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/vdev_impl.h>
40 #include <sys/zap_leaf.h>
41 #include <sys/zap_impl.h>
42 #include <ctype.h>
43 #include <sys/zfs_acl.h>
44 #include <sys/sa_impl.h>
45
46 #ifdef _KERNEL
47 #define ZFS_OBJ_NAME "zfs"
48 extern int64_t mdb_gethrtime(void);
49 #else
50 #define ZFS_OBJ_NAME "libzpool.so.1"
51 #endif
52
53 #define ZFS_STRUCT "struct " ZFS_OBJ_NAME "`"
54
55 #ifndef _KERNEL
56 int aok;
57 #endif
58
59 static int
60 getmember(uintptr_t addr, const char *type, mdb_ctf_id_t *idp,
61 const char *member, int len, void *buf)
62 {
63 mdb_ctf_id_t id;
64 ulong_t off;
65 char name[64];
66
67 if (idp == NULL) {
68 if (mdb_ctf_lookup_by_name(type, &id) == -1) {
69 mdb_warn("couldn't find type %s", type);
70 return (DCMD_ERR);
71 }
72 idp = &id;
73 } else {
74 type = name;
75 mdb_ctf_type_name(*idp, name, sizeof (name));
76 }
77
78 if (mdb_ctf_offsetof(*idp, member, &off) == -1) {
79 mdb_warn("couldn't find member %s of type %s\n", member, type);
80 return (DCMD_ERR);
81 }
82 if (off % 8 != 0) {
83 mdb_warn("member %s of type %s is unsupported bitfield",
84 member, type);
85 return (DCMD_ERR);
86 }
87 off /= 8;
88
89 if (mdb_vread(buf, len, addr + off) == -1) {
90 mdb_warn("failed to read %s from %s at %p",
91 member, type, addr + off);
92 return (DCMD_ERR);
93 }
94 /* mdb_warn("read %s from %s at %p+%llx\n", member, type, addr, off); */
95
96 return (0);
97 }
98
99 #define GETMEMB(addr, structname, member, dest) \
100 getmember(addr, ZFS_STRUCT structname, NULL, #member, \
101 sizeof (dest), &(dest))
102
103 #define GETMEMBID(addr, ctfid, member, dest) \
104 getmember(addr, NULL, ctfid, #member, sizeof (dest), &(dest))
105
106 static boolean_t
107 strisprint(const char *cp)
108 {
109 for (; *cp; cp++) {
110 if (!isprint(*cp))
111 return (B_FALSE);
112 }
113 return (B_TRUE);
114 }
115
116 static int verbose;
117
118 static int
119 freelist_walk_init(mdb_walk_state_t *wsp)
120 {
121 if (wsp->walk_addr == NULL) {
122 mdb_warn("must supply starting address\n");
123 return (WALK_ERR);
124 }
125
126 wsp->walk_data = 0; /* Index into the freelist */
127 return (WALK_NEXT);
128 }
129
130 static int
131 freelist_walk_step(mdb_walk_state_t *wsp)
132 {
133 uint64_t entry;
134 uintptr_t number = (uintptr_t)wsp->walk_data;
135 char *ddata[] = { "ALLOC", "FREE", "CONDENSE", "INVALID",
136 "INVALID", "INVALID", "INVALID", "INVALID" };
137 int mapshift = SPA_MINBLOCKSHIFT;
138
139 if (mdb_vread(&entry, sizeof (entry), wsp->walk_addr) == -1) {
140 mdb_warn("failed to read freelist entry %p", wsp->walk_addr);
141 return (WALK_DONE);
142 }
143 wsp->walk_addr += sizeof (entry);
144 wsp->walk_data = (void *)(number + 1);
145
146 if (SM_DEBUG_DECODE(entry)) {
147 mdb_printf("DEBUG: %3u %10s: txg=%llu pass=%llu\n",
148 number,
149 ddata[SM_DEBUG_ACTION_DECODE(entry)],
150 SM_DEBUG_TXG_DECODE(entry),
151 SM_DEBUG_SYNCPASS_DECODE(entry));
152 } else {
153 mdb_printf("Entry: %3u offsets=%08llx-%08llx type=%c "
154 "size=%06llx", number,
155 SM_OFFSET_DECODE(entry) << mapshift,
156 (SM_OFFSET_DECODE(entry) + SM_RUN_DECODE(entry)) <<
157 mapshift,
158 SM_TYPE_DECODE(entry) == SM_ALLOC ? 'A' : 'F',
159 SM_RUN_DECODE(entry) << mapshift);
160 if (verbose)
161 mdb_printf(" (raw=%012llx)\n", entry);
162 mdb_printf("\n");
163 }
164 return (WALK_NEXT);
165 }
166
167 static int
168 mdb_dsl_dir_name(uintptr_t addr, char *buf)
169 {
170 static int gotid;
171 static mdb_ctf_id_t dd_id;
172 uintptr_t dd_parent;
173 char dd_myname[MAXNAMELEN];
174
175 if (!gotid) {
176 if (mdb_ctf_lookup_by_name(ZFS_STRUCT "dsl_dir",
177 &dd_id) == -1) {
178 mdb_warn("couldn't find struct dsl_dir");
179 return (DCMD_ERR);
180 }
181 gotid = TRUE;
182 }
183 if (GETMEMBID(addr, &dd_id, dd_parent, dd_parent) ||
184 GETMEMBID(addr, &dd_id, dd_myname, dd_myname)) {
185 return (DCMD_ERR);
186 }
187
188 if (dd_parent) {
189 if (mdb_dsl_dir_name(dd_parent, buf))
190 return (DCMD_ERR);
191 strcat(buf, "/");
192 }
193
194 if (dd_myname[0])
195 strcat(buf, dd_myname);
196 else
197 strcat(buf, "???");
198
199 return (0);
200 }
201
202 static int
203 objset_name(uintptr_t addr, char *buf)
204 {
205 static int gotid;
206 static mdb_ctf_id_t os_id, ds_id;
207 uintptr_t os_dsl_dataset;
208 char ds_snapname[MAXNAMELEN];
209 uintptr_t ds_dir;
210
211 buf[0] = '\0';
212
213 if (!gotid) {
214 if (mdb_ctf_lookup_by_name(ZFS_STRUCT "objset",
215 &os_id) == -1) {
216 mdb_warn("couldn't find struct objset");
217 return (DCMD_ERR);
218 }
219 if (mdb_ctf_lookup_by_name(ZFS_STRUCT "dsl_dataset",
220 &ds_id) == -1) {
221 mdb_warn("couldn't find struct dsl_dataset");
222 return (DCMD_ERR);
223 }
224
225 gotid = TRUE;
226 }
227
228 if (GETMEMBID(addr, &os_id, os_dsl_dataset, os_dsl_dataset))
229 return (DCMD_ERR);
230
231 if (os_dsl_dataset == 0) {
232 strcat(buf, "mos");
233 return (0);
234 }
235
236 if (GETMEMBID(os_dsl_dataset, &ds_id, ds_snapname, ds_snapname) ||
237 GETMEMBID(os_dsl_dataset, &ds_id, ds_dir, ds_dir)) {
238 return (DCMD_ERR);
239 }
240
241 if (ds_dir && mdb_dsl_dir_name(ds_dir, buf))
242 return (DCMD_ERR);
243
244 if (ds_snapname[0]) {
245 strcat(buf, "@");
246 strcat(buf, ds_snapname);
247 }
248 return (0);
249 }
250
251 static void
252 enum_lookup(char *out, size_t size, mdb_ctf_id_t id, int val,
253 const char *prefix)
254 {
255 const char *cp;
256 size_t len = strlen(prefix);
257
258 if ((cp = mdb_ctf_enum_name(id, val)) != NULL) {
259 if (strncmp(cp, prefix, len) == 0)
260 cp += len;
261 (void) strncpy(out, cp, size);
262 } else {
263 mdb_snprintf(out, size, "? (%d)", val);
264 }
265 }
266
267 /* ARGSUSED */
268 static int
269 zfs_params(uintptr_t addr, uint_t flags, int argc, const mdb_arg_t *argv)
270 {
271 /*
272 * This table can be approximately generated by running:
273 * egrep "^[a-z0-9_]+ [a-z0-9_]+( =.*)?;" *.c | cut -d ' ' -f 2
274 */
275 static const char *params[] = {
276 "arc_reduce_dnlc_percent",
277 "zfs_arc_max",
278 "zfs_arc_min",
279 "arc_shrink_shift",
280 "zfs_mdcomp_disable",
281 "zfs_prefetch_disable",
282 "zfetch_max_streams",
283 "zfetch_min_sec_reap",
284 "zfetch_block_cap",
285 "zfetch_array_rd_sz",
286 "zfs_default_bs",
287 "zfs_default_ibs",
288 "metaslab_aliquot",
289 "reference_tracking_enable",
290 "reference_history",
291 "spa_max_replication_override",
292 "spa_mode_global",
293 "zfs_flags",
294 "zfs_txg_synctime_ms",
295 "zfs_txg_timeout",
296 "zfs_write_limit_min",
297 "zfs_write_limit_max",
298 "zfs_write_limit_shift",
299 "zfs_write_limit_override",
300 "zfs_no_write_throttle",
301 "zfs_vdev_cache_max",
302 "zfs_vdev_cache_size",
303 "zfs_vdev_cache_bshift",
304 "vdev_mirror_shift",
305 "zfs_vdev_max_pending",
306 "zfs_vdev_min_pending",
307 "zfs_scrub_limit",
308 "zfs_no_scrub_io",
309 "zfs_no_scrub_prefetch",
310 "zfs_vdev_time_shift",
311 "zfs_vdev_ramp_rate",
312 "zfs_vdev_aggregation_limit",
313 "fzap_default_block_shift",
314 "zfs_immediate_write_sz",
315 "zfs_read_chunk_size",
316 "zfs_nocacheflush",
317 "zil_replay_disable",
318 "metaslab_gang_bang",
319 "metaslab_df_alloc_threshold",
320 "metaslab_df_free_pct",
321 "zio_injection_enabled",
322 "zvol_immediate_write_sz",
323 };
324
325 for (int i = 0; i < sizeof (params) / sizeof (params[0]); i++) {
326 int sz;
327 uint64_t val64;
328 uint32_t *val32p = (uint32_t *)&val64;
329
330 sz = mdb_readvar(&val64, params[i]);
331 if (sz == 4) {
332 mdb_printf("%s = 0x%x\n", params[i], *val32p);
333 } else if (sz == 8) {
334 mdb_printf("%s = 0x%llx\n", params[i], val64);
335 } else {
336 mdb_warn("variable %s not found", params[i]);
337 }
338 }
339
340 return (DCMD_OK);
341 }
342
343 /* ARGSUSED */
344 static int
345 blkptr(uintptr_t addr, uint_t flags, int argc, const mdb_arg_t *argv)
346 {
347 mdb_ctf_id_t type_enum, checksum_enum, compress_enum;
348 char type[80], checksum[80], compress[80];
349 blkptr_t blk, *bp = &blk;
350 char buf[BP_SPRINTF_LEN];
351
352 if (mdb_vread(&blk, sizeof (blkptr_t), addr) == -1) {
353 mdb_warn("failed to read blkptr_t");
354 return (DCMD_ERR);
355 }
356
357 if (mdb_ctf_lookup_by_name("enum dmu_object_type", &type_enum) == -1 ||
358 mdb_ctf_lookup_by_name("enum zio_checksum", &checksum_enum) == -1 ||
359 mdb_ctf_lookup_by_name("enum zio_compress", &compress_enum) == -1) {
360 mdb_warn("Could not find blkptr enumerated types");
361 return (DCMD_ERR);
362 }
363
364 enum_lookup(type, sizeof (type), type_enum,
365 BP_GET_TYPE(bp), "DMU_OT_");
366 enum_lookup(checksum, sizeof (checksum), checksum_enum,
367 BP_GET_CHECKSUM(bp), "ZIO_CHECKSUM_");
368 enum_lookup(compress, sizeof (compress), compress_enum,
369 BP_GET_COMPRESS(bp), "ZIO_COMPRESS_");
370
371 SPRINTF_BLKPTR(mdb_snprintf, '\n', buf, bp, type, checksum, compress);
372
373 mdb_printf("%s\n", buf);
374
375 return (DCMD_OK);
376 }
377
378 typedef struct mdb_dmu_buf_impl {
379 struct {
380 uint64_t db_object;
381 } db;
382 uintptr_t db_objset;
383 uint64_t db_level;
384 uint64_t db_blkid;
385 struct {
386 uint64_t rc_count;
387 } db_holds;
388 } mdb_dmu_buf_impl_t;
389
390 /* ARGSUSED */
391 static int
392 dbuf(uintptr_t addr, uint_t flags, int argc, const mdb_arg_t *argv)
393 {
394 mdb_dmu_buf_impl_t db;
395 char objectname[32];
396 char blkidname[32];
397 char path[MAXNAMELEN];
398
399 if (DCMD_HDRSPEC(flags))
400 mdb_printf(" addr object lvl blkid holds os\n");
401
402 if (mdb_ctf_vread(&db, ZFS_STRUCT "dmu_buf_impl", "mdb_dmu_buf_impl_t",
403 addr, 0) == -1)
404 return (DCMD_ERR);
405
406 if (db.db.db_object == DMU_META_DNODE_OBJECT)
407 (void) strcpy(objectname, "mdn");
408 else
409 (void) mdb_snprintf(objectname, sizeof (objectname), "%llx",
410 (u_longlong_t)db.db.db_object);
411
412 if (db.db_blkid == DMU_BONUS_BLKID)
413 (void) strcpy(blkidname, "bonus");
414 else
415 (void) mdb_snprintf(blkidname, sizeof (blkidname), "%llx",
416 (u_longlong_t)db.db_blkid);
417
418 if (objset_name(db.db_objset, path)) {
419 return (DCMD_ERR);
420 }
421
422 mdb_printf("%p %8s %1u %9s %2llu %s\n", addr,
423 objectname, (int)db.db_level, blkidname,
424 db.db_holds.rc_count, path);
425
426 return (DCMD_OK);
427 }
428
429 /* ARGSUSED */
430 static int
431 dbuf_stats(uintptr_t addr, uint_t flags, int argc, const mdb_arg_t *argv)
432 {
433 #define HISTOSZ 32
434 uintptr_t dbp;
435 dmu_buf_impl_t db;
436 dbuf_hash_table_t ht;
437 uint64_t bucket, ndbufs;
438 uint64_t histo[HISTOSZ];
439 uint64_t histo2[HISTOSZ];
440 int i, maxidx;
441
442 if (mdb_readvar(&ht, "dbuf_hash_table") == -1) {
443 mdb_warn("failed to read 'dbuf_hash_table'");
444 return (DCMD_ERR);
445 }
446
447 for (i = 0; i < HISTOSZ; i++) {
448 histo[i] = 0;
449 histo2[i] = 0;
450 }
451
452 ndbufs = 0;
453 for (bucket = 0; bucket < ht.hash_table_mask+1; bucket++) {
454 int len;
455
456 if (mdb_vread(&dbp, sizeof (void *),
457 (uintptr_t)(ht.hash_table+bucket)) == -1) {
458 mdb_warn("failed to read hash bucket %u at %p",
459 bucket, ht.hash_table+bucket);
460 return (DCMD_ERR);
461 }
462
463 len = 0;
464 while (dbp != 0) {
465 if (mdb_vread(&db, sizeof (dmu_buf_impl_t),
466 dbp) == -1) {
467 mdb_warn("failed to read dbuf at %p", dbp);
468 return (DCMD_ERR);
469 }
470 dbp = (uintptr_t)db.db_hash_next;
471 for (i = MIN(len, HISTOSZ - 1); i >= 0; i--)
472 histo2[i]++;
473 len++;
474 ndbufs++;
475 }
476
477 if (len >= HISTOSZ)
478 len = HISTOSZ-1;
479 histo[len]++;
480 }
481
482 mdb_printf("hash table has %llu buckets, %llu dbufs "
483 "(avg %llu buckets/dbuf)\n",
484 ht.hash_table_mask+1, ndbufs,
485 (ht.hash_table_mask+1)/ndbufs);
486
487 mdb_printf("\n");
488 maxidx = 0;
489 for (i = 0; i < HISTOSZ; i++)
490 if (histo[i] > 0)
491 maxidx = i;
492 mdb_printf("hash chain length number of buckets\n");
493 for (i = 0; i <= maxidx; i++)
494 mdb_printf("%u %llu\n", i, histo[i]);
495
496 mdb_printf("\n");
497 maxidx = 0;
498 for (i = 0; i < HISTOSZ; i++)
499 if (histo2[i] > 0)
500 maxidx = i;
501 mdb_printf("hash chain depth number of dbufs\n");
502 for (i = 0; i <= maxidx; i++)
503 mdb_printf("%u or more %llu %llu%%\n",
504 i, histo2[i], histo2[i]*100/ndbufs);
505
506
507 return (DCMD_OK);
508 }
509
510 #define CHAIN_END 0xffff
511 /*
512 * ::zap_leaf [-v]
513 *
514 * Print a zap_leaf_phys_t, assumed to be 16k
515 */
516 /* ARGSUSED */
517 static int
518 zap_leaf(uintptr_t addr, uint_t flags, int argc, const mdb_arg_t *argv)
519 {
520 char buf[16*1024];
521 int verbose = B_FALSE;
522 int four = B_FALSE;
523 zap_leaf_t l;
524 zap_leaf_phys_t *zlp = (void *)buf;
525 int i;
526
527 if (mdb_getopts(argc, argv,
528 'v', MDB_OPT_SETBITS, TRUE, &verbose,
529 '4', MDB_OPT_SETBITS, TRUE, &four,
530 NULL) != argc)
531 return (DCMD_USAGE);
532
533 l.l_phys = zlp;
534 l.l_bs = 14; /* assume 16k blocks */
535 if (four)
536 l.l_bs = 12;
537
538 if (!(flags & DCMD_ADDRSPEC)) {
539 return (DCMD_USAGE);
540 }
541
542 if (mdb_vread(buf, sizeof (buf), addr) == -1) {
543 mdb_warn("failed to read zap_leaf_phys_t at %p", addr);
544 return (DCMD_ERR);
545 }
546
547 if (zlp->l_hdr.lh_block_type != ZBT_LEAF ||
548 zlp->l_hdr.lh_magic != ZAP_LEAF_MAGIC) {
549 mdb_warn("This does not appear to be a zap_leaf_phys_t");
550 return (DCMD_ERR);
551 }
552
553 mdb_printf("zap_leaf_phys_t at %p:\n", addr);
554 mdb_printf(" lh_prefix_len = %u\n", zlp->l_hdr.lh_prefix_len);
555 mdb_printf(" lh_prefix = %llx\n", zlp->l_hdr.lh_prefix);
556 mdb_printf(" lh_nentries = %u\n", zlp->l_hdr.lh_nentries);
557 mdb_printf(" lh_nfree = %u\n", zlp->l_hdr.lh_nfree,
558 zlp->l_hdr.lh_nfree * 100 / (ZAP_LEAF_NUMCHUNKS(&l)));
559 mdb_printf(" lh_freelist = %u\n", zlp->l_hdr.lh_freelist);
560 mdb_printf(" lh_flags = %x (%s)\n", zlp->l_hdr.lh_flags,
561 zlp->l_hdr.lh_flags & ZLF_ENTRIES_CDSORTED ?
562 "ENTRIES_CDSORTED" : "");
563
564 if (verbose) {
565 mdb_printf(" hash table:\n");
566 for (i = 0; i < ZAP_LEAF_HASH_NUMENTRIES(&l); i++) {
567 if (zlp->l_hash[i] != CHAIN_END)
568 mdb_printf(" %u: %u\n", i, zlp->l_hash[i]);
569 }
570 }
571
572 mdb_printf(" chunks:\n");
573 for (i = 0; i < ZAP_LEAF_NUMCHUNKS(&l); i++) {
574 /* LINTED: alignment */
575 zap_leaf_chunk_t *zlc = &ZAP_LEAF_CHUNK(&l, i);
576 switch (zlc->l_entry.le_type) {
577 case ZAP_CHUNK_FREE:
578 if (verbose) {
579 mdb_printf(" %u: free; lf_next = %u\n",
580 i, zlc->l_free.lf_next);
581 }
582 break;
583 case ZAP_CHUNK_ENTRY:
584 mdb_printf(" %u: entry\n", i);
585 if (verbose) {
586 mdb_printf(" le_next = %u\n",
587 zlc->l_entry.le_next);
588 }
589 mdb_printf(" le_name_chunk = %u\n",
590 zlc->l_entry.le_name_chunk);
591 mdb_printf(" le_name_numints = %u\n",
592 zlc->l_entry.le_name_numints);
593 mdb_printf(" le_value_chunk = %u\n",
594 zlc->l_entry.le_value_chunk);
595 mdb_printf(" le_value_intlen = %u\n",
596 zlc->l_entry.le_value_intlen);
597 mdb_printf(" le_value_numints = %u\n",
598 zlc->l_entry.le_value_numints);
599 mdb_printf(" le_cd = %u\n",
600 zlc->l_entry.le_cd);
601 mdb_printf(" le_hash = %llx\n",
602 zlc->l_entry.le_hash);
603 break;
604 case ZAP_CHUNK_ARRAY:
605 mdb_printf(" %u: array", i);
606 if (strisprint((char *)zlc->l_array.la_array))
607 mdb_printf(" \"%s\"", zlc->l_array.la_array);
608 mdb_printf("\n");
609 if (verbose) {
610 int j;
611 mdb_printf(" ");
612 for (j = 0; j < ZAP_LEAF_ARRAY_BYTES; j++) {
613 mdb_printf("%02x ",
614 zlc->l_array.la_array[j]);
615 }
616 mdb_printf("\n");
617 }
618 if (zlc->l_array.la_next != CHAIN_END) {
619 mdb_printf(" lf_next = %u\n",
620 zlc->l_array.la_next);
621 }
622 break;
623 default:
624 mdb_printf(" %u: undefined type %u\n",
625 zlc->l_entry.le_type);
626 }
627 }
628
629 return (DCMD_OK);
630 }
631
632 typedef struct dbufs_data {
633 mdb_ctf_id_t id;
634 uint64_t objset;
635 uint64_t object;
636 uint64_t level;
637 uint64_t blkid;
638 char *osname;
639 } dbufs_data_t;
640
641 #define DBUFS_UNSET (0xbaddcafedeadbeefULL)
642
643 /* ARGSUSED */
644 static int
645 dbufs_cb(uintptr_t addr, const void *unknown, void *arg)
646 {
647 dbufs_data_t *data = arg;
648 uintptr_t objset;
649 dmu_buf_t db;
650 uint8_t level;
651 uint64_t blkid;
652 char osname[MAXNAMELEN];
653
654 if (GETMEMBID(addr, &data->id, db_objset, objset) ||
655 GETMEMBID(addr, &data->id, db, db) ||
656 GETMEMBID(addr, &data->id, db_level, level) ||
657 GETMEMBID(addr, &data->id, db_blkid, blkid)) {
658 return (WALK_ERR);
659 }
660
661 if ((data->objset == DBUFS_UNSET || data->objset == objset) &&
662 (data->osname == NULL || (objset_name(objset, osname) == 0 &&
663 strcmp(data->osname, osname) == 0)) &&
664 (data->object == DBUFS_UNSET || data->object == db.db_object) &&
665 (data->level == DBUFS_UNSET || data->level == level) &&
666 (data->blkid == DBUFS_UNSET || data->blkid == blkid)) {
667 mdb_printf("%#lr\n", addr);
668 }
669 return (WALK_NEXT);
670 }
671
672 /* ARGSUSED */
673 static int
674 dbufs(uintptr_t addr, uint_t flags, int argc, const mdb_arg_t *argv)
675 {
676 dbufs_data_t data;
677 char *object = NULL;
678 char *blkid = NULL;
679
680 data.objset = data.object = data.level = data.blkid = DBUFS_UNSET;
681 data.osname = NULL;
682
683 if (mdb_getopts(argc, argv,
684 'O', MDB_OPT_UINT64, &data.objset,
685 'n', MDB_OPT_STR, &data.osname,
686 'o', MDB_OPT_STR, &object,
687 'l', MDB_OPT_UINT64, &data.level,
688 'b', MDB_OPT_STR, &blkid) != argc) {
689 return (DCMD_USAGE);
690 }
691
692 if (object) {
693 if (strcmp(object, "mdn") == 0) {
694 data.object = DMU_META_DNODE_OBJECT;
695 } else {
696 data.object = mdb_strtoull(object);
697 }
698 }
699
700 if (blkid) {
701 if (strcmp(blkid, "bonus") == 0) {
702 data.blkid = DMU_BONUS_BLKID;
703 } else {
704 data.blkid = mdb_strtoull(blkid);
705 }
706 }
707
708 if (mdb_ctf_lookup_by_name(ZFS_STRUCT "dmu_buf_impl", &data.id) == -1) {
709 mdb_warn("couldn't find struct dmu_buf_impl_t");
710 return (DCMD_ERR);
711 }
712
713 if (mdb_walk("dmu_buf_impl_t", dbufs_cb, &data) != 0) {
714 mdb_warn("can't walk dbufs");
715 return (DCMD_ERR);
716 }
717
718 return (DCMD_OK);
719 }
720
721 typedef struct abuf_find_data {
722 dva_t dva;
723 mdb_ctf_id_t id;
724 } abuf_find_data_t;
725
726 /* ARGSUSED */
727 static int
728 abuf_find_cb(uintptr_t addr, const void *unknown, void *arg)
729 {
730 abuf_find_data_t *data = arg;
731 dva_t dva;
732
733 if (GETMEMBID(addr, &data->id, b_dva, dva)) {
734 return (WALK_ERR);
735 }
736
737 if (dva.dva_word[0] == data->dva.dva_word[0] &&
738 dva.dva_word[1] == data->dva.dva_word[1]) {
739 mdb_printf("%#lr\n", addr);
740 }
741 return (WALK_NEXT);
742 }
743
744 /* ARGSUSED */
745 static int
746 abuf_find(uintptr_t addr, uint_t flags, int argc, const mdb_arg_t *argv)
747 {
748 abuf_find_data_t data;
749 GElf_Sym sym;
750 int i;
751 const char *syms[] = {
752 "ARC_mru",
753 "ARC_mru_ghost",
754 "ARC_mfu",
755 "ARC_mfu_ghost",
756 };
757
758 if (argc != 2)
759 return (DCMD_USAGE);
760
761 for (i = 0; i < 2; i ++) {
762 switch (argv[i].a_type) {
763 case MDB_TYPE_STRING:
764 data.dva.dva_word[i] = mdb_strtoull(argv[i].a_un.a_str);
765 break;
766 case MDB_TYPE_IMMEDIATE:
767 data.dva.dva_word[i] = argv[i].a_un.a_val;
768 break;
769 default:
770 return (DCMD_USAGE);
771 }
772 }
773
774 if (mdb_ctf_lookup_by_name(ZFS_STRUCT "arc_buf_hdr", &data.id) == -1) {
775 mdb_warn("couldn't find struct arc_buf_hdr");
776 return (DCMD_ERR);
777 }
778
779 for (i = 0; i < sizeof (syms) / sizeof (syms[0]); i++) {
780 if (mdb_lookup_by_obj(ZFS_OBJ_NAME, syms[i], &sym)) {
781 mdb_warn("can't find symbol %s", syms[i]);
782 return (DCMD_ERR);
783 }
784
785 if (mdb_pwalk("list", abuf_find_cb, &data, sym.st_value) != 0) {
786 mdb_warn("can't walk %s", syms[i]);
787 return (DCMD_ERR);
788 }
789 }
790
791 return (DCMD_OK);
792 }
793
794
795 typedef struct dbgmsg_arg {
796 boolean_t da_verbose;
797 boolean_t da_address;
798 } dbgmsg_arg_t;
799
800 /* ARGSUSED */
801 static int
802 dbgmsg_cb(uintptr_t addr, const void *unknown, void *arg)
803 {
804 static mdb_ctf_id_t id;
805 static boolean_t gotid;
806 static ulong_t off;
807
808 dbgmsg_arg_t *da = arg;
809 time_t timestamp;
810 char buf[1024];
811
812 if (!gotid) {
813 if (mdb_ctf_lookup_by_name(ZFS_STRUCT "zfs_dbgmsg", &id) ==
814 -1) {
815 mdb_warn("couldn't find struct zfs_dbgmsg");
816 return (WALK_ERR);
817 }
818 gotid = TRUE;
819 if (mdb_ctf_offsetof(id, "zdm_msg", &off) == -1) {
820 mdb_warn("couldn't find zdm_msg");
821 return (WALK_ERR);
822 }
823 off /= 8;
824 }
825
826
827 if (GETMEMBID(addr, &id, zdm_timestamp, timestamp)) {
828 return (WALK_ERR);
829 }
830
831 if (mdb_readstr(buf, sizeof (buf), addr + off) == -1) {
832 mdb_warn("failed to read zdm_msg at %p\n", addr + off);
833 return (DCMD_ERR);
834 }
835
836 if (da->da_address)
837 mdb_printf("%p ", addr);
838 if (da->da_verbose)
839 mdb_printf("%Y ", timestamp);
840
841 mdb_printf("%s\n", buf);
842
843 if (da->da_verbose)
844 (void) mdb_call_dcmd("whatis", addr, DCMD_ADDRSPEC, 0, NULL);
845
846 return (WALK_NEXT);
847 }
848
849 /* ARGSUSED */
850 static int
851 dbgmsg(uintptr_t addr, uint_t flags, int argc, const mdb_arg_t *argv)
852 {
853 GElf_Sym sym;
854 dbgmsg_arg_t da = { 0 };
855
856 if (mdb_getopts(argc, argv,
857 'v', MDB_OPT_SETBITS, B_TRUE, &da.da_verbose,
858 'a', MDB_OPT_SETBITS, B_TRUE, &da.da_address,
859 NULL) != argc)
860 return (DCMD_USAGE);
861
862 if (mdb_lookup_by_obj(ZFS_OBJ_NAME, "zfs_dbgmsgs", &sym)) {
863 mdb_warn("can't find zfs_dbgmsgs");
864 return (DCMD_ERR);
865 }
866
867 if (mdb_pwalk("list", dbgmsg_cb, &da, sym.st_value) != 0) {
868 mdb_warn("can't walk zfs_dbgmsgs");
869 return (DCMD_ERR);
870 }
871
872 return (DCMD_OK);
873 }
874
875 /*ARGSUSED*/
876 static int
877 arc_print(uintptr_t addr, uint_t flags, int argc, const mdb_arg_t *argv)
878 {
879 kstat_named_t *stats;
880 GElf_Sym sym;
881 int nstats, i;
882 uint_t opt_a = FALSE;
883 uint_t opt_b = FALSE;
884 uint_t shift = 0;
885 const char *suffix;
886
887 static const char *bytestats[] = {
888 "p", "c", "c_min", "c_max", "size", "duplicate_buffers_size",
889 "arc_meta_used", "arc_meta_limit", "arc_meta_max",
890 NULL
891 };
892
893 static const char *extras[] = {
894 "arc_no_grow", "arc_tempreserve",
895 NULL
896 };
897
898 if (mdb_lookup_by_obj(ZFS_OBJ_NAME, "arc_stats", &sym) == -1) {
899 mdb_warn("failed to find 'arc_stats'");
900 return (DCMD_ERR);
901 }
902
903 stats = mdb_zalloc(sym.st_size, UM_SLEEP | UM_GC);
904
905 if (mdb_vread(stats, sym.st_size, sym.st_value) == -1) {
906 mdb_warn("couldn't read 'arc_stats' at %p", sym.st_value);
907 return (DCMD_ERR);
908 }
909
910 nstats = sym.st_size / sizeof (kstat_named_t);
911
912 /* NB: -a / opt_a are ignored for backwards compatability */
913 if (mdb_getopts(argc, argv,
914 'a', MDB_OPT_SETBITS, TRUE, &opt_a,
915 'b', MDB_OPT_SETBITS, TRUE, &opt_b,
916 'k', MDB_OPT_SETBITS, 10, &shift,
917 'm', MDB_OPT_SETBITS, 20, &shift,
918 'g', MDB_OPT_SETBITS, 30, &shift,
919 NULL) != argc)
920 return (DCMD_USAGE);
921
922 if (!opt_b && !shift)
923 shift = 20;
924
925 switch (shift) {
926 case 0:
927 suffix = "B";
928 break;
929 case 10:
930 suffix = "KB";
931 break;
932 case 20:
933 suffix = "MB";
934 break;
935 case 30:
936 suffix = "GB";
937 break;
938 default:
939 suffix = "XX";
940 }
941
942 for (i = 0; i < nstats; i++) {
943 int j;
944 boolean_t bytes = B_FALSE;
945
946 for (j = 0; bytestats[j]; j++) {
947 if (strcmp(stats[i].name, bytestats[j]) == 0) {
948 bytes = B_TRUE;
949 break;
950 }
951 }
952
953 if (bytes) {
954 mdb_printf("%-25s = %9llu %s\n", stats[i].name,
955 stats[i].value.ui64 >> shift, suffix);
956 } else {
957 mdb_printf("%-25s = %9llu\n", stats[i].name,
958 stats[i].value.ui64);
959 }
960 }
961
962 for (i = 0; extras[i]; i++) {
963 uint64_t buf;
964
965 if (mdb_lookup_by_obj(ZFS_OBJ_NAME, extras[i], &sym) == -1) {
966 mdb_warn("failed to find '%s'", extras[i]);
967 return (DCMD_ERR);
968 }
969
970 if (sym.st_size != sizeof (uint64_t) &&
971 sym.st_size != sizeof (uint32_t)) {
972 mdb_warn("expected scalar for variable '%s'\n",
973 extras[i]);
974 return (DCMD_ERR);
975 }
976
977 if (mdb_vread(&buf, sym.st_size, sym.st_value) == -1) {
978 mdb_warn("couldn't read '%s'", extras[i]);
979 return (DCMD_ERR);
980 }
981
982 mdb_printf("%-25s = ", extras[i]);
983
984 /* NB: all the 64-bit extras happen to be byte counts */
985 if (sym.st_size == sizeof (uint64_t))
986 mdb_printf("%9llu %s\n", buf >> shift, suffix);
987
988 if (sym.st_size == sizeof (uint32_t))
989 mdb_printf("%9d\n", *((uint32_t *)&buf));
990 }
991 return (DCMD_OK);
992 }
993
994 typedef struct mdb_spa_print {
995 pool_state_t spa_state;
996 char spa_name[MAXNAMELEN];
997 } mdb_spa_print_t;
998
999 /*
1000 * ::spa
1001 *
1002 * -c Print configuration information as well
1003 * -v Print vdev state
1004 * -e Print vdev error stats
1005 *
1006 * Print a summarized spa_t. When given no arguments, prints out a table of all
1007 * active pools on the system.
1008 */
1009 /* ARGSUSED */
1010 static int
1011 spa_print(uintptr_t addr, uint_t flags, int argc, const mdb_arg_t *argv)
1012 {
1013 const char *statetab[] = { "ACTIVE", "EXPORTED", "DESTROYED",
1014 "SPARE", "L2CACHE", "UNINIT", "UNAVAIL", "POTENTIAL" };
1015 const char *state;
1016 int config = FALSE;
1017 int vdevs = FALSE;
1018 int errors = FALSE;
1019
1020 if (mdb_getopts(argc, argv,
1021 'c', MDB_OPT_SETBITS, TRUE, &config,
1022 'v', MDB_OPT_SETBITS, TRUE, &vdevs,
1023 'e', MDB_OPT_SETBITS, TRUE, &errors,
1024 NULL) != argc)
1025 return (DCMD_USAGE);
1026
1027 if (!(flags & DCMD_ADDRSPEC)) {
1028 if (mdb_walk_dcmd("spa", "spa", argc, argv) == -1) {
1029 mdb_warn("can't walk spa");
1030 return (DCMD_ERR);
1031 }
1032
1033 return (DCMD_OK);
1034 }
1035
1036 if (flags & DCMD_PIPE_OUT) {
1037 mdb_printf("%#lr\n", addr);
1038 return (DCMD_OK);
1039 }
1040
1041 if (DCMD_HDRSPEC(flags))
1042 mdb_printf("%<u>%-?s %9s %-*s%</u>\n", "ADDR", "STATE",
1043 sizeof (uintptr_t) == 4 ? 60 : 52, "NAME");
1044
1045 mdb_spa_print_t spa;
1046 if (mdb_ctf_vread(&spa, "spa_t", "mdb_spa_print_t", addr, 0) == -1)
1047 return (DCMD_ERR);
1048
1049 if (spa.spa_state < 0 || spa.spa_state > POOL_STATE_UNAVAIL)
1050 state = "UNKNOWN";
1051 else
1052 state = statetab[spa.spa_state];
1053
1054 mdb_printf("%0?p %9s %s\n", addr, state, spa.spa_name);
1055
1056 if (config) {
1057 mdb_printf("\n");
1058 mdb_inc_indent(4);
1059 if (mdb_call_dcmd("spa_config", addr, flags, 0,
1060 NULL) != DCMD_OK)
1061 return (DCMD_ERR);
1062 mdb_dec_indent(4);
1063 }
1064
1065 if (vdevs || errors) {
1066 mdb_arg_t v;
1067
1068 v.a_type = MDB_TYPE_STRING;
1069 v.a_un.a_str = "-e";
1070
1071 mdb_printf("\n");
1072 mdb_inc_indent(4);
1073 if (mdb_call_dcmd("spa_vdevs", addr, flags, errors ? 1 : 0,
1074 &v) != DCMD_OK)
1075 return (DCMD_ERR);
1076 mdb_dec_indent(4);
1077 }
1078
1079 return (DCMD_OK);
1080 }
1081
1082 typedef struct mdb_spa_config_spa {
1083 uintptr_t spa_config;
1084 } mdb_spa_config_spa_t;
1085
1086 /*
1087 * ::spa_config
1088 *
1089 * Given a spa_t, print the configuration information stored in spa_config.
1090 * Since it's just an nvlist, format it as an indented list of name=value pairs.
1091 * We simply read the value of spa_config and pass off to ::nvlist.
1092 */
1093 /* ARGSUSED */
1094 static int
1095 spa_print_config(uintptr_t addr, uint_t flags, int argc, const mdb_arg_t *argv)
1096 {
1097 mdb_spa_config_spa_t spa;
1098
1099 if (argc != 0 || !(flags & DCMD_ADDRSPEC))
1100 return (DCMD_USAGE);
1101
1102 if (mdb_ctf_vread(&spa, ZFS_STRUCT "spa", "mdb_spa_config_spa_t",
1103 addr, 0) == -1)
1104 return (DCMD_ERR);
1105
1106 if (spa.spa_config == 0) {
1107 mdb_printf("(none)\n");
1108 return (DCMD_OK);
1109 }
1110
1111 return (mdb_call_dcmd("nvlist", spa.spa_config, flags,
1112 0, NULL));
1113 }
1114
1115 /*
1116 * ::vdev
1117 *
1118 * Print out a summarized vdev_t, in the following form:
1119 *
1120 * ADDR STATE AUX DESC
1121 * fffffffbcde23df0 HEALTHY - /dev/dsk/c0t0d0
1122 *
1123 * If '-r' is specified, recursively visit all children.
1124 *
1125 * With '-e', the statistics associated with the vdev are printed as well.
1126 */
1127 static int
1128 do_print_vdev(uintptr_t addr, int flags, int depth, int stats,
1129 int recursive)
1130 {
1131 vdev_t vdev;
1132 char desc[MAXNAMELEN];
1133 int c, children;
1134 uintptr_t *child;
1135 const char *state, *aux;
1136
1137 if (mdb_vread(&vdev, sizeof (vdev), (uintptr_t)addr) == -1) {
1138 mdb_warn("failed to read vdev_t at %p\n", (uintptr_t)addr);
1139 return (DCMD_ERR);
1140 }
1141
1142 if (flags & DCMD_PIPE_OUT) {
1143 mdb_printf("%#lr\n", addr);
1144 } else {
1145 if (vdev.vdev_path != NULL) {
1146 if (mdb_readstr(desc, sizeof (desc),
1147 (uintptr_t)vdev.vdev_path) == -1) {
1148 mdb_warn("failed to read vdev_path at %p\n",
1149 vdev.vdev_path);
1150 return (DCMD_ERR);
1151 }
1152 } else if (vdev.vdev_ops != NULL) {
1153 vdev_ops_t ops;
1154 if (mdb_vread(&ops, sizeof (ops),
1155 (uintptr_t)vdev.vdev_ops) == -1) {
1156 mdb_warn("failed to read vdev_ops at %p\n",
1157 vdev.vdev_ops);
1158 return (DCMD_ERR);
1159 }
1160 (void) strcpy(desc, ops.vdev_op_type);
1161 } else {
1162 (void) strcpy(desc, "<unknown>");
1163 }
1164
1165 if (depth == 0 && DCMD_HDRSPEC(flags))
1166 mdb_printf("%<u>%-?s %-9s %-12s %-*s%</u>\n",
1167 "ADDR", "STATE", "AUX",
1168 sizeof (uintptr_t) == 4 ? 43 : 35,
1169 "DESCRIPTION");
1170
1171 mdb_printf("%0?p ", addr);
1172
1173 switch (vdev.vdev_state) {
1174 case VDEV_STATE_CLOSED:
1175 state = "CLOSED";
1176 break;
1177 case VDEV_STATE_OFFLINE:
1178 state = "OFFLINE";
1179 break;
1180 case VDEV_STATE_CANT_OPEN:
1181 state = "CANT_OPEN";
1182 break;
1183 case VDEV_STATE_DEGRADED:
1184 state = "DEGRADED";
1185 break;
1186 case VDEV_STATE_HEALTHY:
1187 state = "HEALTHY";
1188 break;
1189 case VDEV_STATE_REMOVED:
1190 state = "REMOVED";
1191 break;
1192 case VDEV_STATE_FAULTED:
1193 state = "FAULTED";
1194 break;
1195 default:
1196 state = "UNKNOWN";
1197 break;
1198 }
1199
1200 switch (vdev.vdev_stat.vs_aux) {
1201 case VDEV_AUX_NONE:
1202 aux = "-";
1203 break;
1204 case VDEV_AUX_OPEN_FAILED:
1205 aux = "OPEN_FAILED";
1206 break;
1207 case VDEV_AUX_CORRUPT_DATA:
1208 aux = "CORRUPT_DATA";
1209 break;
1210 case VDEV_AUX_NO_REPLICAS:
1211 aux = "NO_REPLICAS";
1212 break;
1213 case VDEV_AUX_BAD_GUID_SUM:
1214 aux = "BAD_GUID_SUM";
1215 break;
1216 case VDEV_AUX_TOO_SMALL:
1217 aux = "TOO_SMALL";
1218 break;
1219 case VDEV_AUX_BAD_LABEL:
1220 aux = "BAD_LABEL";
1221 break;
1222 case VDEV_AUX_VERSION_NEWER:
1223 aux = "VERS_NEWER";
1224 break;
1225 case VDEV_AUX_VERSION_OLDER:
1226 aux = "VERS_OLDER";
1227 break;
1228 case VDEV_AUX_UNSUP_FEAT:
1229 aux = "UNSUP_FEAT";
1230 break;
1231 case VDEV_AUX_SPARED:
1232 aux = "SPARED";
1233 break;
1234 case VDEV_AUX_ERR_EXCEEDED:
1235 aux = "ERR_EXCEEDED";
1236 break;
1237 case VDEV_AUX_IO_FAILURE:
1238 aux = "IO_FAILURE";
1239 break;
1240 case VDEV_AUX_BAD_LOG:
1241 aux = "BAD_LOG";
1242 break;
1243 case VDEV_AUX_EXTERNAL:
1244 aux = "EXTERNAL";
1245 break;
1246 case VDEV_AUX_SPLIT_POOL:
1247 aux = "SPLIT_POOL";
1248 break;
1249 default:
1250 aux = "UNKNOWN";
1251 break;
1252 }
1253
1254 mdb_printf("%-9s %-12s %*s%s\n", state, aux, depth, "", desc);
1255
1256 if (stats) {
1257 vdev_stat_t *vs = &vdev.vdev_stat;
1258 int i;
1259
1260 mdb_inc_indent(4);
1261 mdb_printf("\n");
1262 mdb_printf("%<u> %12s %12s %12s %12s "
1263 "%12s%</u>\n", "READ", "WRITE", "FREE", "CLAIM",
1264 "IOCTL");
1265 mdb_printf("OPS ");
1266 for (i = 1; i < ZIO_TYPES; i++)
1267 mdb_printf("%11#llx%s", vs->vs_ops[i],
1268 i == ZIO_TYPES - 1 ? "" : " ");
1269 mdb_printf("\n");
1270 mdb_printf("BYTES ");
1271 for (i = 1; i < ZIO_TYPES; i++)
1272 mdb_printf("%11#llx%s", vs->vs_bytes[i],
1273 i == ZIO_TYPES - 1 ? "" : " ");
1274
1275
1276 mdb_printf("\n");
1277 mdb_printf("EREAD %10#llx\n", vs->vs_read_errors);
1278 mdb_printf("EWRITE %10#llx\n", vs->vs_write_errors);
1279 mdb_printf("ECKSUM %10#llx\n",
1280 vs->vs_checksum_errors);
1281 mdb_dec_indent(4);
1282 }
1283
1284 if (stats)
1285 mdb_printf("\n");
1286 }
1287
1288 children = vdev.vdev_children;
1289
1290 if (children == 0 || !recursive)
1291 return (DCMD_OK);
1292
1293 child = mdb_alloc(children * sizeof (void *), UM_SLEEP | UM_GC);
1294 if (mdb_vread(child, children * sizeof (void *),
1295 (uintptr_t)vdev.vdev_child) == -1) {
1296 mdb_warn("failed to read vdev children at %p", vdev.vdev_child);
1297 return (DCMD_ERR);
1298 }
1299
1300 for (c = 0; c < children; c++) {
1301 if (do_print_vdev(child[c], flags, depth + 2, stats,
1302 recursive))
1303 return (DCMD_ERR);
1304 }
1305
1306 return (DCMD_OK);
1307 }
1308
1309 static int
1310 vdev_print(uintptr_t addr, uint_t flags, int argc, const mdb_arg_t *argv)
1311 {
1312 int recursive = FALSE;
1313 int stats = FALSE;
1314 uint64_t depth = 0;
1315
1316 if (mdb_getopts(argc, argv,
1317 'r', MDB_OPT_SETBITS, TRUE, &recursive,
1318 'e', MDB_OPT_SETBITS, TRUE, &stats,
1319 'd', MDB_OPT_UINT64, &depth,
1320 NULL) != argc)
1321 return (DCMD_USAGE);
1322
1323 if (!(flags & DCMD_ADDRSPEC)) {
1324 mdb_warn("no vdev_t address given\n");
1325 return (DCMD_ERR);
1326 }
1327
1328 return (do_print_vdev(addr, flags, (int)depth, stats, recursive));
1329 }
1330
1331 typedef struct metaslab_walk_data {
1332 uint64_t mw_numvdevs;
1333 uintptr_t *mw_vdevs;
1334 int mw_curvdev;
1335 uint64_t mw_nummss;
1336 uintptr_t *mw_mss;
1337 int mw_curms;
1338 } metaslab_walk_data_t;
1339
1340 static int
1341 metaslab_walk_step(mdb_walk_state_t *wsp)
1342 {
1343 metaslab_walk_data_t *mw = wsp->walk_data;
1344 metaslab_t ms;
1345 uintptr_t msp;
1346
1347 if (mw->mw_curvdev >= mw->mw_numvdevs)
1348 return (WALK_DONE);
1349
1350 if (mw->mw_mss == NULL) {
1351 uintptr_t mssp;
1352 uintptr_t vdevp;
1353
1354 ASSERT(mw->mw_curms == 0);
1355 ASSERT(mw->mw_nummss == 0);
1356
1357 vdevp = mw->mw_vdevs[mw->mw_curvdev];
1358 if (GETMEMB(vdevp, "vdev", vdev_ms, mssp) ||
1359 GETMEMB(vdevp, "vdev", vdev_ms_count, mw->mw_nummss)) {
1360 return (WALK_ERR);
1361 }
1362
1363 mw->mw_mss = mdb_alloc(mw->mw_nummss * sizeof (void*),
1364 UM_SLEEP | UM_GC);
1365 if (mdb_vread(mw->mw_mss, mw->mw_nummss * sizeof (void*),
1366 mssp) == -1) {
1367 mdb_warn("failed to read vdev_ms at %p", mssp);
1368 return (WALK_ERR);
1369 }
1370 }
1371
1372 if (mw->mw_curms >= mw->mw_nummss) {
1373 mw->mw_mss = NULL;
1374 mw->mw_curms = 0;
1375 mw->mw_nummss = 0;
1376 mw->mw_curvdev++;
1377 return (WALK_NEXT);
1378 }
1379
1380 msp = mw->mw_mss[mw->mw_curms];
1381 if (mdb_vread(&ms, sizeof (metaslab_t), msp) == -1) {
1382 mdb_warn("failed to read metaslab_t at %p", msp);
1383 return (WALK_ERR);
1384 }
1385
1386 mw->mw_curms++;
1387
1388 return (wsp->walk_callback(msp, &ms, wsp->walk_cbdata));
1389 }
1390
1391 /* ARGSUSED */
1392 static int
1393 metaslab_walk_init(mdb_walk_state_t *wsp)
1394 {
1395 metaslab_walk_data_t *mw;
1396 uintptr_t root_vdevp;
1397 uintptr_t childp;
1398
1399 if (wsp->walk_addr == NULL) {
1400 mdb_warn("must supply address of spa_t\n");
1401 return (WALK_ERR);
1402 }
1403
1404 mw = mdb_zalloc(sizeof (metaslab_walk_data_t), UM_SLEEP | UM_GC);
1405
1406 if (GETMEMB(wsp->walk_addr, "spa", spa_root_vdev, root_vdevp) ||
1407 GETMEMB(root_vdevp, "vdev", vdev_children, mw->mw_numvdevs) ||
1408 GETMEMB(root_vdevp, "vdev", vdev_child, childp)) {
1409 return (DCMD_ERR);
1410 }
1411
1412 mw->mw_vdevs = mdb_alloc(mw->mw_numvdevs * sizeof (void *),
1413 UM_SLEEP | UM_GC);
1414 if (mdb_vread(mw->mw_vdevs, mw->mw_numvdevs * sizeof (void *),
1415 childp) == -1) {
1416 mdb_warn("failed to read root vdev children at %p", childp);
1417 return (DCMD_ERR);
1418 }
1419
1420 wsp->walk_data = mw;
1421
1422 return (WALK_NEXT);
1423 }
1424
1425 typedef struct mdb_spa {
1426 uintptr_t spa_dsl_pool;
1427 uintptr_t spa_root_vdev;
1428 } mdb_spa_t;
1429
1430 typedef struct mdb_dsl_dir {
1431 uintptr_t dd_phys;
1432 int64_t dd_space_towrite[TXG_SIZE];
1433 } mdb_dsl_dir_t;
1434
1435 typedef struct mdb_dsl_dir_phys {
1436 uint64_t dd_used_bytes;
1437 uint64_t dd_compressed_bytes;
1438 uint64_t dd_uncompressed_bytes;
1439 } mdb_dsl_dir_phys_t;
1440
1441 typedef struct mdb_vdev {
1442 uintptr_t vdev_parent;
1443 uintptr_t vdev_ms;
1444 uint64_t vdev_ms_count;
1445 vdev_stat_t vdev_stat;
1446 } mdb_vdev_t;
1447
1448 typedef struct mdb_metaslab {
1449 space_map_t ms_allocmap[TXG_SIZE];
1450 space_map_t ms_freemap[TXG_SIZE];
1451 space_map_t ms_map;
1452 space_map_obj_t ms_smo;
1453 space_map_obj_t ms_smo_syncing;
1454 } mdb_metaslab_t;
1455
1456 typedef struct space_data {
1457 uint64_t ms_allocmap[TXG_SIZE];
1458 uint64_t ms_freemap[TXG_SIZE];
1459 uint64_t ms_map;
1460 uint64_t avail;
1461 uint64_t nowavail;
1462 } space_data_t;
1463
1464 /* ARGSUSED */
1465 static int
1466 space_cb(uintptr_t addr, const void *unknown, void *arg)
1467 {
1468 space_data_t *sd = arg;
1469 mdb_metaslab_t ms;
1470
1471 if (GETMEMB(addr, "metaslab", ms_allocmap, ms.ms_allocmap) ||
1472 GETMEMB(addr, "metaslab", ms_freemap, ms.ms_freemap) ||
1473 GETMEMB(addr, "metaslab", ms_map, ms.ms_map) ||
1474 GETMEMB(addr, "metaslab", ms_smo, ms.ms_smo) ||
1475 GETMEMB(addr, "metaslab", ms_smo_syncing, ms.ms_smo_syncing)) {
1476 return (WALK_ERR);
1477 }
1478
1479 sd->ms_allocmap[0] += ms.ms_allocmap[0].sm_space;
1480 sd->ms_allocmap[1] += ms.ms_allocmap[1].sm_space;
1481 sd->ms_allocmap[2] += ms.ms_allocmap[2].sm_space;
1482 sd->ms_allocmap[3] += ms.ms_allocmap[3].sm_space;
1483 sd->ms_freemap[0] += ms.ms_freemap[0].sm_space;
1484 sd->ms_freemap[1] += ms.ms_freemap[1].sm_space;
1485 sd->ms_freemap[2] += ms.ms_freemap[2].sm_space;
1486 sd->ms_freemap[3] += ms.ms_freemap[3].sm_space;
1487 sd->ms_map += ms.ms_map.sm_space;
1488 sd->avail += ms.ms_map.sm_size - ms.ms_smo.smo_alloc;
1489 sd->nowavail += ms.ms_map.sm_size - ms.ms_smo_syncing.smo_alloc;
1490
1491 return (WALK_NEXT);
1492 }
1493
1494 /*
1495 * ::spa_space [-b]
1496 *
1497 * Given a spa_t, print out it's on-disk space usage and in-core
1498 * estimates of future usage. If -b is given, print space in bytes.
1499 * Otherwise print in megabytes.
1500 */
1501 /* ARGSUSED */
1502 static int
1503 spa_space(uintptr_t addr, uint_t flags, int argc, const mdb_arg_t *argv)
1504 {
1505 mdb_spa_t spa;
1506 uintptr_t dp_root_dir;
1507 mdb_dsl_dir_t dd;
1508 mdb_dsl_dir_phys_t dsp;
1509 uint64_t children;
1510 uintptr_t childaddr;
1511 space_data_t sd;
1512 int shift = 20;
1513 char *suffix = "M";
1514 int bytes = B_FALSE;
1515
1516 if (mdb_getopts(argc, argv, 'b', MDB_OPT_SETBITS, TRUE, &bytes, NULL) !=
1517 argc)
1518 return (DCMD_USAGE);
1519 if (!(flags & DCMD_ADDRSPEC))
1520 return (DCMD_USAGE);
1521
1522 if (bytes) {
1523 shift = 0;
1524 suffix = "";
1525 }
1526
1527 if (GETMEMB(addr, "spa", spa_dsl_pool, spa.spa_dsl_pool) ||
1528 GETMEMB(addr, "spa", spa_root_vdev, spa.spa_root_vdev) ||
1529 GETMEMB(spa.spa_root_vdev, "vdev", vdev_children, children) ||
1530 GETMEMB(spa.spa_root_vdev, "vdev", vdev_child, childaddr) ||
1531 GETMEMB(spa.spa_dsl_pool, "dsl_pool",
1532 dp_root_dir, dp_root_dir) ||
1533 GETMEMB(dp_root_dir, "dsl_dir", dd_phys, dd.dd_phys) ||
1534 GETMEMB(dp_root_dir, "dsl_dir",
1535 dd_space_towrite, dd.dd_space_towrite) ||
1536 GETMEMB(dd.dd_phys, "dsl_dir_phys",
1537 dd_used_bytes, dsp.dd_used_bytes) ||
1538 GETMEMB(dd.dd_phys, "dsl_dir_phys",
1539 dd_compressed_bytes, dsp.dd_compressed_bytes) ||
1540 GETMEMB(dd.dd_phys, "dsl_dir_phys",
1541 dd_uncompressed_bytes, dsp.dd_uncompressed_bytes)) {
1542 return (DCMD_ERR);
1543 }
1544
1545 mdb_printf("dd_space_towrite = %llu%s %llu%s %llu%s %llu%s\n",
1546 dd.dd_space_towrite[0] >> shift, suffix,
1547 dd.dd_space_towrite[1] >> shift, suffix,
1548 dd.dd_space_towrite[2] >> shift, suffix,
1549 dd.dd_space_towrite[3] >> shift, suffix);
1550
1551 mdb_printf("dd_phys.dd_used_bytes = %llu%s\n",
1552 dsp.dd_used_bytes >> shift, suffix);
1553 mdb_printf("dd_phys.dd_compressed_bytes = %llu%s\n",
1554 dsp.dd_compressed_bytes >> shift, suffix);
1555 mdb_printf("dd_phys.dd_uncompressed_bytes = %llu%s\n",
1556 dsp.dd_uncompressed_bytes >> shift, suffix);
1557
1558 bzero(&sd, sizeof (sd));
1559 if (mdb_pwalk("metaslab", space_cb, &sd, addr) != 0) {
1560 mdb_warn("can't walk metaslabs");
1561 return (DCMD_ERR);
1562 }
1563
1564 mdb_printf("ms_allocmap = %llu%s %llu%s %llu%s %llu%s\n",
1565 sd.ms_allocmap[0] >> shift, suffix,
1566 sd.ms_allocmap[1] >> shift, suffix,
1567 sd.ms_allocmap[2] >> shift, suffix,
1568 sd.ms_allocmap[3] >> shift, suffix);
1569 mdb_printf("ms_freemap = %llu%s %llu%s %llu%s %llu%s\n",
1570 sd.ms_freemap[0] >> shift, suffix,
1571 sd.ms_freemap[1] >> shift, suffix,
1572 sd.ms_freemap[2] >> shift, suffix,
1573 sd.ms_freemap[3] >> shift, suffix);
1574 mdb_printf("ms_map = %llu%s\n", sd.ms_map >> shift, suffix);
1575 mdb_printf("last synced avail = %llu%s\n", sd.avail >> shift, suffix);
1576 mdb_printf("current syncing avail = %llu%s\n",
1577 sd.nowavail >> shift, suffix);
1578
1579 return (DCMD_OK);
1580 }
1581
1582 typedef struct mdb_spa_aux_vdev {
1583 int sav_count;
1584 uintptr_t sav_vdevs;
1585 } mdb_spa_aux_vdev_t;
1586
1587 typedef struct mdb_spa_vdevs {
1588 uintptr_t spa_root_vdev;
1589 mdb_spa_aux_vdev_t spa_l2cache;
1590 mdb_spa_aux_vdev_t spa_spares;
1591 } mdb_spa_vdevs_t;
1592
1593 static int
1594 spa_print_aux(mdb_spa_aux_vdev_t *sav, uint_t flags, mdb_arg_t *v,
1595 const char *name)
1596 {
1597 uintptr_t *aux;
1598 size_t len;
1599 int ret, i;
1600
1601 /*
1602 * Iterate over aux vdevs and print those out as well. This is a
1603 * little annoying because we don't have a root vdev to pass to ::vdev.
1604 * Instead, we print a single line and then call it for each child
1605 * vdev.
1606 */
1607 if (sav->sav_count != 0) {
1608 v[1].a_type = MDB_TYPE_STRING;
1609 v[1].a_un.a_str = "-d";
1610 v[2].a_type = MDB_TYPE_IMMEDIATE;
1611 v[2].a_un.a_val = 2;
1612
1613 len = sav->sav_count * sizeof (uintptr_t);
1614 aux = mdb_alloc(len, UM_SLEEP);
1615 if (mdb_vread(aux, len, sav->sav_vdevs) == -1) {
1616 mdb_free(aux, len);
1617 mdb_warn("failed to read l2cache vdevs at %p",
1618 sav->sav_vdevs);
1619 return (DCMD_ERR);
1620 }
1621
1622 mdb_printf("%-?s %-9s %-12s %s\n", "-", "-", "-", name);
1623
1624 for (i = 0; i < sav->sav_count; i++) {
1625 ret = mdb_call_dcmd("vdev", aux[i], flags, 3, v);
1626 if (ret != DCMD_OK) {
1627 mdb_free(aux, len);
1628 return (ret);
1629 }
1630 }
1631
1632 mdb_free(aux, len);
1633 }
1634
1635 return (0);
1636 }
1637
1638 /*
1639 * ::spa_vdevs
1640 *
1641 * -e Include error stats
1642 *
1643 * Print out a summarized list of vdevs for the given spa_t.
1644 * This is accomplished by invoking "::vdev -re" on the root vdev, as well as
1645 * iterating over the cache devices.
1646 */
1647 /* ARGSUSED */
1648 static int
1649 spa_vdevs(uintptr_t addr, uint_t flags, int argc, const mdb_arg_t *argv)
1650 {
1651 mdb_arg_t v[3];
1652 int errors = FALSE;
1653 int ret;
1654
1655 if (mdb_getopts(argc, argv,
1656 'e', MDB_OPT_SETBITS, TRUE, &errors,
1657 NULL) != argc)
1658 return (DCMD_USAGE);
1659
1660 if (!(flags & DCMD_ADDRSPEC))
1661 return (DCMD_USAGE);
1662
1663 mdb_spa_vdevs_t spa;
1664 if (mdb_ctf_vread(&spa, "spa_t", "mdb_spa_vdevs_t", addr, 0) == -1)
1665 return (DCMD_ERR);
1666
1667 /*
1668 * Unitialized spa_t structures can have a NULL root vdev.
1669 */
1670 if (spa.spa_root_vdev == NULL) {
1671 mdb_printf("no associated vdevs\n");
1672 return (DCMD_OK);
1673 }
1674
1675 v[0].a_type = MDB_TYPE_STRING;
1676 v[0].a_un.a_str = errors ? "-re" : "-r";
1677
1678 ret = mdb_call_dcmd("vdev", (uintptr_t)spa.spa_root_vdev,
1679 flags, 1, v);
1680 if (ret != DCMD_OK)
1681 return (ret);
1682
1683 if (spa_print_aux(&spa.spa_l2cache, flags, v, "cache") != 0 ||
1684 spa_print_aux(&spa.spa_spares, flags, v, "spares") != 0)
1685 return (DCMD_ERR);
1686
1687 return (DCMD_OK);
1688 }
1689
1690 /*
1691 * ::zio
1692 *
1693 * Print a summary of zio_t and all its children. This is intended to display a
1694 * zio tree, and hence we only pick the most important pieces of information for
1695 * the main summary. More detailed information can always be found by doing a
1696 * '::print zio' on the underlying zio_t. The columns we display are:
1697 *
1698 * ADDRESS TYPE STAGE WAITER TIME_ELAPSED
1699 *
1700 * The 'address' column is indented by one space for each depth level as we
1701 * descend down the tree.
1702 */
1703
1704 #define ZIO_MAXINDENT 7
1705 #define ZIO_MAXWIDTH (sizeof (uintptr_t) * 2 + ZIO_MAXINDENT)
1706 #define ZIO_WALK_SELF 0
1707 #define ZIO_WALK_CHILD 1
1708 #define ZIO_WALK_PARENT 2
1709
1710 typedef struct zio_print_args {
1711 int zpa_current_depth;
1712 int zpa_min_depth;
1713 int zpa_max_depth;
1714 int zpa_type;
1715 uint_t zpa_flags;
1716 } zio_print_args_t;
1717
1718 typedef struct mdb_zio {
1719 enum zio_type io_type;
1720 enum zio_stage io_stage;
1721 uintptr_t io_waiter;
1722 uintptr_t io_spa;
1723 struct {
1724 struct {
1725 uintptr_t list_next;
1726 } list_head;
1727 } io_parent_list;
1728 int io_error;
1729 } mdb_zio_t;
1730
1731 typedef struct mdb_zio_timestamp {
1732 hrtime_t io_timestamp;
1733 } mdb_zio_timestamp_t;
1734
1735 static int zio_child_cb(uintptr_t addr, const void *unknown, void *arg);
1736
1737 static int
1738 zio_print_cb(uintptr_t addr, zio_print_args_t *zpa)
1739 {
1740 mdb_ctf_id_t type_enum, stage_enum;
1741 int indent = zpa->zpa_current_depth;
1742 const char *type, *stage;
1743 uintptr_t laddr;
1744 mdb_zio_t zio;
1745 mdb_zio_timestamp_t zio_timestamp = { 0 };
1746
1747 if (mdb_ctf_vread(&zio, ZFS_STRUCT "zio", "mdb_zio_t", addr, 0) == -1)
1748 return (WALK_ERR);
1749 (void) mdb_ctf_vread(&zio_timestamp, ZFS_STRUCT "zio",
1750 "mdb_zio_timestamp_t", addr, MDB_CTF_VREAD_QUIET);
1751
1752 if (indent > ZIO_MAXINDENT)
1753 indent = ZIO_MAXINDENT;
1754
1755 if (mdb_ctf_lookup_by_name("enum zio_type", &type_enum) == -1 ||
1756 mdb_ctf_lookup_by_name("enum zio_stage", &stage_enum) == -1) {
1757 mdb_warn("failed to lookup zio enums");
1758 return (WALK_ERR);
1759 }
1760
1761 if ((type = mdb_ctf_enum_name(type_enum, zio.io_type)) != NULL)
1762 type += sizeof ("ZIO_TYPE_") - 1;
1763 else
1764 type = "?";
1765
1766 if (zio.io_error == 0) {
1767 stage = mdb_ctf_enum_name(stage_enum, zio.io_stage);
1768 if (stage != NULL)
1769 stage += sizeof ("ZIO_STAGE_") - 1;
1770 else
1771 stage = "?";
1772 } else {
1773 stage = "FAILED";
1774 }
1775
1776 if (zpa->zpa_current_depth >= zpa->zpa_min_depth) {
1777 if (zpa->zpa_flags & DCMD_PIPE_OUT) {
1778 mdb_printf("%?p\n", addr);
1779 } else {
1780 mdb_printf("%*s%-*p %-5s %-16s ", indent, "",
1781 ZIO_MAXWIDTH - indent, addr, type, stage);
1782 if (zio.io_waiter != 0)
1783 mdb_printf("%-16lx ", zio.io_waiter);
1784 else
1785 mdb_printf("%-16s ", "-");
1786 #ifdef _KERNEL
1787 if (zio_timestamp.io_timestamp != 0) {
1788 mdb_printf("%llums", (mdb_gethrtime() -
1789 zio_timestamp.io_timestamp) /
1790 1000000);
1791 } else {
1792 mdb_printf("%-12s ", "-");
1793 }
1794 #else
1795 mdb_printf("%-12s ", "-");
1796 #endif
1797 mdb_printf("\n");
1798 }
1799 }
1800
1801 if (zpa->zpa_current_depth >= zpa->zpa_max_depth)
1802 return (WALK_NEXT);
1803
1804 if (zpa->zpa_type == ZIO_WALK_PARENT)
1805 laddr = addr + mdb_ctf_offsetof_by_name(ZFS_STRUCT "zio",
1806 "io_parent_list");
1807 else
1808 laddr = addr + mdb_ctf_offsetof_by_name(ZFS_STRUCT "zio",
1809 "io_child_list");
1810
1811 zpa->zpa_current_depth++;
1812 if (mdb_pwalk("list", zio_child_cb, zpa, laddr) != 0) {
1813 mdb_warn("failed to walk zio_t children at %p\n", laddr);
1814 return (WALK_ERR);
1815 }
1816 zpa->zpa_current_depth--;
1817
1818 return (WALK_NEXT);
1819 }
1820
1821 /* ARGSUSED */
1822 static int
1823 zio_child_cb(uintptr_t addr, const void *unknown, void *arg)
1824 {
1825 zio_link_t zl;
1826 uintptr_t ziop;
1827 zio_print_args_t *zpa = arg;
1828
1829 if (mdb_vread(&zl, sizeof (zl), addr) == -1) {
1830 mdb_warn("failed to read zio_link_t at %p", addr);
1831 return (WALK_ERR);
1832 }
1833
1834 if (zpa->zpa_type == ZIO_WALK_PARENT)
1835 ziop = (uintptr_t)zl.zl_parent;
1836 else
1837 ziop = (uintptr_t)zl.zl_child;
1838
1839 return (zio_print_cb(ziop, arg));
1840 }
1841
1842 /* ARGSUSED */
1843 static int
1844 zio_print(uintptr_t addr, uint_t flags, int argc, const mdb_arg_t *argv)
1845 {
1846 zio_print_args_t zpa = { 0 };
1847
1848 if (!(flags & DCMD_ADDRSPEC))
1849 return (DCMD_USAGE);
1850
1851 if (mdb_getopts(argc, argv,
1852 'r', MDB_OPT_SETBITS, INT_MAX, &zpa.zpa_max_depth,
1853 'c', MDB_OPT_SETBITS, ZIO_WALK_CHILD, &zpa.zpa_type,
1854 'p', MDB_OPT_SETBITS, ZIO_WALK_PARENT, &zpa.zpa_type,
1855 NULL) != argc)
1856 return (DCMD_USAGE);
1857
1858 zpa.zpa_flags = flags;
1859 if (zpa.zpa_max_depth != 0) {
1860 if (zpa.zpa_type == ZIO_WALK_SELF)
1861 zpa.zpa_type = ZIO_WALK_CHILD;
1862 } else if (zpa.zpa_type != ZIO_WALK_SELF) {
1863 zpa.zpa_min_depth = 1;
1864 zpa.zpa_max_depth = 1;
1865 }
1866
1867 if (!(flags & DCMD_PIPE_OUT) && DCMD_HDRSPEC(flags)) {
1868 mdb_printf("%<u>%-*s %-5s %-16s %-16s %-12s%</u>\n",
1869 ZIO_MAXWIDTH, "ADDRESS", "TYPE", "STAGE", "WAITER",
1870 "TIME_ELAPSED");
1871 }
1872
1873 if (zio_print_cb(addr, &zpa) != WALK_NEXT)
1874 return (DCMD_ERR);
1875
1876 return (DCMD_OK);
1877 }
1878
1879 /*
1880 * [addr]::zio_state
1881 *
1882 * Print a summary of all zio_t structures on the system, or for a particular
1883 * pool. This is equivalent to '::walk zio_root | ::zio'.
1884 */
1885 /*ARGSUSED*/
1886 static int
1887 zio_state(uintptr_t addr, uint_t flags, int argc, const mdb_arg_t *argv)
1888 {
1889 /*
1890 * MDB will remember the last address of the pipeline, so if we don't
1891 * zero this we'll end up trying to walk zio structures for a
1892 * non-existent spa_t.
1893 */
1894 if (!(flags & DCMD_ADDRSPEC))
1895 addr = 0;
1896
1897 return (mdb_pwalk_dcmd("zio_root", "zio", argc, argv, addr));
1898 }
1899
1900 typedef struct txg_list_walk_data {
1901 uintptr_t lw_head[TXG_SIZE];
1902 int lw_txgoff;
1903 int lw_maxoff;
1904 size_t lw_offset;
1905 void *lw_obj;
1906 } txg_list_walk_data_t;
1907
1908 static int
1909 txg_list_walk_init_common(mdb_walk_state_t *wsp, int txg, int maxoff)
1910 {
1911 txg_list_walk_data_t *lwd;
1912 txg_list_t list;
1913 int i;
1914
1915 lwd = mdb_alloc(sizeof (txg_list_walk_data_t), UM_SLEEP | UM_GC);
1916 if (mdb_vread(&list, sizeof (txg_list_t), wsp->walk_addr) == -1) {
1917 mdb_warn("failed to read txg_list_t at %#lx", wsp->walk_addr);
1918 return (WALK_ERR);
1919 }
1920
1921 for (i = 0; i < TXG_SIZE; i++)
1922 lwd->lw_head[i] = (uintptr_t)list.tl_head[i];
1923 lwd->lw_offset = list.tl_offset;
1924 lwd->lw_obj = mdb_alloc(lwd->lw_offset + sizeof (txg_node_t),
1925 UM_SLEEP | UM_GC);
1926 lwd->lw_txgoff = txg;
1927 lwd->lw_maxoff = maxoff;
1928
1929 wsp->walk_addr = lwd->lw_head[lwd->lw_txgoff];
1930 wsp->walk_data = lwd;
1931
1932 return (WALK_NEXT);
1933 }
1934
1935 static int
1936 txg_list_walk_init(mdb_walk_state_t *wsp)
1937 {
1938 return (txg_list_walk_init_common(wsp, 0, TXG_SIZE-1));
1939 }
1940
1941 static int
1942 txg_list0_walk_init(mdb_walk_state_t *wsp)
1943 {
1944 return (txg_list_walk_init_common(wsp, 0, 0));
1945 }
1946
1947 static int
1948 txg_list1_walk_init(mdb_walk_state_t *wsp)
1949 {
1950 return (txg_list_walk_init_common(wsp, 1, 1));
1951 }
1952
1953 static int
1954 txg_list2_walk_init(mdb_walk_state_t *wsp)
1955 {
1956 return (txg_list_walk_init_common(wsp, 2, 2));
1957 }
1958
1959 static int
1960 txg_list3_walk_init(mdb_walk_state_t *wsp)
1961 {
1962 return (txg_list_walk_init_common(wsp, 3, 3));
1963 }
1964
1965 static int
1966 txg_list_walk_step(mdb_walk_state_t *wsp)
1967 {
1968 txg_list_walk_data_t *lwd = wsp->walk_data;
1969 uintptr_t addr;
1970 txg_node_t *node;
1971 int status;
1972
1973 while (wsp->walk_addr == NULL && lwd->lw_txgoff < lwd->lw_maxoff) {
1974 lwd->lw_txgoff++;
1975 wsp->walk_addr = lwd->lw_head[lwd->lw_txgoff];
1976 }
1977
1978 if (wsp->walk_addr == NULL)
1979 return (WALK_DONE);
1980
1981 addr = wsp->walk_addr - lwd->lw_offset;
1982
1983 if (mdb_vread(lwd->lw_obj,
1984 lwd->lw_offset + sizeof (txg_node_t), addr) == -1) {
1985 mdb_warn("failed to read list element at %#lx", addr);
1986 return (WALK_ERR);
1987 }
1988
1989 status = wsp->walk_callback(addr, lwd->lw_obj, wsp->walk_cbdata);
1990 node = (txg_node_t *)((uintptr_t)lwd->lw_obj + lwd->lw_offset);
1991 wsp->walk_addr = (uintptr_t)node->tn_next[lwd->lw_txgoff];
1992
1993 return (status);
1994 }
1995
1996 /*
1997 * ::walk spa
1998 *
1999 * Walk all named spa_t structures in the namespace. This is nothing more than
2000 * a layered avl walk.
2001 */
2002 static int
2003 spa_walk_init(mdb_walk_state_t *wsp)
2004 {
2005 GElf_Sym sym;
2006
2007 if (wsp->walk_addr != NULL) {
2008 mdb_warn("spa walk only supports global walks\n");
2009 return (WALK_ERR);
2010 }
2011
2012 if (mdb_lookup_by_obj(ZFS_OBJ_NAME, "spa_namespace_avl", &sym) == -1) {
2013 mdb_warn("failed to find symbol 'spa_namespace_avl'");
2014 return (WALK_ERR);
2015 }
2016
2017 wsp->walk_addr = (uintptr_t)sym.st_value;
2018
2019 if (mdb_layered_walk("avl", wsp) == -1) {
2020 mdb_warn("failed to walk 'avl'\n");
2021 return (WALK_ERR);
2022 }
2023
2024 return (WALK_NEXT);
2025 }
2026
2027 static int
2028 spa_walk_step(mdb_walk_state_t *wsp)
2029 {
2030 return (wsp->walk_callback(wsp->walk_addr, NULL, wsp->walk_cbdata));
2031 }
2032
2033 /*
2034 * [addr]::walk zio
2035 *
2036 * Walk all active zio_t structures on the system. This is simply a layered
2037 * walk on top of ::walk zio_cache, with the optional ability to limit the
2038 * structures to a particular pool.
2039 */
2040 static int
2041 zio_walk_init(mdb_walk_state_t *wsp)
2042 {
2043 wsp->walk_data = &wsp->walk_addr;
2044
2045 if (mdb_layered_walk("zio_cache", wsp) == -1) {
2046 mdb_warn("failed to walk 'zio_cache'\n");
2047 return (WALK_ERR);
2048 }
2049
2050 return (WALK_NEXT);
2051 }
2052
2053 static int
2054 zio_walk_step(mdb_walk_state_t *wsp)
2055 {
2056 mdb_zio_t zio;
2057 uintptr_t *spap = wsp->walk_data;
2058
2059 if (mdb_ctf_vread(&zio, ZFS_STRUCT "zio", "mdb_zio_t",
2060 wsp->walk_addr, 0) == -1)
2061 return (WALK_ERR);
2062
2063 if (*spap != 0 && *spap != zio.io_spa)
2064 return (WALK_NEXT);
2065
2066 return (wsp->walk_callback(wsp->walk_addr, &zio, wsp->walk_cbdata));
2067 }
2068
2069 /*
2070 * [addr]::walk zio_root
2071 *
2072 * Walk only root zio_t structures, optionally for a particular spa_t.
2073 */
2074 static int
2075 zio_walk_root_step(mdb_walk_state_t *wsp)
2076 {
2077 mdb_zio_t zio;
2078 uintptr_t *spap = wsp->walk_data;
2079
2080 if (mdb_ctf_vread(&zio, ZFS_STRUCT "zio", "mdb_zio_t",
2081 wsp->walk_addr, 0) == -1)
2082 return (WALK_ERR);
2083
2084 if (*spap != 0 && *spap != zio.io_spa)
2085 return (WALK_NEXT);
2086
2087 /* If the parent list is not empty, ignore */
2088 if (zio.io_parent_list.list_head.list_next !=
2089 wsp->walk_addr +
2090 mdb_ctf_offsetof_by_name(ZFS_STRUCT "zio", "io_parent_list") +
2091 mdb_ctf_offsetof_by_name("struct list", "list_head"))
2092 return (WALK_NEXT);
2093
2094 return (wsp->walk_callback(wsp->walk_addr, &zio, wsp->walk_cbdata));
2095 }
2096
2097 #define NICENUM_BUFLEN 6
2098
2099 static int
2100 snprintfrac(char *buf, int len,
2101 uint64_t numerator, uint64_t denom, int frac_digits)
2102 {
2103 int mul = 1;
2104 int whole, frac, i;
2105
2106 for (i = frac_digits; i; i--)
2107 mul *= 10;
2108 whole = numerator / denom;
2109 frac = mul * numerator / denom - mul * whole;
2110 return (mdb_snprintf(buf, len, "%u.%0*u", whole, frac_digits, frac));
2111 }
2112
2113 static void
2114 mdb_nicenum(uint64_t num, char *buf)
2115 {
2116 uint64_t n = num;
2117 int index = 0;
2118 char *u;
2119
2120 while (n >= 1024) {
2121 n = (n + (1024 / 2)) / 1024; /* Round up or down */
2122 index++;
2123 }
2124
2125 u = &" \0K\0M\0G\0T\0P\0E\0"[index*2];
2126
2127 if (index == 0) {
2128 (void) mdb_snprintf(buf, NICENUM_BUFLEN, "%llu",
2129 (u_longlong_t)n);
2130 } else if (n < 10 && (num & (num - 1)) != 0) {
2131 (void) snprintfrac(buf, NICENUM_BUFLEN,
2132 num, 1ULL << 10 * index, 2);
2133 strcat(buf, u);
2134 } else if (n < 100 && (num & (num - 1)) != 0) {
2135 (void) snprintfrac(buf, NICENUM_BUFLEN,
2136 num, 1ULL << 10 * index, 1);
2137 strcat(buf, u);
2138 } else {
2139 (void) mdb_snprintf(buf, NICENUM_BUFLEN, "%llu%s",
2140 (u_longlong_t)n, u);
2141 }
2142 }
2143
2144 /*
2145 * ::zfs_blkstats
2146 *
2147 * -v print verbose per-level information
2148 *
2149 */
2150 static int
2151 zfs_blkstats(uintptr_t addr, uint_t flags, int argc, const mdb_arg_t *argv)
2152 {
2153 boolean_t verbose = B_FALSE;
2154 zfs_all_blkstats_t stats;
2155 dmu_object_type_t t;
2156 zfs_blkstat_t *tzb;
2157 uint64_t ditto;
2158 dmu_object_type_info_t dmu_ot[DMU_OT_NUMTYPES + 10];
2159 /* +10 in case it grew */
2160
2161 if (mdb_readvar(&dmu_ot, "dmu_ot") == -1) {
2162 mdb_warn("failed to read 'dmu_ot'");
2163 return (DCMD_ERR);
2164 }
2165
2166 if (mdb_getopts(argc, argv,
2167 'v', MDB_OPT_SETBITS, TRUE, &verbose,
2168 NULL) != argc)
2169 return (DCMD_USAGE);
2170
2171 if (!(flags & DCMD_ADDRSPEC))
2172 return (DCMD_USAGE);
2173
2174 if (GETMEMB(addr, "spa", spa_dsl_pool, addr) ||
2175 GETMEMB(addr, "dsl_pool", dp_blkstats, addr) ||
2176 mdb_vread(&stats, sizeof (zfs_all_blkstats_t), addr) == -1) {
2177 mdb_warn("failed to read data at %p;", addr);
2178 mdb_printf("maybe no stats? run \"zpool scrub\" first.");
2179 return (DCMD_ERR);
2180 }
2181
2182 tzb = &stats.zab_type[DN_MAX_LEVELS][DMU_OT_TOTAL];
2183 if (tzb->zb_gangs != 0) {
2184 mdb_printf("Ganged blocks: %llu\n",
2185 (longlong_t)tzb->zb_gangs);
2186 }
2187
2188 ditto = tzb->zb_ditto_2_of_2_samevdev + tzb->zb_ditto_2_of_3_samevdev +
2189 tzb->zb_ditto_3_of_3_samevdev;
2190 if (ditto != 0) {
2191 mdb_printf("Dittoed blocks on same vdev: %llu\n",
2192 (longlong_t)ditto);
2193 }
2194
2195 mdb_printf("\nBlocks\tLSIZE\tPSIZE\tASIZE"
2196 "\t avg\t comp\t%%Total\tType\n");
2197
2198 for (t = 0; t <= DMU_OT_TOTAL; t++) {
2199 char csize[NICENUM_BUFLEN], lsize[NICENUM_BUFLEN];
2200 char psize[NICENUM_BUFLEN], asize[NICENUM_BUFLEN];
2201 char avg[NICENUM_BUFLEN];
2202 char comp[NICENUM_BUFLEN], pct[NICENUM_BUFLEN];
2203 char typename[64];
2204 int l;
2205
2206
2207 if (t == DMU_OT_DEFERRED)
2208 strcpy(typename, "deferred free");
2209 else if (t == DMU_OT_OTHER)
2210 strcpy(typename, "other");
2211 else if (t == DMU_OT_TOTAL)
2212 strcpy(typename, "Total");
2213 else if (mdb_readstr(typename, sizeof (typename),
2214 (uintptr_t)dmu_ot[t].ot_name) == -1) {
2215 mdb_warn("failed to read type name");
2216 return (DCMD_ERR);
2217 }
2218
2219 if (stats.zab_type[DN_MAX_LEVELS][t].zb_asize == 0)
2220 continue;
2221
2222 for (l = -1; l < DN_MAX_LEVELS; l++) {
2223 int level = (l == -1 ? DN_MAX_LEVELS : l);
2224 zfs_blkstat_t *zb = &stats.zab_type[level][t];
2225
2226 if (zb->zb_asize == 0)
2227 continue;
2228
2229 /*
2230 * Don't print each level unless requested.
2231 */
2232 if (!verbose && level != DN_MAX_LEVELS)
2233 continue;
2234
2235 /*
2236 * If all the space is level 0, don't print the
2237 * level 0 separately.
2238 */
2239 if (level == 0 && zb->zb_asize ==
2240 stats.zab_type[DN_MAX_LEVELS][t].zb_asize)
2241 continue;
2242
2243 mdb_nicenum(zb->zb_count, csize);
2244 mdb_nicenum(zb->zb_lsize, lsize);
2245 mdb_nicenum(zb->zb_psize, psize);
2246 mdb_nicenum(zb->zb_asize, asize);
2247 mdb_nicenum(zb->zb_asize / zb->zb_count, avg);
2248 (void) snprintfrac(comp, NICENUM_BUFLEN,
2249 zb->zb_lsize, zb->zb_psize, 2);
2250 (void) snprintfrac(pct, NICENUM_BUFLEN,
2251 100 * zb->zb_asize, tzb->zb_asize, 2);
2252
2253 mdb_printf("%6s\t%5s\t%5s\t%5s\t%5s"
2254 "\t%5s\t%6s\t",
2255 csize, lsize, psize, asize, avg, comp, pct);
2256
2257 if (level == DN_MAX_LEVELS)
2258 mdb_printf("%s\n", typename);
2259 else
2260 mdb_printf(" L%d %s\n",
2261 level, typename);
2262 }
2263 }
2264
2265 return (DCMD_OK);
2266 }
2267
2268 typedef struct mdb_reference {
2269 uintptr_t ref_holder;
2270 uintptr_t ref_removed;
2271 uint64_t ref_number;
2272 } mdb_reference_t;
2273
2274 /* ARGSUSED */
2275 static int
2276 reference_cb(uintptr_t addr, const void *ignored, void *arg)
2277 {
2278 mdb_reference_t ref;
2279 boolean_t holder_is_str = B_FALSE;
2280 char holder_str[128];
2281 boolean_t removed = (boolean_t)arg;
2282
2283 if (mdb_ctf_vread(&ref, "reference_t", "mdb_reference_t", addr,
2284 0) == -1)
2285 return (DCMD_ERR);
2286
2287 if (mdb_readstr(holder_str, sizeof (holder_str),
2288 ref.ref_holder) != -1)
2289 holder_is_str = strisprint(holder_str);
2290
2291 if (removed)
2292 mdb_printf("removed ");
2293 mdb_printf("reference ");
2294 if (ref.ref_number != 1)
2295 mdb_printf("with count=%llu ", ref.ref_number);
2296 mdb_printf("with tag %lx", ref.ref_holder);
2297 if (holder_is_str)
2298 mdb_printf(" \"%s\"", holder_str);
2299 mdb_printf(", held at:\n");
2300
2301 (void) mdb_call_dcmd("whatis", addr, DCMD_ADDRSPEC, 0, NULL);
2302
2303 if (removed) {
2304 mdb_printf("removed at:\n");
2305 (void) mdb_call_dcmd("whatis", ref.ref_removed,
2306 DCMD_ADDRSPEC, 0, NULL);
2307 }
2308
2309 mdb_printf("\n");
2310
2311 return (WALK_NEXT);
2312 }
2313
2314 typedef struct mdb_refcount {
2315 uint64_t rc_count;
2316 } mdb_refcount_t;
2317
2318 typedef struct mdb_refcount_removed {
2319 uint64_t rc_removed_count;
2320 } mdb_refcount_removed_t;
2321
2322 typedef struct mdb_refcount_tracked {
2323 boolean_t rc_tracked;
2324 } mdb_refcount_tracked_t;
2325
2326 /* ARGSUSED */
2327 static int
2328 refcount(uintptr_t addr, uint_t flags, int argc, const mdb_arg_t *argv)
2329 {
2330 mdb_refcount_t rc;
2331 mdb_refcount_removed_t rcr;
2332 mdb_refcount_tracked_t rct;
2333 int off;
2334 boolean_t released = B_FALSE;
2335
2336 if (!(flags & DCMD_ADDRSPEC))
2337 return (DCMD_USAGE);
2338
2339 if (mdb_getopts(argc, argv,
2340 'r', MDB_OPT_SETBITS, B_TRUE, &released,
2341 NULL) != argc)
2342 return (DCMD_USAGE);
2343
2344 if (mdb_ctf_vread(&rc, "refcount_t", "mdb_refcount_t", addr,
2345 0) == -1)
2346 return (DCMD_ERR);
2347
2348 if (mdb_ctf_vread(&rcr, "refcount_t", "mdb_refcount_removed_t", addr,
2349 MDB_CTF_VREAD_QUIET) == -1) {
2350 mdb_printf("refcount_t at %p has %llu holds (untracked)\n",
2351 addr, (longlong_t)rc.rc_count);
2352 return (DCMD_OK);
2353 }
2354
2355 if (mdb_ctf_vread(&rct, "refcount_t", "mdb_refcount_tracked_t", addr,
2356 MDB_CTF_VREAD_QUIET) == -1) {
2357 /* If this is an old target, it might be tracked. */
2358 rct.rc_tracked = B_TRUE;
2359 }
2360
2361 mdb_printf("refcount_t at %p has %llu current holds, "
2362 "%llu recently released holds\n",
2363 addr, (longlong_t)rc.rc_count, (longlong_t)rcr.rc_removed_count);
2364
2365 if (rct.rc_tracked && rc.rc_count > 0)
2366 mdb_printf("current holds:\n");
2367 off = mdb_ctf_offsetof_by_name("refcount_t", "rc_list");
2368 if (off == -1)
2369 return (DCMD_ERR);
2370 mdb_pwalk("list", reference_cb, (void*)B_FALSE, addr + off);
2371
2372 if (released && rcr.rc_removed_count > 0) {
2373 mdb_printf("released holds:\n");
2374
2375 off = mdb_ctf_offsetof_by_name("refcount_t", "rc_removed");
2376 if (off == -1)
2377 return (DCMD_ERR);
2378 mdb_pwalk("list", reference_cb, (void*)B_FALSE, addr + off);
2379 }
2380
2381 return (DCMD_OK);
2382 }
2383
2384 /* ARGSUSED */
2385 static int
2386 sa_attr_table(uintptr_t addr, uint_t flags, int argc, const mdb_arg_t *argv)
2387 {
2388 sa_attr_table_t *table;
2389 sa_os_t sa_os;
2390 char *name;
2391 int i;
2392
2393 if (mdb_vread(&sa_os, sizeof (sa_os_t), addr) == -1) {
2394 mdb_warn("failed to read sa_os at %p", addr);
2395 return (DCMD_ERR);
2396 }
2397
2398 table = mdb_alloc(sizeof (sa_attr_table_t) * sa_os.sa_num_attrs,
2399 UM_SLEEP | UM_GC);
2400 name = mdb_alloc(MAXPATHLEN, UM_SLEEP | UM_GC);
2401
2402 if (mdb_vread(table, sizeof (sa_attr_table_t) * sa_os.sa_num_attrs,
2403 (uintptr_t)sa_os.sa_attr_table) == -1) {
2404 mdb_warn("failed to read sa_os at %p", addr);
2405 return (DCMD_ERR);
2406 }
2407
2408 mdb_printf("%<u>%-10s %-10s %-10s %-10s %s%</u>\n",
2409 "ATTR ID", "REGISTERED", "LENGTH", "BSWAP", "NAME");
2410 for (i = 0; i != sa_os.sa_num_attrs; i++) {
2411 mdb_readstr(name, MAXPATHLEN, (uintptr_t)table[i].sa_name);
2412 mdb_printf("%5x %8x %8x %8x %-s\n",
2413 (int)table[i].sa_attr, (int)table[i].sa_registered,
2414 (int)table[i].sa_length, table[i].sa_byteswap, name);
2415 }
2416
2417 return (DCMD_OK);
2418 }
2419
2420 static int
2421 sa_get_off_table(uintptr_t addr, uint32_t **off_tab, int attr_count)
2422 {
2423 uintptr_t idx_table;
2424
2425 if (GETMEMB(addr, "sa_idx_tab", sa_idx_tab, idx_table)) {
2426 mdb_printf("can't find offset table in sa_idx_tab\n");
2427 return (-1);
2428 }
2429
2430 *off_tab = mdb_alloc(attr_count * sizeof (uint32_t),
2431 UM_SLEEP | UM_GC);
2432
2433 if (mdb_vread(*off_tab,
2434 attr_count * sizeof (uint32_t), idx_table) == -1) {
2435 mdb_warn("failed to attribute offset table %p", idx_table);
2436 return (-1);
2437 }
2438
2439 return (DCMD_OK);
2440 }
2441
2442 /*ARGSUSED*/
2443 static int
2444 sa_attr_print(uintptr_t addr, uint_t flags, int argc, const mdb_arg_t *argv)
2445 {
2446 uint32_t *offset_tab;
2447 int attr_count;
2448 uint64_t attr_id;
2449 uintptr_t attr_addr;
2450 uintptr_t bonus_tab, spill_tab;
2451 uintptr_t db_bonus, db_spill;
2452 uintptr_t os, os_sa;
2453 uintptr_t db_data;
2454
2455 if (argc != 1)
2456 return (DCMD_USAGE);
2457
2458 if (argv[0].a_type == MDB_TYPE_STRING)
2459 attr_id = mdb_strtoull(argv[0].a_un.a_str);
2460 else
2461 return (DCMD_USAGE);
2462
2463 if (GETMEMB(addr, "sa_handle", sa_bonus_tab, bonus_tab) ||
2464 GETMEMB(addr, "sa_handle", sa_spill_tab, spill_tab) ||
2465 GETMEMB(addr, "sa_handle", sa_os, os) ||
2466 GETMEMB(addr, "sa_handle", sa_bonus, db_bonus) ||
2467 GETMEMB(addr, "sa_handle", sa_spill, db_spill)) {
2468 mdb_printf("Can't find necessary information in sa_handle "
2469 "in sa_handle\n");
2470 return (DCMD_ERR);
2471 }
2472
2473 if (GETMEMB(os, "objset", os_sa, os_sa)) {
2474 mdb_printf("Can't find os_sa in objset\n");
2475 return (DCMD_ERR);
2476 }
2477
2478 if (GETMEMB(os_sa, "sa_os", sa_num_attrs, attr_count)) {
2479 mdb_printf("Can't find sa_num_attrs\n");
2480 return (DCMD_ERR);
2481 }
2482
2483 if (attr_id > attr_count) {
2484 mdb_printf("attribute id number is out of range\n");
2485 return (DCMD_ERR);
2486 }
2487
2488 if (bonus_tab) {
2489 if (sa_get_off_table(bonus_tab, &offset_tab,
2490 attr_count) == -1) {
2491 return (DCMD_ERR);
2492 }
2493
2494 if (GETMEMB(db_bonus, "dmu_buf", db_data, db_data)) {
2495 mdb_printf("can't find db_data in bonus dbuf\n");
2496 return (DCMD_ERR);
2497 }
2498 }
2499
2500 if (bonus_tab && !TOC_ATTR_PRESENT(offset_tab[attr_id]) &&
2501 spill_tab == NULL) {
2502 mdb_printf("Attribute does not exist\n");
2503 return (DCMD_ERR);
2504 } else if (!TOC_ATTR_PRESENT(offset_tab[attr_id]) && spill_tab) {
2505 if (sa_get_off_table(spill_tab, &offset_tab,
2506 attr_count) == -1) {
2507 return (DCMD_ERR);
2508 }
2509 if (GETMEMB(db_spill, "dmu_buf", db_data, db_data)) {
2510 mdb_printf("can't find db_data in spill dbuf\n");
2511 return (DCMD_ERR);
2512 }
2513 if (!TOC_ATTR_PRESENT(offset_tab[attr_id])) {
2514 mdb_printf("Attribute does not exist\n");
2515 return (DCMD_ERR);
2516 }
2517 }
2518 attr_addr = db_data + TOC_OFF(offset_tab[attr_id]);
2519 mdb_printf("%p\n", attr_addr);
2520 return (DCMD_OK);
2521 }
2522
2523 /* ARGSUSED */
2524 static int
2525 zfs_ace_print_common(uintptr_t addr, uint_t flags,
2526 uint64_t id, uint32_t access_mask, uint16_t ace_flags,
2527 uint16_t ace_type, int verbose)
2528 {
2529 if (DCMD_HDRSPEC(flags) && !verbose)
2530 mdb_printf("%<u>%-?s %-8s %-8s %-8s %s%</u>\n",
2531 "ADDR", "FLAGS", "MASK", "TYPE", "ID");
2532
2533 if (!verbose) {
2534 mdb_printf("%0?p %-8x %-8x %-8x %-llx\n", addr,
2535 ace_flags, access_mask, ace_type, id);
2536 return (DCMD_OK);
2537 }
2538
2539 switch (ace_flags & ACE_TYPE_FLAGS) {
2540 case ACE_OWNER:
2541 mdb_printf("owner@:");
2542 break;
2543 case (ACE_IDENTIFIER_GROUP | ACE_GROUP):
2544 mdb_printf("group@:");
2545 break;
2546 case ACE_EVERYONE:
2547 mdb_printf("everyone@:");
2548 break;
2549 case ACE_IDENTIFIER_GROUP:
2550 mdb_printf("group:%llx:", (u_longlong_t)id);
2551 break;
2552 case 0: /* User entry */
2553 mdb_printf("user:%llx:", (u_longlong_t)id);
2554 break;
2555 }
2556
2557 /* print out permission mask */
2558 if (access_mask & ACE_READ_DATA)
2559 mdb_printf("r");
2560 else
2561 mdb_printf("-");
2562 if (access_mask & ACE_WRITE_DATA)
2563 mdb_printf("w");
2564 else
2565 mdb_printf("-");
2566 if (access_mask & ACE_EXECUTE)
2567 mdb_printf("x");
2568 else
2569 mdb_printf("-");
2570 if (access_mask & ACE_APPEND_DATA)
2571 mdb_printf("p");
2572 else
2573 mdb_printf("-");
2574 if (access_mask & ACE_DELETE)
2575 mdb_printf("d");
2576 else
2577 mdb_printf("-");
2578 if (access_mask & ACE_DELETE_CHILD)
2579 mdb_printf("D");
2580 else
2581 mdb_printf("-");
2582 if (access_mask & ACE_READ_ATTRIBUTES)
2583 mdb_printf("a");
2584 else
2585 mdb_printf("-");
2586 if (access_mask & ACE_WRITE_ATTRIBUTES)
2587 mdb_printf("A");
2588 else
2589 mdb_printf("-");
2590 if (access_mask & ACE_READ_NAMED_ATTRS)
2591 mdb_printf("R");
2592 else
2593 mdb_printf("-");
2594 if (access_mask & ACE_WRITE_NAMED_ATTRS)
2595 mdb_printf("W");
2596 else
2597 mdb_printf("-");
2598 if (access_mask & ACE_READ_ACL)
2599 mdb_printf("c");
2600 else
2601 mdb_printf("-");
2602 if (access_mask & ACE_WRITE_ACL)
2603 mdb_printf("C");
2604 else
2605 mdb_printf("-");
2606 if (access_mask & ACE_WRITE_OWNER)
2607 mdb_printf("o");
2608 else
2609 mdb_printf("-");
2610 if (access_mask & ACE_SYNCHRONIZE)
2611 mdb_printf("s");
2612 else
2613 mdb_printf("-");
2614
2615 mdb_printf(":");
2616
2617 /* Print out inheritance flags */
2618 if (ace_flags & ACE_FILE_INHERIT_ACE)
2619 mdb_printf("f");
2620 else
2621 mdb_printf("-");
2622 if (ace_flags & ACE_DIRECTORY_INHERIT_ACE)
2623 mdb_printf("d");
2624 else
2625 mdb_printf("-");
2626 if (ace_flags & ACE_INHERIT_ONLY_ACE)
2627 mdb_printf("i");
2628 else
2629 mdb_printf("-");
2630 if (ace_flags & ACE_NO_PROPAGATE_INHERIT_ACE)
2631 mdb_printf("n");
2632 else
2633 mdb_printf("-");
2634 if (ace_flags & ACE_SUCCESSFUL_ACCESS_ACE_FLAG)
2635 mdb_printf("S");
2636 else
2637 mdb_printf("-");
2638 if (ace_flags & ACE_FAILED_ACCESS_ACE_FLAG)
2639 mdb_printf("F");
2640 else
2641 mdb_printf("-");
2642 if (ace_flags & ACE_INHERITED_ACE)
2643 mdb_printf("I");
2644 else
2645 mdb_printf("-");
2646
2647 switch (ace_type) {
2648 case ACE_ACCESS_ALLOWED_ACE_TYPE:
2649 mdb_printf(":allow\n");
2650 break;
2651 case ACE_ACCESS_DENIED_ACE_TYPE:
2652 mdb_printf(":deny\n");
2653 break;
2654 case ACE_SYSTEM_AUDIT_ACE_TYPE:
2655 mdb_printf(":audit\n");
2656 break;
2657 case ACE_SYSTEM_ALARM_ACE_TYPE:
2658 mdb_printf(":alarm\n");
2659 break;
2660 default:
2661 mdb_printf(":?\n");
2662 }
2663 return (DCMD_OK);
2664 }
2665
2666 /* ARGSUSED */
2667 static int
2668 zfs_ace_print(uintptr_t addr, uint_t flags, int argc, const mdb_arg_t *argv)
2669 {
2670 zfs_ace_t zace;
2671 int verbose = FALSE;
2672 uint64_t id;
2673
2674 if (!(flags & DCMD_ADDRSPEC))
2675 return (DCMD_USAGE);
2676
2677 if (mdb_getopts(argc, argv,
2678 'v', MDB_OPT_SETBITS, TRUE, &verbose, TRUE, NULL) != argc)
2679 return (DCMD_USAGE);
2680
2681 if (mdb_vread(&zace, sizeof (zfs_ace_t), addr) == -1) {
2682 mdb_warn("failed to read zfs_ace_t");
2683 return (DCMD_ERR);
2684 }
2685
2686 if ((zace.z_hdr.z_flags & ACE_TYPE_FLAGS) == 0 ||
2687 (zace.z_hdr.z_flags & ACE_TYPE_FLAGS) == ACE_IDENTIFIER_GROUP)
2688 id = zace.z_fuid;
2689 else
2690 id = -1;
2691
2692 return (zfs_ace_print_common(addr, flags, id, zace.z_hdr.z_access_mask,
2693 zace.z_hdr.z_flags, zace.z_hdr.z_type, verbose));
2694 }
2695
2696 /* ARGSUSED */
2697 static int
2698 zfs_ace0_print(uintptr_t addr, uint_t flags, int argc, const mdb_arg_t *argv)
2699 {
2700 ace_t ace;
2701 uint64_t id;
2702 int verbose = FALSE;
2703
2704 if (!(flags & DCMD_ADDRSPEC))
2705 return (DCMD_USAGE);
2706
2707 if (mdb_getopts(argc, argv,
2708 'v', MDB_OPT_SETBITS, TRUE, &verbose, TRUE, NULL) != argc)
2709 return (DCMD_USAGE);
2710
2711 if (mdb_vread(&ace, sizeof (ace_t), addr) == -1) {
2712 mdb_warn("failed to read ace_t");
2713 return (DCMD_ERR);
2714 }
2715
2716 if ((ace.a_flags & ACE_TYPE_FLAGS) == 0 ||
2717 (ace.a_flags & ACE_TYPE_FLAGS) == ACE_IDENTIFIER_GROUP)
2718 id = ace.a_who;
2719 else
2720 id = -1;
2721
2722 return (zfs_ace_print_common(addr, flags, id, ace.a_access_mask,
2723 ace.a_flags, ace.a_type, verbose));
2724 }
2725
2726 typedef struct acl_dump_args {
2727 int a_argc;
2728 const mdb_arg_t *a_argv;
2729 uint16_t a_version;
2730 int a_flags;
2731 } acl_dump_args_t;
2732
2733 /* ARGSUSED */
2734 static int
2735 acl_aces_cb(uintptr_t addr, const void *unknown, void *arg)
2736 {
2737 acl_dump_args_t *acl_args = (acl_dump_args_t *)arg;
2738
2739 if (acl_args->a_version == 1) {
2740 if (mdb_call_dcmd("zfs_ace", addr,
2741 DCMD_ADDRSPEC|acl_args->a_flags, acl_args->a_argc,
2742 acl_args->a_argv) != DCMD_OK) {
2743 return (WALK_ERR);
2744 }
2745 } else {
2746 if (mdb_call_dcmd("zfs_ace0", addr,
2747 DCMD_ADDRSPEC|acl_args->a_flags, acl_args->a_argc,
2748 acl_args->a_argv) != DCMD_OK) {
2749 return (WALK_ERR);
2750 }
2751 }
2752 acl_args->a_flags = DCMD_LOOP;
2753 return (WALK_NEXT);
2754 }
2755
2756 /* ARGSUSED */
2757 static int
2758 acl_cb(uintptr_t addr, const void *unknown, void *arg)
2759 {
2760 acl_dump_args_t *acl_args = (acl_dump_args_t *)arg;
2761
2762 if (acl_args->a_version == 1) {
2763 if (mdb_pwalk("zfs_acl_node_aces", acl_aces_cb,
2764 arg, addr) != 0) {
2765 mdb_warn("can't walk ACEs");
2766 return (DCMD_ERR);
2767 }
2768 } else {
2769 if (mdb_pwalk("zfs_acl_node_aces0", acl_aces_cb,
2770 arg, addr) != 0) {
2771 mdb_warn("can't walk ACEs");
2772 return (DCMD_ERR);
2773 }
2774 }
2775 return (WALK_NEXT);
2776 }
2777
2778 /* ARGSUSED */
2779 static int
2780 zfs_acl_dump(uintptr_t addr, uint_t flags, int argc, const mdb_arg_t *argv)
2781 {
2782 zfs_acl_t zacl;
2783 int verbose = FALSE;
2784 acl_dump_args_t acl_args;
2785
2786 if (!(flags & DCMD_ADDRSPEC))
2787 return (DCMD_USAGE);
2788
2789 if (mdb_getopts(argc, argv,
2790 'v', MDB_OPT_SETBITS, TRUE, &verbose, TRUE, NULL) != argc)
2791 return (DCMD_USAGE);
2792
2793 if (mdb_vread(&zacl, sizeof (zfs_acl_t), addr) == -1) {
2794 mdb_warn("failed to read zfs_acl_t");
2795 return (DCMD_ERR);
2796 }
2797
2798 acl_args.a_argc = argc;
2799 acl_args.a_argv = argv;
2800 acl_args.a_version = zacl.z_version;
2801 acl_args.a_flags = DCMD_LOOPFIRST;
2802
2803 if (mdb_pwalk("zfs_acl_node", acl_cb, &acl_args, addr) != 0) {
2804 mdb_warn("can't walk ACL");
2805 return (DCMD_ERR);
2806 }
2807
2808 return (DCMD_OK);
2809 }
2810
2811 /* ARGSUSED */
2812 static int
2813 zfs_acl_node_walk_init(mdb_walk_state_t *wsp)
2814 {
2815 if (wsp->walk_addr == NULL) {
2816 mdb_warn("must supply address of zfs_acl_node_t\n");
2817 return (WALK_ERR);
2818 }
2819
2820 wsp->walk_addr +=
2821 mdb_ctf_offsetof_by_name(ZFS_STRUCT "zfs_acl", "z_acl");
2822
2823 if (mdb_layered_walk("list", wsp) == -1) {
2824 mdb_warn("failed to walk 'list'\n");
2825 return (WALK_ERR);
2826 }
2827
2828 return (WALK_NEXT);
2829 }
2830
2831 static int
2832 zfs_acl_node_walk_step(mdb_walk_state_t *wsp)
2833 {
2834 zfs_acl_node_t aclnode;
2835
2836 if (mdb_vread(&aclnode, sizeof (zfs_acl_node_t),
2837 wsp->walk_addr) == -1) {
2838 mdb_warn("failed to read zfs_acl_node at %p", wsp->walk_addr);
2839 return (WALK_ERR);
2840 }
2841
2842 return (wsp->walk_callback(wsp->walk_addr, &aclnode, wsp->walk_cbdata));
2843 }
2844
2845 typedef struct ace_walk_data {
2846 int ace_count;
2847 int ace_version;
2848 } ace_walk_data_t;
2849
2850 static int
2851 zfs_aces_walk_init_common(mdb_walk_state_t *wsp, int version,
2852 int ace_count, uintptr_t ace_data)
2853 {
2854 ace_walk_data_t *ace_walk_data;
2855
2856 if (wsp->walk_addr == NULL) {
2857 mdb_warn("must supply address of zfs_acl_node_t\n");
2858 return (WALK_ERR);
2859 }
2860
2861 ace_walk_data = mdb_alloc(sizeof (ace_walk_data_t), UM_SLEEP | UM_GC);
2862
2863 ace_walk_data->ace_count = ace_count;
2864 ace_walk_data->ace_version = version;
2865
2866 wsp->walk_addr = ace_data;
2867 wsp->walk_data = ace_walk_data;
2868
2869 return (WALK_NEXT);
2870 }
2871
2872 static int
2873 zfs_acl_node_aces_walk_init_common(mdb_walk_state_t *wsp, int version)
2874 {
2875 static int gotid;
2876 static mdb_ctf_id_t acl_id;
2877 int z_ace_count;
2878 uintptr_t z_acldata;
2879
2880 if (!gotid) {
2881 if (mdb_ctf_lookup_by_name("struct zfs_acl_node",
2882 &acl_id) == -1) {
2883 mdb_warn("couldn't find struct zfs_acl_node");
2884 return (DCMD_ERR);
2885 }
2886 gotid = TRUE;
2887 }
2888
2889 if (GETMEMBID(wsp->walk_addr, &acl_id, z_ace_count, z_ace_count)) {
2890 return (DCMD_ERR);
2891 }
2892 if (GETMEMBID(wsp->walk_addr, &acl_id, z_acldata, z_acldata)) {
2893 return (DCMD_ERR);
2894 }
2895
2896 return (zfs_aces_walk_init_common(wsp, version,
2897 z_ace_count, z_acldata));
2898 }
2899
2900 /* ARGSUSED */
2901 static int
2902 zfs_acl_node_aces_walk_init(mdb_walk_state_t *wsp)
2903 {
2904 return (zfs_acl_node_aces_walk_init_common(wsp, 1));
2905 }
2906
2907 /* ARGSUSED */
2908 static int
2909 zfs_acl_node_aces0_walk_init(mdb_walk_state_t *wsp)
2910 {
2911 return (zfs_acl_node_aces_walk_init_common(wsp, 0));
2912 }
2913
2914 static int
2915 zfs_aces_walk_step(mdb_walk_state_t *wsp)
2916 {
2917 ace_walk_data_t *ace_data = wsp->walk_data;
2918 zfs_ace_t zace;
2919 ace_t *acep;
2920 int status;
2921 int entry_type;
2922 int allow_type;
2923 uintptr_t ptr;
2924
2925 if (ace_data->ace_count == 0)
2926 return (WALK_DONE);
2927
2928 if (mdb_vread(&zace, sizeof (zfs_ace_t), wsp->walk_addr) == -1) {
2929 mdb_warn("failed to read zfs_ace_t at %#lx",
2930 wsp->walk_addr);
2931 return (WALK_ERR);
2932 }
2933
2934 switch (ace_data->ace_version) {
2935 case 0:
2936 acep = (ace_t *)&zace;
2937 entry_type = acep->a_flags & ACE_TYPE_FLAGS;
2938 allow_type = acep->a_type;
2939 break;
2940 case 1:
2941 entry_type = zace.z_hdr.z_flags & ACE_TYPE_FLAGS;
2942 allow_type = zace.z_hdr.z_type;
2943 break;
2944 default:
2945 return (WALK_ERR);
2946 }
2947
2948 ptr = (uintptr_t)wsp->walk_addr;
2949 switch (entry_type) {
2950 case ACE_OWNER:
2951 case ACE_EVERYONE:
2952 case (ACE_IDENTIFIER_GROUP | ACE_GROUP):
2953 ptr += ace_data->ace_version == 0 ?
2954 sizeof (ace_t) : sizeof (zfs_ace_hdr_t);
2955 break;
2956 case ACE_IDENTIFIER_GROUP:
2957 default:
2958 switch (allow_type) {
2959 case ACE_ACCESS_ALLOWED_OBJECT_ACE_TYPE:
2960 case ACE_ACCESS_DENIED_OBJECT_ACE_TYPE:
2961 case ACE_SYSTEM_AUDIT_OBJECT_ACE_TYPE:
2962 case ACE_SYSTEM_ALARM_OBJECT_ACE_TYPE:
2963 ptr += ace_data->ace_version == 0 ?
2964 sizeof (ace_t) : sizeof (zfs_object_ace_t);
2965 break;
2966 default:
2967 ptr += ace_data->ace_version == 0 ?
2968 sizeof (ace_t) : sizeof (zfs_ace_t);
2969 break;
2970 }
2971 }
2972
2973 ace_data->ace_count--;
2974 status = wsp->walk_callback(wsp->walk_addr,
2975 (void *)(uintptr_t)&zace, wsp->walk_cbdata);
2976
2977 wsp->walk_addr = ptr;
2978 return (status);
2979 }
2980
2981 typedef struct mdb_zfs_rrwlock {
2982 uintptr_t rr_writer;
2983 boolean_t rr_writer_wanted;
2984 } mdb_zfs_rrwlock_t;
2985
2986 static uint_t rrw_key;
2987
2988 /* ARGSUSED */
2989 static int
2990 rrwlock(uintptr_t addr, uint_t flags, int argc, const mdb_arg_t *argv)
2991 {
2992 mdb_zfs_rrwlock_t rrw;
2993
2994 if (rrw_key == 0) {
2995 if (mdb_ctf_readsym(&rrw_key, "uint_t", "rrw_tsd_key", 0) == -1)
2996 return (DCMD_ERR);
2997 }
2998
2999 if (mdb_ctf_vread(&rrw, "rrwlock_t", "mdb_zfs_rrwlock_t", addr,
3000 0) == -1)
3001 return (DCMD_ERR);
3002
3003 if (rrw.rr_writer != 0) {
3004 mdb_printf("write lock held by thread %lx\n", rrw.rr_writer);
3005 return (DCMD_OK);
3006 }
3007
3008 if (rrw.rr_writer_wanted) {
3009 mdb_printf("writer wanted\n");
3010 }
3011
3012 mdb_printf("anonymous references:\n");
3013 (void) mdb_call_dcmd("refcount", addr +
3014 mdb_ctf_offsetof_by_name(ZFS_STRUCT "rrwlock", "rr_anon_rcount"),
3015 DCMD_ADDRSPEC, 0, NULL);
3016
3017 mdb_printf("linked references:\n");
3018 (void) mdb_call_dcmd("refcount", addr +
3019 mdb_ctf_offsetof_by_name(ZFS_STRUCT "rrwlock", "rr_linked_rcount"),
3020 DCMD_ADDRSPEC, 0, NULL);
3021
3022 /*
3023 * XXX This should find references from
3024 * "::walk thread | ::tsd -v <rrw_key>", but there is no support
3025 * for programmatic consumption of dcmds, so this would be
3026 * difficult, potentially requiring reimplementing ::tsd (both
3027 * user and kernel versions) in this MDB module.
3028 */
3029
3030 return (DCMD_OK);
3031 }
3032
3033 /*
3034 * MDB module linkage information:
3035 *
3036 * We declare a list of structures describing our dcmds, and a function
3037 * named _mdb_init to return a pointer to our module information.
3038 */
3039
3040 static const mdb_dcmd_t dcmds[] = {
3041 { "arc", "[-bkmg]", "print ARC variables", arc_print },
3042 { "blkptr", ":", "print blkptr_t", blkptr },
3043 { "dbuf", ":", "print dmu_buf_impl_t", dbuf },
3044 { "dbuf_stats", ":", "dbuf stats", dbuf_stats },
3045 { "dbufs",
3046 "\t[-O objset_t*] [-n objset_name | \"mos\"] "
3047 "[-o object | \"mdn\"] \n"
3048 "\t[-l level] [-b blkid | \"bonus\"]",
3049 "find dmu_buf_impl_t's that match specified criteria", dbufs },
3050 { "abuf_find", "dva_word[0] dva_word[1]",
3051 "find arc_buf_hdr_t of a specified DVA",
3052 abuf_find },
3053 { "spa", "?[-cv]", "spa_t summary", spa_print },
3054 { "spa_config", ":", "print spa_t configuration", spa_print_config },
3055 { "spa_space", ":[-b]", "print spa_t on-disk space usage", spa_space },
3056 { "spa_vdevs", ":", "given a spa_t, print vdev summary", spa_vdevs },
3057 { "vdev", ":[-re]\n"
3058 "\t-r display recursively\n"
3059 "\t-e print statistics",
3060 "vdev_t summary", vdev_print },
3061 { "zio", ":[cpr]\n"
3062 "\t-c display children\n"
3063 "\t-p display parents\n"
3064 "\t-r display recursively",
3065 "zio_t summary", zio_print },
3066 { "zio_state", "?", "print out all zio_t structures on system or "
3067 "for a particular pool", zio_state },
3068 { "zfs_blkstats", ":[-v]",
3069 "given a spa_t, print block type stats from last scrub",
3070 zfs_blkstats },
3071 { "zfs_params", "", "print zfs tunable parameters", zfs_params },
3072 { "refcount", ":[-r]\n"
3073 "\t-r display recently removed references",
3074 "print refcount_t holders", refcount },
3075 { "zap_leaf", "", "print zap_leaf_phys_t", zap_leaf },
3076 { "zfs_aces", ":[-v]", "print all ACEs from a zfs_acl_t",
3077 zfs_acl_dump },
3078 { "zfs_ace", ":[-v]", "print zfs_ace", zfs_ace_print },
3079 { "zfs_ace0", ":[-v]", "print zfs_ace0", zfs_ace0_print },
3080 { "sa_attr_table", ":", "print SA attribute table from sa_os_t",
3081 sa_attr_table},
3082 { "sa_attr", ": attr_id",
3083 "print SA attribute address when given sa_handle_t", sa_attr_print},
3084 { "zfs_dbgmsg", ":[-va]",
3085 "print zfs debug log", dbgmsg},
3086 { "rrwlock", ":",
3087 "print rrwlock_t, including readers", rrwlock},
3088 { NULL }
3089 };
3090
3091 static const mdb_walker_t walkers[] = {
3092 { "zms_freelist", "walk ZFS metaslab freelist",
3093 freelist_walk_init, freelist_walk_step, NULL },
3094 { "txg_list", "given any txg_list_t *, walk all entries in all txgs",
3095 txg_list_walk_init, txg_list_walk_step, NULL },
3096 { "txg_list0", "given any txg_list_t *, walk all entries in txg 0",
3097 txg_list0_walk_init, txg_list_walk_step, NULL },
3098 { "txg_list1", "given any txg_list_t *, walk all entries in txg 1",
3099 txg_list1_walk_init, txg_list_walk_step, NULL },
3100 { "txg_list2", "given any txg_list_t *, walk all entries in txg 2",
3101 txg_list2_walk_init, txg_list_walk_step, NULL },
3102 { "txg_list3", "given any txg_list_t *, walk all entries in txg 3",
3103 txg_list3_walk_init, txg_list_walk_step, NULL },
3104 { "zio", "walk all zio structures, optionally for a particular spa_t",
3105 zio_walk_init, zio_walk_step, NULL },
3106 { "zio_root",
3107 "walk all root zio_t structures, optionally for a particular spa_t",
3108 zio_walk_init, zio_walk_root_step, NULL },
3109 { "spa", "walk all spa_t entries in the namespace",
3110 spa_walk_init, spa_walk_step, NULL },
3111 { "metaslab", "given a spa_t *, walk all metaslab_t structures",
3112 metaslab_walk_init, metaslab_walk_step, NULL },
3113 { "zfs_acl_node", "given a zfs_acl_t, walk all zfs_acl_nodes",
3114 zfs_acl_node_walk_init, zfs_acl_node_walk_step, NULL },
3115 { "zfs_acl_node_aces", "given a zfs_acl_node_t, walk all ACEs",
3116 zfs_acl_node_aces_walk_init, zfs_aces_walk_step, NULL },
3117 { "zfs_acl_node_aces0",
3118 "given a zfs_acl_node_t, walk all ACEs as ace_t",
3119 zfs_acl_node_aces0_walk_init, zfs_aces_walk_step, NULL },
3120 { NULL }
3121 };
3122
3123 static const mdb_modinfo_t modinfo = {
3124 MDB_API_VERSION, dcmds, walkers
3125 };
3126
3127 const mdb_modinfo_t *
3128 _mdb_init(void)
3129 {
3130 return (&modinfo);
3131 }