xref: /titanic_50/usr/src/cmd/mdb/common/modules/zfs/zfs.c (revision 75e454957f9291aa383758b0f444a8ca503c73cf)
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 2006 Sun Microsystems, Inc.  All rights reserved.
23  * Use is subject to license terms.
24  */
25 
26 #pragma ident	"%Z%%M%	%I%	%E% SMI"
27 
28 #include <mdb/mdb_ctf.h>
29 #include <sys/zfs_context.h>
30 #include <sys/mdb_modapi.h>
31 #include <sys/dbuf.h>
32 #include <sys/dmu_objset.h>
33 #include <sys/dsl_dir.h>
34 #include <sys/dsl_pool.h>
35 #include <sys/metaslab_impl.h>
36 #include <sys/space_map.h>
37 #include <sys/list.h>
38 #include <sys/spa_impl.h>
39 #include <sys/vdev_impl.h>
40 #include <sys/zio_compress.h>
41 
42 #ifndef _KERNEL
43 #include "../genunix/list.h"
44 #endif
45 
46 #ifdef _KERNEL
47 #define	ZFS_OBJ_NAME	"zfs"
48 #else
49 #define	ZFS_OBJ_NAME	"libzpool.so.1"
50 #endif
51 
52 static char *
53 local_strdup(const char *s)
54 {
55 	char *s1 = mdb_alloc(strlen(s) + 1, UM_SLEEP);
56 
57 	(void) strcpy(s1, s);
58 	return (s1);
59 }
60 
61 static int
62 getmember(uintptr_t addr, const char *type, mdb_ctf_id_t *idp,
63     const char *member, int len, void *buf)
64 {
65 	mdb_ctf_id_t id;
66 	ulong_t off;
67 	char name[64];
68 
69 	if (idp == NULL) {
70 		if (mdb_ctf_lookup_by_name(type, &id) == -1) {
71 			mdb_warn("couldn't find type %s", type);
72 			return (DCMD_ERR);
73 		}
74 		idp = &id;
75 	} else {
76 		type = name;
77 		mdb_ctf_type_name(*idp, name, sizeof (name));
78 	}
79 
80 	if (mdb_ctf_offsetof(*idp, member, &off) == -1) {
81 		mdb_warn("couldn't find member %s of type %s\n", member, type);
82 		return (DCMD_ERR);
83 	}
84 	if (off % 8 != 0) {
85 		mdb_warn("member %s of type %s is unsupported bitfield",
86 		    member, type);
87 		return (DCMD_ERR);
88 	}
89 	off /= 8;
90 
91 	if (mdb_vread(buf, len, addr + off) == -1) {
92 		mdb_warn("failed to read %s from %s at %p",
93 		    member, type, addr + off);
94 		return (DCMD_ERR);
95 	}
96 	/* mdb_warn("read %s from %s at %p+%llx\n", member, type, addr, off); */
97 
98 	return (0);
99 }
100 
101 #define	GETMEMB(addr, type, member, dest) \
102 	getmember(addr, #type, NULL, #member, sizeof (dest), &(dest))
103 
104 #define	GETMEMBID(addr, ctfid, member, dest) \
105 	getmember(addr, NULL, ctfid, #member, sizeof (dest), &(dest))
106 
107 static int
108 getrefcount(uintptr_t addr, mdb_ctf_id_t *id,
109     const char *member, uint64_t *rc)
110 {
111 	static int gotid;
112 	static mdb_ctf_id_t rc_id;
113 	ulong_t off;
114 
115 	if (!gotid) {
116 		if (mdb_ctf_lookup_by_name("struct refcount", &rc_id) == -1) {
117 			mdb_warn("couldn't find struct refcount");
118 			return (DCMD_ERR);
119 		}
120 		gotid = TRUE;
121 	}
122 
123 	if (mdb_ctf_offsetof(*id, member, &off) == -1) {
124 		char name[64];
125 		mdb_ctf_type_name(*id, name, sizeof (name));
126 		mdb_warn("couldn't find member %s of type %s\n", member, name);
127 		return (DCMD_ERR);
128 	}
129 	off /= 8;
130 
131 	return (GETMEMBID(addr + off, &rc_id, rc_count, *rc));
132 }
133 
134 static int
135 read_symbol(char *sym_name, void **bufp)
136 {
137 	GElf_Sym sym;
138 
139 	if (mdb_lookup_by_obj(MDB_TGT_OBJ_EVERY, sym_name, &sym)) {
140 		mdb_warn("can't find symbol %s", sym_name);
141 		return (DCMD_ERR);
142 	}
143 
144 	*bufp = mdb_alloc(sym.st_size, UM_SLEEP);
145 
146 	if (mdb_vread(*bufp, sym.st_size, sym.st_value) == -1) {
147 		mdb_warn("can't read data for symbol %s", sym_name);
148 		mdb_free(*bufp, sym.st_size);
149 		return (DCMD_ERR);
150 	}
151 
152 	return (DCMD_OK);
153 }
154 
155 static int verbose;
156 
157 static int
158 freelist_walk_init(mdb_walk_state_t *wsp)
159 {
160 	if (wsp->walk_addr == NULL) {
161 		mdb_warn("must supply starting address\n");
162 		return (WALK_ERR);
163 	}
164 
165 	wsp->walk_data = 0;  /* Index into the freelist */
166 	return (WALK_NEXT);
167 }
168 
169 static int
170 freelist_walk_step(mdb_walk_state_t *wsp)
171 {
172 	uint64_t entry;
173 	uintptr_t number = (uintptr_t)wsp->walk_data;
174 	char *ddata[] = { "ALLOC", "FREE", "CONDENSE", "INVALID" };
175 	int mapshift = SPA_MINBLOCKSHIFT;
176 
177 	if (mdb_vread(&entry, sizeof (entry), wsp->walk_addr) == -1) {
178 		mdb_warn("failed to read freelist entry %p", wsp->walk_addr);
179 		return (WALK_DONE);
180 	}
181 	wsp->walk_addr += sizeof (entry);
182 	wsp->walk_data = (void *)(number + 1);
183 
184 	if (SM_DEBUG_DECODE(entry)) {
185 		mdb_printf("DEBUG: %3u  %10s: txg=%llu  pass=%llu\n",
186 		    number,
187 		    ddata[SM_DEBUG_ACTION_DECODE(entry)],
188 		    SM_DEBUG_TXG_DECODE(entry),
189 		    SM_DEBUG_SYNCPASS_DECODE(entry));
190 	} else {
191 		mdb_printf("Entry: %3u  offsets=%08llx-%08llx  type=%c  "
192 		    "size=%06llx", number,
193 		    SM_OFFSET_DECODE(entry) << mapshift,
194 		    (SM_OFFSET_DECODE(entry) + SM_RUN_DECODE(entry)) <<
195 		    mapshift,
196 		    SM_TYPE_DECODE(entry) == SM_ALLOC ? 'A' : 'F',
197 		    SM_RUN_DECODE(entry) << mapshift);
198 		if (verbose)
199 			mdb_printf("      (raw=%012llx)\n", entry);
200 		mdb_printf("\n");
201 	}
202 	return (WALK_NEXT);
203 }
204 
205 
206 static int
207 dataset_name(uintptr_t addr, char *buf)
208 {
209 	static int gotid;
210 	static mdb_ctf_id_t dd_id;
211 	uintptr_t dd_parent;
212 	char dd_myname[MAXNAMELEN];
213 
214 	if (!gotid) {
215 		if (mdb_ctf_lookup_by_name("struct dsl_dir",
216 		    &dd_id) == -1) {
217 			mdb_warn("couldn't find struct dsl_dir");
218 			return (DCMD_ERR);
219 		}
220 		gotid = TRUE;
221 	}
222 	if (GETMEMBID(addr, &dd_id, dd_parent, dd_parent) ||
223 	    GETMEMBID(addr, &dd_id, dd_myname, dd_myname)) {
224 		return (DCMD_ERR);
225 	}
226 
227 	if (dd_parent) {
228 		if (dataset_name(dd_parent, buf))
229 			return (DCMD_ERR);
230 		strcat(buf, "/");
231 	}
232 
233 	if (dd_myname[0])
234 		strcat(buf, dd_myname);
235 	else
236 		strcat(buf, "???");
237 
238 	return (0);
239 }
240 
241 static int
242 objset_name(uintptr_t addr, char *buf)
243 {
244 	static int gotid;
245 	static mdb_ctf_id_t osi_id, ds_id;
246 	uintptr_t os_dsl_dataset;
247 	char ds_snapname[MAXNAMELEN];
248 	uintptr_t ds_dir;
249 
250 	buf[0] = '\0';
251 
252 	if (!gotid) {
253 		if (mdb_ctf_lookup_by_name("struct objset_impl",
254 		    &osi_id) == -1) {
255 			mdb_warn("couldn't find struct objset_impl");
256 			return (DCMD_ERR);
257 		}
258 		if (mdb_ctf_lookup_by_name("struct dsl_dataset",
259 		    &ds_id) == -1) {
260 			mdb_warn("couldn't find struct dsl_dataset");
261 			return (DCMD_ERR);
262 		}
263 
264 		gotid = TRUE;
265 	}
266 
267 	if (GETMEMBID(addr, &osi_id, os_dsl_dataset, os_dsl_dataset))
268 		return (DCMD_ERR);
269 
270 	if (os_dsl_dataset == 0) {
271 		strcat(buf, "mos");
272 		return (0);
273 	}
274 
275 	if (GETMEMBID(os_dsl_dataset, &ds_id, ds_snapname, ds_snapname) ||
276 	    GETMEMBID(os_dsl_dataset, &ds_id, ds_dir, ds_dir)) {
277 		return (DCMD_ERR);
278 	}
279 
280 	if (ds_dir && dataset_name(ds_dir, buf))
281 		return (DCMD_ERR);
282 
283 	if (ds_snapname[0]) {
284 		strcat(buf, "@");
285 		strcat(buf, ds_snapname);
286 	}
287 	return (0);
288 }
289 
290 static void
291 enum_lookup(char *out, size_t size, mdb_ctf_id_t id, int val,
292     const char *prefix)
293 {
294 	const char *cp;
295 	size_t len = strlen(prefix);
296 
297 	if ((cp = mdb_ctf_enum_name(id, val)) != NULL) {
298 		if (strncmp(cp, prefix, len) == 0)
299 			cp += len;
300 		(void) strncpy(out, cp, size);
301 	} else {
302 		mdb_snprintf(out, size, "? (%d)", val);
303 	}
304 }
305 
306 /* ARGSUSED */
307 static int
308 zio_pipeline(uintptr_t addr, uint_t flags, int argc, const mdb_arg_t *argv)
309 {
310 	mdb_ctf_id_t pipe_enum;
311 	int i;
312 	char stage[1024];
313 
314 	if (mdb_ctf_lookup_by_name("enum zio_stage", &pipe_enum) == -1) {
315 		mdb_warn("Could not find enum zio_stage");
316 		return (DCMD_ERR);
317 	}
318 
319 	for (i = 0; i < 32; i++) {
320 		if (addr & (1U << i)) {
321 			enum_lookup(stage, sizeof (stage), pipe_enum, i,
322 			    "ZIO_STAGE_");
323 			mdb_printf("    %s\n", stage);
324 		}
325 	}
326 
327 	return (DCMD_OK);
328 }
329 
330 /* ARGSUSED */
331 static int
332 zfs_params(uintptr_t addr, uint_t flags, int argc, const mdb_arg_t *argv)
333 {
334 	/*
335 	 * This table can be approximately generated by running:
336 	 * egrep "^[a-z0-9_]+ [a-z0-9_]+( =.*)?;" *.c | cut -d ' ' -f 2
337 	 */
338 	static const char *params[] = {
339 		"arc_reduce_dnlc_percent",
340 		"zfs_arc_max",
341 		"zfs_arc_min",
342 		"arc_kmem_reclaim_shift",
343 		"zfs_mdcomp_disable",
344 		"zfs_prefetch_disable",
345 		"zfetch_max_streams",
346 		"zfetch_min_sec_reap",
347 		"zfetch_block_cap",
348 		"zfetch_array_rd_sz",
349 		"zfs_default_bs",
350 		"zfs_default_ibs",
351 		"metaslab_aliquot",
352 		"reference_tracking_enable",
353 		"reference_history",
354 		"zio_taskq_threads",
355 		"spa_max_replication_override",
356 		"spa_mode",
357 		"zfs_flags",
358 		"txg_time",
359 		"zfs_vdev_cache_max",
360 		"zfs_vdev_cache_size",
361 		"zfs_vdev_cache_bshift",
362 		"vdev_mirror_shift",
363 		"zfs_vdev_max_pending",
364 		"zfs_vdev_min_pending",
365 		"zfs_scrub_limit",
366 		"zfs_vdev_time_shift",
367 		"zfs_vdev_ramp_rate",
368 		"zfs_vdev_aggregation_limit",
369 		"fzap_default_block_shift",
370 		"zfs_immediate_write_sz",
371 		"zfs_read_chunk_size",
372 		"zil_disable",
373 		"zfs_nocacheflush",
374 		"zio_gang_bang",
375 		"zio_injection_enabled",
376 		"zvol_immediate_write_sz",
377 	};
378 	int i;
379 
380 	for (i = 0; i < sizeof (params) / sizeof (params[0]); i++) {
381 		int sz;
382 		uint64_t val64;
383 		uint32_t *val32p = (uint32_t *)&val64;
384 
385 		sz = mdb_readvar(&val64, params[i]);
386 		if (sz == 4) {
387 			mdb_printf("%s = 0x%x\n", params[i], *val32p);
388 		} else if (sz == 8) {
389 			mdb_printf("%s = 0x%llx\n", params[i], val64);
390 		} else {
391 			mdb_warn("variable %s not found", params[i]);
392 		}
393 	}
394 
395 	return (DCMD_OK);
396 }
397 
398 /* ARGSUSED */
399 static int
400 blkptr(uintptr_t addr, uint_t flags, int argc, const mdb_arg_t *argv)
401 {
402 	blkptr_t bp;
403 	dmu_object_type_info_t *doti;
404 	zio_compress_info_t *zct;
405 	zio_checksum_info_t *zci;
406 	int i;
407 	char buf[MAXPATHLEN];
408 
409 	if (mdb_vread(&bp, sizeof (blkptr_t), addr) == -1) {
410 		mdb_warn("failed to read blkptr_t");
411 		return (DCMD_ERR);
412 	}
413 
414 	if (read_symbol("dmu_ot", (void **)&doti) != DCMD_OK)
415 		return (DCMD_ERR);
416 	for (i = 0; i < DMU_OT_NUMTYPES; i++) {
417 		mdb_readstr(buf, sizeof (buf), (uintptr_t)doti[i].ot_name);
418 		doti[i].ot_name = local_strdup(buf);
419 	}
420 
421 	if (read_symbol("zio_checksum_table", (void **)&zci) != DCMD_OK)
422 		return (DCMD_ERR);
423 	for (i = 0; i < ZIO_CHECKSUM_FUNCTIONS; i++) {
424 		mdb_readstr(buf, sizeof (buf), (uintptr_t)zci[i].ci_name);
425 		zci[i].ci_name = local_strdup(buf);
426 	}
427 
428 	if (read_symbol("zio_compress_table", (void **)&zct) != DCMD_OK)
429 		return (DCMD_ERR);
430 	for (i = 0; i < ZIO_COMPRESS_FUNCTIONS; i++) {
431 		mdb_readstr(buf, sizeof (buf), (uintptr_t)zct[i].ci_name);
432 		zct[i].ci_name = local_strdup(buf);
433 	}
434 
435 	/*
436 	 * Super-ick warning:  This code is also duplicated in
437 	 * cmd/zdb.c .   Yeah, I hate code replication, too.
438 	 */
439 	for (i = 0; i < BP_GET_NDVAS(&bp); i++) {
440 		dva_t *dva = &bp.blk_dva[i];
441 
442 		mdb_printf("DVA[%d]: vdev_id %lld / %llx\n", i,
443 		    DVA_GET_VDEV(dva), DVA_GET_OFFSET(dva));
444 		mdb_printf("DVA[%d]:       GANG: %-5s  GRID:  %04x\t"
445 		    "ASIZE: %llx\n", i, DVA_GET_GANG(dva) ? "TRUE" : "FALSE",
446 		    DVA_GET_GRID(dva), DVA_GET_ASIZE(dva));
447 		mdb_printf("DVA[%d]: :%llu:%llx:%llx:%s%s%s%s\n", i,
448 		    DVA_GET_VDEV(dva), DVA_GET_OFFSET(dva), BP_GET_PSIZE(&bp),
449 		    BP_SHOULD_BYTESWAP(&bp) ? "e" : "",
450 		    !DVA_GET_GANG(dva) && BP_GET_LEVEL(&bp) != 0 ? "i" : "",
451 		    DVA_GET_GANG(dva) ? "g" : "",
452 		    BP_GET_COMPRESS(&bp) != 0 ? "d" : "");
453 	}
454 	mdb_printf("LSIZE:  %-16llx\t\tPSIZE: %llx\n",
455 	    BP_GET_LSIZE(&bp), BP_GET_PSIZE(&bp));
456 	mdb_printf("ENDIAN: %6s\t\t\t\t\tTYPE:  %s\n",
457 	    BP_GET_BYTEORDER(&bp) ? "LITTLE" : "BIG",
458 	    doti[BP_GET_TYPE(&bp)].ot_name);
459 	mdb_printf("BIRTH:  %-16llx   LEVEL: %-2d\tFILL:  %llx\n",
460 	    bp.blk_birth, BP_GET_LEVEL(&bp), bp.blk_fill);
461 	mdb_printf("CKFUNC: %-16s\t\tCOMP:  %s\n",
462 	    zci[BP_GET_CHECKSUM(&bp)].ci_name,
463 	    zct[BP_GET_COMPRESS(&bp)].ci_name);
464 	mdb_printf("CKSUM:  %llx:%llx:%llx:%llx\n",
465 	    bp.blk_cksum.zc_word[0],
466 	    bp.blk_cksum.zc_word[1],
467 	    bp.blk_cksum.zc_word[2],
468 	    bp.blk_cksum.zc_word[3]);
469 
470 	return (DCMD_OK);
471 }
472 
473 /* ARGSUSED */
474 static int
475 dbuf(uintptr_t addr, uint_t flags, int argc, const mdb_arg_t *argv)
476 {
477 	mdb_ctf_id_t id;
478 	dmu_buf_t db;
479 	uintptr_t objset;
480 	uint8_t level;
481 	uint64_t blkid;
482 	uint64_t holds;
483 	char objectname[32];
484 	char blkidname[32];
485 	char path[MAXNAMELEN];
486 
487 	if (DCMD_HDRSPEC(flags)) {
488 		mdb_printf("        addr object lvl blkid holds os\n");
489 	}
490 
491 	if (mdb_ctf_lookup_by_name("struct dmu_buf_impl", &id) == -1) {
492 		mdb_warn("couldn't find struct dmu_buf_impl_t");
493 		return (DCMD_ERR);
494 	}
495 
496 	if (GETMEMBID(addr, &id, db_objset, objset) ||
497 	    GETMEMBID(addr, &id, db, db) ||
498 	    GETMEMBID(addr, &id, db_level, level) ||
499 	    GETMEMBID(addr, &id, db_blkid, blkid)) {
500 		return (WALK_ERR);
501 	}
502 
503 	if (getrefcount(addr, &id, "db_holds", &holds)) {
504 		return (WALK_ERR);
505 	}
506 
507 	if (db.db_object == DMU_META_DNODE_OBJECT)
508 		(void) strcpy(objectname, "mdn");
509 	else
510 		(void) mdb_snprintf(objectname, sizeof (objectname), "%llx",
511 		    (u_longlong_t)db.db_object);
512 
513 	if (blkid == DB_BONUS_BLKID)
514 		(void) strcpy(blkidname, "bonus");
515 	else
516 		(void) mdb_snprintf(blkidname, sizeof (blkidname), "%llx",
517 		    (u_longlong_t)blkid);
518 
519 	if (objset_name(objset, path)) {
520 		return (WALK_ERR);
521 	}
522 
523 	mdb_printf("%p %8s %1u %9s %2llu %s\n",
524 	    addr, objectname, level, blkidname, holds, path);
525 
526 	return (DCMD_OK);
527 }
528 
529 /* ARGSUSED */
530 static int
531 dbuf_stats(uintptr_t addr, uint_t flags, int argc, const mdb_arg_t *argv)
532 {
533 #define	HISTOSZ 32
534 	uintptr_t dbp;
535 	dmu_buf_impl_t db;
536 	dbuf_hash_table_t ht;
537 	uint64_t bucket, ndbufs;
538 	uint64_t histo[HISTOSZ];
539 	uint64_t histo2[HISTOSZ];
540 	int i, maxidx;
541 
542 	if (mdb_readvar(&ht, "dbuf_hash_table") == -1) {
543 		mdb_warn("failed to read 'dbuf_hash_table'");
544 		return (DCMD_ERR);
545 	}
546 
547 	for (i = 0; i < HISTOSZ; i++) {
548 		histo[i] = 0;
549 		histo2[i] = 0;
550 	}
551 
552 	ndbufs = 0;
553 	for (bucket = 0; bucket < ht.hash_table_mask+1; bucket++) {
554 		int len;
555 
556 		if (mdb_vread(&dbp, sizeof (void *),
557 		    (uintptr_t)(ht.hash_table+bucket)) == -1) {
558 			mdb_warn("failed to read hash bucket %u at %p",
559 			    bucket, ht.hash_table+bucket);
560 			return (DCMD_ERR);
561 		}
562 
563 		len = 0;
564 		while (dbp != 0) {
565 			if (mdb_vread(&db, sizeof (dmu_buf_impl_t),
566 			    dbp) == -1) {
567 				mdb_warn("failed to read dbuf at %p", dbp);
568 				return (DCMD_ERR);
569 			}
570 			dbp = (uintptr_t)db.db_hash_next;
571 			for (i = MIN(len, HISTOSZ - 1); i >= 0; i--)
572 				histo2[i]++;
573 			len++;
574 			ndbufs++;
575 		}
576 
577 		if (len >= HISTOSZ)
578 			len = HISTOSZ-1;
579 		histo[len]++;
580 	}
581 
582 	mdb_printf("hash table has %llu buckets, %llu dbufs "
583 	    "(avg %llu buckets/dbuf)\n",
584 	    ht.hash_table_mask+1, ndbufs,
585 	    (ht.hash_table_mask+1)/ndbufs);
586 
587 	mdb_printf("\n");
588 	maxidx = 0;
589 	for (i = 0; i < HISTOSZ; i++)
590 		if (histo[i] > 0)
591 			maxidx = i;
592 	mdb_printf("hash chain length	number of buckets\n");
593 	for (i = 0; i <= maxidx; i++)
594 		mdb_printf("%u			%llu\n", i, histo[i]);
595 
596 	mdb_printf("\n");
597 	maxidx = 0;
598 	for (i = 0; i < HISTOSZ; i++)
599 		if (histo2[i] > 0)
600 			maxidx = i;
601 	mdb_printf("hash chain depth	number of dbufs\n");
602 	for (i = 0; i <= maxidx; i++)
603 		mdb_printf("%u or more		%llu	%llu%%\n",
604 		    i, histo2[i], histo2[i]*100/ndbufs);
605 
606 
607 	return (DCMD_OK);
608 }
609 
610 typedef struct dbufs_data {
611 	mdb_ctf_id_t id;
612 	uint64_t objset;
613 	uint64_t object;
614 	uint64_t level;
615 	uint64_t blkid;
616 	char *osname;
617 } dbufs_data_t;
618 
619 #define	DBUFS_UNSET	(0xbaddcafedeadbeefULL)
620 
621 /* ARGSUSED */
622 static int
623 dbufs_cb(uintptr_t addr, const void *unknown, void *arg)
624 {
625 	dbufs_data_t *data = arg;
626 	uintptr_t objset;
627 	dmu_buf_t db;
628 	uint8_t level;
629 	uint64_t blkid;
630 	char osname[MAXNAMELEN];
631 
632 	if (GETMEMBID(addr, &data->id, db_objset, objset) ||
633 	    GETMEMBID(addr, &data->id, db, db) ||
634 	    GETMEMBID(addr, &data->id, db_level, level) ||
635 	    GETMEMBID(addr, &data->id, db_blkid, blkid)) {
636 		return (WALK_ERR);
637 	}
638 
639 	if ((data->objset == DBUFS_UNSET || data->objset == objset) &&
640 	    (data->osname == NULL || (objset_name(objset, osname) == 0 &&
641 		strcmp(data->osname, osname) == 0)) &&
642 	    (data->object == DBUFS_UNSET || data->object == db.db_object) &&
643 	    (data->level == DBUFS_UNSET || data->level == level) &&
644 	    (data->blkid == DBUFS_UNSET || data->blkid == blkid)) {
645 		mdb_printf("%#lr\n", addr);
646 	}
647 	return (WALK_NEXT);
648 }
649 
650 /* ARGSUSED */
651 static int
652 dbufs(uintptr_t addr, uint_t flags, int argc, const mdb_arg_t *argv)
653 {
654 	dbufs_data_t data;
655 	char *object = NULL;
656 	char *blkid = NULL;
657 
658 	data.objset = data.object = data.level = data.blkid = DBUFS_UNSET;
659 	data.osname = NULL;
660 
661 	if (mdb_getopts(argc, argv,
662 	    'O', MDB_OPT_UINT64, &data.objset,
663 	    'n', MDB_OPT_STR, &data.osname,
664 	    'o', MDB_OPT_STR, &object,
665 	    'l', MDB_OPT_UINT64, &data.level,
666 	    'b', MDB_OPT_STR, &blkid) != argc) {
667 		return (DCMD_USAGE);
668 	}
669 
670 	if (object) {
671 		if (strcmp(object, "mdn") == 0) {
672 			data.object = DMU_META_DNODE_OBJECT;
673 		} else {
674 			data.object = mdb_strtoull(object);
675 		}
676 	}
677 
678 	if (blkid) {
679 		if (strcmp(blkid, "bonus") == 0) {
680 			data.blkid = DB_BONUS_BLKID;
681 		} else {
682 			data.blkid = mdb_strtoull(blkid);
683 		}
684 	}
685 
686 	if (mdb_ctf_lookup_by_name("struct dmu_buf_impl", &data.id) == -1) {
687 		mdb_warn("couldn't find struct dmu_buf_impl_t");
688 		return (DCMD_ERR);
689 	}
690 
691 	if (mdb_walk("dmu_buf_impl_t", dbufs_cb, &data) != 0) {
692 		mdb_warn("can't walk dbufs");
693 		return (DCMD_ERR);
694 	}
695 
696 	return (DCMD_OK);
697 }
698 
699 typedef struct abuf_find_data {
700 	dva_t dva;
701 	mdb_ctf_id_t id;
702 } abuf_find_data_t;
703 
704 /* ARGSUSED */
705 static int
706 abuf_find_cb(uintptr_t addr, const void *unknown, void *arg)
707 {
708 	abuf_find_data_t *data = arg;
709 	dva_t dva;
710 
711 	if (GETMEMBID(addr, &data->id, b_dva, dva)) {
712 		return (WALK_ERR);
713 	}
714 
715 	if (dva.dva_word[0] == data->dva.dva_word[0] &&
716 	    dva.dva_word[1] == data->dva.dva_word[1]) {
717 		mdb_printf("%#lr\n", addr);
718 	}
719 	return (WALK_NEXT);
720 }
721 
722 /* ARGSUSED */
723 static int
724 abuf_find(uintptr_t addr, uint_t flags, int argc, const mdb_arg_t *argv)
725 {
726 	abuf_find_data_t data;
727 	GElf_Sym sym;
728 	int i;
729 	const char *syms[] = {
730 		"ARC_mru",
731 		"ARC_mru_ghost",
732 		"ARC_mfu",
733 		"ARC_mfu_ghost",
734 	};
735 
736 	if (argc != 2)
737 		return (DCMD_USAGE);
738 
739 	for (i = 0; i < 2; i ++) {
740 		switch (argv[i].a_type) {
741 		case MDB_TYPE_STRING:
742 			data.dva.dva_word[i] = mdb_strtoull(argv[i].a_un.a_str);
743 			break;
744 		case MDB_TYPE_IMMEDIATE:
745 			data.dva.dva_word[i] = argv[i].a_un.a_val;
746 			break;
747 		default:
748 			return (DCMD_USAGE);
749 		}
750 	}
751 
752 	if (mdb_ctf_lookup_by_name("struct arc_buf_hdr", &data.id) == -1) {
753 		mdb_warn("couldn't find struct arc_buf_hdr");
754 		return (DCMD_ERR);
755 	}
756 
757 	for (i = 0; i < sizeof (syms) / sizeof (syms[0]); i++) {
758 		if (mdb_lookup_by_name(syms[i], &sym)) {
759 			mdb_warn("can't find symbol %s", syms[i]);
760 			return (DCMD_ERR);
761 		}
762 
763 		if (mdb_pwalk("list", abuf_find_cb, &data, sym.st_value) != 0) {
764 			mdb_warn("can't walk %s", syms[i]);
765 			return (DCMD_ERR);
766 		}
767 	}
768 
769 	return (DCMD_OK);
770 }
771 
772 void
773 abuf_help(void)
774 {
775 	mdb_printf("::abuf_find dva_word[0] dva_word[1]\n");
776 }
777 
778 /*
779  * ::spa
780  *
781  * 	-c	Print configuration information as well
782  * 	-v	Print vdev state
783  * 	-e	Print vdev error stats
784  *
785  * Print a summarized spa_t.  When given no arguments, prints out a table of all
786  * active pools on the system.
787  */
788 /* ARGSUSED */
789 static int
790 spa_print(uintptr_t addr, uint_t flags, int argc, const mdb_arg_t *argv)
791 {
792 	spa_t spa;
793 	char poolname[MAXNAMELEN];
794 	const char *statetab[] = { "ACTIVE", "EXPORTED", "DESTROYED",
795 		"UNINIT", "UNAVAIL" };
796 	const char *state;
797 	int config = FALSE;
798 	int vdevs = FALSE;
799 	int errors = FALSE;
800 
801 	if (mdb_getopts(argc, argv,
802 	    'c', MDB_OPT_SETBITS, TRUE, &config,
803 	    'v', MDB_OPT_SETBITS, TRUE, &vdevs,
804 	    'e', MDB_OPT_SETBITS, TRUE, &errors,
805 	    NULL) != argc)
806 		return (DCMD_USAGE);
807 
808 	if (!(flags & DCMD_ADDRSPEC)) {
809 		if (mdb_walk_dcmd("spa", "spa", argc, argv) == -1) {
810 			mdb_warn("can't walk spa");
811 			return (DCMD_ERR);
812 		}
813 
814 		return (DCMD_OK);
815 	}
816 
817 	if (flags & DCMD_PIPE_OUT) {
818 		mdb_printf("%#lr\n", addr);
819 		return (DCMD_OK);
820 	}
821 
822 	if (DCMD_HDRSPEC(flags))
823 		mdb_printf("%<u>%-?s %9s %-*s%</u>\n", "ADDR", "STATE",
824 		    sizeof (uintptr_t) == 4 ? 60 : 52, "NAME");
825 
826 	if (mdb_vread(&spa, sizeof (spa), addr) == -1) {
827 		mdb_warn("failed to read spa_t at %p", addr);
828 		return (DCMD_ERR);
829 	}
830 
831 	if (mdb_readstr(poolname, sizeof (poolname), (uintptr_t)spa.spa_name)
832 	    == -1) {
833 		mdb_warn("failed to read pool name at %p", spa.spa_name);
834 		return (DCMD_ERR);
835 	}
836 
837 	if (spa.spa_state < 0 || spa.spa_state > POOL_STATE_UNAVAIL)
838 		state = "UNKNOWN";
839 	else
840 		state = statetab[spa.spa_state];
841 
842 	mdb_printf("%0?p %9s %s\n", addr, state, poolname);
843 
844 	if (config) {
845 		mdb_printf("\n");
846 		mdb_inc_indent(4);
847 		if (mdb_call_dcmd("spa_config", addr, flags, 0,
848 		    NULL) != DCMD_OK)
849 			return (DCMD_ERR);
850 		mdb_dec_indent(4);
851 	}
852 
853 	if (vdevs || errors) {
854 		mdb_arg_t v;
855 
856 		v.a_type = MDB_TYPE_STRING;
857 		v.a_un.a_str = "-e";
858 
859 		mdb_printf("\n");
860 		mdb_inc_indent(4);
861 		if (mdb_call_dcmd("spa_vdevs", addr, flags, errors ? 1 : 0,
862 		    &v) != DCMD_OK)
863 			return (DCMD_ERR);
864 		mdb_dec_indent(4);
865 	}
866 
867 	return (DCMD_OK);
868 }
869 
870 /*
871  * ::spa_config
872  *
873  * Given a spa_t, print the configuration information stored in spa_config.
874  * Since it's just an nvlist, format it as an indented list of name=value pairs.
875  * We simply read the value of spa_config and pass off to ::nvlist.
876  */
877 /* ARGSUSED */
878 static int
879 spa_print_config(uintptr_t addr, uint_t flags, int argc, const mdb_arg_t *argv)
880 {
881 	spa_t spa;
882 
883 	if (argc != 0 || !(flags & DCMD_ADDRSPEC))
884 		return (DCMD_USAGE);
885 
886 	if (mdb_vread(&spa, sizeof (spa), addr) == -1) {
887 		mdb_warn("failed to read spa_t at %p", addr);
888 		return (DCMD_ERR);
889 	}
890 
891 	if (spa.spa_config == NULL) {
892 		mdb_printf("(none)\n");
893 		return (DCMD_OK);
894 	}
895 
896 	return (mdb_call_dcmd("nvlist", (uintptr_t)spa.spa_config, flags,
897 	    0, NULL));
898 }
899 
900 void
901 vdev_help(void)
902 {
903 	mdb_printf("[vdev_t*]::vdev [-er]\n"
904 		"\t-> -e display vdev stats\n"
905 		"\t-> -r recursive (visit all children)\n");
906 }
907 
908 /*
909  * ::vdev
910  *
911  * Print out a summarized vdev_t, in the following form:
912  *
913  * ADDR             STATE	AUX            DESC
914  * fffffffbcde23df0 HEALTHY	-              /dev/dsk/c0t0d0
915  *
916  * If '-r' is specified, recursively visit all children.
917  *
918  * With '-e', the statistics associated with the vdev are printed as well.
919  */
920 static int
921 do_print_vdev(uintptr_t addr, int flags, int depth, int stats,
922     int recursive)
923 {
924 	vdev_t vdev;
925 	char desc[MAXNAMELEN];
926 	int c, children;
927 	uintptr_t *child;
928 	const char *state, *aux;
929 
930 	if (mdb_vread(&vdev, sizeof (vdev), (uintptr_t)addr) == -1) {
931 		mdb_warn("failed to read vdev_t at %p\n", (uintptr_t)addr);
932 		return (DCMD_ERR);
933 	}
934 
935 	if (flags & DCMD_PIPE_OUT) {
936 		mdb_printf("%#lr", addr);
937 	} else {
938 		if (vdev.vdev_path != NULL) {
939 			if (mdb_readstr(desc, sizeof (desc),
940 			    (uintptr_t)vdev.vdev_path) == -1) {
941 				mdb_warn("failed to read vdev_path at %p\n",
942 				    vdev.vdev_path);
943 				return (DCMD_ERR);
944 			}
945 		} else if (vdev.vdev_ops != NULL) {
946 			vdev_ops_t ops;
947 			if (mdb_vread(&ops, sizeof (ops),
948 			    (uintptr_t)vdev.vdev_ops) == -1) {
949 				mdb_warn("failed to read vdev_ops at %p\n",
950 				    vdev.vdev_ops);
951 				return (DCMD_ERR);
952 			}
953 			(void) strcpy(desc, ops.vdev_op_type);
954 		} else {
955 			(void) strcpy(desc, "<unknown>");
956 		}
957 
958 		if (depth == 0 && DCMD_HDRSPEC(flags))
959 			mdb_printf("%<u>%-?s %-9s %-12s %-*s%</u>\n",
960 			    "ADDR", "STATE", "AUX",
961 			    sizeof (uintptr_t) == 4 ? 43 : 35,
962 			    "DESCRIPTION");
963 
964 		mdb_printf("%0?p ", addr);
965 
966 		switch (vdev.vdev_state) {
967 		case VDEV_STATE_CLOSED:
968 		    state = "CLOSED";
969 		    break;
970 		case VDEV_STATE_OFFLINE:
971 		    state = "OFFLINE";
972 		    break;
973 		case VDEV_STATE_CANT_OPEN:
974 		    state = "CANT_OPEN";
975 		    break;
976 		case VDEV_STATE_DEGRADED:
977 		    state = "DEGRADED";
978 		    break;
979 		case VDEV_STATE_HEALTHY:
980 		    state = "HEALTHY";
981 		    break;
982 		default:
983 		    state = "UNKNOWN";
984 		    break;
985 		}
986 
987 		switch (vdev.vdev_stat.vs_aux) {
988 		case VDEV_AUX_NONE:
989 			aux = "-";
990 			break;
991 		case VDEV_AUX_OPEN_FAILED:
992 			aux = "OPEN_FAILED";
993 			break;
994 		case VDEV_AUX_CORRUPT_DATA:
995 			aux = "CORRUPT_DATA";
996 			break;
997 		case VDEV_AUX_NO_REPLICAS:
998 			aux = "NO_REPLICAS";
999 			break;
1000 		case VDEV_AUX_BAD_GUID_SUM:
1001 			aux = "BAD_GUID_SUM";
1002 			break;
1003 		case VDEV_AUX_TOO_SMALL:
1004 			aux = "TOO_SMALL";
1005 			break;
1006 		case VDEV_AUX_BAD_LABEL:
1007 			aux = "BAD_LABEL";
1008 			break;
1009 		default:
1010 			aux = "UNKNOWN";
1011 			break;
1012 		}
1013 
1014 		mdb_printf("%-9s %-12s %*s%s\n", state, aux, depth, "", desc);
1015 
1016 		if (stats) {
1017 			vdev_stat_t *vs = &vdev.vdev_stat;
1018 			int i;
1019 
1020 			mdb_inc_indent(4);
1021 			mdb_printf("\n");
1022 			mdb_printf("%<u>       %12s %12s %12s %12s "
1023 			    "%12s%</u>\n", "READ", "WRITE", "FREE", "CLAIM",
1024 			    "IOCTL");
1025 			mdb_printf("OPS     ");
1026 			for (i = 1; i < ZIO_TYPES; i++)
1027 				mdb_printf("%11#llx%s", vs->vs_ops[i],
1028 				    i == ZIO_TYPES - 1 ? "" : "  ");
1029 			mdb_printf("\n");
1030 			mdb_printf("BYTES   ");
1031 			for (i = 1; i < ZIO_TYPES; i++)
1032 				mdb_printf("%11#llx%s", vs->vs_bytes[i],
1033 				    i == ZIO_TYPES - 1 ? "" : "  ");
1034 
1035 
1036 			mdb_printf("\n");
1037 			mdb_printf("EREAD    %10#llx\n", vs->vs_read_errors);
1038 			mdb_printf("EWRITE   %10#llx\n", vs->vs_write_errors);
1039 			mdb_printf("ECKSUM   %10#llx\n",
1040 			    vs->vs_checksum_errors);
1041 			mdb_dec_indent(4);
1042 		}
1043 
1044 		if (stats)
1045 			mdb_printf("\n");
1046 	}
1047 
1048 	children = vdev.vdev_children;
1049 
1050 	if (children == 0 || !recursive)
1051 		return (DCMD_OK);
1052 
1053 	child = mdb_alloc(children * sizeof (void *), UM_SLEEP | UM_GC);
1054 	if (mdb_vread(child, children * sizeof (void *),
1055 	    (uintptr_t)vdev.vdev_child) == -1) {
1056 		mdb_warn("failed to read vdev children at %p", vdev.vdev_child);
1057 		return (DCMD_ERR);
1058 	}
1059 
1060 	for (c = 0; c < children; c++) {
1061 		if (do_print_vdev(child[c], flags, depth + 2, stats,
1062 		    recursive))
1063 			return (DCMD_ERR);
1064 	}
1065 
1066 	return (DCMD_OK);
1067 }
1068 
1069 static int
1070 vdev_print(uintptr_t addr, uint_t flags, int argc, const mdb_arg_t *argv)
1071 {
1072 	int recursive = FALSE;
1073 	int stats = FALSE;
1074 
1075 	if (mdb_getopts(argc, argv,
1076 	    'r', MDB_OPT_SETBITS, TRUE, &recursive,
1077 	    'e', MDB_OPT_SETBITS, TRUE, &stats,
1078 	    NULL) != argc)
1079 		return (DCMD_USAGE);
1080 
1081 	if (!(flags & DCMD_ADDRSPEC)) {
1082 		mdb_warn("no vdev_t address given\n");
1083 		return (DCMD_ERR);
1084 	}
1085 
1086 	return (do_print_vdev(addr, flags, 0, stats, recursive));
1087 }
1088 
1089 typedef struct metaslab_walk_data {
1090 	uint64_t mw_numvdevs;
1091 	uintptr_t *mw_vdevs;
1092 	int mw_curvdev;
1093 	uint64_t mw_nummss;
1094 	uintptr_t *mw_mss;
1095 	int mw_curms;
1096 } metaslab_walk_data_t;
1097 
1098 static int
1099 metaslab_walk_step(mdb_walk_state_t *wsp)
1100 {
1101 	metaslab_walk_data_t *mw = wsp->walk_data;
1102 	metaslab_t ms;
1103 	uintptr_t msp;
1104 
1105 	if (mw->mw_curvdev >= mw->mw_numvdevs)
1106 		return (WALK_DONE);
1107 
1108 	if (mw->mw_mss == NULL) {
1109 		uintptr_t mssp;
1110 		uintptr_t vdevp;
1111 
1112 		ASSERT(mw->mw_curms == 0);
1113 		ASSERT(mw->mw_nummss == 0);
1114 
1115 		vdevp = mw->mw_vdevs[mw->mw_curvdev];
1116 		if (GETMEMB(vdevp, struct vdev, vdev_ms, mssp) ||
1117 		    GETMEMB(vdevp, struct vdev, vdev_ms_count, mw->mw_nummss)) {
1118 			return (WALK_ERR);
1119 		}
1120 
1121 		mw->mw_mss = mdb_alloc(mw->mw_nummss * sizeof (void*),
1122 		    UM_SLEEP | UM_GC);
1123 		if (mdb_vread(mw->mw_mss, mw->mw_nummss * sizeof (void*),
1124 		    mssp) == -1) {
1125 			mdb_warn("failed to read vdev_ms at %p", mssp);
1126 			return (WALK_ERR);
1127 		}
1128 	}
1129 
1130 	if (mw->mw_curms >= mw->mw_nummss) {
1131 		mw->mw_mss = NULL;
1132 		mw->mw_curms = 0;
1133 		mw->mw_nummss = 0;
1134 		mw->mw_curvdev++;
1135 		return (WALK_NEXT);
1136 	}
1137 
1138 	msp = mw->mw_mss[mw->mw_curms];
1139 	if (mdb_vread(&ms, sizeof (metaslab_t), msp) == -1) {
1140 		mdb_warn("failed to read metaslab_t at %p", msp);
1141 		return (WALK_ERR);
1142 	}
1143 
1144 	mw->mw_curms++;
1145 
1146 	return (wsp->walk_callback(msp, &ms, wsp->walk_cbdata));
1147 }
1148 
1149 /* ARGSUSED */
1150 static int
1151 metaslab_walk_init(mdb_walk_state_t *wsp)
1152 {
1153 	metaslab_walk_data_t *mw;
1154 	uintptr_t root_vdevp;
1155 	uintptr_t childp;
1156 
1157 	if (wsp->walk_addr == NULL) {
1158 		mdb_warn("must supply address of spa_t\n");
1159 		return (WALK_ERR);
1160 	}
1161 
1162 	mw = mdb_zalloc(sizeof (metaslab_walk_data_t), UM_SLEEP | UM_GC);
1163 
1164 	if (GETMEMB(wsp->walk_addr, struct spa, spa_root_vdev, root_vdevp) ||
1165 	    GETMEMB(root_vdevp, struct vdev, vdev_children, mw->mw_numvdevs) ||
1166 	    GETMEMB(root_vdevp, struct vdev, vdev_child, childp)) {
1167 		return (DCMD_ERR);
1168 	}
1169 
1170 	mw->mw_vdevs = mdb_alloc(mw->mw_numvdevs * sizeof (void *),
1171 	    UM_SLEEP | UM_GC);
1172 	if (mdb_vread(mw->mw_vdevs, mw->mw_numvdevs * sizeof (void *),
1173 	    childp) == -1) {
1174 		mdb_warn("failed to read root vdev children at %p", childp);
1175 		return (DCMD_ERR);
1176 	}
1177 
1178 	wsp->walk_data = mw;
1179 
1180 	return (WALK_NEXT);
1181 }
1182 
1183 typedef struct mdb_spa {
1184 	uintptr_t spa_dsl_pool;
1185 	uintptr_t spa_root_vdev;
1186 } mdb_spa_t;
1187 
1188 typedef struct mdb_dsl_dir {
1189 	uintptr_t dd_phys;
1190 	uint64_t dd_used_bytes;
1191 	int64_t dd_space_towrite[TXG_SIZE];
1192 } mdb_dsl_dir_t;
1193 
1194 typedef struct mdb_dsl_dir_phys {
1195 	uint64_t dd_used_bytes;
1196 	uint64_t dd_compressed_bytes;
1197 	uint64_t dd_uncompressed_bytes;
1198 } mdb_dsl_dir_phys_t;
1199 
1200 typedef struct mdb_vdev {
1201 	uintptr_t vdev_parent;
1202 	uintptr_t vdev_ms;
1203 	uint64_t vdev_ms_count;
1204 	vdev_stat_t vdev_stat;
1205 } mdb_vdev_t;
1206 
1207 typedef struct mdb_metaslab {
1208 	space_map_t ms_allocmap[TXG_SIZE];
1209 	space_map_t ms_freemap[TXG_SIZE];
1210 	space_map_t ms_map;
1211 	space_map_obj_t ms_smo;
1212 	space_map_obj_t ms_smo_syncing;
1213 } mdb_metaslab_t;
1214 
1215 typedef struct space_data {
1216 	uint64_t ms_allocmap[TXG_SIZE];
1217 	uint64_t ms_freemap[TXG_SIZE];
1218 	uint64_t ms_map;
1219 	uint64_t avail;
1220 	uint64_t nowavail;
1221 } space_data_t;
1222 
1223 /* ARGSUSED */
1224 static int
1225 space_cb(uintptr_t addr, const void *unknown, void *arg)
1226 {
1227 	space_data_t *sd = arg;
1228 	mdb_metaslab_t ms;
1229 
1230 	if (GETMEMB(addr, struct metaslab, ms_allocmap, ms.ms_allocmap) ||
1231 	    GETMEMB(addr, struct metaslab, ms_freemap, ms.ms_freemap) ||
1232 	    GETMEMB(addr, struct metaslab, ms_map, ms.ms_map) ||
1233 	    GETMEMB(addr, struct metaslab, ms_smo, ms.ms_smo) ||
1234 	    GETMEMB(addr, struct metaslab, ms_smo_syncing, ms.ms_smo_syncing)) {
1235 		return (WALK_ERR);
1236 	}
1237 
1238 	sd->ms_allocmap[0] += ms.ms_allocmap[0].sm_space;
1239 	sd->ms_allocmap[1] += ms.ms_allocmap[1].sm_space;
1240 	sd->ms_allocmap[2] += ms.ms_allocmap[2].sm_space;
1241 	sd->ms_allocmap[3] += ms.ms_allocmap[3].sm_space;
1242 	sd->ms_freemap[0] += ms.ms_freemap[0].sm_space;
1243 	sd->ms_freemap[1] += ms.ms_freemap[1].sm_space;
1244 	sd->ms_freemap[2] += ms.ms_freemap[2].sm_space;
1245 	sd->ms_freemap[3] += ms.ms_freemap[3].sm_space;
1246 	sd->ms_map += ms.ms_map.sm_space;
1247 	sd->avail += ms.ms_map.sm_size - ms.ms_smo.smo_alloc;
1248 	sd->nowavail += ms.ms_map.sm_size - ms.ms_smo_syncing.smo_alloc;
1249 
1250 	return (WALK_NEXT);
1251 }
1252 
1253 /*
1254  * ::spa_space [-b]
1255  *
1256  * Given a spa_t, print out it's on-disk space usage and in-core
1257  * estimates of future usage.  If -b is given, print space in bytes.
1258  * Otherwise print in megabytes.
1259  */
1260 /* ARGSUSED */
1261 static int
1262 spa_space(uintptr_t addr, uint_t flags, int argc, const mdb_arg_t *argv)
1263 {
1264 	mdb_spa_t spa;
1265 	uintptr_t dp_root_dir;
1266 	mdb_dsl_dir_t dd;
1267 	mdb_dsl_dir_phys_t dsp;
1268 	uint64_t children;
1269 	uintptr_t childaddr;
1270 	space_data_t sd;
1271 	int shift = 20;
1272 	char *suffix = "M";
1273 	int bits = FALSE;
1274 
1275 	if (mdb_getopts(argc, argv, 'b', MDB_OPT_SETBITS, TRUE, &bits, NULL) !=
1276 	    argc)
1277 		return (DCMD_USAGE);
1278 	if (!(flags & DCMD_ADDRSPEC))
1279 		return (DCMD_USAGE);
1280 
1281 	if (bits) {
1282 		shift = 0;
1283 		suffix = "";
1284 	}
1285 
1286 	if (GETMEMB(addr, struct spa, spa_dsl_pool, spa.spa_dsl_pool) ||
1287 	    GETMEMB(addr, struct spa, spa_root_vdev, spa.spa_root_vdev) ||
1288 	    GETMEMB(spa.spa_root_vdev, struct vdev, vdev_children, children) ||
1289 	    GETMEMB(spa.spa_root_vdev, struct vdev, vdev_child, childaddr) ||
1290 	    GETMEMB(spa.spa_dsl_pool, struct dsl_pool,
1291 	    dp_root_dir, dp_root_dir) ||
1292 	    GETMEMB(dp_root_dir, struct dsl_dir, dd_phys, dd.dd_phys) ||
1293 	    GETMEMB(dp_root_dir, struct dsl_dir,
1294 	    dd_used_bytes, dd.dd_used_bytes) ||
1295 	    GETMEMB(dp_root_dir, struct dsl_dir,
1296 	    dd_space_towrite, dd.dd_space_towrite) ||
1297 	    GETMEMB(dd.dd_phys, struct dsl_dir_phys,
1298 	    dd_used_bytes, dsp.dd_used_bytes) ||
1299 	    GETMEMB(dd.dd_phys, struct dsl_dir_phys,
1300 	    dd_compressed_bytes, dsp.dd_compressed_bytes) ||
1301 	    GETMEMB(dd.dd_phys, struct dsl_dir_phys,
1302 	    dd_uncompressed_bytes, dsp.dd_uncompressed_bytes)) {
1303 		return (DCMD_ERR);
1304 	}
1305 
1306 	mdb_printf("dd_space_towrite = %llu%s %llu%s %llu%s %llu%s\n",
1307 	    dd.dd_space_towrite[0] >> shift, suffix,
1308 	    dd.dd_space_towrite[1] >> shift, suffix,
1309 	    dd.dd_space_towrite[2] >> shift, suffix,
1310 	    dd.dd_space_towrite[3] >> shift, suffix);
1311 	mdb_printf("dd_used_bytes = %llu%s\n",
1312 	    dd.dd_used_bytes >> shift, suffix);
1313 
1314 	mdb_printf("dd_phys.dd_used_bytes = %llu%s\n",
1315 	    dsp.dd_used_bytes >> shift, suffix);
1316 	mdb_printf("dd_phys.dd_compressed_bytes = %llu%s\n",
1317 	    dsp.dd_compressed_bytes >> shift, suffix);
1318 	mdb_printf("dd_phys.dd_uncompressed_bytes = %llu%s\n",
1319 	    dsp.dd_uncompressed_bytes >> shift, suffix);
1320 
1321 	bzero(&sd, sizeof (sd));
1322 	if (mdb_pwalk("metaslab", space_cb, &sd, addr) != 0) {
1323 		mdb_warn("can't walk metaslabs");
1324 		return (DCMD_ERR);
1325 	}
1326 
1327 	mdb_printf("ms_allocmap = %llu%s %llu%s %llu%s %llu%s\n",
1328 	    sd.ms_allocmap[0] >> shift, suffix,
1329 	    sd.ms_allocmap[1] >> shift, suffix,
1330 	    sd.ms_allocmap[2] >> shift, suffix,
1331 	    sd.ms_allocmap[3] >> shift, suffix);
1332 	mdb_printf("ms_freemap = %llu%s %llu%s %llu%s %llu%s\n",
1333 	    sd.ms_freemap[0] >> shift, suffix,
1334 	    sd.ms_freemap[1] >> shift, suffix,
1335 	    sd.ms_freemap[2] >> shift, suffix,
1336 	    sd.ms_freemap[3] >> shift, suffix);
1337 	mdb_printf("ms_map = %llu%s\n", sd.ms_map >> shift, suffix);
1338 	mdb_printf("last synced avail = %llu%s\n", sd.avail >> shift, suffix);
1339 	mdb_printf("current syncing avail = %llu%s\n",
1340 	    sd.nowavail >> shift, suffix);
1341 
1342 	return (DCMD_OK);
1343 }
1344 
1345 /*
1346  * ::spa_verify
1347  *
1348  * Given a spa_t, verify that that the pool is self-consistent.
1349  * Currently, it only checks to make sure that the vdev tree exists.
1350  */
1351 /* ARGSUSED */
1352 static int
1353 spa_verify(uintptr_t addr, uint_t flags, int argc, const mdb_arg_t *argv)
1354 {
1355 	spa_t spa;
1356 
1357 	if (argc != 0 || !(flags & DCMD_ADDRSPEC))
1358 		return (DCMD_USAGE);
1359 
1360 	if (mdb_vread(&spa, sizeof (spa), addr) == -1) {
1361 		mdb_warn("failed to read spa_t at %p", addr);
1362 		return (DCMD_ERR);
1363 	}
1364 
1365 	if (spa.spa_root_vdev == NULL) {
1366 		mdb_printf("no vdev tree present\n");
1367 		return (DCMD_OK);
1368 	}
1369 
1370 	return (DCMD_OK);
1371 }
1372 
1373 /*
1374  * ::spa_vdevs
1375  *
1376  * 	-e	Include error stats
1377  *
1378  * Print out a summarized list of vdevs for the given spa_t.
1379  * This is accomplished by invoking "::vdev -re" on the root vdev.
1380  */
1381 /* ARGSUSED */
1382 static int
1383 spa_vdevs(uintptr_t addr, uint_t flags, int argc, const mdb_arg_t *argv)
1384 {
1385 	spa_t spa;
1386 	mdb_arg_t v;
1387 	int errors = FALSE;
1388 
1389 	if (mdb_getopts(argc, argv,
1390 	    'e', MDB_OPT_SETBITS, TRUE, &errors,
1391 	    NULL) != argc)
1392 		return (DCMD_USAGE);
1393 
1394 	if (!(flags & DCMD_ADDRSPEC))
1395 		return (DCMD_USAGE);
1396 
1397 	if (mdb_vread(&spa, sizeof (spa), addr) == -1) {
1398 		mdb_warn("failed to read spa_t at %p", addr);
1399 		return (DCMD_ERR);
1400 	}
1401 
1402 	/*
1403 	 * Unitialized spa_t structures can have a NULL root vdev.
1404 	 */
1405 	if (spa.spa_root_vdev == NULL) {
1406 		mdb_printf("no associated vdevs\n");
1407 		return (DCMD_OK);
1408 	}
1409 
1410 	v.a_type = MDB_TYPE_STRING;
1411 	v.a_un.a_str = errors ? "-re" : "-r";
1412 
1413 	return (mdb_call_dcmd("vdev", (uintptr_t)spa.spa_root_vdev,
1414 	    flags, 1, &v));
1415 }
1416 
1417 typedef struct txg_list_walk_data {
1418 	uintptr_t lw_head[TXG_SIZE];
1419 	int	lw_txgoff;
1420 	int	lw_maxoff;
1421 	size_t	lw_offset;
1422 	void	*lw_obj;
1423 } txg_list_walk_data_t;
1424 
1425 static int
1426 txg_list_walk_init_common(mdb_walk_state_t *wsp, int txg, int maxoff)
1427 {
1428 	txg_list_walk_data_t *lwd;
1429 	txg_list_t list;
1430 	int i;
1431 
1432 	lwd = mdb_alloc(sizeof (txg_list_walk_data_t), UM_SLEEP | UM_GC);
1433 	if (mdb_vread(&list, sizeof (txg_list_t), wsp->walk_addr) == -1) {
1434 		mdb_warn("failed to read txg_list_t at %#lx", wsp->walk_addr);
1435 		return (WALK_ERR);
1436 	}
1437 
1438 	for (i = 0; i < TXG_SIZE; i++)
1439 		lwd->lw_head[i] = (uintptr_t)list.tl_head[i];
1440 	lwd->lw_offset = list.tl_offset;
1441 	lwd->lw_obj = mdb_alloc(lwd->lw_offset + sizeof (txg_node_t),
1442 	    UM_SLEEP | UM_GC);
1443 	lwd->lw_txgoff = txg;
1444 	lwd->lw_maxoff = maxoff;
1445 
1446 	wsp->walk_addr = lwd->lw_head[lwd->lw_txgoff];
1447 	wsp->walk_data = lwd;
1448 
1449 	return (WALK_NEXT);
1450 }
1451 
1452 static int
1453 txg_list_walk_init(mdb_walk_state_t *wsp)
1454 {
1455 	return (txg_list_walk_init_common(wsp, 0, TXG_SIZE-1));
1456 }
1457 
1458 static int
1459 txg_list0_walk_init(mdb_walk_state_t *wsp)
1460 {
1461 	return (txg_list_walk_init_common(wsp, 0, 0));
1462 }
1463 
1464 static int
1465 txg_list1_walk_init(mdb_walk_state_t *wsp)
1466 {
1467 	return (txg_list_walk_init_common(wsp, 1, 1));
1468 }
1469 
1470 static int
1471 txg_list2_walk_init(mdb_walk_state_t *wsp)
1472 {
1473 	return (txg_list_walk_init_common(wsp, 2, 2));
1474 }
1475 
1476 static int
1477 txg_list3_walk_init(mdb_walk_state_t *wsp)
1478 {
1479 	return (txg_list_walk_init_common(wsp, 3, 3));
1480 }
1481 
1482 static int
1483 txg_list_walk_step(mdb_walk_state_t *wsp)
1484 {
1485 	txg_list_walk_data_t *lwd = wsp->walk_data;
1486 	uintptr_t addr;
1487 	txg_node_t *node;
1488 	int status;
1489 
1490 	while (wsp->walk_addr == NULL && lwd->lw_txgoff < lwd->lw_maxoff) {
1491 		lwd->lw_txgoff++;
1492 		wsp->walk_addr = lwd->lw_head[lwd->lw_txgoff];
1493 	}
1494 
1495 	if (wsp->walk_addr == NULL)
1496 		return (WALK_DONE);
1497 
1498 	addr = wsp->walk_addr - lwd->lw_offset;
1499 
1500 	if (mdb_vread(lwd->lw_obj,
1501 	    lwd->lw_offset + sizeof (txg_node_t), addr) == -1) {
1502 		mdb_warn("failed to read list element at %#lx", addr);
1503 		return (WALK_ERR);
1504 	}
1505 
1506 	status = wsp->walk_callback(addr, lwd->lw_obj, wsp->walk_cbdata);
1507 	node = (txg_node_t *)((uintptr_t)lwd->lw_obj + lwd->lw_offset);
1508 	wsp->walk_addr = (uintptr_t)node->tn_next[lwd->lw_txgoff];
1509 
1510 	return (status);
1511 }
1512 
1513 /*
1514  * ::walk spa
1515  *
1516  * Walk all named spa_t structures in the namespace.  This is nothing more than
1517  * a layered avl walk.
1518  */
1519 static int
1520 spa_walk_init(mdb_walk_state_t *wsp)
1521 {
1522 	GElf_Sym sym;
1523 
1524 	if (wsp->walk_addr != NULL) {
1525 		mdb_warn("spa walk only supports global walks\n");
1526 		return (WALK_ERR);
1527 	}
1528 
1529 	if (mdb_lookup_by_obj(ZFS_OBJ_NAME, "spa_namespace_avl", &sym) == -1) {
1530 		mdb_warn("failed to find symbol 'spa_namespace_avl'");
1531 		return (WALK_ERR);
1532 	}
1533 
1534 	wsp->walk_addr = (uintptr_t)sym.st_value;
1535 
1536 	if (mdb_layered_walk("avl", wsp) == -1) {
1537 		mdb_warn("failed to walk 'avl'\n");
1538 		return (WALK_ERR);
1539 	}
1540 
1541 	return (WALK_NEXT);
1542 }
1543 
1544 static int
1545 spa_walk_step(mdb_walk_state_t *wsp)
1546 {
1547 	spa_t	spa;
1548 
1549 	if (mdb_vread(&spa, sizeof (spa), wsp->walk_addr) == -1) {
1550 		mdb_warn("failed to read spa_t at %p", wsp->walk_addr);
1551 		return (WALK_ERR);
1552 	}
1553 
1554 	return (wsp->walk_callback(wsp->walk_addr, &spa, wsp->walk_cbdata));
1555 }
1556 
1557 /*
1558  * MDB module linkage information:
1559  *
1560  * We declare a list of structures describing our dcmds, and a function
1561  * named _mdb_init to return a pointer to our module information.
1562  */
1563 
1564 static const mdb_dcmd_t dcmds[] = {
1565 	{ "blkptr", ":", "print blkptr_t", blkptr },
1566 	{ "dbuf", ":", "print dmu_buf_impl_t", dbuf },
1567 	{ "dbuf_stats", ":", "dbuf stats", dbuf_stats },
1568 	{ "dbufs",
1569 	"\t[-O objset_t*] [-n objset_name | \"mos\"] [-o object | \"mdn\"] \n"
1570 	"\t[-l level] [-b blkid | \"bonus\"]",
1571 	"find dmu_buf_impl_t's that meet criterion", dbufs },
1572 	{ "abuf_find", "dva_word[0] dva_word[1]",
1573 	"find arc_buf_hdr_t of a specified DVA",
1574 	abuf_find },
1575 	{ "spa", "?[-cv]", "spa_t summary", spa_print },
1576 	{ "spa_config", ":", "print spa_t configuration", spa_print_config },
1577 	{ "spa_verify", ":", "verify spa_t consistency", spa_verify },
1578 	{ "spa_space", ":[-b]", "print spa_t on-disk space usage", spa_space },
1579 	{ "spa_vdevs", ":", "given a spa_t, print vdev summary", spa_vdevs },
1580 	{ "vdev", ":[-re]", "vdev_t summary", vdev_print },
1581 	{ "zio_pipeline", ":", "decode a zio pipeline", zio_pipeline },
1582 	{ "zfs_params", "", "print zfs tunable parameters", zfs_params },
1583 	{ NULL }
1584 };
1585 
1586 static const mdb_walker_t walkers[] = {
1587 	/*
1588 	 * In userland, there is no generic provider of list_t walkers, so we
1589 	 * need to add it.
1590 	 */
1591 #ifndef _KERNEL
1592 	{ LIST_WALK_NAME, LIST_WALK_DESC,
1593 		list_walk_init, list_walk_step, list_walk_fini },
1594 #endif
1595 	{ "zms_freelist", "walk ZFS metaslab freelist",
1596 		freelist_walk_init, freelist_walk_step, NULL },
1597 	{ "txg_list", "given any txg_list_t *, walk all entries in all txgs",
1598 		txg_list_walk_init, txg_list_walk_step, NULL },
1599 	{ "txg_list0", "given any txg_list_t *, walk all entries in txg 0",
1600 		txg_list0_walk_init, txg_list_walk_step, NULL },
1601 	{ "txg_list1", "given any txg_list_t *, walk all entries in txg 1",
1602 		txg_list1_walk_init, txg_list_walk_step, NULL },
1603 	{ "txg_list2", "given any txg_list_t *, walk all entries in txg 2",
1604 		txg_list2_walk_init, txg_list_walk_step, NULL },
1605 	{ "txg_list3", "given any txg_list_t *, walk all entries in txg 3",
1606 		txg_list3_walk_init, txg_list_walk_step, NULL },
1607 	{ "spa", "walk all spa_t entries in the namespace",
1608 		spa_walk_init, spa_walk_step, NULL },
1609 	{ "metaslab", "given a spa_t *, walk all metaslab_t structures",
1610 		metaslab_walk_init, metaslab_walk_step, NULL },
1611 	{ NULL }
1612 };
1613 
1614 static const mdb_modinfo_t modinfo = {
1615 	MDB_API_VERSION, dcmds, walkers
1616 };
1617 
1618 const mdb_modinfo_t *
1619 _mdb_init(void)
1620 {
1621 	return (&modinfo);
1622 }
1623