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