186 return (DCMD_ERR);
2187 }
2188
2189 if (mdb_getopts(argc, argv,
2190 'v', MDB_OPT_SETBITS, TRUE, &verbose,
2191 NULL) != argc)
2192 return (DCMD_USAGE);
2193
2194 if (!(flags & DCMD_ADDRSPEC))
2195 return (DCMD_USAGE);
2196
2197 if (GETMEMB(addr, "spa", spa_dsl_pool, addr) ||
2198 GETMEMB(addr, "dsl_pool", dp_blkstats, addr) ||
2199 mdb_vread(&stats, sizeof (zfs_all_blkstats_t), addr) == -1) {
2200 mdb_warn("failed to read data at %p;", addr);
2201 mdb_printf("maybe no stats? run \"zpool scrub\" first.");
2202 return (DCMD_ERR);
2203 }
2204
2205 tzb = &stats.zab_type[DN_MAX_LEVELS][DMU_OT_TOTAL];
2206 if (tzb->zb_gangs != 0) {
2207 mdb_printf("Ganged blocks: %llu\n",
2208 (longlong_t)tzb->zb_gangs);
2209 }
2210
2211 ditto = tzb->zb_ditto_2_of_2_samevdev + tzb->zb_ditto_2_of_3_samevdev +
2212 tzb->zb_ditto_3_of_3_samevdev;
2213 if (ditto != 0) {
2214 mdb_printf("Dittoed blocks on same vdev: %llu\n",
2215 (longlong_t)ditto);
2216 }
2217
2218 mdb_printf("\nBlocks\tLSIZE\tPSIZE\tASIZE"
2219 "\t avg\t comp\t%%Total\tType\n");
2220
2221 for (t = 0; t <= DMU_OT_TOTAL; t++) {
2222 char csize[NICENUM_BUFLEN], lsize[NICENUM_BUFLEN];
2223 char psize[NICENUM_BUFLEN], asize[NICENUM_BUFLEN];
2224 char avg[NICENUM_BUFLEN];
2225 char comp[NICENUM_BUFLEN], pct[NICENUM_BUFLEN];
2226 char typename[64];
2227 int l;
2228
2229
2230 if (t == DMU_OT_DEFERRED)
2231 strcpy(typename, "deferred free");
2232 else if (t == DMU_OT_OTHER)
2233 strcpy(typename, "other");
2234 else if (t == DMU_OT_TOTAL)
2235 strcpy(typename, "Total");
2236 else if (mdb_readstr(typename, sizeof (typename),
2237 (uintptr_t)dmu_ot[t].ot_name) == -1) {
2238 mdb_warn("failed to read type name");
2239 return (DCMD_ERR);
2240 }
2241
2242 if (stats.zab_type[DN_MAX_LEVELS][t].zb_asize == 0)
2243 continue;
2244
2245 for (l = -1; l < DN_MAX_LEVELS; l++) {
2246 int level = (l == -1 ? DN_MAX_LEVELS : l);
2247 zfs_blkstat_t *zb = &stats.zab_type[level][t];
2248
2249 if (zb->zb_asize == 0)
2250 continue;
2251
2252 /*
2253 * Don't print each level unless requested.
2254 */
2255 if (!verbose && level != DN_MAX_LEVELS)
2256 continue;
2257
2258 /*
2259 * If all the space is level 0, don't print the
2260 * level 0 separately.
2261 */
2262 if (level == 0 && zb->zb_asize ==
2263 stats.zab_type[DN_MAX_LEVELS][t].zb_asize)
2264 continue;
2265
2266 mdb_nicenum(zb->zb_count, csize);
2267 mdb_nicenum(zb->zb_lsize, lsize);
2268 mdb_nicenum(zb->zb_psize, psize);
2269 mdb_nicenum(zb->zb_asize, asize);
2270 mdb_nicenum(zb->zb_asize / zb->zb_count, avg);
2271 (void) snprintfrac(comp, NICENUM_BUFLEN,
2272 zb->zb_lsize, zb->zb_psize, 2);
2273 (void) snprintfrac(pct, NICENUM_BUFLEN,
2274 100 * zb->zb_asize, tzb->zb_asize, 2);
2275
2276 mdb_printf("%6s\t%5s\t%5s\t%5s\t%5s"
2277 "\t%5s\t%6s\t",
2278 csize, lsize, psize, asize, avg, comp, pct);
2279
2280 if (level == DN_MAX_LEVELS)
2281 mdb_printf("%s\n", typename);
2282 else
2283 mdb_printf(" L%d %s\n",
2284 level, typename);
2285 }
2286 }
2287
2288 return (DCMD_OK);
2289 }
2290
2291 /* ARGSUSED */
2292 static int
2293 reference_cb(uintptr_t addr, const void *ignored, void *arg)
2294 {
2295 static int gotid;
2296 static mdb_ctf_id_t ref_id;
2297 uintptr_t ref_holder;
2298 uintptr_t ref_removed;
2299 uint64_t ref_number;
2300 boolean_t holder_is_str = B_FALSE;
2301 char holder_str[128];
2302 boolean_t removed = (boolean_t)arg;
2303
2304 if (!gotid) {
2305 if (mdb_ctf_lookup_by_name("struct reference", &ref_id) == -1) {
2306 mdb_warn("couldn't find struct reference");
2307 return (WALK_ERR);
2308 }
2309 gotid = TRUE;
2310 }
2311
2312 if (GETMEMBID(addr, &ref_id, ref_holder, ref_holder) ||
2313 GETMEMBID(addr, &ref_id, ref_removed, ref_removed) ||
2314 GETMEMBID(addr, &ref_id, ref_number, ref_number))
2315 return (WALK_ERR);
2316
2317 if (mdb_readstr(holder_str, sizeof (holder_str), ref_holder) != -1)
2318 holder_is_str = strisprint(holder_str);
2319
2320 if (removed)
2321 mdb_printf("removed ");
2322 mdb_printf("reference ");
2323 if (ref_number != 1)
2324 mdb_printf("with count=%llu ", ref_number);
2325 mdb_printf("with tag %p", (void*)ref_holder);
2326 if (holder_is_str)
2327 mdb_printf(" \"%s\"", holder_str);
2328 mdb_printf(", held at:\n");
2329
2330 (void) mdb_call_dcmd("whatis", addr, DCMD_ADDRSPEC, 0, NULL);
2331
2332 if (removed) {
2333 mdb_printf("removed at:\n");
2334 (void) mdb_call_dcmd("whatis", ref_removed,
2335 DCMD_ADDRSPEC, 0, NULL);
2336 }
2337
2338 mdb_printf("\n");
2339
2340 return (WALK_NEXT);
2341 }
2342
2343 /* ARGSUSED */
2344 static int
2345 refcount(uintptr_t addr, uint_t flags, int argc, const mdb_arg_t *argv)
2346 {
2347 uint64_t rc_count, rc_removed_count;
2348 uintptr_t rc_list, rc_removed;
2349 static int gotid;
2350 static mdb_ctf_id_t rc_id;
2351 ulong_t off;
2352 boolean_t released = B_FALSE;
2353
2354 if (!(flags & DCMD_ADDRSPEC))
2355 return (DCMD_USAGE);
2356
2357 if (mdb_getopts(argc, argv,
2358 'r', MDB_OPT_SETBITS, B_TRUE, &released,
2359 NULL) != argc)
2360 return (DCMD_USAGE);
2361
2362 if (!gotid) {
2363 /*
2364 * The refcount structure is different when compiled debug
2365 * vs nondebug. Therefore, we want to make sure we get the
2366 * refcount definition from the ZFS module, in case it has
2367 * been compiled debug but genunix is nondebug.
2368 */
2369 if (mdb_ctf_lookup_by_name(ZFS_STRUCT "refcount",
2370 &rc_id) == -1) {
2371 mdb_warn("couldn't find struct refcount");
2372 return (DCMD_ERR);
2373 }
2374 gotid = TRUE;
2375 }
2376
2377 if (GETMEMBID(addr, &rc_id, rc_count, rc_count))
2378 return (DCMD_ERR);
2379
2380 if (GETMEMBID(addr, &rc_id, rc_removed_count, rc_removed_count)) {
2381 mdb_printf("refcount_t at %p has %llu current holds\n",
2382 addr, (longlong_t)rc_count);
2383 return (DCMD_OK);
2384 }
2385
2386 mdb_printf("refcount_t at %p has %llu current holds, "
2387 "%llu recently released holds\n",
2388 addr, (longlong_t)rc_count, (longlong_t)rc_removed_count);
2389
2390 if (rc_count > 0)
2391 mdb_printf("current holds:\n");
2392 if (mdb_ctf_offsetof(rc_id, "rc_list", &off) == -1)
2393 return (DCMD_ERR);
2394 rc_list = addr + off/NBBY;
2395 mdb_pwalk("list", reference_cb, (void*)B_FALSE, rc_list);
2396
2397 if (released) {
2398 if (rc_removed_count > 0)
2399 mdb_printf("released holds:\n");
2400 if (mdb_ctf_offsetof(rc_id, "rc_removed", &off) == -1)
2401 return (DCMD_ERR);
2402 rc_removed = addr + off/NBBY;
2403 mdb_pwalk("list", reference_cb, (void*)B_TRUE, rc_removed);
2404 }
2405
2406 return (DCMD_OK);
2407 }
2408
2409 /* ARGSUSED */
2410 static int
2411 sa_attr_table(uintptr_t addr, uint_t flags, int argc, const mdb_arg_t *argv)
2412 {
2413 sa_attr_table_t *table;
2414 sa_os_t sa_os;
2415 char *name;
2416 int i;
2417
2418 if (mdb_vread(&sa_os, sizeof (sa_os_t), addr) == -1) {
2419 mdb_warn("failed to read sa_os at %p", addr);
2420 return (DCMD_ERR);
2421 }
2422
2423 table = mdb_alloc(sizeof (sa_attr_table_t) * sa_os.sa_num_attrs,
2424 UM_SLEEP | UM_GC);
2425 name = mdb_alloc(MAXPATHLEN, UM_SLEEP | UM_GC);
2426
2427 if (mdb_vread(table, sizeof (sa_attr_table_t) * sa_os.sa_num_attrs,
2428 (uintptr_t)sa_os.sa_attr_table) == -1) {
2429 mdb_warn("failed to read sa_os at %p", addr);
2430 return (DCMD_ERR);
2431 }
2432
2433 mdb_printf("%<u>%-10s %-10s %-10s %-10s %s%</u>\n",
2434 "ATTR ID", "REGISTERED", "LENGTH", "BSWAP", "NAME");
2435 for (i = 0; i != sa_os.sa_num_attrs; i++) {
2436 mdb_readstr(name, MAXPATHLEN, (uintptr_t)table[i].sa_name);
2437 mdb_printf("%5x %8x %8x %8x %-s\n",
2438 (int)table[i].sa_attr, (int)table[i].sa_registered,
2439 (int)table[i].sa_length, table[i].sa_byteswap, name);
2440 }
2441
2442 return (DCMD_OK);
2443 }
2444
2445 static int
2446 sa_get_off_table(uintptr_t addr, uint32_t **off_tab, int attr_count)
2447 {
2448 uintptr_t idx_table;
2449
2450 if (GETMEMB(addr, "sa_idx_tab", sa_idx_tab, idx_table)) {
2451 mdb_printf("can't find offset table in sa_idx_tab\n");
2452 return (-1);
2453 }
2454
2455 *off_tab = mdb_alloc(attr_count * sizeof (uint32_t),
2456 UM_SLEEP | UM_GC);
2457
2458 if (mdb_vread(*off_tab,
2459 attr_count * sizeof (uint32_t), idx_table) == -1) {
2460 mdb_warn("failed to attribute offset table %p", idx_table);
2461 return (-1);
2462 }
2463
2464 return (DCMD_OK);
2465 }
2466
2467 /*ARGSUSED*/
2468 static int
2469 sa_attr_print(uintptr_t addr, uint_t flags, int argc, const mdb_arg_t *argv)
2470 {
2471 uint32_t *offset_tab;
2472 int attr_count;
2473 uint64_t attr_id;
2474 uintptr_t attr_addr;
2475 uintptr_t bonus_tab, spill_tab;
2476 uintptr_t db_bonus, db_spill;
2477 uintptr_t os, os_sa;
2478 uintptr_t db_data;
2479
2480 if (argc != 1)
2481 return (DCMD_USAGE);
2482
2483 if (argv[0].a_type == MDB_TYPE_STRING)
2484 attr_id = mdb_strtoull(argv[0].a_un.a_str);
2485 else
2486 return (DCMD_USAGE);
2487
2488 if (GETMEMB(addr, "sa_handle", sa_bonus_tab, bonus_tab) ||
2489 GETMEMB(addr, "sa_handle", sa_spill_tab, spill_tab) ||
2490 GETMEMB(addr, "sa_handle", sa_os, os) ||
2491 GETMEMB(addr, "sa_handle", sa_bonus, db_bonus) ||
2492 GETMEMB(addr, "sa_handle", sa_spill, db_spill)) {
2493 mdb_printf("Can't find necessary information in sa_handle "
2494 "in sa_handle\n");
2495 return (DCMD_ERR);
2496 }
2497
2498 if (GETMEMB(os, "objset", os_sa, os_sa)) {
2499 mdb_printf("Can't find os_sa in objset\n");
2500 return (DCMD_ERR);
2501 }
2502
2503 if (GETMEMB(os_sa, "sa_os", sa_num_attrs, attr_count)) {
2504 mdb_printf("Can't find sa_num_attrs\n");
2505 return (DCMD_ERR);
2506 }
2507
2508 if (attr_id > attr_count) {
2509 mdb_printf("attribute id number is out of range\n");
2510 return (DCMD_ERR);
2511 }
2512
2513 if (bonus_tab) {
2514 if (sa_get_off_table(bonus_tab, &offset_tab,
2515 attr_count) == -1) {
2516 return (DCMD_ERR);
2517 }
2518
2519 if (GETMEMB(db_bonus, "dmu_buf", db_data, db_data)) {
2520 mdb_printf("can't find db_data in bonus dbuf\n");
2521 return (DCMD_ERR);
2522 }
2523 }
2524
2525 if (bonus_tab && !TOC_ATTR_PRESENT(offset_tab[attr_id]) &&
2526 spill_tab == NULL) {
2527 mdb_printf("Attribute does not exist\n");
2528 return (DCMD_ERR);
2529 } else if (!TOC_ATTR_PRESENT(offset_tab[attr_id]) && spill_tab) {
2530 if (sa_get_off_table(spill_tab, &offset_tab,
2531 attr_count) == -1) {
2532 return (DCMD_ERR);
2533 }
2534 if (GETMEMB(db_spill, "dmu_buf", db_data, db_data)) {
2535 mdb_printf("can't find db_data in spill dbuf\n");
2536 return (DCMD_ERR);
2537 }
2538 if (!TOC_ATTR_PRESENT(offset_tab[attr_id])) {
2539 mdb_printf("Attribute does not exist\n");
2540 return (DCMD_ERR);
2541 }
2542 }
2543 attr_addr = db_data + TOC_OFF(offset_tab[attr_id]);
2544 mdb_printf("%p\n", attr_addr);
2545 return (DCMD_OK);
2546 }
2547
2548 /* ARGSUSED */
2549 static int
2550 zfs_ace_print_common(uintptr_t addr, uint_t flags,
2551 uint64_t id, uint32_t access_mask, uint16_t ace_flags,
2552 uint16_t ace_type, int verbose)
2553 {
2554 if (DCMD_HDRSPEC(flags) && !verbose)
2555 mdb_printf("%<u>%-?s %-8s %-8s %-8s %s%</u>\n",
2556 "ADDR", "FLAGS", "MASK", "TYPE", "ID");
2557
2558 if (!verbose) {
2559 mdb_printf("%0?p %-8x %-8x %-8x %-llx\n", addr,
2560 ace_flags, access_mask, ace_type, id);
2561 return (DCMD_OK);
2562 }
2563
2564 switch (ace_flags & ACE_TYPE_FLAGS) {
2565 case ACE_OWNER:
2566 mdb_printf("owner@:");
2567 break;
2568 case (ACE_IDENTIFIER_GROUP | ACE_GROUP):
2569 mdb_printf("group@:");
2570 break;
2571 case ACE_EVERYONE:
2572 mdb_printf("everyone@:");
2573 break;
2574 case ACE_IDENTIFIER_GROUP:
2575 mdb_printf("group:%llx:", (u_longlong_t)id);
2576 break;
2577 case 0: /* User entry */
2578 mdb_printf("user:%llx:", (u_longlong_t)id);
2579 break;
2580 }
2581
2582 /* print out permission mask */
2583 if (access_mask & ACE_READ_DATA)
2584 mdb_printf("r");
2585 else
2586 mdb_printf("-");
2587 if (access_mask & ACE_WRITE_DATA)
2588 mdb_printf("w");
2589 else
2590 mdb_printf("-");
2591 if (access_mask & ACE_EXECUTE)
2592 mdb_printf("x");
2593 else
2594 mdb_printf("-");
2595 if (access_mask & ACE_APPEND_DATA)
2596 mdb_printf("p");
2597 else
2598 mdb_printf("-");
2599 if (access_mask & ACE_DELETE)
2600 mdb_printf("d");
2601 else
2602 mdb_printf("-");
2603 if (access_mask & ACE_DELETE_CHILD)
2604 mdb_printf("D");
2605 else
2606 mdb_printf("-");
2607 if (access_mask & ACE_READ_ATTRIBUTES)
2608 mdb_printf("a");
2609 else
2610 mdb_printf("-");
2611 if (access_mask & ACE_WRITE_ATTRIBUTES)
2612 mdb_printf("A");
2613 else
2614 mdb_printf("-");
2615 if (access_mask & ACE_READ_NAMED_ATTRS)
2616 mdb_printf("R");
2617 else
2618 mdb_printf("-");
2619 if (access_mask & ACE_WRITE_NAMED_ATTRS)
2620 mdb_printf("W");
2621 else
2622 mdb_printf("-");
2623 if (access_mask & ACE_READ_ACL)
2624 mdb_printf("c");
2625 else
2626 mdb_printf("-");
2627 if (access_mask & ACE_WRITE_ACL)
2628 mdb_printf("C");
2629 else
2630 mdb_printf("-");
2631 if (access_mask & ACE_WRITE_OWNER)
2632 mdb_printf("o");
2633 else
2634 mdb_printf("-");
2635 if (access_mask & ACE_SYNCHRONIZE)
2636 mdb_printf("s");
2637 else
2638 mdb_printf("-");
2639
2640 mdb_printf(":");
2641
2642 /* Print out inheritance flags */
2643 if (ace_flags & ACE_FILE_INHERIT_ACE)
2644 mdb_printf("f");
2645 else
2646 mdb_printf("-");
2647 if (ace_flags & ACE_DIRECTORY_INHERIT_ACE)
2648 mdb_printf("d");
2649 else
2650 mdb_printf("-");
2651 if (ace_flags & ACE_INHERIT_ONLY_ACE)
2652 mdb_printf("i");
2653 else
2654 mdb_printf("-");
2655 if (ace_flags & ACE_NO_PROPAGATE_INHERIT_ACE)
2656 mdb_printf("n");
2657 else
2658 mdb_printf("-");
2659 if (ace_flags & ACE_SUCCESSFUL_ACCESS_ACE_FLAG)
2660 mdb_printf("S");
2661 else
2662 mdb_printf("-");
2663 if (ace_flags & ACE_FAILED_ACCESS_ACE_FLAG)
2664 mdb_printf("F");
2665 else
2666 mdb_printf("-");
2667 if (ace_flags & ACE_INHERITED_ACE)
2668 mdb_printf("I");
2669 else
2670 mdb_printf("-");
2671
2672 switch (ace_type) {
2673 case ACE_ACCESS_ALLOWED_ACE_TYPE:
2674 mdb_printf(":allow\n");
2675 break;
2676 case ACE_ACCESS_DENIED_ACE_TYPE:
2677 mdb_printf(":deny\n");
2678 break;
2679 case ACE_SYSTEM_AUDIT_ACE_TYPE:
2680 mdb_printf(":audit\n");
2681 break;
2682 case ACE_SYSTEM_ALARM_ACE_TYPE:
2683 mdb_printf(":alarm\n");
2684 break;
2685 default:
2686 mdb_printf(":?\n");
2687 }
2688 return (DCMD_OK);
2689 }
2690
2691 /* ARGSUSED */
2692 static int
2693 zfs_ace_print(uintptr_t addr, uint_t flags, int argc, const mdb_arg_t *argv)
2694 {
2695 zfs_ace_t zace;
2696 int verbose = FALSE;
2697 uint64_t id;
2698
2699 if (!(flags & DCMD_ADDRSPEC))
2700 return (DCMD_USAGE);
2701
2702 if (mdb_getopts(argc, argv,
2703 'v', MDB_OPT_SETBITS, TRUE, &verbose, TRUE, NULL) != argc)
2704 return (DCMD_USAGE);
2705
2706 if (mdb_vread(&zace, sizeof (zfs_ace_t), addr) == -1) {
2707 mdb_warn("failed to read zfs_ace_t");
2708 return (DCMD_ERR);
2709 }
2710
2711 if ((zace.z_hdr.z_flags & ACE_TYPE_FLAGS) == 0 ||
2712 (zace.z_hdr.z_flags & ACE_TYPE_FLAGS) == ACE_IDENTIFIER_GROUP)
2713 id = zace.z_fuid;
2714 else
2715 id = -1;
2716
2717 return (zfs_ace_print_common(addr, flags, id, zace.z_hdr.z_access_mask,
2718 zace.z_hdr.z_flags, zace.z_hdr.z_type, verbose));
2719 }
2720
2721 /* ARGSUSED */
2722 static int
2723 zfs_ace0_print(uintptr_t addr, uint_t flags, int argc, const mdb_arg_t *argv)
2724 {
2725 ace_t ace;
2726 uint64_t id;
2727 int verbose = FALSE;
2728
2729 if (!(flags & DCMD_ADDRSPEC))
2730 return (DCMD_USAGE);
2731
2732 if (mdb_getopts(argc, argv,
2733 'v', MDB_OPT_SETBITS, TRUE, &verbose, TRUE, NULL) != argc)
2734 return (DCMD_USAGE);
2735
2736 if (mdb_vread(&ace, sizeof (ace_t), addr) == -1) {
2737 mdb_warn("failed to read ace_t");
2738 return (DCMD_ERR);
2739 }
2740
2741 if ((ace.a_flags & ACE_TYPE_FLAGS) == 0 ||
2742 (ace.a_flags & ACE_TYPE_FLAGS) == ACE_IDENTIFIER_GROUP)
2743 id = ace.a_who;
2744 else
2745 id = -1;
2746
2747 return (zfs_ace_print_common(addr, flags, id, ace.a_access_mask,
2748 ace.a_flags, ace.a_type, verbose));
2749 }
2750
2751 typedef struct acl_dump_args {
2752 int a_argc;
2753 const mdb_arg_t *a_argv;
2754 uint16_t a_version;
2755 int a_flags;
2756 } acl_dump_args_t;
2757
2758 /* ARGSUSED */
2759 static int
2760 acl_aces_cb(uintptr_t addr, const void *unknown, void *arg)
2761 {
2762 acl_dump_args_t *acl_args = (acl_dump_args_t *)arg;
2763
2764 if (acl_args->a_version == 1) {
2765 if (mdb_call_dcmd("zfs_ace", addr,
2766 DCMD_ADDRSPEC|acl_args->a_flags, acl_args->a_argc,
2767 acl_args->a_argv) != DCMD_OK) {
2768 return (WALK_ERR);
2769 }
2770 } else {
2771 if (mdb_call_dcmd("zfs_ace0", addr,
2772 DCMD_ADDRSPEC|acl_args->a_flags, acl_args->a_argc,
2773 acl_args->a_argv) != DCMD_OK) {
2774 return (WALK_ERR);
2775 }
2776 }
2777 acl_args->a_flags = DCMD_LOOP;
2778 return (WALK_NEXT);
2779 }
2780
2781 /* ARGSUSED */
2782 static int
2783 acl_cb(uintptr_t addr, const void *unknown, void *arg)
2784 {
2785 acl_dump_args_t *acl_args = (acl_dump_args_t *)arg;
2786
2787 if (acl_args->a_version == 1) {
2788 if (mdb_pwalk("zfs_acl_node_aces", acl_aces_cb,
2789 arg, addr) != 0) {
2790 mdb_warn("can't walk ACEs");
2791 return (DCMD_ERR);
2792 }
2793 } else {
2794 if (mdb_pwalk("zfs_acl_node_aces0", acl_aces_cb,
2795 arg, addr) != 0) {
2796 mdb_warn("can't walk ACEs");
2797 return (DCMD_ERR);
2798 }
2799 }
2800 return (WALK_NEXT);
2801 }
2802
2803 /* ARGSUSED */
2804 static int
2805 zfs_acl_dump(uintptr_t addr, uint_t flags, int argc, const mdb_arg_t *argv)
2806 {
2807 zfs_acl_t zacl;
2808 int verbose = FALSE;
2809 acl_dump_args_t acl_args;
2810
2811 if (!(flags & DCMD_ADDRSPEC))
2812 return (DCMD_USAGE);
2813
2814 if (mdb_getopts(argc, argv,
2815 'v', MDB_OPT_SETBITS, TRUE, &verbose, TRUE, NULL) != argc)
2816 return (DCMD_USAGE);
2817
2818 if (mdb_vread(&zacl, sizeof (zfs_acl_t), addr) == -1) {
2819 mdb_warn("failed to read zfs_acl_t");
2820 return (DCMD_ERR);
2821 }
2822
2823 acl_args.a_argc = argc;
2824 acl_args.a_argv = argv;
2825 acl_args.a_version = zacl.z_version;
2826 acl_args.a_flags = DCMD_LOOPFIRST;
2827
2828 if (mdb_pwalk("zfs_acl_node", acl_cb, &acl_args, addr) != 0) {
2829 mdb_warn("can't walk ACL");
2830 return (DCMD_ERR);
2831 }
2832
2833 return (DCMD_OK);
2834 }
2835
2836 /* ARGSUSED */
2837 static int
2838 zfs_acl_node_walk_init(mdb_walk_state_t *wsp)
2839 {
2840 if (wsp->walk_addr == NULL) {
2841 mdb_warn("must supply address of zfs_acl_node_t\n");
2842 return (WALK_ERR);
2843 }
2844
2845 wsp->walk_addr +=
2846 mdb_ctf_offsetof_by_name(ZFS_STRUCT "zfs_acl", "z_acl");
2847
2848 if (mdb_layered_walk("list", wsp) == -1) {
2849 mdb_warn("failed to walk 'list'\n");
2850 return (WALK_ERR);
2851 }
2852
2853 return (WALK_NEXT);
2854 }
2855
2856 static int
2857 zfs_acl_node_walk_step(mdb_walk_state_t *wsp)
2858 {
2859 zfs_acl_node_t aclnode;
2860
2861 if (mdb_vread(&aclnode, sizeof (zfs_acl_node_t),
2862 wsp->walk_addr) == -1) {
2863 mdb_warn("failed to read zfs_acl_node at %p", wsp->walk_addr);
2864 return (WALK_ERR);
2865 }
2866
2867 return (wsp->walk_callback(wsp->walk_addr, &aclnode, wsp->walk_cbdata));
2868 }
2869
2870 typedef struct ace_walk_data {
2871 int ace_count;
2872 int ace_version;
2873 } ace_walk_data_t;
2874
2875 static int
2876 zfs_aces_walk_init_common(mdb_walk_state_t *wsp, int version,
2877 int ace_count, uintptr_t ace_data)
2878 {
2879 ace_walk_data_t *ace_walk_data;
2880
2881 if (wsp->walk_addr == NULL) {
2882 mdb_warn("must supply address of zfs_acl_node_t\n");
2883 return (WALK_ERR);
2884 }
2885
2886 ace_walk_data = mdb_alloc(sizeof (ace_walk_data_t), UM_SLEEP | UM_GC);
2887
2888 ace_walk_data->ace_count = ace_count;
2889 ace_walk_data->ace_version = version;
2890
2891 wsp->walk_addr = ace_data;
2892 wsp->walk_data = ace_walk_data;
2893
2894 return (WALK_NEXT);
2895 }
2896
2897 static int
2898 zfs_acl_node_aces_walk_init_common(mdb_walk_state_t *wsp, int version)
2899 {
2900 static int gotid;
2901 static mdb_ctf_id_t acl_id;
2902 int z_ace_count;
2903 uintptr_t z_acldata;
2904
2905 if (!gotid) {
2906 if (mdb_ctf_lookup_by_name("struct zfs_acl_node",
2907 &acl_id) == -1) {
2908 mdb_warn("couldn't find struct zfs_acl_node");
2909 return (DCMD_ERR);
2910 }
2911 gotid = TRUE;
2912 }
2913
2914 if (GETMEMBID(wsp->walk_addr, &acl_id, z_ace_count, z_ace_count)) {
2915 return (DCMD_ERR);
2916 }
2917 if (GETMEMBID(wsp->walk_addr, &acl_id, z_acldata, z_acldata)) {
2918 return (DCMD_ERR);
2919 }
2920
2921 return (zfs_aces_walk_init_common(wsp, version,
2922 z_ace_count, z_acldata));
2923 }
2924
2925 /* ARGSUSED */
2926 static int
2927 zfs_acl_node_aces_walk_init(mdb_walk_state_t *wsp)
2928 {
2929 return (zfs_acl_node_aces_walk_init_common(wsp, 1));
2930 }
2931
2932 /* ARGSUSED */
2933 static int
2934 zfs_acl_node_aces0_walk_init(mdb_walk_state_t *wsp)
2935 {
2936 return (zfs_acl_node_aces_walk_init_common(wsp, 0));
2937 }
2938
2939 static int
2940 zfs_aces_walk_step(mdb_walk_state_t *wsp)
2941 {
2942 ace_walk_data_t *ace_data = wsp->walk_data;
2943 zfs_ace_t zace;
2944 ace_t *acep;
2945 int status;
2946 int entry_type;
2947 int allow_type;
2948 uintptr_t ptr;
2949
2950 if (ace_data->ace_count == 0)
2951 return (WALK_DONE);
2952
2953 if (mdb_vread(&zace, sizeof (zfs_ace_t), wsp->walk_addr) == -1) {
2954 mdb_warn("failed to read zfs_ace_t at %#lx",
2955 wsp->walk_addr);
2956 return (WALK_ERR);
2957 }
2958
2959 switch (ace_data->ace_version) {
2960 case 0:
2961 acep = (ace_t *)&zace;
2962 entry_type = acep->a_flags & ACE_TYPE_FLAGS;
2963 allow_type = acep->a_type;
2964 break;
2965 case 1:
2966 entry_type = zace.z_hdr.z_flags & ACE_TYPE_FLAGS;
2967 allow_type = zace.z_hdr.z_type;
2968 break;
2969 default:
2970 return (WALK_ERR);
2971 }
2972
2973 ptr = (uintptr_t)wsp->walk_addr;
2974 switch (entry_type) {
2975 case ACE_OWNER:
2976 case ACE_EVERYONE:
2977 case (ACE_IDENTIFIER_GROUP | ACE_GROUP):
2978 ptr += ace_data->ace_version == 0 ?
2979 sizeof (ace_t) : sizeof (zfs_ace_hdr_t);
2980 break;
2981 case ACE_IDENTIFIER_GROUP:
2982 default:
2983 switch (allow_type) {
2984 case ACE_ACCESS_ALLOWED_OBJECT_ACE_TYPE:
2985 case ACE_ACCESS_DENIED_OBJECT_ACE_TYPE:
2986 case ACE_SYSTEM_AUDIT_OBJECT_ACE_TYPE:
2987 case ACE_SYSTEM_ALARM_OBJECT_ACE_TYPE:
2988 ptr += ace_data->ace_version == 0 ?
2989 sizeof (ace_t) : sizeof (zfs_object_ace_t);
2990 break;
2991 default:
2992 ptr += ace_data->ace_version == 0 ?
2993 sizeof (ace_t) : sizeof (zfs_ace_t);
2994 break;
2995 }
2996 }
2997
2998 ace_data->ace_count--;
2999 status = wsp->walk_callback(wsp->walk_addr,
3000 (void *)(uintptr_t)&zace, wsp->walk_cbdata);
3001
3002 wsp->walk_addr = ptr;
3003 return (status);
3004 }
3005
3006 typedef struct mdb_zfs_rrwlock {
3007 kthread_t *rr_writer;
3008 boolean_t rr_writer_wanted;
3009 } mdb_zfs_rrwlock_t;
3010
3011 static uint_t rrw_key;
3012
3013 /* ARGSUSED */
3014 static int
3015 rrwlock(uintptr_t addr, uint_t flags, int argc, const mdb_arg_t *argv)
3016 {
3017 mdb_zfs_rrwlock_t rrw;
3018
3019 if (rrw_key == 0) {
3020 if (mdb_ctf_readsym(&rrw_key, "uint_t", "rrw_tsd_key", 0) == -1)
3021 return (DCMD_ERR);
3022 }
3023
3024 if (mdb_ctf_vread(&rrw, "rrwlock_t", "mdb_zfs_rrwlock_t", addr,
3025 0) == -1)
3026 return (DCMD_ERR);
3027
3028 if (rrw.rr_writer != NULL) {
3029 mdb_printf("write lock held by thread %p\n", rrw.rr_writer);
3030 return (DCMD_OK);
3031 }
3032
3033 if (rrw.rr_writer_wanted) {
3034 mdb_printf("writer wanted\n");
3035 }
3036
3037 mdb_printf("anonymous references:\n");
3038 (void) mdb_call_dcmd("refcount", addr +
3039 mdb_ctf_offsetof_by_name(ZFS_STRUCT "rrwlock", "rr_anon_rcount"),
3040 DCMD_ADDRSPEC, 0, NULL);
3041
3042 mdb_printf("linked references:\n");
3043 (void) mdb_call_dcmd("refcount", addr +
3044 mdb_ctf_offsetof_by_name(ZFS_STRUCT "rrwlock", "rr_linked_rcount"),
3045 DCMD_ADDRSPEC, 0, NULL);
3046
3047 /*
3048 * XXX This should find references from
3049 * "::walk thread | ::tsd -v <rrw_key>", but there is no support
3050 * for programmatic consumption of dcmds, so this would be
3051 * difficult, potentially requiring reimplementing ::tsd (both
3052 * user and kernel versions) in this MDB module.
3053 */
3054
3055 return (DCMD_OK);
3056 }
3057
3058 /*
3059 * MDB module linkage information:
3060 *
3061 * We declare a list of structures describing our dcmds, and a function
3062 * named _mdb_init to return a pointer to our module information.
3063 */
3064
3065 static const mdb_dcmd_t dcmds[] = {
3066 { "arc", "[-bkmg]", "print ARC variables", arc_print },
3067 { "blkptr", ":", "print blkptr_t", blkptr },
3068 { "dbuf", ":", "print dmu_buf_impl_t", dbuf },
3069 { "dbuf_stats", ":", "dbuf stats", dbuf_stats },
3070 { "dbufs",
3071 "\t[-O objset_t*] [-n objset_name | \"mos\"] "
3072 "[-o object | \"mdn\"] \n"
3073 "\t[-l level] [-b blkid | \"bonus\"]",
3074 "find dmu_buf_impl_t's that match specified criteria", dbufs },
3075 { "abuf_find", "dva_word[0] dva_word[1]",
3076 "find arc_buf_hdr_t of a specified DVA",
3077 abuf_find },
3078 { "spa", "?[-cv]", "spa_t summary", spa_print },
3079 { "spa_config", ":", "print spa_t configuration", spa_print_config },
3080 { "spa_verify", ":", "verify spa_t consistency", spa_verify },
3081 { "spa_space", ":[-b]", "print spa_t on-disk space usage", spa_space },
3082 { "spa_vdevs", ":", "given a spa_t, print vdev summary", spa_vdevs },
3083 { "vdev", ":[-re]\n"
3084 "\t-r display recursively\n"
3085 "\t-e print statistics",
3086 "vdev_t summary", vdev_print },
3087 { "zio", ":[cpr]\n"
3088 "\t-c display children\n"
3089 "\t-p display parents\n"
3090 "\t-r display recursively",
3091 "zio_t summary", zio_print },
3092 { "zio_state", "?", "print out all zio_t structures on system or "
3093 "for a particular pool", zio_state },
3094 { "zfs_blkstats", ":[-v]",
3095 "given a spa_t, print block type stats from last scrub",
3096 zfs_blkstats },
3097 { "zfs_params", "", "print zfs tunable parameters", zfs_params },
3098 { "refcount", ":[-r]\n"
3099 "\t-r display recently removed references",
3100 "print refcount_t holders", refcount },
3101 { "zap_leaf", "", "print zap_leaf_phys_t", zap_leaf },
3102 { "zfs_aces", ":[-v]", "print all ACEs from a zfs_acl_t",
3103 zfs_acl_dump },
3104 { "zfs_ace", ":[-v]", "print zfs_ace", zfs_ace_print },
3105 { "zfs_ace0", ":[-v]", "print zfs_ace0", zfs_ace0_print },
3106 { "sa_attr_table", ":", "print SA attribute table from sa_os_t",
3107 sa_attr_table},
3108 { "sa_attr", ": attr_id",
3109 "print SA attribute address when given sa_handle_t", sa_attr_print},
3110 { "zfs_dbgmsg", ":[-va]",
3111 "print zfs debug log", dbgmsg},
3112 { "rrwlock", ":",
3113 "print rrwlock_t, including readers", rrwlock},
3114 { NULL }
3115 };
3116
3117 static const mdb_walker_t walkers[] = {
3118 { "zms_freelist", "walk ZFS metaslab freelist",
3119 freelist_walk_init, freelist_walk_step, NULL },
3120 { "txg_list", "given any txg_list_t *, walk all entries in all txgs",
3121 txg_list_walk_init, txg_list_walk_step, NULL },
3122 { "txg_list0", "given any txg_list_t *, walk all entries in txg 0",
3123 txg_list0_walk_init, txg_list_walk_step, NULL },
3124 { "txg_list1", "given any txg_list_t *, walk all entries in txg 1",
3125 txg_list1_walk_init, txg_list_walk_step, NULL },
3126 { "txg_list2", "given any txg_list_t *, walk all entries in txg 2",
3127 txg_list2_walk_init, txg_list_walk_step, NULL },
3128 { "txg_list3", "given any txg_list_t *, walk all entries in txg 3",
3129 txg_list3_walk_init, txg_list_walk_step, NULL },
3130 { "zio", "walk all zio structures, optionally for a particular spa_t",
3131 zio_walk_init, zio_walk_step, NULL },
3132 { "zio_root",
3133 "walk all root zio_t structures, optionally for a particular spa_t",
3134 zio_walk_init, zio_walk_root_step, NULL },
3135 { "spa", "walk all spa_t entries in the namespace",
3136 spa_walk_init, spa_walk_step, NULL },
3137 { "metaslab", "given a spa_t *, walk all metaslab_t structures",
3138 metaslab_walk_init, metaslab_walk_step, NULL },
3139 { "zfs_acl_node", "given a zfs_acl_t, walk all zfs_acl_nodes",
3140 zfs_acl_node_walk_init, zfs_acl_node_walk_step, NULL },
3141 { "zfs_acl_node_aces", "given a zfs_acl_node_t, walk all ACEs",
3142 zfs_acl_node_aces_walk_init, zfs_aces_walk_step, NULL },
3143 { "zfs_acl_node_aces0",
3144 "given a zfs_acl_node_t, walk all ACEs as ace_t",
3145 zfs_acl_node_aces0_walk_init, zfs_aces_walk_step, NULL },
3146 { NULL }
3147 };
3148
3149 static const mdb_modinfo_t modinfo = {
3150 MDB_API_VERSION, dcmds, walkers
3151 };
3152
3153 const mdb_modinfo_t *
3154 _mdb_init(void)
3155 {
3156 return (&modinfo);
3157 }